Stromal cells generated in long-term cultures appear to follow a vascular smooth muscle differentiation pathway. Such a pathway, comprising several steps hallmarked by the expression of cytoskeletal and extracellular matrix markers, is found not only for bone marrow stromal cells, but also for stromal cells generated from the different developmental sites of hematopoiesis (yolk sac, aorta-gonad-mesonephros region, fetal liver, and spleen). Factors responsible for this differentiation pathway and its functional significance are discussed. The mesenchymal founder cell might be, at least for bone marrow, a mesenchymal stem cell (MSC), giving rise to stromal cells, endothelial cells, adipocytes, osteoblasts, and chondrocytes. A feature that distinguishes the MSC lineage from that of the hematopoietic stem cell lineage is that differentiation pathways are not strictly delineated, since even apparently fully differentiated cells from a given lineage have the potential to convert into another lineage (phenotype “plasticity”) and intermediate cell phenotypes are observed. A stochastic Repression/Induction model that would account for this plasticity is proposed.
Stromal cells constitute the cell population that assists the hematopoietic stem cell (HSC) and its progeny, i.e., the set of cells modulating quiescence, self-renewal, and commitment of HSC and the proliferation, maturation, and apoptosis of more mature hematopoietic cells. Stromal cells are easily defined in culture, forming the nonhematopoietic adherent cell component from long-term cultures. In vivo, they make up the microenvironment of hematopoiesis, comprising the set of nonhematopoietic cells from the different hematopoietic sites. This concise review is focused on recent data providing insight into the phenotype of the stromal cells and the relationships among stromal cells, the vascular smooth muscle lineage, and mesenchymal stem cells (MSCs). In a number of recent and excellent reviews on stromal stem cells [1–, 3] or MSCs [4,, 5], the reader will find views emphasizing other aspects of stromal cell physiology, such as stromal cells in vivo, the ontogeny of stroma, and gene transfer into stromal cells.
Mesenchyme to Vascular Smooth Muscle Cells (VSMC), a Differentiation Pathway
Mesenchymal cells are primordial cells of mesodermal origin, forming the connective tissue throughout the body. Mesenchymal cells would therefore give rise to fibroblasts, osteoblasts, chondrocytes, smooth muscle cells, and endothelial cells, as well as stromal cells with hematopoietic support. We have shown, by studying the expression of a number of markers on cells from human and murine long-term cultures, that stromal cells are the progeny of mesenchymal cells following a VSMC differentiation pathway.
Stromal Cells are Vascular Smooth Muscle-Like
The above conclusion, reached from the study of human long-term marrow cultures and murine cell lines, is based on two major lines of evidence: A) in human bone marrow primary cultures, cultures of colony-derived stromal cell lines (CDCLs), and cultures of stromal cells derived from Stro-1+ cells, there is sequential appearance of a set of cytoskeletal and extracellular matrix (ECM) proteins specific for VSMCs [6–, 9], and B) murine immortalized lines from the different hematopoietic sites (bone marrow, fetal liver, spleen, aorta-gonad-mesonephros region, and yolk sac) can be categorized into a VSMC lineage based on the expression of these markers [10,, 11].
In human cultures, the earliest markers are an actin isoform specific for VSMC, α-smooth muscle actin (ASMA), and the fibronectin isoform comprising the spliced extra-domain (ED) a, (encoded 3′ to the major arginine-glycine-aspartic acid [RGD]-dependent cell-binding site). Cytoskeletal VSMC-specific markers, which appear later, are: metavinculin and h-caldesmon, splice variants specific for VSMC of the actin-binding proteins vinculin and l-caldesmon, h1-calponin, SM22α, SM α-actinin (recognized by the monoclonal antibody 1E12), and SM myosin (comprising the 204 kDa SM1 heavy chain and, more rarely, the 200 kDa SM2 chain). In some murine lines, the muscle-specific intermediate filament component, desmin, is also detected. Stromal cells also express ECM proteins observed on developing VSMC: thrombospondin-1 and, at least for human stroma, the fibronectin variant comprising the EDb splice domain (encoded 5′ to the major RGD site) and the β2 isoform of laminin.
The expression of a relatively large number of markers is highly relevant, since no single marker allows a cell to be identified as belonging to the vascular smooth muscle lineage . During development, VSMCs from arteries progressively acquire their phenotype by expressing an increasing number of cytoskeletal and ECM molecules. It is only after birth that VSMCs express SM myosin and present a full contractile apparatus suitable for hemodynamic resistance . Studies on murine stromal lines [10,, 11] have shown that some of the VSMC markers are correlated. Remarkably, factors given by the statistical analysis associate markers characteristic of early differentiated VSMCs (ASMA, EDa+ fibronectin, thrombospondin-1), of VSMCs at midstage of differentiation (SM22-α, SM-actinin, h1-calponin, h-caldesmon, metavinculin), or of fully differentiated VSMCs (SM-1 myosin heavy chain, desmin). Stromal cell differentiation in culture appears to recapitulate the developmental program of VSMC differentiation. The existence of several steps along the VSMC differentiation pathway, where differentiation is arrested in some stromal cell lines, implies a heterogeneity of the stromal population. Such heterogeneity is also a characteristic of mesenchymal cells that acquire VSMC characteristics in physiological or pathological conditions .
Founder Mesenchymal Cells
Observations of the early stage of long-term cultures from human stroma have revealed a consistent pattern of proteins expressed by all stromal cells [6–, 9]. Moreover, stromal cells from all lines, regardless of their hematopoietic origin [10,, 11], express vimentin, laminin-β1, fibronectin, and osteopontin. Vimentin is an intermediate filament protein widely expressed by cells of mesenchymal origin. Laminin-β1 and fibronectin are detected prior to gastrulation, and osteopontin is expressed during gastrulation [14–, 16]. Due to their distribution in vivo, vimentin, laminin-β1, fibronectin, and osteopontin (in mouse) appear to be adequate markers for mesenchymal cells that constitute the first adherent cells appearing in long-term cultures.
Human stromal precursors can be sorted by their expression of membrane antigens recognized by the Stro-1 monoclonal antibody , Thy-1 , vascular cell adhesion molecule-1 , endoglin , the α1 integrin subunit , and MUC-18/CD146 . In the mouse, one may sort precursors using Sca-1 expression . Precursors can be cloned either in liquid culture (growth of colony-forming units-fibroblasts) , or in methyl-cellulose (CDCL) . Whether the stromal precursors obtained using the methods described above are equivalent to founder mesenchymal cells has yet to be demonstrated.
Recent data suggest that endothelial cells give rise to VSMCs. Some endothelial lines and primary cultures may shift from an endothelial phenotype (cells expressing von Willebrand factor [vWf]) to a VSMC phenotype (cells expressing ASMA) under the influence of transforming growth factor-β (TGF-β) or depletion of angiogenic factors [25,, 26]. Remarkably, transitional cells expressing both vWf and ASMA were observed after a few days' exposure to TGF-β . Embryonic endothelial cells from the aortic wall of birds appear to transdifferentiate into ASMA+ cells during migration to the subendothelial region [27,, 28]. Embryonic stem cells expressing the receptor-1 for vascular endothelial growth factor (VEGF) are able to differentiate into endothelial cells in response to VEGF or into VSMCs in response to platelet-derived growth factor (PDGF) . These reports suggest that endothelial cells might serve as precursors of VSMCs, at least at some anatomical sites, or that both lineages derive from a common primitive cell.
Within hematopoietic sites, mesenchymal or endothelial cells may differentiate under the autocrine influence of cytokines and ECM molecules. Such a hypothesis would fit with the model of marrow stromal cell formation advocated by Bianco et al. , where vascular invasion precedes hematopoiesis. It would also be in agreement with our observation that a “primary logette,” comprising endothelial and myoid cells, is formed in the bone marrow of the human fetus 2 weeks prior to hematopoietic colonization .
Potential Inducers of the Differentiation Pathway
Different mediators expressed and/or secreted by endothelial cells appear to play a role in the recruitment and differentiation of VSMCs [12–, 14,, 31–, 33]. The recruitment of VSMCs appears to result from the interplay between PDGF, tissue factor, and endoglin  and, as yet, unknown proteins resulting from the induction of transcription factors, such as myocyte-specific enhancer-binding factor-2C (MEF-2C) . The differentiation of VSMCs appears to be regulated by an interplay between stimulatory molecules, such as TGF-β, insulin-like growth factor, endothelin I, angiotensin II, heparan sulfate, laminin, collagen IV, and retinoids and inhibitory molecules, such as PDGF, γ-interferon, and fibronectin. In addition, cell shape plays an important role in VSMC differentiation. For example, in response to adherence and stretching on an ECM, mesenchymal cells express ASMA and other VSMC cytoskeletal proteins via a still unknown mechanism involving mitogen-activated protein (MAP) kinase . The ras-GTPase, rhoA, is the molecular switch for stress fiber formation and has been implicated as part of the shape of the cell signaling pathway .
The study of growth conditions of stromal precursors [24,, 38–, 42] has highlighted the importance of various cytokines (in particular, PDGF, epidermal growth factor, interleukins-1 and -6, tumor-necrosis factor-α, TGF-β, and fibroblast growth factor-2) and shown that cytokine effects are dependent on the origin of the stromal cells. Cytokine effects also vary depending upon the presence of serum and of adhesion molecules, whether cells are plated in liquid or semisolid medium, and whether stromal precursors are assayed on fresh bone marrow or after culture of a primary layer. Whether these cytokines play a role in stromal differentiation in addition to growth has yet to be clarified. From what is known for VSMCs, it is probable that TGF-β and PDGF modulate the stromal differentiation process in coordination with ECM molecules. TGF-β may be especially relevant since it is secreted by stromal cells , and a hypothesized mechanism for its action  would explain the correlations we observed among ASMA, thrombospondin-1, and EDa+ fibronectin expression [10,, 11]. In cultures of myofibroblasts, TGF-β appears to induce ASMA expression via an, as yet, uncharacterized EDa+ fibronectin receptor. In addition, thrombospondin-1, a molecule that also stimulates ASMA expression in fibroblasts, is a well-known activator of TGF-β. TGF-β might, therefore, be the underlying factor in a regulatory pathway of ASMA expression, involving thrombospondin-1 and EDa+ fibronectin in murine and human stromal cells [10,, 11].
Functional Significance of the Differentiation Pathway
A VSMC differentiation pathway for hematopoietic stroma may appear surprising and unrelated to known functions of VSMCs. Further analysis reveals a functional significance. Stromal cells involved in hematopoietic support synthesize a number of cytokines and adhesion molecules . Stromal cells are also very likely involved in the trafficking of hematopoietic cells since, at the sinus level, they may modify the tightness of endothelial junctions , and since stromal cells express the chemokine stromal-derived factor-1 . This dual function, the synthesis of mediators and regulation of trafficking, is reminiscent of the two functions described for VSMCs able to modulate their expression from a synthetic phenotype to a contractile phenotype . The synthetic phenotype is typical of more immature (or “dedifferentiated”) VSMCs, in particular, those found after balloon injury in animals or in the intima in humans, while the contractile phenotype characterizes fully mature cells as found in the media.
A recent report indicates that some of the intimal VSMCs developing after experimental injury in mice are blood borne, which suggests that marrow-derived cells are recruited in vascular healing as a complementary source of VSMCs . A marrow origin of these repair cells would be a straightforward explanation for the phenotypic and functional similarity between marrow stromal cells and VSCMs, since both populations might be the progeny of the same stem cell population located in the bone marrow. Such a hypothesis is consistent with the recent report of Kuznetsov et al. , wherein circulating mesenchymal precursors gave rise to cells with VSMC, osteogenic, and adipocytic features.
Stromal Cells Derive from MSCs
Investigators involved in orthopedic research have suggested the existence of bone marrow MSCs that give rise to adipocytes, chondrocytes, osteoblasts, hematopoietic support stroma, and even sarcomeric muscle [50,, 51]. A multipotential cell giving rise to adipocytes, chondrocytes, and osteoblasts has been described in human marrow , and a multipotential cell line providing osteoclast support as well has been obtained from the mouse . We have recently derived cells from these different lineages, including stroma with hematopoietic support, from Stro-1+ marrow cells . These data indicate the presence in human and murine bone marrow of a mesenchymal multipotent precursor cell. A recent work indicates the existence of MSCs in human first trimester fetal liver . Whether the fetal liver, which is the major site of HSC in vivo expansion, is also the site for expansion of MSCs, remains to be determined. Recently, Reyes et al.  reported the presence of a mesodermal stem cell in the bone marrow of humans that would give rise to endothelial cells in addition to the lineages previously described for MSCs. Such a mesodermal stem cell would therefore generate all cellular components of the marrow microenvironment (stromal cells with a VSMC differentiation, endothelial cells, adipocytes, and osteoblasts).
It is possible, from the presently available data, to draw a tree for MSC potential as shown in Figure 1. At first glance, it is very similar to the hematopoietic tree described for the HSC. In particular, a distinction is made between an MSC compartment, where major proliferation takes place, and a differentiation compartment, where cells committed to a specific lineage or lineages mature along differentiation pathways. The analysis of stromal cell differentiation potential has been aided by the discovery of specific in vitro conditions for differentiation along different mesenchymal lineages. Osteogenic factors usually used in cultures are a combination of dexamethasone (except for mouse cultures), β-glycerophosphate and ascorbic acid phosphate, while adipogenic factors combine dexamethasone, isobutyl-methylxanthine, and indomethacin. Cartilage is obtained by aggregating the cells in the presence of TGF-β. Stroma formation is observed by culturing in a medium containing a high concentration of screened horse and fetal calf serum; cellular attachment and stretching, TGF-β, and PDGF are probably essential for this differentiation pathway (see above).
However, as already indicated by Bianco et al. [1,, 2], there are major differences (outlined as follows) between the mesenchymal tree and the hematopoietic lineage that need to be emphasized.
Self-renewal capacity has yet to be demonstrated for MSCs; a feature that may be difficult to demonstrate due to the low to nil turnover under stationary conditions (Feature 2). Indeed, plasticity (Feature 3) of the mesenchymal lineage suppresses the absolute theoretical requirement for a compartment of stem cells with extensive self-renewal capacity.
Fluxes toward the different lineages are of a lower magnitude in the mesenchymal lineage. In the hematopoietic system there is a strict requirement for perennial cell renewal due to a daily loss of 109 to 1011 cells per kilogram of body weight (in humans). In the mesenchymal system, such steady-state loss is nonexistent in the adult under stationary conditions. There is turnover of bone, but the turnover in not precisely a steady-state phenomenon. While a certain amount of bone turnover can be detected in an adult skeleton at any given time, this bone turnover is a local event, involving the coordinated excavation and refilling of bone by hematopoietic lineage-derived osteoclasts and mesenchymal-derived osteoprogenitor cells. In addition, not all types of bone are turned over equally. Medullary bone is replaced at a far greater rate than is cortical bone . The localized formation of bone also occurs in response to trauma, while systemic changes in bone only occur in certain disease states, as has been demonstrated in mice with a collagen mutation leading to an osteogenesis imperfecta phenotype , in response to systemically active osteogenic factors, such as parathyroid hormone fragments , and in injuries to bone marrow [60,, 61].
Lineages from MSCs are not as strictly delineated as lineages from HSCs. It is possible, at the clonal level, to induce a switch from the adipocytic to the osteoblastic lineage  and hypertrophic chondrocytes express all of the accepted “markers” for osteoblasts, including osteocalcin, alkaline phosphatase, the formation of a mineralized matrix [63–, 65], and even the expression of the transcription factor core-binding factor A1 (Cbfa1). Moreover, certain features of the stromal cell suggest the existence of intermediate cell isoforms. Human stromal cells with VSMC differentiation express alkaline phosphatase and collagen I, considered early markers of the osteoblastic lineage. In addition, we have observed that some human stromal cells have the morphology of an adipocytic cell at one pole (lipid-filled vesicles) and a VSMC morphology at the other pole, featuring bundles of myofilaments, caveolae, and fibronexus .
These peculiar features of mesenchymal cell expression lead us to propose the stochastic Repression/Induction model described below.
Stem Cell Models
Different models arise from different conceptions of the MSC compartment. A hierarchical model of MSCs has been proposed based on the in vitro differentiation potential of human MSCs [5,, 66]. In that study, a limited number of multipotential MSCs were observed (for example, no bipotential adipogenic/chondrogenic clones were detected), and it was possible to draw a hierarchical tree that was consistent with these data. It is possible that the hierarchical model works for human-derived MSCs, but may not be applicable to other species. Another possibility is that the expansion, required for multipotential analysis in the experiments of Muraglia et al. , resulted in the selection, or loss, of some subpopulations of multiprogenitors that could not be expanded sufficiently to be assayed. Regardless, there is no hierarchical diagram that effectively delineates the differentiation potentials of mouse-derived MSCs, and a hierarchical diagram does not account for plasticity and the functional overlap that exists between lineages , such as adipocytic and/or osteoblastic and stromal lineages.
The stochastic Repression/Induction model (Fig. 2) would account for the plasticity observed for the mesenchymal lineages. Following a hypothesis made for HSCs , MSCs would be characterized by a “molecular ground state” of mesenchymal differentiation potential. This “molecular ground state” could arise from a series of gene silencing events occurring during development, which would result in a collection of stem cells that differ in the number of nonsilenced genes available for activation. Gene silencing events could occur through various mechanisms, including DNA methylation, deacetylation, modifications in chromatin structure, and other mechanisms that are still hypothetical . The mesenchymal stem cells would proliferate and undergo change in gene expression with, at different steps, either repression or activation of genes (we consider here four sets of genes: stromagenic, osteogenic, chondrogenic, and adipogenic). We propose: A) that the repression events are stochastic, so repression does not follow a defined lineage pathway, and B) that activation, also a stochastic event, does not mean an activated cell must lose potential for other phenotypes; the repression and activation steps are distinct and not mutually exclusive. Many pathways would be possible, and the nature of the committed (determined) cells entering the differentiation compartment would reflect what has occurred in the MSC compartment (Fig. 2). For example, a linear pathway of progressive repression would lead to a cell committed to a single, nonrepressed lineage. On the contrary, pathways with several repression/activation steps would lead to cells committed to several lineages, i.e., cells with several ongoing differentiation programs expressed simultaneously or sequentially. While these genetic events, involving transcription factors, would be foremost for MSCs in the MSC compartment, within the Differentiation compartment, external factors (hormones, cytokines, or adhesion molecules, see above) would regulate the shift between lineages for cells committed to several lineages.
Activation of progenitor cells along a particular phenotypic pathway might be due to the expression of a master regulatory gene. The concept of master regulatory gene was first brought to light by the identification of MyoD, a muscle transcription factor capable of inducing the expression of a bank of muscle-specific genes . Subsequent studies have shown that myogenic differentiation is regulated by a set of transcription factors that includes MyoD, Myf5, myogenin, and MRF4. Candidates for master genes for other mesenchymal lineages have been identified, such as peroxisome proliferator-activated receptor-gamma 2 (PPAR-γ2) for adipocytes and Cbfa1 for osteoblasts. PPAR-γ2 was shown to be sufficient for induction of adipogenesis in fibroblasts . However, recent results indicate that, like for muscle, the regulation of adipogenesis would be under the control of multiple transcription factors, including CCAAT-enhancer binding protein-alpha/beta (C/EBP-α/β), retinoic X receptor-alpha , and these, in turn, may be under repressive control of yet another gene, wnt-10b . Cbfa1 was shown to be necessary for bone development in knockout mice, and expression of Cbfa1 alone is sufficient to induce osteogenesis in nonosteogenic cells [74,, 75]. While these results implicate Cbfa1 as a likely candidate as the master gene for osteogenesis, there must be other regulatory factors at play, since Cbfa1 is also expressed in some nonosteogenic cells, such as nonosteogenic bone marrow-derived cell lines .
Lineage repression appears to be a common regulatory mechanism of mesenchymal lineage progression. One example of a regulatory factor that modulates differentiation via repression is nuclear factor of activated T cells (NFAT-p). In NFAT-p knockout mice, spontaneous expression of chondrogenesis was observed in extraarticular connective tissue . In that study, a detailed examination of NFAT-p expression in chondrocytes and MSCs showed that chondrogenic induction correlates with NFAT-p repression, and that in chondrocytes, NFAT-p overexpression can extinguish chondrogenesis. In a study of adipocyte differentiation, Jaiswal et al.  showed that inhibition of MAP kinase expression in human MSCs cultured in osteogenic medium leads to an increased expression of adipocytes, including the upregulation of the expression of the adipocyte transcription factor PPAR-γ2; in another study , stable transfection of the PPAR-γ 2 gene was shown to inhibit Cbfa1 expression in osteogenic cells and repress osteoblastic expression. Another example is the repressive role of wnt-10b  and GATA-2 and GATA-3 in adipogenesis .
Thus, for the induction of mesenchymal progenitor cells, no simple set of individual master genes for each mesenchymal lineage has been identified that can, alone, account for differentiation along all mesenchymal pathways. Rather, MSC differentiation is under the influence of multiple inductive and repressive factors (nuclear factors shown in Table 1) that, often simultaneously, can lead to the expression of “end-stage” markers for more than one lineage. The stochastic Repression/Induction model appears, therefore, to be a paradigm for mapping the complex interactions of factors that regulate the progression of MSCs toward the expression of a mature, yet plastic, mesenchymal phenotype.
Table Table 1.. Examples of activating or repressing nuclear factors of mesenchymal lineages
*By analogy with factors involved in vascular smooth muscle cell differentiation/recruitment (see text).
Plasticity, a major characteristic of the MSC system, is best explained by a stochastic model where the plastic nature of the committed cells entering the differentiation compartment would be the result of stochastic activation and repression events affecting transcription factor genes within MSCs. Recent data suggest that plasticity should even be extended to nonmesenchymal lineages of neuroectodermal (neurons, astrocytes, and oligodendrocytes) [82,, 83] or endodermal (hepatocytes)  origin. Such plasticity would remove the theoretical need for stem cells with indefinite self-renewal capacity, since some cells in the progeny of the mother cell would express, anew, genes that were transitorily suppressed. Therefore, the MSC system would be distinct from other classic stem cell systems (hematopoietic, intestinal, or epidermal), wherein the existence of a stem cell compartment is the required correlate to the irreversible differentiation of the progeny.
Once MSCs are determined (or committed) toward one or several lineages, differentiation programs would proceed under the control of external factors (cytokines, hormones, adhesion molecules). Hence, MSCs would give rise to the different populations of the hematopoietic microenvironment (stromal cells, endothelial cells, osteoblasts, adipocytes). Stromal cells follow a VSMC differentiation program that has multiple steps. The VSMC differentiation program may be interrupted at multiple points along this pathway, thus conferring heterogeneity to the stromal compartment. The VSMC differentiation of the stroma may explain its dual functional role, that of hematopoietic support and control of trafficking.
Work supported by grants from: Association pour la Recherche sur le Cancer (ARC No. 5837), Association pour la Recherche sur les Myopathies/Institut National de la Santé et de la Recherche Médicale (AFM/INSERM No. 4CS02F), and grants from the National Institutes of Health.