Two conceptual pillars of developmental biology have been rocked in the last few years. The long standing dogma that neurons in the adult brain do not regenerate has been challenged by evidence that new neurons are born in “germinal zones” of the hippocampus and subventricular zone in rodents and humans throughout life (Eriksson et al., 1998; Gage, 2000). The second concept under attack is that a cell committed to a specific phenotypic fate by virtue of residence in a mature organ cannot change its destiny. A growing body of experimental evidence proves that cells derived from one adult tissue can change into cellular phenotypes not normally found in that tissue. This suggests that cells from mature tissues can be reprogrammed (“transdifferentiated”) to change into a completely distinct phenotype from that found normally in that tissue. An alternative explanation is that mature tissues harbor a small number of pluripotent stem cells with a greater differentiation potential than appreciated previously.
A stem cell is defined by its functional capacity to both self-renew and to generate a large number of differentiated progeny (McKay, 1997; Gordon and Blackett, 1998; Scheffler et al., 1999). The truly totipotent cell, capable of generating all cell types and constructing a complete organism, is the fertilized egg. During development and morphogenesis cells proliferate, migrate and differentiate, but throughout this process there is a residual of quiescent, uncommitted cells that are believed to be the result of asymmetric division ensuring self-renewal of the stem cell population. Stem cells can be isolated from the developing embryo and from specific adult tissues, but their potential for differentiation into all cell types becomes restricted gradually to a more limited range of cells typical of the mature tissue in which the stem cell (or progenitor cell) resides.
Over the last few years, tissue-specific stem cells have been shown to give rise to cells not normally found in the organ or tissue of residence. For example, neural stem cells can give rise to blood cell lines (Bjornson et al., 1999) and bone marrow stromal cells can generate skeletal muscle (Wakitani et al., 1995; Ferrari et al., 1998), cardiac muscle (Orlic et al., 2001a,b; Makino et al., 1999), oval hepatocytes (Petersen et al., 1999), as well as glia and neuron-like cells. The existence in postnatal tissue of stem/progenitor cells with ample proliferation and differentiation potential opens up the possibility of using autologous adult stem cells to treat degenerative, post-traumatic and hereditary diseases. This article reviews studies in which cells from postnatal bone marrow or umbilical cord blood were induced to proliferate and differentiate into glia and neurons, cellular lineages that are not in their normal repertoire.
MULTIPOTENT NON-HEMATOPOIETIC PROGENITORS IN BONE MARROW
Bone marrow stromal cells (BMSC) provide the structural and functional support for the generation of blood cell lineages from hematopoietic stem cells. BMSC consist of morphologically and biochemically distinct cell types: bone marrow fibroblast–reticular cells, adipocytes, osteoblasts, macrophages, and endothelial cells. BMSC can be cultivated in vitro and contain progenitors capable of generating bone, cartilage, fat, and other connective tissues. These non-hematopoietic precursors found in bone marrow stromal cells are also known as colony forming unit fibroblasts (CFU-f) and mesenchymal stem cells (MSC). (Prockop, 1997). The BMSC per se have been reported to have many characteristics of MSC; hence, the terms are often used synonymously. MSC, a homogenous population of fibroblast-like cells purified by Percoll gradient and expanded in vitro can generate progeny that differentiate into multiple cell lineages, including bone, fat, tendon and cartilage (Pittenger et al., 1999). Several recent reports demonstrate that under specific experimental conditions BMSC can also differentiate into cells that are not part of their normal repertoire: skeletal and cardiac muscle, hepatocytes, glia and neurons (Azizi et al., 1998; Ferrari et al., 1998; Makino et al., 1999; Petersen et al., 1999; Sanchez-Ramos et al., 2000; Jackson et al., 2001; Kohyama et al., 2001; Orlic et al., 2001a,b).
Reports on neural cell differentiation from bone marrow or umbilical cord blood can be grouped for analysis: the in vitro studies (Table I) and the transplantation studies (Table II). To facilitate comparison of experimental results from different groups, Table I lists salient experimental variables used in each study. Most researchers relied on the separation of stromal cells by adherence to plastic, and in most of the in vitro studies hematopoeitic progenitors were depleted or removed from the marrow cells. Nevertheless, one study identified the presence of CD34+ cells in the subclone of cells that could give rise to neural cells. The agents used to induce neural differentiation in vitro included retinoic acid, growth factors (alone or in combination), antioxidants, a demethylating agent, compounds which increase intracellular cyclic AMP (cAMP) and a physiological neural inducer, noggin. Although both neuron-like cells and glia were generated in most of the in vitro studies reviewed, glial cells were generated exclusively in one study (Nakano et al., 2001). Neurons with no glia were uniquely produced in another study (Woodbury et al., 2000). One in vitro study provided physiological evidence of neuronal function after differentiation from BMSC. Many experiments were performed in vivo and involved transplantation of untreated or modified marrow cells into a host animal. Table II summarizes the experimental variables in most of the transplantation studies. Bone marrow was obtained from mice, rats or humans. In mice, some researchers pretreated animals with 5-fluorouracil to increase the generation of marrow stem cells. Cells were implanted directly into the brain or ventricles, or administered systemically via a tail vein. Various animal models were used: rats with ischemic infarct, irradiated mice, MPTP-induced parkinsonism and immunocompromised mice. Improvement in neurologic deficits were demonstrated in the stroke, trauma and Parkinson disease animal models.
Table I. Differentiation of Bone Marrow Cells to Neural Lineages In Vitro: Comparison of Study Parameters
Population of cells
Induction of differentiation
Markers of neural lineages
Human BMSC depleted of CD34+ and mouse BMSC depleted of Scal+ cells; plastic adherent cells
DMEM + FBS + EGF + FGF
RA + BDNF + NGF; on polyethylimine-coated substrate or bed of murine fetal midbrain cells
DIFFERENTIATION OF BONE MARROW CELLS INTO MICROGLIA, ASTROCYTES AND OLIGODENDROGLIA
The earliest work on the relationship between marrow cells and glial cells focused on the natural trafficking of microglia between bone marrow and brain. Microglia are believed by many to derive from an hematopoietic line (monocytes), whereas astrocytes and oligodendroglia are considered to be derived from embryonic neuroectoderm and are developmentally distinct from microglia (Ling and Wong, 1993; Ling, 1994). Some microglia, however, are thought to have a neuroectodermal origin (Neuhaus and Fedoroff, 1994). Eglitis and Mezey (1997) sought to determine the extent to which cells outside the central nervous system contribute to the maintenance of microglia in adult mice. Bone marrow cells were labeled with a retroviral vector carrying the gene for neomycin resistance (neoR). A second approach for tracking marrow-derived cells in the recipient relied on in situ hybridization with a probe specific to the Y-chromosome of male donor marrow cells; the neoR- labeled male bone marrow cells were then infused by tail vein into sublethally irradiated female mice (WBB6F1/J-KitW/KitW-v). Over the next few days to weeks later, there was an influx of labeled cells into the brain of the recipients (Eglitis and Mezey, 1997). Marrow-derived cells were found throughout all regions of the brain from cortex to brainstem. They appeared to reside within the parenchyma because perfusion with phosphate-buffered saline (PBS) did not remove them. Occasional marrow-derived cells were found in association with vascular structures. The densities of donor cells in the recipient brain parenchyma paralleled the vascularity of a given region. Cortex, with few capillaries, had a lower cell density of marrow-derived cells than the more vascularized chorid plexus. Area postrema had the highest density of marrow-derived cells within the parenchyma. Some bone marrow-derived cells were positive for the microglial antigenic marker F4/80. Other marrow-derived cells expressed the astroglial marker glial fibrillary acidic protein (GFAP). Approximately 10% of the marrow-derived cells in the brain expressed either the microglial F4/80 antigen or GFAP; the identity of the remaining 90% of the marrow-derived cells in the recipient brain is unknown. These results indicated that some microglia and astroglia arose from a precursor that is a normal constituent of adult bone marrow. The authors considered the appearance of marrow-derived astroglia a normal process because the numbers of marrow-derived cells detected in brain increased over time, and their appearance did not appear to be a consequence of the transplantation procedure. In addition, the radiation did not appear necessary for the marrow cells to migrate to brain. Male donor cells engrafted and persisted for greater than two months in recipients that had received no irradiation. Furthermore, there were as many Y-chromosome/GFAP double-stained cells seen in the animals without radiation as seen in animals with irradiation. An interesting observation made by the authors was the appearance of cells with marrow markers in the ependymal layer of the ventricles. The finding of bone marrow-derived cells suggest that these cells home in and differentiate in response to signals from the subependymal zone.
In another set of experiments using the male bone marrow to female recipient paradigm, Eglitis et al. (1999) demonstrated a preferential homing of marrow-derived progenitors to the site of an hypoxic/ischemic injury in rat brain. In an acute unilateral middle cerebral artery occlusion model, 2.8% of the total DAPI-stained nuclei counted in the ischemic lesioned side of the brain were derived from bone marrow (i.e., Y-chromosome+), whereas 1.8% of the total nuclei were bone-marrow derived in the intact unlesioned side. Thus, the ischemic side of the brain had attracted 55% more marrow-derived cells than the non-ischemic side. No such difference was found between the two hemispheres in brain sections obtained from two intact animals grafted with bone marrow cells. The percentage of total GFAP+ cells that were double labeled (Y-chromosome+/GFAP+) was 161% greater in the lesioned hemisphere compared to the unlesioned side. Of the total GFAP+ cells on the lesioned side, 4.7% were bone marrow-derived, and on the unlesioned side, 1.8% were bone marrow-derived.. There was clearly a preferential targeting of the marrow-derived astrocytes to a region of cerebral injury. Astrocytic proliferation has been shown to occur in both the ischemic regions of brain as well as in regions that are undamaged by ischemia. The findings of Eglitis et al. (1999) suggest that an additional source of astrocytes is related to increased migration and differentiation of cells derived from bone marrow.
Other researchers have reported that infusion of human BMSC into rodent brain resulted in engrafting, migration and survival of cells (Azizi et al., 1998; Kopen et al., 1999). A subset of human marrow cells (separated on the basis of adherence to plastic) labeled with bisbenzamide were injected directly into the corpus striatum of rat. From 5–72 days later, brain sections were examined for the presence of donor cells; approximately 20% of the infused cells had engrafted in the host brain. The cells had migrated from the injection site to corpus callosum, contralateral cerebral cortex and ipsilateral temporal lobe. After engraftment, these cells lost markers typical of marrow stromal cells in culture, such as immunoreactivity to antibodies against collagen and fibronectin. BMSC developed many of the characteristics of astrocytes, and their engraftment and migration contrasted markedly with fibroblasts that continue to produce collagen after implantation.
Grafting of a subset of BMSC into the lateral ventricle of neonatal mice also resulted in their migration throughout the forebrain and cerebellum without disruption of host brain architecture (Kopen et al., 1999). In these experiments, the bone marrow stromal cells were depleted of cells that express the cell surface receptor CD11b, a marker of myelopoietic cells. The grafted BMSC were labeled with bisbenzamide or bromodeoxyuridine (BrdU) to track the fate of the cells. In the forebrain, a large number of donor cells were found ipsilateral to the injection site throughout the striatum, from the anterior commissure to the cingulate cortex. BMSC were also reported to line white matter tracts, including the corpus callosum and the external capsule, suggesting that their distribution throughout the forebrain was an ordered process of migration. Many BMSC were detected lining the ependyma throughout the ventricles. The presence of cells double-labeled for BrdU and GFAP suggested that some of the BMSC within the corpus striatum, the molecular layer of the hippocampus and the cerebellum had differentiated into astrocytes. Interestingly, BMSC were also located in areas undergoing active postnatal neurogenesis, including the Islands of Calleja in the ventral forebrain and the subependyma of the olfactory bulb. A large number of BrdU-labeled cells were found integrated within the folia of the cerebellum. The majority of the cells were localized to the external granular layer, the internal granular layer and to a lesser extent, the molecular layer. The Purkinje cells in the cerebellum were not labeled with BrdU, consistent with their earlier maturation during embryogenesis. Most of the BrdU-labeled cells in the cerebellum were found ipsilateral to the side of injection, but some were found on the contralateral side. Many BrdU-labeled BMSC lined the fourth ventricle uniformly, and small foci were seen in the white matter tracts adjacent to the dorsal horns of the fourth ventricle. The authors suggested that the BMSC gained access to the external granular layer and the reticular formation of the brain stem by following a pathway similar to that used by neural progenitors at the time when they emigrate out of the rhombic lip, into the primordial external granular layer during embryogenesis. In rare sections, occasional neurofilament positive BMSC were found in the brainstem suggesting that some BMSC differentiated into a neuronal phenotype (Kopen et al., 1999).
Nakano et al. (2001) demonstrated the capacity of murine bone marrow cells to differentiate into three distinct glial phenotypes (oligodendrocytes, astrocytes and microglia) after direct injection into the corpus striatum of irradiated mice. Systemic infusion of bone marrow cells resulted only in the appearance of marrow-derived microglia in brain, however, demonstrating the importance of the brain microenvironment in providing instructive signals for astroglial and oligodendroglial differentiation. The marrow cells were tracked by prelabeling with a retroviral vector containing the green fluorescence protein gene. Before labeling, the bone marrow cells were prestimulated for 48 hr in fibronectin-coated culture plates and media containing human interleukin-6 (IL-6) and murine stem cell factor. The bone marrow cells were transplanted into irradiated mice by either systemic infusion or direct injection into the corpus striatum of brain. To identify cell types, brain sections were stained with specific antibodies against neuronal cell markers: neuron specific enolase (NSE) for neurons, GFAP for astrocytes, carbonic anhydrase II (CAII) for oligodendrocytes, and ionized calcium binding adapter molecule 1 (Iba1) for microglia. Twenty-four weeks after systemic infusion, transplanted cells expressed Iba1 but none of the other brain cell markers. These results are in contrast to those reported by others (Eglitis and Mezey, 1997), who found both microglia and astrocytes in brains of mice when bone marrow cells were administered systemically. The population of bone marrow cells differed in these two studies. Because Nakano et al. (2001) prestimulated the marrow cells with stem cell factor and IL-6, it is likely the population of cells administered was not the same as that used by Eglitis and Mezey (1997).
DIFFERENTIATION OF BONE MARROW CELLS INTO NEURONS IN VITRO
Studies by Sanchez-Ramos et al. have demonstrated that a subset of both human and murine bone marrow has the capacity to differentiate into neural cells and, in particular, to express markers of early neuron development (Sanchez-Ramos et al., 1998, 2000). BMSC were separated from whole bone marrow by adherence to polyethylene culture flasks. Proliferation of cells was maintained by use of epidermal growth factor (EGF). Before induction of differentiation, cultures were enriched in fibronectin immunoreactive (ir) cells and depleted of mouse hematopoietic stem cell (Sca1+) or human hematopoietic stem cells (CD34+). Treatment of the cultures with retinoic acid (RA) and brain derived neurotrophic factor (BDNF) resulted in decreased number of fibronectin-ir cells. This was due to gradual loss of the large, flat fibronectin-ir cells and the appearance of smaller oval cells with short processes that did not stain with fibronectin (Fig.1, Frame 1). Analysis of BMSC lysates prepared from cultures treated with either proliferation medium or differentiation medium demonstrated the presence of nestin, neuron-specific nuclear protein (NeuN) and GFAP. Treatment with RA or RA + BDNF decreased the expression of nestin protein. Microscopic examination of the cultures after immunocytochemical processing revealed a small proportion of NeuN-ir and GFAP-ir cells (0.5 and 1%, respectively, of the BMSC cells).
To assess the influence of factors released by developing neural tissues and cell–cell interactions, BMSC obtained from transgenic lac-Z mice, which express E. coli β-gal, were co-cultured with fetal mouse mesencepahlic cells or neonatal rat forebrain glia cell cultures. Co-culturing of BMSC with fetal mouse midbrain increased significantly the percentage of BMSC that expressed NeuN and GFAP (markers of neurons and astroglia, respectively). These results were confirmed in a second set of experiments, which utilized BMSC labeled with a fluorescent vital stain (Fig.1, Frame 9). The co-culture experiments support the hypothesis that cell–cell contact, in addition to signaling with trophic factors and cytokines, plays an important role in differentiation of these BMSC. The neural cells produced from BMSC in the co-cultures did not exhibit the morphology of mature neurons or glia, nor did they express microtubule-associated protein (MAP)-2, a marker of mature neurons. This may due to the short duration of incubation (maximum of two weeks) and a slower maturation rate for human-derived cells. The expression of NeuN, but not MAP-2, in BMSC-derived cells is consistent with current knowledge regarding the timepoints of expression of neuronal proteins: MAP-2 is expressed at a later developmental stage than NeuN. Immunohistochemically detectable NeuN protein has been reported to first appear at developmental timepoints which correspond with the withdrawal of the neuron from the cell cycle or with the initiation of terminal differentiation of the neuron (Mullen et al., 1992). In contrast MAP-1, -2 and -3 undergo a number of significant changes during development, with the expression of MAP-2 considered to occur “late,” particularly between 10–20 days in the postnatal rat pup (Riederer and Matus, 1985).
Whereas Sanchez-Ramos et al. (2000) used RA in combination with growth factors to induce differentiation of BMSC to neural phenotypes (neuron-like cells and glia), others found that β-mercaptoethanol (BME) treatment could rapidly induce differentiation into neuron-like cells, but not glial cells (Woodbury et al., 2000; Black and Woodbury, 2001). BMSC (from either rat or human) were propagated in vitro for over 20 passages, attesting to their proliferative capacity. To induce differentiation to neuron-like cells, the BMSC were transferred to serum-free medium containing 1–10 mM BME. Within 1 hr of exposure to BME, changes in morphology of some of the cells were apparent. Responsive cells assumed neuronal morphological traits progressively over the first 3 hr. Cells exhibited increased expression of the neuronal cytoplasmic marker NSE within 30 min of treatment. The authors documented a remarkable metamorphosis in the cells over the next several hours: the cell bodies became increasingly spherical and refractile and exhibited a typical neuronal perikaryal appearance. BMSC-derived neurons displayed distinct neuronal morphologies ranging from simple bipolar to large, extensive branched multipolar cells (Fig. 1, Frames 10,11). Rare neurons exhibited pyramidal cell morphologies; neurons elaborating long processes with evident varicosities were more common. To address the issue of long term differentiation, expression of nestin was monitored immunocytochemically and by Western blot analysis of the cultures at 5 hr, 1 day, and 6 days post-differentiation. High levels of nestin protein was seen in a subset of BMSC-derived neural cells at 5 hr, and the proportion of nestin-positive cells decreased with time. By Day 6 post-induction, there was no detectable nestin expression in the cells, a finding consistent with ongoing maturation of neurons. Concomitant with these changes, TrkA, the high affinity nerve growth factor receptor, was detectable at 5 hr, and persisted through the 6 days.
Other agents, in addition to BME, were found to be effective in changing BMSC into a neural phenotype. Dimethyl sulfoxide (DMSO), butylated hydroxyanisole (BHA) and butylated hydroxytoluene (BHT) alone or in combination were found to be effective inducing agents. The most effective treatment was found to be 2% DMSO and 200 μM BHA, which resulted in as high as 78% of the cells expressing NSE. Further analysis also detected other neuronal markers, including NeuN, NF-m (an intermediate neurofilament) and tau protein (Fig. 1, Frames 12–15). Interestingly, none of these protocols resulted in the differentiation into glial cells, because no GFAP immunoreactive cells were found.
Kohyama et al. (2001) described two additional methods for inducing BMSC to differentiate into neurons in vitro. One protocol utilized 5-azacytidine (5-Aza-C), a demethylating agent capable of altering gene expression (Holiday, 1996), in a medium containing a combination of growth factors. The induction medium contained 5-Aza-C (10 μM) supplemented with the growth factors NGF, BDNF and NT3. After 96 hr of induction, the medium was replaced with B27-supplemented Dulbecco's Modified Eagle Medium (DMEM)/F12 containing the same three growth factors. An alternative protocol utilized the neural inducer noggin, a diffusible factor that mediates neural induction during early embryogenesis as well as during adult neurogenesis by antagonizing bone morphogenetic protein (BMP) signaling (Smith and Harland, 1992). Noggin also converts embryonic stem cells into primitive neural stem cells by inhibition of BMP-related signaling (Tropepe et al., 1999). The neurons generated from marrow stroma using either of these two protocols formed neurites, expressed neuron-specific markers and genes, and started to respond to depolarizing stimuli as functional mature neurons. Among stromal cells, the researcher found that isolated mature osteoblasts with strong in vivo osteogenic activity could be converted efficiently into functional neurons
Deng et al. (2001) reported that agents that increase intracellular cAMP levels, such as isobutylmethylxanthine (IBMX) and dibutyryl cAMP (db-cAMP), augmented the proportion of human BMSC that differentiate into a neural cell morphology (Fig. 1, Frame 2). They noted however, that even undifferentiated human BMSC express some markers characteristic of neural cells, such as MAP1B, neuron specific tubulin (TuJ-1), NSE and vimentin. The treatment with the enhancers of cAMP for six days increased the expression of NSE and vimentin mRNA without increasing transcription of either MAP1B or TuJ-1. There was no detectable expression of NF-m, MAP2, tau, GFAP and myelin basic protein before or after differentiation. These results are somewhat different from those obtained in two earlier reports using different culture conditions (Sanchez-Ramos et al., 1998; Woodbury et al., 2000). The change to a neuron-like morphology was similar in all three studies, but the number of neural cells varied widely. The results of Deng et al.(2001) were similar to those of Woodbury et al. (2000), in that no GFAP was detected, whereas Sanchez-Ramos et al. (1998) observed GFAP expression both before and after differentiation.
FUNCTIONAL ACTIVITY OF BONE MARROW-DERIVED NEURONS
Although BMSC have been shown to express proteins found typically at various stages of neuronal development, it was not clear that these neuron-like cells possessed the functional activities of true neurons. So far, only one laboratory has published strong evidence for functional activity in vitro. Bone marrow-derived neuronal cells were analyzed by whole-cell patch clamp recording (Kohyama et al., 2001). Treatment of BMSC with 5-Aza-C led to a decrease in resting membrane potential to −20 mV on Day 14 and −50 mV on Day 28. A similar resting membrane potential was also observed in neurons that served as a positive control. Ionic currents were also measured by the patch clamp method under the voltage clamp condition. With the increase in voltage, rectifying current was clearly detected, indicating the presence of voltage-dependent K+ current. The K+ channels started to be expressed concomitantly with the morphological change and increased expression of neuron-specific markers. Moreover, the ability of the bone marrow-derived neurons to respond to depolarizing stimuli was documented (Kohyama et al., 2001). These cells showed a rapid and reversible calcium increase in response to acetylcholine, a response characteristic of neurons (Fig. 1, Frames 16,17).
NEUROGENIC CELLS MAY BE ISOLATED FROM UMBLICAL CORD STROMA
Human umbilical cord blood may also harbor cells capable of differentiation into neural lineages (Sanchez-Ramos et al., 2001). Human umbilical cord blood cells were plated in culture dishes and incubated with DMEM + 10% fetal calf serum (FCS) for 2 days then media was changed to either RA + NGF or maintained in fresh DMEM + FCS for 4–7 days. Microscopic examination of immunostained cultures treated with RA + NGF revealed a heterogeneous mixture of cell types, ranging from large flat epithelioid cells to small spindle-shaped cells with fine branching neuritic processes. Treatment of the cultures with RA + NGF resulted in a decrease in total number of cells visualized under phase contrast microscopy compared to controls treated with DMEM. This treatment also increased the proportion of the Musashi-1 immunoreactive cells to 6.2% of the total cells (Fig.1, Frame 4). RA + NGF treatment also increased the proportion of cells (+18.7%) that exhibited β-tubulin III immunoreactivity. Approximately 8% of the cells incubated with DMEM showed β-tubulin III immunoreactivity. Thirty-four percent of DMEM-treated cultures were immunoreactive for GFAP, a marker of astrocytes, but treatment with RA + NGF increased the proportion of GFAP-ir cells to 66.2%. Sixty percent of the cells were immunoreactive for BrdU, indicating that the cells were continuing to proliferate.
Human umbilical cord blood cells prepared as above were also transplanted into the developing rat brain, which is known to provide a conducive environment for development of neural phenotypes (Zigova et al., 2002). Neonatal pups (1-day-old) received unilateral injection of a cell suspension from cord blood cell cultures containing either DMEM or RA + NGF into the anterior part of the subventricular zone (SVZ). One month after transplantation, animals were perfused, brains cryosectioned and immunocytochemistry was carried out for identification of neural phenotypes. Approximately 20% of transplanted cells, regardless of their pretreatment, survived within the neonatal brain. The majority of the grafted cells were found in the SVZ with some dispersion into adjacent cortex and corpus callosum. Cells grafted from the DMEM + FBS cultures were considerably more dispersed than those from the RA + NGF cultures. Double-labeling with cell type-specific markers revealed that approximately 2% of the umbilical cord blood-derived cells (from both RA + NGF and DMEM-treated cultures) expressed GFAP, and rare cord blood-derived cells (<0.2%) expressed the neuronal marker β-tubulin III. There was limited migration away from the SVZ but there was no migration along the rostral migratory pathway to the olfactory bulb, the usual destination of neural progenitors in the SVZ. The majority of these cells, regardless of pretreatment, appeared not to behave like typical neural progenitors. This may be a result of the injection of a heterogenous cell population already committed to a more differentiated state before grafting and a poor survival rate of the human cord blood cells grafted into non-immunosuppressed rat brain. Future work with this model should utilize neural progenitors isolated and enriched from umbilical cord and transplanted into an immunosuppressed animal.
In a study of gene expression using DNA microarray technology, umbilical cord blood cells were incubated under the same conditions (DMEM + FCS vs. RA + NGF) as described previously (Sanchez-Ramos et al., 2001). Cord blood cells cultured in the presence of RA + NGF exhibited significant changes (>2-fold) in the expression of 322 genes of a total of 12,600 human genes represented on the Affymetrix DNA Chip®. The majority of these genes were not related directly to the process of neurogenesis; however, at least 20 of these genes could be linked by literature searches to products found in neurons, glia or developing neural cells. For example, the greatest degree of upregulation (44-fold increase) was seen in the mRNA for neurite outgrowth extension protein or pleiotrophin. Several other transcripts associated with early neuronal development, such as glypican-4, neuronal pentraxin II, neuronal PAS1, and neuronal growth associated protein 43 (GAP43) all increased significantly. Musashi-1, the earliest marker of neural precursors, was upregulated 1.5-fold. Concomitant with the increased expression of markers indicative of neural development, there was a decrease in expression of many genes associated with development of blood cell lines. The greatest changes occurred in the expression of HLA Class I locus C heavy chain, macrophage receptor MARCO, secreted T-cell activation protein Attractin (attractin), leucocyte immunoglobulin-like receptor-8 (LIR-8), thymocyte antigen CD1c, erythropoietin receptor and erythropoietin.
These findings suggest that human umbilical cord blood contains cells that can be induced to neural antigenic and morphological phenotype. Musashi-1 is an RNA-binding protein that has been found in the developing or adult CNS tissues of frogs, birds, rodents, and humans (Kaneko et al., 2000). The anti-Musashi-1 monoclonal antibody used in this study was provided by H. Okano and has been shown to react with undifferentiated, proliferative cells of the SVZ in the CNS of all vertebrates tested (Kaneko et al., 2000). Both the upregulation and the post-translational processing of Class-III β-tubulin, one of the most specialized tubulins found in neurons (Fanarraga et al., 1999), are believed to be essential throughout neuronal differentiation (Laferriere and Brown, 1996; Laferriere et al., 1997). The cord blood cultures treated with RA + NGF also increased expression of other genes reported to be specific for neural cells, but we selected only a few of them for further study. The increased expression of pleiotrophin (neurite outgrowth promoting protein), an extracellular matrix-associated protein that enhances axonal growth in perinatal cerebral neurons (Raulo et al., 1992), was independently confirmed by RT-PCR and Western blot analysis Glypican-4 expression was upregulated in the DNA microarray study, but RT-PCR analysis revealed that it was also present in the untreated cord blood cells. Glypican-4 has been reported to be expressed in cells immunoreactive for nestin and the D1.1 antigen, markers of neural precursor cells, but has not been detected in early postmitotic or fully differentiated neurons (Hagihara et al., 2000). Necdin mRNA, a marker specific for neurons (Taniura et al., 1998; Yoshikawa, 2000), was detected in the RT-PCR analysis in treated and untreated cells, and its protein expression (NeuN) was shown to be increased by 19% in the Western blot analysis.
GENE EXPRESSION IN NEURAL PROGENITOR AND HEMATOPOIETIC CELLS DIFFERENTIATION: CLUES TO META-DIFFERENTIATION?
The conversion of cells of mesodermal origin (osteoblasts) into cells that normally have ectodermal origin (neural cells) has been termed “meta-differentiation” (Kohyama et al., 2001). The mechanism for meta-differentiation may involve at least two basic processes, operating alone or in concert: 1) a reprogramming of the gene expression profile of a differentiated cell into that of a pluripotent cell, and 2) proliferation and differentiation of a pluripotent progenitor/stem cell already harbored in the BMSC. The reprogramming of a differentiated cell into a pluripotent one has been termed transdifferentiation, not to be confused with dedifferentiation, a process that also results in reversion of a mature cell into a more primitive state as a preliminary to neoplastic transformation. The best example of transdifferentiation is the reprogramming that occurs in a somatic nucleus transferred into the environment of an enucleated egg (as in reproductive cloning). The cytoplasm of the egg contains instructive signals that reset the old program of the epithelial cell nucleus into that of a fresh totipotent cell capable of generating a complete organism. Little is known about the signaling molecules in egg cytoplasm responsible for this process of reprogramming. More is known about compounds used to induce meta-differentiation of bone marrow cells into neurons, such as the demethylating agent 5-Aza-C or RA in conjunction with growth factors (Umezawa et al., 1992; Holiday, 1996). These molecules have direct and indirect actions on gene transcription, supporting the concept that reprogramming the genome is the key to understanding meta-differentiation.
Clues to the molecular mechanism responsible for transdifferentiation can be found in the systematic study of gene expression profiles that orchestrate differentiation of stem cells into specific phenotypes. Gene expression studies in stem cell populations have been facilitated by combining a powerful genetic subtraction technique, representational difference analysis (RDA), with cDNA microarray analysis, and validated by in situ hybridization to see where the gene(s) are expressed (Geschwind et al., 2001). This method overcomes the limitation of working with heterogeneous populations containing not only stem cells, but also numerous committed progenitor cells and differentiated cells. An RDA subtraction was performed in which cDNA from neural stem cell cultures that had been differentiated for 24 hr was subtracted from cDNA of sister cultures that were maintained as stem/progenitor cultures (the neurosphere condition). This analysis identified known and novel genes enriched in neural progenitor cultures. Using in situ hybridization technique, many genes were found to be expressed preferentially in the germinal regions of brain (the ventricular and subventricular zone). Several genes were enriched in hematopoietic stem cells (HSC) suggesting an overlap of gene expression in neural and hematopoietic progenitors (Geschwind et al., 2001). In a converse study, a large set of HSC genes were found to be expressed in mouse neurospheres, a population containing neural progenitor cells (Terskikh et al., 2001). Many of these are genes known to be involved in cell cycling, DNA repair and signaling machinery. Some of these genes are candidate markers for HSC and neural progenitors, useful for future experiments to identify neural progenitors at the earliest stages. Could it be that expression of some early neural proteins in bone marrow derived cells is simply an epi-phenomenon, reflecting shared genetic programs characteristic of all stem/progenitor cells? This would suggest that the bone marrow cells expressing a few neural proteins are really marrow cells artificially caught in early stages of differentiation, and are not truly neurons. There is compelling evidence for the conversion of osteoblasts into mature neurons, however, complete with electrophysiologic function typical of neurons (Kohyama et al., 2001). The gene expression profiles of stem cells, regardless of their pedigree, does shed light on the molecular basis of “stemness.” The genes expressed in both HSC and neural stem cells very likely evolved to participate in fundamental stem cell functions, including mechanisms for regulating proliferation, differentiation and protection of the genome during lifelong cell renewal (Terskikh et al., 2001). Although the results from these gene expression studies provide some clues to the molecular basis for transdifferentiation, the mechanism remains poorly understood. With the advances in cDNA microarray technology and functional genomics, the tools are now available to address the problem systematically.
MIGRATION AND IN SITU DIFFERENTIATION OF BONE MARROW CELLS INTO A NEURONAL PHENOTYPE AFTER TRANSPLANTATION
An interesting property of BMSC is their proclivity to migrate. Two elegant studies have shown that transplanted adult bone marrow cells can migrate from systemic circulation to the brain, where they undergo transdifferentiation into neurons as well as microglia and astroglia. Adult bone marrow cells from male mice were administered systemically to females of a mouse strain (PU.1 knockouts) that lacks macrophages, neutrophils, mast cells, osteoclasts and B- and T-cells at birth (Mezey et al., 2000). These animals require a bone marrow transplant within 48 hr of birth if they are to survive. Within 24 hr after birth, PU.1 homozygous recipients were given intraperitoneal injections of unprocessed bone marrow cell suspensions (containing 107 cells) from wild-type male mice. Between 1–4 months after transplantation, marrow-derived cells were present in the brains of all the transplanted mice examined. Between 2.3–4.6% of all cells (all identifiable nuclei) were Y-chromosome positive (derived from donor bone marrow). The Y-chromosome bearing cells were evenly distributed throughout different brain regions. The Y-chromosome was present in 0.3–2.3% of neurons, marked by immunoreactivity for neuron specific nuclear protein (NeuN). In the brains of transplanted female mice, all the Y-chromosome+ neuronal nuclei also expressed NSE. These studies demonstrate that bone marrow cells administered systemically can migrate into the brain and differentiate into cells that express neuron-specific antigens.
In a similar study, another group found that bone marrow cells infused into irradiated mice migrated into the brain and differentiated into cells that expressed neuronal antigens (Brazelton et al., 2000). Adult marrow was harvested from transgenic mice that ubiquitously express enhanced green fluorescent protein (GFP). GFP-expressing (GFPpos) bone marrow was administered by tail vein (6 million cells/recipient) into lethally irradiated, isogenic mice. Brains harvested several months after the transplant and examined by light microscopy revealed the presence of GFPpos cells throughout the brain, including the olfactory bulb, hippocampus, cortical areas and cerebellum. Examination of dissociated brain and bone marrow cells from the recipients revealed that essentially all of the GFPpos cells that engrafted in the host bone marrow also expressed CD45 (surface marker of all nucleated mature blood lineages). A significant subset (up to 20%) of the GFPpos cells that engrafted in the brain, however, lacked both CD45 and CD11b (surface markers expressed by all myelomonocytic cells). These findings suggested that exposure to a brain microenvironment led a subpopulation of bone marrow-derived cells to acquire novel phenotypes. Using confocal microscopy, the researchers determined that individual cells co-expressed GFP and neuron-specific antigens. The olfactory bulb was selected for indepth quantification and revealed that 0.2–0.3% of the total number of neurons were derived from bone marrow by 8–12 weeks after transplantation. A substantial proportion of marrow-derived cells co-expressed multiple neuron-specific gene products including two neuronal proteins, 200-kD neurofilament (Nf-H) and Class III β-tubulin, but did not express the glial cell marker GFAP.
SPONTANEOUS CELL FUSION: A POTENTIAL EXPLANATION FOR THE CHANGE IN PHENOTYPE AFTER TRANSPLANTATION?
Two recent independent studies suggest that transdifferentiation, or conversion to a neural phenotype, after transplantation of Y-chromosome or GFP-labeled BMSC may be due to a cell fusion phenomenon. These reports have raised some doubt concerning the plasticity of adult tissue-derived stem cells. One of the studies demonstrated that adult murine bone marrow cells can fuse spontaneously with embryonic stem cells when co-cultured in media containing interleukin 3 (Terada et al., 2002). The resulting hybrid cells exhibited multipotentiality, an observation that mimics the phenotypic conversion of BMSC to unexpected cell lineages. The other study showed mouse neural progenitor cells, when co-cultured with pluripotent embryonic stem cells, gave rise to multipotent cells that expressed genetic markers of the neural cell. Ostensibly it would appear that the neural cells dedifferentiated into a multipotent cell line or transdifferentiated into unexpected phenotypes, but the investigators were able to demonstrate these cells were tetraploid hybrids with full pluripotent character (Ying et al., 2002). The authors of these two studies suggest that cell fusion might be responsible for the observations attributed to an intrinsic plasticity of adult tissue stem cells. This is certainly a possibility in studies where adult cells are co-incubated with multipotent stem cells in vitro. It might also occur when adult BMSC are grafted into adult brain and there is fusion of the donor marrow cells with neural progenitors located in neurogenic zones (SVZ). The results of five in vitro studies cannot be explained by cell fusion, however, because adult tissue-derived BMSC were not co-cultured with any cells (Table I). Only one in vitro study reviewed here utilized co-cultures. In the very first demonstration of neural cells generated from BMSC in vitro, the low frequency of neuron-like cells generated from BMSC in monoculture was increased by co-culturing the BMSC with fetal brain tissue(Sanchez-Ramos et al., 2000), Although cell fusion may be the mechanism for enhanced conversion of phenotype, unknown humoral factors elaborated by the fetal cells and cell–cell contact may also have played a role in the transdifferentiation process. In any case, the spontaneous cell fusion phenomenon serves to caution researchers who study plasticity of adult stem cells. Future work involving co-culture or transplant studies will need to test for the possibility of cell fusion and generation of cell hybrids.
ISOLATION AND CHARACTERIZATION OF THE SUBPOPULATIONS OF BONE MARROW CELLS CAPABLE OF DIFFERENTIATION INTO A NEURAL LINEAGE
The cell, or subpopulation of cells, that is the source of neurons and glia or that actually enters the brain and differentiates in situ into neurons remains unclear, because many studies used heterogeneous or poorly characterized populations of bone marrow cells. Some investigators made an effort to eliminate hematopoietic precursors from the marrow cells before inducing differentiation in vitro or before transplantation. Hematopoeitic stem cells (CD34+ cells from human bone marrow or Sca1+ cells from mouse bone marrow) were separated by immunomagnetic bead sorting before preparing stromal cell cultures (Sanchez-Ramos et al., 2000). Other researchers utilized fluorescent cell sorting with flow cytometry of bone marrow stromal cells after the first passage to demonstrate that the BMSC were negative for CD11b and CD45, surface markers associated with lympohematopoietic cells (Black and Woodbury, 2001). Although neural cells were differentiated from cultures depleted of hematopoietic progenitor cells in these two studies, the BMSC cultures consisted of heterogenous cell types. From these studies it was not possible to determine whether a primitive multipotent stem cell resides in adult bone marrow or whether a transdifferentiation process occurred. Subcloning by limiting dilution provides a better approach to this problem because it is assumed that all the cells are derived from a single multipotent stem cell that resides in adult bone marrow. Several clones of BMSC were generated from female mice by limiting dilution (Kohyama et al., 2001). The clones could differentiate into osteogenic, adipogenic and myogenic lineages under specific culture conditions reported earlier by other investigators (Pittenger et al., 1999). Without any treatment, the BMSC clones exhibited a fibroblastic appearance. All three clones exhibited high alkaline phosphatase activity, indicating they had an osteogenic potential. All three of the clones also expressed the following surface markers: Sca1, (but not c-kit), CD29 (Integrin1), CD140 (platelet-derived growth factor receptor α), CD44 (Pgp-1/Ly-24), and Ly6c (marker of osteoblasts). Neural differentiation was induced by incubation with 5-Aza-C (10 μM) for 96 hr and continued incubation for NGF/NT3/BDNF, as described earlier. The BMSC changed from a fibroblastic appearance to neuron-like cells. In two of the clones, 20% of the cells formed neurite-like processes and exhibited a neuron-like morphology by Day 9 after treatment. Immunocytochemistry revealed the cells were positive for neuronal markers including NeuN, Tuj1, and Hu. These cultures also were positive for markers of astroglia (GFAP) and oligodendrocytes (Gal-C). The number of positive cells differed among the three cell lines, suggesting that each cell line contained several committed progenitors that independently give rise to neurons, astrocytes and oligodendrocytes. The fact that all three clones expressed alkaline phosphatase (a marker of osteogenic potential) and yet could be induced to differentiate into neural cells after treatment with 5-Aza-C, a methylation modifier, lends support to the transdifferentiation process. This hypothesis suggests that reprogramming of the nucleus of osteoblastic progenitors is responsible for the activation of neuroectodermal genetic programs. Lineage interconversion between distinct adult committed progenitors is unexpected but not impossible. Expression of neurogenic phenotypes have been reported in bone-originated sarcomas (Ewing sarcoma), an uncommon neoplasm of bone and extraosseous tissue (Sugimoto et al., 1997; Gardner et al., 1998). These neoplastic sarcoma cells occasionally express neuroectodermal markers, such as NSE, neurofilament and nerve growth factor receptor, suggesting a relationship of Ewing sarcoma to neuroectodermal tumors (Sugimoto et al., 1997; Chung et al., 1998; Gardner et al., 1998).
Another approach for producing a homogeneous population of bone marrow-derived cells capable of generating neural lineages was described by the laboratory of Verfaillie (Reyes et al., 2002). Human bone marrow mononuclear cells obtained by Ficoll-Pague density gradient centrifugation were depleted of CD45+ and glycophorin-A+ cells, (cells of hematopoeitic lineage) by immunomagnetic sorting. After two depletion steps, the mononuclear cells were greater than 99.5% free of CD45+, GlyA+ cells. The remaining cells (0.5% of the mononuclear fraction), were propagated for at least 20 doublings, resulting in a homogeneous population of cells termed multipotent adult progenitor cells (MAPC). By 10–15 cell doublings, MAPC did not express CD10, CD31, CD34, CD36, CD38, CD50, CD62E, CD106, CD117, HIP12, fibroblast antigen, HLA-DR, class I HLA, CD45, Tie, or Tek. These cells did not change morphology and phenotype for up to 70 doublings. The MAPCs were shown to differentiate under specific culture conditions to mesenchymal cell types (cartilage, bone, fibroblasts, and adipocytes), and to cells of almost all other mesodermal lineages (skeletal, smooth and cardiac myoblasts, and endothelial cells). Most interesting, the MAPC were induced to differentiate to cells of neuroectodermal lineage, including β-tubulin-III, neurofilament, NSE and glutamate-positive neurons, GFAP+ astrocytes and galactocerebroside+ oligodendrocytes (Reyes and Verfaillie, 2001). Based on the number of steps and passages used in the enrichment procedure, the frequency of the MAPC was calculated to be extremely low (1 in 107 or 108). Their origin remains speculative, but it has been suggested that they are descendants from primordial germ cells, whose function remains unknown (Reyes and Verfaillie, 2001). It is also possible, however, that the frequent passaging resulted in the creation of a new cell line, through a poorly understood process of dedifferentiation. Verfaillie notes that the MAPC cultures stop growing, however, when they become confluent and hence do not behave like a neoplastic cell line (Reyes and Verfaillie, 2001).
BONE MARROW STROMAL CELLS FOR THERAPY OF STROKE, TRAUMA, AND PARKINSON DISEASE
Anticipating the therapeutic potential of these fundamental observations, bone marrow cells have been shown to hasten recovery of neurologic deficits in rodent models of stroke, brain and spinal cord trauma and Parkinson Disease. Researchers have demonstrated that direct intracerebral grafting and intravenous infusion of BMSC are effective in hastening recovery from the neurological deficit induced by middle cerebral artery occlusion (Li et al., 2000; Chen et al., 2001a). One of the first reports described direct transplantation of adult BMSC, prelabeled with BrdU, into the striatum after embolic middle cerebral artery occlusion (Li et al., 2000). The mice were killed 28 days after stroke. BrdU-reactive cells survived and migrated a distance of approximately 2.2 mm from the grafting areas toward the ischemic areas. NeuN was co-expressed in 1% of BrdU stained cells and the astrocytic specific protein GFAP in 8% of the BrdU stained cells. Functional recovery from a rotarod test and modified neurologic severity score tests (including motor, sensory, and reflex) were improved significantly in the mice receiving the BMSC graft. Similar functional recovery was achieved after intravenous infusion of BMSC into rats after a middle cerebral artery embolic stroke (Chen et al., 2001a). In both of these studies there were few marrow-derived cells that expressed neuron-specific markers, and the size of the infarct was not altered in animals that received the BMSC compared to stroked animals without the BMSC. It is difficult to explain the mechanism of neurologic recovery other than to suggest that these cells elaborated a number of trophic factors and cytokines that promote recovery in the host. In support of this hypothesis, Chen et al. (2000) found that a composite graft of fresh BMSC along with brain-derived neurotrophic factor (BDNF), transplanted into the ischemic boundary zone of rat brain, improved survival and differentiation of the BMSC. This combined treatment also hastened functional recovery after middle cerebral artery occlusion. These studies show clearly a therapeutic benefit in a stroke model, but the correlation between the differentiation of neural cells and recovery is poor. Other characteristics of these cells that may play a role in functional recovery from stroke remain to be explored.
Transplantation of BMSC into the spinal cord after a contusion injury was also reported to enhance recovery based on standardized assessments (Chopp et al., 2000). Rats were subjected to a weight-driven implant injury. BMSC or PBS was injected into the spinal cord one week after injury. Sections of tissue were analyzed by double-labeled immunohistochemistry for BMSC identification. Functional outcome measurements were carried out weekly, up to 5 weeks post-injury. The data showed a significant improvement in functional outcome in animals treated with BMSC transplantation compared to control animals. Scattered cells derived from BMSC expressed neural protein markers, but like the stroke studies, it is unlikely that reconstruction of injured spinal cord by differentiating marrow cells is responsible for the improvement. The enhanced recovery, according to speculation by the authors, is most likely mediated by humoral factors released by marrow cells.
Treatment of traumatic brain injury by intravenous infusion or direct intracerebral grafting of BMSC also enhanced recovery from the neurologic deficit (Mahmood, 2001; Mahmood et al., 2001). More specifically, there was a significant improvement in motor function, 14 and 28 days after transplantation, in the transplanted rats compared to control rats. Histological examination of the rat brains revealed that marrow-derived cells survived, proliferated, and migrated toward the injury site. A small proportion of the BrdU-labeled BMSC cells expressed neuronal (NeuN) and glial (GFAP) markers. Like the other studies in this series, it is unlikely that the differentiation of BMSC into neuron-like cells and glia explains the recovery.
The therapeutic potential of BMSC for the treatment of Parkinson disease was given an impetus by a recent publication from the same group of researchers (Li et al., 2001). BMSC prelabeled with BrdU were grafted into the striatum of MPTP-treated mice. The grafted MPTP-treated mice exhibited a significant improvement on the rotarod test at 35 days after transplant, compared to nongrafted controls. Immunohistochemistry revealed BrdU reactive cells in the striatum of the grafted MPTP-treated mice at least four weeks after transplantation. Scattered BrdU-reactive cells expressed tyrosine hydroxylase (TH) immunoreactivity. Although the BMSC injected intrastriatally survive, express TH immunoreactivity, and promote some functional recovery, much more work is required to understand the mechanism for this recovery. It is not known whether the grafted cells increase production of dopamine or whether other processes, such as the secretion of neurotrophic factors by the marrow-derived cells, mediate the improvement in motor function.
The conversion of bone marrow or umbilical cord blood cells into neurons and glia may be explained by several mechanisms that are not mutually exclusive and may act in concert: 1) The stem cells found in adult tissues are true multipotent stem cells that arrive in the adult tissue early in development (or perhaps migrate to the organ later in development), but retain “stemness” (self-renewal, multipotency) in the adult tissue throughout life; and 2) tissue-specific progenitor cells, with limited differentiation potential, normally designed to replace cells specific to the organ of residence, have the capacity under experimental conditions to transdifferentiate into cells characteristic of other tissues and organs. This latter explanation implies a genetic reprogramming of the nucleus of the progenitor by instructive molecular signals, similar to what occurs when a mature somatic cell nucleus is inserted into an egg to generate a totipotent cell capable of producing all cells that make up a complete organism (as in reproductive cloning). Recent in vitro observations point out a confounding phenomenon of fusion of adult tissue cells with embryonic stem cells that should raise a flag of caution when interpreting transdifferentiation experiments. Without understanding completely the molecular mechanisms responsible for adult tissue stem cell plasticity, it is still possible to move forward in the translation of these findings to clinical applications. The precedent for using bone marrow and umbilical cord blood to replace bone marrow of patients with leukemia is well established, providing additional impetus for exploring bone marrow transplantation in the treatment of neurodegenerative diseases, trauma and stroke.