Hierarchization of Myogenic and Adipogenic Progenitors Within Human Skeletal Muscle§


  • Didier F. Pisani,

    1. Institute of Developmental Biology and Cancer, University of Nice Sophia-Antipolis, CNRS, UMR6543, Nice, France
    Search for more papers by this author
  • Noémie Clement,

    1. Institute of Developmental Biology and Cancer, University of Nice Sophia-Antipolis, CNRS, UMR6543, Nice, France
    2. Centre de Référence Maladies Neuromusculaires et de la SLA, CHU de Nice, Nice, France
    Search for more papers by this author
  • Agnès Loubat,

    1. Plateforme de Cytométrie, IFR50, Faculté de Médecine, University of Nice Sophia-Antipolis, Nice, France
    Search for more papers by this author
  • Magali Plaisant,

    1. Institute of Developmental Biology and Cancer, University of Nice Sophia-Antipolis, CNRS, UMR6543, Nice, France
    Search for more papers by this author
  • Sabrina Sacconi,

    1. Institute of Developmental Biology and Cancer, University of Nice Sophia-Antipolis, CNRS, UMR6543, Nice, France
    2. Centre de Référence Maladies Neuromusculaires et de la SLA, CHU de Nice, Nice, France
    Search for more papers by this author
  • Jean-Yves Kurzenne,

    1. Services de Chirurgie Pédiatrique, CHU de Nice, Hôpital l'Archet, Nice, France
    Search for more papers by this author
  • Claude Desnuelle,

    1. Institute of Developmental Biology and Cancer, University of Nice Sophia-Antipolis, CNRS, UMR6543, Nice, France
    2. Centre de Référence Maladies Neuromusculaires et de la SLA, CHU de Nice, Nice, France
    Search for more papers by this author
  • Christian Dani,

    1. Institute of Developmental Biology and Cancer, University of Nice Sophia-Antipolis, CNRS, UMR6543, Nice, France
    Search for more papers by this author
  • Claude A. Dechesne

    Corresponding author
    1. Institute of Developmental Biology and Cancer, University of Nice Sophia-Antipolis, CNRS, UMR6543, Nice, France
    • Institute of Developmental Biology and Cancer, CNRSUMR6543. Tour Pasteur, 28 avenue Valombrose, 06,107 Nice, France
    Search for more papers by this author
    • Telephone: +33493377082; Fax: +33493377058

  • Author contributions: D.F.P. and C.A.D.: collection and/or assembly of data, data analysis and interpretation, manuscript writing, conception and design, final approval of manuscript; C.D.: financial support, administrative support, data analysis and interpretation; N.C., A.L., and M.P.: collection and/or assembly of data; S.S., C.D., and J.Y.K.: collection and/or assembly of data, provision of study material or patients.

  • First published online in STEM CELLS EXPRESS October 8, 2010.

  • §

    Disclosure of potential conflicts of interest is found at the end of this article.


Skeletal muscle cells constitute a heterogeneous population that maintains muscle integrity through a high myogenic regenerative capacity. More unexpectedly, this population is also endowed with an adipogenic potential, even in humans, and intramuscular adipocytes have been found to be present in several disorders. We tested the distribution of myogenic and adipogenic commitments in human muscle-derived cells to decipher the cellular basis of the myoadipogenic balance. Clonal analysis showed that adipogenic progenitors can be separated from myogenic progenitors and, interestingly, from myoadipogenic bipotent progenitors. These progenitors were isolated in the CD34+ population on the basis of the expression of CD56 and CD15 cell surface markers. In vivo, these different cell types have been found in the interstitial compartment of human muscle. In vitro, we show that the proliferation of bipotent myoadipogenic CD56+CD15+ progenitors gives rise to myogenic CD56+CD15 progenitors and adipogenic CD56CD15+ progenitors. A cellular hierarchy of muscle and fat progenitors thus occurs within human muscle. These results provide cellular bases for adipogenic differentiation in human skeletal muscle, which may explain the fat development encountered in different muscle pathological situations. STEM CELLS 2010;28:2182–2194


MSCs are considered as tissue resident stem cells that are preferentially localized in the stromal, interstitial, or perivascular regions. These cells have been identified in many tissues, and extensively studied in bone marrow, adipose tissue, and skeletal muscle [1–5].

Many studies on the characterization of postnatal MSCs and mesenchymal progenitors refer to specific populations with particular morphological, molecular, and differentiation potential characteristics [6–8]. As the hierarchization of these cells is still unclear, they cannot be classified into a unifying scheme [2]. This is essentially due to the lack of common markers to classify them [6, 7]. It has been speculatively suggested that a hierarchization must exist in mesenchymal tissues, with a multipotent stem cell located upstream of various mesenchymal cell intermediates, including multipotent progenitors and committed precursors [2, 9]. The findings of various studies support this hypothesis with the characterization of different committed progenitors derived from “clonal” populations [10–12]. However, there are still various general problems, that is, a lack (a) of cell surface markers restricted to a particular cell intermediate, (b) of characterization of the anatomical origin of each characterized cell, and (c) of physiological function. For comparison, in the hematopoietic system, several surface markers are known that enable both hierarchization and prospective purification of hematopoietic stem and progenitor cells [13].

Other studies have described human MSCs and progenitors, with a clear anatomical origin, along with markers that allow their prospective purification directly from tissues. This is especially the case for skeletal muscle, where various resident stem cells have been characterized, alternatively to satellite cells, such as myoendothelial cells, pericytes, mesoangioblasts, CD133+ cells, and side population cells [3, 9, 14, 15]. All of these cells could be prospectively purified in human skeletal muscle using one or more specific surface markers such as CD34, CD56, CD31, CD133, CD144, and CD146 or according to their endogenous specificity such as alkaline phosphatase activity or Hoechst stain exclusion. These cells are able to differentiate into various lineages, but there are still no known intermediates between stem cells and differentiated stages, thus hampering hierarchical classification of these different lineages.

Adipogenic lineage in skeletal muscle is of high interest for muscle physiopathology. Indeed, severe muscle dystrophies involve significant intramuscular adipose development. This may induce deregulation of muscle homeostasis, with potential insulin resistance, regeneration deregulation, and physical activity dysfunction. The cellular origin of muscle fat formation is still unclear. Very recently, two major studies described a population of skeletal muscle resident progenitors in mouse (CD31, CD45, CD34+, Sca-1+, CD90+, CD140a+) displaying a unique white adipogenic potential [16, 17]. These cells seem to represent the main lineage involved in muscle adipose deposition in a mouse model of glycerol-induced fatty muscle [16–18].

Characterization of an equivalent population in humans is crucial to gain further insight and progress in the treatment of adipose deposition in pathological muscles. In humans, it is recognized that CD56 antigen neural cell adhesion molecule (NCAM) is a cell surface marker related to myogenic cells [19], while recently CD15 antigen (stage specific antigen 1 [SSEA1], Lewis X) has been found in skeletal muscle-derived adipogenic cells [20]. Moreover, we recently demonstrated that skeletal muscle cells expressing CD34 display a high adipogenic potential in vitro and in vivo [21]. In this study, we demonstrate, via cell cloning and cell separation according to the expression of CD15, CD34, and CD56 markers, the existence of a myoadipogenic lineage hierarchy of resident progenitors in human skeletal muscle. Interestingly, none of these progenitors were found to belong neither to the satellite cell lineage nor to the brown fat lineage. Moreover, based on this combination of cell surface markers, we were able to prospectively purify committed adipogenic and/or myogenic progenitors from skeletal muscles. These important findings are the first step toward a clear characterization of molecular events triggering adipogenic versus myogenic commitment in human skeletal muscle, thus substantially boosting the overall understanding of intramuscular fat, particularly in severely dystrophic muscles.


Reagents and Antibodies

Cell culture media, serum, buffer, and trypsin were purchased from Lonza Verviers (Verviers, Belgium, http://www.lonza.com/group/en/company/sites/europe/lonza_verviers_sprl.html) and cell culture reagents from Sigma-Aldrich Chimie (Saint-Quentin Fallavier, France, http://www.lyon-spiral.org/campagne_2003/b_ nord_isere/f_sigma_stquentin.html).

Monoclonal mouse antimyosin heavy chain (A4.1025) was purchased from Developmental Studies Hybridoma Bank (University of Iowa, Iowa City, Iowa, http://dshb.biology.uiowa.edu/) and mouse antinestin (VMA5326) from AbCys (Paris, France, http://www.abcysonline.com/Default.asp). Polyclonal rabbit antilaminin (L9393) was purchased from Sigma-Aldrich Chimie and anti-fatty acid binding protein 4 (FABP4) was a gift from D.A. Bernlohr (University of Minnesota). The polyclonal guinea pig antibody against perilipin (RDI-PROGP29) was purchased from Research Diagnostic Inc. (Flanders, NJ, http://www.fitzgerald-fii.com/). Antibodies against CD15 (555400) and CD56 (559043) were purchased from BD Biosciences (Le Pont-De-Claix, France, http://www.bd.com/fr/biosciences/products/). All secondary antibodies (Alexa Fluor 405, 488, and 594) were purchased from Invitrogen (Cergy Pontoise, France, http://www.invitrogen.com/site/us/en/home/Global/Contact-Us/RegionalContactUs/France.html).

Antibodies for fluorescence-activated cell sorting (FACS) and flow cytometry were purchased from BD Biosciences (CD15, 555401; CD31, 130-092-652; CD34, 555822; CD44, 555479; CD45, 555485; CD49b, 555669; CD56, 555,516 and 555517; CD90, 555596; CD106, 551647; CD117, 550412; CD146, 550315), R&D Systems (STRO-1, MAB1038; Abingdon, England, http://www.rndsystems.com/) and Miltenyi Biotec (CD133, 130-090-826; CD271, 130-091-885; marrow stromal cell antigen-1 [MSCA-1], 130-093-589; Paris, France, http://www.miltenyibiotec. fr/en/default.aspx). MACS® μbeads-conjugated antibodies for cell sorting (CD15, 130-046-601; CD34, 130-046-702 and CD56, 130-050-401) were purchased from Miltenyi Biotec.

Skeletal Muscle Cell Isolation

A total of 27 skeletal muscle samples were obtained as res nullus from surgeries or diagnostic biopsies (10 healthy adult donors: mean age = 39.0 years; range = 17–66 years; 10 healthy young donors: mean age = 3.8 years; range = 3 months to 11 years; 12 pathological donors: mean age = 53.6; range = 38–75 years). Ten skeletal muscle samples were particularly used in the study and were chosen for their variability to exclude individual bias (Supporting Information Table 2). All protocols were approved by the Centre Hospitalier Universitaire de Nice Review Board, according to the rules of the French Regulatory Health Authorities. Samples were placed in F10 medium and transferred to the laboratory.

Skeletal muscle cells were isolated by a standard method [22]. Briefly, skeletal muscle was minced into fragments of 1 mm3 and digested at 37°C under agitation, first using Liberase (collagenase mix, Roche Diagnostics, Meylan, France, http://roche.fr/portal/eipf/france/rochefr/recherche/home) for 1 hour and then 0.25% trypsin-EDTA (Lonza Verviers) for 20 minutes. The enzymatic reaction was stopped by adding 1% fetal bovine serum (FBS). The suspension was homogenized and filtered through 100-μm and then 40-μm cell strainers (BD Biosciences), and finally, the cells were pelleted by centrifugation.

Cell Separation, Culture, and Differentiation

Cell separation was performed with a magnetic μbead MS column (MACS) using CD34-, CD15- and/or CD56-μbeads (MACS) according to the manufacturer's instructions or with a FACSAria cell sorter (Becton Dickinson, Le Pont-De-Claix, France) using relevant antibodies. Sorted cells were reanalyzed in all experiments (see examples in Fig. 2A and Supporting Information Fig. 1A). Single cell sortings were performed on a 96-well plate using FACSAria cell sorter coupled to an “Automated Cell Deposition Unit” (Becton Dickinson). Proliferating clones were transferred onto a 24-well-plate after 2 weeks and then analyzed. The purity of cell cloning by FACS was checked by cloning a mixed population of CD56+GFP+ and CD56+GFP cells (50/50). All amplified clones were fully GFP+ (40/40 clones) or GFP (30/30 clones; Supporting Information Fig. 1B).

Fresh or sorted cells were cultured in F10 medium complemented with 20% FBS, 2 mM glutamine, 10 mM Hepes, antibiotics (100 U/ml of penicillin and 100 mg/ml of streptomycin), 10−6 M dexamethasone and 2.5 ng/ml basic fibroblast growth factor (bFGF) at 37°C under 5% CO2.

Myogenic differentiation was induced by withdrawing FBS from confluent cells and adding 10 μg/ml insulin, 5 μg/ml transferring, and 2% horse serum. Adipogenic differentiation was induced by the addition of Dulbecco's modified Eagle's medium (DMEM)/F12 50/50 v/v complemented with 10% FBS, 1 μM dexamethasone, 0.5 mM 1-methyl-3-isobutylmethyl-xanthine, and 5 μg/ml insulin. Three days later, cells were placed in DMEM/F12 complemented with 10% FBS, 5 μg/ml insulin, 0.2 nM Triiodothyronine, and 100 nM rosiglitazone [peroxisome proliferator-activated receptor γ (PPARγ) agonist]. Dual myoadipogenic differentiation was induced by mixing (1/1 v/v) myogenic and adipogenic medium.

Low-density cultures for clone analysis were performed by plating 500 cells in a 100-mm-diameter Petri dish (9 cells/cm2). After 2 weeks under growth conditions, clones were clearly visible and individualized.

For differentiated cell staining assays, plated cells were fixed in 4% paraformaldehyde (PAF) for 10 minutes and treated with the following filtered solutions for 15 minutes: Oil Red O (60% of a stock solution at 0.5% w/v in isopropanol and 40% of distilled water) or crystal violet (0.1% w/v in distilled water). All steps were performed at room temperature (RT) and a PBS wash was performed twice between all steps.

Fluorescence Analysis

For flow cytometry analysis, cells were treated sequentially under agitation with primary antibodies or the relevant isotype control (30 minutes, 4°C) and, if necessary, with immunofluorescent secondary antibodies (15 minutes, 4°C) until analysis with a FACSCalibur flow cytometer and CellQuestPro software (BD Biosciences).

For immunofluorescence analysis of cell cultures, plated cells were fixed with PAF 4% for 10 minutes, permeabilized with 0.1% Triton X-100 for 10 minutes, and then sequentially incubated with primary antibody for 1 hour and the relevant secondary antibody for 30 minutes. All steps were performed at RT and a PBS wash was performed twice between all steps.

For immunofluorescence analysis of skeletal muscle sections, we used 8-μm-cryostat sections. All steps were performed at RT and a PBS wash was performed twice between all steps. Sections were permeabilized in PBS 0.2% Triton X-100 (PBT) at RT for 20 minutes, saturated in PBT 3% bovine serum albumin (BSA) (30 minutes, RT), and incubated with primary antibody (1 hour, RT) and then with a secondary antibody (45 minutes, RT).

Cells or sections were finally mounted in Mowiol and visualized with an Axiovert microscope (Carl Zeiss, Le Pecq, France) under oil immersion and pictures were captured and treated with AxioVision software (Carl Zeiss).

RNA Extracts and Reverse Transcription Polymerase Chain Reaction/Quantitative Polymerase Chain Reaction

Total RNA was extracted using TRI-reagent® (Euromedex, Souffelweyersheim, France, http://www.euromedex.com/) according to the manufacturer's instructions from plated cells or primary cells. RNA was treated with DNase I (Promega, Charbonnieres, France) for 30 minutes before reverse transcription. First strand cDNA was generated on 2 μg of RNA with moloney murine leukemia virus reverse transcriptase (M-MLV-RT) (Promega) in the presence of 12.5 ng/l random primers for 2 hours at 37°C.

Polymerase chain reactions (PCRs) were performed on 0.5 μl of cDNA with GoTaq polymerase (Promega), 2 μM of each dNTP, and 200 nM each primer. β-actin cDNA or 18S rRNA was amplified as a reference. For each PCR experiment, a water control and a control without M-MLV-RT were served as negative controls, whereas adequate cDNA templates served as positive controls (not shown). For quantitative PCR (QPCR), the final reaction volume was 20 μl including specific primers (0.3 μM), 5 ng of reverse transcribed RNA, and 10 μl of SYBR Green Master Mix (Applied Biosystems, France). The QPCR conditions were as follows: 2 minutes at 50°C, 10 minutes at 95°C, followed by 40 cycles of 15-second at 95°C, 1 minute at 60°C. QPCR assays were run on an ABI Prism 7000 real-time PCR machine (Applied Biosystems, France). Quantification was performed using the comparative-ΔCt method. The housekeeping gene TATA box binding protein was used as reference. All primer sequences were designed using Primer Express software (Applied Biosystems, Courtaboeuf, France; Supporting Information Table 1). For QPCR analysis, primers were validated by testing the PCR efficiency using standard curves (85% < efficiency < 115%).

Statistical Analysis

Differences between data groups were evaluated for significance using the two-tailed unpaired Student's t test. A p value < .05 was considered significant. The data are presented as mean ± SEM of measurements of independent skeletal muscle samples or cell clones, and their number is indicated. Pearson's correlation coefficient was performed to determine correlations between two measures.


The Ranges of Myogenic and Adipogenic Differentiation Potentials of Muscle-Derived Cells Were Correlated with the Percentages of CD56+ and CD15+ Cells, Respectively

Populations of cells purified from human skeletal muscle samples differentiate into myotubes as well as adipocytes when cultured under suitable conditions [20, 23]. Recently, we showed that the myogenic and adipogenic potentials are in fact restricted to CD34+ muscle cells [21]. To further investigate this bipotentiality, myogenic and adipogenic differentiations were examined at the single clone level after plating CD34+ muscle cells from several donors at very low density (biopsies b1 to b5, Supporting Information Table 2). CD34+ cells were derived from fresh digested muscle samples by μbead cell sorting. We obtained a percentage of clonogenicity of 25.46 ± 6.27 (n = 11). Each sample gave rise to three clone types in variable proportions, that is, (a) myogenic clones identified by the presence of abundant elongated multinucleated myotubes, (b) adipogenic clones recognized by their content in adipocytes with typical lipid-filled vacuoles stained with Oil Red O, and (c) myoadipogenic clones exhibiting a combination of myotubes and adipocytes (Fig. 1A). The proportions of the three clone types generally stayed in the same range after different passages in culture but markedly varied between muscle samples (Fig. 1B, upper panel). Muscle cells thus contained a mixture of monopotent and bipotent cells, regarding the myogenic and adipogenic differentiation potential. At this point, we cannot rule out that the bipotentiality was at least in part due to the clone mixture, but the very low density used for cell cloning strongly suggests that all the myoadipogenic clones could not be due to cloning artifacts.

Figure 1.

Clonal differentiation analysis of skeletal muscle-derived CD34+ cells according to CD15 and CD56 expression. This analysis was conducted on unsorted or CD15 and CD56 μbead-sorted cells prepared from five muscle biopsies (b1 to b5). For each cellular fraction, 500 cells were plated on a 100-mm dish (9 cells/cm2) for 2 weeks and then differentiated in a myoadipogenic medium for 2 weeks. Clones were stained with crystal violet and Oil Red O to detect multinucleated myotubes and triglyceride accumulation, respectively. An average of 100 ± 9 differentiated cell clones derived from each muscle biopsy was counted per assay. (A): The left picture shows a representative view of differentiated clones; the other pictures show a high-magnification view of adipogenic and bipotent myoadipogenic clones. Scale bar = 100 μm. (B): The upper histogram represents the quantification of the amounts of each type of clone. The lower histogram displays CD15 and CD56 expression in the respective analyzed cell fractions. Columns of each muscle sample refer to clones differentiated at different culture passages (passages 0–6) after sorting. (C): Correlation between CD15 or CD56 expression and number of adipogenic or myogenic clones, respectively.

We tentatively used CD56 and CD15 cell surface markers to determine the cellular origins of the three types of clone. CD56 is known to be expressed by satellite cells and other promyogenic cells within skeletal muscle [20, 23, 24]. No adipogenic-specific cell surface markers have been characterized to date. CD15 is an antigen displayed by several glycoproteins, and it is essentially known to be present in hematopoietic and neural cells, but it has been recently suggested to be displayed by an adipogenic subset of skeletal muscle-derived cells [20, 23, 24]. CD15 and CD56 marker expression was measured by flow cytometry (FACS) on CD34+ cell populations concomitantly with cloning assays. The relative proportions of each clone type were clearly related to the expression of CD15 and CD56 in the populations at the time of cloning (Fig. 1B, lower panel): the percentages of myogenic clones correlated with the percentages of CD56+ cells, the percentages of adipogenic clones with the percentages of CD15+ cells, and the percentages of myoadipogenic clones with the percentages of CD15+CD56+ cells. A close correlation between CD15 expression and adipogenic clone frequency (r = 0.9318) as well as between CD56 expression and myogenic clone frequency was demonstrated (r = 0.9673) by linear regression analysis (Fig. 1C). No significant correlation was found between CD15+CD56+ double expression and myoadipogenic clone frequency, but this frequency was very low (not shown).

In addition, we subfractionated CD34+-derived cells (passage 2) according to the expression of CD15 and/or CD56 by μbead cell sorting. One or three passages later (passage 3 or 6), we reanalyzed CD15 and CD56 expressions (Fig. 1B, lower panel) and performed new differentiation assays on low-density-plated cells (Fig. 1B, upper panel). We obtained a clonogenicity percentage of 30.73 ± 2.48 (n = 10). Interestingly, despite the fact that the correlation between CD15 and/or CD56 expressions and the differentiation potential was maintained (Fig. 1C), no enrichment was found in CD15+CD56+ cells or in myoadipogenic clones.

Differentiation Potential of CD34+CD15+CD56, CD34+CD15CD56+, and CD34+CD15+CD56+ Cell Populations

To assess the predictive value of these correlations, we sorted CD15+CD56, CD15CD56+, and CD15+CD56+ cell fractions of CD34+ cells from three muscle samples (b1, b4, b5, Supporting Information Table 2) by FACS. Two of them came from young and adult healthy donors and one from a donor suffering of facioscapulohumeral muscular dystrophy (FSH). Cell purity was checked directly after the sorting and showed a sort efficiency of 91%–96% (Fig. 2A). Adipogenic and myogenic potentials were tested in appropriate culture media after cell amplification. CD15+CD56 cells differentiated only into adipocytes and CD15CD56+ cells only into myotubes, whereas CD15+CD56+ cells differentiated in both adipocytes and myotubes (Fig. 2B). Adipogenic and myogenic differentiations were characterized on the basis of the cell morphology, specific staining, and specific marker expression. Mature adipocytes exhibited triglyceride specific Oil Red O staining and expressed late markers of adipogenesis, such as FABP4, adipsin, and leptin (Fig. 2C and Supporting Information Fig. 2A). None of the adipogenic populations expressed the brown adipocyte markers uncoupling protein 1 (UCP1) or cell death activator-A (CIDEA), indicating that adipogenic differentiation led only to white adipocytes (Supporting Information Fig. 2A). Elongated multinucleated myotubes expressed early and late markers of myogenesis such as Myf5, MyoD, myogenin, MCK, and dystrophin (Fig. 2C and Supporting Information Fig. 2A). Finally, the distinct differentiation potentials were also confirmed by differentiation assay in a dual myoadipogenic medium. CD15+CD56 cells differentiated only into adipocytes expressing FABP4 or perilipin (Fig. 2D), CD15CD56+ cells differentiated only into myotubes expressing myosin (Fig. 2E), whereas CD15+CD56+ cells concomitantly differentiated into adipocytes expressing perilipin and FABP4 and into myotubes expressing myosin (Fig. 2F).

Figure 2.

CD34+CD15±CD56± cells committed to adipogenic and/or myogenic lineages. CD34+CD15±CD56± cells were sorted by fluorescence-activated cell sorting from three different muscle biopsies (b1, b4, b5) and cultivated for 7 days in myogenic medium or 10 days in adipogenic medium. (A): Flow cytometry profiles of muscle-derived cells from biopsy b4 before (unsorted cells) and after FACS cell sorting (sorted cells). (B): Cells were stained with Oil Red O (to detect triglyceride accumulation) and counterstained with crystal violet (to detect multinucleated cells). Higher magnifications display classic lipid containing vesicles. (C): Expressions of myogenic markers (MCK and Myf5) and adipogenic markers (FABP4 and adiponectin) were analyzed by reverse transcription polymerase chain reactions for CD15+-, CD56+-, and CD15+CD56+-sorted cells. Expression was determined using the cycle threshold (Ct) mean of the TBP housekeeping gene as reference. The results are expressed as a mean ± SEM of values found for cells derived from three independent biopsies. (D, E, F): Sorted cells were cultured for 10 days in myoadipogenic medium and then fixed for immunofluorescence assay. Antibodies against MyHC (green) and FABP4 (red) or perilipin (red) were used to detect myotubes and adipocytes, respectively. Nuclei were counterstained with DAPI. Higher magnification views display classic staining for MyHC (striated), FABP4 (cytoplasmic), and perilipin (vesicle periphery). Scale bar = 100 μm. Abbreviations: FABP4, fatty acid binding protein 4; FITC, fluoresceine iso thio cyanate; FLI-H, FL1 high detector; FSC-H, forward scatter high detector; MyHC, myosin heavy chain; PE, phycoerythrin; SSC-H, side scatter high detector.

Altogether, the results obtained at the cell population and clonal levels indicate that myogenic and adipogenic progenitors, present in human muscle, can be separated on the basis of the expression of CD56 and CD15 cell surface markers. As expected, the myogenic cells specifically expressed CD56 and, for the first time, CD15 was characterized as a specific marker of muscle adipogenic cells. In addition, CD15CD56+ monopotent myogenic and CD15+CD56 monopotent adipogenic progenitors coexisted with CD15+CD56+ bipotent myoadipogenic progenitors.

In Vivo Frequency of CD34+CD15+CD56, CD34+CD15CD56+, and CD34+CD15+CD56+ Cells

We quantified these subpopulations by FACS purification from the same three fresh human skeletal muscle samples as above and two new samples (b7, b8). CD34+ cells represented only 2% of all muscle-extracted cells. They contained CD15+CD56+ cells (3.7% ± 2.8%) and CD15+CD56 cells (7.4% ± 4.6%; Fig. 3A). Conversely, they contained only a fraction of CD15CD56+ cells (16.9% ± 10.7%), which are present in both CD34+ and CD34 cell fractions, as we recently reported [21]. No CD15+ cells were detected in the highly heterogeneous CD34 fraction, even after amplification of adherent CD34 cells (data not shown). This lack of CD34CD15+ cells was correlated with the absence of in vitro adipogenic potential of the CD34 fraction [21]. On the other hand, in culture, high variability was observed in the proportions of CD15±CD56± cells between biopsies (Fig. 3A and Supporting Information Fig. 3). Finally, CD15 and CD56 stains were confirmed by microscopy analysis of FACS-analyzed cells (Fig. 3B).

Figure 3.

In vivo CD34+CD15+/−CD56+/− cell frequency and ex vivo stability. (A): Flow cytometry analysis of two fresh skeletal muscle extracts (biopsies b7, b8). CD34 (APC) was displayed by 2% of whole muscle extract. CD15 (FITC) and CD56 (PE) detection was performed in CD34-positive cells (R6). Relevant isotype analyses are given as controls. The histogram displays the mean CD-expressing cells for five different samples (biopsies b1, b4, b6, b7, b8). (B): Directly after cytometry analysis, cells were plated on a slide for microscopy assessment of labeling. Analysis was performed without permeabilization. Scale bar = 20 μm. (C): CD34+ muscle-derived cells were sorted by FACS from three fresh biopsies (b1, b4, b5). After culture amplification, CD15+/−CD56+/− cells were sorted by FACS and reanalyzed 2 days later for CD15 and CD56 expression. Two sequential sorts (sort 2 and 3) were performed on primary sorted CD15+CD56+ cells and reanalyzed 2 days after each sort. Unlike the two simple positive populations, the double-positive population could not be maintained in culture. Abbreviations: APC, allophycocyanin; DAPI, 4′,6-diamidino-2-phenylindole; FITC, fluorescein isothiocyanate; FSC, forward scatter detector; SSC, side scatter detector.

CD34+CD15+CD56+ Cells Are Lineage Upstream of CD34+CD15+CD56 and CD34+CD15CD56+ Cells

We sorted and cultured the adipogenic CD34+CD15+CD56 population, the myogenic CD34+CD15CD56+ population and the myoadipogenic CD34+CD15+CD56+ population from the same three previous muscle samples. As already shown, the CD34 marker was rapidly lost in culture for all cells [21]. Proliferation had differential impacts on cell phenotype maintenance. CD15 and CD56 expression remained stable for CD15+CD56 cells and CD15CD56+ cells, indicating no cell derivation during proliferation (Fig. 3C). To the contrary, CD15+CD56+ cells could not be maintained proliferative in culture with the same immunophenotype status. FACS analysis as early as the first passage after sorting showed that only a fraction of cells were still positive for both CD15 and CD56, whereas the remaining cells derived in CD15+CD56 cells and CD15CD56+ cells. Second and third sequential sorts of double-positive cells showed the same derivation.

To clarify a lineage scheme within these cell populations, we performed clonal sorting (one cell per well) of CD34+CD15+CD56+ cells from three new muscle samples (b6, b7, b8, Supporting Information Table 2). We obtained a clonogenicity potential of 39.93% (230/576). After amplification, CD34+CD15+CD56+ cell-derived clones were tested for CD15 and CD56 expression along with the differentiation potential. More than 200 clones were tested. To avoid potential sorting contamination by CD15+CD56 or CD15CD56+ cells, we eliminated clones displaying only one of the two markers. Three clone profiles were found, indicating that a single CD34+CD15+CD56+ cell mainly derived to: (a) CD34+CD15CD56+ cells only (≈50%) or (b) CD34+CD15+ CD56 cells only (≈10%) or (c) both CD34+CD15+CD56 and CD34+CD15CD56+ cells (≈40%; Fig. 4A). Double-negative CD34+CD15CD56 cells were also found, especially for profile 2, but their nature was not investigated because they could merely have been due to incomplete CD15 labeling. Clones from the three profiles were submitted to differentiation assay in myoadipogenic medium. The results were fully in accordance with the differentiation observed for the nonclonal population (Fig. 2). Three clone types were found that is, (a) myogenic clones (≈60%), (b) adipogenic clones (≈5%), and (c) bi-potent myoadipogenic clones (≈35%; Fig. 4B). The frequency of clones according to their differentiation potential was closely correlated with the frequency of cell CD15±CD56± profiles (compare panel [A] and panel [B] histograms in Fig. 4). These results demonstrated that a single proliferating CD34+CD15+CD56+ cell derived into CD34+CD15+CD56 and/or CD34+CD15CD56+ cells. Moreover, these results, obtained through clonal sorting by FACS, confirmed the presence of bipotent myoadipogenic clones found in CD34+ muscle cell population plated at very low density.

Figure 4.

Myoadipogenic potential of sorted CD15+CD56+ cells at the single cell level. A total of 576 cells from three different biopsies (b6, b7, b8) were sorted at the single-cell level. A total of 230 amplified relevant clones (proliferative and containing at least CD15+CD56+ cells) were studied. (A): A total of 156 clones were analyzed by flow cytometry to check for CD15 and CD56 expression. Three profiles were found: (profile 1) clones containing CD15CD56+ and CD15+CD56+ cells, (profile 2) clones containing CD15+CD56 and CD15+CD56+ cells, and (profile 3) clones containing CD15CD56+, CD15+CD56, and CD15+CD56+ cells. The histogram displays the mean ± SEM of each clone profile frequency. (B): The other 74 amplified clones were submitted to myoadipogenic conditions for 10 days and then fixed for MyHC (green), nestin (green), FABP4 (red), or perilipin (red) immunofluorescence detection. Myogenic (MyHC or nestin positive), adipogenic (FABP4 or perilipin positive), and myoadipogenic (MyHC/nestin and FABP4/perilipin positives) clones were counted and the mean ± SEM of each differentiation frequency is shown in the histogram. DAPI was used to counterstain nuclei. Oil red O was performed on a part of the relevant differentiated clones to assess the quality of adipogenic differentiation. Higher magnification displays relevant staining. Scale bar = 100 μm. Abbreviations: DAPI, 4′,6-diamidino-2-phenylindole; FABP4, fatty acid binding protein 4; FITC, fluorescein isothiocyanate; FL1-H, forward scatter high detector; MyHC, myosin heavy chain; PE, phycoerythrin.

Lineage Origin of CD34+CD15+CD56+ Cells and Their CD34+CD15+CD56 and CD34+CD15CD56+ Progenies

The relationships of the cells with satellite cells were first clarified by investigating their anatomical position within muscle tissue. We previously demonstrated that CD34+ cells were located exclusively in an interstitial position [21]. Here, we extended our results by studying the expression of CD15 and CD56 markers. Immunofluorescence analyses were performed on muscle sections from a young and an adult healthy donors (b9, b10; supporting information Table 2) and led to identical results. As expected [21, 25], CD56+ cells were frequent and found: (a) in a satellite cell position, along fibers underneath the basal lamina, identified by the presence of laminin, and also, (b) in an interstitial position between the basal lamina of adjacent fibers (Fig. 5A, upper panel). In contrast, CD15+ cells were rare and restricted to the interstitial compartment (Fig. 5A, lower panel), along with double-positive CD15+CD56+ cells (Fig. 5B). The anatomical position of CD15+CD56+ and CD15+CD56 cells suggested that these cells do not correspond to satellite cells, in contrast to many subsarcolemmal CD15CD56+ cells.

Figure 5.

Anatomical localization and cell surface-specific marker expression in the CD15+CD56+ lineage. (A, B): Immunodetection of CD15+ (green) and CD56+ (red) cells on skeletal muscle sections (biopsies b9, b10). Subsarcolemmal (yellow arrow) or interstitial (white arrow) positions were assessed by comparison with laminin (basal lamina, blue or green). CD15+(A) and CD15+CD56+(B) cells were only found in an interstitial position, whereas CD56+(A) cells were also found in the subsarcolemmal position. DAPI was used to counterstain nuclei (blue). Higher magnification views display cell peripheral staining and interstitial/subsarcolemmal staining. Scale bar = 100 μm. (C): Expressions of satellite cell markers (MYF5, PAX7) and muscle interstitial cell marker (PW1) were analyzed by reverse transcription quantitative polymerase chain reactions for CD34+CD15+CD56+ (biopsies b6, b7, b8), CD34+CD15CD56+ (biopsies b7, b8), and CD34CD15CD56+ (biopsies b7, b8) freshly sorted cells. Expression was determined using the Ct mean of the TBP housekeeping gene as reference. The results are expressed relative to CD34CD15CD56+ (b8). (D): Immunophenotype of freshly cultivated (passage 1) CD15+CD56 (R2), CD15CD56+ (R3), and CD15+CD56+ cells (R4). Abbreviations: APC, allophycocyanin; DAPI, 4′,6-diamidino-2-phenylindole; FITC, fluorescein isothiocyanate; PE, phycoerythrin.

In agreement with these data, reverse transcription (RT)-QPCR and RT-PCR analyses showed that CD34+CD15+ CD56+ and CD34+CD15CD56+ cells freshly purified from biopsies b6, b7, and b8 expressed MYF5 but not the satellite cell marker PAX7 (Fig. 5C and Supporting Information Fig. 2B). All of these cells also expressed PW1 (or PEG3, paternally expressed gene three protein), which has recently been detected in mouse PAX7+ satellite cells as well as in a subset of interstitial cells [26]. We previously demonstrated that CD34CD15CD56+ cells were linked to the satellite cell compartment. As expected, CD34CD15CD56+ cells purified from b7 and b8 expressed MYF5, PW1, and PAX7 (Fig. 5C and Supporting Information Fig. 2B). It is interesting to note that CD34+CD15CD56+ cells expressed a higher level of MYF5 compared with CD34+CD15+CD56+. This result demonstrates a commitment of CD34+CD15CD56+ cells in the myogenic lineage. Regarding the brown fat lineage, it was found that CD34+CD15+CD56+ and CD34+CD15CD56+ cells did not express the brown fat progenitor marker PRDM16 (Supporting Information Fig. 2B).

Finally, the cell surface expression profiles of CD34+CD15+CD56, CD34+CD15CD56+, and CD34+ CD15+CD56+ cells were compared (Fig. 5D). All cells were positive for mesenchymal stem or progenitor cell markers CD44, CD146, CD49, CD90, and negative for lineage markers CD45, CD106, CD117, CD133, and STRO-1, and expressed a low level of the endothelial marker CD31. However, differences were observed for two other mesenchymal cell markers, CD271 (low affinity nerve growth factor receptor) and MSCA-1. Cytometry analysis showed a loss of expression of these markers following commitment. CD34+CD15+CD56+ cells strongly expressed both CD271 and MSCA-1, but a fraction of CD34+CD15CD56+ cells lost the expression of CD271, whereas most of the CD34+CD15+CD56 cells lost the expression of both CD271 and MSCA-1. This indicates that at least the CD34+CD15CD56+ cell population could be further fractionated in subpopulations.

Correlation Between the Myoadipogenic Progenitor Frequency and Biopsy Origins

As mentioned earlier (Fig. 3), the relative proportions of CD15+ and CD56+ cells were not markedly altered with culture passages. However, these proportions varied depending on the sample donor (Fig. 6 and Supporting Information Fig. 3). The CD56/CD15 profiles were examined in muscle cells isolated from healthy adult and young donors as well as pathological donors presenting with muscular dystrophies such as FSH, Pompe disease, and type-1 myotonic dystrophy. FACS analyses were performed on unsorted cultivated cells (passages 1–2). Despite substantial variations noted between donor samples, a significantly higher percentage of CD15+ subpopulations and a significantly lower percentage of CD56+ subpopulations were observed in cells originating from pathological muscle samples compared with those from healthy donors. Conversely, no significant variation was found for CD15+CD56+ cells.

Figure 6.

Frequency of myoadipogenic progenitors according to the pathological status. (A): Frequency of CD15+CD56, CD15CD56+, and CD15+CD56+ cells in freshly cultivated skeletal muscle-derived cells (passage 0–2) determined by flow cytometry analysis. (B): The histogram displays the frequency mean ± SEM of the three subsets of cells according to the healthy or dystrophic status. Abbreviations: DM1, type-1 myotonic dystrophy; FSH, facioscapulohumeral muscular dystrophy.


We have characterized, for the first time, a hierarchical organization in a human mesenchymal lineage specified by the loss of expression of specific markers.

The first step to establish a hierarchization according to progenitor commitment is to identify markers of these commitments, particularly cell surface markers, to prospectively purify cells of interest in vivo. No surface markers have been clearly described for mesenchymal commitment, but various informative results have been obtained [7, 20, 23]. On the basis of these results, we have enhanced data on CD56 by demonstrating that muscle CD34+ cells lacking CD56 expression were unable to commit to a myogenic differentiation program in vitro. Conversely, sorting of CD56-expressing muscle cells demonstrated the high myogenic potential of these cells. Within skeletal muscle cells, pericytes seem to be an exception as prospectively isolated CD146+CD34CD56CD45 pericytes were able to differentiate in vitro and in vivo into mature myotubes [14, 27]. Nevertheless, it could be interesting to test the CD56 expression of pericytes after their switch in specific myogenic medium. At this point, these results demonstrate that CD56 is a “marker” of myogenic potential in muscle, but the lack of this marker would not always be associated with a lack of myogenic potential. At this stage, it is unclear if CD56 is a marker of myogenic commitment in nonmuscle tissues. A recent article demonstrated that a CD56+CD271+MSCA-1+ subset of human bone marrow cells displayed myogenic potential, but the authors did not assess whether the loss of CD56 was correlated with a loss of myogenic potential [28]. We tentatively sorted CD56-expressing cells from adult human adipose tissue (unpublished data). These cells were more prone to enter the myogenic program but could not undergo terminal differentiation. Considering the difficulty of clarifying this result, we concluded that the relationship between CD56 and myogenic potential in this tissue is still questionable.

Although adipocytes and adipose tissue have long been studied, no specific surface marker has been described yet for adipose stem cells and preadipocytes. However, a recent study shows that CD15 is expressed by an adipogenic subfraction of skeletal muscle-derived cells [20]. CD15 is a nonprotein antigen, also named Lewis X antigen or SSEA-1, corresponding to a carbohydrate residue (3-fucosyl-N-acetyl-lactosamine) displayed by various cell surface proteins and synthesized by fucosyltransferase enzyme [29, 30]. CD15 was first described on committed embryonic stem cells and then studied essentially in myeloid cells. Recently, the presence of CD15 has also been associated with neural stem cells. It has been included in a specific marker panel (associated to CD24, CD29, or CD133) to purify these cells [31, 32]. In our study, we found that CD15 was present only at the surface of adipogenic cells and that its loss was correlated with the loss of adipogenic potential. This shows that the CD15 marker can be used to study the evolution of adipogenic cells along lineage commitment and to prospectively purify adipogenic cells in skeletal muscle. At this stage, it is not clear whether CD15 is a relevant marker only in skeletal muscle or in all tissues. CD15 could be the first preadipocyte-specific cell surface marker and this is currently under investigation. On the other hand, the close correlation found between the percentages of CD15-positive cells and of adipogenic cells in a whole skeletal muscle extract revealed that CD15 could be a quantitative marker of tissue content in adipogenic cells. This point could be of high interest for pathophysiological studies of muscle dystrophies as dystrophy severity is often correlated with an elevated adipocyte content, whereas the origin of these fat cells is unknown and out of control at this stage [33–35]. Here, we found a higher level of CD15-positive cells, and thus of adipogenic cells, in cultivated cells derived from dystrophic muscles compared with healthy muscles. These cells could be responsible for the fat deposition occurring in dystrophic muscles, and opens new avenues for studying dystrophic muscles and for ultimately determining targets for the control of fat deposition.

It was not surprising to find myogenic and adipogenic cells within human skeletal muscle tissue [20, 23]. However, using CD15 and CD56, we detected and isolated a few cells expressing the two markers in the CD34+ cell fraction. These CD34+CD15+CD56+ cells were able to differentiate in both myotubes and adipocytes, as expected from the presence of the two lineage markers. We have not yet found the conditions required to maintain these cells in culture with an unchanged immunophenotype, starting from the whole population as well as the clonal cell population. Indeed, after a few doublings, CD15+CD56+ cells derived in CD15+CD56 and/or CD15CD56+ cells. This phenomenon was correlated with a loss of differentiation potential, specific to the lost marker. Finally, we have developed a dynamic model that is able to describe a hierarchization between these cells (Fig. 7). This point was clearly confirmed by clonal sorting of CD34+CD15+CD56+ cells. The progeny committed to one restricted lineage, according to the expression of only one of the two markers, displayed only one differentiation potential. Furthermore, it is important to note that all of these cells also displayed osteogenic potential (unpublished data). We thus demonstrated a hierarchization with the loss of one potential and the existence of committed progenitors between stem and precursor cells within a mesenchymal lineage.

Figure 7.

Schematic summary of hierarchization in a myoadipogenic lineage according to CD15 and CD56 expression. The satellite cell lineage is displayed for comparison. Abbreviations: A, adipogenic; M, myogenic; O, osteogenic.

CD34+CD15+CD56+ cells are localized in the interstitial compartment, but not restricted to the vicinity of blood vessels, and they are not present in the subsarcolemmal compartment and do not express PAX7, which excludes identity with satellite cells [36]. Cells from several lineages are located in this position [3]. On the basis of the marker expression, we were able to determine that these cells were not derived from the pericyte lineage. Despite the common expression of CD146, fresh pericytes express neither CD34 nor CD56 [14]. Recently, a new human lineage has been described in a periendothelial location, that is, the brown adipocyte lineage. Brown adipocyte progenitors from human skeletal muscle express CD34 but not CD56 [37]. They could correspond to CD34+CD15+CD56 adipogenic cells, but we did not note any expression of brown fat markers (UCP1, CIDEA) in differentiated adipocytes derived from CD34+CD15+CD56+ cells or in cultivated cells expressing PRDM16, a brown fat/muscle progenitor marker described in mouse [38]. Moreover, CD34+CD15+CD56+ cells do not correspond to a side population or to CD133+ mesodermal progenitors due to the lack of ABCG2 and CD133 expression [39, 40]. Finally, CD34+CD15+CD56+cells express CD146 and a very low level of CD31, that is, two markers of the endothelial lineage. However, they were found to be negative for CD144, by RT-PCR as well as by flow cytometry. Thus, added to the lack of expression of PAX7, these cells seem to differ from the myoendothelial CD144+cells also described in the interstitial compartment and displaying close characteristics [15]. Finally, human CD34+CD15+CD56+ cells could at least partially correspond to the PW1+PAX7 interstitial cells described recently by Mitchell and coworkers in mouse [26]. However, more investigations are needed to assess the extent of similarity between these human and mouse cells.

Altogether, these data are not in favor of an origin of CD34+CD15+CD56+ from a known lineage. Therefore, two possibilities are remaining: these cells are new cells with an undetermined origin in human skeletal muscle or they derive from a known lineage (myoendothelial, endothelial, pericyte, or satellite cell lineages) but have taken an interstitial position. This change in microenvironment could have modified the surface expression pattern of the cells, as this is the case for integrins.

CD34+CD15+CD56 cells could be the human equivalent of the resident adipose progenitors described recently in mouse muscle [16, 17]. Indeed, the mouse cells display a close surface immunophenotype CD31CD45α7-integrin CD34+CD90+CD140a+Sca-1+, are localized in the interstitial space, are not related to a known lineage such as satellite or vascular-derived cells and are only adipogenic. In addition, we previously reported that human skeletal muscle-derived CD34+ cells containing CD34+CD15+CD56 cells are able to form adipocytes in vivo, contrary to CD34 cells, in the same mouse model of injured muscle with adipose deposition used to characterize mouse adipogenic populations [21].

Despite the undetermined origin of CD34+CD15+CD56+ in human skeletal muscle, it is interesting to note that cells with close characteristics are present in other human tissues. Indeed, the CD56+MSCA-1+CD271+ cells described in bone marrow and CD34+CD271+ cells described in adipose tissue displayed common features with CD34+CD15+CD56+cells and also do not belong to classic known lineages [28, 41]. Moreover, CD34+CD15+CD56+ cells should be linked to the CD34+CD31 adipose progenitors described in adipose tissue, and it would be interesting to assess CD15 expression in these preadipocytes [5, 42].


In this study, we described the first in vitro dynamic model of commitment in myogenic and adipogenic lineages associated with identified commitment specific markers. Human CD34+CD15+CD56+ primary cells provide an excellent model for studying the myoadipogenic balance, which we and others have long studied in the C2C12 mouse cell line [43–46]. This human model may allow us to decipher the mechanisms leading commitment in myogenic or adipogenic lineages with the possibility of integrating it in a physiological context. This balance is likely very important in dystrophic muscle where many adipocytes replace muscle tissue in relationship with the pathology severity [33, 34]. It is thus of high interest to determine the pathways involved in the commitment of CD34+CD15+CD56+ within the adipogenic lineage in dystrophic versus healthy skeletal muscles. The presented data could pave a new way for controlling fat development in dystrophic muscle and other physiopathological disorders.


This study was supported by the Centre National de la Recherche Scientifique.


The authors indicate no potential conflicts of interest.