In Vitro High-Capacity Assay to Quantify the Clonal Heterogeneity in Trilineage Potential of Mesenchymal Stem Cells Reveals a Complex Hierarchy of Lineage Commitment§


  • Katie C. Russell,

    1. Department of Chemical and Biomolecular Engineering, Tulane University, New Orleans, Louisiana, USA
    2. Biomedical Sciences Graduate Program,Tulane University, New Orleans, Louisiana, USA
    Search for more papers by this author
  • Donald G. Phinney,

    1. Biomedical Sciences Graduate Program,Tulane University, New Orleans, Louisiana, USA
    2. Center for Gene Therapy, Tulane University, New Orleans, Louisiana, USA
    Search for more papers by this author
  • Michelle R. Lacey,

    1. Department of Mathematics, Tulane University, New Orleans, Louisiana, USA
    Search for more papers by this author
  • Bonnie L. Barrilleaux,

    1. Department of Chemical and Biomolecular Engineering, Tulane University, New Orleans, Louisiana, USA
    2. Biomedical Sciences Graduate Program,Tulane University, New Orleans, Louisiana, USA
    Search for more papers by this author
  • Kristin E. Meyertholen,

    1. Department of Chemical and Biomolecular Engineering, Tulane University, New Orleans, Louisiana, USA
    Search for more papers by this author
  • Kim C. O'Connor

    Corresponding author
    1. Department of Chemical and Biomolecular Engineering, Tulane University, New Orleans, Louisiana, USA
    2. Biomedical Sciences Graduate Program,Tulane University, New Orleans, Louisiana, USA
    3. Center for Gene Therapy, Tulane University, New Orleans, Louisiana, USA
    • Department of Chemical and Biomolecular Engineering, Tulane University, Boggs Center Room 300, New Orleans, Louisiana 70118, USA
    Search for more papers by this author
    • Telephone: 504-865-5740; Fax: 504-865-6744

  • Author contributions: K.R.: Collection and/or assembly of data, data analysis and interpretation, manuscript writing; D. P.: Conception and design, provision of study materials or patients, data analysis and interpretation; M.L.: Data analysis and interpretation; B.B. and K.M.: Collection and/or assembly of data; K.O.: Conception and design, financial support, administrative support, data analysis and interpretation, manuscript writing, final approval of manuscript.

  • First published online in STEM CELLS EXPRESS February 1, 2010.

  • §

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


In regenerative medicine, bone marrow is a promising source of mesenchymal stem cells (MSCs) for a broad range of cellular therapies. This research addresses a basic prerequisite to realize the therapeutic potential of MSCs by developing a novel high-capacity assay to quantify the clonal heterogeneity in potency that is inherent to MSC preparations. The assay utilizes a 96-well format to (1) classify MSCs according to colony-forming efficiency as a measure of proliferation capacity and trilineage potential to exhibit adipo-, chondro-, and osteogenesis as a measure of multipotency and (2) preserve a frozen template of MSC clones of known potency for future use. The heterogeneity in trilineage potential of normal bone marrow MSCs is more complex than previously reported: all eight possible categories of trilineage potential were detected. In this study, the average colony-forming efficiency of MSC preparations was 55–62%, and tripotent MSCs accounted for nearly 50% of the colony-forming cells. The multiple phenotypes detected in this study infer a more convoluted hierarchy of lineage commitment than described in the literature. Greater cell amplification, colony-forming efficiency, and colony diameter for tri- versus unipotent clones suggest that MSC proliferation may be a function of potency. CD146 may be a marker of multipotency, with ∼2-fold difference in mean fluorescence intensity between tri- and unipotent clones. The significance of these findings is discussed in the context of the efficacy of MSC therapies. The in vitro assay described herein will likely have numerous applications given the importance of heterogeneity to the therapeutic potential of MSCs. STEM CELLS 2010;28:788–798


Bone marrow is a promising source of mesenchymal stem cells (MSCs) for regenerative medicine. They proliferate readily in culture, differentiate into various cell lineages, regulate the immune response, and promote the growth of host cells [1, 2]. For example, the trilineage potential to exhibit adipo-, chondro-, and osteogenesis is a basic criterion for defining multipotent MSCs [3]. Thus, MSC therapies are under development to treat a broad range of diseases including myocardial infarction [4], renal failure [5], and osteoarthritis [6]. The efficacy of these therapeutic applications is highly dependent on the intrinsic heterogeneity of MSC preparations [2]. Clonal analysis has revealed that MSCs are a heterogeneous mixture of cells that differ in their stage of lineage commitment and extent of differentiation [7, 8, 9]. Despite its importance in defining potency, there have been only limited investigations of the heterogeneity in trilineage potential of MSCs and underlying hierarchical relationships. This research addresses the deficiency through the development of a high-capacity assay that classifies MSCs according to their proliferative capacity and trilineage potential.

The hierarchy of MSC lineage commitment was initially described as a sequential loss of adipogenic and then chondrogenic potential to yield osteogenic progenitors [7, 10]. Conversely, a subsequent study generated MSC clones that exhibited adipogenesis but not chondrogenesis, suggesting that the hierarchical relationships may be more complex than originally proposed [11]. This latter study is supported by further research with multipotent fibroblasts derived from human dermis, which revealed a bifurcated hierarchy that produces bipotent progenitors with either an osteo-adipogenic or osteo-chondrogenic phenotype [12]. The assay described in this research will resolve the discrepancy between these opposing results.

Many challenges arise when developing an assay to characterize MSC heterogeneity. Without standardized procedures to isolate and culture MSCs, a specific immunophenotype that is indicative of potency has yet to be identified [13, 14]. Likewise, the transcriptome and proteome expression profiles of MSCs are highly dependent on their culture conditions [15]. Sorting MSCs according to light-scattering properties during flow cytometry provides only a partial enrichment of multipotent cells [16]. These obstacles can be overcome by characterizing MSCs by their functional ability to proliferate and differentiate. Individual MSC clones have been isolated and their potency evaluated [10]; however, a high-capacity format is required to obtain statistically significant results. High-capacity assays in ≥96-well microplates have been developed to assess proliferation [16] and differentiation to a specific lineage, such as chondrocytes [17], but they do not evaluate multipotency nor present a strategy to retain large numbers of clones for additional analysis once function has been determined. The high-capacity assay presented here is unique in its ability to quantify the trilineage potential of MSC clones, couple measurements of potency and proliferative potential, and preserve a template of frozen clones for future use.

In developing the in vitro assay, there were substantial differences in the extent of MSC differentiation when a variety of popular culture conditions were compared. The high-capacity assay detected a clonal heterogeneity in MSC potency that is consistent with a nonlinear hierarchy of lineage commitment. Tripotent MSCs were prevalent in culture, and lineage commitment was found to diminish cell proliferation, colony-forming efficiency, and colony diameter. Additionally, CD146 expression was markedly higher in clones that were tripotent than in unipotent clones. The significance of these findings is discussed in the context of the efficacy of MSC therapies. The assay has the potential to be a fundamental tool in stem-cell technology with many clinical and basic research applications.


MSC Cultivation

Primary MSCs were harvested from 2 ml of iliac crest bone marrow aspirate from healthy adult volunteers by the Center for Preparation and Distribution of Adult Stem Cells formerly at Tulane University School of Medicine (New Orleans, LA) and currently at Texas A&M Health Science Center (Temple, TX) as previously described [18]. Cell culture supplies were obtained from Invitrogen, Carlsbad, CA, except where noted. Plastic-adherent cells were inoculated at 50–100 cells/cm2 and amplified in complete culture medium (CCMA) consisting of α-mimimum essential medium (α-MEM) with 2 mM L-glutamine supplemented with 17% FBS (HyClone, Logan, UT, and an additional 2 mM L-glutamine. Furthermore, 100 U/ml of penicillin and 100 μg/ml of streptomycin were added to all media in this study. MSCs were maintained at less than 50% confluence and subcultured with 0.25% trypsin/1 mM EDTA. Passage 2 cells were used in all experiments and maintained in a 37°C humidified incubator at 5% CO2 with medium exchange every 3–4 days. Viable cell density was measured by trypan blue staining and hemocytometer counting.


Trypsinized MSCs were washed by centrifugation in PBS. Nonspecific antigens were blocked by incubating the cells at 106 cells/ml in PBS containing 1% bovine serum albumin for 20 minutes at 37°C. Aliquots of 100 μl cell suspension were incubated at 4°C for 20 minutes with one of the fluorochrome-conjugated, anti-human monoclonal antibodies listed in supporting information Table 1. Labeled samples were washed by centrifugation in three volumes of phosphate-buffered saline (PBS). Isotype controls were run in parallel at the same concentration used for each antibody. The immunophenotype of MSCs was evaluated with a FC500 flow cytometer (Beckman Coulter, Fullerton, CA,

Cell Cloning

To detect single cells during cloning, MSCs were sterilely labeled in situ for 10 minutes with 5 μM CellTracker Green, 5-chloromethylfluorescein diacetate (λexem = 492/517 nm), in serum-free CCMA prewarmed to 37°C, according to Invitrogen's instructions. After trypsinization, MSC clones were generated by limiting dilution into a 96-well microplate containing 50 μl/well of fresh CCMA and 75 μl/well of CCMA conditioned by MSCs for 48 hours and sterile-filtered (0.2 μm pore size) to remove any suspended cells. Each well was examined with a fluorescent Olympus IX50 microscope (Olympus America, Center Valley, PA, to determine the plating efficiency as the percentage of wells inoculated with a single cell. Fifty microliters of fresh CCMA was added to each well 3 days after inoculation and 50 μl of medium was replaced with fresh CCMA after an additional 3 days. On day 7 of cultivation, colony-forming efficiency was calculated as the percentage of MSC colonies originating from a single cell, divided by the fraction of viable cells in the inoculum. Clonal colonies containing at least 300 cells/well were subcultured at a 1:4 ratio to evaluate trilineage potential and preserve a master plate of frozen clones.

Osteogenic Differentiation

After subculturing, MSC clonal colonies were expanded for 7 days in 96-well microplates containing 150 μl/well of CCMA until ∼75% confluent. Osteogenesis was induced with one of two differentiation media (ODM1 and ODM2). The former consists of low-glucose Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovice serum (FBS), 100 nM dexamethasone (Sigma-Aldrich, St. Louis, MO,, 10 mM β-glycerophosphate (Sigma-Aldrich), and 50 μM L-ascorbic acid 2-phosphate (Sigma-Aldrich) [19]. ODM2 was prepared by supplementing CCMA with 10 nM dexamethasone, 10 mM β-glycerophosphate, and 50 μg/ml of L-ascorbic acid 2-phosphate [20]. On day 21 of cultivation in osteogenic medium, MSCs were fixed in 4% paraformaldehyde for 20 minutes and stained with 1% Alizarin Red S (pH 4.2, Sigma-Aldrich) for 20 minutes to detect calcified extracellular matrix. After qualitative analysis by microscopy, the stain was extracted with 100 μl/well of 10% cetylpyridinium chloride in 10 mM sodium phosphate buffer (pH 7.0) for 15 minutes, and the spectral absorbance of the solution was measured at 562 nm [21].

Adipogenic Differentiation

MSC clonal colonies were amplified in CCMA for 7 days as described above and then differentiated over 21 days in an adipogenic medium (ADM1) of CCMA supplemented with 0.5 μM dexamethasone, 0.5 mM isobutylmethylxanthine (Sigma-Aldrich), and 50 μM indomethacin (Sigma-Aldrich) [20]. Alternatively, clonal colonies were initially cultivated in an adipogenic induction medium (ADM2i) of high-glucose DMEM supplemented with 10% FBS, 10 μg/ml of insulin, 1 μM dexamethasone, 0.5 mM isobutylmethylxanthine, and 0.1 mM indomethacin. After 72 hours, ADM2i was replaced with adipogenic maintenance medium (ADM2m): high-glucose DMEM supplemented with 10% FBS and 10 μg/ml of insulin. After an additional 24 hours, this sequence was repeated three times, followed by cultivation in ADM2m for 1 week [22]. To visually detect lipid accumulation, cultures were fixed in 4% paraformaldehyde for 20 minutes and stained with 0.3% Oil Red O (Sigma-Aldrich) in 0.6% isopropanol for 1 hour. The content of Oil Red O in samples was quantified by extraction with 100 μl/well of isopropanol for 5 minutes, followed by spectrophotometry at 510 nm [18]. For more sensitive detection of adipogenesis to evaluate potency, 5 μl of AdipoRed reagent (Lonza, Walkersville, MD, was added to MSC samples in 200 μl of PBS/well and mixed immediately. After 10 minutes, the fluorescence was measured with excitation of 485 nm and emission of 535 nm.

Chondrogenic Differentiation

Chondrogenesis was induced in monolayer and pellet cultures. Clonal colonies were amplified for 7 days in 96-well microplates containing 150 μl/well of CCMA for monolayer cultures or, alternatively, for nearly 2 weeks in 6-well plates containing 2 ml of CCMA/well for pellet cultures. Amplified clones were cultivated as pellet cultures in 96-well, V-bottom polypropylene microplates (Thermo Fisher Scientific, Fremont, CA, that had been inoculated with 2 ± 0.2 × 105 cells/well and contained 200 μl/well of CCMA [17]. Cell pellets formed by gravity sedimentation within the first 24 hours after inoculation and did not adhere to the well during cultivation. To induce differentiation in monolayer and pellet cultures, CCMA was replaced with one of two chondrogenic media: CDM1, high-glucose DMEM supplemented with 100 ng/ml of bone morphogenetic protein-2 (R&D Systems Inc., Minneapolis, MN,, 10 ng/ml of transforming growth factor-β3, 100 nM dexamethasone, 50 μg/ml of L-ascorbic acid 2-phosphate, 100 μg/ml of pyruvate (Sigma-Aldrich), 40 μg/ml of proline (Sigma-Aldrich), and 10 μl/ml of ITS+ (BD Biosciences, San Diego, CA, [23]; and CDM2, high-glucose DMEM supplemented with 10 ng/ml of transforming growth factor-β1, 100 nM dexamethasone, 37.5 μg/ml of L-ascorbic acid 2-phosphate, 1 mM pyruvate, and 10 μl/ml of ITS+ [24].

Three weeks after cultivation in differentiation medium, monolayer cultures were fixed in 4% paraformaldehyde for 20 minutes and then stained overnight at 25°C with 1% Alcian Blue 8-GX (Sigma-Aldrich) in 0.1 N HCl (pH 1.0) to detect matrix deposition of sulfated glycosaminoglycans (GAGs) [25]. Pellet cultures were fixed in 10% buffered formalin, embedded in paraffin, sectioned at 5 μm, deparaffinized, and hydrated before staining. After visual inspection, Alcian Blue in monolayer cultures was extracted in 1% Sodium dodecyl sulfate (SDS), and the concentration in solution was evaluated by absorbance measurement at 605 nm [26]. To determine GAG content in pellet cultures, cell aggregates were digested at 65°C for 1 hour in 200 μl/well of 300 μg/ml papain (Sigma-Aldrich) in 50 mM sodium phosphate buffer containing 2 mM cysteine and 2 mM EDTA (pH 6.5). One hundred microliters of the papain digest was added to 1 ml of 16 μg/ml 1,9-dimethylmethylene blue (DMMB, Sigma-Aldrich) in 0.5% ethanol/0.2% formic acid containing 2 mg/ml of sodium formate. The absorbance of the solution was read immediately at 535 nm [27]. To evaluate potency, the GAG–DMMB complex was precipitated and then solubilized according to the method of Barbosa et al. [28] for more sensitive GAG detection. Chondroitin sulfate A was used as a standard.

DNA Quantitation

DNA content of MSC cultures was quantified as previously described [17]. Briefly, 100 μl of papain-digested cell samples was mixed with 200 μl 0.1 M NaOH, incubated at room temperature for 30 minutes, and then neutralized by the addition of 200 μl 0.1 M HCl in 5 M NaCl, 100 mM sodium phosphate. In a 96-well microplate, 100 μl of the sample was added to 100 μl of 0.7 μg/ml Hoechst 33258 (Sigma-Aldrich) in 10 mM Tris buffer containing 200 mM NaCl and 1 mM EDTA (pH 7.4). Fluorescence was measured with excitation of 340 nm and emission of 465 nm. A standard curve was prepared with calf thymus DNA (Sigma-Aldrich).

Freezing Clones

A master plate of frozen clones was prepared immediately after the initial subculture of MSC clonal colonies by adding 50 μl of the trypsinized cell suspension to a fresh 96-well microplate containing 50 μl/well of 2× freezing media (65% CCMA, 27% FBS, and 8% dimethyl sulfoxide (DMSO)). To inhibit degassing of CO2 and medium evaporation, 100 μl of filter-sterilized light paraffin oil was added to the top of each well, and the lid was secured with parafilm. The plate was transferred to a styrofoam box, frozen overnight at −80°C, and then placed directly into a −80°C freezer for long-term storage. As little as 75 cells/well were frozen by this method. Plates were defrosted in a 37°C incubator, and thawed cell suspensions were transferred to 24-well plates containing 1 ml of fresh CCMA/well. A day after inoculation, the medium was exchanged with fresh CCMA for routine cell amplification.

Image Analysis

Images of stained histological samples of differentiated MSCs were captured with an Optronics DEI-750 digital camera (Optronics, Goleta, CA, mounted onto an Olympus IX50 microscope. Staining was evaluated in terms of the product of the percent stained area and its optical density [29]. Areas of positive staining were traced using a Graphire 4 CTE-640 tablet (Wacom Technology Corp., Vancouver, WA, and analyzed with the Area and Optical Density options in the Count/Size and Draw tools of Image-Pro Plus software Version 6.1 (Media Cybernetics, Crofton, MD, [30]. Optical density was calculated on a scale of 0 (white) to 2.5 (black) relative to the corresponding negative control using the Background Correction option. The diameter of MSC clonal colonies was determined after tracing images of colonies stained with Crystal Violet using the Mean Diameter option of Image-Pro Plus software. Imaging results are reported as a mean value from 20–50 randomly selected images on average per culture sample.

Statistical Analysis

To account for experimental variation among culture wells, measurements for each sample were normalized to a set of negative controls. All samples were centered and scaled using the mean and standard deviation (SD) computed for the controls, so that reported differentiation measurements provide the standard score for each sample relative to the control values. To determine appropriate cut-off values for positive differentiation, the R open source statistical software package (version 2.7, was employed to generate Gaussian kernel density estimates for the distribution of measurements observed for the undifferentiated MSC controls. Values corresponding to the 95th percentiles from the fitted control densities were selected as lower limits of detection for differentiation.

Cell cloning by limiting dilution was analyzed by the goodness of fit to a Poisson distribution with a chi-squared test. Comparisons of differentiation of matched MSC clones under alternative culture conditions and marker expression were analyzed with a two-tailed, paired Student's t-test. Proliferation and colony-forming properties of MSC clones were compared using the Kruskal-Wallis nonparametric analysis of variance, followed by Dunn's multiple comparisons test. p-values of 0.05 or less were considered statistically significant.


Immunophenotype and Colony-Forming Efficiency

The cells employed in this study exhibited an immunophenotype that is characteristic of MSCs (e.g., CD44+, CD90+, and CD166+) and were negative for cell-surface epitopes indicative of hematopoietic cells (e.g., CD34, CD36, and CD45), in agreement with previous findings [1] (supporting information Table 1). Colony-forming efficiency was evaluated by limiting dilution of fluorescently labeled MSCs (Figs. 1A–1D), which readily formed colonies under these clonogenic conditions (Fig. 1E). The tracker dye CellTracker Green facilitated cell detection in wells on a microplate as shown in the fluorescence micrographs and corresponding phase-contrast images in Figures 1A–1D. The dye had negligible cytotoxicity: the viability of the inoculum was 90 ± 4% (Fig. 1F). Cell cloning into 96-well plates followed a Poisson distribution. Specifically, a single cell was inoculated into a well with a plating efficiency of 26 ± 4%; the remaining wells contained multiple MSCs or were void of cells (Fig. 1F). For the representative MSC donor preparation shown in Figure 1, wells initially harboring a single MSC formed colonies after a week of cultivation with an efficiency of 55 ± 7% (n > 3), which is consistent with published values by other methods [18].

Figure 1.

Plating and colony-forming efficiency during cloning for a representative mesenchymal stem cell (MSC) donor preparation (A–D): MSCs were detected with CellTracker Green during limiting dilution. Fluorescence micrographs (A, B) and corresponding phase-contrast micrographs (C, D) of the labeled MSCs. Scale bar: 100 μm. (E): Cell colonies in 96-well microplates stained with Crystal Violet after 21 days in culture. (F): Viability of the inoculum, efficiency of plating clones, and colony-forming efficiency reported for triplicate microplates.

Culture Conditions for Differentiation

In addition to measuring colony-forming efficiency, our high-capacity assay classifies MSCs according to potency, specifically their trilineage potential to exhibit adipo-, chondro-, and osteogenesis. To develop our in vitro assay, we compared popular differentiation media using matched monolayer cultures from MSC clones and based our selection on the extent of mineralization during osteogenesis, lipid accumulation during adipogenesis, and sulfated GAG deposition during chondrogenesis. There was no significant difference in Alizarin Red S staining of calcified extracellular matrix when matched cultures produced from the same clonal colony were incubated for 21 days in ODM1 or ODM2, even though the osteogenic media differs by 10-fold in the concentration of dexamethasone (Figs. 2A–2D). For adipogenic differentiation, more lipids were visible in matched cultures exposed continuously to ADM1 versus a combination of alternating ADM2i and ADM2m (Figs. 2E–2G). Absorbance readings from extracted Oil Red O confirmed these observations (p < .05) (Fig. 2H). CDM1 containing bone morphogenetic protein-2 and transforming growth factor-β3 produced greater Alcian Blue staining for GAG relative to CDM2 containing transforming growth factor-β1 (Figs. 2I–2K). The standardized absorbance of extracted Alcian Blue was 25–100% higher for 6 of the 10 MSC clonal cultures differentiated in CDM1 (p < .01) (Fig. 2L). Media producing the greatest degree of differentiation were chosen for our assay. DNA content of these confluent cultures varied by ≤25% for different clones in a given medium (e.g., 0.6 ± 0.1 μg/well in ODM1).

Figure 2.

Comparison of osteo- (A–D), adipo- (E–H), and chondrogenesis (I–L) for MSCs cultivated in different differentiation media for 21 days. Representative matched clonal cultures stained with Alizarin Red S (A–C) to detect mineralization induced with ODM1 (A) or ODM2 (B), Oil Red O to detect lipid accumulation induced with ADM1 (E) or ADM2i and m (F), and Alcian Blue to detect GAG deposition induced with CDM1 (I) or CDM2 (J). Negative control: matched clones maintained in CCMA (C, G, K). Scale bars: 100 μm. Standardized absorbance of extracted histological stains for matched clonal colonies: (D), ODM1 (black) and ODM2 (white); (H), ADM1 (black) and ADM2i and m (white); and (L), CDM1 (black) and CDM2 (white). Measurements were standardized with a mean ± SD of 0.25 ± 0.04 (D), 0.13 ± 0.02 (H), and 0.11 ± 0.01 (L) absorbance units for the negative control.

For chondrogenesis, MSCs in V-bottom polypropylene microplates containing CDM1 formed aggregated cell pellets with a diameter of approximately 1 mm (Figs. 3A, 3B). Pellet cultures deposited substantially more GAG per microgram of DNA than monolayers of matched clones differentiated in the same medium (Fig. 3C), consistent with previous findings [17]. Standardized GAG/DNA content was ≥2-fold higher in MSC pellets for half of the clonal cultures examined (p < .01). On the basis of these findings, our high-capacity assay utilized pellet cultures in V-bottom microplates to assess the chondrogenic potential of MSCs.

Figure 3.

Effect of pellet culture on chondrogenesis. (A): MSC clones were plated at 2 ± 0.2 × 105 cells per well into 96-well, V-bottom polypropylene microplates to induce pellet formation. (B): Representative pellet cultivated in CDM1. Scale bar: 1.0 mm. (C): Standardized μg GAG/μg DNA of pellets (black) compared with matched clonal colonies cultivated as monolayers (white). GAG/DNA content was standardized with a mean ± SD of 1.1 ± 0.1 μg GAG/μg DNA for MSC monolayers in complete culture medium as the negative control. Abbreviation: GAG, glycosaminoglycan.

Trilineage Potential of MSC Clones

The trilineage potential of clonal cultures of MSCs was evaluated using the differentiation conditions specified above. Clonal colonies were subcultured at a 1:4 ratio into replicate microplates: three plates were employed to evaluate trilineage potential; the fourth was frozen to preserve a template of clones for future use. Figure 4 depicts the extent of differentiation for a representative set of 96 MSC clones cultivated in osteo-, adipo-, and chondrogenic medium and 24 negative controls of the parent MSC preparation maintained in CCMA growth medium. In particular, the density plots in Figure 4 provide the fitted probability distributions of clones in the 96-well assay and negative controls, so that the area under the curve above a selected interval represents the estimated portion of clones that have a standardized differentiation measurement within that interval. MSC clones were designated as positive for differentiation if their standardized measurement exceeded the 95th percentile of the estimated probability density for the negative controls. As an example, a MSC clone with a standardized absorbance ≥1.6 was designated as positive for osteogenesis for the representative data in Figure 4A; clones with lower absorbance readings were classified as negative. Over three experiments that each included 24–26 negative controls, the average cutoff values were 1.9 ± 0.2, 2.1 ± 0.1, and 1.6 ± 0.3 standard deviations for osteo-, adipo-, and chondrogenesis, respectively.

Figure 4.

Representative clonal variation in the extent of trilineage differentiation of mesenchymal stem cells (MSCs). Fitted kernel density estimates and standardized differentiation measurements for 24 negative controls (light gray) and 96 MSC clones (dark gray) cultivated in osteo- (ODM1, A), adipo- (ADM1, B), and chondrogenic medium (CDM1, C) as monolayer (A, B) and pellet cultures (C) for 21 days. Negative control: MSC monolayers cultivated in complete culture medium during the same period. The 95th percentile threshold values for positive differentiation are represented by vertical dashed lines. Standardized measurements are calculated with a mean ± SD of 0.25 ± 0.04 absorbance units (A), 2,800 ± 500 RFU (B), and 0.29 ± 0.06 μg GAG/μg DNA (C). Abbreviations: GAG, glycosaminoglycan; RFU, relative fluorescence units.

Correlations were established to relate standardized differentiation measurements (Fig. 4) to histological staining of differentiated MSCs (Fig. 5). The histological samples were monolayers for osteo- and adipogenesis (Figs. 5A–5J) and sectioned pellets for chondrogenesis (Figs. 5K–5O). Staining was evaluated in terms of the product of the percent stained area and its optical density [29]. This area-times-intensity score was ≤30 for osteogenesis (Figs. 5A–5E), ≤ 10 for adipogenesis (Figs. 5F–5J), and ≤25 for chondrogenesis (Figs. 5K–5O). The higher percentages for osteo- and chondrogenesis may be due to staining extracellular versus intracellular markers and/or a greater degree of differentiation. Linear regression models were sufficient to fit the data in all cases (Figs. 5E, 5J, 5O). From these equations, the 95th percentile threshold, which was described in the previous paragraph, corresponds to an area-times-intensity score of 1.0 ± 0.1 (Fig. 5E), above which clonal cultures were designated as positive for osteogenesis. The cutoff values for adipo- and chondrogenesis were equivalent to scores of 0.20 ± 0.01 and 2.9 ± 0.2, respectively (Figs. 5J, 5O). In all three cases, the threshold area-times-intensity scores were at least 8-fold less than the maximum values observed for positive clones.

Figure 5.

Correlations between histological staining of mesenchymal stem cells (MSCs) and standardized measurements of osteo- (A–E), adipo- (F–J), and chondrogenesis (K–O). Representative histological samples stained with Alizarin Red (A–D), AdipoRed (F–I), and Alcian Blue (K–N): negative control (A, F, K); osteogenic MSC clones with average standardized absorbance of 4.3 (B), 30 (C), and 44 (D); adipogenic clones with average standardized fluorescence of 73 (G), 190 (H), and 510 (I); and chondrogenic clones with average standardized μg GAG/μg DNA of 4.4 (L), 10 (M), and 37 (N). Scale bars: 100 μm. Correlation graphs: clones are designated positive (▴) or negative (○) for differentiation based on a 95% confidence level with standardized scores of 1.9 ± 0.2 for osteogenesis (E), 2.1 ± 0.1 for adipogenesis (J), and 1.6 ± 0.3 for chondrogenesis (O). Inserted graphs are plotted on an expanded scale. Culture conditions for differentiation are the same as for Figure 4. Negative control: MSCs in complete culture medium. Abbreviation: GAG, glycosaminoglycan.

The threshold criteria were applied to standardized differentiation measurements for 96 clonal cultures from two MSC donor preparations with colony-forming efficiencies of 55 ± 7% and 62 ± 10% (n > 3). All eight possible categories of trilineage potential were observed (Fig. 6A). Tripotent MSCs accounted for nearly half of colony-forming cells in our study. For bipotent populations, there were notable differences between donors. For one donor preparation, ∼20% of the colony-forming MSCs exhibited the osteo-chondrogenic phenotype, whereas the second donor had an equivalent percentage of the osteo-adipogenic progenitors. In both cases, this was at least four times more prevalent than the remaining two bipotent phenotypes in culture. Unipotent osteoprogenitors represented on the order of 10% of the colony-forming cells in culture, whereas unipotent MSCs with an adipogenic phenotype were detected infrequently. MSC clones that did not exhibit the potential to differentiate into any of the three lineages were rare. These findings suggest a complex hierarchical relationship among MSC populations (Fig. 6B). Tripotent MSCs commit to a specific lineage via one of three possible bipotent states: osteo-chondrogenic, osteo-adipogenic, or adipo-chondrogenic. Further loss of lineage potential from these intermediate phenotypes yields osteo-, adipo-, and chondrogenic unipotent progenitors.

Figure 6.

Heterogeneity in potency of mesenchymal stem cell clones from two donors (A) and corresponding hierarchy of lineage commitment (B). High-capacity assay detects tripotent, bipotent (OC, OA, and AC), and unipotent (O, C, A) clones, and clones that do not differentiate into any of the three lineages. Abbreviations: A, adipogenic phenotype; C, chondrogenic phenotype; O, osteogenic phenotype.

Recovery of Frozen Clones

One of the challenges with the high-capacity assay is that MSCs of known potency are identified retrospectively, after 3 weeks of differentiation. This necessitates the preservation of frozen clones for additional analysis once lineage potential has been determined. To facilitate clone storage, a methodology is presented to freeze MSC clonal colonies in situ within a microplate. As demonstrated in Figure 7A, tri-, bi-, and unipotent MSCs that are frozen in this manner can be subsequently amplified. After thawing and a 3-day recovery, 100 ± 10 MSCs were seeded into 24-well microplates and amplified for 6 days. At the end of the expansion period, cultures were ≤60% confluent. Amplification of tripotent clonal colonies ranged from 9- to 76-fold with a median of 28-fold. This was significantly higher than that for unipotent MSCs, which were expanded 2- to 44-fold during this period with a median of 5-fold (p < .05). The same trend in cell expansion applies to clones of different potency prior to freezing and, thus, is not an artifact of cryopreservation (supporting information Fig. 1). This suggests that MSC proliferation may be a function of potency.

Figure 7.

Initial characterization of mesenchymal stem cell (MSC) clones of known potency. Clones were frozen in situ within 96-well microplate, thawed, and cultured for 3 days before further use. To evaluate proliferative potential (A, B), 100 ± 10 cells/well were seeded into (A) 24-well plates containing 1 ml complete culture medium (CCMA)/well and amplified for 6 days (≤60% confluent) or (B) 10-cm dishes each containing 13 ml CCMA and cultivated for 14 days. To evaluate colony-forming efficiency (B), MSCs were stained with Crystal Violet. Median ([BOND]) calculated from n = 10 clones (A, B). To evaluate CD146 expression (C–E), 10–15 clones of a given potency were amplified to 2–3 × 104 cells, potency of the expanded cultures was confirmed, and cultures were pooled to form a sample of ∼3 × 105 cells for flow cytometry. Representative histograms of CD146 expression in pooled tripotent clones (C, dashed curve) and unipotent clones (D, dashed curve) relative to parent MSCs (solid curve) and isotype control (gray). Mean fluorescent intensity (E) of MSC groups labeled with CD146 antibody (black) or isotype control (white) is reported as mean ± SD from two independent experiments. ∗, p < .05 and , p ≤ .01 versus tripotent clones (A, B); ∗, p < .05 versus parent MSCs (E).

Colony-forming efficiency had a similar dependence on potency (Fig. 7B). Tripotent clones formed colonies with a median efficiency of 50% as compared with 14% and 1% for bi- and unipotent clones, respectively (p < .01). Likewise, colonies that formed from tripotent clones had the largest median diameter (supporting information Fig. 2). Combined, these data suggest that lineage commitment diminishes proliferative potential, which manifests itself in a lower colony-forming efficiency and smaller colony size.

Expression of Cell-Surface Markers

On the basis of our findings in Figures 7A and B, we investigated the expression level as a function of potency of a representative subset of cell-surface antigens in supporting information Table 1 that had been reported in the literature to have differential expression as a function of proliferation rate and/or colony-forming efficiency: CD44, CD73, CD146, and CD271 [31, 32, 33]. Of this group, only CD146 exhibited a correlation between marker expression and potency. To generate the data in Figures 7C–7E, we amplified 10–15 clonal colonies of a given potency to 2–3 × 104 cells, confirmed the potency of the expanded MSCs, and pooled the cultures to form a sample size of ∼3 × 105 cells for analysis by flow cytometry. Histograms from the pooled tripotent clones were shifted to higher CD146 expression relative to the parent MSC preparation from which the clones were generated (Fig. 7C), whereas histograms of parent MSCs and their unipotent progeny were similar (Fig. 7D). In particular, the mean fluorescent intensity of tripotent clones was nearly twice the value for their unipotent counterparts: 79 ± 4.6 compared with 41 ± 6.4 (p < .05) (Fig. 7E).

There are conflicting data in the literature concerning CD44 and CD73 expression. Higher levels of expression for these markers on faster-growing MSCs have been reported [32], but others did not detect this correlation [34]. In our study, there were no noticeable differences in the expression levels of CD44 and CD73 among our sample groups of MSCs (data not shown). Expression of CD271, a possible marker for colony-forming MSCs [33], was negligible in our sample groups and donor preparations (supporting information Table 1).


To overcome current limitations in MSC identification, a novel high-capacity assay has been presented that quantifies the clonal heterogeneity inherent to MSCs by analyzing individual cells for their colony-forming efficiency as a measure of proliferative potential and their clonal progeny for trilineage potential as a measure of potency. A master plate of frozen clones enables selective amplification based on potency. The in vitro assay revealed a highly heterogeneous MSC culture in which tripotent cells were the most prevalent population. Clonal heterogeneity observed in this study is consistent with a nonlinear hierarchy of lineage commitment. Recovery and amplification of frozen clones suggest a relationship between MSC proliferation and potency that was supported by the colony-forming properties of the clones. An initial immunophenotyping of clones revealed differential expression of CD146 as a function of potency. As discussed in further detail below, the ability of the assay to isolate cells of known potency and quantify clonal heterogeneity has utility to improve the efficacy of MSC therapies.

MSC Heterogeneity

Our investigation of potency revealed that the heterogeneity in trilineage potential of normal bone marrow MSCs is more complex than previously reported. All eight possible categories of trilineage potential were observed in our investigation, whereas previous research detected only subsets of these categories. For example, most of the studies with normal bone marrow MSCs reported the existence of tripotent, osteo-chondrogenic, and osteogenic phenotypes [7, 10, 35]. These three groups were among the most prevalent in our study. Another investigation of normal bone marrow MSCs isolated six clones: all were osteogenic, five were adipogenic, and two were chondrogenic [19]. Previous research with immortalized bone marrow MSCs detected seven of the eight categories reported here [36]. In particular, the adipo-chondrogenic phenotype was not detected in the analysis of 100 immortalized clones [36]. Likewise, this category of cells was in the minority in our study with normal MSCs. More than 60% of immortalized clones did not exhibit the potential to differentiate into any of the three lineages, and less than 5% of immortalized clones were tripotent [36]. This trend was reversed in our investigation. Discrepancies among the studies can be attributed to differences in culture conditions, immortalization, detection of differentiation, number of clones examined, passage number, and/or donors. In our own study, for example, MSC donor preparations with comparable colony-forming efficiencies differed in their relative content of the osteo-adipogenic and osteo-chondrogenic progenitors.

There are numerous applications for an assay that quantifies clonal heterogeneity and preserves clones of known potency for future use. Efficacy of MSC therapies has been traditionally attributed to their potency [16]. Our assay can be employed, for instance, to monitor MSC preparations in the clinic for consistent content of multipotent cells and to determine the variability in heterogeneity among different donors, with age and under different culture conditions. Furthermore, there is a growing body of evidence that suggests that the heterogeneity of MSCs may contribute to their broad therapeutic efficacy, with multiple cell populations participating in tissue repair through diverse mechanisms that include the regulation of inflammation and apoptosis [2]. Frequently these repair mechanisms are examined with the entire MSC preparation rather than its constituent populations [37]. Our assay may help resolve the contribution of MSC populations to specific therapeutic properties as a function of potency. Last, it has been postulated that some cytokines, such as fibroblast growth factor-2, act on MSCs in a stage-specific manner, eliciting different responses as the cells become committed to a particular lineage [38]. Here too, the assay can be useful in testing this hypothesis.

Hierarchy of Lineage Commitment

Although the hierarchical model for generating blood cells from hematopoietic progenitors is well established [39], far less is known about the lineage commitment of MSCs derived from bone marrow. Previously, a linear hierarchy for MSCs was proposed in which tripotent cells beget osteoprogenitors through sequential loss of adipogenic and then chondrogenic potential [7, 10]. Our data suggest a greater variety of bi- and unipotent cells as shown in Figure 6. It is noteworthy that three of the four most prevalent populations in our study are the components of the linear model: tripotent cells, bipotent cells with osteo-chondrogenic potential, and osteoprogenitors. The prevalence of the osteo-adipogenic phenotype for one of the MSC donor preparations in our study suggests a more complex hierarchical model. Fate decisions of stem cells to commit to a specific pathway have been modeled as stochastic events that occur randomly [40], as deterministic processes that are influenced by the microenvironment [41] or as a combination of these two approaches [42]. Although tripotent MSCs may progress to an osteo-chondrogenic, osteo-adipogenic, or adipo-chondrogenic phenotype, the regulatory factors in the microenvironment of the bone marrow and/or during ex vivo amplification may favor a subset of these pathways. This is consistent with the rarity of adipo-chondrogenic progenitors for the two MSC donor preparations in our study. Until recently, lineage commitment had been considered primarily an irreversible process. Evidence of stem cell plasticity [43] and reprogramming of somatic cells to a stem-like state [44] suggest that hierarchical pathways may be reversible under some conditions.

Knowledge of these hierarchical relationships will facilitate the rational design of culture conditions for the ex vivo amplification of MSCs. Specifically, the hierarchy serves as a template for the development of computational models that predict the dynamic interactions among the distinct cell populations that give rise to the overall potency of an MSC preparation [45]. Predictive models can aid in the selection of amplification conditions to control the heterogeneity and, thus, the efficacy of MSC therapies. Additionally, the hierarchy in Figure 6 provides a framework with which to elucidate the components of transcriptional pathways that govern MSC lineage commitment, such as TAZ (transcriptional coactivator with a Pro-Pro-X-Tyr-binding motif), which regulates adipogenesis and osteogenesis [46]. Identification of transcriptional modulators of MSCs will present new molecular targets to control the heterogeneity of MSC therapies.

Relationship Between Potency and Proliferation

Differences in the amplification and colony-forming properties of clonal MSCs in our research suggest a relationship between cell potency and proliferation. These findings are consistent with previous results that note variation in the proliferative capacity among MSC clones [47]. A recent study grouped MSC clones into fast- and slow-growing categories. Similar to the trend reported here, all except one of the fast-growing clones were tripotent, whereas the slow-growing clones showed limited differentiation potential [48]. Consistent with our results on colony-forming efficiency, a highly clonogenic fraction of MSCs isolated from bone marrow was multipotent, exhibiting adipo- and osteogenesis [16]. It may be possible to exploit these differential growth kinetics to enrich multipotent cells in a heterogeneous MSC preparation during ex vivo amplification for clinical use. In fact, the greater proliferative potential of tripotent MSCs may have contributed to their higher frequency in our MSC preparations, which were cultivated at low cell density.

CD146 As a Cell-Surface Marker of Potency

The identification of cell-surface epitopes as markers of MSC potency has remained elusive. Our initial immunophenotyping of clonal MSCs revealed that tripotent clones express higher levels of CD146 than their unipotent counterparts. CD146 (also known as MCAM, Mel-CAM, S-Endo-1, A32 antigen, and MUC18) is expressed on several cell types (e.g., MSCs, endothelial cells, and melanoma cells) and participates in heterotypic intercellular adhesion [31, 49]. In agreement with our findings, CD146 has been reported to be upregulated on highly proliferative MSCs, which are additionally capable of trilineage differentiation [31, 50]. MSCs share similar functional and gene-expression profiles with CD146-positive perivascular cells [51], supporting the concept that the postnatal MSC niche is the perivascular site within microvessels [52]. In this microenvironment, CD146 may not only participate in heterotypic intercellular adhesion but also regulate proliferation through transmembrane signaling. Overexpression of CD146 in CD146-negative primary melanoma cells results in increased tumor growth [53]. This correlation between CD146 expression and proliferation is consistent with our results.


The research presented here addresses a basic deficiency in stem cell technology by developing a quantitative and high-capacity assay to characterize the clonal heterogeneity of MSC trilineage potential. To date, the assay has revealed a complex hierarchy of lineage commitment in which proliferation capacity and CD146 expression diminish with loss of trilineage potential. This methodology has numerous applications given the importance of heterogeneity to the therapeutic potential of MSCs. The insight into MSC populations that can be attained with this in vitro assay will ultimately enable control over the composition and, thus, the efficacy of MSC therapies.


We thank Alan Tucker for his assistance with flow cytometry, Dina Gaupp for preparing samples for histology, and Prof. Darwin Prockop for his helpful conversations and suggestions about this project. This research was supported with grants from the National Institutes of Health (1R03 EB007281) and National Science Foundation (BES-0514242). D.P. present address is Department of Molecular Therapeutics, The Scripps Research Institute, Jupiter, FL 33458, USA.


The authors indicate no potential conflicts of interest.