Isolation and characterization of bone marrow multipotential mesenchymal progenitor cells




There is an increased interest in rheumatology in mesenchymal progenitor/stem cells (MPCs) and their roles in rheumatic diseases, but little is known about the phenotype of these cells in vivo. The aim of this study was to isolate and characterize human bone marrow (BM) MPCs.


Fluorescence microscopy was used to identify putative MPCs among adherent BM cells. To purify them, a positive selection with antifibroblast microbeads was used, combined with fluorescence-activated cell sorting (FACS) for microbead+,CD45low cells. A more detailed phenotype of these cells was determined using 4-color flow cytometry, and standard chondrogenic, osteogenic, and adipogenic assays were used to investigate their differentiation potentials.


Putative MPCs microscopically identified as large, fibroblast-like, D7-FIB+ cells were purified using positive selection with D7-FIB–conjugated (antifibroblast) microbeads followed by FACS for specifically bound microbead+,CD45low cells. These cells represented 0.01% of mononuclear cells in the BM. They were uniformly positive for CD105, LNGFR, HLA–DR, CD10, CD13, CD90, STRO-1, and bone morphogenetic protein receptor type IA (BMPRIA) and were negative for CD14, CD34, CD117, and CD133. Only cells with this phenotype could proliferate and produce adherent cell monolayers capable of chondrogenic, osteogenic, and adipogenic differentiation. D7-FIB− cells in the BM lacked any MPC activity. Uncultured skin fibroblasts had a phenotype similar to that of BM MPCs, but were negative for LNGFR, STRO-1, HLA–DR, and BMPRIA.


This study shows the distinct phenotype, morphology, and method of isolation of BM MPCs. The findings may have implications for defining the physiologic roles of MPCs in arthritis, bone diseases, and joint regeneration.

There is a growing interest in rheumatology in the role of mesenchymal progenitor cells (MPCs) in the pathogenesis of the rheumatic diseases (1–4). These primitive progenitors possess high proliferative potential and can differentiate into several mesenchymal lineages, including bone and cartilage (5–10). It is thought that MPCs may participate in joint repair, and their possible role in regeneration of damaged joints has been suggested (1, 4). Despite recent progress in the field of MPC biology, the in vivo characteristics of these cells remain largely unknown (7–9), and their presence in various tissues, including synovium, is established retrospectively by lengthy expansion/differentiation assays (3). As a consequence, their precise numbers, anatomic distribution within the joint, possible routes of entering the synovium, and relationship to other types of progenitor and stem cells remain poorly defined (3).

The phenotype of MPCs expanded in culture has been previously described (6). Cultured MPCs derived from the bone marrow (BM) (5, 6) and synovium (3), as well as from peripheral (11), cord (12, 13), and fetal (14) blood, all appear as monolayers of fibroblastic cells capable of multilineage mesenchymal differentiation. These cultured MPCs are known to derive from individual clonogenic cells, termed colony-forming units–fibroblastic (CFU-Fs) (6, 11). In the BM, CFU-Fs were documented to have a fibroblastic appearance, characteristic prominent nucleoli (5, 15, 16), and an average frequency of 1 cell per 104–105 mononuclear cells (MNCs) (15, 17). In human peripheral blood, they are even less frequent (11). The extreme rarity of these cells remains the main reason for the current lack of data on their phenotype, and the inability to purify CFU-Fs still leads to ambiguity regarding their precise relationship to hematopoietic and other stem cells.

To date, only a few surface antigenic markers have been described that were used in the purification of MPCs/CFU-Fs from the BM (18–20). The molecules commonly seen as specific for cultured MPCs, such as CD44 and CD29 (13), have in fact broad cell reactivity and, hence, are unsuitable for the detection of MPCs in vivo. STRO-1, the first antibody used to partially enrich CFU-Fs from human BM, also cross-reacts with other cells (erythroblasts) (18). Another potential candidate MPC molecule, bone morphogenetic protein receptor type IA (BMPRIA), was recently used to detect putative mesenchymal precursors in the synovium (2). However, the proliferation and differentiation potentials of BMPRIA+ cells have not been reported, and the BMPRIA status of BM MPCs is still unknown. Therefore, there is a clear need for novel markers and methods of detection, enumeration, and isolation of MPCs from the BM and other tissues as a prerequisite for establishing their roles in joint homeostasis and arthritis.

In the present study, we isolated and characterized MPCs in the BM and established their phenotype and relationship to hematopoietic progenitors. We purified BM MPCs and showed that these cells were adherent, fibroblast-like, and capable of proliferation and osteogenic, adipogenic, and chondrogenic lineage progression. We demonstrated that BM MPCs expressed several unique antigens that distinguished them from dermal fibroblasts, cells lacking multilineage differentiation potentials. These findings should provide a basis for identification of MPCs in the joints and further our understanding of their roles in joint physiology and in diseases such as rheumatoid arthritis, osteoarthritis, and osteoporosis.



BM was obtained from the posterior iliac crest of 25 allogeneic donors (age 5–45 years) following written consent from the local Ethics Committee. BM MNCs were separated using Lymphoprep (Nycomed Pharma, Oslo, Norway) and frozen in liquid nitrogen until required. Frozen BM MNCs from 10 normal donors were purchased from BioWhittaker (Wokingham, UK). Cells were thawed in Dulbecco's modified Eagle's medium (DMEM)/10% fetal calf serum (FCS) (Gibco, Paisley, UK) containing 20 units/ml DNase (Sigma, Dorset, UK) and used for CFU-F assay, flow cytometry, or magnetic isolation as described below.

Normal human skin fibroblasts (American Type Culture Collection [ATCC], Rockville, MD) were cultured in DMEM with 10% FCS. Primary cultures of skin fibroblasts were obtained by culturing explants of human dermis in DMEM with 10% FCS. Uncultured dermal fibroblasts were obtained by digesting small pieces of human dermis with 0.25% collagenase (Stem Cell Technologies, Vancouver, Canada). STRO-1 hybridoma cells were kindly provided by Dr. Paul Simmons (Stem Cell Biology Laboratory, Peter MacCallum Cancer Institute, Melbourne, Victoria, Australia).

Generation of cultured MPCs and CFU-F assay.

Cultured MPCs were generated from the plastic-adherent fraction of BM according to standard methods (6). The growth media consisted of DMEM, 100 units/ml penicillin, 100 μg/ml streptomycin (all from Gibco) and 10% preselected FCS (the standard media). Preselected FCS (Hyclone, Logan, UT) was provided by Drs. B. Deans and A. Mackay (Osiris Therapeutics, Baltimore, MD). CFU-F assay was performed as described previously (21); the cell seeding density was 5 × 104/cm2.

Immunofluorescence microscopy.

A confocal microscope (Leica, Milton Keynes, UK) was used for immunofluorescence analysis of individual CFU-Fs within 24 hours of seeding. For this, 15-mm diameter glass coverslips (BDH, Poole, UK) were coated with 5 μg/cm2 fibronectin to facilitate cell adherence to glass. Fibronectin-coated coverslips were placed in a 12-well plate, and BM MNCs (5 × 106) were added to each well. Adherent fibroblast-like cells were analyzed for immunoreactivity with D7-FIB (a marker of human fibroblasts), CD45 (a pan-leukocyte marker), and STRO-1 (a known CFU-F marker). Purified antibodies D7-FIB (Serotec, Kidlington, UK) and CD45 (clone 9.4; ATCC) at 10 μg/ml and the STRO-1 hybridoma culture supernatant (neat) were used. Secondary antibodies were horse anti-mouse IgG–tetramethylrhodamine isothiocyanate (1:100 dilution; Vector, Burlingame, CA) and goat anti-mouse IgM–fluorescein isothiocyanate (FITC) (1:100 dilution; Sigma). For isotype controls, purified mouse IgG2a and IgM were purchased from Serotec and Coulter (High Wycombe, UK), respectively. Incubations with both primary and secondary antibodies were performed for 1 hour at room temperature. Coverslips were mounted on glass slides (BDH) using Mowiol (CN Biosciences, Nottingham, UK). For phase-contrast microscopy, fibronectin-coated gridded coverslips (BDH) were also used.

Enrichment of putative MPCs by positive selection with D7-FIB–conjugated microbeads.

D7-FIB–conjugated microbeads were purchased from Miltenyi Biotec (Bisley, UK). The buffer consisted of phosphate buffered saline (PBS), 0.5% bovine serum albumin (BSA), and 20 units/ml DNase (all from Sigma). BM MNCs (100 × 106) were used for each separation, and the cell/microbead mixture was incubated for 15 minutes at room temperature. Cells were subsequently spun down, resuspended in 0.5 ml of buffer, and applied to a single MiniMACS column (Miltenyi Biotec). Columns were rinsed with 6 ml of buffer prior to release of the bound fraction. The depletion/enrichment of putative MPCs in both flow-through (unbound) and positive (bound) fractions was assessed by the CFU-F assay as outlined above.

Identification of the putative MPC population in the positive fraction.

Flow cytometry with an XL/MCL (Coulter) was used to compare unseparated, unbound, and bound fractions and to identify a population present only in the positive (bound) fraction. Microbead+ cells were detected with secondary antibody that recognized a D7-FIB–microbead conjugate on the cell surface (goat anti-mouse IgG–phycoerythrin [PE]; Serotec). To exclude brightly fluorescent dead cells, propidium iodide (PI; Sigma) at 2 μg/ml was used. D7-FIB microbead+,CD45low cells were sorted using CD45–FITC antibody (Dako, High Wycombe, UK) and a FACSVantage cell sorter (Becton Dickinson, Oxford, UK). The morphologic purity of the sorted cells was assessed following seeding onto 8-well chamber slides (BDH) and staining with a quick Giemsa stain (Sigma) 4–12 hours after sorting.

Detection of D7-FIB+,CD45low putative MPCs in the MNC fraction and red blood cell–depleted fraction of BM and analysis of their extended phenotype by 4-color flow cytometry.


The following commercially available antibodies were used: CD117–PE, LNGFR–PE, and CD90–PE (Becton Dickinson); CD133/AC133–PE (Miltenyi Biotec); CD105–PE (Serotec); and CD45–FITC and glycophorin A (GPA)–PE (Dako). BMPRIA expression was detected using a biotinylated polyclonal goat antibody (R&D Systems, Abingdon, UK) at 500 ng/ml followed by streptavidin–PE (1:200 dilution; Becton Dickinson). The following antibodies were purified and conjugated in-house: CD10–PE, CD13–PE, CD14–PE, CD34–PE, HLA–DR–PE, D7-FIB–PE, FITC- and PE-labeled isotype controls, and D7-FIB–allophycocyanin (APC). APC-conjugated control antibody was purchased from Becton Dickinson. STRO-1+ cells were detected using concentrated STRO-1 culture supernatant (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City) followed by 10 μl of goat anti-mouse IgM–PE (1:100 dilution; Sigma).

Method sensitivity.

Due to the very low frequency of putative MPCs in the BM (∼0.01% as estimated by the CFU-F assay), the sensitivity of flow cytometry for detecting such a rare population was determined by spiking 106 peripheral blood lymphocytes with 105, 104, 103, and 102 cultured skin fibroblasts. The CD45/D7-FIB antibody combination was used to distinguish between the two cell types, and fibroblasts representing as little as 0.01% of total cells were reliably detected in a single test with not less than 5 × 105 total cells acquired. Based on this result, and in order to collect a sufficient number of events for the phenotyping of putative MPCs, triplicate tests were conducted and the data were pooled.

Cell staining and analysis.

A total of 2 × 106 frozen–thawed MNCs were stained with FITC-, PE-, and APC-conjugated antibodies, and 2 μg/ml of PI was added immediately before collection. Cells were acquired using a FACSort system with CellQuest software version 3.1 (BD Biosciences, San Jose, CA). Putative MPCs were identified using two sequential gates: gate R1, for D7-FIB+,CD45low cells; and gate R2, for PI− cells with high side-scatter (SSChigh,PI− cells) characteristics within gate R1. The PE fluorescence profiles of the cells confined to gate R2 were analyzed and compared with corresponding negative controls. The dot plots for triplicate tests created with the WinMDI program version 2.8 (Scripps Research Institute, La Jolla, CA) were overlaid using Paint Shop Pro version 5 (Jasc Software, Eden Prairie, MN), and the final plots represented 1.5 × 106 pooled events.

It was possible that the manipulation of BM MNCs (including processing, freezing, storage, and thawing) might have affected the MPC phenotype. To address this, BM samples from 10 subjects were aspirated, processed immediately, and analyzed. Red blood cell lysis was performed using 0.86% ammonium chloride (Sigma) for 5 minutes at 37°C, and cells were then stained and processed as outlined above.

Fluorescence-activated cell sorting (FACS) and expansion of putative BM MPCs.

Three-color FACS was performed using D7-FIB microbead–enriched cells, and putative MPCs were sorted using a combination of antibodies, including CD45–FITC, GPA–PE, LNGFR–PE, and CD13–PE. The sorted cells (5,000–15,000, recovered from a single purification experiment) were seeded into 6-well cluster plates in standard medium and allowed to grow to confluence. Cells were subsequently trypsinized and expanded for a further 5 passages until four 150-cm2 flasks of confluent cells were obtained.

Osteogenic, adipogenic, and chondrogenic differentiation.

Osteogenic differentiation.

Osteogenic differentiation of confluent MPC monolayers obtained by the standard method (from plastic-adherent BM cells) or derived from purified putative MPCs (D7-FIB+,CD45low,LNGFR+ cells) was induced by 100 nM dexamethasone, 0.05 mML-ascorbic acid-2-phosphate, and 10 mM β-glycerophosphate (all from Sigma) as previously described (22). Alkaline phosphatase activity and matrix mineralization were detected using the Sigma kit 82 and 1% alizarin red (Sigma), respectively.

Adipogenic differentiation.

Adipogenic differentiation was induced in DMEM/10% FCS supplemented with 0.5 mM isobutylmethylxanthine (Sigma), 60 μM indomethacin (ICN, Basingstoke, UK), and 0.5 mM hydrocortisone (Sigma). Accumulation of lipid vacuoles was visualized with 0.5% oil red, as previously described (6).

Chondrogenic differentiation.

For chondrogenic differentiation, cells (2.5 × 105) were placed in serum-free medium consisting of high-glucose DMEM (Gibco), 100 μg/ml sodium pyruvate, 40 μg/ml proline, 50 μg/ml L-ascorbic acid-2-phosphate, 1 mg/ml BSA, 1× insulin–transferrin–selenium plus, 100 nM dexamethasone (all from Sigma), and 10 ng/ml transforming growth factor β3 (TGFβ3; R&D Systems) (23). Medium was changed every other day. Micromasses were harvested at week 3, and frozen sections (5-μm thick) were prepared.

Sulfated glycosaminoglycan was visualized with 1% toluidine blue (Sigma). Deposition of type II (cartilage-specific) collagen was detected with a mouse anti-human type II collagen monoclonal antibody (Chemicon International, Harrow, UK) at 1 μg/test. Mouse IgG (Sigma) at 1 μg/test was used as negative control. Sections were fixed in 4% formalin, rehydrated in PBS, digested with 40 mU/ml chondroitinase ABC (Sigma), and treated with 10% goat serum (Dako) to block nonspecific binding. Nonspecific peroxidase was blocked with ChemMate peroxidase blocking solution (Dako), and antibody binding was detected using a ChemMate detection kit (Dako). Slides were counterstained with hematoxylin and then mounted.


D7-FIB molecule expressed on BM CFU-Fs/putative BM MPCs.

BM clonogenic fibroblastic cells (CFU-Fs) have been shown to be morphologically distinguishable from adherent hematopoietic cells, including monocytes (15, 16). We examined morphologic types of BM cells adherent after 4 hours of culture and identified distinctive cells that were flat and large (up to 50 μm) and that had characteristic long projections (Figure 1A). In culture, these “fibroblast-like” cells were capable of migration and cell division, giving rise to daughter cells that had the typical spindle morphology of cultured fibroblasts (Figure 1B). The number of original “fibroblast-like” cells prior to division correlated with the number of subsequent fibroblastic colonies (data not shown).

Figure 1.

Morphologic and immunofluorescence analysis of human colony-forming units–fibroblastic (CFU-Fs) and their fibroblastic progeny. A, Phase-contrast photomicrograph of bone marrow (BM) cells adherent to fibronectin-coated glass after 4 hours of incubation in standard medium (grid size 55 μm). Arrow indicates a candidate CFU-F. B, Phase-contrast photomicrographs of BM cells showing the division history (days 1–3) of an individual CFU-F monitored using a gridded coverslip. Arrows indicate a single CFU-F and its progeny. C–E, Membrane fluorescence staining of a CFU-F/putative mesenchymal progenitor cell (MPC). C, Staining with D7-FIB antibody, the human fibroblast marker. D, Staining with STRO-1 antibody, the human CFU-F marker. E, Staining with CD45 antibody, the pan-leukocyte marker. Several hematopoietic cells display bright staining with CD45, compared with weak fluorescence of a single CFU-F/putative MPC. F, Positive fluorescence staining of CFU-F–derived adherent cell monolayers with D7-FIB antibody. (Original magnification × 400 in A; × 100 in B; × 600 in CF.)

By immunofluorescence microscopy, “fibroblast-like” cells were strongly positive for D7-FIB, a marker of human fibroblasts (Figure 1C). Similar levels of expression were observed with the STRO-1 antibody, the known marker of human CFU-Fs (18) (Figure 1D). “Fibroblast-like” cells displayed very low CD45 fluorescence compared with hematopoietic cells which were much smaller, less adherent, and brightly positive (Figure 1E). The latter cells were negative for D7-FIB and STRO-1.

Upon expansion in culture, “fibroblast-like” cell–derived colonies merged, forming monolayers of fibroblasts that continued to express the D7-FIB antigen (Figure 1F). Taken together, these data showed that CFU-Fs/putative MPCs in the human BM could be identified based on the expression of D7-FIB and the lack of expression/low expression of CD45.

Purification of putative MPCs using positive selection with D7-FIB–conjugated microbeads.

In order to purify putative BM MPCs, positive selection with D7-FIB–conjugated microbeads was undertaken. The CFU-F assay was used to assess fibroblast-forming potentials of different fractions. The majority of CFU-Fs (97%; range 93–100%) were recovered in the bound (positive) fraction, while the unbound (flow-through) fraction was depleted of CFU-Fs (n = 20 experiments). Results of a representative experiment are shown in Figure 2, demonstrating monolayers of fibroblasts growing from the bound fraction (Figure 2A) compared with the lack of growth from the unbound fraction (Figure 2B). Cells grown from plastic-adherent cells without separation are shown in Figure 2C. In terms of total cell numbers, the bound fraction represented 0.50 ± 0.15% of the total recovered cells (mean ± SD), and CFU-F enrichment in this fraction was ∼100-fold. A typical colony grown from the column-bound fraction is shown in Figure 2D. These experiments confirmed that CFU-Fs/putative MPCs were D7-FIB+.

Figure 2.

Fibroblast-forming potential of bone marrow fractions separated using D7-FIB microbeads. Photomicrographs were taken on day 10 of culture. A, Fraction bound to column. B, Flow-through/unbound fraction (arrows indicate macrophages). C, Unseparated cells. D, A representative colony grown from the column-bound fraction. (Original magnification × 100.)

Putative MPCs in the positive fraction were next identified using flow cytometry. True D7-FIB microbead+ cells could be detected using a PE-labeled goat anti-mouse antibody (Figure 3). The microbead+ cells represented an average of 5% of live cells in the bound fraction (range 1–10%; n = 15 experiments), were absent in the unbound fraction, and were undetectable prior to magnetic separation with the standard 10,000 events collected (Figure 3A). Taking into account an average MNC yield of 0.5% in the positive fraction and an MNC recovery of 80% after selection, this population had an average frequency of 0.02% prior to separation, comparable with an average CFU-F frequency of 0.01%. D7-FIB microbead+ cells were always SSChigh and always displayed relatively high forward-scatter characteristics (Figure 3A). Consistent with the results of immunofluorescence microscopy, D7-FIB microbead+ cells were CD45low (Figure 3B).

Figure 3.

Detection and initial characterization of D7-FIB microbead+ cells in the column-bound fraction using secondary anti-mouse antibody and flow cytometry. A, Microbead+ cells (gate R2) have special high forward-scatter (FSC) and side-scatter (SSC) characteristics, while microbead− (contaminating) cells (gate R1) consist of a mixture of lymphoid, myeloid, and granulocytic cells. B, Microbead+ cells enriched in the column-bound fraction are CD45low. C and D, Typical morphology of sorted D7-FIB microbead+,CD45low cells observed at 4 hours in culture. Cells have characteristic long projections and prominent nucleoli. C, Phase-contrast photomicrograph showing long projections. D, Giemsa-stained cells, showing prominent nucleoli. FITC = fluorescein isothiocyanate. (Original magnification × 400.)

Sorted microbead+,CD45low cells were adherent and displayed uniform “fibroblast-like” morphology several hours after seeding (Figures 3C and D). All cells had characteristic long processes (Figure 3C) and prominent nucleoli (Figure 3D). After several days in culture, proliferation of these cells and formation of colonies/monolayers of typical spindle-shaped fibroblasts were observed. In a manner similar to microbead− cells in the flow-through fraction, microbead− cells nonspecifically trapped in columns (Figure 3A, region R1) did not have any fibroblast-forming potential. Thus, in these experiments, purification of putative BM MPCs was achieved, and their basic phenotype (D7-FIB+,CD45low) was confirmed by flow cytometry.

Phenotype analysis of BM D7-FIB+,CD45low cells using 4-color flow cytometry.

D7-FIB antibody was conjugated to APC, and the phenotype of D7-FIB+,CD45low cells was analyzed using PE-conjugated antibodies specific for different cell surface markers. Both MNC fraction and the red blood cell–depleted fraction of BM were used in these experiments, and two sequential gates were set up for the analysis. Gate R1 was drawn around the D7-FIB+,CD45low cells, and gate R2 was set to remove dead cells within gate R1 (Figure 4A). A distinct population of D7-FIB+,CD45low cells with an average frequency of 0.01% was found in all experiments performed with the MNC fraction (n = 15 experiments) (Figure 4A). The same population was detected in the red blood cell–depleted fraction (n = 10 experiments), although at a lower frequency (as low as 0.003%), reflecting the effect of dilution with granulocytic cells that was absent in the MNC fraction. There was no population with the same D7-FIB+,CD45low phenotype and a similar frequency in the peripheral blood (data not shown).

Figure 4.

Detection and immunophenotype of D7-FIB+,CD45low cells in the mononuclear cell fraction of bone marrow (BM) by 4-color flow cytometry. A, Gating strategy. APC = allophycocyanin; SSC = side scatter; FITC = fluorescein isothiocyanate; PI = propidium iodide. B, Phenotype of gated D7-FIB+,CD45low,SSChigh,PI− cells. Numbers in the upper right corner of each plot are the phycoerythrin (PE) mean fluorescence intensity. The following cell surface–specific antibodies were used: IgG1 + IgG2a (PE-labeled isotype controls), IgM + 2° (IgM isotype control plus PE-labeled secondary antibody against human IgM), streptavidin–PE (sPE), CD13 (aminopeptidase N), CD90 (Thy-1; marker of hematopoietic progenitors and fibroblasts), CD105 (endoglin/SH2; expressed on cultured mesenchymal progenitor cells [MPCs] and endothelial cells), LNGFR (low-affinity nerve growth factor receptor p75; specific for BM adventitial reticular cells), CD10 (neutral endopeptidase), STRO-1 + 2° (stromal cell precursors plus PE-labeled secondary antibody against human IgM), HLA–DR (major histocompatibility complex class II molecule), CD34 (hematopoietic progenitor/stem cell marker/endothelial marker), CD133 (AC133; marker of hematopoietic and endothelial progenitors), CD117 (c-kit; stem cell factor receptor expressed on early hematopoietic progenitors), BMPRIA + sPE (bone morphogenetic protein receptor type IA; potential MPC marker), and CD14 (monocyte-specific lipopolysaccharide receptor).

D7-FIB+,CD45low cells were positive for CD13, CD90 (Thy-1), CD105 (endoglin/SH2), LNGFR, CD10, STRO-1, and HLA–DR (Figure 4B). They uniformly lacked the hematopoietic progenitor markers CD34 and CD133, and the majority of cells were negative for CD117 (c-kit) (Figure 4B). D7-FIB+,CD45low cells were clearly negative for CD14 and lacked other lineage markers, including GPA, CD4, CD7, CD16, and CD21 (data not shown). D7-FIB+,CD45low cells were positive for BMPRIA, although its expression was quite low (Figure 4B).

Optimized purification of putative BM MPCs.

A secondary goat anti-mouse antibody used to detect D7-FIB microbead+,CD45low cells in the enriched fraction could have produced some nonspecific binding that was avoided when directly labeled antibodies were used. In these experiments, we used LNGFR–PE, CD13–PE (brightly positive on microbead+,CD45low cells), and GPA–PE (negative on microbead+,CD45low cells) and analyzed their expression profiles before and after enrichment with D7-FIB microbeads (Figure 5A). Distinct populations of CD45low,LNGFR+, CD45low,CD13+, or CD45low,GPA− cells were clearly detected after enrichment (indicated by arrows in the “Bound” plots of Figure 5A). Sorted cells with these phenotypes had “fibroblast-like” morphology and produced growing monolayers of fibroblasts in all experiments. CD45−,GPA+ erythroblasts and CD45+ lymphoid and myeloid cells nonspecifically trapped in columns consistently failed to produce growing adherent cell monolayers (Figure 5B). A combination of CD45–FITC and LNGFR–PE provided the best resolution with the least cross-reaction (Figures 5A and B) and was selected for all subsequent experiments. Thus, in these experiments, we confirmed some of the phenotype data obtained with 4-color flow cytometry and optimized the FACS step in MPC purification.

Figure 5.

Optimized sorting and expansion of putative BM MPCs. A, Detection of D7-FIB microbead+,CD45low cells enriched in the column-bound fraction using direct staining with CD13, glycophorin A (GPA), and LNGFR. Arrows indicate putative MPC population. B, Only D7-FIB+,CD45low,LNGFR+ cells are capable of forming growing cultures of adherent cells. C, Phenotype of cells expanded from sorted D7-FIB+,CD45low,LNGFR+ cells after 4 passages in culture. See Figure 4 for other definitions.

Multilineage differentiation upon expansion of sorted D7-FIB+,CD45low,LNGFR+ cells.

D7-FIB microbead+,CD45low,LNGFR+ cells were purified and allowed to grow in the standard medium, and the phenotype of resulting cell monolayers was investigated (Figure 5C). Continuous expression of CD13 and D7-FIB (Figure 5C) and of CD10 and CD105 (data not shown) confirmed the fibroblastic nature of the progeny of purified D7-FIB+,CD45low,LNGFR+ cells. The disappearance of STRO-1, LNGFR, and HLA–DR (Figure 5C) suggested that the expression of the latter molecules was induced in vivo by the BM microenvironment and that the necessary stimuli were absent in the standard culture medium.

Confluent monolayers of D7-FIB+,CD45low,LNGFR+ cell–derived fibroblasts were next subjected to conditions promoting osteogenesis, adipogenesis, and chondrogenesis. Cultured MPCs obtained from the same donors by the standard method (from BM plastic-adherent cells) were used as a positive control of differentiation (Figure 6). Commercially available cultured skin fibroblasts, as well as primary cultures of human skin fibroblasts derived from fibroblastic outgrowth of dermal explants, were used as a negative control of differentiation (Figure 6).

Figure 6.

Differentiation of cell monolayers derived from sorted D7-FIB+,CD45low,LNGFR+ cells into osteogenic, adipogenic, and chondrogenic lineages. Left panels show cultured BM MPCs (positive control). Middle panels show D7-FIB+,CD45low,LNGFR+-derived cells. Right panels show cultured skin fibroblasts (negative control). A and B, Osteogenic differentiation of cultures grown in osteoinductive conditions. A, Alkaline phosphatase staining of 2-week-old cultures. B, Calcium deposition in 3-week-old cultures (alizarin red stain). C, Adipogenic differentiation of cultures grown in adipoinductive conditions. Shown is oil red staining of lipid-laden cells in 3-week-old cultures. D and E, Chondrogenic differentiation of micromass cultures grown in serum-free medium supplemented with transforming growth factor β3 and analyzed on day 21 of culture. D, Toluidine blue staining of extracellular anionic proteoglycans. E, Extracellular matrix deposition of type II collagen. See Figure 4 for definitions. (Original magnification × 100.)

D7-FIB+,CD45low,LNGFR+-derived and positive control cells treated with osteogenic medium produced alkaline phosphatase and calcium deposits, which were not seen in similarly treated cultures of skin fibroblasts (Figures 6A and B). Accumulation of lipid vacuoles upon adipogenic induction was observed in control MPCs and in D7-FIB+,CD45low,LNGFR+-derived cells, and not in cultures of fibroblasts (Figure 6C). TGFβ3-treated micromasses obtained from both control cultured MPCs and monolayers of cells derived from D7-FIB+,CD45low,LNGFR+ cells displayed cartilage-specific metachromasia with toluidine blue staining, which was not seen in similarly cultured skin fibroblasts (Figure 6D). The extracellular deposition of type II collagen confirmed that chondrogenic differentiation was taking place in the positive control and D7-FIB+,CD45low,LNGFR+-derived micromasses and not in the fibroblast-derived micromasses (Figure 6E). Thus, these data demonstrated that BM D7-FIB+,CD45low,LNGFR+ cells possessed mesenchymal multilineage differentiation potentials upon expansion; that is, they were true BM MPCs. The purification/cell sorting procedures did not affect their ability to differentiate into chondrocytes, osteoblasts, and adipocytes.

BM MPC population phenotypically different from uncultured skin fibroblasts.

Four-color flow cytometry revealed that the phenotype of BM MPCs was largely fibroblastic based on the expression of D7-FIB, CD13, CD10, CD105, and CD90, the molecules present on human fibroblasts grown in culture. However, BM MPCs were uniformly positive for HLA–DR, LNGFR, and STRO-1, the molecules absent on cultured MPCs (Figure 5C) and cultured fibroblasts (data not shown). It was not clear whether the expression of these molecules was specific for BM MPCs or whether it was common for all uncultured fibroblastic cells and simply down-regulated during culture. To investigate this, we analyzed the expression of several MPC antigens on uncultured skin fibroblasts obtained from collagenase-digested pieces of dermis (Figure 7).

Figure 7.

Phenotype of uncultured skin fibroblasts present in collagenase-digested tissue specimens. A, Gating strategy. B, Phenotype analysis of cells confined to gate R1 + gate R2. Solid arrows indicate a population of fibroblasts. Compared with bone marrow mesenchymal progenitor cells, skin fibroblasts lack STRO-1, BMPRIA, HLA–DR, and LNGFR. Broken arrows indicate forward-scatter (FSC)/side-scatter (SSC) profiles of gated fibroblasts and debris. See Figure 4 for other definitions.

The gating strategy involved the exclusion of PI+ dead cells (gate R1) and subsequent removal of CD45+ (hematopoietic) cells (gate R2) (Figure 7A). Within gate R2, two subpopulations of cells were found: large CD45low cells that consistently expressed D7-FIB, CD13, and CD105 (fibroblasts), and exceptionally small CD45− cells that were negative for all of the markers studied (debris). CD45low,CD13+,CD105+,D7-FIB+ resident skin fibroblasts did not display the bright positivity for LNGFR, HLA–DR, or STRO-1 characteristic of BM MPCs, and BMPRIA expression was also absent (Figure 7B). These data demonstrated that the expression of LNGFR, HLA–DR, STRO-1, and to a lesser extent, BMPRIA distinguished BM MPCs from skin fibroblasts (cells of a similar nature but incapable of multilineage differentiation). Incubation of BM MNCs or red blood cell–depleted BM cells with trypsin or collagenase did not affect the expression of LNGFR, HLA–DR, and STRO-1 by BM D7-FIB+,CD45low cells (data not shown). This indicated that the epitopes recognized by the latter antibodies were not trypsin- or collagenase-sensitive.


MPCs may be of major pathogenic and therapeutic importance in the rheumatic diseases, but to date, they have only been indirectly characterized following expansion in culture (1–4, 6). The phenotype and topography of MPCs in vivo and their relationship to other progenitor and stem cells are poorly defined (3, 7–10). The purpose of this study was to isolate and characterize MPCs resident in human BM. We showed that all of the BM MPC activity was confined to a rare, phenotypically distinct, and homogeneous population of D7-FIB+,CD45low cells that are different from both skin fibroblasts and hematopoietic progenitors. Representing only 0.01% of BM MNCs, D7-FIB+,CD45low cells were purified here using a combination of magnetic selection and FACS, resulting in levels of enrichment that had not been reported before.

Using multiparameter flow cytometry, we obtained a detailed phenotype profile of BM MPCs. Published data on the MPC phenotype in vivo are very limited (18–20, 24). Our findings confirmed that BM MPCs expressed STRO-1 and CD105 antigens (18–20). Both markers, however, are known to be present on other types of BM cells and therefore are not very selective for MPCs. STRO-1 is expressed on erythroblasts (17–19), and CD105/SH2/endoglin (25, 26) is expressed on endothelial cells (27) and on pre–B cells (28). In comparison, the D7-FIB molecule showed minimal coexpression on other BM cells. It therefore appears superior to STRO-1 or CD105 as a marker of MPCs in the BM.

Apart from the D7-FIB antigen, which is a fibroblast-specific molecule of yet-unknown function, BM MPCs were found to express other antigens present on human fibroblasts, the peptidases CD10 and CD13, and CD90 (Thy-1) (29, 30). The molecules strongly expressed on BM MPCs and absent on uncultured skin fibroblasts were LNGFR, STRO-1, and HLA–DR. LNGFR expression was previously reported on adventitial reticular cells (ARCs), an interconnected network of nonhematopoietic star-like cells in the BM (31, 32). This indicated that ARCs were the likely in vivo equivalents of MPCs in the BM. The function of LNGFR on MPCs is unclear, but it may have a morphogenic role in the development of the BM cavity, kidney, and other organs (31–33).

The expression of HLA–DR on BM MPCs was unexpected. It is possible that HLA–DR may play a role in hematopoietic progenitor cell maturation, since HLA–DR was previously shown on stromal cells in the developing thymus (34). Interestingly, thymic stroma was also shown to express TGFβ receptors (35), a large superfamily of molecules that includes BMPRs (36). BMPs are known to regulate the growth, differentiation, and apoptosis of various cell types, including osteoblasts, chondroblasts, and neural and epithelial cells (36). In this study, one member of this family, BMPRIA, was found to be expressed on BM MPCs and not on skin fibroblasts. This strengthens the idea that BMPRIA+ cells described in inflamed synovial tissue are related to MPCs (2). The expression of other members of the BMPR family on BM MPCs is currently under investigation.

To date, the relationship between MPCs and hematopoietic progenitor cells has not been fully established. Our study clearly demonstrated that BM MPCs did not express CD34 or CD133, the most commonly used markers of hematopoietic progenitors. Morphologically, fibroblast-like MPCs also differed markedly from hematopoietic progenitors, which are known to be small, blast-like cells. This indicated that hematopoietic and mesenchymal progenitors in the BM are morphologically and phenotypically different cell types.

There is a great deal of debate in the literature as to what constitutes a stem or progenitor cell as well as the degree of plasticity of various progenitor cells. Stem cells are capable of self-renewal and can differentiate into all cell types (pluripotential) or to more than one differentiated cell type (multipotential) (37). Progenitor cells are the intermediate step between stem cells and fully differentiated cells (38). Early progenitors can still be multipotential cells, but unlike stem cells, they are not capable of self-renewal (38). Because MPCs have been described as multipotential cells, many investigators have adopted the term mesenchymal stem cells, in one view somewhat prematurely, since no suitable assay to assess their self-renewal has yet been developed (39).

A broad range of alternative terms (such as “marrow stromal stem cells” or “osteogenic stem cells”) historically used to define cells with MPC properties tends to add even more confusion to the contentious nature of this field (6–10, 40). In addition, recent findings suggest that BM may contain putative stem cells more primitive than MPCs (40) as well as pluripotent cells capable of producing progeny with characteristics of mesoderm, neuroectoderm, and endoderm (41). Such cells can be copurified with MPCs from the CD45−,GPA− cell fraction of human BM, but their more detailed phenotype has not yet been reported. It may be possible in the future to isolate these primitive cells using the D7-FIB–based strategy described in this report. With regard to the MPCs themselves, further investigation is required to determine whether D7-FIB+,CD45low cells are a self-renewing cell population or whether they are derived from more primitive progenitors with a yet-unknown phenotype.

It has recently been shown that MPCs are present in normal synovium, and their role in the development of arthritis has been postulated (1–4). However, in the absence of a “phenotypic fingerprint,” the nature and roles of MPCs in bone and joint homeostasis have remained speculative. In this study, we identified and described the phenotype of these exceptionally rare cells in human BM. On the basis of this research, precise quantification of BM MPCs in such diseases as osteoporosis and osteoarthritis could be performed. The phenotype data from the present study could allow identification of MPCs in the joint, based on the D7-FIB+,CD45low,LNGFR+ phenotype. This will lead to further progress in the understanding of the distribution and plasticity of MPCs in vivo and the roles they may play in the joint physiology in health and disease.


We thank the staff of The Geoffrey Giles Theatre of St. James's University Hospital for providing bone marrow. We thank Drs. Bob Deans, Alastair Mackay, and Don Simonetti from Osiris Therapeutics for advice and kind provision of preselected serum, antibodies, and reagents, Rachel Grayson for providing dermis specimens, and Dr. Erika de Wynter for reviewing the manuscript.