Direct selection of human bone marrow mesenchymal stem cells using an anti-CD49a antibody reveals their CD45med,low phenotype


Dr F. Deschaseaux, Institut d'Etude et de Transfert de Gènes, 240 Route de Dole, 25000 Besançon, France. E-mail:


Summary. Human bone marrow mesenchymal stem cells (MSC) generate, via a fibroblast colony-forming unit (CFU-F), osteo-chondroblastic cells as well as adipocytes and stromacytes. To date, these stem cells are isolated indirectly using a cell culture method and phenotyped as CD45 negative while the in vivo counterparts are undetermined. Our aim was to develop a direct selection method and to determine the phenotype of the MSC isolated in this way. Mesenchymal cells were selected with anti-CD49a and/or anti-CD45 antibodies using either flow cytometry or a magnetic beads method. All CFU-F were always detected in the small population of CD49a-positive cells. These CFU retained their differentiation potential and gave rise to osteo-chondroblastic cells, adipocytes and stromacytes. Phenotypic studies on uncultured cells revealed a CD45med,low, CD34low, HLA-II cell population. Flow cytometry cell sorting showed that MSC with CFU-F potential were obtained only from a CD49a+/CD45med,low population. In addition, when cultured, they clearly became CD45, CD34, HLA-II, CD49a+. These results confirmed that MSC can be directly selected easily from human bone marrow using magnetic beads without altering their differentiation potential. These cells expressed mildly the haematopoietic marker CD45, which was dramatically downregulated by in vitro culture. The expression of CD45 coupled to CD49a thus enabled direct selection of the MSC.

In humans, the bone marrow (BM) is the site where blood- and bone-forming cells are generated for the development and maintenance of these tissues. A heterogeneous population of cells from stem cells to differentiated cells, e.g. stromal cells (stromacytes), is involved in these mechanisms. Therefore, BM contains two kinds of stem cells: (1) the mesenchymal stem cells (MSC), the source of bone- and haematopoietic microenvironment-forming cells, and (2) the haematopoietic stem cells (HSC), which give rise to the blood-forming cells (Potten, 1997; Weissman, 2000). MSC are normally isolated indirectly from BM aspirates utilizing their strong capacity to adhere to plastic culture flasks. In vitro, these cells are readily characterized, among the adherent haematopoietic cells, by their ability to generate single cell-derived colonies or fibroblast colony-forming units (CFU-F) (Castro-Malaspina et al, 1980; Bruder et al, 1997; Prockop, 1997; Pittenger et al, 1999). In turn, cells from the CFU can be expanded and differentiated in vitro into osteoblastic, chondroblastic, adipocytic and stromacytic cells using defined media (Bruder et al, 1997; Prockop, 1997; Pittenger et al, 1999). The latter cells are required for the long-term support of the growth and differentiation of HSC in vitro as well as in vivo (Galmiche et al, 1993; Almeida-Porada et al, 2000). There is thus a close relationship and interplay between the two stem cell types. Owing to their multipotentiality, MSC are currently used for developing cell and gene therapy protocols for the treatment of a number of human diseases (Bruder et al, 1994; Baxter et al, 2002). Recently, Horwitz et al (2002) showed an engraftment of cultured MSC in bone of children with osteogenesis imperfecta and an increase in growth velocity during the first 6 months post infusion.

Using indirect selection methods, the phenotype and the reconstitution potential of expanded or cultured MSC have been well characterized (Conget & Minguell, 1999; Pittenger et al, 1999; Deans & Moseley, 2000). Cultured MSC are known to be both CD45 and CD34 negative (CD45, CD34) whereas HSC are CD45 and CD34 positive (CD45+, CD34+) (Berardi et al, 1995; Prockop, 1997; Conget & Minguell, 1999; Pittenger et al, 1999; Ziegler et al, 1999; Okuno et al, 2002; Verfaillie, 2002). Unfortunately, there are very few data on their in vivo counterparts. Simmons and Torok-Storb (1991a) directly selected a population of cells containing some stromal precursors accompanied by a majority of erythroblastic cells and B-lymphocytes by using the Stro-1 antibody and flow cytometry.

Our previous studies have shown that stromal precursors can differentiate into myofibroblastic stromacytes that are phenotypically similar to vascular smooth muscle (VSM) cells of the aortic intima (Galmiche et al, 1993). We demonstrated that primate, human and non-human BM myofibroblasts from cell lines or from primary cultures have a strong CD49a protein expression. In addition, we isolated the stromacyte precursors in CD49a+ BM fractions (Deschaseaux & Charbord, 2000; Deschaseaux et al, 2001). This CD49a direct selection has provided an insight on the earliest molecular events that occur during stromacytic differentiation. Indeed, at the beginning of the culture, all CD49a+-derived stromal cells were recognized by the Stro-1 antibody. After 3 weeks of culture, Stro-1 staining was negative while the expression of CD49a remained stable. The CD49a molecule is the α1-integrin subunit from very-late antigen 1 (VLA-1) integrin. This adhesion molecule, specific to the VSM lineage, acts as the receptor for both collagen (CO) and laminin (Belkin et al, 1990; Glukhova & Koteliansky, 1995; Gotwals et al, 1996; Pozzi et al, 1998; Wang et al, 1998). Other molecules are also specific to VSM, such as the cytoskeletal proteins αSM-actin, h-caldesmon, SM22α, calponin and SM myosin. Remarkably, the genes encoding these proteins are all dependant on promoters that contain at least one CarG box, the binding site for serum response factor (Obata et al, 1997). Therefore, the CD49a molecule has a crucial role in stromacytic or myofibroblastic differentiation from the mesenchymal compartment. We, therefore, investigated whether it was possible to develop a method that directly selects for MSC using the CD49a molecule.

In this study, we demonstrated that all human multipotential BM MSC, when directly selected, were contained within the CD49a+ cell population. The haematopoietic marker CD45 was shown to be mildly expressed by unprocessed or fresh MSC while cultured MSC were clearly CD45. Moreover, before the induction of differentiation some lineage-specific markers were detected, suggesting a heterogeneity in the differentiation state among the CFU-F that were derived from this population. In conclusion, all CFU-F generated by immature and more mature mesenchymal progenitors can be directly selected from human BM when using antibodies against the CD49a molecule. The significance of CD45 expression by MSC is functionally unclear but proves their BM origin. We also assessed the capacity of these cells to be cryopreserved for stem cell banking. This direct selection will be of benefit for studies on MSC lineages as well as for subsequent clinical use.

Materials and methods

Cell isolations and cultures.  After obtaining informed consent, BM samples were collected from patients undergoing thoracic surgery. Marrow cells from aspirates were centrifuged on Hypaque–Ficoll density gradients, and the interface mononuclear cells were collected. After two washes with phosphate-buffered saline (PBS) containing 0·1% (w/v) human serum albumin (HSA) (LFFB; Courtaboeuf, France), cells were incubated for 30 min at 4°C under gentle agitation with the monoclonal antibodies anti-CD49a (clone TS2/7 from T-Cell Science, Cambridge, MA, USA) (Kern et al, 1994) or clone SR84 (Pharmingen, San Diego, CA, USA) or control antibody of the same isotype at the appropriate concentration (0·2 µg/107 cells). After two washes, cells were incubated for 30 min with beads (three beads per cell) coated with human anti-mouse immunoglobulin (Ig)G and a spacer of DNA fragment (CellectionTM, Pan Mouse IgG Kit; Dynal, Oslo, Norway), according to the manufacturer's instructions. Uncoated cells were removed after four washes with PBS/HSA. Coated cells were incubated for 15 min at 37°C in McCoy's 5A medium (Gibco BRL, Gaitherburg, MA, USA) supplemented with 1% (v/v) FCS (fetal calf serum) (StemCell Technology, Vancouver, Canada) containing DNAse (Dynal). The beads were then removed from rosetted cells by flushing the suspension several times through a pipette and by placing the tube in the magnetic particle concentrator (MPC-1; Dynal). The capacity of the MSC to be banked was assessed and some fractions of CD49a+ cells were cryopreserved. Unfractionated (initial fraction), fractionated and cryopreserved cells were cultured in expansion medium, consisting of Dulbecco's modified Eagles' medium (DMEM; Biomedia, Boussens, France) with 10% (v/v) FCS, at a concentration of 1 × 104− 1 × 105 cells/cm2. Immunomagnetic selections of CD45+ cells were performed using the anti-CD45 antibodies (clone PD7/26; Dakopatts, Glostrup, Denmark; or clone B-A11; Diaclone, Besançon, France) at the concentration described above. For some experiments, CD49a+ cells were isolated from the CD45-rejected cell population. Some CD34+ cells were also selected from BM mononuclear cells and from CD49a+ cells. The CD34+ cells were isolated using a magnetic separation column according to the manufacturer's instructions (mini-Macs; Miltenyi Biotec, Bergisch Gladbach, Germany). The average final purity of the CD34+ fraction was 98% ± 1 [mean ± standard error of the mean (SEM), 20 experiments]. Cultures were maintained in a humidified atmosphere with 5% CO2. After 15 d, the number of CFU-F that comprised more than 10 elongated cells was determined from each fraction. When culture dishes became near-confluent, cells were detached with 0·25% trypsin, containing 1 mmol/l EDTA (Gibco BRL) for 5 min at 37°C, counted and subsequently replated at 5 × 103 cells/cm2 in 96-wells plate for the differentiation step.

In vitro differentiation of cells from CFU-F.  The potential of BM MSC to differentiate, via the CFU-F, into osteogenic, chondrogenic, adipogenic and stromacytic lineages was verified after the first passage. Each well or dish was kept in expansion medium until confluence was reached. Cultures were stimulated with the appropriate differentiating medium, as previously described (Jaiswal et al, 1997; Banfi et al, 2000). Briefly, osteogenic and adipogenic media comprised of expansion media supplemented with 100 × 10−9 mol/l dexamethasone, 0·05 × 10−3 mol/l ascorbic acid and 10 × 10−3 mol/l β-glycerophosphate for osteoinduction. Dexamethasone (1 × 10−6 mol/l) and 100 µg isobutylmethylxanthin (IBMX; Sigma, St Louis, MO) were added for adipogenic culture conditions. Chondrogenic culture conditions were set up using F-12 standard medium (BioMedia) with 10% (v/v) FCS, 0·05 × 10−3 mol/l ascorbic acid, 1 ng/ml human recombinant transforming growth factor β1 (TGF-β1) (RD, Mineapolis, MN, USA). Stromacytic culture medium was established as described by Gartner and Kaplan (1980) with McCoy's 5A medium (Biomedia), 12·5% (v/v) FCS, 12·5% (v/v) horse serum (Sigma), 10−6 mol/l hydrocortisone and 5 × 10−5 mol/l β-mercaptoethanol. As control, cultures were maintained in expansion medium. After 2 weeks of culture, cells were examined for the expression of lineage-specific molecules.

Cell characterization. Uncultured cells from the initial, CD49a+, CD49a fractions before or after cryopreservation were examined using flow cytometry. The phenotypes of cells cultured in expansion medium and in differentiation conditions were analysed using in situ immunofluorescence (IF), enzyme-linked immunosorbent assay (ELISA) and flow cytometry. The following antibodies were used: mouse monoclonal antibodies (mAb) anti-human CD3 conjugated to fluorescein isothiocyanate (FITC) (clone B-B11), CD14 FITC (B-A8), human leucocyte antigen (HLA) class II conjugated to phycoerythrin (PE) (B-F1), HLA class I FITC (W6/32-HL), CD31 FITC (B-B38), glycophorin A FITC, CD54 FITC (B-H17), CD50 FITC (B-R1), CD10 FITC (B-E3), CD117 PE (BK15) and anti-CD11b PE (44) (Diaclone); mAb anti-human CD38 FITC (clone T16), CD49d FITC (HP2/1), CD49e FITC (SAM-1), CD36 FITC (FA6-32), CD45 (J33) conjugated to phycocyanin-5 (PC-5) (Coulter-Immunotech, Marseille, France); mAb anti-CD73 PE (clone ad2), CD49a PE (SR84), CD90 FITC (5E10) and anti-CD34 PE (HPCA-2) (Becton Dickinson, Mountain View, CA, USA; Pharmingen). Antibodies used for indirect IF antiα-SM actin (1A4), smooth muscle myosin heavy chain (hSM-V), fibronectin (FN-15), vimentin (Vim 13·2) were purchased from Sigma; antibodies anti-human collagen I (CO I; polyclonal), CO II (polyclonal), osteocalcin (1–22) came from Chemicon (Temecula, CA, USA).

Flow cytometric studies were performed on fresh or cryopreserved cells and cells from the CFU-F (resuspended by treatment with collagenase 250 U/ml; Sigma). Cells were washed in PBS/HSA, and primary antibodies (see above) at the appropriate concentration were added. Antibodies of the same isotype were used as negative controls (Diaclone, Immunotech, Becton Dickinson and Chemicon). The cells were passed through the Becton Dickinson FACSort, equipped with the cellquest software program. The fluorescence histogram for each mAb was displayed along with that of the corresponding control antibody. Cytoplasmic antigens were analysed on cells that had been fixed and permeabilized using a 3·7% formaldehyde/0·1% Triton X100 solution (Sigma).

Immunofluorescence in situ studies were performed as previously described (Deschaseaux & Charbord, 2000). Briefly, confluent adherent layers of cells were grown on eight-well chamber slides and then fixed for 30 min at 4°C. Different fixations were used: 3·7% (w/v) formaldehyde with 0·5% (w/v) glutaraldehyde for extracellular matrix (ECM) antigens, and an absolute methanol for cytoskeleton antigens. Primary antibodies, described above, were added and stained by either goat anti-mouse polyvalent Ig conjugated to FITC (Sigma) for mAb or goat anti-rabbit IgG conjugated to tetra-methylrhodamine-B-isothiocyanate (TRITC) (Sigma) for polyclonal antibodies. Lipid droplets within adipocytic cells were stained with Oil Red-O solution (Sigma). Slides were examined using a microscope equipped for fluorescence (Leitz Aristoplan; Leica, Weltzar, Germany).

The expression of some ECM proteins was analysed using an ELISA. Cells cultured in 96-well plates, in differentiation media (osteogenic and chondrogenic) and in control medium, were studied for the extracellular expression of osteocalcin, CO I and CO II. The cells were washed three times and fixed for 45 min using 3·7% formaldehyde. After three washes, unsaturated sites were blocked for 2 h at 37°C under gentle agitation using a PBS/Tween 0·05% (w/v), 3% HSA and 1 mmol/l CaCl2 solution. Primary antibodies were added for 1 h at 37°C. The cells were washed four times and incubated with the goat anti-rabbit antibody conjugated to horseradish peroxidase (Bio-Rad, Hercules, CA, USA). After three washes, peroxidase activity was detected using a tetramethylbenzidine (TMB) peroxidase enzyme immunoassay (EIA) substrate kit (Bio-Rad). Expression was determined by reading the plates at 450 nm on a microplate reader (Victor; Wallac, Perkin Elmer, Turku, Finland), and values obtained from cells cultured in differentiation media were compared with those of cells cultured in control medium.

Alkaline phosphatase enzyme activity of cells cultured in differentiation media in 96-well plates was measured, after rinsing cells twice with Hank's balanced salt solution (Gibco BRL), using the alkaline phosphatase diagnostic Kit (Sigma), according to the manufacturer's instructions.

Cobblestone area forming cell (CAFC) assay. In order to study the haematopoiesis-sustaining function of stromacytes, cobblestone area forming cell (CAFC) assays were performed, as described previously (Breems et al, 1998). Cells from CFU-F were seeded into 96-well plates in complete long-term medium, described above, and cultured until confluence (feeder layer). Serial half-dilutions of BM CD34+ cells in 100 µl medium (15 parallel wells) were seeded over feeder cells. Cultures were maintained at 37°C with weekly half medium changes. Positive wells were defined by the presence of at least five phase-dark cells formed under the stromal layer (CAFC). Negative wells were counted under an inverted phase–contrast microscope (Zeiss, Oberkochen, Germany). The number of CAFC was calculated for 6 weeks of co-culture using a Poisson statistic. We assumed, on average, that one CAFC had been seeded per well when the percentage of negative wells was 37%.

Flow cytometric cell sorting.  Bone marrow mononuclear cells were stained using the mAb anti-human CD49a PE (Pharmingen) and anti-human CD45 PC5 (Immunotech). After three washes, cells were passed through a Beckmann Coulter cell sorter Epics Altra (with expo32 software) using PBS in the sheath fluid. To sort, two types of gates were used: the first was assigned to SA cells, and the second was determined by the CD45 versus CD49a fluorescence. Sorted cells, whose purity ranged from 95% to 99%, were cultured into expansion medium.

Statistical analysis.  Values are given as mean ± SEM. Comparisons between means were made using Student's t-test or a non-parametric Wilcoxon's W-test. The difference was statistically significant when P < 0·05.


Isolation of bone marrow CD49a+ cells and CFU-F potential study

CD49a+ cells were selected from BM mononuclear cells by magnetic beads. Cells were not recovered using the control antibody. The percentage CD49a+ fraction retained was 3·6% ± 0·4 (n = 34). No significant difference was found in the percentage of CD49a+ mononuclear cells detected by flow cytometry (2·7% ± 0·5, P = 0·7). As noted in previous reports (Castro-Malaspina et al, 1980; Pittenger et al, 1999), human BM MSC are characterized by their ability to form colonies comprising non-refringent spindle-shaped cells deriving from single cells (CFU-F) when plated in culture flasks. Before selection, half of the BM sample was used to measure the CFU-F potential. After 10–14 d of culture, all CFU-F were recovered in the CD49a+ fraction (94% total CFU-F potential of BM sample). Accordingly, no colonies were observed in the CD49a fraction. The cloning efficiency (CE: number of CFU-F/105 cells seeded) was 30·5 ± 3 (one CFU-F/3278 cells seeded) for the CD49a+ fraction versus 1·4 ± 0·3 for the unseparated cell fraction. The enrichment factor was therefore 20. Moreover, the total recovery for this type of separation was 84% ± 3. No differences were observed when using the TS2/7 or SR84 mAb.

In order to determine whether CD49a+ cells could be banked, the cells were assessed for their CFU-F potential after cryopreservation. The CE for cryopreserved CD49a+ cells increased (101 ± 28 or one CFU-F/990 cells seeded). Therefore, the CFU-F potential was conserved after cell banking.

CFU-F from mesenchymal progenitors have been also generated by an adhesion protocol (Pittenger et al, 1999). We applied this method to the CD49a+ cells. After 24 h of culture, non-adherent cells were removed. We were unable to detect CFU-F in the non-adherent fraction. Thus, MSC with CFU-F potential were contained exclusively in CD49a+ fraction, and possessed a capacity to adhere strongly to culture dishes and could be cryopreserved.

Phenotypic studies of fresh BM CD49a+ cells

Cell culture and cell adhesion are both known to alter antigen expression. The phenotype of fresh MSC when selected by the indirect selection method is largely unknown. As all MSC were included in the BM CD49a+ population, it should be possible to study the phenotype of freshly isolated MSC before and after cryopreservation.

The CD49a-positive cells were studied using flow cytometry. In the freshly isolated fraction, three types of cell population were detected (Fig 1A). One comprised large [high forward scatter (FSC)] and granular [high sideways scatter (SSC)] CD66b+/CD45+ cells corresponding to the phenotype of granulocytic cells. The second fraction comprised large and moderate SSC CD36+/CD45+ and CD14+/CD45+ monocytic cells. The third population comprised small agranular (SA) cells. SA cells represented 48% ± 1·5 of the fresh CD49a+ cell fraction, while granulocytic cells represented 38% ± 1 and monocytic cells represented 14% ± 1. Only SA cells were detected in the cryopreserved fraction. Indeed, 50% of the total cells from the CD49a+ fraction were lost with a bias against granulocytic and monocytic cells (Fig 1B), as shown by the lack of detection of CD66b, CD36 and CD14 (Table I). These results explained the increased CE that was observed for cryopreserved cells. Flow cytometric studies on the SA fraction have shown that fresh cells comprised T (CD3+) and B (CD19+) lymphocytes (9% ± 9 and 1% ± 0·5 respectively), and erythroblastic glycophorin A+ cells (Table I). T and B lymphocytes were enriched in cryopreserved cells (19 ± 5 and 11 ± 2 respectively) while the percentage of erythroblasts was stable (23% ± 9 and 30%). HLA-II was exclusively detected on CD19+ cells. Moreover, stromal markers CD90 and CD166 were also expressed. A few cells (7% ± 1) expressed the CD73 molecule which contains the epitopes recognized by SH3 and SH4 antibodies. CD34 and the c-kit (stem cell factor receptor) molecules were not detected on the fresh SA fraction but cryopreserved cells were weakly positive (4% ± 0·5 for CD34 and 8% ± 4 for c-kit).

Figure 1.

Flow cytometric histograms of uncultured BM CD49a+ cells. Haematopoietic markers were examined in freshly separated (A) and cryopreserved (B) CD49a+ cells. Fifty per cent of granulo-monocytic cells were detected in the freshly selected cell population while the remaining cells were small and agranular (SA cells). After cryopreservation, only SA cells were found (B).

Table I.  Flow cytometric studies of SA cells.
AntigensPercentage (mean ± SEM) of positive cells from
Fresh fractionCryopreserved fraction
  1. NT, not tested.

CD39 ± 919 ± 5
CD191 ± 0·511 ± 2
CD14, CD66b,CD11b00
Glycophorin A23 ± 930
CD902035 ± 9
HLA-IINT17 ± 2
c-kit08 ± 4
CD3404 ± 0·5
CD385 ± 314 ± 4
CD4588 ± 270 ± 3
CXCR-462 ± 366 ± 4
CD49d73 ± 482 ± 3
CD49e17 ± 219 ± 5

Some authors have described a MSC generated by a small population of BM CD34+ cells (Simmons & Torok-Storb, 1991b). Therefore, we investigated whether all the CFU-F were generated by CD49a+/CD34+ cells. We purified human BM CD34+ from mononuclear and from CD49a+ cells then quantified their CFU-F potential. From BM mononuclear cells, the cloning efficiency (CE = 13 ± 0·2) was significantly lower than that obtained with the BM CD49a+ cells (P < 0·01). Moreover, a lot of CFU-F were detected in the CD34 fraction. From the CD49a+ cells, almost the entire CFU-F potential was detected in the CD49a+/CD34 fraction. The proportion of CFU-F was significantly higher in the CD49a+/CD34 fraction than that of the CD49a+/CD34+ fraction (84% ± 5 and 16% ± 5 respectively, P < 0·001). Thus, the CD49a molecule is a more specific marker for MSC than CD34. However, a small proportion of CD34+ cells were able to generate CFU-F and should be included in the CD49a+ fraction.

Other molecules involved in cell adhesion mechanisms were examined. Adhesion molecules are crucial for the expansion, self-renewal and differentiation of a stem cell. Molecules from the integrin and immunoglobulin superfamilies have been shown to have a significant role in these mechanisms (Verfaillie, 1998; Arroyo et al, 1999; Cortes et al, 1999; Zhu et al, 1999; Potocnik et al, 2000; Rosu-Myles et al, 2000; Levesque et al, 2001). For example, the VLA-4 (CD49d/CD29) and VLA-5 (CD49e/CD29) integrins (CD106 and fibronectin receptors respectively) are expressed by stem cells from both the haematopoietic and the mesenchymal compartment (Conget & Minguell, 1999; Pittenger et al, 1999; Deans & Moseley, 2000). Moreover, the homing of a stem cell has been shown to be dependent on the membrane expression of CXCR-4, the stromal cell-derived factor 1 (SDF-1) receptor (Peled et al, 1999a,b). CXCR-4, CD49d and CD49e molecules have also been detected on SA cells.

In conclusion, no CFU-F have ever been generated from CD3+, CD19+ and glycophorin A+ cells (Simmons & Torok-Storb, 1991b; and data not shown), 40% to 60% of SA cells were able to generate CFU-F in culture. After cryopreservation, the MSC-containing compartment represented 0·7% of the total BM mononuclear cells. This population was found to be CD34low, lin, HLA-II. In addition, only CD45 cells, detected in the SA cell fraction, were glycophorin A+(Fig 2A), and no CD73+ cells were CD45 (Fig 2B). Therefore, the CFU-F were generated by CD45+ cells.

Figure 2.

Phenotypic characterization of SA cells using flow cytometry. SA cells from the BM CD49a+ fraction were analysed. All CD45-negative cells were glycophorin A + erythroblastic cells (A). The expression of CD73, specific to MSC, was also assessed. No CD73+/CD45 cells were detected (B).

The receptor tyrosine phosphatase molecule CD45 (leucocyte common antigen) is present on all haematopoietic cells except mature erythroblastic glycophorin A+ cells (Penninger et al, 2001). Our phenotypic studies on SA cells showed an expression of the CD45 molecule by unprocessed MSC. To confirm this result, we performed both an immunomagnetic and a flow cytometric sorting (see Materials and methods). Using an indirect immunomagnetic selection method and CD45 antibody (clone PD/26), an equal number of CFU-F was found in both the CD45 (rejected cells) and CD45+ (retained cells) fractions. Indeed, no significant difference was observed between the two CE (3 ± 1 for CD45 fraction and 4 ± 1 for CD45+ fraction, n = 6) with 82% ± 7 of total recovery. We used another CD45 selection method involving serial immunomagnetic selections. Beads were coated with a low concentration (1 µg/107 beads) of an alternative mAb anti-CD45 (clone B-A11). BM mononuclear cells were stained directly with these types of beads. The CD45 cells were then incubated with the anti-CD49a mAb. CD49a+, CD49a and CD45+ cells were cultured for their CFU-F potential. All CFU-F were recovered in the CD45+ fraction (n = 3) except for a few CFU-F found in the CD49a+/CD45 fraction. The CD45 direct selection enabled the retention of more MSC than the indirect protocol. The flow cytometric method, using CD49a/CD45 double staining, detected mainly three populations (Fig 3): CD49a/CD45, CD49a/CD45+ and CD49a+/CD45+. Very few cells were found in CD49a+/CD45 fraction. Within the CD49a+ fraction, two cell populations were observed with the anti-CD45 mAb: 33% ± 5 cells were CD49a+/CD45med,low and 66% ± 5 cells were CD49a+/CD45high (1% ± 0·5 and 2% ± 0·5 in total BM mononuclear cells respectively). These populations were sorted from BM SA cells to avoid cell aggregates and were cultured for 15 d. No CFU-F was detected in the CD49a (confirming the results from CD49a+ immunomagnetic sorting), CD49a+/CD45high and CD49a+/CD45 fractions. On the other hand, CFU-F were only generated from the CD49a+/CD45med,low cell population. The CE for this fraction was 227 ± 75 or one CFU-F/440 CD49a+/CD45med,low cells seeded. In conclusion, MSC with CFU-F potential that were contained in the small CD49a+/CD45med,low, HLA-II cell population also showed weak expression of CD34.

Figure 3.

Flow cytometric cell sorting. Bone marrow mononuclear cells were stained by anti-CD49a and anti-CD45 (pan leukocytic marker) mAb and analysed by flow cytometry. Two distinct populations of cells were found that were positive for both CD49a and CD45 when compared with the negative control (isotype control). In the R1 (2% mononuclear cells) region, cells were CD45high while cells in the R2 region, 1% of mononuclear cells, expressed this molecule at a low-to-moderate level. After sorting, all CFU-F were detected in the R2 region.

Proliferation and phenotypic studies of cultured mesenchymal cells from CD49a+fraction

Adherent mesenchymal progenitor cells were shown to proliferate without differentiation when cultured in expansion medium (Pittenger et al, 1999). Bone marrow CD49a+ cells were plated at 1 × 104 cells per cm2 and also cultured in this medium. After 15 d of culture, the cells from the CFU-F were separated for phenotypic analysis, and for proliferation and differentiation assays.

The number of MSC generated was determined throughout the 8 weeks of culture in a non-clonal system. During the first 3 weeks, the cells had a high proliferation rate (log phase) which slowed considerably between 3 and 8 weeks (Fig 4). At this plating density, the cells expanded by up to 10 000-fold over 3–4 weeks and produced more than 1·1 × 106 ± 0·2 mesenchymal cells from 1 ml of normal BM. The number of doublings (15 ± 0·5) and the time for one division (39 ± 1 h) were relatively moderate. However, these values are similar to other reports for non-clonal systems (Bruder et al, 1997; Conget & Minguell, 1999; Colter et al, 2001).

Figure 4.

Proliferation of mesenchymal cells from the BM CD49a+ fraction. Cells from CFU-F were seeded at 1 × 104 cells/cm2 and cultured in expansion medium until 80% confluence was reached. The cells were resuspended, counted and reseeded at the same density. Between 5 and 8 weeks of culture, more than 1 × 106 mesenchymal cells were produced, corresponding to a 10 000-fold expansion.

From 2 to 5 weeks of culture, cells from the CFU-F were tested for the expression of membrane molecules (Table II) using flow cytometry. These cells were found to be HLA-II and HLA-I+, as previously reported (Pittenger et al, 1999; Deans & Moseley, 2000). The CD166, CD44 (hyaluronate receptor) and CD49e molecules were strongly expressed. The expression of CD54, CD50, CD31 and CD49d were moderate. The CXCR-4 receptor was not detected on the cell surface but was found at the intracytoplasmic level. The CD49a molecule was consistently expressed. Haematopoietic markers were not detected (CD45, CD14, CD34 and CD38) except for CD10. In addition, Thy-1 (CD90), Jagged (Notch receptor) were detected whereas c-kit was not. The expression of molecules that are characteristic of the expanded mesenchymal progenitors (Pittenger et al, 1999; Deschaseaux & Charbord, 2000) was assessed using in situ IF and flow cytometry. The membrane markers Stro-1, CD73, 1B10, and the cytoskeletal molecules vimentin, αSM-actin and ECM fibronectin were expressed (Table II).

Table II.  Phenotype studies of expanded CD49a+ MSC.
Detection of antigens from
Adhesion molecules*Haematopoietic markers* HLA molecules* Others*
  • *

    The molecules were detected by flow cytometry

  • †except for

  • α

    SM-actin, vimentin and fibronectin proteins, which were detected by in situ IF.

  • +, most of the cells were positive; ±, a few cells were positive; –, negative; (+), positive for intracellular antigen.

CD50±CD34  Jagged+
CD31±CD45  Stro-1+
CD44+CD38  SH3+
CD49a+    1B10+
CD49d± (+)    αSM-actin+
CD49e+    Vimentin+
CXCR4– (+)    Fibronectin+

Mesenchymal stem cell multipotential studies

The CFU-F derived from MSC have been shown to generate all cells of the mesenchymal lineage in vivo and in vitro (Deans & Moseley, 2000). Osteoblastic cells can be characterized in vitro by the specific expression of osteocalcin and by a strong secretion of CO I. A high alkaline phosphatase activity has also been described (Bruder et al, 1997). In contrast, chondroblastic cells express CO II at the end of differentiation process (Barry et al, 2001a), while lipid droplets, stained with oil red O, are easily detected within adipocytic cells. Cells from the same BM sample were cultured in both differentiation medium and expansion medium (negative control). Differentiation into the osteoblastic, chondroblastic and adipocytic lineages was induced. Osteoblastic cells, generated from the CFU-F, were detected after 2–3 weeks of culture in osteogenic induction medium. Strong expressions of both extracellular osteocalcin and CO I were seen with in situ IF compared with the weak staining that was detected in the negative control (Table III). The expression of these molecules, as quantified by ELISA and histochemistry, are shown in Fig 5A. Collagen I and osteocalcin were expressed significantly more (P < 0·006) than the negative control. These low expressions of the osteoblastic markers by control cells confirmed our in situ IF results and were in agreement with other investigations (Banfi et al, 2000). In addition, we detected a significantly (P < 0·0001) stronger alkaline phosphatase activity compared with the controls. The chondroblastic cells were assayed by the same immunodetection protocol. Collagen II expression (Fig 5B) was observed exclusively on the stroma from chondrogenic induction cultures. A large number of adipocytic oil red O-positive cells was also detected only when cells were cultured in the adipogenic medium (Table III). Finally, BM MSC cultured in long-term medium (Dexter cultures) generate stromacytic or myofibroblastic colonies (Galmiche et al, 1993). These stromacytes can form a microenvironment which sustains haematopoiesis when seeded with CD34+ HSC (Gartner & Kaplan, 1980). The stromacytic differentiation pathway detailed in our previous report (Deschaseaux & Charbord, 2000) was confirmed by a quantitative functional study. Nevertheless, some clusters of well-differentiated myofibroblasts, expressing the smooth muscle myosin heavy chain (hSM-V+ cells), were detected with in situ IF in the stromas cultured within expansion medium (Table III). Quantification of the long-term haematopoiesis-sustaining function of stromacytic cells was undertaken using the CAFC assay. After 6 weeks of co-culture, the number of CAFC generated by BM CD34+ haematopoietic progenitor cells was determined by limiting dilution. Then, 124 CD34+ cells (n = 5; r2 = 0·74) were seeded to obtain one CAFC. This value was in agreement with that of a previous report (Mazini et al, 1998).

Table III. In situ immunofluorescence studies.
AntigensCells cultured in differentiation medium promoting lineage ofControl medium
  • *Stromacyte lineage was studied in our previous work (Deschaseaux & Charbord, 2000).

  • FN-15 mAb recognizes a common region of the differently spliced FN.

  • Vascular smooth muscle myosin heavy chain.

  • §

    A few clusters of VSMMyo-positive cells were observed among negative stromal cells.

  • NT, not tested. The detected fluorescence was weak (±) to strong (+); –, negative.

CO I+±++±
Nile red oil_+_±_
Figure 5.

Differentiation potential tests. Before induction of differentiation, cells from CFU-F were cultured in 96-well plates in expansion medium until confluence was reached. The negative control was a cell fraction cultured in unsupplemented expansion medium. Alkaline phosphatase activity, expression of CO I and osteocalcin were quantified by using colorimetric methods performed on cells incubated in osteogenic culture conditions (A). CO II expression is shown for cells cultured in chondrogenic medium (B). For all assays, a basal expression of CO I and osteocalcin was detected in stroma maintained in control medium. The calculated absorbance ratio values between these two culture conditions were 1·8 ± 0·2 for the alkaline phosphatase activity, 1·4 ± 0·1 for CO I and 1·3 ± 0·04 for osteocalcin (A). These increases were significant (*P < 0·0001, **P < 0·006). Collagen II was detected (0·1 ± 0·01; *P < 0·0001) only in cells cultured in chondrogenic medium (B).

These results demonstrated that the cells from CFU-F generated osteo-chondroblastic, adipocytic and functional stromacytic cells. In conclusion, all multipotential MSC with CFU-F potential were generated from CD49a+ cells. However, some lineage-specific markers were detected before induction of differentiation, suggesting that either more mature cells were included in the freshly selected CD49a+ fraction, or the expansion medium induced the differentiation.


Multipotential BM MSC generate osteo-chondroblastic cells, adipocytes and stromacytes (Pittenger et al, 1999). This cell type is usually selected indirectly using its adherence capacity to plastic culture dishes and is characterized by an ability to generate CFU-F in expansion medium. The function and phenotype of fresh or unprocessed MSC have not been fully characterized. However, identification of these cells is crucial for the study of the normal development of all mesenchymal-deriving cells and their potential changes in diseases. In this study, we have described an original method to directly select MSC from human BM. By using immunomagnetic beads directed to CD49a+ cells, we were able to recover efficiently all the CFU-F within a small population of cells. Our data showed that the selection of mesenchymal progenitors using anti-CD49a mAb was as efficient as the selection using the Stro-1 antibody (Simmons & Torok-Storb, 1991a). In addition, cells from the CFU-F were able to generate osteoblastic and chondroblastic cells, adipocytes and functional stromacytes as described by others using the indirect procedure (Pittenger et al, 1999; Deans & Moseley, 2000). Mesenchymal stem cells, therefore, belong to the CD49a+ BM fraction and can be directly selected using this immunomagnetic method. Although the freshly separated cell fraction contained mainly mature haematopoietic cells (granulocytic and erythroblastic cells), an enrichment of CFU-F was observed after cryopreservation. This indicated that MSC were preferentially preserved by this method, confirming a previous report (Bruder et al, 1997).

Several descriptions of indirectly selected MSC being CD45 have been reported (Conget & Minguell, 1999; Pittenger et al, 1999; Deans & Moseley, 2000). However, our results showed that fresh MSC expressed CD45. This discrepancy can be explained by the phenotypic changes induced by cell culture: unprocessed MSC were CD45+ and became CD45 when cultured. These results are in agreement with the previous report of Clark et al (2001). Using flow cytometry, they sorted MSC from human BM exclusively in a small CD45med,low population of cells that also expressed an antigen recognized by the SH2 antibody. SH2, SH3 and SH4 antibodies have been described to be specifically expressed by MSC (Haynesworth et al, 1992). The SH3 and SH4 epitopes are found on the CD73 molecule (Barry et al, 2001b). Our experiments have also shown that the CD49a+/CD73+ cells did not contain CD45 cells. Nevertheless, some investigators have used a negative selection based on the CD45 marker in order to deplete haematopoietic cells (or non-MSC) in fresh BM using an immunomagnetic method (Tondreau et al, 2001). Our study strongly argues against such a selection as the CFU-F were always detected in the CD45+ fraction. Our hypothesis to explain this discordance was that the CD45med,low cells could not be retained by the magnetic field. Some technical reports have shown that the recovery using magnetic selection depends on the magnetic field, on the number of beads bound to the cell membrane or on the number of targeted membrane proteins (Chalmers et al, 1998; Comella et al, 2001). Thus, CFU-F obtained from the CD45 fraction can be explained by the fact that the immunomagnetic procedure has not retained MSC because of the weak expression of the CD45 protein. Using the indirect immunomagnetic CD45 selection, we obtained equivalent CFU-F in both positive and negative fractions. The use of a secondary antibody theoretically increases the number of bound antibodies that may reduce the specificity. In addition, the mAb used within the direct selection protocol could be more efficient in recognizing the antigen than the mAb used in the indirect method. The CD45 molecule also exists in different isoforms (Penninger et al, 2001), so we can hypothesize that MSC could be excluded by some other mAb. Therefore, unprocessed MSC were CD45+, while cultured or more mature mesenchymal cells were CD45. The functional significance of this downregulation is unclear. However, CD45 is also a negative regulator of both stem cell factor and interleukin 3 receptors (Penninger et al, 2001). This latter receptor is expressed on cultured MSC while c-kit is not (Deans & Moseley 2000). Although these cytokines are specific to HSC (Potten, 1997), they could have a role in the MSC compartment. Surprisingly, there are three types of stem cells that reside in the BM and express CD45: MSC, HSC and a stem cell with the SP phenotype determined by flow cytometry (Goodell et al, 1997). The expression of CD45 by HSC and MSC and their BM localization show that these cells are perhaps more closely related than previous reports indicate. This is supported by a previous report that showed this close relationship. Singer et al (1987) generated BM stromal cell lines that expressed haematopoietic markers. Indeed, the SV40 transformation of adherent cells permitted the detection of the CD45 marker on stromal-like cells. Thus, the literature and our results assume that MSC are able to express the pan-haematopoietic marker CD45. Nevertheless, Reyes et al (2001) reported a BM CD45 stem cell more primitive than CD49a+ MSC. The multipotent adult progenitor cell (MAPC) was detected within the CD45 fraction after immunomagnetic selection. The MAPC could be another BM stem cell that does not express the CD45 molecule or that moderately expresses this protein but is rejected by the immunomagnetic method as explained above.

Other membrane molecule expression studies in freshly selected MSC have shown that the HSC marker CD34 was barely detectable. Mesenchymal stem cells have been reported by some investigators to be CD34+ (Simmons & Torok-Storb, 1991b). However, we have shown that only a few CFU-F were derived from BM CD49a+/CD34+ cells compared with BM CD49a+/CD34 cells. The CD34+/CD49a+ cells with CFU-F potential could thus represent a subpopulation of MSC. Finally, HLA class II was consistently not expressed by MSC. Therefore, we showed that an unprocessed CD49a+ population of cells, containing the primitive MSC, has a CD34low, CD45med,low, HLA-II phenotype.

Results of our phenotypic studies performed during the expansion phase of the cells were in accordance with previous reports using the indirect selection method (Conget & Minguell, 1999; Pittenger et al, 1999; Deans & Moseley, 2000). However, the in vitro step decreased the expression of CD49d and CXCR4 molecules. The weak detection of CD49d on cultured MSC is not consistent with the report of Pittenger et al (1999) but is in agreement with the results of Conget and Minguell (1999) and our previous studies (Deschaseaux & Charbord, 2000). The different conditions of culture, e.g. different FCS or the use of different antibodies, could explain these discordant results. The CD49d and CXCR4 proteins were shown to be crucial for the homing of the stem cell to the BM (Simmons & Torok-Storb, 1991b; Mazini et al, 1998). Therefore, altering the phenotype could decrease the homing efficiency and the reconstitution process as described previously for HSC (Peled et al, 1999a,b).

In situ immunofluorescence or ELISA examination of expanded cells demonstrated the weak expression of specific osteoblastic differentiation markers. This result could signify that either more differentiated mesenchymal progenitors, e.g. osteoblastic progenitors, can be isolated with MSC or that the osteoblastic lineage could be induced by the expansion medium. The latter hypothesis was proposed by Banfi et al (2000) who reported expression of osteocalcin in unstimulated cultures detected by reverse transcription polymerase chain reaction. The presence of more mature mesenchymal cells among MSC was supported by data from clonal studies (Muraglia et al, 2000). Moreover, it has been shown that only 30% of CFU-F were tripotential (showing osteocytic, chondrocytic and adipocytic potential). Other CFU-F seemed to be more differentiated because they are only bi or unipotential. As, 100% of CFU-F were recovered in our CD49a+ fraction, it is highly probable that these more mature progenitors were contained in the cultures used for expansion or in non-clonal multipotential studies. In support of this hypothesis is our finding that mature myofibroblastic cell clusters are detected in expansion medium cultures. Previous work has also demonstrated that VLA-1 integrin was expressed throughout the VSM lineage (Glukhova & Koteliansky, 1995). Hence, these more mature cells may have been retained in the CD49a+ fraction and formed small clusters in the expansion cultures. The proliferation study showed the number of doublings was relatively low compared with other studies (Bruder et al, 1997; Banfi et al, 2000). One would assume that the presence of more differentiated CFU-F would decrease these values. This fact has also been shown by Colter et al (2000) who demonstrated that MSC contained two populations: one had a strong proliferation potential (> 40 doublings) and the other had a significantly lower value. The mixture of these two populations gave a relative low number of doubling. In addition, they also showed that expansion of mesenchymal cells is negatively regulated by the density per cm2 of cells seeded. The greatest expansion was obtained when only 3 cells/cm2 were cultured. However, in our work, the concentration of 1 × 104 cells seeded/cm2 was used. This could also help to explain the results found in our proliferation study.

In conclusion, using CD49a selection, both MSC and more differentiated mesenchymal progenitors were isolated. However, in order to demonstrate this heterogeneity, it was necessary to distinguish, phenotypically and functionally, the primitive compartment from the more differentiated one. Gronthos et al (2001) showed that stromal precursors had a higher cloning efficiency when seeded on CO IV, fibronectin, vitronectin and laminin when compared with CO I and III. This differential growth pattern observed on different CO species may reflect the primitive state of the stromal precursors. The mesenchymal compartment heterogeneity can be, therefore, studied through expression of membrane counter receptors. In addition, these results confirmed the MSC specificity of the CD49a protein because the VLA-1 integrin is the CO IV and laminin receptor. Using our data, we suggest a mechanism to summarize the changes in cell phenotype found in the CD49a+ mesenchymal compartment from unprocessed to cultured cells (Fig 6). Unprocessed or fresh MSC were CD45+ while cultured MSC and more mature cells were CD45. The exact phenotype of the most primitive precursor may be only truly determined by the use of CD49a selection and culture conditions which closely mimic the natural microenvironment. In addition, the use of the CD49a marker for the specific selection of unprocessed MSC could be an advantage for manufacturing MSC for clinical use.

Figure 6.

The CD49a+ mesenchymal compartment. A hierarchical model of ‘mesengenesis’ is shown. MSC generated in vitro fibroblastic colonies (CFU-F). Cells from these CFU-F generated osteo-chondroblastic, adipocytic and stromacytic cells in differentiation-inducing media. MSC and cells from CFU-F were CD49a+/HLA-II. Only unprocessed MSC showed the expression of the CD45 haematopoietic marker. Phenotypic studies and culture of BM CD49a+ cells showed a subpopulation of unprocessed MSC expressing the CD34 protein (CD34low). The CD49d and CXCR4 molecules, involved in adhesion and migration processes, were downregulated by the in vitro culture.