Multilineage Potential of Adult Human Mesenchymal Stem Cells

See allHide authors and affiliations

Science  02 Apr 1999:
Vol. 284, Issue 5411, pp. 143-147
DOI: 10.1126/science.284.5411.143

This article has a correction. Please see:


Human mesenchymal stem cells are thought to be multipotent cells, which are present in adult marrow, that can replicate as undifferentiated cells and that have the potential to differentiate to lineages of mesenchymal tissues, including bone, cartilage, fat, tendon, muscle, and marrow stroma. Cells that have the characteristics of human mesenchymal stem cells were isolated from marrow aspirates of volunteer donors. These cells displayed a stable phenotype and remained as a monolayer in vitro. These adult stem cells could be induced to differentiate exclusively into the adipocytic, chondrocytic, or osteocytic lineages. Individual stem cells were identified that, when expanded to colonies, retained their multilineage potential.

Recently, pluripotent stem cells have been cultured from human fetal tissue and have shown the ability to give rise to a variety of differentiated cell types found in embryonic germ layers (1). Many adult tissues contain populations of stem cells that have the capacity for renewal after trauma, disease, or aging. The cells may be found within the tissue or in other tissues that serve as stem cell reservoirs. For example, although bone marrow is the major source of adult hematopoietic stem cells (HSCs) that renew circulating blood elements, these cells can be found in other tissues (2). The adult bone marrow also contains mesenchymal stem cells (MSCs), which contribute to the regeneration of mesenchymal tissues such as bone, cartilage, muscle, ligament, tendon, adipose, and stroma (3–5). In vitro and animal implant studies (5–10) have indicated that there is either a multipotent MSC or the populations are mixtures of committed progenitor cells, each with restricted potential. We report the isolation, expansion, and characterization of the multipotent human MSC (hMSC).

We characterized an isolated population of homogeneous human mesenchymal cells from bone marrow taken from the iliac crest (5) (see Web Fig. 1, available Previous studies have identified selection criteria for fetal bovine serum (FBS) that allows the expansion of a marrow cell population with MSC potential after implantation (5, 11). The mesenchymal cells described here were characterized by their ability to proliferate in culture with an attached well-spread morphology (Fig. 1, A and B), by the presence of a consistent set of marker proteins on their surface (Fig. 1, C and D) (12–14), and by their extensive consistent differentiation to multiple mesenchymal lineages under controlled in vitro conditions (Fig. 2).

Figure 1

Characterization of isolated marrow stromal cells. Cells were cultured from marrow after density fractionation and are shown (A) at 48 hours after plating and (B) at 10 days after plating. (C) Flow cytometry shows the enrichment of these cells as they were cultured. Results shown were obtained at days 2, 5, and 14 of culture with antibodies SH2 and SH3, which were raised against surface markers (11). At 14 days, the cells were 95 to 99% homogeneous and were negative for reactivity to antigens CD14, CD34 (Becton-Dickinson), or CD45 (Pharmingen), which are common on cells of the hematopoietic lineages. (D) Homogeneity and reproducibility of the isolation procedure was demonstrated by flow cytometry. See Web Fig. 1 for additional information.

Figure 2

Isolated marrow-derived stem cells differentiate to mesenchymal lineages. Cultured cells from donors were tested for the ability to differentiate in vitro to multiple lineages. Donors (A through C) 158, (Dthrough F) 177, and (G throughI) 260 were each shown to differentiate appropriately to the adipogenic (Adipo), chondrogenic (Chondro), and osteogenic (Osteo) lineages. Adipogenesis was indicated by the accumulation of neutral lipid vacuoles that stain with oil red O (A, D, and G), and such changes were not evident (J) with Hs27 newborn skin fibroblasts or (M) with 1087Sk adult mammary tissue fibroblasts. Chondrogenesis was shown by staining with the C4F6 monoclonal antibody to type II collagen and by morphological changes (B, E, and H), which were not seen by similarly culturing (K) Hs27 or (N) 1087Sk cells. Osteogenesis was indicated by the increase in alkaline phosphatase (C, F, and I) and calcium deposition, which was not seen in the (L) Hs27 or (O) 1087Sk cells (see Web Fig. 4).

A density gradient was used in the isolation procedure to eliminate unwanted cell types that were present in the marrow aspirate. A small percentage (estimated at about 0.001 to 0.01%) of cells isolated from the density interface of 1.073 g/ml attached and grew as fibroblastic cells that developed into visible symmetric colonies at about 5 to 7 days after initial plating. HSCs and nonadherent cells were removed with changes in medium. The isolated cultured mesenchymal cells comprised a single phenotypic population (95 and 98% homogeneous at passages 1 and 2, respectively) by flow cytometric analysis of expressed surface antigens; whether or not trypsin was used to remove the cells from the substrate did not change the results. These expanded attached mesenchymal cells were uniformly positive for SH2, SH3, CD29, CD44, CD71, CD90, CD106, CD120a, CD124, and many other surface proteins. Immunofluorescence microscopy evaluation showed no unlabeled cells. No subpopulations of marrow-derived mesenchymal cells could be discerned morphologically by microscopic observation or by fluorescence cytometry, size, and granularity criteria, or with more than 50 available antibodies (11, 13, 14) (see Web Fig. 1). Several of these antibodies were used to routinely characterize the expanded mesenchymal cell population. In contrast, the mesenchymal cells were negative for other markers of the hematopoietic lineage, including the lipopolysaccharide receptor CD14, CD34, and the leukocyte common antigen CD45 (Fig. 1D). Hematopoietic cells were never identified in the expanded cultures. The mesenchymal cells appear to be different from marrow fibroblastic cells isolated by other cells that remained positive for the CD34 surface antigen (15). As many as 50 million to 375 million cells were generated by passage 2 from a 10-ml marrow aspirate that was obtained from 19- to 57-year-old donors (see Web Fig. 1). This range of yield is expected for the expansion of primary cells. We isolated these cells from the marrow of over 350 donors, and the expanded mesenchymal cells from at least 50 donors were tested in three lineage assays, with all tested donors responding positively in these assays. The cells did not differentiate spontaneously during culture expansion. Similarly isolated marrow mesenchymal cells maintained a normal karyotype and telomerase activity, even at passage 12 (see Web Fig. 1).

The differentiation potential of the mesenchymal cells from three adult donors were compared to two strains of human fibroblasts (16), which are other cells of mesenchymal lineage, by culturing the cells under conditions that were favorable for adipogenic, chondrogenic, or osteogenic differentiation. After 1 to 3 weeks in lineage-specific culture conditions, the expanded cells from the three donors were highly differentiated (the phenotype of lineage-specific cell types was easily distinguished), without evidence of the other lineages (Fig. 2, A through I).

Adipogenic differentiation was induced in the expanded mesenchymal cell cultures by treatment with 1-methyl-3-isobutylxanthine, dexamethasone, insulin, and indomethacin (17). Induction was apparent by the accumulation of lipid-rich vacuoles within cells (Fig. 2, A, D, and G). These adipocytes expressed peroxisome proliferation-activated receptor γ2 (PPARγ2), lipoprotein lipase (LPL), and the fatty acid–binding protein aP2 (18) (see WebFig. 2, available at Multiple induction treatments resulted in more than 95% of the cells committing to this lineage, and the lipid vacuoles continued to develop over time, coalesced, and eventually filled the cell. These adipocytes remained healthy in culture for at least 3 months.

To promote chondrogenic differentiation, we gently centrifuged the isolated mesenchymal cells to form a pelleted micromass and cultured the cells without serum and with transforming growth factor–β3 (19). The cell pellets developed a multilayered matrix-rich morphology and histologically showed an increased proteoglycan-rich extracellular matrix during culture. Type II collagen (Fig. 2, B, E, and H), which is typical of articular cartilage, was detected at 10 to 14 days with monoclonal C4F6 (20). For each donor, the cells showed a positive stain for chondrocytic markers over an increasing proportion of the pellet from 10 to 21 days, and this persisted without decreasing. Similar cultures have been maintained for 3 months. Chondrocyte-like lacunae were evident in histological sections, and the extensive extracellular matrix was rich in aggrecan and type II collagen (see Web Fig. 3, available

Figure 3

Flow cytometry characterization of clonal cells. Marrow stromal cells from donor 398 were plated at low density, and individual attached cells were identified on the culture dish. When a colony was evident, a cloning ring was placed around it, and the colony was subcultured to an individual well of a 24-well culture plate. The expanded cells were transferred to a 75-cm2flask, analyzed when they approached confluence, and compared to the parent cells. Colony cells were stained with surface markers SH3 and SH2 or CD14, CD34, or CD45 and found to be homogeneous and indistinguishable from the parent culture. Cells of colonies 3 and 12 were multipotential in the differentiation assays (as shown in Fig. 4).

The osteogenic differentiation of cultured human mesenchymal cells that were derived from adult bone marrow has been characterized (3,21). Under the influence of dexamethasone, β-glycerol phosphate, and ascorbate and in the presence of 10% v/v FBS, the isolated mesenchymal cells formed aggregates or nodules and increased their expression of alkaline phosphatase (Fig. 2, C, F, and I). Quantitative assays revealed a 4- to 10-fold increase in alkaline phosphatase activity, and calcium accumulation was evident after 1 week and increased over time (see Web Fig. 4, available

Figure 4

Comparison of the differentiation potential of parental and clonally derived cells. (A) Cells of colonies 3 and 12 were placed in the differentiation conditions and were shown to give rise to adipogenic, chondrogenic, and osteogenic lineages. The response was very similar to the parental cell population. The colonies also gave a similar response in the quantitative assays for (B) adipogenic lipid accumulation and (C) osteogenic calcium deposition (OS).

As differentiation to the three lineages occurred, there was no evidence of necrosis or apoptosis; it is apparent when groups of dead cells are present because there are no scavenger cells to remove them. However, we cannot rule out the fact that apoptosis may occur on a single-cell basis.

In contrast, cultured normal skin fibroblasts did not undergo any such differentiation when cultured for as long as 28 days, although the cells appeared healthy and nonnecrotic in these different conditions (Fig. 2, K and N). Cultured hMSCs, therefore, had a differentiation capacity that was not present in primary fibroblasts, which are mature mesenchymal cells.

The entire culture of expanded mesenchymal cells appeared to progress to the desired lineage, on the basis of phenotypic characterization and histological analysis. Multiple lineages were not present. To further test the possibility that a barely detectable population of cells could differentiate to lineages other than those desired in the in vitro assays, we isolated RNA from cells after differentiation, and reverse transcriptase–polymerase chain reaction (RT-PCR) was used to amplify expressed transcripts from multiple genes whose expression is restricted to the adipocytic (PPARγ2 and aP2), chondrocytic (type II and IX collagens), or osteocytic (osteopontin and alkaline phosphatase) lineage (see Web Fig. 5, available The sensitive RT-PCR assays, in addition to the immunological and histochemical data, showed that neither the adipogenic nor the osteogenic cultures were composed of cells of multiple lineages. The chondrogenic cultures did not appear histologically to contain cells that were of the other lineages, but they did express osteopontin and alkaline phosphatase.

These experiments showed that the isolated expanded hMSC culture would differentiate, in a controlled manner, to multiple lineages. To test the hypothesis that marrow contains individual hMSCs that could differentiate to all three of the lineages and to rule out that marrow contains only committed progenitor cells, we performed the same experiments with expanded colonies of cells derived from single cells. For this, the Percoll-isolated marrow cells were plated at 1.5 million to 6 million cells per 150-mm dish. Individual adherent cells were identified at 24 hours and grown into colonies of 30 to 100 cells, and the colonies were isolated with cloning rings. Isolated colonies were treated with trypsin and transferred to individual wells of a 24-well tissue culture plate and then transferred to a 75-cm2 flask as they expanded. Approximately 800,000 to 1,500,000 cells were obtained from the flasks at harvest, representing 19 to 21 in vitro population doublings from the original cell. The cells were subjected to flow cytometric analysis and in vitro differentiation assays.

Flow analysis of the colonies revealed the same surface molecules as those found on the cultured undifferentiated hMSCs from which they were derived (Fig. 3) and on the other donors that were studied (Fig. 1). Of the six colonies derived from clonal cells, all underwent osteogenic differentiation, five underwent adipogenic differentiation, and two also underwent chondrogenic differentiation (Fig. 4). One other colony was weakly but clearly positive for the chondrocytic phenotype. The fact that some of the marrow-derived adult human mesenchymal cells, which were grown and expanded as individual colonies, possessed a multipotent capacity for lineage differentiation supports our conclusion that these cells represent human mesenchymal stem cells (hMSCs). The observation that some colonies displayed limited differentiation under these experimental conditions has several possible interpretations. For example, the clonal cell isolation, 19 to 21 population doublings, and in vitro culture conditions may have caused some loss of multilineage potential that was originally present in these cells. Another interpretation is that some of these cells might represent progenitor cells, with restricted differentiation potential, that are present in adult marrow.

The specific environmental cues to initiate the proliferation and differentiation of hMSCs in vivo are not understood. The ability to isolate, expand the culture, and direct the differentiation of hMSCs in vitro to particular lineages provides the opportunity to study events associated with commitment and differentiation. From the different assay conditions required for efficient differentiation, we conclude that basal nutrients, cell density, spatial organization, mechanical forces, and growth factors and cytokines have a profound influence on hMSC differentiation. The hMSCs cultured in each of the differentiation conditions may also produce autocrine and paracrine factors that are essential for lineage progression. As for the differentiation to other mesenchymal lineages, expanded mesenchymal cells have been shown to function as stromal cells for the in vitro support of HSCs (13), and rabbit MSCs have been shown to be effective for the regeneration of a severed tendon in a rabbit model (8).

The hMSCs described here have the ability to proliferate extensively, and they maintain the ability to differentiate into multiple cell types in vitro, establishing their stem cell nature. Their cultivation and selective differentiation should provide further understanding of this important progenitor of multiple tissue types and the potential of new therapeutic approaches for the restoration of damaged or diseased tissue.

  • * To whom correspondence should be addressed. E-mail: mpittenger{at}


View Abstract

Navigate This Article