Mesenchymal Stem Cells Spontaneously Express Neural Proteins in Culture and Are Neurogenic after Transplantation
Reports of neural transdifferentiation of mesenchymal stem cells (MSCs) suggest the possibility that these cells may serve as a source for stem cell–based regenerative medicine to treat neurological disorders. However, some recent studies controvert previous reports of MSC neurogenecity. In the current study, we evaluate the neural differentiation potential of mouse bone marrow–derived MSCs. Surprisingly, we found that MSCs spontaneously express certain neuronal phenotype markers in culture, in the absence of specialized induction reagents. A previously published neural induction protocol that elevates cytoplasmic cyclic AMP does not upregulate neuron-specific protein expression significantly in MSCs but does significantly increase expression of the astrocyte-specific glial fibrillary acidic protein. Finally, when grafted into the lateral ventricles of neonatal mouse brain, MSCs migrate extensively and differentiate into olfactory bulb granule cells and periventricular astrocytes, without evidence of cell fusion. These results indicate that MSCs may be “primed” toward a neural fate by the constitutive expression of neuronal antigens and that they seem to respond with an appropriate neural pattern of differentiation when exposed to the environment of the developing brain.
Previous studies have shown that adult stem cells exist in a variety of tissues, and these tissue-specific stem cells may have the capacity for transdifferentiating into cell types of other lineages [1–6]. The apparent multipotency of some adult stem cells has generated tremendous interest in their potential therapeutic value, because these cells may be used in autologous treatments and do not raise the ethical concerns associated with human embryonic stem cells. However, because some laboratories have failed to repeat several significant experiments demonstrating transdifferentiation [7, 8], the early exuberance surrounding the first reports of adult stem cell plasticity has given way to serious concerns about whether the early reports were describing true transdifferentiation or rather epiphenomena mediated, perhaps, by cell contamination or fusion [9–12].
Among the adult stem cells, bone marrow–derived mesenchymal stem cells (MSCs) may represent the best hope for autologous stem cell–based replacement therapies because, in addition to their potency and accessibility, these cells should not elicit graft versus host disease [13, 14]. For this reason, the MSC is one of the most extensively studied adult stem cells with respect to transdifferentiation potential [2, 14–17], especially toward neural differentiation in the hope of developing therapeutics for neurodegenerative diseases [18–20]. However, due to the lack of universally defined cell surface markers to characterize the MSC [21–23], it remains enigmatic with regard to both its identity and qualification as a true stem cell [21, 22]. Many previous reports of neural transdifferentiation of MSCs are also shadowed by some recent inconsistent results. For instance, using a transgenic mouse line carrying a green fluorescence protein expression vector under control of the glial fibrillary acidic protein (GFAP) promoter, Wehner and colleagues  examined the capacity of MSCs to undergo neuralization. After three in vivo and in vitro experiments, they concluded that bone marrow–derived cells could not differentiate along the astrocytic lineage. More recently, two independent groups [24, 25] re-evaluated the rapid and robust neuron-like neurofilament expression by MSCs under a previously reported dimethylsulfoxide/butylated hydroxy anisole (DMSO/BHA) induction protocol. They applied time-lapse imaging analysis and compared cells of different types, including rat epidermal fibroblasts, and PC12 cells. They found that the neuron-like morphological and immunocytochemical changes of MSCs after induction are not the result of genuine neurofilament extension but represent cell shrinkage and actin cytoskeleton retraction in response to chemical stress, because these cells showed similar morphological changes in the presence of Triton X-100 or sodium hydroxide. Although these results cannot account for all the neural transdifferentiation of MSCs reported so far and although they cannot explain the in vivo neuralization of the MSCs, they do raise serious questions about the generality of the neural differentiation potential of the MSCs and temper the hope of potentially applying MSCs in the treatment of brain disorders.
In an attempt to clarify some of these issues, we derived MSCs following a common protocol that uses the plastic adherence property of MSCs [26, 27]. We assessed MSC “stemness” by examining their clonality and multipotency in vitro and in vivo. Our findings suggest that (a) the MSCs constitutively display some neural properties in culture; (b) the induced “neuronal” morphological transformation is probably a generic response to chemical stress induced by dibutyryl cyclic AMP (dbcAMP)/isobutylmethylxanthine (IBMX); (c) the MSCs possess some stem cell characteristics in vitro, because they maintain multipotency in single-cell clones; and (d) after transplant, the MSCs appear to respond to environmental cues within the developing nervous system by differentiating into astrocytes and neurons, with no evidence of cell fusion.
Materials and Methods
C57/B6 and C57/B6GFP adult mice (8 weeks) were used to establish MSC cultures, using the physical property of plastic adherence [26, 27]. Briefly, mice were given a lethal dose of Phenobarbital, and the tibias and femurs were removed. A 22-gauge needle filled with Dulbecco's modified Eagle's medium (DMEM) was used to flush out whole bone marrow. The recovered cells were then mechanically dissociated, filtered through a 70-μm mesh, and plated in 35-mm tissue culture dishes containing DMEM supplemented with 20% fetal bovine serum (FBS), 0.5% gentamicin, and 1,000 units/ml of leukemia inhibitory factor (LIF), as per Jiang et al. . After 24 hours, the nonadherent cells were removed, and the culture medium was completely replaced. At confluency, MSCs were passaged (1:3 dilution) with fresh medium.
To generate clonal cultures, we grew single MSCs in conditioned medium collected from confluent MSC cultures. Conditioned medium was centrifuged at 2,600g for 10 minutes and filtered through a 0.22-μm mesh to eliminate cellular components. We created a dilution series with MSCs to reach a cell density of one to two cells per 5 μl and plated 5 μl of the cell suspension in each well of a 48-well plate. Immediately after plating, we examined each well with phase microscopy and excluded those wells containing more than one cell. We then added 100 μl of mixed medium (50% conditioned medium +50% fresh medium). To ensure single-cell clonality, we again examined each well after an additional 24 hours and discarded those containing more than one cell. Clonal MSC cultures were maintained in the mixed medium until confluency, at which point the cells were maintained in fresh, nonconditioned medium.
Fluorescence-Activated Cell Sorting Analysis of MSCs
Immunofluorescence with a variety of antibodies against surface antigens was used to characterize MSCs: directly conjugated anti-Sca1, anti-CD34, anti-CD45, and directly conjugated anti-mouse IgG2a (used as control) (1:500; BD PharMingen, San Diego, http://www.bdbiosciences.com/pharmingen). In addition, the following unconjugated antibodies were used: antic-Kit, anti-CD9, anti-CD31, anti-CD105 (1:500; BD PharMingen), and anti-CD11b (1:300; Serotec Ltd, Oxford, U.K., http://www.serotec.com). Primary antibodies were applied for 30 minutes at room temperature, followed by washing and application of fluorescent-conjugated secondary antibodies for an additional 30 minutes for the unconjugated antibodies. Cells were then centrifuged at 200g and washed twice in phosphate-buffered saline (PBS) to eliminate unbound antibodies. Approximately 106 cells per ml cell suspension was run through a flow cytometer (CELLQuest, Becton Dickinson FACScan; Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com).
Immunolabeling and Quantification
Immunolabeling was performed on MSCs plated on uncoated glass coverslips. Cells were fixed in ethanol/acetic acid (95:5) for 15 minutes, washed with PBS containing 0.1% Triton X-100 (PBST), and blocked for 30 minutes in PBST supplemented with 10% FBS. Cells were then incubated with primary antibodies overnight at 4°C, washed, and incubated in secondary antibodies for 1 hour at room temperature. Cell counting was performed under a fluorescence microscope (Olympus BX51; Tokyo, http://www.olympus-global.com). The ratios of positive cells were obtained by averaging three different experiments for both control and treatment groups. In each experiment, five randomly chosen views were counted and averaged. Free-floating, 40-μm brain sections were immunolabeled with the following antibodies, as previously described : nestin (1:50; Developmental Studies Hybridoma Bank, Iowa City, IA, http://www.uiowa.edu/∼dshbwww), GFAP (1 drop/0.5 ml; Shandon Immunon, Waltham, MA, https://www.thermo.com), neurofilament medium subunit (NFM; 1:500; EnCor Biotechnology Inc., Alachua, FL, http://www.encorbio.com), β-III-tubulin (1:1,500; Promega, Madison, WI, http://www.promega.com), S100-β (1:500; Sigma, St. Louis, http://www.sigmaaldrich.com), and polysialylated-NCAM (PSA-NCAM; 1:100; Chemicon, Temecula, CA, http://www.chemicon.com). Confocal laser scanning microscopic analysis of immunolabeling was performed on a Leica TCS SP2 confocal laser imaging system (Leica Microsystems, Wetzlar, Germany, http://www.leica.com). Every fifth sagittal section through the forebrain was selected for quantification of β-III tubulin+ cells in the hemisphere receiving cell engraftment. Total numbers of cells for the hemisphere were then obtained by multiplying by 5.
Elevating Cytoplasmic cAMP of MSCs
Our protocol for elevating intracellular cAMP was modified from Deng et al. . In addition to primary induction medium (0.5 mM IBMX/1 mM dbcAMP [Sigma] in DMEM/F12) used for the first 24 hours of treatment, a cocktail of growth factors (10 ng/ml of brain-derived neurotrophic factor [Pepro Tech Inc., Rocky Hill, NJ, http://www.peprotech.com]), nerve growth factor (Invitrogen, Carlsbad, CA, http://www.invitrogen.com), epidermal growth factor (Pepro Tech), basic fibroblast growth factor (Pepro Tech), and N2 Supplements (Gibco) was added to the primary induction medium for treatments longer than 24 hours.
In Situ Hybridization for GFAP mRNA
To generate GFAP riboprobes, we used reverse transcription–polymerase chain reaction (RT-PCR) to amplify a 401-bp DNA fragment of the GFAP gene (GenInfo Identifier: 26080421) from mouse brain tissue with a pair of primers designed using the Primer3 program (Whitehead Institute for Biomedical Research, Cambridge, MA, http://www.wi.mit.edu, and Howard Hughes Medical Institute, Chevy Chase, MD, http://www.hhmi.org) (forward: 5′-GCCACCAGTAACATGCA AGA-3′; reverse: 5′-ATGGTGATGCGGTTTTCTTC-3′). The PCR product was then cloned into the PCR4 TOPO vector (Invitrogen). After linearization, plasmids extracted from clones of both directions were used as templates to synthesize digoxigenin (DIG)-labeled GFAP sense and antisense probes using T7 RNA polymerase. In situ hybridization followed the protocol of Braissant and Wahli  with small modifications. The probe concentration was 400 ng/ml, and the hybridization temperature was set at 45°C.
For Western blotting, approximately 20 μg of protein from cell lysates was electrophoretically separated by 8% SDS-PAGE. After transfer to a nitrocellulose membrane, we applied anti-GFAP (1:30; Immunon) antibody and a chemiluminescence method for detection (ECL; Amersham Biosciences, Piscataway, NJ, http://www.amersham.com). We then incubated the membrane in striping solution at 56°C for 30 minutes and incubated it again using anti-actin (1:2,000; Abcam, Cambridge, U.K., http://www.abcam.com) antibody.
Transplantation of MSCs into Neonatal Mouse Brain
MSCs were trypsinized and labeled with the fluorescent carbocyanine dye, DiI (Molecular Probes Inc., Eugune, OR, http://probes.invitrogen.com), according to a protocol adapted from Laywell et al. . Briefly, cells were centrifuged for 5 minutes at 1,000 rpm and resuspended in fresh medium. DiI was dissolved in absolute ethanol (2.5 mg/ml) and added to the cell suspension such that the final concentration was 40 μg/ml. The cells were incubated in the DiI-containing medium for 30 minutes at 37°C before being washed three times in PBS.
DiI-labeled MSCs were transplanted into the lateral ventricle of postnatal day 1–4 (PN1–4) wild-type C57BL6 mice as described previously . Under hypothermia anesthesia, approximately 1 × 105 MSCs in 1 μl of PBS was injected into the left lateral ventricle. After 10 days of survival, mice were euthanized with an overdose of Avertin and perfused transcardially with 4% paraformaldehyde in PBS. The brain tissue was excised, post-fixed overnight in perfusate, and sectioned through the sagittal plane into 40-μm slices with a vibratome.
Y-Chromosome Painting for Cell Fusion Detection
Sections of 20 μm vibratome were used for assaying possible fusion events between donor MSCs and indigenous host cells in the neonatal mouse brain. Brain sections were first treated with 0.2 N HCl for 30 minutes and retrieved in 1 M sodium thiocyanate (NaSCN) for 30 minutes at 85°C. The sections were then digested with 4 mg/ml pepsin (Sigma; diluted in 0.9% NaCl [pH 2.0]) for 60 minutes at 37°C. After equilibrating in 2× SSC for 1 minute, the sections were dehydrated through graded alcohols. The tissue was then incubated with fluorescein isothiocyanate–conjugated Y-chromosome probes (denatured for 43 minutes at 37°C; Cambio, Dry Drayton, U.K., http://www.cambio.co.uk) using Hybrite (Vysis Inc., Des Plaines, IL, http://www.vysis.com) for 20.5 hours after a denaturing step of 6 minutes at 75°C. After hybridization, cells were washed first in 1:1 formamide/2× SSC, then in 2× SSC before being coverslipped in mountant containing 4′6-diamidino-2-phenylindole (DAPI; Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com). To evaluate potential cell fusion events, we first used a fluorescence microscope (Olympus BX51) to eliminate easily identified cells with a single Y chromosome in the nucleus, and we applied a confocal laser imaging system (Leica TCS SP2) to further inspect cells that potentially host more than one Y chromosomes within the boundary of the cell nucleus.
MSC Culture and Characterization
We established five different viable cultures of MSCs (two from C57/B6, three from C57/B6GFP) from the tibia and femur of adult mice according to the adhesive property of MSCs [26, 27]. Approximately 30 days after plating, the appearance of fast growing MSCs with fibroblast-like morphology could be observed amidst more slowly growing, round or polygonal cells that appeared first in the initial bone marrow dissociates culture. At approximately day 45, we observed stable fibroblast-like MSC culture (Fig. 1A), which can endure long-term culture (>50 passages). Besides observing the morphological change, we observed GFP silencing concomitantly in all of the cultures from the bone marrow of three C57/B6GFP mice used to establish MSCs, indicating a change of gene expression profile in the process of establishing MSC cultures (Fig. 1A).
To characterize the MSCs, we performed fluorescence-activated cell sorting (FACS) analysis using a battery of markers for characterizing MSCs (Fig. 1B). We found that MSCs are negative for the hematopoietic markers CD34, CD45, and Mac1 and for the stem cell marker c-kit, but positive for Sca1 (18.6% of total cells), and negative for the endothelial marker CD31, but positive for CD105 (19.1%) and CD9 (97%).
MSC Cultures Spontaneously Express Neural Markers
We evaluated the noninduced expression of “neural” proteins by MSCs and found that, in all five cultures, nearly 100% of MSCs are positive for the intermediate filament protein nestin. In addition, subsets of the MSCs are positive for several neuron-specific proteins, including β-III tubulin (12%) and NFM (13.2%); negative for PSA-NCAM, a surface protein expressed on migratory neuroblasts; positive for the astrocyte-specific protein, S100-β (15%), but negative for the astrocyte intermediate filament proteins, GFAP and vimentin (Fig. 2B). These properties of the MSCs remain unchanged between cells from early (<5) and late (>50) passages and are similar among all five of our MSC cultures.
Cyclic AMP Elevation Upregulates Astrocyte-Specific Protein
Several studies have used cytoplasmic elevation of cAMP to induce neural differentiation from MSCs [30, 33–35]. To test the same protocol on our MSCs, we treated the cells with 0.5 mM IBMX/1mM dbcAMP. We found that, as reported , cytoplasmic cAMP elevation does induce a significant morphological change of MSCs, in which the flat, fibroblast-like cells become neuron-like with rounded somas and long spindly processes. However, we observed no detectable change in the expression of most neural markers induced by the treatment using immunolabeling (Fig. 2B). Furthermore, when we treated NIH3T3 cells with the same protocol, we observed a similar morphological change without detecting neural marker expression (Fig. 2B, 2C). Elevation of cytoplasmic cAMP did, however, lead to an upregulation of the astrocyte intermediate filament protein, GFAP (Figs. 2B, 3A). This enhanced GFAP expression was confirmed using both in situ hybridization with digoxigenin-labeled GFAP riboprobes (Fig. 3B) and Western blotting (Fig. 3C).
Single Cell–Derived MSC Clones Generate Neuronal and Astrocytic Progenies
We cloned MSCs from single cells by limiting dilution in conditioned medium. Single cell–derived MSCs recapitulated the cell surface marker expression profile of their ancestor population by FACS analysis (data not shown). In the cells treated with dbcAMP/IBMX protocol, we found that cells immunopositive for both neuron- and astrocyte-specific protein were present in the same MSC clone (Fig. 4A). Furthermore, when we probed the cloned MSCs with antibodies against NFM, we failed to observe any immunopositive cells in clones consisting of fewer than five cells (n = 10); however, in clones consisting of 10 or more cells (n = 13), we did see labeling with this marker (Fig. 4Ba). This may imply that there are both symmetric and asymmetric divisions in MSCs that are regulated by cell density and/or replication number (Fig. 4Bb; Discussion).
MSCs Exhibit Neural Differentiation upon Grafting into the Neonatal Mouse Brain
To test the in vivo transdifferentiation capacity of MSCs, we grafted the non-treated cells into the lateral ventricles of neonatal mouse brain. We observed a migration of MSCs along the rostral migratory stream (RMS) from the lateral ventricle to the olfactory bulb. Whereas most of the grafted cells maintained a spindle-like appearance similar to their in vitro morphology, some cells exhibited morphological characteristics of astrocytes around the ventricle and penetrated into the overlying parenchyma (Fig. 5A). Immunolabeling shows that some of these cells are positive for GFAP antibody (Fig. 5B). A number of cells within the RMS are immunopositive for the neuronal marker β-III tubulin (Fig. 5B), and more significantly, we consistently observed a small number of MSCs possessing typical characteristics of granule cells within the granule cell layer (GCL) of the olfactory bulb (Fig. 5A). In the four animals we evaluated, 355 DiI+/β-III tubulin+ cells were found in the RMS, representing 46.4% of all donor cells found in the RMS. Immunolabeling reveals that these cells, like the indigenous migratory neuroblasts that normally repopulate the olfactory bulb interneuron population, are positive for PSA-NCAM (Fig. 5B). To confirm the expression of neuronal proteins by these donor MSCs, we used confocal laser scanning microscopy to verify that the expression of the proteins are indeed in the same focal layers of the DiI used to label the cells (Fig. 6). To control for the possible leakage of the DiI, we grafted identically labeled NIH3T3 cells into the ventricles of a different set of animals (n = 4). In this case, we observed labeled cells only near the site of injection within the subependymal zone of the lateral ventricle, with no labeled cells detected within the RMS or olfactory bulb (data not shown). Furthermore, when we applied Y-chromosome painting on female pups receiving male donor MSCs, we observed that the presence of a single Y chromosome in the nucleus of a cell accorded well with the DiI-labeling (Fig. 5Ab).
Chromosome Analysis Reveals No Evidence of Cell Fusion
To evaluate the possibility that cell fusion between donor MSCs and differentiated host cells is responsible for the co-expression of neuronal proteins and DiI labeling, we grafted DiI-labeled, male MSCs into neonatal mouse brain and analyzed tissue sections for the presence of cells with more than one Y chromosome in the male recipients. We optimized the Y-chromosome painting such that a high efficiency of detection (>99%) was achieved in cells with an intact nucleus, using DAPI counterstaining and fluorescence microscopy. From analysis of three different animals with fluorescence and confocal microscopy, we observed that all DiI-labeled cells (n = 165) contained only one Y chromosome. A representative analysis of Y-chromosome position is shown in Figure 7B. We conclude, therefore, that the presence of donor-derived olfactory cells with the morphology and immunophenotype of olfactory bulb interneurons is not the result of fusion between donor and host cells.
We have demonstrated that MSCs from adult mice constitutively express several neural markers in vitro under standard culture conditions. The previously reported neural induction protocol of applying dbcAMP/IBMX to elevate the cytoplasmic cAMP induces upregulation of astrocyte-specific protein GFAP in MSCs, although it seems ineffective in upregulating neuron-specific proteins as assessed by immunolabeling with several neuronal-specific proteins. In vitro, single-cell MSC clones spontaneously generate cells that express neuronal-specific proteins and cells that express astrocyte-specific protein under chemical induction. Nonfused MSCs also have the capacity to generate neurons and astrocytes upon grafting into the neonatal mouse brain. These cells seemingly behave similarly to indigenous neural progenitors, given that donor cells are seen to migrate along the RMS to the olfactory bulb, where they differentiate into olfactory interneurons.
MSC Cell Surface Antigen Profile Shows Usual Characteristics
We believe that the MSCs described in the present study are similar to the MSCs used in previous studies, based on the cell surface antigen profile [36, 37]. The absence of CD34, CD45, and CD11b has been widely accepted as the major difference between MSCs and hematopoietic stem cells (HSCs) . The expression of some endothelial cell markers, including CD105 and CD9, has also been reported in MSCs [38–41]. Whereas the expression of the stem cell marker Sca1 mirrors other reports , the lack of c-kit expression by our MSCs is anomalous , and this discrepancy may reflect the loose definition of MSCs currently in vogue. Although there is a wide range of surface markers that have been tested to characterize MSCs, at present there is no single set of phenotypic markers used to unequivocally identify an MSC. As a result, there may be unidentified subtypes of MSCs that differ slightly from one to another, and this may account for the variation in marker expression, as well as the inconsistent results regarding the transdifferentiation of MSCs from different laboratories. The use of FBS as the main–or only–source of growth factors to establish the cell population has been standard practice for more than 15 years. It is simple and effective, but the lack of positive selection markers–as used for HSCs–may result in the inclusion of undefined cell types, which may underlie the interlaboratory variability seen with these types of protocols. We noticed that our MSCs in culture have an irregular growth rate at different periods of the culture (data not shown). We also noticed that MSCs derived from GFP transgenic mouse lose ubiquitous GFP expression during the course of culture (Fig. 1A), indicating a genetic re-makeup during the course of transforming into stable cell populations that allow long-term culture.
MSCs Spontaneously Express Neural Proteins
The spontaneous expression of neural-specific proteins demonstrated by our MSCs casts doubt on some previously reported protocols that claim neural induction but fail to show the pre-induction level of neural-specific proteins. However, rather than calling into question the neural transdifferentiation potential of MSCs, this clarification actually strengthens it by showing the vigorous, spontaneous acquisition of neural properties by uninduced MSCs. The neural property exhibited by MSCs may be explained by the neural differentiation propensity of stem cells reflected in the development of the nervous system during embryogenesis. It is generally believed that unspecified ectoderm cells differentiate into neural lineage by default unless inhibited by ventralizing factors, such as bone morphogenetic protein-4 (BMP4) . So-called neuralizing factors such as noggin, chordin, and follistatin promote neuro-ectoderm specification by inhibiting BMP4 . The embryonic stem cells also show active spontaneous neural differentiation unless inhibited by BMP in vitro . Therefore, it is not surprising that MSCs, as multipotent stem cells, may exhibit a neural property in their default state of differentiation in vitro, where there are no pro-mesoderm inhibitors such as BMP4. The expression of some neural markers by pre-induced MSCs is a matter of some controversy in the literature. With the exception of neuron-specific enolase (NSE), Woodbury et al.  did not observe any neural-specific protein expression. Sanchez-Ramos et al.  reported low levels of NeuN, nestin, and GFAP expression detectable with immunocytochemistry, whereas Deng et al.  have previously reported expression of vimentin, Map1b, and β-III tubulin but no NFM, GFAP, or S-100-β. A more recent paper by Tondreau et al.  corroborates our finding by reporting significant expression of several neuronal markers, including nestin, β-III tubulin, Map2, and tyrosine hydroxylase by noninduced MSCs.
Neuronal Morphology Induced by dbcAMP/IBMX Protocol May Not Indicate Neuronalization
We found that the use of the dbcAMP/IBMX induction protocol, despite causing a vigorous neuron-like morphological change, does not seem to change the neuron-specific protein expression profile in MSCs evaluated by immunolabeling. Furthermore, we have also observed a rapid and dramatic morphological change in NIH3T3, similar to that of MSCs upon dbcAMP/IBMX treatment, but without expression of neuronal marker. In the initial paper describing the protocol , Deng et al. found no expression of NFM, which differs from our results, but they did report equal levels of MAP1b and β-III tubulin with and without induction, as we have found in our MSCs. Tondreau et al.  recently reported an unchanged nestin expression pre- and post-induction of more than 90%, as we have demonstrated, but reported a decreased β-III tubulin expression level from more than 90% to approximately 30% after 10 days of treatment with lower concentration of dbcAMP (5 μM). Our results, together with previous reports, suggest that the dramatic neuron-like morphological transformation of MSCs under dbcAMP/IBMX treatment is an unreliable indicator of neuronalization, supporting previous analyses of a DMSO/BHA induction protocol [24, 25]. Although we observed no significant increase of several neuron-specific proteins using immunolabeling after the treatment of dbcAMP/IBMX protocol, we caution that more sensitive detection methods may reveal subtle changes of the gene expression.
MSCs Can Express Astrocyte-Specific Protein GFAP In Vitro and In Vivo
Along with many other inconsistent reports of neural-specific protein expression with or without induction in MSCs, the expression of astrocyte-specific protein GFAP has also been controversial. Despite the early findings of MSC transdifferentiation into GFAP-expressing astrocyte in vitro and in vivo [28, 46, 48], Wehner et al.  reported that there was no GFAP expression from MSCs derived from a mouse strain carrying a GFP expression vector driven by the GFAP promoter cassette, and this paper partially supports the original paper describing the dbcAMP/IBMX induction of neural differentiation from MSCs, in which no GFAP expression has been detected before or after the treatment . However, our immunolabeling, in situ hybridization, and Western blotting data unequivocally demonstrate that GFAP expression is upregulated by cytoplasmic cAMP elevation. In agreement with this, Tondreau et al.  also reported the upregulation of GFAP by MSCs after prolonged exposure to a low concentration of dbcAMP/IBMX. Besides the demonstration of GFAP expression in vitro, we also observed MSC differentiation into GFAP-expressing cells after transplantation into the neonatal mouse brain. As shown in Figure 1A, MSCs undergo GFP gene silencing during the establishment of the long-term culturing MSC population. It may, therefore, be speculated that the gene-silencing event could have interfered with the GFP expression cassette in the previous study of Wehner et al.  and thus resulted in a failure to detect GFAP expression in MSCs.
MSCs Demonstrate “Stemness” by Clonality and Multipotency
We have shown that clonal MSC cultures give rise to cell populations that are identical to the parent population. These clones exhibit multipotency by differentiating into cells of neuronal and astrocytic lineages. Pittenger et al.  reported a similar clonal property of bone marrow–derived multipotent human MSCs in differentiating into adipogenic, chondrogenic, and osteogenic lineage. Using the immunophenotyping of our MSC clones of different sizes as a basis, we propose a working model that may reflect the symmetric and asymmetric cell division pattern in the MSCs (Fig. 4Ab). We suggest that at least three cell types exist in the MSC population, each with different potency: multipotent, neuron-restricted, and astrocyte-restricted. The fact that we did not observe neural marker expression in the small clones (<5 cells) may mean that only multipotent cells, which do not express neural markers, can renew themselves by symmetric division, whereas neuron-restricted and astrocyte-restricted cells do not survive or proliferate under clonal culture conditions. The fact that we start to observe neural-specific protein expression in larger clones (>10 cells) may mean that cell division number, or cell density, triggers asymmetric division that generates cells with restricted potentials.
Nonfused MSCs Give Rise to Neurons in the Neonatal Mouse Brain
Although the neural differentiation capability of MSCs in vitro has been widely explored, the in vivo response of this cell type upon direct engraftment into the brain has not been adequately assessed. Our finding that noninduced MSCs integrate into the postnatal neurogenic pathway of the RMS/olfactory bulb system by migrating appropriately and differentiating into olfactory granule cells supports the conclusion that the bone marrow–derived adult stem cells indeed possess neural transdifferentiation capability under the influence of environment cues from the brain. The fact that MSCs can migrate along RMS and differentiate into mature neurons at a distant site may indicate a potential therapeutic use for these cells in acting as neural progenitor cells and replacing lost neural tissue after injuries. Munoz-Elias et al.  reported a wide scope of migration of MSCs after transplantation into the embryonic rat brain, and grafted cells appeared to express the appropriate neuron marker calbindin within the olfactory bulb. Zhao et al.  demonstrated that human MSCs expressed astrocyte markers and some neuronal markers after grafting to the site of ischemic injury in rat brains. Further work in various injury models, designed to fully assess the ability of MSCs to functionally integrate into neural circuitry, will determine the potential therapeutic value of these cells in the treatment of neurological injury and disease.
In summary, we have demonstrated that MSCs acquire neural progenitor-like properties by expressing neuron- and astrocyte-specific proteins both spontaneously and after chemical induction. Although the previously reported neural induction protocol using dbcAMP/IBMX does not seem to promote neuronal differentiation in our MSCs, it may be able to drive astrocyte differentiation, as shown by GFAP upregulation. However, the neuron-like morphological transformation under the protocol may not be a reliable criterion for evaluating neuronalization, due to the fact the NIH3T3 also acquired identical morphology without neuron-specific protein expression. We have also used a confocal scanning imaging system and Y-chromosome painting techniques to demonstrate, unequivocally, that neural transdifferentiation from a stem cell of mesenchymal origin does exist and may be able to develop into the ideal cell type for cell-replacement therapy in the future.
This study was supported by the National Institutes of Health (DAS HL70143). We thank Dongqi Tang for his constructive discussions that led to the initiation of this work and Doug Smith and Bhavna Bhardwaj for their help in confocal microscopy imaging and FACS analysis.
E.L., B.P., and D.S. own stock in Reg Med.