Research Article

Asymmetric division of clonal muscle stem cells coordinates muscle regeneration in vivo

See allHide authors and affiliations

Science  08 Jul 2016:
Vol. 353, Issue 6295, aad9969
DOI: 10.1126/science.aad9969

Dividing asymmetrically to fix muscle

Resident tissue stem cells called satellite cells repair muscle after injury. However, how satellite cells operate inside living tissue is unclear. Gurevich et al. exploited the optical clarity of zebrafish larvae and used a series of genetic approaches to study muscle injury. After injury, satellite cells divide asymmetrically to generate a progenitor pool for muscle replacement and at the same time “self-renew” the satellite stem cell. This results in regeneration that is highly clonal in nature, validating many decades of in vitro analyses examining the regenerative capacity of skeletal muscle.

Science, this issue p. 136

Structured Abstract


Mammalian skeletal muscle harbors tissue-specific stem cells that are triggered to replace damaged fibers after injury. Genetic ablation of satellite cells in the mouse results in a failure to regenerate muscle, which indicates that these cells are the major (and possibly only) mediators for repair of skeletal muscle. Further evidence for the central role of satellite cells in muscle regeneration comes from transplantation experiments with genetically marked cells, which demonstrate that satellite cells are highly proliferative myogenic precursors capable of self‐renewal and the resumption of quiescence, properties deemed important in a cell population responsible for muscle repair. Considerable in vitro evidence, derived from cultured fibers and myoblasts, is suggestive of a role for asymmetric division in generating both a self-renewing “immortal” stem cell and a differentiation-competent progenitor cell that proliferates and ultimately replaces damaged muscle. However, asymmetric division of satellite cells has not been documented in vivo. Furthermore, considerable doubt remains over how accurately in vitro studies can model satellite cell behavior, given that the isolation and culture of individual muscle fibers and cells stimulates satellite cell proliferation. Finally, it is not clear whether the environment an activated satellite cell encounters in a single fiber explant, or in culture, mimics the molecular and biophysical architecture of a regenerating muscle injury in vivo. Consequently, what role, if any, the wound environment itself plays in regeneration and self-renewal is difficult to address in these systems.


Using the optical clarity and genetic tractability of the zebrafish system, we developed tools to track and image the regeneration of living muscle tissue after injury. Marking muscle stem and progenitor cells with transgenes and using long-term imaging and lineage-tracing modalities enabled us to visualize cell movements and behaviors during regeneration in vivo.


In vivo cell tracking permitted high-resolution imaging of the entire process of muscle regeneration, from injury to fiber replacement. Using this approach, we were able to determine the morphological, cellular, and genetic basis for zebrafish muscle regeneration. Our analysis identified a stem cell niche in the zebrafish myotome that is equivalent to the mammalian satellite cell system, revealing that this evolutionarily ancient stem cell is probably present throughout the vertebrate phylogeny. Complex interactions were observed between satellite cells and both injured and uninjured fibers within the wound environment. Among the most notable of these was the identification of filopodia-like projections, emanating from uninjured fibers, which adhere to and “lasso” the activated satellite cell to guide it to the wound edge. Furthermore, we documented the in vivo occurrence of asymmetric satellite cell division, a process that drives both self-renewal and regeneration via a clonally restricted progenitor pool.


Asymmetric divisions occur during in vivo muscle regeneration to generate clonally related progenitors required for muscle repair. This finding resolves a long-term debate surrounding the existence of this mechanism of stem cell self-renewal and muscle repair in vivo. Our results also reveal the highly dynamic nature of the wound environment, where uninjured fibers at the wound edge play a crucial role in directing differentiating progenitors to regions of the wound that are most in need of new fiber addition.

Mechanism of in vivo muscle repair.

(A to C) Muscle regeneration is clonal. Regenerating fibers (outlined in white) express the same color after fluorescent lineage tracing, indicating clonal derivation from a single stem cell. Sagittal, transverse, and coronal sections are shown in (A) to (C), respectively. (D) Regeneration dynamics in vivo. Quiescent satellite cells, activated upon injury, undergo asymmetric division, which results in self-renewing or proliferating cells. Proliferative cells undergo myogenesis to generate de novo immature fibers.


Skeletal muscle is an example of a tissue that deploys a self-renewing stem cell, the satellite cell, to effect regeneration. Recent in vitro studies have highlighted a role for asymmetric divisions in renewing rare “immortal” stem cells and generating a clonal population of differentiation-competent myoblasts. However, this model currently lacks in vivo validation. We define a zebrafish muscle stem cell population analogous to the mammalian satellite cell and image the entire process of muscle regeneration from injury to fiber replacement in vivo. This analysis reveals complex interactions between satellite cells and both injured and uninjured fibers and provides in vivo evidence for the asymmetric division of satellite cells driving both self-renewal and regeneration via a clonally restricted progenitor pool.

Adult muscle stem cells or satellite cells play a critical role in amniote skeletal muscle repair. Normally quiescent and located between the mature muscle fiber and the overlying basal lamina, satellite cells are characterized by the expression of a number of specific markers, including the paired homeobox genes Pax3 and Pax7, the myogenic regulatory factor (MRF) gene Myf5, and the hepatocyte growth factor membrane receptor cMet (1, 2). Genetic ablation of Pax7-expressing satellite cells from adult muscle results in a failure to regenerate, demonstrating that satellite cells are the major and possibly only mediators of murine skeletal muscle repair (3). The ability of satellite cells to self-renew has been demonstrated in transplantation experiments using genetically marked cells (4). Transplantation of a single intact myofiber with its complement of less than 10 satellite cells has been shown to produce more than 100 new myofibers containing 25,000 nuclei, as well as a contribution to the host pool of satellite cells from the donor tissue (4). Such experiments demonstrate that satellite cells are proliferative myogenic cells capable of self-renewal and reassumption of quiescence, all of which are important requirements for a stem cell population responsible for muscle repair. However, exactly how self-renewal is regulated remains an intense area of investigation. The majority of studies have focused on a possible role for asymmetric divisions of satellite cells in generating a self-renewing “immortal” stem cell and a progenitor cell required for the generation of a differentiation-competent population that proliferates and ultimately replaces damaged muscle (59). Although considerable in vitro evidence derived from cultured fibers and myoblasts is suggestive of a role for asymmetric division during satellite cell self-renewal, asymmetric division of satellite cells has not been documented to date in any in vivo environment. Furthermore, considerable doubt remains over how accurately current in vitro–based paradigms can model satellite cell behavior. This is because the in vitro culture of muscle fibers and the isolation of individual cells is inherently an activating process for satellite cells. Finally, it is not clear whether the environment an activated satellite cell encounters in a single fiber explant or in culture mimics the molecular and biophysical architecture of a regenerating muscle injury in vivo. Consequently, what role, if any, the wound environment itself plays in regeneration and self-renewal is difficult to address. Recent studies using intravital imaging within injured and uninjured mouse muscle have begun to examine these questions in vivo, but technical constraints have prevented the examination of asymmetric cell division and self-renewal (10).

De novo muscle repair mechanisms in zebrafish

In this study, we subjected transgenic zebrafish to distinct muscle injuries and followed regeneration by simple optical inspection of muscle tissue. Of the injury models tested (1113), the one that proved to be the most rapid, reproducible, and amenable to gene expression analyses was the needle-stick larval injury. In this model, an epaxial myotome at the end of the yolk extension of larvae at 4 days postfertilization (dpf) is injured with a single puncture, using a 30-gauge needle held at an angle of 75° relative to the body axis (Fig. 1, B to D, and fig. S1, A to E). This form of injury resulted in the loss of about half of all fibers in the epaxial myotome at all medial lateral levels at the point of injury. The repair process can be directly monitored in the living larvae via the use of transgenes that mark muscle fibers and by using the light-refracting properties of intact muscle sarcomeres (termed birefringence) to quantitate muscle integrity. Our analyses revealed that needle-stick muscle injury in zebrafish larvae results in the rapid activation of endogenous muscle repair, a process that plateaus and concludes by 7 days postinjury (dpi) (fig. S1, A, B, and G, n = 20 fish; fig. S1, C to E, n = 7 fish; and movie S1). Fiber loss occurs immediately after injury and continues for several days. There is little evidence of wound contraction contributing to repair because the wound site remains stable during the regenerative process until it is repaired by new fiber addition (fig. S2, n = 8 fish; and movie S2).

Fig. 1 In toto imaging of muscle regeneration in vivo.

(A to D) Uninjured muscle [Tg(actc1b:mCherry), red] contains myf5-positive cells [Tg(myf5:GFP), green], a subset of which are activated postinjury and exhibit a highly bipolar extended shape [arrowheads in (B) to (D)]. The remainder of the myf5-positive cells fail to be activated and do not migrate to the injury (arrows). Representative images taken from n = 12 independent fish. Scale bar, 50 μm. (E to G) Newly regenerated muscle initially expresses slow MyHC, shown by Tg(smyhc1:GFP) (green), despite the injury predominately affecting deeper fast MyHC–expressing muscle [Tg(actc1b:BFP), blue]. Representative images taken from n = 14 independent fish. Arrows indicate slow MyHC-expressing fibers in the fast muscle domain. Scale bar, 50μ m. (H) Stereotypical phases of muscle regeneration. Maximum intensity projection through Tg(myf5:GFP) and Tg(actcb1:mCherry) double-transgenic larvae, post–laser wound injury, time-lapsed over 48 hours after injury. Arrows mark cell divisions and arrowheads marks the filopodial extension from the adjacent uninjured muscle. T = time in minutes. Representative images taken from n = 8 independent fish. Scale bar, 50μm. (I) Quantitation of the length of the phases described in (H). (J to N′) Inhibition of filopodia prevents regeneration. (J) Quantitation of birefringence after needle-stab injury in 3-dpf larvae soaked in a low-concentration range of cytochalasin D, shown to specifically inhibit filopodial extensions in vivo and in vitro. All concentrations inhibit regeneration by 4 dpi; the first time point that a recovery in birefringence is detected in untreated siblings. Error bars indicate mean + SEM. Statistics: t test, two-tailed numbers (n) are indicated at each time point and concentration. ****P < 0.0001. (K to N′) Individual larvae followed over the entire period of regeneration were imaged for birefringence each day. (K to L′) Control untreated siblings recover rapidly form needle-stick injury. (M to N′) Larvae treated with 50 mM cytochalasin D fail to regenerate and reestablish birefringence by 4 dpi. (K′ to N′) High-magnification views of the regions boxed in (K) to (N), respectively.

To study the cell biological basis of this repair, we performed needle-stick injuries in fish that were transgenic for both Tg(actc1b:mcherry) and Tg(myf5:GFP) (GFP, green fluorescent protein), which mark differentiated muscle fibers in red and myf5 expression in green (Fig. 1, A to D, n = 12 fish; figs. S2 and S3, A to D, n = 38 fish; and movies S1 and S2). Myf5 in amniotes is expressed in both quiescent and activated satellite cells and transit-amplifying progenitors during muscle regeneration and, consequently, marks an extended phase of the regenerative process (2, 14). We theorized, therefore, that the Tg(actc1b:mcherry) and Tg(myf5:GFP) transgene combination may allow us to image stem cell activation to muscle differentiation in continuous time-lapse analyses. In uninjured larvae, myf5-positive cells are located throughout the myotome, interstitial to differentiated muscles fibers, and first appear at 4 dpf (Fig. 1A). Injury induces a series of dynamic cell shape changes within a proportion of myf5-positive cells; these changes include the extension of long filopodial processes into the wound, reminiscent of the behavior deployed by muscle progenitors within injured adult mouse muscle (10). Consequently, migrating myf5-positive cells exhibit an elongated shape, entering the wound site from multiple levels and positions within the injury (movie S3), following their membrane extensions into the wound site (Fig. 1, B to D, and movie S1). This activated subset of myf5-positive cells is chiefly characterized by their polarized, migratory phenotype, with activated cells often elongating and migrating over nonextended myf5-GFP cells that are initially closer to the wound (Fig. 1, B to D). By 2 dpi, myf5-positive cells within the wound site have begun to elongate as myocytes, with evidence by 3 dpi of fully differentiated muscle fibers replacing those that have been lost (fig. S1, C to E, n = 7 fish; and fig. S3, A to D). myf5 expression peaks at 3 dpi and declines steadily, returning to preinjury levels by 7 dpi, when regeneration is complete (fig. S3, A to D).

To examine the nature of the de novo myogenic event induced by muscle injury, we investigated the expression of endogenous genes involved in muscle repair in amniotes. Specifically, we examined the expression of pax7, which marks both quiescent and activated muscle stem cells and is the most universally accepted marker of satellite cells in postnatal murine muscle (5). We also investigated the expression of the MRF genes myf5, myoD, and myogenin, which are sequentially activated during regeneration in mammals (2). For all myogenic regulators, expression is specifically induced at the injury site, in a temporal fashion consistent with their roles in controlling a local de novo myogenic event after injury (12) (fig. S3, E to H, n = 10 fish per group).

Furthermore, regenerating fibers initially express an embryonic slow myosin heavy-chain (MyHC) isoform during differentiation. This observation reinforces the notion that different processes, distinct to those that occur during muscle growth, coordinate muscle regeneration. Regenerating fibers express the slow MyHC1 gene (15) and protein (16), and quantitation of this expression (fig. S1, H to R, n = 6 fish) as well as analysis of the expression of the slow MyHC1:GFP transgene [Tg(smyhc1:GFP)] during injury (Fig. 1, E to G, n = 14 fish), revealed that this is a transient expression exhibited specifically by newly regenerating fibers. Embryonic slow MyHC1 expression occurs despite the fact that the injury encompasses a region of the myotome that is mainly composed of fast MyHC-expressing fibers (Fig. 1, E to G, and fig. S1, H to R). Transient embryonic slow MyHC expression, independent of the predominant fiber type of the injured muscle, is also a feature of regenerating fibers in amniote injury models (1720). We therefore concluded that muscle injury in zebrafish results in activation of a specific de novo myogenic program.

Stereotypical phases of muscle repair in vivo

Next we induced muscle injuries via laser ablation of muscle fibers in the Tg(actc1b:mcherry) and Tg(myf5:GFP) double-transgenic line. Laser ablation results in single muscle fibers or small groups of fibers being ablated in a more defined area of injury when compared with the needle-stick injury used above. Regeneration of this laser injury is more rapid and occurs over a 3 days (fig. S1F, n = 20 fish). This allows the entire wound to be captured at resolutions that can distinguish individual cells. We were able to capture the entirety of the regenerative process, from the initial injury to the differentiation of replacement muscle cells. This analysis, coupled with those described above, allowed the definition and quantification of specific phases in the regenerative processes (Fig. 1, H and I, and movie S4, n = 8 fish).

The initial phase (migration and contact) is characterized by migration of the myf5-positive cell populations into the wound area. Migration of myf5-positive cells is limited to within the injured myotome itself and the neighboring myotomes on either side. Migration to the wound site can occur from any position within these myotomes. However, we never see dorsal-ventral traverse of the horizontal myosepta for discrete wounds within either the epaxial or hypaxial myotomes where the septum remains intact after injury (movies S1 to S7). Once in the wound, cells associate as rounded cells with severed and dying fibers (Fig. 1, H and I, and movie S4, green arrows). In the second phase (proliferation and activation), myf5-positive cells undergo division to generate muscle progenitors (Fig. 1, H and I, and movie S4, cyan arrows), similar to events and behaviors exhibited by myogenic progenitors in early muscle repair in mice (10). Quantitation of the total number of proliferative cells confirmed that muscle regeneration was associated with a rapid proliferative burst during the 2 days immediately after injury (fig. S6, A, C, and I, n = 6 fish). EdU pulse labeling after needle-stick injury confirmed the proliferative nature of the myf5-positive cells detectable within the wound site, in line with our time-lapse observations of laser ablation injuries (fig. S6, F and I, n = 6 fish). In the third phase (bipolar-extension and interaction), myf5-positive cells extend to a bipolar shape, increase myf5:GFP expression, and lose contact with the dying cells around them. Bipolar myf5-positive cells in the wound then actively migrate toward each other, extend filopodia to make contact, and then recoil and move rapidly apart (Fig. 1, H and I, and movie S4, blue arrows). Next, resident uninjured differentiated muscle fibers extend filopodial-like membrane protrusions, which adhere and “lasso” around the activated progenitor cells (Fig. 1, H and I, movie S4, yellow arrows, and fig. S4). Filopodia can be observed in every time-lapse analysis of sufficient length that we performed for both types of injury: needle stick and laser ablation (n = 8 fish, Fig. 1H, fig. S4). Filopodia are invariably associated with the arrival of the myf5-positive cells into the wound site and extend to contact them. The size and nature of the filopodia are dependent on the location and distance between the uninjured muscle at the wound and the position of the myf5-positive cells relative to the muscle cell at the wound edge (Fig. 1H, movie S4, and fig. S4). The timing and directionality of the extending filopodia suggest that signals from the migrating and dividing stem and progenitor cells may elicit and guide the filopodia. Consequent filopodial retraction appears to guide the lassoed elongating myocytes to align next to the differentiated muscle fiber from which the contacting filopodia extends (Fig. 1, H and I, and movie S4).

To define the functional role of filopodia in the regenerative process, we inhibited their formation after injury. Low-dose cytochalasin D selectively suppresses filopodia formation by capping actin filament barbed ends while sparing the formation of lamellipodia, another F-actin–dependent membrane extension process (21, 22). Furthermore, cytochalasin D has recently been applied in vivo to specifically inhibit filopodia in developing mouse and zebrafish embryos (23, 24). Using a similar approach, we applied low-dose cytochalasin D to animals injured by needle-stab and assayed regeneration under these conditions. Addition of low-dose cytochalasin D had no overall effect on larval morphology and muscle integrity in uninjured animals (Fig. 1, M to N′, n = 59). However, injured animals showed a marked defect in regeneration upon cytochalasin D treatment (n = 17, 5 nM; n = 17, 25 nM; n = 19, 50 nM), which suggests that impeding filopodia formation leads to inhibited regeneration (Fig. 1, J to N′).

After alignment against resident uninjured fibers, myocytes elongate and initiate expression of Tg(actc1b:mcherry), signaling the differentiation of the myf5-positive progenitor pool into muscle fibers that heralds the end of the final phase (resident fiber capture and differentiation) (Fig. 1, H and I, movie S4, white arrows). Thus, our analyses have determined that muscle regeneration is associated with a series of stereotypical morphogenetic interactions between cells at the wound site and the activated stem cells.

Muscle regeneration is impaired in myogenin mutants

Next we examined whether muscle regeneration deployed a genetic program distinct from that required during embryonic myogenesis. In mice, the MRF genes Myf5 and MyoD are individually required to direct regeneration of skeletal muscle (25, 26). Mutations in each of the zebrafish MRF genes (myf5, myoD, mrf4, and myog) have been previously generated (27, 28). Using birefringence (fig. S5, B and F, n = 10 fish per condition), gene expression, and muscle differentiation markers (fig. S5, C to E, n = 6 fish per probe per condition) to systematically monitor regeneration in homozygous mutants, we determined that only homozygous myogenin (myogfh265) mutants possessed a muscle-repair deficit (fig. S5). Myogenin is required for the differentiation of muscle cells during embryonic amniote myogenesis and is consequently a prenatal lethal mutation in mice (29). By contrast, zebrafish myog mutants possess no apparent defect in embryonic myogenesis, as myod and myog act redundantly to coordinate embryonic fast muscle myogenesis (27, 28). Similar numbers of stem cells are present at the wound site initially, with the deficit in repair being accompanied by a subsequent defect in proliferation at the injury (fig. S6, D, E, and H, n = 6 fish). To determine more precisely the phase of regeneration that was deficient in myog mutants, homozygous myogfh265 larvae transgenic for both Tg(myf5:GFP) and Tg(actc1b:mCherry) were subjected to laser ablation muscle injury and continuous time-lapse imaging, as described above (Fig. 2A and movie S5, n = 7 fish). myf5-expressing cells in myogfh265-mutant fish migrate effectively to the injury site. However, these cells failed to fully extend and mature to a bipolar shape and were eliminated from the wound site (Fig. 2, A and B, and movie S5), consistent with the marker analyses described above. Specifically, myogfh265-mutant fish possessed dying myf5-expressing cells only within the injury site (Fig. 2, C to H, n = 6 fish), which indicates that myog acts as a survival checkpoint for myoblasts involved in muscle repair. This analysis therefore defines a genetically distinct set of myogenin-dependent cells that are specifically required for muscle regeneration in zebrafish.

Fig. 2 Myogenin activity is specifically required for muscle regeneration as a survival checkpoint for proliferating myoblasts.

(A) Maximum intensity projection of a myogfh265 mutant fish, double-transgenic for Tg(myf5:GFP) (green) and Tg(actcb1:mCherry) (red), time-lapsed over 35 hours after injury t, time in minutes.. myf5-expressing cells (arrows) exhibit defects in differentiation. Specifically, cells enter the wound site and initiate bipolar extension but fail to progress to differentiation and die within the wound. Representative images were taken from n = 7 independent fish. (B) Quantitation of the length of the phases described in Fig. 1 for WT injuries, as related to the phases observed in the myogfh265 mutant context, from (A). (C to F′′′) Maximum intensity projection of needle-stick–injured WT and myogfh265 mutant fish, transgenic for Tg(myf5:GFP) (green) and stained with TUNEL (red). myogfh265 mutant fish possess myf5-positive cells that are also positive for TUNEL staining (arrowheads). Examples of non–myf5-positive TUNEL-positive cells (triangles) and myf5-positive non–TUNEL-positive (open arrowheads) cells are also evident. White boxes denote the specific area of the wound viewed in the subsequent three single-channel images. Representative images taken were from n = 6 independent fish. DAPI, 4′,6-diamidino-2-phenylindole. (G) Quantification of the total amount of TUNEL-positive cells. Both WT and mutant fish have comparable levels of TUNEL-positive cells throughout muscle regeneration. Scores shown are means for each measured day post– needle-stick injury (DPI) from n = 6 independent fish per condition (mutant or WT). Error bars indicate mean + SEM. Statistics: t test, two-tailed. NS, not significant. (H) Quantification of the number of myf5-expressing cells costained with TUNEL, expressed as a percentage the total number of myf5-positive cells in the wound site. Scores shown are means for each measured day post injury from n = 6 independent fish per condition (mutant or WT). Error bars indicate mean + SEM. Statistics: t test, two-tailed. ***P < 0.001.

Specific stem cell populations control muscle repair

Although these initial studies reveal that a genetically distinct population of cells contribute to the regenerative process, they do not formally eliminate the possibility that these cells arise through dedifferentiation at the wound site and consequently contribute to muscle regeneration via the de novo myogenesis process described above. To determine whether injured skeletal muscles can undergo dedifferentiation and contribute to repair, we injected the construct actc1b:nlsGFP into the Tg(actc1b:mCherryCAAX) transgenic fish, which results in mosaic genetically marked myonuclei (green) in a fish with all muscle fiber membranes marked (red) (fig. S7 A′). We performed laser ablation of muscle cells containing these green nuclei and showed that the nuclei of injured muscle fibers are cleared from the site of the wound and subsequently do not play a role in repair via dedifferentiation (fig. S7, A to D′, n = 6 fish). To determine whether surrounding uninjured myofibers were capable of dedifferentiating and playing a role in injury repair, we performed needle-stick injuries to fish double-transgenic for Tg(actc1b:Cre) and Tg(ef1a1:LOXP-eGFP-LOXP-mCherry) (eGFP, enhanced GFP), marking all tissue in the fish as green, except for mature myofibers that expressed actc1b (marked as red). Examination of these injured fish reveals that the regenerate consists of green cells early in the repair process, indicating that these cells are not derived from dedifferentiated mature myofibers (fig. S7, E to G, n = 6 fish). Together, these results show that skeletal muscle regeneration in zebrafish does not largely occur by dedifferentiation of injured or uninjured muscle cells at the injury site. Skeletal muscle regeneration is therefore fundamentally different from the regeneration of the other striated muscle type, cardiac muscle, which in zebrafish is controlled by the reentry of cardiomyocytes into the cell cycle in response to injury (30).

We next looked for a source of potential stem cells within the myotome. Although myf5 expression encompasses the broadest temporal phase of an activated stem cell in amniote muscle repair, it occurs extensively throughout all phases of myogenesis and is not a marker specific for the stem cell compartment. We therefore sought to define markers that were specific for muscle stem cells deployed during regeneration. Our analysis revealed that pax7a- and pax3a-positive cells evident deep within the myotome, beginning at 4 dpf in a similar location to the myf5 cells described above (Fig. 3, A to F′′′, n = 6 fish). However, these genes are also expressed by cells of the external cell layer (ECL), a progenitor pool for muscle growth located on the superficial surface of the myotome, evident during embryonic and larval periods. Therefore, the expression of these genes is not specific to a muscle stem cell niche (11, 12, 31, 32) (Fig. 3, A and D). cMet is a marker of quiescent satellite cells in postnatal rodent muscle and is thought to be specifically required for the activation of satellite cells in response to injury (2, 33). In line with its proposed function, cmet expression initiates at 4 dpf, later than either pax3 or pax7; is expressed in only a small subset of the deeper cells of the myotome; and is not expressed in the progenitor layer of the ECL (34) (Fig. 3, A to F′′′, and fig. S8, n = 5 fish). cmet is also prominently expressed in nonmuscle progenitor populations, including primary motor neurons and the lateral line nerve (35) (Fig. 3, A and D, and fig. S8, H to J). Furthermore, the expression of cmet persists into adulthood in a fiber-associated niche (fig. S8, A to G′), similar to that described for the mammalian satellite cell. This observation is consistent with the regenerative capacity of adult zebrafish muscle, as well as data from previous studies of adult fiber isolation in zebrafish (13, 31, 36). To better characterize these cells in vivo, we generated a bacterial artificial chromosome (BAC) transgenic line that expressed both KALTA4 and mCherry from the cmet promoter. This Tg(cmet:mCherry-T2A-KALTA4)-modified BAC faithfully recapitulated the expression of the cmet gene upon germline transmission (Fig. 3, A to F′′′, and fig. S8, A to G′). Furthermore, an antibody against cMet (36) colocalizes with mCherry expression in the myotome evident within the cmet BAC transgenic line, validating that this line reports endogenous cmet expression (Fig. 3 and fig. S8, H to J′). This transgenic line demonstrated a marked increase in cmet-positive cell number per myotome section throughout the early development of the fish, leading up to metamorphosis (up to 15 mm in total length), after which this population maintained a stable number throughout postmetamorphic development (15 to 30 mm total length), suggesting a consistent requirement of this cell type within the myotome throughout the life span of the fish (fig. S8, K and L). Using this transgenic line, we established that cmet and pax3a double-positive and cmet and pax7a double-positive cells constituted a minor subset of the interstitial myotomal cells (~15% of non-ECL, myotomal, pax3- or pax7-positive cells) (Fig. 3, A to J, n = 3 per transgenic combination). These cells express a cyclin-dependent kinase inhibitor and therefore probably represent a quiescent population (Fig. 3, K to N′′′, n = 8 fish).

Fig. 3 cmet expression specifically marks satellite-like cells required for zebrafish muscle regeneration.

(A) Maximum-projection confocal images of Tg(pax3a:GFP) (green) and Tg(cmet:mCherry-T2A-KALTA4) (red) double-transgenic fish. (B and C) Single confocal sections through the myotome show pax3a and cmet double-labeled cells [(B) parasagittal, (C) transverse]. (B′ to B′′′ and C′ to C′′′) High-magnification views of regions boxed in (B) and (C), respectively (arrowheads, pax3a-high and cmet-high double-positive cells; arrows, pax3a-high and cmet-low–expressing cells). (D) Maximum-projection confocal images of Tg(pax7a:GFP) (green) and Tg(cmet:mCherry-T2A-KALTA4) (red) double-transgenic fish. (E and F) Single confocal sections through the myotome show pax7a and cmet double-labeled cells within the myotome [(E) parasagittal, (F) transverse]. (E′ to E′′′ and F′ to F′′′) High-magnification views of regions boxed in (E) and (F), respectively (arrowheads, double-positive cells; arrows, single-positive cells). (G and H) Quantification of pax3a-only versus pax3a and cmet double-labeled cells (G) or pax7a-only versus pax7a and cmet double-labeled cells (H) per myotome. (I and J) Quantification of pax3a and cmet double-positive cells as a percentage of total pax3a-positive cells (I) or pax7a and cmet double-positive cells as a percentage of total pax7a-positive cells (J). For both pax3a and pax7a, cmet colabels 10 to 15% of the total population. Error bars indicate mean + SEM. (A to F′′) Representative images were selected from n = 6 independent fish. (G to J) Numbers are derived from total myotomal counts of three individual larvae per transgene combination. (K to N′′′) Cross section of 4-dpf [(K) to (K′′′)] and 10-mm [(L) to (N′′′)] Tg(cmet:mCherry-T2A-KALTA4) larvae costaining with an antibody against the cyclin-dependent kinase inhibitor ink4b. Arrows highlight colocalizing ink4b nuclear staining that occurs in cmet-positive cells (representative images from n = 8 sectioned embryos).

Next, we performed in vivo ablation of the aforementioned cmet cells to determine whether these cells are required for regeneration in zebrafish. We crossed the Tg(cmet:mCherry-T2A-KALTA4) line to Tg(uas:E1b:Eco.NfsB-mCherry), which results in the expression of the nitroreductase enzyme specifically within cmet-positive cells. Addition of metronidazole (Met) to these embryos ablated cmet-positive cells (Fig. 4, A to D, n = 30 fish) and resulted in a failure to initiate regeneration and de novo myogenesis at the wound site (Fig. 4, E to N, n > 20 per treatment and probe). As cmet is also expressed within motor neurons, it remained possible that the coablation of neurons with cmet-positive muscle stem cells could also directly or indirectly affect the regeneration of muscle cells that they innervate. To control for this possibility, we made use of a previously generated transgenic line, Tg(cmetMN:GFP-T2A-KALTA4), which contains only a subset of the cmet regulatory regions evident in the full-length BAC clone and expresses KALTA4 and GFP only within the motor neuron expression domain (35). Crossing this line to Tg(uas:E1b:Eco.NfsB-mCherry) results in the expression of the nitroreductase enzyme specifically within cmet-positive motor neurons. Addition of Met to these embryos resulted in complete ablation of primary motor neurons (Fig. 4, O to P′). In parallel, we also performed ablations in the cmet BAC line, which ablates both cmet-positive motor neurons and muscle stem cells. To assay the effect of these two different ablation strategies on muscle regeneration, we performed needle-stab injuries on ablated animals and assayed the regenerative capacity of injured larvae. Although the chemical ablations carried out in the BAC transgenic line resulted in defects in muscle regeneration, ablation of cmet-positive motor neurons had no effect on the ability of larvae to regenerate (Fig. 4, Q to U). These studies collectively define cmet-positive cells as the functional equivalent of mammalian satellite cells, specifically required for muscle regeneration within the zebrafish myotome.

Fig. 4 cmet-positive cells are required for regeneration.

(A to C) Maximum-projection confocal images of larvae transgenic for Tg(pax7a:GFP); Tg(cmet:KALTA4-T2A-mCherry); Tg(uas:E1b:Eco.NfsB-mCherry), which results in the expression of the nitroreductase enzyme specifically within cmet-positive cells. Addition of the prodrug metronidazole (Met) to these triple-transgenic larvae [(B) and (C)] results in ablation of cmet-expressing cells [arrowheads in (A)] but does not affect myoblasts expressing pax7a alone. The images from the cmet ablation are quantified in (D), indicating that 1 mM of Met is sufficient to induce significant reduction in cmet-positive cell numbers. mCherry-positive cellular debris collects superficially in metronidazole-treated larvae [(B) and (C)], which results not only from myoblast cell death but also degradation of the lateral line [asterisks in (A)] and motor neurons [arrows mark cell bodies in (A)], which also express cmet. The appearance of this mCherry-positive cellular debris is used as a positive control to show that the ablation has worked efficiently. Error bars in (D) indicate mean + SEM. ***P < 0.001. (E to N) Ablation of cmet-positive cells through metronidazole addition [(K) to (M)] results in a failure to initiate de novo myogenesis, which is indicated as failure to initiate the expression of myogenic regulatory factor gene expression at the wound site. Such a failure does not occur in control animals that have not received metronidazole [-Met, (H) to (J)] or those larvae that do receive metronidazole (+Met) but do not carry the Tg(uas:E1b:Eco.NfsB-mCherry) transgene and therefore do not express nitroreductase (-nfsb) [(E) to (G)]. Representative images of n > 20 animals per treatment group and probe. Insets show high-magnification views. (N) Quantitation of myogenic gene expression of ablated and nonablated larvae. Larvae were placed into arbitrary bins of expression (high, medium, low, or none), examples of which are shown in (E) to (M). (O to U) Ablation of cmet-positive motor neurons inhibits regeneration. (O) Transgenic line Tg(cmetMN:GFP-T2A-KALTA4) containing a subset of the cmet-regulatory regions evident in the full-length BAC clone expresses KALTA4 only within the motor neuron–expression domain. Crossing this line to Tg(uas:E1b:Eco.NfsB-mCherry) results in the expression of the nitroreductase enzyme (red) specifically within cmet-positive motor neurons. (P) Addition of Met to these embryos results in complete ablation of primary motor axons (positions marked by arrows) and their arbors, leaving only nonspecific mCherry cellular debris after ablation. NT, neural tube. (Q to U) Larval muscle regeneration is not impaired in motor neuron–ablated animals [+Met +injury, (S)] and is identical to injured nonablated controls [-Met +injury, (R)]. By contrast, ablation of both motor neurons and cmet-positive muscle stem cells in the cmet BAC transgenic line impairs regeneration (+Met +injury). (Q to T) Representative images of larvae under indicated conditions imaged for birefringence under polarized light. (U) Quantitation of birefringence levels relative to uninjured controls reveals that motor neuron–ablated larvae regenerate similarly to nonablated controls. However, ablation of both motor neurons and cmet-positive muscle stem cells in the cmet BAC transgenic line results in significantly lower birefringence, which points to impairment of muscle regeneration. Error bars indicate mean + SEM. Statistics: t test, two-tailed. ****P < 0.001; ***P < 0.001; **P < 0.001.

Asymmetric divisions generate clonally restricted progenitors

To examine how the cmet-positive cell population is activated during injury, we undertook continuous time-lapse analyses of the cmet-positive cells in several transgenic backgrounds that had undergone laser ablation injury. Analyses within injured animals double-transgenic for Tg(myf5:GFP) and Tg(cmet:mCherry-T2A-KALTA4) revealed that cmet and myf5 double-positive cells are associated with the injury site and appear as doublets within cmet-only cells (Fig. 5, A to D, and fig. S9E, n = 6 time-lapse analyses of injured larvae, n = 35 doublets identified). These results suggest that cmet-positive cells are activated by injury to divide asymmetrically to produce a myf5-positive progenitor population that, in turn, generates the de novo myogenesis evident during muscle regeneration described above. Collectively, our results suggest that the myf5-positive cells observed in earlier experiments contained in this paper (Figs. 1H and 2A and movies S4 and S5) are, in fact, cmet-low/myf5-high daughter cells that are responding to signals from the injury site and from regenerating myofibers. Next, we undertook time-lapse imaging within injured animals double-transgenic for Tg(pax7a:GFP) and Tg(cmet:mCherry-T2A-KALTA4) to capture the earliest division of cmet-positive cells. cmet and pax7a double-positive cells migrate to the wound site and proliferate, indicating that cmet marks injury-responding cells during muscle regeneration. Division of activated cmet and pax7a cells was asymmetric, such that after each division a rounded cmet-high–/pax7a-low–expressing cell was maintained and a pax7a-high– or cmet-low–expressing progenitor population was created (Fig. 5E, movies S6 and S7, and fig. S9). Individual cmet-positive cells were identified that gave rise to multiple cmet-low– and pax7a-high–expressing progeny (n = 6 time lapses of injured larvae, n = 38 doublets identified; fig. S9 and movies S6 and S7). Furthermore, our analyses indicate that pax7 cells are proliferative within the wound site at 2 dpi, suggesting that cmet-low/pax7a-high cells proliferate, elongate, and undergo differentiation in the manner we have described above (Fig. 5E; movie S6; and fig. S6, D and H). Collectively, these observations suggest that the cmet-high/pax7a-low cells evident within our time-lapse analyses represent a regeneration-specific stem cell compartment that divides asymmetrically to self-renew the stem cell compartment and produce proliferative daughters cells required to effect regeneration.

Fig. 5 Zebrafish satellite-like cells represent a lineage-restricted injury-responsive stem cell that undergoes asymmetric division during regeneration.

(A to D) Maximum-intensity-projection confocal images of Tg(myf5:GFP) (green) and Tg(cmet:mCherry-T2A-KALTA4) (red) double-transgenic fish 2 days after needle injury. Double-positive cells (arrows), cells with Tg(myf5:GFP) alone (asterisks), and cells with Tg(cmet:mCherry-T2A-KALTA4) alone (arrowheads) are marked. Insets reveal the existence of cells with Tg(cmet:mCherry-T2A-KALTA4) alone and double-positive doublet cells associated with the injury site, illustrated in the region framed in the combined differential interference contrast image in (D). Representative images were taken from n = 6 independent fish. (E) Asymmetric division generates a cmet-high/pax7-low stem cell compartment and a cmet-low/pax7-high differentiation-competent progenitor population after laser ablation injury. Continuous confocal time-lapse imaging of 4-dpf larvae immediately after laser injury shows cells transgenic for Tg(cmet:mCherry-T2A-KALTA4) (red) and Tg(pax7a:GFP) (green), as well as those imaged in the brightfield channel (insets without brightfield). A single cmet-high/pax7-low cell (arrow) maintains cmet expression through several rounds of division that generate cmet-low/pax7-high cells (arrowheads), which subsequently elongate and differentiate. Time (t) is given in minutes. Scale bar, 50μ m. (F to H) Needle-stick injury of larvae transgenic for Tg(ubi:zebrabow) and Tg(cmet:mCherry-T2A-KALTA4) injected with a UAS:CreERT2 DNA construct. Tamoxifen addition results in limited recombination events generating a highly restricted set of colors in regenerating fibers and myoblasts. Red is the default expression color that appears in every cell; the expression of any other fluorophore represents a specific recombination. Only two regenerating lineages are evident: one from a recombination of the zebrabow cassette that generates “blue” fibers [marked by arrows in (G)] and myoblasts and a second that generates “pink” fibers [marked by arrowheads in (G)] and myoblasts. (G) High-magnification view of the region boxed in (F). (H) Cross section at the level marked in (G). Representative images are from n = 14 imaged larvae. Scale bar in (F), 50μ m. (I to Q) Injured larvae transgenic for Tg(msgn1:CreERT2) and Tg(ubi:zebrabow), to which tamoxifen addition results in clonal labeling of all somite-derived cells. (I) Quantitation of the clonality of muscle regeneration. To statistically evaluate the clonal nature of regeneration, the hue (radians) and saturation (percentage) values were manually determined for each fiber and plotted such that each coordinate represents the color identity of a particular fiber. Distances between all coordinates were then calculated to provide a statistically evaluated measure of clonality. Error bars indicate mean + SEM. Statistics: t test, two-tailed. ***P < 0.001. (J to M) Confocal sections through an uninjured larvae at 4 dpf, revealing the specific differentiated muscle fibers clonally marked by individual zebrabow rearrangements. (N to Q) The same larvae as in (J) to (M) (at 3 dpi and 7 dpf), revealing that regenerating fibers all express the same zebrabow green rearrangement (regenerated fibers bounded by white line), which indicates that regenerating fibers are clonally related and derive from a single stem cell {sagittal [(J) and (N)], coronal [(L) and (P)], and transverse [(K) and (O)] sections at the levels indicated}. Scale bars, 50 μm. Representative images were taken from n = 13 independent fish. (M and Q) Plotted spectral values (dots) of individual fibers from the same larvae within uninjured (M) and regenerated (Q) muscle, revealing the comparative clonality of regenerating fibers after injury.

In vitro studies have suggested that satellite cells divide in an apical-basal fashion relative to the plane of the muscle fiber, a process that subsequently dictates the fate of daughter cells (5). More recent analyses have defined a role for the dystrophin-associated glycoprotein complex in the control of this process whereby dystrophin localization within the stem cell niche interacts with the PAR complex to regulate asymmetric division of the satellite cell (37). To determine whether a similar mechanism could control asymmetric division seen in our in vivo model, we used a gene-trap line driving the citrine fluorophore under the native dystrophin locus, Gt(dmd-citrine) (38), in combination with our Tg(cmet:mCherry-T2A-KALTA4) line. Examination of quiescent satellite cells showed us that dystrophin is specifically expressed on cmet-positive cells and is localized asymmetrically, but this asymmetry appears random relative to muscle fiber orientation (fig. S8, M to N′′′), which suggests that the initial asymmetric cell divisions of the activated stem cell would likewise be in multiple orientations. This suggests that local “micro niches” could provide specific extracellular matrix cues that dictate the localization and polarity of the stem cell divisions initially.

Given the low number of these cells evident in the larval myotome (~15% of pax7a- or pax3a-positive cells in the myotome; Fig. 3, I and J), the asymmetric model of stem cell division indicates that a finite number of regeneration-specific stem cells, represented by cmet expression, are activated after injury, and that these cells would generate a limited set of clonally related lineages to effect repair. Clonality can also be predicted from several of the current models of satellite cell action in amniote muscle repair that propose the existence of a small subset of satellite cells with true stem cells properties (79, 14). Specifically, the asymmetric model suggests that particular satellite cells maintain the ability to return to quiescence via asymmetric DNA segregation followed by differential gene expression, cell cycle entry, and metabolic status. Thus, this model indicates that all myogenic cells generated during repair would be clonally related to rare “stem” satellite cells (79, 14). However, this clonal model of stem cell action lacks in vivo validation. To provide evidence for this mode of satellite cell action and validate our own observations, we crossed the Tg(cmet:mcherry-T2A-KALTA4) line with the ubiquitously expressed Tg(ubi:zebrabow) line (Fig. 5, F to H, n = 22 fish) (39). These double-transgenic embryos were further injected with a UAS:CreERT2 construct, which is responsive to the KALTA4 promoter. We then performed needle-stick injuries on these animals. Two outcomes were possible from this experiment. If cmet expression simply reflects a progenitor phase of muscle proliferation, then Cre-activated deletion in cmet-positive cells would result in multiple deletions within independently proliferating cells and, hence, a wide range of distinct zebrabow rearrangements leading to many different colored myoblast lineages. However, if cmet expression denotes a stem population that generates a clonally related progenitor compartment, Cre-mediated deletion would generate rearrangements within a limited number of stem cells. In turn, these stem cells would generate the bulk of proliferating myoblasts in the regenerate, with progeny of individual zebrabow-recombined stem cells all expressing an identical color upon differentiation into muscle cells. Upon needle-stick injury of the Tg(cmet:mcherry-T2A-KALTA4), Tg(ubi:zebrabow) double-transgenic line into which UAS:CreERT2 was injected, only a very few distinct colors are evident that cluster in homogeneously expressed and regionally localized clones of myoblasts and muscle cells produced during regeneration (Fig. 5, F to H, n = 22 fish). This result indicates that the cmet-positive cells represent the injury-responsive clonal stem cell compartment of the zebrafish myotome (fig. S10J).

To more broadly assay the clonal nature of muscle regeneration, independent of the cmet stem cell compartment, we performed needle-stick muscle injuries in animals that were doubly transgenic for Tg(ubi:zebrabow) and Tg(msgn1:CreERT2). The addition of tamoxifen to these fish during embryogenesis results in zebrabow rearrangements in all cells derived from the somites, as the mesogenin (msgn1) promoter is somite-specific (40). Because the somite is the embryonic compartment from which all muscle stem and progenitor cells derive, this combination of transgenes enables the generation of a “musclebow” fish, allowing the clonal relationships of muscle cells to be assessed (Fig. 5, J to Q; fig. S10, A to I, n = 13 fish; and movie S8) (39). Tamoxifen addition to uninjured embryos early during embryogenesis results in the labeling of muscle fibers within the embryo and reveals random deletions of the zebrabow array, leading to individual fibers expressing unique combinations of fluorophores (Fig. 5, I to M; fig. S10, A to I; and movies S8 and S10, n = 13 fish). No bias in any specific zebrabow deletion is detected by this analysis, with individual fibers displaying random spectral values upon differentiation (Fig. 5, I to M; fig. S10, A to I; and movies S8 and S10). Therefore, injury in these animals allows for global assay of the number of somite-derived lineages that contribute to zebrafish muscle regeneration. To statistically evaluate the clonal nature of regeneration, we manually determined the hue (radians) and saturation (percentage) values for each fiber and plotted them such that each coordinate represents the color identity of a particular fiber. We then calculated the distances between all coordinates to provide a statistically evaluated measure of clonality (Fig. 5I; see methods for computational methodology). Our analysis indicated that, in stark contrast to muscle fibers that result from growth, regenerated muscle fibers are highly clonal, as a limited numbers of recombined fluorescent phenotypes contribute to regeneration, consistent with a stem cell origin for these cells (Fig. 5, N to Q; fig. S10, A′ to I′; and movies S9 and S10). However, multiple clones, contributing individual single clonally restricted lineages, can contribute to regeneration of larger wounds, eliminating the possibility that a single clonal lineage marked early in development generates all satellite cells in zebrafish (e.g., fig. S10, E, E′, I, and I′). Furthermore, our analysis confirmed that the stem cells that contribute to regeneration derive from the somite, as these cells exhibit a recombined fluorescent phenotype generated from the expression of the somite-specific meosgenin Cre line. Collectively, our analyses, combined with previously published results (79, 14), indicate that vertebrate muscle regeneration results from the activation of limited set of somite-derived stem cells that generate a clonally restricted proliferative myoblast population, which in turn differentiates to effect muscle repair.


We analyzed the morphological, cellular, and genetic basis for muscle regeneration in zebraish. Using transgenes that mark muscle stem and progenitor cells, in combination with long-term imaging modalities, we imaged the entire process of regeneration, from the initial injury to the formation of newly differentiated muscle fibers at the wound site in a vertebrate. We show that a stem cell niche equivalent to the mammalian satellite cell system operates within the zebrafish myotome. This observation suggests that the satellite cell is part of an evolutionary ancient stem cell system that is present throughout the vertebrate phylogeny and may well be a basal feature of gnathostomes.

Our time-lapse analyses enabled us to characterize a number of stereotypical phases of repair involving specific cell cohorts (Fig. 6 and fig. S10J). Although many of these morphogenetic events are canonical ones, in terms of known stem cell–driven regenerative processes (activation, proliferation of a progenitor population, and differentiation), the dynamic nature of the morphogenetic changes of both the activated stem cell and the proliferating progenitor population is surprising. Although migration of activated satellite cells on fibers has been documented in vitro (41), there has been no indication from such paradigms that the migration of activated muscle stem cells would deploy such a polarized phenotype. The polarized shape that myf5-positive cells adopt to migrate to the site of injury is therefore suggestive of a novel directional migratory mode. A recent examination of the role of activated pax7-positive cells during in vivo muscle repair in the mouse revealed similar polarized phenotypes for activated migrating murine myogenic cells (10), indicating the potential relevance of our observation to mammalian muscle biology. Due to current limitations of examining this process in vivo, little other information exists as to the mechanism used by migrating activated satellite cells to track to the wound site in amniotes. These limitations are largely overcome by use of the zebrafish larval system.

Fig. 6 A model for in vivo muscle regeneration.

(A to J) In vivo repair of muscle from a defined stem population occurs through a number of stereotypical phases. The initial phase (migration and contact) is characterized by migration of the satellite-like cell populations into the wound area. Once in the wound, cells immediately associate as rounded cells with severed and dying fibers and remain in contact for several hours. In the second phase (proliferation and activation), activated stem cells undergo an asymmetric division to generate the proliferating compartment that undergoes further division. In the next phase (bipolar extension and interaction), proliferating cells extend to a bipolar shape and lose contact with the dying cells around them. Bipolar cells in the wound then actively migrate toward each other, extend filopodia to make contact, and then recoil and move rapidly apart from one another. Next, resident uninjured differentiated muscle cells extend filopodial-like membrane protrusions, which adhere and lasso the activated cells. Filopodial retraction guides the captured cell to align next to the differentiated muscle cell from which the filopodia extends. Cells then elongate and differentiate into muscle fibers, marking the end of the final phase (resident fiber capture and differentiation). Within 7 days, muscle fiber repair is complete, and the injury site is indistinguishable from surrounding uninjured fibers.

Furthermore, the interplay between the proliferating myogenic cells that enter the wound site with injured fibers in the initial contact phase of regeneration and secondarily with the undamaged remaining fibers at the injury site was also unexpected. A lack of information is available about how the injury site itself acts to direct the architecture and behavior of the regenerating fibers during skeletal muscle regeneration, as existing studies have almost exclusively focused on the factors that influence the stem cell compartment. However, the importance of the wound environment has been emphasized in recent in vivo investigations performed during early muscle repair in the mouse, which showed that myogenic progenitors at the injury site use the extracellular matrix left from dead and damaged muscle fibers to provide orientation cues for cell division and alignment (10). In this regard, the “lassoing” event that appears to guide differentiating cells at the wound by the surrounding uninjured fibers was an extension of these observations. It suggests that instead of being passive bystanders in the regenerative process, uninjured fibers themselves may play a role in directing differentiating progenitors to regions of the wound that are most in need of new fiber addition. It also suggests that the kinetics of repair at the wound, and even aspects of myocyte differentiation, may partly be dictated by the nature and availability of intact fibers to participate in this process.

Using the cmet and pax7 transgenes as markers of the stem cell population, we have been able to examine activated cells immediately after injury by continuous time-lapse analyses. Our analyses detected the existence of two clonally related sets of cells during regeneration: a cmet-high/pax7-low stem cell population and a pax7-high/cmet-low progenitor population. Time-lapse analyses revealed that the cmet-high population divided asymmetrically to generate the differentiation-competent cmet-low population, as well as a cmet-high cell that maintained the ability of producing both cmet-low proliferative daughter cells and a self-renewing cmet-high undifferentiated stem cell. Previous analyses of murine single-fiber explants have suggested that the plane of division of a satellite cell, relative to the underlying basal lamina, dictates the fate of individual satellite cell daughters (5). Specifically, division of a satellite cell in the apical-basal plane relative to the muscle fiber generates an apically positioned, differentiation-competent progenitor and a basally located self-renewing stem cell. A more recent imaging study of in vivo mouse muscle repair has suggested that activated myogenic progenitors at the wound site divide in a planar fashion, although this study was unable to assay the process of satellite cell activation (10). Although the highly migratory nature of the activated satellite cells in vivo precludes a definitive assignment of relative polarity of division, our observations suggest that the initial division of activated satellite cells probably occurs in a random polarity with respect to the plane of the muscle fiber in zebrafish. Furthermore, the clonality of regeneration we observe in response to muscle injury is a major finding of this work. Clonality is a direct prediction of the self-renewing immortal stem cell hypothesis generated from seminal in vitro observations (59), and our work provides direct in vivo validation of this hypothesis. We conclude that asymmetric divisions occur during muscle regeneration to generate the clonally related progenitors required for muscle repair, resolving a long-term debate surrounding the existence of this mechanism of stem cell self-renewal and muscle repair in vivo.

Materials and methods

Zebrafish strains and maintenance

Wild-type (WT) and transgenic lines Tg(-80.0myf5:eGFP) [referred to as Tg(myf5:GFP)] (42), Tg(actc1b:mCherry) (43), Tg(actc1b:BFP) (43), Tg(actc1b:mCherryCAAX) (44) ,Tg(smyhc1:GFP) (45), TgBAC(pax7a:GFP) [referred to as Tg(pax7:GFP)] (12), TgBAC(pax3a:GFP) [referred to as Tg(pax3:GFP)] (12), Tg(ubi:zebrabow) (39), Tg(msgn1:CreERT2) (40), Tg(UAS:nsfB-mCherry) (46), Tg(cmetMN:GFP-T2A-KALTA4) (34), Gt(dmd-citrine)ct90a [(referred to as Gt(dmd-citrine)] (38), Tg(actc1b:Cre), and Tg(ef1a1:LOXP-eGFP-LOXP-mCherry) were maintained on TE WT background. Staging and husbandry were performed as previously described (47). A new transgenic line produced by microinjection into TE WT fish according to standard procedures (48) and used in this study is TgBAC_CH211-105D03(cmet:mCherry-T2A-KALTA4) [hereafter referred to as Tg(cmet:mCherry-T2A-KALTA4)], a BAC construct. Mutant strains used were myf5hu2022 (28), myf6hu2041 (28), myogfh265 (27), and myodhu2024 (49), all maintained on AB background. These mutants were genotyped by sequencing of polymerase chain reaction products amplified from fin clip or embryo genomic DNA using primers 5′-CATATAAGAAGGGCCGTATG-3′ and 5′-ATGCAGAGCTTCACTGTAGG-3′ for myf5, 5′-AAATGTTTAACTGTTATGCAAAG-3′ and 5′-TGTCGTTAAAGGTCGGATTC-3′ for myod, 5′-AGCGCTGGATAGAATTTCAC-3′ and 5′-CTGTGCGCAGAGTCAAAG-3′ for myf6, and 5′-AGGCTCTGAAGAGGAGCACA-3′ and 5′-GTCAGAAATAAAGCTGTCACTGAG-3′ for myog.

Gateway cloning and mosaic transgenesis

Gateway cloning was performed using the Gateway Tol2kit to create transgenic constructs, according to standard Invitrogen protocols (50). The UAS:CreERT2 and actc1b:nlsGFP constructs were created using the Gateway method. To generate mosaically transgenic zebrafish, individual plasmid DNA–containing constructs flanked by tol2 sites were delivered by microinjection into embryos at the single-cell stage, directly into the cell. The plasmid DNA concentration was typically at 25 ng/μl. The Tg(actc1b:Cre) and Tg(ef1a1:LOXP-eGFP-LOXP-mCherry) transgenic fish were generated from constructs also made using the Gateway method. To generate stable transgenic zebrafish, a mixture of mRNA-encoding tol2 recombinase and a DNA plasmid containing a construct flanked by tol2 sites was delivered by microinjection directly into the cell at the one-cell stage of development. mRNA was delivered at 20 ng/μl and the DNA construct at 25 ng/μl.

Needle-stick injury

For needle-stick injury induction, tricaine-anaesthetized 4-dpf larval zebrafish were placed on a 24.5 mm–by–76.2 mm glass slide in a drop of water and maneuvered into a lateral lying position with their heads pointing toward the left. A single needle-stick injury into the dorsal epaxial musculature above the cloaca was performed at an angle of 75°, using a 30-gauge needle. Subsequently, fish were immediately transferred to fresh embryo water (E3) to recover. The ventral side of the fish relative to the site of injury was used as an uninjured internal control.

Laser injury

For laser-induced injury, tricaine-anaesthetized 4-dpf larval zebrafish were placed in grooves on a specialized silicon block, similar to the one used for microinjection of embryos, mounted in a petri dish. A thin layer of low–melting point agarose [0.5% weight/volume agarose in E3 (Invitrogen)], maintained in a molten state at 36°C, was applied to each fish. As soon as the low–melting point agarose was set, the petri dish was filled with E3 containing tricaine to maintain the anaesthetized state throughout the injury procedure and for subsequent time-lapse photomicroscopy. A Micropoint laser connected to a Zeiss Axioplan microscope was used for injuring, as previously described (51), with a 40X water immersion objective and a laser pulse at a wavelength of 435 nm for cell ablation. Time-lapse recordings were generated on a Zeiss LSM 710 Live Duo confocal microscope (Carl Zeiss Microimaging).

Metronidazole ablation experiments

Cell ablation using this system was performed as previously described (52), with some modifications. Briefly, the prodrug metronidazole (Sigma) was dissolved in E3 water as a 5 or 1 mM working solution, as indicated, along with added 0.2% dimethyl sulfoxide and 0.003% 1-phenyl-2-thiourea in 10% Hank’s saline (PTU). Transgenic zebrafish were incubated in this solution immediately after injury (4 dpf) at 28.5°C, solutions were changed every day, and fish were euthanized and fixed at 7 dpf for whole-mount in situ hybridization (WISH). Control fish were subjected to the same injury and treatment procedure without metronidazole.

WISH and immunohistochemistry

In situ mRNA hybridization and immunohistochemistry were performed as previously described (31). The digoxigenin-tagged probes used were pax7a (53), cmet (31), myf5 (54), mrf4 (55), and myod and myog (56). Analyses of expression levels after in situ hybridization involved randomly mounting stained embryos, which were blinded to treatment group. Larvae were photographed and scored into arbitrary bins of “none,” “low,” “medium,” and “high,” on the basis of the experimental spread of expression. Images representative of the overall phenotype were then presented in the data. Primary antibodies used were rabbit anti-GFP [1:500 (Invitrogen)], mouse anti-Pax7 [1:10 (Developmental Studies Hybridoma Bank, DSHB)], mouse F59 anti–slow myosin heavy chain [1:10 (DSHB)], and rabbit anti-phospho-histone-H3 [1:1000 (Sigma)]. The p16 INK antibody (1:4) (57) probably recognizes proteins that are produced from a single locus ancestral to mammalian CDKN2A and -2B, as only a single syntenic gene is present in zebrafish. Hence, this antibody could recognize a range of potential cyclin inhibitors. Secondary antibodies used include goat anti-rabbit Alexa Fluor 488 [1:1000 (Invitrogen)], goat anti-rabbit Alexa Fluor 568 [1:1000 (Invitrogen)], goat anti-mouse Alexa Fluor 488 [1:1000 (Invitrogen)], goat anti-mouse Alexa Fluor 568 [1:1000, (Invitrogen)].

EdU labeling

Cell proliferation was determined by 5-ethynyl-2′-deoxyuridine EdU labeling according to established conditions (58), using 7 mM EdU (Invitrogen) dissolved in E3 medium. Briefly, larval fish were incubated for a 48-hour block before fixation. Fish were euthanized, fixed, and subsequently processed using the Click-iT EdU Imaging Kit (Invitrogen). This processing occurred after antibody staining, using the Click-iT solution.

TUNEL staining

Cell apoptosis was determined by terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate biotin nick end labeling (TUNEL) staining, performed on whole larvae as previously described (52), with some modifications. To stain for apoptosis, a TMR Red In Situ Cell Detection Kit (Roche) was used according to manufacturer’s specifications. Samples were incubated in stain solution in the dark for 2 hours at 37°C before processing and observation by fluorescence and confocal microscopy.

Birefringence imaging

Birefringence is the property of intact muscle tissue to diffract polarized light (damaged muscle tissue lacks this quality). Birefringence was measured as previously described (59). Briefly, anaesthetized zebrafish were placed on the Leica DMIRB (Leica Microsystems) microscope stage in a glass-bottom FluoroDish (World Precision Instruments). The exposure levels of the attached polarizing filters were automatically adjusted by the integrated Abrio Software [CRI (Hinds Instruments)], and the resulting final picture shows an averaged polarizing effect. The resultant birefringence images were analyzed for birefringence densitometry over the required area using the software FIJI.

Zebrabow analyses

Tamoxifen treatments were conducted as previously described for the Tg(ubi:zebrabow), Tg(msgn1:CreERT2) double-transgenic larvae (40). At 4 dpf, the fish were imaged and subsequently injured by needle stick. At 7 dpf, fish were reimaged at the injury site. Optical transverse cross sections were then generated from confocal stacks with Imaris (Bitplane), and color profiles of regenerated fibers at the injured site were measured in Photoshop (Adobe) as previously described (39). For clonal analysis, images were processed as described in (39), with modifications. Images were opened in Photoshop CS5.1, and the eyedropper tool (sample size set at 5 by 5 pixels) was used to measure the fiber of interest. Color display mode was set to hue saturation brightness, and the hue and saturation values in the color window were manually recorded. Hue (radians) and saturation (percentage) values were converted into x-y coordinates to be plotted on a Cartesian graph using basic trigonometry (hue, degrees; saturation, hypotenuse; adjacent or opposite side, point of interest). Each coordinate represented the color identity of a particular fiber. Distances between all clones were then calculated with following formula: Embedded Image. Distance values were then imported into Prism software to conduct a Student’s t test. For the Tg(cmet: mCherry-T2A-KALTA4); (UAS:CreERT2); Tg(ubi:zebrabow) transgenic larvae, tamoxifen was added at 4 dpf, immediately before injury, and samples were incubated postinjury for a further 8 hours before washing into embryo media.

Supplementary Materials

Figs. S1 to S10


Movies S1 to S10

References and Notes

  1. Acknowledgments: We thank P. Ingham, A. Schier, G. Lieschke, L. Bally-Cuif, and the Zebrafish International Resource Center for the supply of fish strains and reagents; R. Cheney for technical advice; N. Cole for technical assistance; and C. Marcelle and T. Rando for comments on the manuscript. This work was supported by a National Health and Medical Research Council of Australia grant to P.D.C. The Australian Regenerative Medicine Institute is supported by funds from the state government of Victoria and the Australian federal government. cmet transgenic lines are available from P.D.C. under a material transfer agreement with Monash University.
View Abstract

Navigate This Article