Prospective heterotopic ossification progenitors in adult human skeletal muscle

Bone 71 (2015) 164–170

Contents lists available at ScienceDirect

Bone

j ourna l homepage: www.e lsev ie r .com/ locate /bone

Rapid Communication

Prospective heterotopic ossification progenitors in adult humanskeletal muscle

Jennifer Downey a, Dominique Lauzier b, Peter Kloen c, Klaus Klarskov a,d, Martin Richter a,e, Reggie Hamdy b,f,Nathalie Faucheux a,g, Anthony Scimè h, Frédéric Balg a,i, Guillaume Grenier a,i,⁎a CHUS Clinical Research Centre, Université de Sherbrooke, Sherbrooke, QC, Canadab Shriners Hospital for Children, Montreal, QC, Canadac Department of Orthopedic Surgery, Academic Medical Centre, Amsterdam, The Netherlandsd Department of Pharmacology, Faculty of Medicine, Université de Sherbrooke, Sherbrooke, QC, Canadae Department of Pediatrics, Faculty of Medicine, Université de Sherbrooke, Sherbrooke, QC, Canadaf Department of Surgery, Orthopedic Surgery Division, McGill University, Montreal, QC, Canadag Department of Chemical and Biotechnological Engineering, Faculty of Engineering, Université de Sherbrooke, Sherbrooke, QC, Canadah Stem Cell Research Group, Faculty of Health, York University, Toronto, ON, Canadai Department of Orthopedic Surgery, Faculty of Medicine, Université de Sherbrooke, Sherbrooke, QC, Canada

Abbreviations:HO, heterotopic ossification;MSC,mesehumanmuscle residentMSC; UCP1, uncoupling protein 1.⁎ Corresponding author at: 3001–12th AvenueNorth, Sh

Fax: +1 819 820 6410.E-mail address: [email protected] (G

http://dx.doi.org/10.1016/j.bone.2014.10.0208756-3282/© 2014 The Authors. Published by Elsevier Inc

a b s t r a c t

Article history:Received 23 May 2014Revised 11 October 2014Accepted 25 October 2014Available online 1 November 2014

Edited by Sakae Tanaka

Keywords:Human skeletal muscleMesenchymal stromal cellsHeterotopic ossificationMultilineage differentiationBrown adipocytes

Skeletal muscle has strong regenerative capabilities. However, failed regeneration can lead to complicationswhere aberrant tissue forms as is the case with heterotopic ossification (HO), inwhich chondrocytes, osteoblastsand white and brown adipocytes can arise following severe trauma. In humans, the various HO cell types likelyoriginate frommultipotentmesenchymal stromal cells (MSCs) in skeletalmuscle, which have not been identifiedin humans until now. In the present study, adherent cells from freshly digested skeletal muscle tissue were ex-panded in defined culture medium and were FACS-enriched for the CD73+CD105+CD90− population, whichdisplayed robust multilineage potential. Clonal differentiation assays confirmed that all three lineages originatedfrom a single multipotent progenitor. In addition to differentiating into typical HO lineages, human muscle resi-dent MSCs (hmrMSCs) also differentiated into brown adipocytes expressing uncoupling protein 1 (UCP1). Char-acterizing this novel multipotent hmrMSC population with a brown adipocyte differentiation capacity hasenhanced our understanding of the contribution of non-myogenic progenitor cells to regeneration and aberranttissue formation in human skeletal muscle.

© 2014 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Skeletal muscle possesses a remarkable capacity to regenerate fol-lowing trauma, mainly through myogenic stem cells [1]. However, effi-cient tissue repair also requires the activation of resident cells withinthe stroma, notably mesenchymal stromal cells (MSCs). Inappropriateactivation can lead to aberrant tissue formation such as heterotopic os-sification (HO), where extra-skeletal bone forms, most commonly inmuscle, through an endochondral process [2–4]. While HO can arisefrom fibrodysplasia ossificans progressiva (FOP), an uncommon heredi-tary disease, most cases result from a local trauma (surgery, musculartrauma, fractures) or neurological injury [5]. Traumatic HO has been

nchymal stromal cell; hmrMSC,

erbrooke, QC, Canada, J1H 5N4.

. Grenier).

. This is an open access article under

thought to result from the inappropriate differentiation of muscle-resident progenitor cells, induced by a pathological imbalance of localor systemic factors [6]. However, the cellular origin of the ectopicbone in HO remains a matter of debate [7].

While local muscle resident MSCs are a logical candidate as HO pro-genitors, other cells have been proposed. Some studies have implicatedvascular endothelial cells as a potential source for HO progenitors [8].Constitutively activated ACVRI in FOP change the morphology of endo-thelial cells to mesenchymal-like cells and induce the co-expression ofmesenchymal markers in vitro, a process that resembles the endotheli-al–mesenchymal transition [8]. Moreover, endothelial marker Tie2 hasbeen histologically observed in heterotopic lesions from patients withFOP. In addition, lineage tracing studies using Tie2-Cre reportermice in-dicated that these cells generate approximately half the chondrocytesand osteoblasts found in skeletal muscle lesions [8,9]. However, Tie2 isnot specific to endothelial cells and is also expressed in a number ofnon-endothelial cell types, including perivascular cells [10,11]. It hasalso been shown in vivo that the endothelial fraction of murine Tie2

the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

165J. Downey et al. / Bone 71 (2015) 164–170

cells (Tie2+CD31+) does not participate in HO whereas the non-endothelial fraction of Tie2 cells (Tie2+CD31−) does [12]. These recent-ly published findings strongly suggest that the Tie2 progenitors ob-served in HO are not of endothelial origin [7]. Indeed, more than 90%of Tie2+CD31− cells are also PDGFRα+Sca1+, pointing to a mesenchy-mal rather than an endothelial origin [12], which supports the findingsof Leblanc et al., who showed that a Sca1+CD31− muscle resident stro-mal cell population contributes to HO [2].

In humans, PDGFRα has been reported to be a specificmarker for in-terstitial mesenchymal progenitors that are distinct from CD56+ myo-genic cells and that possess adipogenic and fibrogenic potentials [13].While human skeletal muscle PDGFRα+ cells display osteogenic poten-tial in vivo [14], the confirmation of their osteogenic activity came fromsubcutaneous-implanted cell-loaded PLGA-hydroxyapatite blocks,which are not likely representative of the HO environment. In addition,their osteogenic activity was comparable to CD56 myogenic cells [14],suggesting that PDGFRαmay not be a marker that is exclusive to osteo-genic progenitors. Other human studies have shown that a fraction ofskeletal muscle adherent cells can give rise to osteoblasts and that thispotential is greatly increased following trauma [15,16]. A multipotentmyo-endothelial cell population in human skeletal muscle has beencharacterized based on the presence of myogenic (CD56) and endothe-lial (CD34, CD144) cell surface markers and the ability to differentiateinto mesenchymal lineages [17].

Interestingly, the brown adipogenic potential of these putative HOprogenitors has not been investigated, although it has been shownthat brown adipocytes can promote endochondral ossification in anHOmouse model by regulating oxygen availability and inducing a hyp-oxic microenvironment [18,19]. Until now, an adult human skeletalmuscle-derived progenitor cell population giving rise to brown adipo-cytes has not been characterized. While uncoupling protein 1 (UCP1)mRNA expression in adult human whole skeletal muscle has been re-ported, the identity of the responsible progenitors is not known [20].

Given the varied tissuemake-up of HO, no adult human skeletalmus-cle resident progenitor cells have been identified that can differentiateinto mesenchymal as well as brown adipogenic lineages. We enrichedhuman muscle resident mesenchymal stromal cells (hmrMSCs) and,for the first time, showed that hmrMSCs are clonally capable of efficientdifferentiation toward osteogenic, chondrogenic and adipogenic line-ages. Interestingly, these hmrMSCs were also able to differentiate intoUCP1-expressing brown adipocytes, cells that we also detected inhuman HO samples, which lends credence to a possible role for themin the development of HO. A better understanding of the cellular originresponsible for HO will provide a potential therapeutic target to treat,mitigate, or prevent this debilitating condition.

2. Materials and methods

2.1. Human tissue

Healthy human skeletal muscle tissue samples (gracilis andsemitendinosus) were obtained from patients (34 ± 8 years of age;54%male and 46% female) undergoing anterior cruciate ligament recon-struction surgery. HO tissue was obtained from a 21-year-old male pa-tient who had developed a mass in the gluteal muscle following amid-shaft femur fracture (Table S1). The sampleswere collected follow-ing resection surgery. The protocols were approved by the CentreHospitalier de l'Université de Sherbrooke Ethics Committee (#11-122and #13-164), and written consent was obtained from the patients.

2.2. Cell isolation and culture

Carefully dissected skeletal muscle samples were minced and thendigested for 30 min at 37 °C with 1 mg/mL of collagenase type I(Sigma) in DMEM containing 10% FBS. The tissue slurry was dilutedwith medium, passed through 70-μm and 40-μm cell strainers (Becton

Dickenson) and centrifuged at 325 g for 6 min at 4 °C. Primary humanskeletal muscle cells were seeded in tissue culture plates coated withMesencult-SF® attachment substrate and were expanded as adherentcells in Mesencult-XF® medium (StemCell Technologies). After 7 days,an average of 7 × 105 adherent cells were recovered per gram of tissue.The cells were trypsinized at 80% confluence and were centrifuged andresuspended inMesencult-XF®mediumasfirst passage cells, with freshmedium changes every 3–4 days. The cells were sub-cultured at a den-sity of 4 × 103 cells/cm2.

2.3. Fluorescent activated cell sorting

First passage cells were detached with the Accutase™ Cell Detach-ment solution (BD Biosciences), centrifuged and resuspended at ~1× 106 cells per ml in cold sorting buffer (PBS, 1 mM EDTA, 25 mMHEPES, pH 7.0, 1% FBS). The cells were incubated for 20 min on icewith the appropriate primary antibodies (Table S2) according to themanufacturers' instructions. During the cell sorting experiment, livecells were distinguished from dead cells using LIVE/DEAD® VioletViability/Vitality kits (Invitrogen). Fluorescence was compensatedusing BD CompBeads Set Anti-Mouse Ig, κ (BD Biosciences). The cellswere sorted a BD FACSAria™ cell sorter (BD Biosciences) equippedwith four lasers and a 100-μm nozzle set at 20 psi. Sorting gates weredefined based on unstained controls. The cells were analyzed usingFlowJo 7.9 software (Treestar Inc.). A population of unsorted cells wasused as a control. Unsorted and sorted fractions were then expandedas described above.

2.4. Differentiation protocols

The osteogenic, chondrogenic and white and brown adipogenic dif-ferentiation protocols were adapted from published protocols [21-24]and are presented in Table S3. Briefly, for the osteogenic, adipogenic(white and brown) and myofibroblastic assays, the cells were seededat a density of 8 × 103 cells per well in 24-well collagen-coated(Millipore) plates (4000 cells/cm2) in Mesencult-XF® medium and in-cubated at 37 °C in a CO2 incubator until they reached confluence.

For osteogenic differentiation, the cells were cultured in osteogenicmedium (Table S3) for 21 days. Unstimulated cells were cultured in os-teogenic basal medium (DMEM, 5% horse serum [HS]). To assessminer-alization, calciumdeposits in cultureswere stainedwith 40mMAlizarinRed-S, pH 4.1).

For white adipogenic differentiation, the cells were cultured inadipogenic inductionmedium for 3 days and then in adipogenic growthmedium (Table S3) for a further 18 days for oil Red O staining, or11 days for gene expression analyses. Unstimulated cells were culturedin adipogenic induction/growth basalmedium(DMEM, 3%/10% FBS). Anoil red O solution (0.5% oil red O in isopropyl alcohol; Sigma) was usedto detect triglycerides in the lipid droplets of mature adipocytes.

Alizarin red- and oil red O-stained area was quantified using ImageJsoftware (version 1.46, National Institute of Health) [25].

For brown adipogenic differentiation, the cells were incubated inadipogenic induction medium for 3 days and then in brown adipogenicgrowth medium (Table S3) for a further 11 days. Unstimulated cellswere cultured in the same adipogenic basal media as the stimulatedcells (DMEM, 3%/10% FBS).

To stimulate chondrogenesis, ~2.5 × 105 cells were pelleted by cen-trifugation (350 g, 6 min, 4 °C) and were resuspended in chondrogenicculture medium (Table S3). Unstimulated cells were cultured inchondrogenic basal medium (serum-free DMEM). The cells were har-vested by centrifugation on day 21. The pellets were fixed in 4% phos-phate buffered formalin and were embedded in paraffin. Sections(5 μm) cut using an HM325 microtome (Micron) were immersed inanAlcian blue solution (1% Alcian blue in 3% acetic acid; Acros Organics)to stain highly sulfated proteoglycans that characterize the cartilaginousmatrix.

166 J. Downey et al. / Bone 71 (2015) 164–170

To stimulate myofibroblastic differentiation, the cells were incubat-ed in myofibroblastic differentiation medium (Table S3) for 5 days.The TGFβ was omitted for the unstimulated controls.

The differentiation media used for all the procedures were changedtwice every 3 or 4 days.

2.5. Clonal growth and differentiation

CD73+CD105+CD90− hmrMSC clones were established by limitingdilution. Briefly, second passage cells were resuspended at a concentra-tion of less than 1 cell per 200 μl in Mesencult-XF® medium and wereplated in Mesencult-SF® attachment substrate-coated 96-well plates(200 μl per well). After 72 h, wells with a single cell were identified.After 2–3 weeks, single cell-derived clones were passaged, expandedand differentiated in osteogenic, adipogenic, or chondrogenic medium(Table S3) for 21 days.

2.6. Quantitative PCR

qPCR was performed as previously described [2]. Total RNA wasextracted using TRIzol® (Invitrogen) according to the manufacturer'sinstructions. TheRNAwas precipitatedwith isopropanol and1 μg of gly-cogen, rinsed with ethanol and resuspended in RNAse-free water. TheRNA was reverse-transcribed using RT Superscript II kits (Invitrogen).The qPCR reactions were prepared with 2× SYBR green master mix(BioRad). The samples were then placed in a RotorGene 6000 (CorbettRobotics). The qPCR conditions were as follows: 10min at 95 °C, 40 cy-cles of 40 s at 95 °C and 40 s at 56 °C. The results were analyzed usingthe 2−ΔΔCT relative quantification method normalized to the TATA-box binding protein (TBP). The primer sets for the adipogenic andchondrogenic genes were selected from other studies [16,21,26]. Com-mercial primers were used for the osteogenic genes SP7 (Hs_SP7_1_SG,QuantiTec Primer Assays) and DLX5 (Hs_DLX5_1_SG, QuantiTec PrimerAssays). The primer sets are listed in Table S4.

2.7. Western blots

Western blots were performed as previously described [27]. Briefly,the cells were lysed on ice in RIPA buffer containing protease inhibitors(Complete™; Roche Molecular Biochemicals). The homogenate wascentrifuged, the supernatant containing the proteins was recoveredand the protein concentrations were determined using the Bradfordmethod (BioRad). Proteins were separated by polyacrylamide gelelectrophoresis (PAGE) and were transferred to PVDF membranes(Millipore). The membranes were incubated with anti-UCP1 (1:1000,ab10983; Abcam) and anti-GAPDH (1:1000, FL-335; Santa-Cruz) anti-bodies overnight at 4 °C. The membranes were rinsed in PBS-T andwere then incubatedwith the appropriate secondary HRP-coupled anti-bodies (1:5000; Amersham) at RT for 1 h. After several rinses with PBS-T, the membranes were incubated in an ECL solution, and the signalswere detected using Biomax ML film (Kodak). The images were digi-tized, and the bands were quantified using ImageJ software.

2.8. Histology and immunohistochemistry

HO tissue was prepared for histology and immunohistochemistryfollowing resection as previously described [28,29]. Half the tissue wasformalin-fixed and was embedded in 4.5% methyl methacrylate(MMA). Sections (6 μm) cut using a Leica Polycut SM2500 (LeicaMicrosystems) were deplastified and stained with Goldner trichromefor comparative histology.

The remaining tissue was decalcified and was immunolabeled withan anti-UCP1 antibody (1:500, ab10983; Abcam). The immunolabelingwas amplified using avidin–biotin (Vector Laboratories) and was re-vealed using DAB peroxidase substrate kits (Vector Laboratories). Thesections were counterstained with Mayer's hematoxylin.

2.9. Immunofluorescence

Cultured cells were immunolabeled as previously described [30].Briefly, PFA-fixed cells were blocked/permeabilized (PBS containing10% goat serum, 1% BSA and 0.2% Triton® X-100) and were thenincubated with anti-UCP1 (1:800, ab10983; Abcam) or anti-α-SMA(1:100, CLSG36501-05, Cedarlane) primary antibodies for 90 min atRT. After several rinses in PBS-Tween, the cells were incubated withAlexa Fluor®594-conjugated secondary antibody (1:1000, Invitrogen).Cell nuclei were stained with DAPI (Sigma-Aldrich). Samples in whichthe primary antibodies were omitted served as controls. Indirect immu-nofluorescence was examined without counterstaining using anAxioskop 2 phase-contrast/epifluorescence microscope (Carl Zeiss,Inc.) or a DMIRE2 inverted microscope (Leica Microsystems). Photomi-crographic imageswere captured using a Retiga SRV cooled color digitalcamera (Qimaging) and were processed using Adobe Photoshop CS5.

3. Results and discussion

3.1. Multiple lineages including UCP1-expressing adipocytes in human HO

HO is characterized by the inappropriate activation of MSCs in skel-etal muscle leading to extra-skeletal bone tissue-containing cells frommultiple lineages [2,29]. Fig. 1A shows an anteroposterior X-ray of HOtissue in human gluteal muscle following orthopedic trauma. Histologicexaminations of Goldner trichrome-stained resin sections confirmedthe presence of several distinct tissue types (Fig. 1B), including maturebone (green) (Fig. 1C), cartilage (orange-red) (Fig. 1D) and adipocyteswith large lipid-filled vacuoles (Fig. 1E). It has been suggested that thepresence of oxidative brown adipocytes in a mouse model of HO sup-ports bone growth by reducing oxygen availability, which contributesto angiogenesis and endochondral ossification [18,19]. Thewhite adipo-cytes were observed in large numbers unlike the small clusters ofmultilocular adipocytes which are UCP1 positive, a specific brownadipogenic marker [31,32]. Brown adipocytes clusters were located ei-ther near muscle fibers or the fibrocartilage and chondrocyte regions(Fig. 1F). Similar results were obtained in three other HO samples(Table S1). These findings confirmed the presence of brown adipocytesin HO, corroborating previous mouse studies[18,19] and provide thefirst evidence of brown fat in a human skeletal muscle regenerativedisorder.

3.2. Isolated CD73+CD105+CD90− hmrMSCs are clonally multipotent

To isolate adult human skeletal muscle MSCs, whichmay be respon-sible for the aberrant tissue types in HO, dissociated cells from sixdonors (Table S1) were independently grown in defined culture medi-um. Adherent cells from each sample were sorted by FACS based onthe differential expression of characteristic mesenchymal (CD73,CD105), hematopoietic (CD34) and endothelial (CD31) cell surfacemarkers (Fig. 2A) [33]. Hematopoietic and endothelial cell types wereexcluded by CD34− and CD31− gating of viable cells. An initial stromalsubset was obtained based on the CD73+ and CD105+ markers. We re-fined our isolation procedure by sorting the CD73+CD105+ cells basedon the presence or absence of CD90. There is, in fact, no consensus onthe status of CD90 as a true MSC marker. Some studies have reportedthat cell populations with high levels of CD90 expression aremultipotent MSCs, whereas others have categorized CD90 as afibroblastic marker [33,34]. Care must be taken in interpreting these re-sults since culture conditions have been shown to modulate theimmunophenotypes of human stem cells in vitro [35]. However, in de-fined culture media, the human skeletal muscle CD73+CD105+CD90−

(or CD90−) and CD73+CD105+CD90+ (or CD90+) populations madeup 11 ± 8% and 41 ± 2% of total viable cells, respectively (Fig. 2A).

We evaluated the osteogenic, adipogenic and chondrogenic differ-entiation potentials of the unsorted, CD90− and CD90+ cell populations

F

Control UCP-1

D ENMCh

Ad

MCh

C

OB

MB

MB

BA

MB

Ch

Ad

25μm200μm

500μm

50μm

200μm 200μm

Fig. 1. Heterotopic ossification in human skeletal muscle contains differentiated cells frommesenchymal and brown adipocyte lineages. Representative images from one (HO-1) offour donors tested. (A) X-ray of heterotopic ossification (HO;white arrowhead) in the glu-teal muscle of a patient one year after the initial trauma. (B) Lowmagnification of the HOtissue stained with Goldner trichrome. Mature bone (MB) in green, cartilage (Ch) in red-orange and adipose (Ad) tissue. (C) Mature bone (MB) with osteon. At higher magnifica-tion, osteoblasts (OB) lining the mature bone can be seen. (D) At higher magnification,mineralized andnon-mineralized chondrocytes (MChandNMCH, respectively) can be dis-tinguished. (E)White adipocyteswith characteristic lipid droplets. (F) Immunostaining forUCP1-expressing adipocytes suggests the presence of brown adipocytes.

167J. Downey et al. / Bone 71 (2015) 164–170

(Fig. 2B) from four independent donors (Table S1). The CD90− cells dif-ferentiated into osteoblasts, as confirmed by the formation of alizarin-stained mineralized nodules; adipocytes, as confirmed by oil red O-stained lipid droplets; and chondrocytes as confirmed by tissuemorphology and Alcian blue staining. In contrast, the unsorted cellsdisplayed mixed differentiation potentials, with alizarin-stained miner-alization similar to that obtained with the CD90− cells, but limitedadipogenic and chondrogenic differentiation. Quantitative densitome-try analyses of alizarin red staining revealed 86.4 ± 7.3% coverage forunsorted cells compared to 95.9 ± 3.7% and 25.1 ± 28.3% for CD90−

and CD90+ cells, respectively, while quantitative analyses of oil red Ostaining revealed 12.5 ± 9.7% coverage for unsorted cells compared to95.4 ± 2.6% and 0.9 ± 1.1% for CD90− and CD90+ cells, respectively.

Our results differ from those reported by Nesti et al., who showed thatthe CD90+-derived subpopulation is enriched in multipotent cells [16]. However, our population was isolated from untraumatized muscleand was expanded in a defined culture medium. The CD90+ cellsdisplayed minimal differentiation towards all three lineages butexpressedα-smoothmuscle actinwhen stimulatedwith TGFβ, suggest-ing that myofibroblastic progenitors were present, as others haveshown [36,37] (data not shown). These results indicated that theCD90− cell subpopulation contains hmrMSCs that are capable of differ-entiating into the lineages observed in HO.

To determinewhether these CD90− hmrMSCs arise from a commonprogenitor, we isolated clones from the CD90− population to determinetheir lineage commitment and differentiation potential. Nine clones de-rived from a single donor (Table S1) were obtained from 576 platedwells by limiting dilution (clonal efficiency of ~1.6%) andwere assessedfor their differentiation potential toward the osteogenic, adipogenic andchondrogenic lineages. Four clonal progenies (~44%) differentiated intoall three lineages (Fig. S1); three (~33%) displayed bipotent capabilitiesand differentiated into the osteogenic and adipogenic or adipogenic andchondrogenic lineages; whereas one (~11%) was unipotent and differ-entiated only into the adipogenic lineage. One of the nine clones failedto differentiate, probably due to senescence. The clonal assay showedthat the CD90− hmrMSCs contained single progenitor cells with multi-ple lineage differentiation capabilities. The fact that certain clones werenot tripotent suggested that other, more committed progenitors arepresent in this population.

3.3. Characterization of CD90− hmrMSC multipotency and differentiationinto UCP1-expressing brown adipocytes

The multipotent differential potential of the CD90− cells was con-firmed by qPCR. Bone morphogenetic proteins (BMPs) play a criticalrole in the commitment of MSCs and the induction of osteoblastic activ-ity [38,39]. To assess the osteogenic differentiation potential, we usedBMP9, the most potent osteogenic BMP [40], which efficiently inducesthe osteogenic program of mouse progenitor muscle resident stromalcells [2] and for which a role in the development of human HO wasproposed [29]. BMP9 significantly increased the expression of the oste-ogenic markers SP7 andDLX5 in CD90− cells compared to unstimulatedcells (Fig. 3A).

The chondrogenic potential of the CD90− population was also veri-fied under standard chondrogenic conditions using TGFβ, a knownchondrogenic inductor [41]. Compared to the unstimulated control,TGFβ significantly increased cartilage-specific collagen II (Col2A1) andproteoglycan core aggrecan (ACAN) gene expression within 3 and14 days, respectively (Fig. 3B).

We also assessed the white and brown adipogenic potentials of theCD90− population. Unlike white adipocytes, brown adipocytes are spe-cialized in adaptive thermogenesis inwhichUCP1 plays a key role and isa specific marker of this cell type [31,32]. Since UCP1-expressing adipo-cytes are present in human HO (Fig. 1F) and since the CD90− hmrMSCpopulation has a strong adipogenic potential in vitro (Fig. 2B), we deter-mined whether this population could give rise to white adipocytes orUCP1-expressing brown adipocytes. Human adipose-derived stemcells can differentiate into white or brown adipocytes depending onthe length of rosiglitazone (ROS) treatment in adipogenic differentia-tion medium [42]. We used this approach with the CD90− cells todrive white and brown adipocyte formation. Gene expression analysesrevealed that the levels of the general adipogenic factors FABP4, ADIPOQand PPARγ were higher in the white (ROS 3d) and brown (ROS 14d)adipogenic conditions than in the unstimulated control (Fig. 3C). Atday 14, brown adipocyte marker UCP1 mRNA levels were significantlyhigher in the cell preparations treated to induce white (ROS 3d) andbrown (ROS14d) adipocyte formation (38- and4900-fold, respectively)than in the unstimulated control (Fig. 3C). The increase in UCP1 expres-sion was confirmed by immunofluorescence and Western blotting

B

PE-Cy7 CD31

Per

CP

Cy5

.5 C

D10

5

APC CD73

SS

C-A

FITC CD90

V50

0

V450

FSC

-W

FSC-HFSC-H

PE

CD

34

250k

200k

150k

100k

50k

00 50k 100k 150k 200k 250k 0 50k 100k 150k 200k 250k

41±2%11±8%

100%

105

101

102

103

104

SS

C-A

105

102

103

104

105

102

103

104

00

105

101

102

103

104

105

102

103

104

0

0 102

103

104

105

0 102

103

104

105

0 102

103

104

105

0 102

103

104

105

A

Adip

oCh

ondr

oO

steo

Unsorted CD73+CD105+CD90- CD73+CD105+CD90+

Fig. 2. FACS enrichment and multilineage differentiation potential of hmrMSCs. Adherent cells from the digested skeletal muscles of six donors were independently expanded, passagedand then enriched by FACS. (A) Representative gating strategy for donorM38 to sort hmrMSCs. Cells were forward (FSC) and side-scatter (SSC) gated to exclude debris and doublets. Totallive cellswere gated based on calcein violet fluorescence (V450) and aqua dye exclusion (V500). Hematopoietic (CD34) and endothelial (CD31) cellswere excluded from the analysis. Twopopulations of CD73+CD105+ cells were sorted based on their differential expression of CD90. The percentages of each cell population are shown in the panels and are expressed as themeans of three independent experiments. (B) CD73+CD105+CD90− cells display the same osteogenic differentiation potential in vitro as the unsorted population. However, theiradipogenic and chondrogenic differentiation capacities are greatly enriched compared to the unsorted control.Mineralized, multilayered nodules in cultures grown in osteogenicmediumstainedwith alizarin red (scale bar, 250 μm). Triglycerides inmature adipocytes stainedwith oil redO (scale bar, 250 μm).Alcian blue-positive cartilage inparaffin sections of pellets of cellsgrown in chondrogenicmedium (scale bar, 100 μm). Representative images are shown for donor M38. Each experimentwas performed at least three times with cells from four indepen-dent donors (M37, M38, M58 and M60).

168 J. Downey et al. / Bone 71 (2015) 164–170

(Figs. 3D, E). However, UCP1 protein levels were similar in the whiteand brown differentiation conditions (Fig. 3E), suggesting a mixedadipogenic population of white and brown adipocytes.

Olmsted-Davis et al. reported that brown adipocyteswere present ina murinemodel of HO triggered by BMPs, which drive the early steps ofheterotopic endochondral ossification by lowering oxygen tension inadjacent tissue [19]. Moreover, recent results have suggested thatBMPs also induce neurogenic inflammation,which enhances adrenergicstimulation by the sympathetic nervous system (SNS) [43]. The SNS isknown to induceUCP1-expressing brown adipocytes, likely through ad-renergic stimulation [44]. Further studies are required to gain a better

understanding of the effect of SNS stimulation on human muscle resi-dent progenitor cells in HO.

4. Conclusions

Weare thefirst to show thatUCP1-positive adipocytes are present inhuman HO. We are also the first to enrich a subpopulation of CD90−

hmrMSCs from adult human skeletal muscle that can give rise to allthe lineages present in HO (osteogenic, chondrogenic and adipogenic(white and brown)). The characterization of this progenitor cell sub-population is essential for understandingmuscle regenerative disorders

B

D

E

0

5

10

15

PPARγ

*******

020406080

100

5 000

10 000 ****

UCP1

***

***

048

20 00060 000

100000300000500 000

ADIPOQ

****

********

Rel

ativ

e m

RN

Aex

pres

sion

026

105001500250060008000

10 000

FABP4

****

********

Unstimulated ROS 3d ROS 14d

DAPI

/ UC

P1

Unstimulated ROS 3d ROS 14d

0.0

0 .5

1 .0

1 .5

2 .0

ROS3d 14d-R

elat

ive

prot

ein

expr

essi

on

GAPDH

UCP1

ROS3d 14d- UCP1

*****

A

Rel

ativ

e m

RN

Aex

pres

sion

0

2

4

6

0

2

450

100

150

0

2

4

6

8

10

SP7 DLX5

C

***

ACAN

0

5

10

15

20

Col2A1Chondrogenic

Unstimulated BMP9 Unstimulated TGFβ

Rel

ativ

e m

RN

Aex

pres

sion

****

Osteogenic

Adipogenic

*** ****

Fig. 3. Characterization of themultilineage differentiation potential of CD73+CD105+CD90− hmrMSCs progenitors. (A) qPCR of osteogenic genes after 7 days of BMP9 stimulation. Geneswere normalized to TATA-binding protein (TBP). Data are expressed asmeans± SEM (n=4,N=3). ***P b 0.002; ****P b 0.0001. (B) qPCR of early chondrogenic Col2A1 gene expressionfollowing 3 days of TGFβ stimulation, and late chondrogenic ACAN gene expression following 14 days of TGFβ stimulation. Data are expressed as means ± SEM (n = 4, N = 3).***P b 0.001; ****P b 0.0001. (C) qPCR of classic white and brown adipocyte genes after 14 days in adipogenicmedium following 0, 3, or 14 days of ROS treatment. Genes were normalizedto TBP. Data are expressed asmeans± SEM (n=4;N=3). *P b 0.015; **P b 0.004; ***P b 0.006; ****P b 0.0001. (D) Representative UCP1 immunofluorescent staining of hmrMSCs (M68),UCP1 (red) and nucleus (DAPI, blue) (scale bar, 250 μm). (E)Western blotting for UCP1 (30 μg protein/lane) in hmrMSCs differentiated for 14 days following 0, 3, or 14 days of ROS stim-ulation in adipogenic medium. UCP1 protein expression was normalized to GAPDH. Data are expressed as means ± SEM (n= 3; N= 3). *P b 0.035. Each experiment was performed atleast three times with cells from three independent donors (M38, M58, M68).

169J. Downey et al. / Bone 71 (2015) 164–170

170 J. Downey et al. / Bone 71 (2015) 164–170

such as HO. In addition, due to their easy accessibility and reproducibil-ity across donors, CD90− hmrMSCs will be a valuable source of progen-itor cells for further studies and future therapies.

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.bone.2014.10.020.

Authors' roles

Study design: JD, DL, MR, NF, AS, FB and GG. Study conduct: JD, DL,PK, MR, RH, FB and GG. Data collection: JD, DL, PK, MR, RH, FB and GG.Data analysis: JD, DL, PK,MR, RH, NF, AS, FB and GG. Data interpretation:JD, DL, PK, MR, RH, NF, AS, FB and GG. Drafting the manuscript: JD, DL,NF, AS, FB and GG. Financial support: PK, KK, MR, RH, NF, FB and GG. Re-vising the content of the manuscript: JD, DL, PK, KK, MR, RH, NF, AS, FBand GG. Approving the final version of the manuscript: JD, DL, PK, KK,MR, RH, NF, AS, FB and GG. JD, DL, PK, KK, MR, RH, NF, AS, FB and GGtake responsibility for the integrity of the data analysis.

Disclosure of potential conflicts of interest

The authors declare no potential conflicts of interest.

Acknowledgments

We thank the orthopedic service at CHUS and Dr. Svotelis for theircollaboration as well as G. Bourgeau for proofreading our manuscript.JD is the recipient of scholarships from FREOS and PROTEO. GG holds aNew Investigator Award from FRQS. This work was supported by grantsfrom CFI, FRQS-ThéCell and CIHR.

References

[1] Yin H, Price F, Rudnicki MA. Satellite cells and the muscle stem cell niche. PhysiolRev 2013;93:23–67.

[2] Leblanc E, Trensz F, Haroun S, Drouin G, Bergeron E, Penton CM. BMP-9-inducedmuscle heterotopic ossification requires changes to the skeletal muscle microenvi-ronment. J Bone Miner Res 2011;26:1166–77.

[3] Potter BK, Burns TC, Lacap AP, Granville RR, Gajewski DA. Heterotopic ossificationfollowing traumatic and combat-related amputations. Prevalence, risk factors, andpreliminary results of excision. J Bone Joint Surg Am 2007;89:476–86.

[4] Shore EM, Kaplan FS. Inherited human diseases of heterotopic bone formation. NatRev Rheumatol 2010;6:518–27.

[5] Sullivan MP, Torres SJ, Mehta S, Ahn J. Heterotopic ossification after central nervoussystem trauma: a current review. Bone Joint Res 2013;2:51–7.

[6] Ji Y, Christopherson GT, Kluk MW, Amrani O, Jackson WM, Nesti LJ. Heterotopic os-sification following musculoskeletal trauma: modeling stem and progenitor cells intheir microenvironment. Adv Exp Med Biol 2011;720:39–50.

[7] Kan L, Kessler JA. Evaluation of the cellular origins of heterotopic ossification. Ortho-pedics 2014;37:329–40.

[8] Medici D, Shore EM, Lounev VY, Kaplan FS, Kalluri R, Olsen BR. Conversion of vascu-lar endothelial cells into multipotent stem-like cells. Nat Med 2010;16:1400–6.

[9] Lounev VY, Ramachandran R, Wosczyna MN, Yamamoto M, Maidment AD, ShoreEM, et al. Identification of progenitor cells that contribute to heterotopicskeletogenesis. J Bone Joint Surg Am 2009;91:652–63.

[10] Kisanuki YY, Hammer RE, Miyazaki J, Williams SC, Richardson JA, Yanagisawa M.Tie2-Cre transgenic mice: a new model for endothelial cell-lineage analysisin vivo. Dev Biol 2001;230:230–42.

[11] De Palma M, Venneri MA, Galli R, Sergi Sergi L, Politi LS, Sampaolesi M, et al. Tie2identifies a hematopoietic lineage of proangiogenic monocytes required for tumorvessel formation and a mesenchymal population of pericyte progenitors. CancerCell 2005;8:211–26.

[12] Wosczyna MN, Biswas AA, Cogswell CA, Goldhamer DJ. Multipotent progenitors res-ident in the skeletal muscle interstitium exhibit robust BMP-dependent osteogenicactivity and mediate heterotopic ossification. J Bone Miner Res 2012;27:1004–17.

[13] Uezumi A, Fukada S, Yamamoto N, Ikemoto-Uezumi M, Nakatani M, Morita M, et al.Identification and characterization of PDGFRalpha + mesenchymal progenitors inhuman skeletal muscle. Cell Death Dis 2014;5:e1186.

[14] Oishi T, Uezumi A, Kanaji A, Yamamoto N, Yamaguchi A, Yamada H, et al. Osteogenicdifferentiation capacity of human skeletal muscle-derived progenitor cells. PLoS One2013;8:e56641.

[15] Jackson WM, Lozito TP, Djouad F, Kuhn NZ, Nesti LJ, Tuan RS. Differentiation and re-generation potential of mesenchymal progenitor cells derived from traumatizedmuscle tissue. J Cell Mol Med 2011;15:2377–88.

[16] Nesti LJ, Jackson WM, Shanti RM, Koehler SM, Aragon AB, Bailey JR, et al. Differenti-ation potential of multipotent progenitor cells derived from war-traumatized mus-cle tissue. J Bone Joint Surg Am 2008;90:2390–8.

[17] Zheng B, Cao B, Crisan M, Sun B, Li G, Logar A, et al. Prospective identification ofmyogenic endothelial cells in human skeletal muscle. Nat Biotechnol 2007;25:1025–34.

[18] Dilling CF, Wada AM, Lazard ZW, Salisbury EA, Gannon FH, Vadakkan TJ, et al. Vesselformation is induced prior to the appearance of cartilage in BMP-2-mediated hetero-topic ossification. J Bone Miner Res 2010;25:1147–56.

[19] Olmsted-Davis E, Gannon FH, OzenM, IttmannMM, Gugala Z, Hipp JA, et al. Hyp-oxic adipocytes pattern early heterotopic bone formation. Am J Pathol 2007;170:620–32.

[20] Crisan M, Casteilla L, Lehr L, Carmona M, Paoloni-Giacobino A, Yap S, et al. A reser-voir of brown adipocyte progenitors in human skeletal muscle. Stem Cells 2008;26:2425–33.

[21] Pisani DF, Djedaini M, Beranger GE, Elabd C, Scheideler M, Ailhaud G, et al. Differen-tiation of Human Adipose-Derived Stem Cells into "Brite" (Brown-in-White) Adipo-cytes. Front Endocrinol 2011;2:87.

[22] Lecourt S, Marolleau JP, Fromigue O, Vauchez K, Andriamanalijaona R, Ternaux B,et al. Characterization of distinct mesenchymal-like cell populations from humanskeletal muscle in situ and in vitro. Exp Cell Res 2010;316:2513–26.

[23] Vermette M, Trottier V, Menard V, Saint-Pierre L, Roy A, Fradette J. Production of anew tissue-engineered adipose substitute from human adipose-derived stromalcells. Biomaterials 2007;28:2850–60.

[24] Lee CH, Shah B, Moioli EK, Mao JJ. CTGF directs fibroblast differentiation from humanmesenchymal stem/stromal cells and defines connective tissue healing in a rodentinjury model. J Clin Invest 2010;120:3340–9.

[25] Abramoff MD, Magelhaes PJ, Ram SJ. Image Processing with ImageJ. Biophotonics Int2004;11:36–42.

[26] Chang CB, Han SA, Kim EM, Lee S, Seong SC, Lee MC. Chondrogenic potentials ofhuman synovium-derived cells sorted by specific surface markers. OsteoarthritisCartilage 2013;21:190–9.

[27] De Sousa M, Porras DP, Perry CG, Seale P, Scime A. p107 is a crucial regulator for de-termining the adipocyte lineage fate choices of stem cells. Stem Cells 2014;32:1323–36.

[28] Kloen P, Lauzier D, Hamdy RC. Co-expression of BMPs and BMP-inhibitors in humanfractures and non-unions. Bone 2012;51:59–68.

[29] Grenier G, Leblanc E, Faucheux N, Lauzier D, Kloen P, Hamdy RC. BMP-9 expressionin human traumatic heterotopic ossification: a case report. Skelet Muscle 2013;3:29.

[30] Grenier G, Scime A, Le Grand F, Asakura A, Perez-Iratxeta C, Andrade-Navarro MA,et al. Resident endothelial precursors in muscle, adipose, and dermis contribute topostnatal vasculogenesis. Stem Cells 2007;25:3101–10.

[31] Shabalina IG, Petrovic N, de Jong JM, Kalinovich AV, Cannon B, Nedergaard J. UCP1 inbrite/beige adipose tissue mitochondria is functionally thermogenic. Cell Rep 2013;5:1196–203.

[32] Frontini A, Cinti S. Distribution and development of brown adipocytes in the murineand human adipose organ. Cell Metab 2010;11:253–6.

[33] Dominici M, Le Blanc K, Mueller I, Slaper-Cortenbach I, Marini F, Krause D, et al. Min-imal criteria for defining multipotent mesenchymal stromal cells. The InternationalSociety for Cellular Therapy position statement. Cytotherapy 2006;8:315–7.

[34] ChungMT, Liu C, Hyun JS, Lo DD, Montoro DT, HasegawaM, et al. CD90 (Thy-1)-pos-itive selection enhances osteogenic capacity of human adipose-derived stromal cells.Tissue Eng Part A 2013;19:989–97.

[35] Mitchell JB, McIntosh K, Zvonic S, Garrett S, Floyd ZE, Kloster A, et al.Immunophenotype of human adipose-derived cells: temporal changes instromal-associated and stem cell-associated markers. Stem Cells 2006;24:376–85.

[36] Scharenberg MA, Pippenger BE, Sack R, Zingg D, Ferralli J, Schenk S, et al. TGF-beta-induced differentiation intomyofibroblasts involves specific regulation of two MKL1isoforms. J Cell Sci 2014;127:1079–91.

[37] Koumas L, Smith TJ, Feldon S, Blumberg N, Phipps RP. Thy-1 expression in human fi-broblast subsets defines myofibroblastic or lipofibroblastic phenotypes. Am J Pathol2003;163:1291–300.

[38] Nishimura R, Hata K, Matsubara T, Wakabayashi M, Yoneda T. Regulation of boneand cartilage development by network between BMP signalling and transcriptionfactors. J Biochem 2012;151:247–54.

[39] Chen G, Deng C, Li YP. TGF-beta and BMP signaling in osteoblast differentiation andbone formation. Int J Biol Sci 2012;8:272–88.

[40] Kang Q, Sun MH, Cheng H, Peng Y, Montag AG, Deyrup AT, et al. Characterization ofthe distinct orthotopic bone-forming activity of 14 B.P. using recombinantadenovirus-mediated gene delivery. Gene Ther 2004;11:1312–20.

[41] Wang MK, Sun HQ, Xiang YC, Jiang F, Su YP, Zou ZM. Different roles of TGF-beta inthe multi-lineage differentiation of stem cells. World J Stem Cells 2012;4:28–34.

[42] Elabd C, Chiellini C, Carmona M, Galitzky J, Cochet O, Petersen R, et al. Humanmultipotent adipose-derived stem cells differentiate into functional brown adipo-cytes. Stem Cells 2009;27:2753–60.

[43] Salisbury EA, Lazard ZW, Ubogu EE, Davis AR, Olmsted-Davis EA. Transient brownadipocyte-like cells derive from peripheral nerve progenitors in response to bonemorphogenetic protein 2. Stem Cells Transl Med 2012;1:874–85.

[44] Mattsson CL, Csikasz RI, Chernogubova E, Yamamoto DL, Hogberg HT, Amri EZ, et al.beta(1)-Adrenergic receptors increase UCP1 in human MADS brown adipocytes andrescue cold-acclimated beta(3)-adrenergic receptor-knockout mice via nonshiveringthermogenesis. Am J Physiol Endocrinol Metab 2011;301:E1108–18.