Neural transdifferentiation of mesenchymal stem cells – a critical review


  • Invited Review.

    See also: Hanson & Caisander, Semb, Valdimarsdottir & Mummery.

Morten Meyer, Anatomy and Neurobiology, Institute of Medical Biology, University of Southern Denmark, Winsløwparken 21, DK-5000 Odense C, Denmark. e-mail:


The classic concept of stem cell differentiation can be illustrated as driving into a series of one-way streets, where a given stem cell through generations of daughter cells becomes correspondingly restricted and committed towards a definitive lineage with fully differentiated cells as end points. According to this concept, tissue-derived adult stem cells can only give rise to cells and cell lineages found in the natural, specified tissue of residence. During the last few years it has, however, been reported that under certain experimental conditions adult stem cells may lose their tissue or germ layer-specific phenotypes and become reprogrammed to transdifferentiate into cells of other germ layers and tissues. The transdifferentiation process is referred to as “stem cell plasticity”. Mesenchymal stem cells, present in various tissues, including bone marrow, have – besides differentiation into bone, cartilage, smooth muscle and skeletal muscle – also been reported to transdifferentiate into skin, liver and brain cells (neurons and glia). Conversely, neural stem cells have been reported to give rise to blood cells. The actual occurrence of transdifferentiation is currently much debated, but would have immense clinical potential in cell replacement therapy and regenerative medicine. Controlled neural differentiation of human mesenchymal stem cells might thus become an important source of cells for cell therapy of neurodegenerative diseases, since autologous adult mesenchymal stem cells are more easily harvested and effectively expanded than corresponding neural stem cells. This article provides a critical review of the reports of neural transdifferentiation of mesenchymal stem cells, and proposes a set of criteria to be fulfilled for validation of transdifferentiation.


adult stem cells


brain-derived neurotrophic factor


basic fibroblast growth factor


butylated hydroxyanisole




central nervous system


dibutyryl cyclic AMP




epidermal growth factor


embryonic stem cells


fumarylacetoacetate hydrolase


fibroblast growth factor


γ-aminobutyric acid


glial fibrillary acidic protein


haematopoietic stem cells




interleukin 3


microtubule-associated protein 2


multipotent adult progenitor cells


myelin basic protein


mesenchymal stem cells


neuronal nuclei


neurofilament M


nerve growth factor


neural stem cells


neuron-specific enolase


neurotrophin 3


retinoic acid


tyrosine hydroxylase






Stem cells are undifferentiated cells without mature tissue-specific characteristics, which in response to proper stimuli are able to proliferate to reproduce themselves and to produce generations of progenitor cells that can differentiate into one or more cell types (1). Stem cells are typically classified according to site of location. Embryonic stem cells (ESCs) are thus found in the inner cell mass of the blastocyst (early embryo), while adult or tissue-derived (somatic) stem cells (ASCs) are found in already developed tissues of the fetus or the newborn, juvenile and adult organism. Another way of characterizing stem cells is by their potential, through one or more generations of daughter cells, to give rise to one or more types of specialized progeny. The most potential stem cells, the ESCs, which in principle can give rise to every cell type in the body, are called pluripotential, followed by various degrees of multipotential stem cells, such as cells in the germ layers and immature and mature tissues, and finally unipotential cells that are restricted to giving rise to only one cell lineage and cell type, such as B-lymphocytes. This restriction in potential reflects a corresponding specialization, which occurs through several stages during embryonic and fetal development, and for the last cell and tissue-specific stages is maintained to different extents by ASCs residing in the mature tissues of the adult organism (2, 3). The presence and potential of ASCs in a given tissue reflects the natural dynamics in turnover of cells in that tissue as well as the capacity for repair following disease or injury, illustrated by the high cellular turnover and repair capacity of blood and the blood-forming (haematopoietic) tissue and the very limited turnover and cell replacement capacity of the central nervous system (brain and spinal cord). Within a given tissue the most potent or basic ASCs appear to be located to microenvironmental niches, where they rarely divide, but do so in response to various stimuli to maintain the population and to produce more rapidly dividing “amplifying” cells, which in turn may differentiate into all cell types found in the tissue concerned (3).

For many years, the general belief was that the tissue-residing, multipotent ASCs were developmentally restricted only to differentiating into cell lineages of the specified tissue where they resided. In mammals the only exceptions were pathological differentiations into tumour or cancer cells. Reports of experimental findings have, however, during the last few years indicated that at least some ASCs may have the capacity to transdifferentiate into cells of completely different cell and tissue lineages or germ layers. Reports include differentiation of bone marrow cells into muscle cells (4), cardiac muscle (5), hepatocytes (6), astrocytes (7), and neurons (8–21), differentiation of neural stem cells into haematopoietic cells (22) and myogenic elements (23), and differentiation of hepatic oval cells into neural cells (24).

In this review we focus on the capacity of mesenchymal stem cells (MSCs) to transdifferentiate into neurons and glial cells. The MSCs are of interest in this respect because of both the reported observations of transdifferentiation and the doubts raised concerning these, and the clinical perspectives a potential transdifferentiation would have by allowing generation of, for example, neural progenitor cells for cell replacement therapy from MSCs from the very same person who needs therapy (autologous stem cell-based therapy). The aim is to provide a critical review of the reported observations, supplemented by our own results, and to present alternative explanations for functional effects attributed to transdifferentiation of transplanted cells. Finally, a set of recommendations is presented, which should facilitate evaluation of stem cell plasticity and transdifferentiation.


The term transdifferentiation was originally used by developmental biologists to describe the ability of apparently fully differentiated cells derived from a given tissue to change into cells with characteristics of a different tissue in response to either cell culture or surgical removal of adjacent tissue. Today the term “transdifferentiation” is commonly used to describe the plastic ability of ASCs to differentiate into cell lineages of tissues different from the one in which the somatic stem cell resides, and even into cells originating from other germ layers. Transdifferentiation sensu stricto must involve genetic reprogramming with turning off of some sets of genes and turning on of others. If such transformation occurred rapidly, transient co-expression of products of both sets of genes might be found in the cell. If, on the other hand, the process of transformation lasted some time, there might be intermediate stages of cells, where sets of original cell type-characteristic genes were inactive, while new sets of genes had not yet been activated. Clearly true transdifferentiation can only be considered to have taken place when a new characteristic cellular phenotype is stably established (25).

In developmental biology several studies indicate that activation or suppression of so-called “master genes” is essential for the differentiation of cells from one stage to another. Pax5 has been found to act as a master gene in lymphocyte development for commitment to the B-cell lineage. Loss of Pax5 function in cells thus results in reversion of B-lymphoid commitment and expression of T-lymphoid and myeloerythroid differentiation potentials (26). Similarly, it has been proposed that the homeobox gene PDX-1 is important for initiation of development of pancreatic insulin-producing β-cells, even from hepatocytes (27, 28), and the myogenic regulator MyoD to induce myogenesis in various differentiated cell lines (29). Recently, Lee et al. (30) found that the transcriptions factor Foxa2 acts upstream of PDX-1 in the PDX-1 cascade and hence controls the expression of PDX-1. Knowledge of master genes is central in developmental biology and attempts to understand and experimentally control stem cell differentiation. At the same time identification of master genes appears essential for clarifying the occurrence of transdifferentiation and its molecular basis.


The adult bone marrow contains two prototypical stem cell populations: haematopoietic stem cells (HSCs) and mesenchymal stem cells (MSCs) also called bone marrow stromal cells. Both HSCs and MSCs are of mesoderm origin. Whereas HSCs give rise to the various blood cells, MSCs can differentiate into mesenchymal derivatives, including osteocytes, chondrocytes, adipocytes, and myocytes (4, 31–38). In addition, MSCs play a role in providing the stromal support system for HSCs in the bone marrow.

MSCs are only found in low numbers in the bone marrow (1 MSC per 105 mononuclear marrow cells), but can be purified by isolation from other cells in the bone marrow and expanded with high efficiency (34).

Several methods exist for isolation of MSCs from an aspirate of bone marrow. The aspirate is most often harvested from the superior iliac crest of the pelvis (32, 34). However, MSCs have also been isolated from tibial and femoral marrow compartments (32, 39, 40) and thoracic and lumbar spine (41). Furthermore, similar multipotent stromal cells can be isolated from organs such as adipose tissue (42), muscle (43), synovial membranes (44), vascular elements in the deciduous teeth (45, 46), peripheral blood (47), and umbilical cord blood (48). The traditional method of isolation and purification of MSCs from a heterogeneous cell aspirate takes advantage of the selective adhesion of these cells, compared to other cells in the aspirate, to plastic surfaces (31, 33, 34, 49). In the case of bone marrow the aspirate is fractioned on a density gradient solution (often Percoll or Ficoll) to eliminate some unwanted cell types and debris present in the aspirate. The cells in the upper low-density fraction are then plated and enriched using standard cell culture techniques. The primary cell cultures are most often maintained for 12–16 days, during which the non-adherent haematopoietic cells are removed by medium replacement (34, 50). One disadvantage of this simple method of isolation is the cellular heterogeneity of the cultures. To avoid this, alternative methods based on the cell's immunological or combined immunological and physical properties have been used (49, 51). When MSCs are cultured they have an average initial doubling time of 12 to 60 h, dependent on different factors, such as harvesting procedure, frequency of MSCs in the bone marrow and donor age. When cultures reach a high density, cells stop dividing and transform from cells with spindle-shaped morphology into cells with a larger flat and oblong fibroblast-like morphology (40, 41, 48, 49). MSCs are found negative for haematopoietic surface markers, such as CD34, CD45, CD14, CD31, CD133, and positive for CD105, CD166, CD54, CD55, CD13, CD29, CD44, SH2, and SH3. Differences in surface marker characteristics do, however, exist among reported studies, which may be explained by differences in culture methods, donor age, and differentiation stage of the cells (32, 34, 49, 52), but also make comparisons among studies difficult.


MSCs are like other stem cells capable of self-renewal, but it has been proposed that stem cell self-renewal is less crucial for mesenchymal tissue functionality due to multipotentiality and phenotypic flexibility (53). With such plasticity, commitment and differentiation might be reversible in response to environmental cues, just as MSCs might have maintained a higher degree of stem cell plasticity with abilities to convert from one lineage stem cell type to another at a later stage than other tissue-derived, multipotent ASCs. Such plasticity is, for example, essential in bone growth and remodeling.

Under various defined culture conditions and with addition of stimulatory compounds, MSCs have been reported to exhibit transdifferentiation plasticity (Fig. 1). Woodbury et al. (16) thus found that populations and clonal lines of MSCs expressed ectodermal, endodermal and mesodermal genes, assuming a far broader differentiational potential or plasticity of MSCs than expected. This is in line with reported observations of MSCs differentiating into various mesodermal cell types (chondroblasts, adipocytes, osteoblasts, cardiac myocytes and vascular endothelial cells), ectodermal cell types (neural cells and skin) and endodermal cell types (hepatocytes, lung cells and gut cells) when treated with a selected combination of growth factors and substrates (4–21). Using FACS analysis, Western blots and RT-PCR, Tondreau et al. (55) have reported that more than 80% of cultured MSCs constituently express native immature neuronal proteins, such as nestin and β-tubulin III. Moreover, MSCs were found to express more mature neuronal or glial proteins, such as tyrosine hydroxylase (TH), microtubule-associated protein 2 (MAP-2) and glial fibrillary acidic protein (GFAP), after five passages without any specific induction.

Figure 1.

ifferentiation of mesenchymal stem cells (MSCs). MSCs naturally differentiate into cells of the mesodermal germ layer, and differentiation into several specialized mesodermal cell types has also been induced experimentally after cell transplantation and in vitro after treatment with various growth factors and substrates. Transdifferentiation into cell types of the ectodermal and endodermal germ layers has been reported, but with questionable results.


Transdifferentiation of MSCs into neurons has aroused considerable interest. Studies of neural transdifferentiation of MSCs can be divided into two groups, namely in vitro studies (10, 11, 14–18, 20, 21) and transplantation studies (8, 9, 12, 13, 19). Below we will review some of the available data from these studies as well as show data from one of our own studies on neural transdifferentiation of human MSCs (Fig. 2). Doubts raised about the validity of the observations and alternative explanations for the observations are subsequently dealt with in a separate section.

Figure 2.

Morphology of β-tubulin III-positive, neuron-like cells derived from a telomerase immortalized human mesenchymal stem cell line (hMSC-TERT). Cells were grown for 12 days in serum-free medium with 100 nM TPA, 100 μM dbcAMP, 25 μM forskolin and 100 ng/ml aFGF. Most cells displayed compact cell bodies and were either simply bipolar (A) or complex multipolar (B). Both types had branched processes (arrow). The complex multipolar cell was in intimate contact with transitional cells.

Neuronal transdifferentiation of MSCs in vitro

Neuronal transdifferentiation of MSCs in vitro attracted great attention in 2000 in relation to the publications by Woodbury et al. (10) and Sanchez-Ramos et al. (11). Woodbury et al. (10) reported having induced rat and human MSCs to differentiate into cells displaying neuronal characteristics by treating the MSCs with neuronal induction media composed of DMEM and different combinations of β-mercaptoethanol (BME), dimethylsulfoxide (DMSO) and butylated hydroxyanisole (BHA). Within a few hours almost 80% of the treated MSCs turned into cells with neuronal morphology. Performing immunocytochemical staining for nestin, an intermediate filament protein expressed in neuroepithelial neuronal precursor stem cells with decrease of expression with neuronal maturation, the treated MSCs stained positive for nestin after 5 h, while no nestin staining was observed after 6 days. Few of the treated cells showed immunoreactivity for the neural markers TrkA, Tau and neuron-specific enolase (NSE), and none of the cells stained positive for the astroglial marker glial fibrillary acidic protein (GFAP). Corresponding observations were made by Sanchez-Ramos et al. (11), who reported that mouse and human MSCs could be induced to express neuronal markers such as neuron-specific nuclear protein (NeuN), nestin, and GFAP when cultured in the presence of epidermal growth factor (EGF), retinoic acid (RA) or RA together with brain-derived neurothrophic factor (BDNF). Sanchez-Ramos et al. also assessed the influence of factors released from developing neural tissue and cell-cell interactions by co-culturing of MSCs from transgenic lac-Z mice with fetal mouse mesencephalic cells. They found that co-culturing of MSCs with mouse mesencephalic cells increased the number of NeuN- and GFAP-expressing cells. Consistent with these results Aboulfetouh et al. (21) later reported that co-culturing of MSCs with hippocampal brain slices induced MSCs to differentiate into neuron-like cells, suggesting that cell-to-cell contact, in addition to trophic factors and cytokines, plays an important role in neural differentiation of MSCs.

Following these studies an increasing number of independent groups have reported on the neuronal differentiation potential of MSCs in vitro. Deng et al. (15) treated human MSCs with isobutylmethylxanthine (IBMX) and dibutyryl cyclic AMP (dbcAMP), two compounds known to increase the intracellular concentration of cAMP, and reported that 25% of the MSCs differentiated into neuron-like cells expressing NSE and vimentin after 6 days in culture. They did not, however, detect any expression of neurofilament M (NF-M), MAP2, Tau, S-100 and myelin basic protein (MBP) in either untreated or IBMX/dbcAMP-treated MSC cultures, and like Woodbury et al. (10), but unlike Sanchez-Ramos et al. (11), they saw no GFAP expression. The data were interpreted to indicate that conditions that increase intracellular cAMP could induce MSCs to become early neural progenitors. The role of cAMP as a “neural inducer” was also observed by Bang et al. (56), who reported that elevation of cAMP induces a neuronal morphology in human prostatic adenocarcinoma cells. These changes included G1 synchronization, growth arrest, and loss of clonogenicity, indicating terminal differentiation. In line with this, Cox et al. (57) reported that physiological agents increasing intracellular cAMP induced prostate tumour cells to assume the characteristics of neuroendocrine cells. Interestingly, withdrawal of these agents resulted in rapid loss of the neuroendocrine phenotype and expression of neural markers and made the cells re-enter the cell cycle, suggesting chronic cAMP-mediated signalling was required to block proliferation and to maintain a neuroendocrine phenotype.

Kohyama et al. (14) used two different methods for inducing MSCs to differentiate into neurons. First, they differentiated mouse MSCs in neural induction medium containing the demethylating agent 5-azacytidine (5-azaC), nerve growth factor (NGF), neurotrophin 3 (NT-3), and BDNF. Following this they investigated the role of noggin as a neural inducer of murine MSCs by transfecting the MSCs with noggin. These cells were then differentiated in the neural induction medium. Both methods resulted in neuron-like cells, which responded to depolarizing stimuli as functional neurons. Furthermore, cells from both experimental cell groups expressed neural markers, such as Tuj1, NeuN, MAP2, TrkA, TrkB, TrkC, NCAM, GAP-43 and the astrocyte markers GFAP and Gal-C. The proportion of GFAP-expressing cells was lowered, and the number of neurons increased in MSCs cultures transfected with noggin compared to non-transfected cultures. Noggin is a BMP2/4 antagonist, which has been found to be involved in creating a microenvironmental niche for adult neurogenesis in the subventricular zone of the mammalian brain (58). The observation that noggin lowered the number of GFAP-expressing cells could be caused by the inhibition of BMP2, which normally suppresses neurogenesis and promotes astrogenesis from neural stem cells (59).

The role of culture substrates in neural differentiation of MSCs was investigated by Kim et al. (17), who used different combinations of basic fibroblast growth factor (bFGF), NGF, RA, and different substrates and extracellular matrix components, such as laminin, gelatine, collagen, fibronectin, and polyornithine for neuronal induction of human MSCs. The best inductive effect potential was reported to result from a combination of RA and bFGF using fibronectin as a substrate. This induced 40% of the cultured cells to gain a neuron-like morphology and to stain for the neural lineage marker NF-M, TuJ-1, and vimentin, but not the mature neuronal marker Tau or the astroglial marker GFAP. Qian & Saltzman (20) also observed that culture substrate played an important role for neuronal differentiation of MSCs when testing the neuronal inducing potential of different culture substrates in combination with neural induction medium consisting of DMSO, BHA, KCl, forskolin, hydrocortisone, insulin, and valproic acid on human MSCs. They found that the neural induction medium resulted in an apparent morphological change of the cultured cells on all substrates. However, using matrigel (a combination of laminin, collagen type IV, entactin, heparan sulphate proteoglycans and different growth factors) as substrate resulted in the highest number of cells (69%) expressing the neural marker NSE as well as the most extensive branching of the cells. They also reported that increasing the coating density further enhanced the neuronal differentiation of the MSCs, which together with their other results demonstrates the importance of extracellular matrix components for cell development and differentiation.

Using cultured mouse MSCs, Jin et al. (18) examined the effect of different growth factors and combinations thereof on the expression of neuronal marker proteins and found that cultures treated for up to 5 weeks with EGF, bFGF, RA, and NGF contained many cells displaying neuron-like cellular processes and expressing neuronal markers, including NeuN, MAP2, Tau, synaptophysin, α1A, and α1B calcium channel subunits, NR2A glutamate receptor subunits, and α-aminobutyric acid (GABA). Although detection of these markers indicates that the mouse MSCs displayed some neuronal characteristics, the intracellular distribution of these markers was different from the distribution in mature neurons. Jin et al. (18) concluded that a variety of growth factors could drive MSCs toward neuronal phenotypes; but morphological neuronal features and ectopic expressions of neuronal markers are insufficient markers of neuronal differentiation. Besides morphological studies and immunocytochemistry it is accordingly recommended that experiments on neural induction also include electrophysiological parameters, such as resting membrane potentials, depolarizations in response to changes in membrane permeability to potassium and sodium, and induction of actions potentials. These parameters can be measured in neural stem cell-derived presumptive neurons using in vitro recording techniques (60). Until now most in vitro studies, with few exceptions (14), have not employed electrophysiological parameters to demonstrate that neuron-like cells have been induced from MSCs.

Transplantation studies of MSCs

Cell and tissue transplantation to the central nervous system (CNS) has a long tradition in exploring the development, plasticity, and regeneration of the brain and spinal cord. Neurotransplantation has moreover been used for repair and restoration of diseased and damaged nervous tissue, using a variety of cell types and tissue for transplantation, including stem cells or developing tissue harbouring stem cells. In the following, studies employing engrafting of MSCs into the brain or blood stream with entry into the brain will be reviewed. Transplantation of MSCs into the brain of experimental animals with the purpose of testing survival, integration and differentiation was initiated by Azizi et al. (8) using direct injections of human MSCs into the rat brain striatum. Following transplantation they reported that the MSCs migrated through the host brain tissue in a manner similar to that of implanted neural stem cells (NSCs). They also reported that the grafted and migrating MSCs lost expression of markers typical of MSCs in culture. These observations were supported by Kopen et al. (9), who reported that murine MSCs grafted into the lateral ventricles of neonatal mouse brains resulted in migration of these cells throughout the forebrain and cerebellum. Some of the MSCs in the striatum and hippocampus were reported to express GFAP, just as occasional neurofilament-positive MSCs were reported to occur in the brain stem, suggesting that some of the injected MSCs differentiated into a neuronal phenotype.

Transplantation of MSCs into embryonic brains was investigated by Muñoz-Elias et al. (19), who introduced rat MSCs into the telencephalic ventricles of rat embryos in utero to study survival, migration, definitive localization and phenotypic expression of the donor cells. They reported that the transplanted MSCs were able to survive in the CNS for up to 2 months, to assume neuronal morphology, and to express a variety of region-specific neuronal genes.

In vivo studies of neuronal transdifferentiation of MSCs have also involved systemic administration of MSCs. Mezey et al. (12) injected male HSCs intraperitoneally into 1-day-old female mice, lacking the PU1 transcription factor, which is vital for leukocyte development. Since these mice do not produce their own bone marrow cells they need to be engrafted with such cells to survive. The recipient mice were examined at the age of 1 and 4 months and bone marrow-derived male cells were reported to be evenly distributed throughout the brain of all transplanted animals. Chromosome Y-bearing cells expressing the neuronal markers NeuN and NSE were moreover reported to be present in the cerebral cortex, hypothalamus, hippocampus, amygdala and striatum. Consistent with these observations Brazelton et al. (13) reported that intravascular delivery of mouse bone marrow cells genetically marked with GFP into the tail vein of lethally irradiated normal adult mice led to development of donor-derived cells expressing neuronal proteins in the CNS of the recipients. The donor-derived cells were furthermore reported to exhibit phosphorylation of the transcription factor CREB in a manner similar to that of the neurons cells surrounding them.


Although several studies have reported transdifferentiation of MSCs into neuronal cell types (8–21), the findings and interpretations have been challenged. The main arguments against – and alternative explanations – are reviewed below.

Cell fusion

Fusion of cells occurs as part of normal mammalian tissue development. The development of skeletal muscle cells and bone tissue osteoclasts and formation of giant cells from mononuclear phagocytes involves both cell fusion and differentiation. Cell fusion has also been put forward to explain the observations interpreted as transdifferentiation of MSCs. Two independent studies involving co-culturing of murine MSCs with ESCs (61) and transgenic mouse NSCs with ESCs (62) thus showed that cell fusions did occur between ASCs and ESCs. Terada et al. (61) showed that murine MSCs fused spontaneously with ESCs in cultures that contained interleukin-3 (IL-3), and that these spontaneously fused cells adopted the phenotype of the ESCs. This finding was confirmed by Ying et al. (62), who co-cultured neural progenitor cells from transgenic mice with modified ESCs and found that the resulting multipotent cells expressed genetic markers of neural cells, making it appear that de-differentiation of the neural cells into a multipotent cell line had occurred. That fusion had occurred was shown by chromosome and karyotypic analyses, demonstrating the presence of tetraploid hybrids with full pluripotent character. Cell fusion has also been found to occur after in vivo transplantation of MSCs. Vassilopoulos et al. (63) thus found that grafting of fumarylacetoacetate hydrolase (Fah)-expressing (Fah+) mice MSCs into livers of (Fah/) mice resulted in restoration of normal liver function due to fusion between Fah+ MSCs and Fah/ recipient hepatocytes. In the brain, bone marrow-derived cells have been shown to fuse with Purkinje neurons from cerebellum in both human and mice (64–66). This suggests that Purkinje cells, like liver cells, could have a unique fusion capacity.

Since cell fusion is a likely explanation for some of the plasticity observed in co-culturing and transplantation studies, interpreted as transdifferentiation, future transdifferentiation studies should include careful genotyping of potentially transdifferentiated cells. Cell fusion mechanisms cannot, however, explain the results of the in vitro studies, where MSCs are not co-cultured with other cell types or tissues (11, 15–17, 20), just as the occurrence of cell fusion by itself does not rule out the possibility that genuine transdifferentiation of MSCs can occur under certain conditions (67, 68).

Contamination by other cell types

Another explanation for current observations interpreted as transdifferentiation of ASCs could be the presence of quiescent primordial stem cells in adult tissues, which would “contaminate” the cell samples and cultures used for MSCs transdifferentiation studies. These quiescent cells might, if exposed to the right conditions, such as an in vitro culture system and trophic factors and cytokines, be triggered to differentiate into cell types not normally found in the tissue of residence (69). Several studies have indicated that such pluripotent cells might exist in the bone marrow (70–74). Reyes et al. (72) thus identified and isolated a rare cell from human MSC cultures derived from bone marrow that could be expanded for more than 80 population doublings. These cells, referred to as multipotent adult progenitor cells (MAPCs), were reported to differentiate not only into mesenchymal lineage cells, but also into cells with visceral mesodermal, neuroectodermal and endodermal characteristics in vitro. Such differentiation required that the MAPCs were cultured under low cell density conditions in medium without serum, EGF, PDGF-BB, and LIF, but with lineage-specific cytokines (71, 72, 74). Furthermore, after having microinjected single MAPCs into 3.5 day-old blastocysts of C57BL/6 mice and transferred the blastocysts to foster mothers for fetal development and birth, Jiang et al. (71) reported that MAPCs had contributed to most cell types in the chimeric mice, which is a property normally thought to be restricted to ESCs. In a recent study Jiang et al. (75) reported that cells with MAPC characteristics could also be isolated from brain and muscle tissue, supporting the concept of the presence of quiescent primordial stem cells in adult tissues (76).

Cytotoxic cell changes

An important argument against the rapid – within hours – transdifferentiation of MSCs reported from culture studies by Woodbury et al. (10, 16) is that the changes reported as signs of transdifferentiation are, in fact, cytotoxic changes induced by the media and staining procedures. When Lu et al. (77) exposed MSCs, fibroblasts, keratinocytes, HEK293 cells, and PC-12 cells to various stressors, including the induction medium used by Woodbury et al. (10) (containing BME, BHA, DMSO), other detergents, high-molarity sodium chloride, and extremes of pH, they found that all of these treatments resulted in cell shrinkage and adoption of neuron-like morphologies of the cells within a few hours. The cell shrinkage was followed by increases in immunolabeling for the neuronal markers NSE and NeuN, though this immunoreactivity was not accompanied by a corresponding occurrence of mRNAs in RT-PCR. The adoption of neuron-like morphologies, moreover, occurred despite inhibition of protein synthesis with cyclohexamide. These observations were confirmed by Neuhuber et al. (78), who found that changes in morphology of MSCs within 2 h after treatment of the cells using the protocol of Woodbury et al. (16) resulted in disruption of the actin cytoskeleton and a retraction of the cell edge. BME, one of the chemicals used for neuronal induction, was shown to induce neuron-like morphology in a concentration-dependent manner, with induction of cell death at high concentrations. Also BHA, used in transdifferentiation media, has been found to induce cell death (79). Although the cytotoxic concentrations of BHA were higher than the concentration used by Woodbury et al. (10, 16), combination of BHA with serum-free medium could initiate cell shrinkage as an early event of cell death that might be reversed after short-term exposure (77). Applying the protocols of Woodbury et al. (10) and Deng et al. (11) to rat MSCs for comparison, Rismanchi et al. (80) examined cell viability, number of CNS-like cells, and neural antigen presentation. Whereas none of the MSCs cells was observed to express NeuN before transdifferentiation treatment, there was, using the Woodbury protocol, a significant increase in NeuN expression at 4, 5, and 6 h after induction. Using the Deng protocol there was a tendency only for increase in NeuN expression after 6 days. The Woodbury protocol resulted in significantly increased cell death of “differentiated” cells compared to “non-differentiated” cells, while no difference in cell death was observed among “differentiated” and “non-differentiated” cells using the Deng protocol. For both protocols it was found that maintenance of neuron-like morphology required continued culturing in maintenance media (80).

The rapidity by which the neuron-like morphology is both gained and lost in these experiments strongly argues against physiological cell differentiation. It is thus very unlikely that changes in cellular organization and gene and protein expression related to differentiation commitment can become so advanced within a few hours of exposure to inductive molecules (78). Changes in cell morphology suggesting transdifferentiation of MSCs, but occurring within a few hours of treatment, must accordingly be considered with great caution and most probably reflect cytotoxic changes (77, 78, 80).


Human neurodegenerative disorders, such as stroke, Parkinson's and Alzheimer's diseases, amyotrophic lateral sclerosis, epilepsy, trauma and intoxications, are all characterized by neuronal cell loss, associated with corresponding loss of functions and disabilities.

Cell replacement therapy

In Parkinson's disease, with a localized loss of dopaminergic nigrostriatal projection neurons, there is both experimental and clinical proof-of-concept that intrastriatal grafting of fetal dopaminergic neuroblasts from the developing substantia nigra can provide long-term restoration of functional abilities (81). Positive results of cell replacement therapy have also been reported following immature neural cell grafting in Huntington's disease and experimental animal models of stroke (82, 83). Clinical experimental cell therapy for Parkinson's disease and other neurodegenerative disorders does, however, suffer from both logistic and ethical constraints related to the procurement of fetal human brain donor tissue from legal abortions, just as there for Parkinson's disease are unsolved problems regarding full survival/differentiation of the dopaminergic neurons with only survival/differentiation of approximately 10% of the dopaminergic neurons in the grafts. The search for alternative donor cells is currently focusing on differentiation of transplantable dopaminergic neuroblasts from ESCs or tissue-derived ASCs isolated and propagated to sufficiently large numbers from fetal brains. One additional cell source might be transdifferentiated, non-neural stem cells, such as MSCs, provided that a functional effect of such cells is biologically possible. One potential advantage of the use of transdifferentiated cells for cell replacement therapy, if possible, would be that cells could be isolated directly from the potential recipient, transdifferentiated and grafted as an autologous cell graft without subsequent risks of immune reactions and need for immunosuppressive treatment and associated risks of side effects (84). Studies of transplantation of MSCs to animal models of neurodegenerative disorders, including Parkinson's disease (84, 85) and ischaemic brain injury (86–91), have resulted in improvements in the functional deficits, but the specific mechanisms responsible for these effects are at present largely unknown.

Cell factory therapy

For neurodegenerative diseases, and other diseases with cell loss, such as juvenile diabetes (type I diabetes), cell replacement therapy is a late-stage therapy for patients where irreversible cell loss and functional ability or risk of side effects have reached a level beyond the reach of pharmaceutical treatment. Aiming at earlier stages of disease a great deal of research has been and is being performed to develop means by which an ongoing or threatening cell loss can be restricted or prevented. In such approaches the use of stem cells as cell factories for delivery of protective biologically active compounds, such as growth or trophic factors and protective or anti-inflammatory chemokines/cytokines, has been introduced. Accordingly, the restorative effects originally attributed to transdifferentiation of MSCs in experimental animal studies may – at least in part – be mediated by release of beneficial protective factors from these cells or their normal lineage derivatives (87–89).


From the above review it is clear that a number of fundamental issues need further examination and clarification before MSC transdifferentiation can be scientifically proven and accepted and therapeutic application of such transdifferentiated cells can be envisaged.

First, it has to be resolved whether transdifferentiation of stem cells residing in established, mature tissues is at all biologically possible, i.e. is a fact or an artifact. This will require general scientific acceptance of a set of criteria, which should include:

  • • use of objectively defined, homogeneous cell sources

  • • cross-testing of reproducibility of cell isolation and cell propagation procedures and differentiation protocols in different laboratories

  • • definition of well-defined and accepted sets of genotypic and phenotypic biomarkers and cellular functions characteristic of the cell type intended to result from transdifferentiation. For neural transdifferentiation electrophysiological characterization of resting membrane potentials and formation and propagation of axon potentials must be included, as well as expression of functional neurotransmitter receptors

  • • demonstration of stable (ideally long-term) expression of the “end point” biomarkers and cell characteristics

  • • exclusion of expression of the chosen “end point” cellular biomarkers by the cell sources prior to potential transdifferentiation

  • • exclusion of cell fusion interpreted as transdifferentiation

  • • exclusion of cellular leakage and uptake by other cells of cell tracking markers, such as transgene fluorescent proteins

  • • confirmation in vivo of transdifferentiation results obtained in vitro - in accordance with the above criteria

From this review on neural transdifferentiation of mesenchymal stem cells, it is evident that very few reports – either individually or together – fulfil the proposed criteria, meaning that the question of fact or artifact is still unanswered. Moreover, at the moment there seems to be some uncertainty as to which cell type (”general” mesenchymal stem cell, bone marrow stromal cell, bone marrow haematopoietic stem cell or bone marrow-derived multipotent adult progenitor cell (MAPC)) is the actual candidate for MSC transdifferentiation. In the meantime, real progress is being made regarding the functional, neuroprotective or reparative role of MSCs and HSCs and their progeny as “cell factories” in relation to neurodegenerative disorders.

The expert technical help of Dorte Lyholmer and Maibritt Vang Damm is gratefully acknowledged. This work was supported by grants from the Danish Medical Research Council to MM, JZ and the Danish Stem Cell Research Center. We thank Professor Moustapha Kassem, Odense University Hospital, Odense, Denmark for providing the human mesenchymal stem cells used for our transdifferentiation studies.