Mononuclear Cells from Dedifferentiation of Mouse Myotubes display Remarkable Regenerative Capability

Abstract

Certain lower organisms achieve organ regeneration by reverting differentiated cells into tissue-specific progenitors that re-enter embryonic programs. During muscle regeneration in the urodele amphibian, post-mitotic multinucleated skeletal myofibers transform into mononucleated proliferating cells upon injury, and a transcription factor-msx1 plays a role in their reprograming. Whether this powerful regeneration strategy can be leveraged in mammals remains unknown, as it has not been demonstrated that the dedifferentiated progenitor cells arising from muscle cells overexpressing Msx1 are lineage-specific and possess the same potent regenerative capability as their amphibian counterparts. Here we show that ectopic expression of Msx1 reprograms post-mitotic, multinucleated, primary mouse myotubes to become proliferating mononuclear cells. These dedifferentiated cells reactivate genes expressed by embryonic muscle progenitor cells and generate only muscle tissue in vivo both in an ectopic location and inside existing muscle. More importantly, distinct from adult muscle satellite cells, these cells appear both to fuse with existing fibers and to regenerate myofibers in a robust and time-dependent manner. Upon transplantation into a degenerating muscle, these dedifferentiated cells generated a large number of myofibers that increased over time and replenished almost half of the cross-sectional area of the muscle in only 12 weeks. Our study demonstrates that mammals can harness a muscle regeneration strategy used by lower organisms when the same molecular pathway is activated.

Introduction

Unlike the strategies which have led to the current nuclear reprogramming protocols to create pluripotent cells from differentiated cells or converting lineage committed cells to mature cells of other lineages[1–3], urodele amphibians and zebrafish regenerate lost organs using a different nuclear reprogramming strategy. In response to injury, their differentiated cells re-enter the cell cycle and instead of acquiring pluripotency, the dedifferentiated cells retain their original tissue identities and reform these specific lost tissues during regeneration[1, 4, 5]. Whether it is possible to leverage this primitive regenerative strategy to induce new tissue and organ formation in mammals has been a longstanding question, however, there has been much debate about this possibility.

In the past few years, studies using different methods demonstrated that post-mitotic mammalian multinucleated myotubes could be induced to dedifferentiate into mononuclear proliferating cells. Ectopic expression of Msx1, the transcription factor that is up-regulated and drives muscle cell dedifferentiation in urodele amphibians[6], has been shown to dedifferentiate multinucleated C2C12 myotubes into proliferating mononuclear cells[7, 8]. Interestingly, these dedifferentiated mononuclear cells display properties that were more primitive than C2C12 cells. However, due to the tumorigenic nature of C2C12 cells, whether these dedifferentiated mouse muscle cells possessed the regeneration capability of their amphibian counterparts was not explored. In an attempt to induce dedifferentiation without overexpression of Msx1, Pajcini et al. showed that concomitant transient inactivation of Arf and Rb led mammalian myotubes (myocytes) to cellularize and re-enter the cell cycle. The mononuclear clones derived from these myotubes were capable of fusing with existing muscle[9]. It has also been shown that treatment of differentiated muscle cells with small molecules such as the cyclohexylaminopurine reversine, induces a proliferative response, mainly though down-regulation of cyclin-dependent kinase inhibitors or tyrosine phosphatases[10, 11]. These cells have been shown to be multipotent, and are able to fuse into existing muscle after cardiotoxin injury. More recently it was shown that down-regulation of myogenin, one of the myogenic regulatory factors, can reverse the differentiation state of terminally differentiated mouse myotubes and initiate their cleavage into mononucleated cells[12]. However, whether or not these dedifferentiated mammalian muscle cells possessed long term regeneration capability that is similar to their amphibian counterparts was not explored.

We therefore set out to examine whether ectopic overexpression of Msx1 could drive primary multinucleated murine myotubes to re-enter the cell cycle and furthermore to determine if and how these dedifferentiated progenitor cells regenerate skeletal muscles after transplanting them into different in vivo microenvironments.

Materials & Methods

Cell Culture and Gene Transduction

Primary myoblasts were isolated from hind limb muscles of 4-week-old C57BL/10 male mice as described previously[64, 65]. Cells were expanded in Ham’s F10 medium supplemented with 20% fetal calf serum and 5ng/ml basic fibroblast growth factor (bFGF) (Growth Media) on collagen-coated plates. Before transduction, the myogenic identity of cells was verified with anti-desmin antibody through immunocytochemistry.

Retroviral vectors LINX-Msx1-fwd and LINX-Msx1-rev (kind gifts of Dr. Shannon Odelberg) were packaged as described elsewhere and the sequence was driven by a Tet-off inducible system[8]. Primary myoblasts at passage 5 were transduced with either LINX-fwd or -rev virions. The transduced cells were selected using G418 and expanded clonally in the presence of doxycycline 3 μg/ml. A proportion of cells from all 3 groups were transduced with either eGFP or nl-GFP lenti-viral virions as described elsewhere[66]. The nl-GFP is specifically targeted to the nuclei and was used to visualize the myonuclei in real-time microscopic imaging in the present study (supplemental movies). eGFP transduced cells were used for the ectopic and intramuscular cell transplantation experiments in SCID mice.

Induction of Myotube Dedifferentiation

To induce myogenic differentiation of primary myoblasts, cells were cultured in differentiation medium (DMEM with 2% horse serum) with 3μg/ml doxycycline on collagen-coated plates. On the 4th day differentiation in the presence of doxycycline, myotubes were removed from the plate with trypsin, filtered through a 100μm filter and then through a 40μm secondary filter. The remaining myotubes on the sieve were collected and then diluted with differentiation medium. Single myotubes were picked up manually by a pipette and placed in 48 well plates with 1 myotube/well. On the next day all wells were examined carefully by visual inspection for the presence of any residual mononuclear cells (Fig. S2). Contaminated mononuclear cells were destroyed using a 25Ga needle, or the wells were excluded from further observation. The induction of dedifferentiation was then initiated by withdrawing doxycycline and switching to growth medium with addition of 1x Insulin-Transferrin-Selenium(Invitrogen). The area containing each myotube was circled on the bottom of the plate (Fig. S2).

In a separate set of experiments, the myotubes were cultured in Methylcellulose semisolid medium. MethoCult (StemCell Technologies, British Columbia, Canada) mixed with liquid growth medium in 40:60 ratio with the final concentrations of FCS and growth factors adjusted to be the same as liquid growth medium.

Once the dedifferentiation occurred, the mononuclear proliferating cells in the well were picked out after a typical clone was formed. These cells (MIDCs) were then cultured in the growth medium supplemented with LIF 100 μl/10ml on non-coated plates.

Real-time Microscopic Image Acquisition of Myotube Dedifferentiation

For microscopic time-lapse studies on a 37°C heated stage (Omega Engineering Inc., Stamford, CT), 100–200 myotubes were individually placed in a 35mm collagen-coated dish and cultured in HEPES buffered Hams F10 media with no HCO₃ (Invitrogen Inc., Carlsbad, CA) having a mineral oil lid. For studies in an on-stage mini-CO₂ incubator (Pathology Devices, Westminster, MD), myotubes were plated in collagen-coated 48 well plastic dishes in standard growth media. Imaging was accomplished on an automated Nikon TE2000E inverted fluorescence microscope (Nikon USA, Inc.). Excitation illumination was supplied by an X-Cite mercury lamp (EXFO, Mississauga, Ontario) and attenuated by a neutral density filter. Images were captured on a Roper HQ CCD camera (Photometrics, Inc., Tucson, AZ) using 4×4 binning for fluorescence and 2×2 binning for phase contrast images. IPLab software (BioVision Inc., Exton, PA) controlled automated image capture of 200–400 20X fields at each time-point. Time-lapse image acquisition was begun 8–12 hours after induction of msx-1 expression and continued for 25–66 hours at 30-minute intervals and was analyzed using IPLab software.

Multipotency Assays of MIDCs in Vitro

For adipogenic assays, cells were cultured in conditions described previously[67] and then subjected to Oil Red labeling. In osteogenic assays, cells were cultured in conditions described elsewhere without addition of BMP2[68] and then stained for alkaline phosphotase. In chondrogenic assays, cells were seeded on PEG polymer and cultured in the medium described previously[69] and examined for the presence of collagen II. For the myogenic assay, cells were cultured in differentiation medium for 4 days followed by immunocytochemistry for the expression of myosin heavy chain. During all of these assays, the expression of Msx1 was suppressed by using doxycycline beginning at day 0 for each assay.

Gene Expression Analysis by RT-PCR, Western Blot, Immunocytochemistry and Immunohistochemistry

For RT-PCR, total RNA was extracted using a GenEluteTM Mammalian Total RNA Miniprep kit (Sigma) and reverse-transcribed into cDNA using TaqMan Reverse Transcription Reagents (Roche). Primer sequences are listed in Supplemental Figure 18. For qRT-PCR, SYBR Premix Ex Taq™ Kit (TaKaRa) on ABI 7300 PCR and Detection System (Applied Biosystems) were used for qRT-PCR. Mouse GAPDH was amplified as an internal standard. The relative values were calculated using ΔΔCt method and normalized against endogenous GAPDH. Primer sequences are listed in Supplemental Figure 19.

For Western blot analysis, total proteins were extracted and expression levels of Pax3, Pax7, Myf5 and MyoD were measured using antibodies of anti-Pax3 (MAB2457, R&D, MN), Pax7 (Contributed by Kawakami, Hybridoma Bank, IO), MyoD (Sc-760, Santa Cruz, CA) in 1:1000, and Myf5 (ARP32134, Avivasysbio, CA) in 1:2000 dilutions. For immunocyto- and immunohisto-chemistry: Myotubes were immunostained with anti-mouse Myosin Heavy Chain (MEDCLA66, Accurate Chemical & Scientific Corp, NY) followed by incubation with a CyTm3-conjugated secondary antibody (Jackson Lab) (All secondary antibodies here and below were used at a dilution of 1:500). Muscle sections from injected mdx mice were incubated with anti-dystrophin (ab15277, Abcam, MA) in 1:500 dilution followed by incubation with a CyTm3-conjugated secondary antibody. Muscle sections of SCID mice were incubated with mouse-anti-Pax7 antibody in 1:50 dilution followed by the CyTm3-conjugated secondary antibody and then double labeled with rabbit-anti-Laminin (Sigma, MO) in 1:500 followed by incubation with Alexa-488-conjugated secondary antibody (Invitrogen).

Cell Transplantation, in vivo Fluorescence Imaging and Tissue Preparation

For cell transplantation, the cultured MIDCs and myoblasts were trypsinized and collected. Animal experiments were carried under the guidance and approval of the Harvard Medical School standing committee on animals. Three days prior to cell transplantation, the hind legs of 4–8 week old mdx mice (Jackson Lab) were given 18 Gy irradiation. Animals were anesthetized with Ketamine/Xylazine (100 mg/15 mg in 5 ml saline) at a dosage of 100 ml/20g of body weight and then 1×10⁵ MIDCs or the same number of control myoblasts in 10 μl HBSS were injected into TA muscles of mdx mice at 3 positions. Similarly, 1×10⁵ eGFP-labeled MIDCs were injected into the TA muscles of un-irradiated 8–12 week old SCID mice or nude mice (Taconic). The TA muscles of SCID mice were injured with cardiotoxin in a dosage 10 ng/muscle 3 days prior to cell injection. 1×10⁵ eGFP-labeled MIDCs or control myoblasts were injected subcutaneously at the thigh reign of SCID mice at one position. At 60 days post cell injection, cardiotoxin (10ng/leg, Sigma) was injected to selectively destroy myofibers formed by injected cells. All the animals were given doxycycline 1 mg/ml starting 3 days prior to cell transplantation. Suppression of Msx1 after transplantation was confirmed by RT-PCR (Fig. S17). The fate of eGFP-labeled cells was monitored by in vivo fluorescence imaging periodically. For in vivo fluorescence imaging, the injected SCID mice were anesthetized as described above. The imaging procedure and parameters were described as our previous report [39].

Mid-TA cross sections and in some cases serial cross sectional tissue samples were harvested at 2-, 4-, 8-and 12-weeks post cell injection and in age matched uninjected muscles from the 12 week group for histology analysis including analysis of whether donor derived cells formed fat, cartilage or osteogenic tissue. Mice were deeply anesthetized with xylazine and ketamine and had their left ventricles perfused with PBS and 4% (w/v) paraformaldehyde (Sigma). The TA muscles and sites of the subcutaneous engraftment were then carefully removed, post-fixed in 4% PFA over 12 hours, submersed in 30% sucrose overnight, frozen in OCT embedding compound, and Cryo-sectioned coronally (10 microns). Cryo-sections were thaw mounted onto gelatin-coated slides and stored at −20°C.

Statistical Analysis

Quantitative data are presented as mean±standard deviation (SD). A student’s t-test (assuming equal variances) was performed to determine the statistical significance between two experimental groups. A p value less than 0.05 was considered to be statistically significant.

Results

1. Ectopic overexpression of Msx1 reprograms terminally differentiated mouse myotubes to become proliferating mononuclear cells

We first explored whether differentiated mouse myotubes could be induced to reenter the cell cycle and undergo cellularization to form proliferating mononuclear cells by expressing Msx1. Primary myoblasts were transduced with retrovirus where either Msx1 expression was under the control of a tetracycline-off promoter or the vector containing a reversed Msx1 insert (Fig. S1). Permanently transduced clonal cells were expanded and differentiated into myotubes in the presence of doxycycline. Single myotubes were then isolated and cultured in either liquid or semisolid medium after exclusion of all mononucleated cells (Fig. S2). Analysis of individual myotubes over time revealed that 1 to 5% of these myotubes produced proliferating mononucleated cells in 1 to 5 days in both types of media, after initiation of the expression of Msx1 by removal of doxycycline and addition of growth media. No mononuclear cells were produced by myotubes expressing the reversed vector or that were not transduced (Fig. S3, Table 1).

Using static fluorescence microscopy (Fig. 1A) and real-time microscopic imaging, we were able to capture the detailed kinetics of transformation as Msx1 transduced myotubes underwent cellularization (Movies S4 to S6), and confirmed that this does not occur in control myotubes (Movie S7). Analysis of these recordings revealed that Msx1-induced cellularization results from extension of the leading edge of the myotube cells accompanied by nuclear migration, and finally by pinching off of a small portion of the cytoplasm containing a single nucleus. These newly formed mononucleated cells can translocate along the surface of existing multinucleated myotubes, or move freely over the culture substrate as single cells (Movies S4 & S6); Some of them also can undergo repeated rounds of cell division and form new mononucleated daughter cells that retain the ability to migrate individually (Movie S5 and S6). These results provide unequivocal evidence that expression of Msx1 can induce primary mouse myotubes to revert into proliferating mononuclear cells, as previously observed in myotubes formed by immortalized C2C12 cells [7, 8].

Figure 1. Evidence of Msx-1-induced dedifferentiation.

(A) Representative pictures of the kinetics of cellularization of Msx1-expressing myotubes. Cellularization begins 24 hr after induction of Msx-1 expression. This continues through day 2 and daughter cell expansion begins. By Day 3 cellularization is complete and daughter cell expansion continues, with a doubling time of less than 24 hours and a clone of daughter cells is ready for further expansion by day 5. For real time observations over 49.5 hours see Movie S5. (B) Western blot showing the expression levels of MyoD, Myf5, Pax7 and Pax3 in control myoblasts (Cont Mb), Msx1-expressing myoblasts (Msx1-Mb) and two MIDC clones (MIDC-C1 and C2). (C) Quantitative western blot analysis normalized with GAPDH. * P<0.05 and ω P< 0.01 vs Cont-Mb Student’s T test. n= 3 to 4. (D) Myotubes formed by Cont-Mbs and MIDCs stained with adult myosin heavy chain (MyHC) antibody. Scale Bar = 100μm.

2. Dedifferentiated mononuclear cells reactivate embryonic muscle progenitor genes

Comparison of gene expression profiles of mononucleated cells formed by Msx1 induced dedifferentiation of myotubes (MIDCs) with their both their parental myoblasts (where Msx1 expression was turned on but cells had not undergone differentiation/dedifferentiation) and control primary myoblasts demonstrated that the MIDCs exhibit a distinct phenotype. The myogenic determinant factor MyoD was down-regulated in MIDCs, whereas genes expressed at high levels in earlier stage myogenic progenitor cells, such as Myf5, Pax3 and Pax7, were up-regulated (Fig. 1B and C). Interestingly, expression of Msx1 in undifferentiated primary myoblasts also up-regulated these genes but to a lesser extent than the levels found in MIDCs (Fig. 1B and C). MIDCs also expressed the progenitor cell marker Sox2, which is not expressed in primary myoblasts or myoblasts expressing Msx1 (Fig. S8). Intriguingly, some MIDC clones transiently expressed low levels of pluripotent genes, Nanog, Oct4 and Rex1 but their expression diminished over time with passaging (Fig. S8).

With addition of specific growth factors MIDCs could be induced to redifferentiate into cells with adipogenic, chondrogenic or osteogenic characteristics, while primary myoblasts could not (Fig. S9). Attempts to induce forward differentiation of MIDCs into hepatocytes and neuronal cells failed, suggesting that their differentiation potential is restricted to mesodermal derivatives. Most importantly, when MIDCs were cultured under conditions (low concentration of serum and free of growth factors) that are used to differentiate primary myoblasts, they uniformly differentiated into multinucleated myotubes as indicated by their expression of myosin heavy chain (MyHC) (Fig.1D).

3. MIDCs only generate muscle tissue in vivo

To determine whether MIDCs are capable of forming de novo muscle tissue in vivo we injected eGFP-labeled MIDCs or control myoblasts subcutaneously in the thigh region of Severe Combined Immunodeficiency(SCID) mice (n=6/group). Using in vivo fluorescence imaging, we observed that the eGFP-signal reached a peak around 4-weeks after mice were injected with MIDCs (Fig. 2A). To prove that the e-GFP⁺ donor-derived cells were myofibers, 2 months after cell injection, we injected cardiotoxin into the eGFP⁺ region and followed the change by in vivo fluorescence imaging. After cardiotoxin injection, we observed a rapid decline in the eGFP signal, which was followed by recovery of the fluorescence after two weeks (Figs. 2A and S10). Immunohistochemical analysis confirmed that the injected MIDCs formed striated MyHC⁺ myofibers within the subcutaneous space, almost all of which had peripherally located myonuclei (Figs. 2C and D and S11). We found no donor-derived cells having migrated into the muscle below nor any donor-derived cells that were immuno-positive for adipogenic, osteogenic or chondrogenic markers, or T-cell factor 4 (TCF4) [13], a gene expressed by mesoderm cells other than muscle during embryonic development (data not shown). In contrast, in mice injected with control myoblasts, the eGFP signal rapidly declined and at 60 days post injection was completely absent (Fig. 2B). The GFP signal could not be recovered in response to cardiotoxin injection and postmortem histological examination revealed no eGFP⁺ ectopic muscle in these mice.

Figure 2. Generation of myofibers in an ectopic location.

(A) and (B) Representative in vivo fluorescence images after subcutaneous transplantation of eGFP-labeled MIDCs (A) or Cont-Mbs (B). (C) Longitudinal section of a MIDC-derived subcutaneous muscle harvested 3 months post-cell-injection. Upper panel, low magnification. Lower panel, high magnification of the framed area. Dual staining with MyHC and DAPI (for the location of myonuclei). (D) Cross section of 3-month MIDC-derived subcutaneous muscle. The right panels are high magnification of the framed area in left picture. The right most picture demonstrates peripheral myonuclei. Scale Bars=50μm.

4. MIDCs generate myofibers in a niche and time-dependent manner

To assess whether MIDCs engraft into already formed adult muscle, we injected MIDCs or control myoblasts intramuscularly into pre-irradiated tibialis anterior (TA) muscles of dystrophin-deficient mdx mice. In muscles harvested at 2, 4 and 8 weeks following injection, we detected an average of 166±12, 1283±158, and 3340±304 (n=5) dystrophin⁺ myofibers in each MIDC-injected muscle analyzed, whereas only 10±3 (n=5) dystrophin⁺ fibers were observed in muscles injected with control myoblasts at 2 weeks and this number did not increase with time (Fig. 3A). The dystrophin⁺ myofibers in the MIDC-injected group varied widely in size and formed clusters of dystrophin positive fibers. Although some of them had the size of host fibers, the majority of the fibers were small at earlier time points and some of them did not align properly. To further quantify the degree of regeneration by donor-derived fibers, in a separate group, we injected MIDCs into irradiated mdx TA muscles and analyzed dystrophin⁺ muscle fibers at 4 and 12 weeks post injection and compared the 12 week samples with cross sections of age matched mdx control muscles. We measured the total cross sectional area (CSA) occupied by dystrophin⁺ fibers and found that the %CSA occupied by dystrophin⁺ fibers increased significantly in the 12-week animals (44.2±3.7%, n=3) versus 4-week animals (26.6±2.9 n=3 p<0.02 vs 12 week group). We also measured and analyzed the CSA of individual dystrophin⁺ fibers in both groups and found a significant increase in the number of large dystrophin⁺ fibers and improvement in the alignment of dystrophin⁺ fibers in the 8 and 12 week groups (Figs. 3B, C, S12 and S13). Interestingly whereas the majority of dystropin⁻ host fibers (52%±15%, 168/337) and dystrophin⁺ fibers in PM injected muscles (92%, 60/65) have centrally located myonuclei, only a small percentage of dystrophin⁺ fibers in MIDC injected muscles have centrally located nuclei (2.4%±1%, 65/2658, p<0.01 vs both untreated and PM treated muscles) (Figs. 3A and S14). At 12 weeks post cell-injection we counted the number of total myofibers from the mid cross section of each TA muscle harvested. There were 5345±373 (n=3) fibers in MIDC-injected TA muscles of which (4321±491) are dystrophin⁺, and there were 2899±273 (n=3) in control TA muscles (Fig. S15) indicative of new fiber growth in the MIDC injected muscles.

Figure 3. De novo fiber formation in pre-irradiated mdx muscle.

(A) Merged dystrophin and DAPI staining pictures of MIDC- (top) and Cont-Mb- (bottom) injected muscles harvested at 2, 4 and 8 weeks. (B) Sections showing the size and number of dystrophin⁺ fibers in MIDC-injected muscles harvested at 4 and 12 weeks post-cell-injection. (C) Quantitative analysis of fiber size distribution in MIDC-injected muscles harvested at 4 and 12 weeks post-cell-injection (n=3 muscles). Scale Bars=100μm.

When GFP-labeled MIDCs were injected into the muscles of SCID (Fig. 4A and B) or nude (Fig. S16) mice that have no degeneration/regeneration signals, the GFP signal quickly reached a plateau and remained stable up to 6 months, the longest time point tested. In addition to eGFP⁺ myofibers, eGFP⁺/Pax7⁺ mononuclear cells could be detected in muscles transplanted with eGFP-labeled MIDCs when measured at 2 months. The majority of these eGFP⁺/Pax7⁺ cells (75.8%, 232/306) were located beneath the myofiber basement membrane (the typical location of satellite cells)(Fig. 4C). The remaining cells (24.2%, 74/306) appeared to be in the interstitium encircled by their own basement membrane (Fig. 4D).

Figure 4. Evidence of environmentally regulated MIDC regeneration and satellite cell formation.

(A) Representative in vivo fluorescence images of eGFP-labeled MIDCs transplanted in SCID tibialis anterior muscles injected with cardiotoxin 3 days prior to cell-transplantation. (B) Graph of fluorescence signal intensity over time. n=4. Baseline fluorescence prior to injection was null. (C) EGFP⁺/Pax7⁺ cells residing underneath basal membrane. (D) An eGFP⁺/Pax7⁺ cell residing in the interstitium enclosed by its own basal membrane. Scale Bar=20μm.

Discussion

In the present study, by conditionally expressing the transcription factor Msx1 in primary mouse myotubes, we found that they could be induced to re-enter the cell cycle and fragment into mononuclear cells that were lineage-specific proliferating myogenic progenitors. Furthermore, these dedifferentiated mononuclear cells displayed robust, myogenic progenitor-like, highly specialized differentiation potential following in vivo transplantation, indicating that our novel approach may have translational impact towards combating degenerative muscle disorders and volumetric muscle loss. Previous studies have already demonstrated that dedifferentiation can efficiently regenerate impaired organs including fin, heart etc. in zebrafish [1, 15–17] and lost limbs in salamanders [5]. At least in the case of newt myofibers this natural process of regeneration is dependent on the induction of Msx1 expression [6]. In the past few years, several examples of dedifferentiation in mammals have been verified either from in vitro or in vivo experiments [18–20]. A recent study shows that fate-restricted progenitor populations expressing Msx1 are responsible for regenerating the amputated digit tip in vivo in mice [21], which with our results provides evidence that mammals can harness a muscle regeneration strategy if the same Msx1 pathway used by urodele amphibians can be activated.

Mammalian embryonic myogenesis and postnatal muscle repair are two distinct and well-defined processes. During myogenesis, the majority of muscles including limb muscles are formed by Pax3⁺ embryonic myogenic progenitors and Pax3⁺/Pax7⁺ fetal myogenic progenitors [22–24]. With rare exceptions, all the myofibers that build up an adult muscle are formed during embryonic myogenesis [25, 26]. Post-natal growth and repair of mammalian muscle mainly relies on mobilization of satellite cells, the adult form of myogenic progenitors [27]. Unlike embryonic myogenic progenitors, satellite cells, although capable of repairing damaged myofibers, appear to have a limited capacity to initiate new fiber formation, partly due to their specific location. Satellite cells are located beneath the basal membrane that wraps the surface of each myofiber [28]. In response to various destructive assaults on myofibers, satellite cells activate, proliferate and fuse to repair the partially or completely damaged myofiber [29, 30] and, at the same time, attain self-renewal to repopulate a satellite cell pool [31–34]. The process of satellite cell-mediated regeneration is rapid and efficient. Thus, as a collective result of individual fiber replacement, skeletal muscles have an impressive ability to repair themselves after damage. The regeneration of damaged fibers by satellite cells is referred as a tissue level of muscle regeneration by Carlson [35] to distinguish it from the organ regeneration of a partially or entirely lost muscle seen in some urodele amphibians and other lower organisms. Studies have shown that despite the fact that satellite cells of mammalian can migrate beneath the basal membrane of the myofiber with which they are associated[36–39], they can only migrate across the basal membrane to regenerate other fibers during the developmental period when the basal membrane is still fragile, or in some diseases and in extreme injury conditions [40, 41]. Thus, myofiber regeneration occurs by expansion of satellite cell progeny within the original empty endomysial tubes and very few, if any, new fibers are ever formed as a result of de novo regeneration by satellite cells [42–44].

Regeneration of a lost or even partially lost muscle by satellite cells has not been demonstrated in mammals. Strategies to generate artificial muscles using satellite cells and their progeny have had very limited if any success which has posed a major obstacle to achieving that goal [45–52]. Animal experiments show that transplantation of the entire content of a muscle that has been removed, minced and placed back to its original muscle bed resulted in regeneration of only ~30% of the original muscle in a somewhat discordant fashion. Due to this limited or lack of ability for de novo fiber formation, mammals respond to growth demand by hypertrophy instead of hyperplasia [53].

Transplantation of primary myoblasts, in both un-injured muscle and muscle injured with cardiotoxin, results in an initial formation of the donor-derived fibers by fusion with each other or with host myofibers. Once this fusion has occurred, however, the number of donor-derived fibers either remains unchanged over time or decreases because of immuno-rejection [54–56]. Although using highly purified subpopulations of satellite cells or freshly isolated satellite cells to engraft muscle have achieved much more robust regeneration compared with cultured myoblasts, the majority of engrafted fibers (98%) are donor-host hybrids and no report has shown an increase in the number of donor-derived myofibers after initial engraftment [57] by isolated satellite cells unlike our findings in this study. In addition it has been shown that the transplantation of the most primitive satellite cells (Pax7⁺/Myf5⁻) results in extensive repopulation of the satellite cell niche without interstitial retention and new fiber formation [31]. Collectively, these experiments demonstrate that isolated satellite cells have a strong tendency to home to and remain in their native niches and thereafter behave no differently after engraftment than native satellite cells. Thus it is likely that the limited capability of satellite cells to form de novo myofibers is responsible for the inability of mammals to regenerate lost muscle.

In contrast to the tissue level of muscle regeneration, in organ level muscle regeneration all the myofibers regenerate de novo [35]. Organ level of muscle regeneration has only been observed in amphibians and other lower organisms such as zebra fish. In organ level muscle regeneration, new muscles are formed in coordination with the regeneration of surrounding structures such as bones, cartilages, tendons, nerves and blood vessels. Interestingly, it is noticed that myofibers formed during organ level of muscle regeneration have peripheral located nuclei [35] unlike myofibers regenerated by satellite cells which are initially central-nucleated in many species including human [58] and remain central-nucleated in rodents indefinitely [59]. The cell sources of organ level of muscle regeneration have been the subject of intense debate[60]. Studies using different organisms yielded diverse results [17, 61]. A recent study using Cre-loxP-based genetic fate mapping technique has shown that even two closely related salamander species use different cell sources for muscle regeneration and confirmed that in the newt multinuclear muscle cells do revert back to proliferating mononuclear cells after amputation and subsequently contribute to muscle regeneration but in Axilotil it is regenerated by resident Pax7⁺ satellite cells [4].

It has been established by several groups that post-mitotic multinuclear mammalian muscle cells can be induced to become proliferating mononuclear cells [9–11]. Whether these dedifferentiated mammalian muscle cells simply regain proliferating property and take one step back revert to the cells that form post mitotic muscles or whether they obtain the myogenic capability of earlier progenitors or the regeneration capability of dedifferentiated newt muscle cells is a critical question for the applicability of dedifferentiation approach to the regenerative medicine. In the present study, we provide evidence for the first time showing that at least in the case of Msx1 induced dedifferentiation, dedifferentiated cells reactivated genes expressed by embryonic myogenic progenitors and display a degree of myogenic regenerative capability only observed in embryonic/fetal myogenic progenitors and in dedifferentiated newt muscle cells.

Remarkably, in pre-irradiated mdx muscles, injected MIDCs gradually regenerated 300 times more myofibers over a 3 month-period than were generated in muscles injected with their parental satellite cell derived myoblasts, and that over 12 weeks time these MIDCs donor-derived fibers made up as much as half of the cross sectional area of the transplanted muscles and increased the total number of fibers per cross sectional area 1.8 fold. Despite the fact that the dystrophin⁺ fibers occupy only half of the total cross-area they account for 80% of the total number of fiber in the total cross section. This is because many new fibers still remain smaller than normal. Given the fact that the increase in the total number of myofibers coincides with the decrease in the number of dystrophin⁻ host fibers in the MIDC-injected muscle, from 2900 fibers in the uninjected muscle to 1024 fibers/muscle, it is likely that MIDCs have both fused with the existing host fibers and formed de novo fibers. (Fig S11). The unprecedented regenerative power exhibited by MIDCs suggests that if proper cues and niches including supporting structures are provided, these reprogrammed muscle progenitors could potentially reconstruct a fully functional muscle. This offers an attractive therapeutic strategy for treatment of genetic myopathies and volumetric muscle loss due to severe trauma or surgery with ex-vivo treated and expanded autologous cells [46, 47, 62, 63].

Although why this regenerative pathway is activated by injury in urodele amphibians, but rarely if ever in mammals, remains unknown. Whether our findings can be extended to mammalian tissues other than skeletal muscle is also unclear. Nevertheless, we demonstrate that it is possible to directly reprogram terminally differentiated mammalian cells back into a lineage specific progenitor state, rather than having to induce them to pass through a pluripotent state or to transit from one adult state to another. Our studies demonstrate that MIDCs are capable of de novo fiber formation when injected in an ectopic location and the increase in dystrophin⁺ fibers over time in irradiated mdx muscles suggests that de novo fiber formation is also possible in the context of a degenerating muscle although these data do not conclusively prove this. These data shift the current paradigm of mammalian cell reprogramming, and have the potential to provide a shortcut for the regenerative medicine as well.