CD146 expression on mesenchymal stem cells is associated with their vascular smooth muscle commitment

Bone marrow mesenchymal stem cells (MSCs) are plastic adherent cells that can differentiate into various tissue lineages, including osteoblasts, adipocytes and chondrocytes. However, this progenitor property is not shared by all cells within the MSC population. In addition, MSCs vary in their proliferation capacity and expression of markers. Because of heterogeneity of CD146 expression in the MSC population, we compared CD146−/Low and CD146High cells under clonal conditions and after sorting of the non-clonal cell population to determine whether this expression is associated with specific functions. CD146−/Low and CD146High bone marrow MSCs did not differ in colony-forming unit-fibroblast number, osteogenic, adipogenic and chondrogenic differentiation or in vitro haematopoietic-supportive activity. However, CD146−/Low clones proliferated slightly but significantly faster than did CD146High clones. In addition, a strong expression of CD146 molecule was associated with a commitment to a vascular smooth muscle cell (VSMC) lineage characterized by a strong up-regulation of calponin-1 and SM22α expression and an ability to contract collagen matrix. Thus, within a bone marrow MSC population, certain subpopulations characterized by high expression of CD146, are committed towards a VSMC lineage.


Introduction
Mesenchymal stem/stromal cells (MSC) are multipotent progenitors that give rise to skeletal cells (osteoblasts, chondrocytes, haematopoietic-supportive stromal cells) and adipocytes [1][2][3][4]. They are relatively easy to expand, and their progenitor nature is evaluated by colony-forming unit-fibroblast (CFU-F) assay [5]. Although MSCs are often characterized after a period of culture, recently, many studies have aimed to define their native form in bone marrow aspirates. These works emphasize different membrane molecules as native MSC markers. Among them, CD200, a marker for MSCs appeared to be a good marker to reproducibly purify native MSCs. The in vitro adipogenic, osteogenic and chondrogenic potential of CD200 + cells was similar to that for cells separated by adherence [6]. Sacchetti et al., from cultured and native cells by using sorting and clonal methods, found CD146 as a key marker for bone marrow MSCs (BM-MSCs).
Culture-expanded clonogenic CD146+ cells and not CD146À cells were osteogenic in vivo and could re-establish the haematopoietic microenvironment in a xenotransplantation model [7]. A homogenous CD45À/CD146+ multipotent MSC population deriving from human BM exhibited haematopoiesis-supporting abilities, an extensive 12week proliferation and the ability to differentiate in osteoblasts, chondrocytes and adipocytes [8]. Another study suggested that CD146 might be a marker of clonal multipotent cultured MSCs [9].
However, from CD271+/CD45À BM fractions Tormin et al. observed CD146À/low and CD146+ populations with CFU-F potential with any differences in multilineage differentiation capacity. Moreover, the CD146 expression by BM-MSCs depended on the environment: upregulated under normoxia and down-regulated under hypoxia. This O 2related expression was associated with the in situ localization: CD146 expressing reticular cells were located in perivascular regions, whereas cells close to the bone surface were CD146À [10]. In addition, adipose-tissue perivascular cells with MSC properties, such as pericytes of microvessels and capillaries, were identified to have high CD146 expression, whereas other cells from tunica adventitia from large vessels lacked CD146 [11,12]. Therefore, the MSC expression of CD146 is heterogeneous and may depend on the tissue and the molecular environment. This observation suggests functional differences. Therefore, in this study, we investigated BM-MSC functions in terms of CD146 expression. We compared sorted and clonogenic CD146 À/Low and CD146 High cells after in vitro expansion and examined the different properties of MSC such as osteogenic, chondrogenic, adipogenic and vascular smooth muscle cell (VSMC) differentiation; haematopoiesis support; proliferation; CFU-F formation; transcriptome and phenotype to distinguish between these two BM-MSC subpopulations.

Material and methods
Preparation of single cell-derived clonal-cultured MSCs and sorted MSCs Human BM-MSCs were isolated by culture from the BM of healthy donors obtained during the preparation of allogeneic haematopoietic stem cell grafts. This tissue is considered waste material in France and does not require informed consent for use, in accordance with French ethical and legal regulations. Briefly, BM-MSCs were obtained from unprocessed BM without red blood lysis nor density-gradient method and seeded at 5 9 10 4 cells/cm 2 into a 150-cm 2 flask with minimum essential medium a (aMEM; Life Technologies, Saint Aubin, France) supplemented with 10% foetal calf serum (FCS; Lonza, Levallois-Perret, France) and 10 lg/ml ciprofloxacin (Bayer, Puteaux, France). For all MSC cultures, the medium was renewed twice a week until cells reached confluence (P1). Cells were then detached with trypsin (Gibco, Life Technologies) [13,14].
For clonal studies, total BM cells were seeded in 24-well plates at 1-2 9 10 4 /cm 2 /well in aMEM supplemented with 10% FCS and ciprofloxacin. Wells were screened every day starting at day 7 to identify wells with one clone. After 10-14 days, the wells containing only one colony (CFU-F) of MSCs were selected and cells were expanded. After expansion, MSC clones were detached by use of trypsin (Invitrogen), counted and submitted to phenotypic characterization. Cytometry was used to select clones on the basis of CD146 expression: CD146 mean fluorescence intensity was used to determine CD146 Low and CD146 High clones. The doubling population number was calculated by considering the number of CFU-Fs as the number of initiating cells at day 0.
For non-clonal studies, BM cells were seeded at 5 9 10 4 cells/cm 2 in the same medium at day 0 for 21 days (P1). After trypsin treatment, MSCs at passage P1 were used for phenotypic and functional studies. To sort CD146+ and CD146À MSCs, MSCs were incubated with anti-FITC-CD146 antibody (Becton-Dickinson, Le Pont de Claix, France) and separated by use of a fluorescent activated cell sorter (BD ARIA II, "Service commun de cytometrie de Toulouse Purpan"). Sorted cells were seeded at 1000 cells/cm 2 and grown to confluence to calculate population doubling.

CFU-F assay
To evaluate the frequency of CFU-Fs generated, clonal BM-MSCs were seeded at 8 cells/cm 2 for 10 days. Cultures were then washed with PBS, fixed with methanol and stained with Giemsa (Oxoid, Dardilly, France); colonies with more than 50 cells were counted.

Incorporation of 5-ethynyl-2′-deoxyuridine (EdU)
During last 3 days of culture, non-clonal and clonal BM-MSCs were incubated with 5 lM Click-iT EdU (Molecular Probes, Life Technologies). After treatment, clonal BM-MSCs were stained with APC-conjugated anti-CD146 monoclonal antibody, fixed, permeabilized and treated with a Click-iT reaction cocktail for detection following the manufacturer's recommendations. Samples were analysed by use of an ADPCyan flow cytometer and Kaluza software (Beckman-Coulter).

Affymetrix microarrays
After clonal BM-MSC culture and CD146 expression analysis, the RNA in non-clonal and clonal CD146 À/Low and CD146 High MSCs from four donors was extracted by use of the RNeasy Kit (Qiagen, Hilden, Germany). RNA purity and integrity were checked by use of Bioanalyzer 2100 (Agilent, Santa Clara, CA, USA). The RNA integrity number was >9 for all samples used for microarray assays. The accession number in the Geo data set in Medline is available at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE48540.

RNA extraction/cDNA synthesis and quantitative PCR (qPCR)
RNA was extracted by use of the AllPrep ARN/ADN/protein kit (Qiagen, Valencia, CA, USA). cDNA synthesis was performed with 1 lg total RNA with the High Capacity cDNA reverse-transcription kit (Applied Biosystems, Life Technologies) and random hexamers. qPCR was performed on diluted cDNA (equivalent to 25 ng of starting purified RNA) with SSoFast EvaGreen Supermix (Biorad) and 500 nM forward and reverse primers in a total volume of 20 lL on a CFX Real Time System (Bio-Rad, Marnes-la-Coquette, France) at 95°C, 3 min. and 40 cycles of denaturation (95°C, 10 sec.) and primer hybridation and amplification (60°C, 30 sec.). Primer sequences are in Table 1. Each primer couple displayed interpretable PCR efficiencies (95-105%). Melting curves and an appropriate No-RT control were used to validate amplification specificity. Data were analysed by use of the Bio-Rad CFX manager (threshold=0.2) and exported to DataAssist Software v3.0 (Applied Biosystems). Gene expression was calculated by the 2 ÀDCt (or 2 ÀDDCt ) method with PPIA used as an appropriate reference gene because it had the lesser M score.
Cell contraction assay CD146 High/Low clones were incubated in collagen matrices in 24-well plates for 24 hrs during which stress developed. To initiate contraction, we released collagen gels from the sides of the culture dishes. After 5 hrs, the surface area of the matrix contracted was expressed as a percentage of its initial surface area.

Statistical analysis
Data are expressed as mean AE SEM of (n) separate experiments. Statistical comparisons involved Student's t-test and GraphPad Prism software (La Jolla, CA, USA). Differences were considered statistically significant at P < 0.05.

CFU-F and proliferation potential of MSCs and CD146 expression
From BM, we expanded non-clonal MSCs or clonal MSCs and selected them by CD146 expression (CD146 À/Low and CD146 High , Fig. 1A). For non-clonal cells, CD146 expression was heterogeneous (Fig. 1Aa), but clonogenic cells showed homogenous negative-to-low or high expression of CD146 (CD146 À/Low and CD146 High ) ( Fig. 1Ab and c). Clonal CD146 À/Low and CD146 High fractions produced a similar number of CFU-Fs (20 AE 5 versus 27 AE 19, respectively; Fig. 1B). In addition, sorted non-clonal CD146 À/Low and CD146 High cells did not differ in generating CFU-Fs ( Fig. S1A and B).
Proliferation potential was slightly but significantly higher for clonal CD146 À/Low than CD146 High cells (population doubling number: 20.9 AE 0.7 versus 20.3 AE 0.6, P = 0.016) at week 4 ( Fig. 1C) and was similar with sorted non-clonal CD146À/+ cells (Fig. S1C). As confirmed by flow cytometry, the proportion of cycling cells (percentage of EdU-incorporated cells) was significantly higher in the clonal CD146 À/Low than CD146 High fraction (43% versus 24%, Fig. 1D). Of note, CD146 expression was enhanced in clonal CD146 À/Low cells after proliferation period but did not reach the level detected in the clonal CD146 High population.

Functional assessment of CD146 À/Low and CD146 High MSCs
To test whether CD146 À/Low and CD146 High clones exhibited differences in multilineage differentiation capacity, we tested their ability to differentiate into osteoblasts, adipocytes and chondrocytes in vitro. CD146 À/Low and CD146 High cells did not differ in differentiating into osteoblasts and adipocytes with alizarin red and Nile red staining, respectively (Fig. 2Aa-f). To confirm these data, we quantified osteoblasts by expression of Runt-related transcription factor 2 (Runx2), osterix (Osx), alkaline phosphatase (ALPL) and distal-less homeobox 5 (DLX5); adipocytes by expression of peroxisome proliferator-activated nuclear receptor c 2 (PPARc2) and fatty acid binding protein 4 (FABP4); and chondroblasts by expression of as collagen 10 A1 (COLL10A1) differentiation genes by transcriptomic analysis (Fig. 2Ba-g). Results show that CD146 High and CD146 Low groups did not clearly differ in multidifferentiation capacity. In addition, CD146 High and CD146 Low cells showed weaker expression of stemness genes (NANOG), octamer-binding transcription factor 4 (OCT4) and SRY-related HMG-box 2 (SOX2) as compared with non-clonal populations (Fig. S2Aa-c). This difference could be explained by the fact that clones underwent more culture during the selection of clones; thus the doubling number was higher for clonal than non-clonal MSCs (20 versus 12). Of note, the expression of NANOG and OCT4 was slightly higher for CD146 Low than CD146 High cells.
Then, to evaluate haematopoiesis support activities in vitro, cordblood CD34 + haematopoietic stem cells (HSCs) were co-cultured for 28 days on stroma composed of cells from CD146 À/Low or CD146 High clones. Non-adherent cells were counted weekly and seeded in multilineage semisolid medium for clonogenic assay. Proliferation of nonadherent CD34+ haematopoietic cells was higher (not significantly) with co-culture on CD146 À/Low than CD146 High stroma (Fig. 2C a), but total number of CFU-Gs, CFU-Ms, CFU-GMs and BFU-Es were similar on both stromas (Fig. 2Cb-e).

High expression of CD146 in MSCs is associated with commitment to VSM lineage
The levels of the MSC markers CD90, CD73 and CD105 were similar in CD146 À/low and CD146 High populations derived from clones (Fig. 3A). Interestingly, the expression of CD49a (a1-integrin subunit) was upregulated in CD146 High clones. After their expansion, cells were subjected to gene array analysis. We obtained a list of genes up-regulated in CD146 High clones and identified a high expression of CD146, thus confirming our discrimination between studied clones. In this list, we found keratins (KRT7, KRT34, KRT81), CNN1 and LIM-calponin homology domain (LIMCH1) linked to a VSM lineage; insulin-like growth factor binding protein 2 (IGFBP2) expressed in VSMCs; inhibin A (INHBA), a basement membrane protein; and SPARC-related modular calcium binding 1 (SMOC1) and other genes not yet characterized (Fig. 3B, see accession no. GSE48540 in Material and methods).
Because the CD49a molecule is up-regulated in the VSMC commitment program [20,21], we focalized on VSMC-specific molecules in CD146 À/Low and CD146 High cells such as CNN1, which was significantly up-regulated in CD146 High clonal MSCs (Fig. 3B b). To confirm these data, we performed transcriptome analysis by qRT-PCR of CNN1 and found that CNN1 was significantly up-regulated in CD146 High cells (Fig. 3C). We evaluated other VSMC-specific transcripts and found a tendency for up-regulation of early VSMC markers, smooth muscle protein 2a (SM22a) and elastin (ELN) in CD146 High cells (Fig. S2B).
After gene array and transcriptome analysis, we studied CNN1 protein expression in CD146 À/low and CD146 High populations. Most Table 1 List of primers

Forward primer
Reverse primer  Fig. 4A). Finally, after VSMC phenotype analysis in clones, we sought to test CD146 High and CD146 Low subpopulations at a functional level. Thus, cells were assessed for collagen matrix contraction capacity [22,23]. Interestingly, CD146 High clones contracted more matrix than CD146 Low clones at 5 hrs after contraction initiation (51.7 AE 9.4% versus 63.3 AE 2.1% of initial surface of matrix; Fig. 4B). Therefore, CD146 High MSCs had more VSMC characteristics than CD146 Low clones.

Effect of FGF-2 and TGF-b1 on MSC proliferation, CD146 expression and VSMC commitment
Because TGF-b1 is a key cytokine for inducing VSMC differentiation [16,17,[24][25][26][27], we assessed the effect of TGF-b1 on CD146 expression in MSCs and VSMC differentiation. In addition, we used FGF2, a cytokine with mitogenic effect that promotes proliferation of undifferentiated MSCs in vitro [15,28]. From crude bone marrow-derived MSCs treated with FGF2 or TGF-b1, we   In contrast, proliferation was slowed with TGF-b1 (9.5 AE 2.8 population doubling, Fig. 5Aa). Moreover, as determined by flow cytometry with EdU incorporation, the proportion of cycling cells (EdU-incorporated cells) was significantly lower with TGF-b1 than FGF2 treatment (27.8% versus 49.2%, Fig. 5Ab and c). Second, CD146 expression was lower in MSCs with than without FGF2 treatment (9.9 AE 3.7 versus 20.4 AE 8.4), but higher with TGF-b1 treatment (33.1 AE 12.5, Fig. 5B). In parallel and in agreement with Figure 5B, the median fluorescence intensity (MFI) of CD146 staining on EdU-incorporated cells was up-regulated with TGF-b1 as compared with FGF2 (12.1 versus 4.7), so CD146 expression was enhanced by TGF-b1 treatment (Fig. 5A b and c). Furthermore, the presence of a CD146 À population after TGF-b1 treatment can be explained by our use of unfractioned BM whole MSC populations in terms of CD146 expression.
To confirm the VSMC commitment by these treated cells, we tested CNN1 expression in all conditions. Untreated MSCs expressed a basal level of CNN1, which was up-regulated by TGF-b1 as compared with FGF-2 (Fig. 5Ca-c). The expression of SM22a, an early VSMC marker, was strong and stable in all conditions (Fig. 5Cd-f). Therefore, similar to the behaviour of single cell-derived CD146 High clones, a high level of CD146 expression in the MSC population induced by TGF-b1 was associated with low proliferation and commitment to a VSMC lineage.

Discussion
Soluble agents as well as cell density, spatial distribution of cells, and solid components of culture (such as extracellular matrix molecules) are increasingly considered critical to cell fate determination and heterogeneity of cultured MSCs. Some experiments suggested that even when derived from a single cell, the progeny of MSCs can be conditioned to behave differently [29,30]. This heterogeneity could be a result of extensive ex vivo culturing, in vivo heterogeneity and plasticity of phenotype reflecting the natural repertoire of MSCs.
To study the different heterogenous subpopulations of MSCs and define specific markers, we focused on CD146, an endothelial and subendothelial marker well described in the MSC literature. CD146, also known as melanoma cell adhesion molecule (Mel-CAM, MCAM) or MUC18 belongs to the immunoglobulin superfamily. In normal adult tissues, CD146 is primarily expressed by vascular endothelium and smooth muscle cells [31][32][33]. Because of the heterogeneity of CD146 expression in MSC populations, we compared CD146 À/Low and CD146 High cells from clonal and sorted MSCs after culture expansion to determine whether this expression is associated with specific functions. We sorted CD146 High and CD146 À/Low cells with excellent purity from eight healthy donors after primary culture. CD146 High and CD146 À/Low cells did not differ in CFU-F frequency, cell expansion or expression of MSC markers (CD90, CD73, CD44, CD105, CD13, CD166) or adherence molecules (CD106, CD49d, CD49f; data not shown). However, this approach has two potential issues: (i) the surface re-expression of CD146 in sorted CD146 À cells after a short period of culture and (ii) the low number of cells obtained after sorting not being amenable to functional studies. To bypass these issues and to have a sufficient quantity of stable CD146À and CD146+ cells to compare, we generated single cell-derived CD146 + clonogenic MSCs [7]. After expansion of these clonogenic MSCs, most clones expressed CD146 at a moderate level (52% of CD146 int ); 21% were CD146 À/Low and 27% were CD146 High . In this study, we compared clones with extreme levels of CD146 expression, CD146 À/Low and CD146 High .
As previously found, CD146 À/Low and CD146 High MSCs did not differ in CFU-F number or osteogenic, chondrogenic and adipogenic differentiation capacity [10]. Recently, CD146 knockdown in human MSCs was found to disturb proliferation and osteogenic differentiation [34]. This disagreement has a few explanations. First, use of RNA interference or ectopic overexpression can have adverse effects on MSC signalling, whereas in our study, we used 'untouched' physiological clones expressing different levels of CD146 depending on their fate in culture. Second, we used clones with high population doubling number, whereas the previous authors used MSCs passaged only once.
CD146 High and CD146 À/Low clones showed similar in vitro haematopoietic-supportive activity, which agrees with Tormin et al., 111 who demonstrated no difference in stroma-supporting capacity between CD271+/CD45À/CD146 À/Low and CD271+/CD45À/ CD146 +/High fractions in long-term culture initiating cells [10]. However, in a xenotransplantation model, contrary to the CD146À fraction, culture-expanded CD146+ cells could re-establish the haematopoietic microenvironment [7]. Such discrepancies could be explained by the experimental conditions. We and Tormin et al. In addition, we found proliferation slightly but significantly higher for CD146 À/Low than CD146 High MSCs. Accordingly, FGF2 treatment decreased CD146 expression but increased proliferation, as was previously described [7]. Conversely, TGF-b1 treatment was associated with increased CD146 expression and VSMC differentiation. The association of CD146 and the VSMC differentiation lineage is reinforced by the fact that the CD146 promoter contains the gene regulation ele-ment CarGbox, which is present in all promoters of key genes for the VSMC lineage (aSM-actin, SM22a, CD49a) [33]. Alpha1 integrin subunit (CD49a) is one of the SMC-specific marker genes [35]. In our study, CD49a was up-regulated in CD146 High clonal MSCs as was the specific VSMC marker CNN1 at the gene and protein level. In addition, early VSMC markers SM22a and ELN were up-regulated in CD146 High MSCs. Also, CD146 High MSCs exhibited better potential for contraction of collagen matrix than did CD146 À/Low counterparts. Vascular smooth muscle cells are multipotential cells with the ability to differentiate into osteoblasts (resulting in vascular calcification), chondroblasts and adipocytes depending on the inducers (BMP, oxidative stress, etc.) [36]. Because CD146 High MSCs share some properties with VSMCs (e.g. phenotype and matrix contraction potential), they would also exhibit multipotentiality. Therefore, CD146 expression by MSCs seems to be a characteristic and predictive marker for commitment towards the VSMC lineage. Interestingly, the VSMCs were previously related to BM stromal cells with haematopoietic sustaining activities [37]. Indeed, the in vitro model of long-term haematopoiesis requires a layer of stromal cells generated by Dexter-like culture sytems [38,39]. This type of culture constrains BM stromal adherent cells to shift towards the VSMC lineage [38,40]. Despite some reports of osteoblastic localization for the HSC niche, others demonstrated a vascular situation [41]. In this latter context, HSCs are associated with parasinusal cells, which are similar to advential reticular cells and to parasinusal myoid cells [40,42,43]. All of these cells express aSM-actin. In foetal and adult human BM, aSM-actin-expressing cells form haematopoietic niches [44]. Sacchetti et al. described CD146+ cells expressing VSMC linage markers (aSM-actin, CNN1, Desmin) with the ability to form a haematopoietic microenvironment [7]. Recently, Peault's group showed in humans that only CD146+ cells could maintain in vitro HSCs with long-term repopulating potential [45]. Moreover, in agreement with our results, in vitro investigations of haematopoiesissustaining properties revealed no differences between CD146 +/High and CD146 À/Low stromal cells; only in vivo experiments showed differences. Overall, the CD146 +/High MSCs represent a fraction of whole BM-MSCs with few detectable differences as compared with CD146 À/Low MSCs. Nevertheless, the differences concerning the commitment towards a VSMC lineage and a high CD146 expression in the MSC population reflect the start of this commitment.

Supporting information
Additional Supporting Information may be found in the online version of this article: