Functional Neuronal Differentiation of Bone Marrow-Derived Mesenchymal Stem Cells



Recent results have shown the ability of bone marrow cells to migrate in the brain and to acquire neuronal or glial characteristics. In vitro, bone marrow-derived MSCs can be induced by chemical compounds to express markers of these lineages. In an effort to set up a mouse model of such differentiation, we addressed the neuronal potentiality of mouse MSCs (mMSCs) that we recently purified. These cells expressed nestin, a specific marker of neural progenitors. Under differentiating conditions, mMSCs display a distinct neuronal shape and express neuronal markers NF-L (neurofilament-light, or neurofilament 70 kDa) and class III β-tubulin. Moreover, differentiated mMSCs acquire neuron-like functions characterized by a cytosolic calcium rise in response to various specific neuronal activators. Finally, we further demonstrated for the first time that clonal mMSCs and their progeny are competent to differentiate along the neuronal pathway, demonstrating that these bone marrow-derived stem cells share characteristics of widely multipotent stem cells unrestricted to mesenchymal differentiation pathways.


Classic dogmas restrict stem cell differentiation potentialities to lineages that are specific to their tissue of origin. However, numerous recent publications reported evidence that bone marrow-derived cells do not strictly observe this restriction. Notably, they can give rise to brain cells such as neurons of various areas (hippocampus, cortex, cerebellum, olfactory bulb, etc.) after in vivo transplantation, in both mouse and human [1, 2]. Such data suggest the presence in bone marrow grafts of NSCs or progenitors able to migrate from the blood into the brain parenchyma across the blood-brain barrier. Those results open up new enthusiastic perspectives in cell plasticity and cell therapy of the nervous system. The hypothesis of neurogenic stem cells in bone marrow has been challenged by convincing results on cell fusion between graft-derived cells and cerebellum Purkinje cells [3, 4]. Nevertheless, cell fusion has not been demonstrated in other parts of the brain where neuronal differentiation of bone marrow cells occurs. Others reported the failure to obtain any “transdifferentiation” of bone marrow-derived cells [5, 6], provoking extensive discussions on the meaning of the results [7, [8]–9]. This may suggest that such occurrence strongly depends on experimental conditions and on important unidentified factors. Alternatively, another hypothesis could be that hematopoietic stem cells (the presence of which is tested by grafting experiments) cannot differentiate along the neuronal pathway and thus this differentiation may involve another stem cell population housed in the bone marrow such as MSCs.

MSCs derived from adult bone marrow were first described as able to differentiate along three main pathways: osteoblastic, adipocytic, and chondrocytic pathways [10]. Beside these basic potentialities, MSCs may also be able to acquire phenotypes of hematopoietic-supportive cells [11], muscle cells [12, 13], and tenocytes [14], suggesting that these MSCs are widely multipotent but restricted to mesoderm-derived lineages. This basic premise is now challenged by results showing that MSCs can be induced to express neuronal markers in vitro. To this end, various chemical inductors such as butylhydroxytoluene, butylhydroxyanisol (BHA), dimethyl sulfoxide (DMSO), 2-mercaptoethanol, 3-isobutyl-1-methylxanthine (IBMX), 5-aza-cytidine [15, 16] were used alone or in combination. However, differentiation by means of chemical inductors alone or in combination with cytokines should be considered with caution. In some of these studies, differentiation was initiated by a 1-day first step using basic fibroblast growth factor (bFGF) as a preinductive molecule. bFGF was actually previously described as controlling NSC renewal and differentiation [17]. Similarly, multipotent adult progenitors cells (MAPCs) purified from rat or mouse bone marrow can follow neuronal and glial fates under the influence of bFGF [18]. In the mouse model, MAPCs and MSCs cannot be strictly compared in their potentials, because mouse MSCs (mMSCs) were only recently amplified by us and others [19, 20] and MAPCs are not widely available. Therefore, the ability of mMSCs to acquire a neural phenotype in vitro has not been investigated to date. It would be of great interest to firmly establish the presence of neural progenitors in the mammalian bone marrow before investigating their in vivo fate in genetically or chemically induced mouse models of human pathologies.

The major aim of this work was to investigate the ability of mMSCs to differentiate in vitro toward functional neuronal cells in response to a treatment involving bFGF as a physiologic inducer. mMSCs were shown to express nestin, a marker for NSCs in embryo and adult. Neuronal differentiation was characterized by two parallel and complementary means: first, the expression of specific markers and, second, the acquirement of some neuron functions. This dichotomic approach was crucial to fully and unambiguously demonstrate the neuronal differentiation of mMSCs. After induction, the majority of mMSCs adopted a neuron-like morphology in parallel to the expression of neuronal markers NF-L (neurofilament-light, or neurofilament 70 kDa) and class III β-tubulin (β3-tub). This expression of neuronal markers is combined with a functional acquisition of neuronal properties such as an increase of cytosolic calcium concentration in response to various specific neuronal agonists. Our results firmly demonstrate that mMSCs can differentiate along the neuronal pathways toward a functional phenotype. Moreover, we also provide evidence that clonal mMSCs can differentiate along this pathway, demonstrating that these bone marrow-derived stem cells exhibit wide potentialities and are not restricted to differentiate into cell types of mesodermal origin.

Materials and Methods

Culture and Neuronal Differentiation of MSCs

mMSCs were isolated and cultured as previously described [21]. Cells were cultured between 10 and 20 passages. No significant differences in either renewal or differentiation abilities were observed between early and late passages. Cells were retrieved from subconfluent culture by trypsinization (trypsine/EDTA solution; Invitrogen, Carlsbad, CA,, counted, and plated to a density of 3,000 cells per cm2 on poly(lysine)-coated plates. Plates were previously coated overnight with a 10 μg/ml poly(lysine) (Sigma-Aldrich, St. Louis, solution in phosphate-buffered saline (PBS) (Cambrex Bio Science Verviers S.p.r.l., Verviers, Belgium, Cells were subsequently cultured for 7 days in the amplification medium, Dulbecco's modified Eagle's medium-low glucose (Sigma-Aldrich), glutamine 2 mM (Cambrex Bio Science Verviers S.p.r.l.), penicillin/streptomycine (50 U/ml and 50 mg/ml, respectively; Cambrex Bio Science Verviers S.p.r.l.), and fetal calf serum 10% (Sigma-Aldrich) supplemented with bFGF 25 ng/ml (either from R&D Systems Inc., Minneapolis, or Peprotech, Rocky Hill, NJ, For immunofluorescence analysis of markers of expression, mMSCs were plated on poly(lysine)-coated glass or plastic culture slides in the same inductive conditions.

Isolation, Amplification, and Characterization of Monoclonal mMSCs

A cell suspension was retrieved from subconfluent culture of the initial mMSC population by trypsinization and maintained in complete medium supplemented with HEPES 20 mM. Cells were isolated and distributed in a 96-well plate by means of microscopic examination and micromanipulation. Each well was then independently checked for the presence of a solitary cell. After amplification, a part of each cell population was frozen, and another part was tested for differentiation along mesodermal (osteoblastic, adipocytic, and chondrocytic [21]) and neuronal pathways (see above).

RNA Extraction and Analysis

Total RNA was extracted and analyzed as previously described [19]. Primer sequences were as follows: NF-L [21] (forward CCAGGAAGAGCAGACAGAGGT, reverse GTTGGGAATAGGGCTCAATCT, fragment size 302 base pairs [bp]); Nestin primers were adapted from [22] according to a recently validated sequence (GenBank accession number NM_016701) (forward GGAGAGTCGCTTAGAGGTGC, reverse TCAGGAAAGCCAAGAGAAGC, fragment size 327 bp). Hypoxanthine-guanine phosphoribosyltransferase (HPRT) primers were as previously published [19]. All primers were selected in different exons.

Immunofluorescence Analysis

Primary antibodies included mouse anti-nestin (dilution 1:250; BD Pharmingen, San Diego,, mouse anti-β3-tub (dilution 1:500; BAbCO, Richmond, CA,), and mouse anti-NF-L (dilution 1:500; DAKO, Glostrup, Denmark, Donkey anti-mouse secondary antibodies coupled to cyanine 3 were provided by Jackson ImmunoResearch Laboratories (West Grove, PA, (dilution 1:1,000). A nonspecific signal was evaluated as negligible in all experiments using anti-β-galactosidase monoclonal antibody as a control (Promega, Madison, WI, All antibodies were diluted in PBS supplemented with bovine serum albumin (BSA) 3% and donkey serum 2%. Untreated or bFGF-treated mMSCs were rinsed twice with PBS and then fixed 20 minutes in a paraformaldehyde 4% solution in PBS. Slides were treated with a blocking/permeabilizing solution consisting of PBS supplemented with BSA 3%, donkey serum 2%, and Triton X-100 0.3% for at least 20 minutes at room temperature. Slides were then sequentially incubated with primary antibodies overnight at 4°C and with secondary antibodies for 1 hour at room temperature. Cells were finally incubated in a PBS solution containing 1 μg/ml 4,6-diamidino-2-phenylindole to stain nuclear DNA. Washed slides were then mounted with FluorSave (Calbiochem, San Diego, and analyzed using an epifluorescence microscope (Olympus, Tokyo,

Calcium Imaging

Cells were loaded with the calcium indicator dye Fluo-4 by bath application for 30 minutes at 37°C in ACSF-HEPES (artificial cerebral spinal fluid with HEPES, which contained [in mM]: 119 NaCl, 2.5 KCl, 1.3 MgCl2, 2.5 CaCl2, 1 NaH2PO4, 26.2 NaHCO3, 11 d-glucose, 10 HEPES, pH 7.4) containing 5 μM Fluo-4 AM (Molecular Probes Inc., Eugene, OR, and 0.1% DMSO (Sigma-Aldrich). Cells were then placed at room temperature on the stage of an upright compound microscope (Eclipse 600 FN; Nikon Corporation, Tokyo, equipped with a water immersion ×40 objective (numerical aperture 0.8) and a confocal head (PCM 2000; Nikon Corporation). Drugs were perfused in the bath by a multivalve process (VC-6M; Harvard Apparatus, Holliston, MA, Cells were stimulated with 488-nm wavelength light from an argon laser. Emission light was filtered by a 515 ± 15-nm filter. Data acquisition and calcium measurement were performed with the EZ 2000 software (Nikon Corporation). Each frame in a time-lapse sequence was captured every 2 seconds. For each series of images, F0 refers to the mean fluorescence intensity measured throughout all the regions of interest. Relative changes in [Ca2+] over time are expressed as F/F0. The percentage of activated cells is calculated as the number of cells displaying a significant agonist-induced fluorescence intensity increase (>20% F/F0 increase upon application) divided by the total number of cells detected [23].

Statistical Analysis

Data were then exported to Microsoft Excel (Microsoft, Redmond, WA, or SigmaPlot (Systat Software Inc., San Jose, CA, softwares for statistical analysis. All results are presented as mean ± SEM. Statistical tests were performed using either Student's t test or Mann-Whitney rank sum test as appropriate.


mMSCs Acquire a Neuron-Like Shape in the Presence of bFGF and Poly(lysine)

We tested several culture conditions to induce neuronal differentiation of mMSCs. In our hands, previously published protocols using chemical inducers were not able to induce a neuron-like shape in mMSCs, possibly due to species differences (data not shown). We then investigated the efficiency of physiological inducers of neuronal differentiation. We first tested the efficiency of bFGF because of its well-known role in neuron differentiation and physiology. In preliminary experiments, bFGF was added to the culture medium of mMSCs spread on fibronectin-coated dishes but this gave poor results at the morphological level (data not shown), in contrast to the reported effect on mouse MAPCs [18]. Because neuronal cells are classically cultured on poly(lysine)-coated dishes, we hypothesized that mMSC neuronal differentiation may be improved by this way. When plated on poly(lysine), mMSCs showed a smaller size than on fibronectin but their fibroblastoid shape was maintained (Fig. 1A). When bFGF was added to the medium, some short neurite-like extensions were visible after 2 days, easily recognizable after 4 days, and fully developed after 1 week (Fig. 1A). At this time, most cells presented with neuron-like morphology, including a small-cell body and long neurite-like extensions, sometimes a unique long extension and others shorter (Fig. 1B) or bipolar morphologies. From day 7, cells readopted a fibroblastoid shape, suggesting that such conditions could not support long-term cell differentiation.

Figure Figure 1..

Mouse mesenchymal stem cells (mMSCs) can progressively acquire a neuron-like morphology. (A): mMSCs plated on poly(lysine) were treated with basic fibroblast growth factor. Shape change is illustrated by phase-contrast micrographs of various fields at days 0 (D0), 2 (D2), 4 (D4), and 7 (D7). (B): Cells progressively developed neurite-like extensions, with a maximum size at day 7. Some cells presented a unique long extension (arrows) from the small-cell body (arrowheads) as shown in this image of a May-Grunwald Giemsa stained field. Scale bars = 100 μm.

Treatment Increases the Number of Nestin-Positive mMSCs

Nestin is widely considered a specific marker of NSCs and progenitors [24, [25], [26]–27]. Thus, nestin expression was investigated in untreated and bFGF-treated mMSCs as a primary marker that would suggest real neuronal potentiality in this cell population. Reverse transcription-polymerase chain reaction (RT-PCR) analysis showed that the nestin-encoding gene was actively transcribed in both populations at a similar level (Fig. 2A). Our results and those of others [27] suggest that nestin expression is a common feature of mMSCs. Immunolabeling analysis further demonstrated that 50.3% ± 8.2% (mean ± SEM; n = 3) of untreated cells expressed the marker (Fig. 2B). bFGF treatment dramatically increased the number of nestin-positive cells to 94.5% ± 3.4% (mean ± SEM, three independent experiments; p ≤ .02; Fig. 2C). This may be due to the growth of a neural progenitor subpopulation because our culture is heterogeneous in origin or to the commitment of multipotent mMSCs toward a neural fate. Alternatively, nestin could be expressed by all resting cells, but below the detection threshold in 50% of them, and then increased by bFGF treatment. Whatever the mechanism involved, the presence of nestin-positive cells in untreated mMSCs was a clue suggesting that those cells carried a true neuronal potentiality ready to be activated under appropriate conditions.

Figure Figure 2..

Mouse mesenchymal stem cells (mMSCs) expressed nestin before and after treatment. (A): Reverse transcription-polymerase chain reaction analysis of the expression of nestin in amplified subconfluent mMSCs (lane 1: culture on fibronectin in the amplification medium [19]) and treated mMSCs (lane 2: cells cultured for 1 week on poly-lysine in the amplification medium supplemented with basic fibroblast growth factor 25 ng/ml, as described in Materials and Methods). HPRT amplification is shown as a control of mRNA quality. Immunofluorescent analysis of nestin expression without (B) and after (C) treatment of mMSCs. Nestin-specific antibodies labeled a filamentous network in both cells, but with a stronger signal in treated cells, may be due to the concentration of filaments in thin extensions. Scale bars = 100 μm. Abbreviation: HPRT, hypoxanthine-guanine phosphoribosyltransferase.

Treatment Induces Expression of Other Neuronal Markers NF-L and β3-Tub in mMSCs

Nevertheless, shape change and nestin expression may appear as coincidental and without links to a neural phenotype. We therefore attempted to correlate these results with the induction of neuronal marker expression. Neuronal cells are classically characterized by the expression of cytoskeletal proteins such as NF-L and β3-tub, which are both early markers of this lineage. NF-L mRNA was below detection thresholds of our RT-PCR test in undifferentiated mMSCs but was easily amplified from treated cell samples (Fig. 3A). Immunofluorescence analysis showed that the rate of NF-L-positive cells increased from 12.4% ± 11.7% (mean ± SEM, n = 3; Fig. 3B) in the mMSC population to 88.5% ± 1.8% in treated culture (mean ± SEM, n = 4; p ≤ .02; Fig. 3C). NF-L proteins were detected in dotted structures [28] distributed around the nucleus in most cells or in the whole cell body, including some neurite extensions (Fig. 3D). In most cells, NF-L expression was weak in neurites, perhaps because of its phosphorylation status [29]. In contrast to NF-L, β3-tub-specific labeling was found widely distributed in the cell cytoplasm. The proportion of β3-tub-positive cells was also dramatically induced by the treatment, from 2% ± 2% (mean ± SEM, n = 4; Fig. 3E) in mMSCs culture to 65% ± 10.9% (mean ± SEM, n = 4; p ≤ .01; Fig. 3F). Our results suggest the coexpression of nestin and β3-tub in approximately 60% of treated cells, characteristic of an immature neuronal phenotype [30]. Therefore, the treatment engages the cells in the neuronal pathway but does not support full differentiation. No expression of astrocytic and oligodendrocytic markers, respectively, glial fibrillary acidic protein (GFAP), and myelin basic protein, was detected using RT-PCR (N. Platet, unpublished data), supporting the idea that the treatment induced a coordinated genetic program and not a stochastic activation of irrelevant genes. Nevertheless, those results do not rule out the possibility of glial differentiation of mMSCs in other conditions.

Figure Figure 3..

Dramatic induction of major neuronal markers in mouse mesenchymal stem cells (mMSCs) treated with basic fibroblast growth factor. (A): Reverse transcription-polymerase chain reaction analysis of the expression of NF-L in amplified (lane 1) and treated mMSCs (lane 2) showed a strong induction of the gene in cells after treatment. HPRT amplification is shown as a control of mRNA quality. (B–D): Immunofluorescent analysis of NF-L before (B) and after (C, D) treatment of mMSCs. NF-L-specific antibodies labeled punctate structures mainly in the perikaryon zone (C). (D): A higher magnification of bipolar cells presenting NF-L-positive dots distributed both in the cell body and in neurite-like extensions. (E, F): Immunofluorescent analysis of β3-tub expression before (E) and after (F) treatment of mMSCs. β3-tub was found widely distributed in the whole cell, without any preferential localization. Scale bars = 100 μm. Abbreviations: β3-tub, class III β-tubulin; HPRT, hypoxanthine-guanine phosphoribosyltransferase; NF-L, neurofilament-light, or neurofilament 70 kDa.

Treated Cells Can Respond to Classic Neuron Activators by Increasing Their Cytosolic Calcium

To confirm that expression of neuronal markers by mMSCs is the result of a specific program, we investigated the functionality of treated cells by imaging calcium signaling in response to various neuron activators: (a) glutamate, which is able to induce Ca2+ rise through ionotropic or metabotropic receptors activation (reviewed in [31]), (b) veratridine, which is a specific voltage-gated Na+ channel agonist [32] leading to Ca2+ entry via the involvement of Na+/Ca2+ exchanger [33] or voltage-dependent Ca2+ channels [34], and (c) dopamine, which can directly induce a Ca2+ increase [35, 36] or act as a neuromodulator [37]. Differentiated mMSCs responded to 100 μM glutamate (Fig. 4A) and 50 μM veratridine (Fig. 4B) as observed by a large and immediate calcium increase in 51.6% ± 12% (mean ± SEM; n = 6; p ≤ .01) and 68.4% ± 11.6% of cells (mean ± SEM; n = 4; p ≤ .03), respectively. These effects were blocked by 100 μM MK801 (specific N-methyl-d-aspartic acid receptor antagonist) and 1 μM tetrodotoxin (specific Na+ channel antagonist), respectively, confirming the nature of activated receptors (J.-C. Platel, unpublished data). Under the same conditions, undifferentiated mMSCs did not respond to these activators regarding calcium movement. In differentiated mMSCs, dopamine induced a Ca2+ response in 53.7% ± 10.4% of cells (mean ± SEM; n = 3; p ≤ .05) within minutes after agonist addition (Fig. 4C). Response to dopamine was also punctually observed in control cells (approximately 30% responsive cells in one out of three experiments), possibly revealing spontaneous differentiation. These results suggest that dopamine acts as a neuromodulator in differentiated mMSCs. All these results demonstrate that treated mMSCs expressed functional neuronal receptors and voltage-dependent channels. Thus, bFGF-treated mMSCs clearly displayed a neuronal phenotype.

Figure Figure 4..

Classic neuron agonists induce a cytosolic calcium response in mouse mesenchymal stem cells treated with basic fibroblast growth factor. (A–C): Variation of the cytosolic calcium concentration is displayed in the bottom panel as a function of time for the cell(s) pinpointed (arrowheads) in the microphotographs above. Glutamate 100 μM (A) and veratridine 50 μM (B) induced an immediate increase in the cytosolic calcium concentration. Moment of addition is indicated by an arrow. (C): Imaging of calcium response to dopamine 100 μM. Dopamine was added at time = 0. Responses displayed by two characteristic cells (1 and 2, indicated by arrowheads) were shown as an example of dopamine effects. Classic response to this agonist is delayed in time and not synchronous between cells of a same field. Scale bars = 50 μm. Abbreviations: s, seconds; t, time.

Clonally Derived MSCs Can Differentiate Along the Neuronal Pathway

To determine which mMSC subpopulation can differentiate along the neuronal pathway, we performed experiments on clonally derived cultures. We established 17 clonal populations and first tested their ability to differentiate along the three pathways that define MSCs: osteoblastic, adipocytic, and chondrocytic (Table 1). Three clones out of 17 were able to differentiate along all three pathways, demonstrating that they derived from clonal MSCs. Then, we analyzed the expression of neuronal markers in such populations with and without neurogenic induction (Table 1). Before differentiation, all clones were positive for nestin and negative for NF-L. A subset (9 of 17) was also positive for β3-tub. After bFGF treatment, nestin and β3-tub expression was maintained or increased in all clones in which they were already expressed. Treatment also induced de novo expression of NF-L in all clones (Table 1). All three MSC-derived clones (so-called A+O+C in Table 1) were found strongly positive for nestin and NF-L after neurogenic treatment, but only one was weakly positive for β3-tub. This suggests that the treatment acted more efficiently on engaged progenitors than on stem cells. Nevertheless, these results clearly demonstrate for the first time that fully characterized clonally derived MSCs and their progeny can differentiate along the neuronal pathway.

Table Table 1.. Differentiation potentialities of clonally derived mesenchymal stem cells
  1. The absolute and relative numbers of clones with the differentiation potentialities described in the Phenotype column are given in the Ratio column. For every phenotype, the number of clones positive for each marker in the absence and in the presence of the inductive medium is given. After induction, the clones were also all positive for at least two neuronal markers.

  2. Abbreviations: A, adipogenic; C, chondrogenic; O, osteogenic. NF-L, neurofilament-light, or neurofilament 70 kDa.

original image


With the increasing amount of data about gene expression, few previously described differentiation markers remain. To demonstrate the neuronal differentiation of mMSCs, we adopted a “triangulation” strategy involving both marker and function analysis. If nestin is no longer a NSC marker (see above), NF-L and β3-tub expression seems to be limited to neuronal cells. Many cell types express glutamate and dopamine receptors [38, [39]–40], but Ca2+ signaling in response to either of these agonists outside the nervous system has not been addressed to date. Astrocytes, which express glutamate receptors, can elicit such a response, but astrocytic identity of bFGF-treated mMSCs has been definitively ruled out by negative RT-PCR amplification of GFAP. Moreover, glial cells do not express any Na+ channels and could not respond to veratridine. Indeed, differentiated mMSCs respond to voltage-gated Na+ channel activation by increasing their cytoplasmic calcium concentration, demonstrating that those cells belong to an excitable cell type, neuron or muscle cells [32]. Nevertheless, muscles cannot respond to glutamate, and only neurons express β3-tub or NF-L. Altogether, our results clearly demonstrate that mMSCs can acquire a functional neuronal phenotype with early-stage differentiation characteristics.

In the field of neuroscience, nestin has long been considered a specific marker of NSCs in the developing and adult brain. In contrast, publications reported nestin expression in testis [41, 42], developing eye [43], hair follicle [44], skeletal muscle [45, [46]–47], cardiac muscle [48, 49], and pancreas [50, [51]–52] in normal, embryonic, or pathologic situations. Precise identification of nestin-expressing cells has not been systematically performed in those tissues except in a few of them, such as pancreas, in which contradictory results were reported. Interestingly, nestin-positive cells from pancreas, testis, and heart share multipotentiality as a common property, suggesting that it could be a marker for stem cells or a clue for plasticity. To confirm this issue, it would be interesting to address the expression of nestin in native mMSCs, before any amplification. In bone marrow, nestin is faintly expressed by the Lin CD45+ Sca-1+ cell fraction [53], to which mMSCs belong [27, 54], suggesting that the latter express this marker in vivo. In contrast to the poor insight into native MSC nestin expression, it is now firmly established that MSCs amplified in vitro express this marker [27, 55–61, [56], [57], [58], [59], [60], [61]], even if this expression may be acquired after an in vitro maturation (i.e., later passages) [62]. Moreover, Wislet-Gendebien and coworkers [62] underlined the inhibitory effect of serum that we and others did not observe [27, 55]. Because of both the slow growth of mMSCs and the contamination by hematopoietic cells in low-passage cultures [19], we have not been able to investigate nestin expression in either unamplified mMSCs or early passages. In the absence of any protocol to purify homogeneous MSC population from bone marrow, the demonstration of nestin expression by native MSCs remains an unsolved issue.

bFGF has long been known as a key regulator of NSC proliferation and differentiation [17]. Its in vivo importance for neurogenesis has been confirmed by the neural defects observed in bFGF-deficient mice [63]. So far, bFGF has also been extensively tested to improve MSC and stromal bone marrow progenitor culture or differentiation with contrasting results. This cytokine is alternatively described as a mitotic activator [64], a colony forming units-fibroblast surviving factor [65, 66], an activator of either osteoblastic [67, 68] or adipocytic [69] differentiations, and alternatively as inhibitor or inducer of neuronal differentiation [57, 68, 70, 71]. Such discrepancies in bFGF effects may be due to differences in MSC amplification procedures that influence cell behaviors in response to cytokine [72]. In our hands, bFGF induced mMSCs to differentiate toward the neuronal pathway when they were plated onto poly(lysine)-coated culture dishes. To our knowledge, poly(lysine) receptor(s) has (have) not been identified but may participate in cell fate by modulating the bFGF signaling and/or by transducing a differentiating message on its own. It appears in our experience that both messages are necessary for efficiently inducing morphological changes. This may nicely illustrate synergies between growth and adhesive factors in regulation of cell fate but requires further experiments to specifically highlight the molecular mechanisms involved.

The neuronal differentiation of rat or human MSCs was often successfully induced using chemical compounds such as BHA, IBMX, DMSO, or 2-mercaptoethanol. Whether such compounds may induce major perturbations of cell functions is a poorly investigated topic. One of these treatments, mixing β-mercaptoethanol, DMSO, BHA, and other chemicals, clearly increased the number of dead cells in the culture [15]. In contrast, and in the same reports, authors investigated the effect of IBMX and dbcAMP cotreatment without noting a significant effect on cell viability. Nevertheless, if expression of neuronal markers was not strictly linked to cell death, they could be relevant as a symptom of cell stress. In accordance with this hypothesis, Lu and colleagues [73] showed that an artifactual neuronal differentiation could be induced in MSCs by a wide variety of compounds as a stereotypical response to stress. The neuronal morphology previously observed was due to the cytoplasm shrinkage and not to the growth of neurites. In our case, mMSCs plated on poly(lysine) displayed a smaller body size than on fibronectin, and neurite-like structures were then clearly observed growing day after day only in presence of bFGF (Fig. 1). Moreover, in contrast to the results of Lu et al., we confirmed the de novo expression of NF-L by means of RT-PCR and the acquirement of neurons' functional properties by using calcium imaging. Conjunction of structural (expression of markers) and functional (ability to efficiently respond to neuroagonists) evidence allows us to exclude the idea of a nonspecific response but not to definitively rule out the “stress” hypothesis. In that hypothesis, our data actually suggest that such observations would be the expression of a very specific and regulated cell answer using neuron markers and functions in mesenchymal cells, which has not been suggested to date.

Calcium imaging and immunostaining experiments showed that, in bFGF-treated culture, at least 50% of polyclonal mMSCs could differentiate toward the neuronal lineage. As the number of cells clearly increased with time, this phenomenon could result either (a) from the growth of a neurogenic population or (b) from the commitment to the neuronal lineage of individual mMSCs. To distinguish between these two hypotheses, we established and characterized the differentiation potentials of clonal mMSC populations. MSCs were first characterized at the clonal level by Pittenger and co-workers in 1999 [10] as bone marrow cells with wide differentiation potentialities toward at least three mesodermal lineages. To our knowledge, we present herein the first evidence that fully characterized clonal mMSCs have clear neurogenic potentialities. Because of the ectodermal origin of the neuronal lineage, we anticipated that in a hierarchical model the neurogenicity would be lost and progenitors restricted to mesodermal potentialities. Surprisingly, we showed that even cells restricted to only one mesodermal lineage could also express neuronal markers. One hypothesis emphasizes that the mesoderm-ectoderm distinction is not valid in the present instance. MSCs should not be defined as mesodermal stem cells but as multipotent stem cells that resemble neural crest stem cells in their potentialities [74, 75]. The presence of such neural crest-originated stem cells in this tissue has never been suggested or demonstrated, but MSCs might be multipotent stem cells surviving in adulthood with potentialities partly unrelated to the host tissue (i.e., bone marrow) [62].

Despite the presence of functional nerve fibers [76, [77], [78], [79], [80]–81], neuronal cell bodies were never detected in bone marrow, suggesting that there is no need for neuronal regeneration in the bone marrow. Therefore, multipotent stem cells may be resident in adult organs without the need for some of their potentialities, suggesting that some adult cells conserved an indivisible “stem cell package.” The potential given by such a package may be amplified by in vitro conditions as was suggested for NSCs [82, 83]. Our results might suggest not only neuronal potentiality per se but plasticity (i.e., cell reprogramming ability). Multipotent stem cells may be considered the result of a cellular function (plasticity) able to change the cell fate in response to particular conditions, by means of transdifferentiation or dedifferentiation and redifferentiation processes [84, 85]. In this model, plasticity should be another distinctive property of stem cells with a controversial relationship with self-renewal and multipotentiality. Nevertheless, the conditions and molecular mechanisms of plasticity should be carefully studied in the future, leading to an improvement in adult stem cells manipulation.


Our results firmly establish the possibility of functional neuronal differentiation of mMSCs. This in vitro model will allow us to address molecular events inducing and controlling this differentiation. The firm establishment of a mouse model will further allow us to accumulate insights into in vivo neurogenicity of MSCs by transplantation into the brain. This new model with the disposability of hundreds of transgenic mouse strains opens up completely new fields for investigating MSC physiology and potentialities. From a clinical point of view, further investigations should also be initiated on physiological inducers of the neuronal differentiation of human MSCs. In this context, whether such a potentiality allows differentiation after in vivo transplantation in the brain will have to be clarified in the mouse model and others before the elaboration of any bone marrow-based cell therapies of nervous system.


The authors indicate no potential conflicts of interest.


We thank Dr. Michel Villaz for his crucial help for the completion of this work and Dr. Keyoumars Ashkan, Wesley J. Harrison, and Dr. Graham W. Neill for reading and corrections. This work was supported by Association Française contre les Myopathies. M.A. is currently affiliated with Dynamique des Réseaux Neuronaux, INSERM U704, Université Joseph Fourier, Grenoble, France. J.C.P. is currently affiliated with the Department of Neurosurgery, Yale University School of Medicine, New Haven, Connecticut, USA.