Regulation of neural markers nestin and GFAP expression by cultivated bone marrow stromal cells

Stem cells have the capacity to self-replicate or to produce progeny of one or several specific differentiated cell types. They are not committed to a particular type of tissue and can therefore generate cells belonging to different cell types. Two types of stem cells have been identified: embryonic stem (ES) cells, found in the inner cell mass of the early embryo, and tissue-specific, including adult stem cells. Adult or somatic stem cells have been located in bone marrow (Bianco et al.,1999), blood (Domen et al., 1999), cornea and retina(Wu et al., 2001), brain(Davis et al., 1994), skeletal muscles (Seale et al., 2000), dental pulp(Gronthos et al., 2000), liver(Sell, 1990) and skin(Gandarillas and Watt, 1997). Much information concerning the characteristics of adult stem cells is derived from studies of hematopoietic stem cells, because they are easily isolated,abundant and widely used in the treatment of several hematologic diseases and cancers.

Stem cells located in continuously renewing tissues such as skin, gut or bone marrow, are able to regenerate or repair these tissues throughout life. However, in the non-regenerating adult central nervous system (CNS), neural stem cells (NSC) have a poor capacity to generate new neurons or oligodendrocytes to replace cells lost after injury or degeneration. In nervous system disorders in which specific neuronal cell loss occur (e.g. Parkinson's disease) transplantation of neural cells allows for the replacement of lost cells and recovery of some degree of function(Freed et al., 2001).

Embryonic stem cells are totipotent and are thus able to differentiate in any kind of cell type present during development and in adulthood. At present,there is no evidence that adult stem cells are totipotent, but some may have the capacity to differentiate into phenotypes that belong to tissues other than the one from which they originated, a property usually referred to as plasticity or transdifferentiation. A recently reported example of such plasticity is the finding that, after intravascular delivery of genetically labelled adult mouse bone marrow into lethally irradiated adult hosts, donor cells expressing neuronal markers were found in the host CNS(Brazelton et al., 2000). In vitro, a tiny fraction (2-5%) of bone marrow stromal cells cultured in the presence of EGF or BDNF express nestin, glial fibrillary acidic protein (GFAP)and neuron-specific nuclear protein (NeuN)(Sanchez-Ramos et al., 2000). The addition of dibutiryl cyclic AMP has been reported to induce the differentiation of human mesenchymal stem cells (MSCs) into early progenitors of neural cells (Deng et al.,2001).

During the development of the CNS, proliferating neuroepithelial cells express nestin (neural stem cells protein), an intermediate filament protein(Lendahl et al., 1990), which is also expressed by NSC in adult mammals and then used to identify adult neural progenitors in culture (Dahlstrand,et al., 1995). Although nestin is not a specific marker of neural stem cells because it is also transiently expressed in muscle progenitors and in some epithelial derivatives (Mokry and Nemecek, 1998), the analysis of neurospheres obtained from ES cells (which are known to be nestin-negative in vivo) demonstrates that all the cells within those spheres express nestin, suggesting that nestin expression is correlated to or is coincident with the initiation of sphere formation (Tropepe et al.,2001).

The use of MSCs in auto-graft protocols in neurological diseases necessitates the identification of the molecular events that are important for the induction of neural differentiation of MSCs. We found that the presence of serum in the culture medium represses nestin expression by rat stromal cells(rSCs). Moreover, only nestin-positive rSCs are able to form aggregates in suspension when they are transferred to NSCs culture conditions. But, when those spheres or aggregates were placed in culture conditions known to favour neural differentiation of NSCs, only modifications of cell shape were observed. In contrast, when nestin-positive rSCs were grown in co-culture with mouse neural stem cells (mNSCs), heterogenous spheres formed which, when plated on polyornithine-coated surfaces, released 40% rSCs that differentiated into GFAP-positive cells. Nestin-negative rSCs cells, when grown in the same condition, gave rise to less than 5% GFAP-positive rSCs. Nestin expression by rSCs should thus be regarded as a first step in their progression to the neural lineage.

Adult rat bone marrow was obtained from femurs and tibias by aspiration and was resuspended into 5 ml of DEM (Invitrogen, Merelbeke, Belgium)(Azizi et al., 1998). Between 100 and 200×10⁶ marrow cells were plated on 175-cm tissue culture flask in DEM/10% foetal bovine serum (Invitrogen). After 24 hours, the non-adherent cells were removed by replacing the medium. When the rSCs became confluent, they were resuspended with 0.25% Trypsin and 1 mM EDTA and then sub-cultured. To initiate nestin expression, the cultures were washed with PBS and grown in DEM/F12 medium (Invitrogen). After 3 days, the cells were fixed. rSCs were then processed for immunocytochemistry or western blotting as described below.

Green C57BL/6 mice embryos (Jackson Immunoresearch Laboratory, West Grove,PA, USA) were used as a source of mNSC. In the green mouse, GFP expression is under the β-actin promoter activity and NSC can therefore be identified by their green fluorescence (Okabe et al.,1997). The day of conception was determined by the presence of a vaginal plug (embryonic day 0). E15 striata were isolated and triturated in DEM/F12 (Invitrogen) with a sterile Pasteur pipette. The cell suspension was filtered with a 70 μm-pore filter and viable cells were estimated by trypan blue exclusion. The cells were plated (1×10⁶ cells/75-cm tissue culture flask) in DEM/F12 supplemented with epidermal growth factor(EGF, 20 ng/ml, Sigma), N2 and B27 (Neurobasal medium, Invitrogen) which are two multicomponent cell culture supplements devoid of any growth factor. When the size of neurosphere reached approximately 50 cells, they were dissociated to a single cell suspension by trituration and replated in fresh culture medium.

rSCs have been characterized by immunocytochemical labelling for CD45(Pharmingen, The Netherlands; 1:200), CD11b (Pharmingen; 1:200), Thy1.1(Chemicon, Wevelgem, Belgium; 1:200) and P75 NGF-R (Chemicon; 1:200)(Goodell et al., 1997). Fluorescence-activated cell sorting of first passage rSCs was performed with anti-CD45 and anti-CD11b antibodies. Briefly, the cells were suspended with 10 mM ice-cold EDTA (Fluka, Bornem, Belgium) and 500,000 cells were washed in 3 ml of PBS containing 1% foetal calf serum. After centrifugation, the pellet was suspended in 200 μl of the primary antibodies solutions, for 1 hour at room temperature. They were then washed 3 times in PBS containing 1% foetal calf serum and incubated with anti-mouse IgG (1:500) secondary antibody coupled to FITC (Jackson Immunoresearch Laboratory) for 1 hour at room temperature and in the dark. Before analysis using FACS, the cells were fixed by a 15-minute incubation in 1% paraformaldehyde solution. The analysis was performed with a FACSort instrument (Becton Dickinson) and the results were analyzed using the Cellquest program (Becton Dickinson).

The adipogenic differentiation of rSCs was induced by treatment with 1-methyl-3-isobutylxanthine (0.5 mM, Sigma, Belgium), dexamethasone (1 μM,Sigma), bovine insulin (0.01 mg/ml) and indomethacin (0.2 mM, Sigma). rSCs were placed in the above adipocyte induction medium for 24 days. The differentiation was evaluated by accumulation of lipid vacuoles and staining with Oil Red O (Sigma) following fixation with 4% paraformaldehyde. To induce osteocyte differentiation, the rSCs were incubated in DEM containing dexamethasone (0.1 μM, Sigma), ascorbate (0.05 mM, Sigma) andβ-glycerophosphate (10 mM, Sigma) for 12 days. A significant increase in alkaline phosphatase activity was measured with the alkaline phosphatase colorimetric test, following the manufacturer's instructions (Sigma). Chondrogenic differentiation was induced with DEM medium containing dexamethasone (0.1 μM, Sigma), sodium pyruvate (1 mM, Janssen Chemica),ascorbic-2-phosphate acid (0.15 mM, Sigma), proline (0.35 mM, Sigma), bovine insulin (0.25 μg/ml), selenic acid (6.25 μg/ml, Sigma) and linoleic acid(5.35 μg/ml, Sigma). Chondrocytes were obtained when rSCs were grown as a pellet in the induction medium for 20 days. The cell aggregates were fixed with 4% paraformaldehyde, paraffin-embedded, cut at 5 μM sections and stained with toluidine blue.

After being induced to express nestin, rSCs were trypsinized and suspended in DEM/F12 containing N2 and B27 supplements for 24 hours. During this time,the cells aggregated. These aggregates were plated onto polyornithine-coated dishes for 5 days in the same medium and were then processed for immunocytochemistry as described below.

Nestin-positive rSCs were trypsinized and were co-incubated with GFP-positive mNSC (1×10⁶ mNSC and 1×10⁴rSCs) for 48 hours in DEM/F12 containing EGF, N2 and B27 supplements. During this time, the cells aggregated, forming heterogenous spheres. For a good observation of the presence of rSCs into the heterogenous spheres, the rSCs were colored in red with the DiD Vybrant™ cell-labelling solutions(Molecular Probes) following the manufacturer's instructions. These spheres were plated on polyornithine-coated dishes for 5 days, in DEM/F12 containing N2 and B27 supplements and were processed for immunocytochemistry as described below. Nestin-negative rSCs were trypsinized and directly replated with GFP-positive mNSC onto polyornithine-coated dishes for 5 days following the same culture condition as described for the nestin-positive cells. The cells were then processed for immunocytochemistry.

The cultures were fixed with 4% (v/v) paraformaldehyde for 15 minutes at room temperature and washed 3 times in TBS buffer. They were then permeabilized in 1% Triton X-100 (v/v) for 15 minutes and washed 3 times in TBS buffer. Non-specific binding was blocked by a 1-hour treatment in TBST(TBS buffer with 0.1% Tween) containing fat-free milk powder (30 mg/ml). The cells were then incubated for 1 hour at room temperature with either anti-p75/NGF-R, or anti-Thy1.1, or anti-nestin (Rat401, Pharmingen; mouse IgG,dilution 1:1500), or anti-GFAP (Dako; mouse IgG, dilution 1:500), or anti-M2(Developmental Studies Hybridoma Bank; rat IgG, dilution 1:500), or anti-GLAST(Shibata et al., 1997) (rabbit IgG, 1:4000) primary antibodies (diluted in blocking buffer). After 3 washes,cells were incubated in FITC- or Cy5-conjugated anti-mouse IgG (Jackson Immunoresearch; 1:500) or rhodamine-conjugated anti-rat IgG (Jackson Immunoresearch, 1:500) for 1 hour at room temperature and in the dark. The nuclei were stained with ethidium homodimer (0.2 μM, Sigma). The preparations were then mounted in Fluoprep™ (bioMérieux, Marcy L'Etoile, France) and observed using a Bio-Rad MRC1024 laser scanning confocal microscope.

Total protein extracts were obtained from confluent cells that had been cultured in the different media. The cells were harvested by scraping the dish in 500 μl lysis buffer (0.6 M KCl, 5 mM EGTA, 5 mM EDTA, 1% Triton X-100 and 1 mM PMSF in PBS). Extracts were then fractioned into pelletable(insoluble) and non-pelletable (soluble) proteins by centrifugation at 30,000 g for 15 minutes at 4°C. The pellet was resuspended in loading buffer (glycerol 10% v/v; Tris 0.05 M, pH 6.8; SDS 2%, bromophenol blue and 2.5% v/v β-mercaptoethanol) and the suspension was centrifuged at 30,000 g for 15 minutes at 4°C. The supernatant was used for protein concentration measurement using the RC DC Protein Assay(Bio-Rad). The same protein quantities in each lane were separated by electrophoresis using the Phastgel 4-15% SDS gradient (Amersham Pharmacia Biotech) and transferred to a PVDF membrane. The membranes were saturated with 3% gelatin (BioRad), incubated for 1 hour with a monoclonal antibody against rat nestin (Pharmingen, 1:1,000) or β-actin, as control for protein loading (Sigma, 1:5000) at room temperature and then washed several times with PBS-0.1% Tween. The membrane was then incubated in biotinylated goat anti-mouse antibody (Boehringher Mannheim; 1:5000) for 1 hour at room temperature. After several washes in PBS-0.1% Tween, the membrane was incubated with peroxidase-coupled streptavidin (1:100,000, Sigma) at 37°C for 1 hour. Radial glial cells were used as a positive control for nestin expression.

DNA content per cell was determined by FACS analysis after staining the cells with propidium iodide. After 5 days of co-culture, rSCs and mNSCs were trypsinised and fixed into 70% ethanol, at 4°C during 16 hours. The cells were stained with propidium iodide (400 μg/ml, Sigma) just before FACS analysis.

Rat stromal cells were isolated from the femoral and tibial bones of adult rats and propagated in culture. rSCs have been reported to be CD45- and CD11b-negative, but Thy1.1- and P75 NGF-R-positive(Goodell et al., 1997). Fluorescent cell sorting at passage 1 demonstrated that the cells were negative for CD11b (Fig. 1A)and CD45 (Fig. 1B) – two cell surface markers associated with lymphohematopoietic cells. In contrast,immunofluorescent labeling demonstrated that 100% of the rSCs were positive for Thy1.1 (Fig. 1C) and p75 NGF receptor (Fig. 1D). Classically, SCs can differentiate either into adipocytes, or chondrocytes and osteocytes (Pereira et al.,1995; Kuznetsov et al.,2001; Prockop,1997; Majumdar et al.,1998; Pittenger et al.,1999). When our putative rSCs fraction was placed in adipogenic,osteogenic and chondrogenic induction media, adipocytes, osteocytes and chondrocytes, respectively, were obtained(Fig. 1E-G).

To induce neural differentiation of rSCs in long-term cultures, we placed them in a classical NSC culture medium (DEM/F12 supplemented with N2 and B27). After 72 hours most rSCs expressed nestin as revealed by immunocytochemistry(Fig. 2A). When rSCs are placed in DEM/F12 (Fig. 2B),DEM/F12+N2 (Fig. 2C) and DEM/F12+B27 (Fig. 2D), the same result is obtained, suggesting that the removal of serum from the culture medium was actually responsible for the induction of nestin expression by rSCs. These results were confirmed by western blotting(Fig. 2F). In all of these experiments, no alteration in the morphology of nestin-positive SCs was observed. We demonstrated that the absence of serum was necessary but not sufficient for the induction of nestin expression. Indeed, the number of passages of rSCs in vitro also regulates their ability to express nestin when grown in serum-free condition (Fig. 3A-E): a minimum of ten passages is needed before rSCs are able to express nestin when placed in serum-free culture conditions. However, the capacity of these cells to differentiate into adipocytes or osteocytes did not change as a function of the number of passages(Fig. 3F,H).

Recently, Tropepe et al. (Tropepe et al., 2001) suggested that nestin expression by ES cells is correlated with the capacity to form neurospheres and that ES cells have to go through a nestin-expression stage before differentiating into neural cells. When nestin-positive and nestin-negative rSCs were trypsinised and replaced in an NSC growth medium (DEM/F12, N2 and B27) in non-adherent conditions, we observed that passage 15 nestin-positive rSCs aggregated in suspension(Fig. 4A). In contrast, passage 4 nestin-negative rSCs which had been cultivated in serum-free conditions for 48 hours prior to being transferred to NSCs growth medium, remained in suspension and did not form spheres or aggregates(Fig. 4B). Passage 15 rSCs that did not express nestin (because they had been cultivated in serum-containing medium) adhered spontaneously to the dish(Fig. 4C). We then plated the nestin-positive rSCs aggregates on a polyornithine-coated surface for 5 days as is done for neurospheres in order to stimulate cell differentiation. The cell morphology changed from the flat and elongated shape of MSCs(Fig. 4D) to more rounded morphology (Fig. 4E), but immunocytological labelling with antibodies recognizing astrocytic (anti-GFAP,anti-GLAST), oligodendroglial (anti-O4, anti-A2B5) and neuronal markers(anti-NeuN, anti-NSE, anti-Tuj1, anti-MAP2B) were all negative (data not shown). Moreover, after 5 days in these conditions, nestin expression decreased from 80% to 15% (Fig. 4F).

As nestin expression by rSCs appears insufficient to stimulate their differentiation into neural cells, we have co-cultivated nestin-positive rSCs with NSCs because the neural differentiation of MSC has been observed mainly in vivo when cells had been grafted in newborn mice brain. When we co-cultivated nestin-positive red-labelled rSCs for 5 days with NSCs obtained from E15 `green mouse' striata (Okabe et al., 1997), we observed that nestin-positive rSCs formed heterogenous spheres (Fig. 5M)with the NSC. These heterogenous spheres were then transferred on polyornithine-coated dishes for 5 days to allow cellular differentiation. Immunological labeling revealed the differentiation of 40±2.39%nestin-positive rSCs into GFAP-expressing cells (n=3, representing∼3000 cells) (Fig. 5A-C). A similar percentage of GLAST-positive rSCs was also observed in these conditions (Fig. 5G-I). However, no nestin-positive rSCs expressed neuronal (anti-NeuN, anti-NSE,anti-Tuj1, anti-MAP2B) or oligodendroglial markers (anti-O4, anti-A2B5) (data not shown). Moreover, passage 4 nestin-negative rSCs were unable to form heterogenous spheres, and when plated directly with mNSC onto a polyornithine-coated surface, only 4% of them expressed GFAP(Fig. 5D-F). Three justifications could be formulated to explain the GFAP expression by rSC: 1) a decrease of GFP expression in mNSCs, 2) a rSC-mNSC fusion event, and 3)phenotypic plasticity of the rSCs. To exclude a GFP downregulation, we cultivated GFP-positive mNSC alone in the same conditions, and a GFAP immunoreactivity was observed within GFP-positive cells(Fig. 5J-L). To exclude a cell fusion process, we first used antibody M2, which specifically recognizes mouse-specific astrocytes (Lagenaur and Schachner, 1981) to demonstrate that the GFP-negative astrocytes that developed under these conditions were of rat origin: no GFP-negative,GFAP-positive cells were recognized by the M2 antibody, but the GFAP- and GFP-positive cells were (Fig. 5N,O). Furthermore, we analyzed the ploidy of the co-cultivated cells in order to determine if rSCs adopt the astroglial phenotype of the recipient cells by a hypothetical cell fusion event. We compared DNA ploidy of the rSCs maintained in co-culture with mNSC for 5 days(Fig. 5P) with the DNA ploidy of rSCs and mNSC cultivated separately for the same period, and we found the same DNA content in the three cultures, thus excluding a cell fusion event.

Cellular therapies are promising approaches in the treatment of several neurological diseases such as Parkinson's disease(Isacson et al., 2001) or Huntington's disease (Dunnett,2000), but also for spinal cord injury(Hall, 2001), and, in association with gene therapy, the treatment of glioma(Armstrong et al., 2000; Isacson and Sladek, Jr, 1999). One main problem concerns the origin and nature of the cells to be used for such procedures. The ideal cell should exhibit several key properties,including: (1) a high level of proliferation in vitro, allowing the production of a large number of cells from a minimal amount of donor material, (2) a good control of this proliferative activity in vivo, and (3) a phenotypic plasticity allowing the differentiation into appropriate neuronal or glial phenotype. The cells used for transplantation into patients with neurological disorders have hitherto been derived from the foetal human CNS. For example,the cells grafted in Parkinson's disease patients have been obtained from foetal donors. They are isolated as post-mitotic neurons from foetal mesencephalic tissue in which dopaminergic neurons are normally found. This approach has several pitfalls: a limited number of cells is available, grafted patients need to be immunosuppressed, the material is not easily available and there are a number of ethical problems(Freed et al., 2001).

Several recent reports suggest that bone marrow stromal cells could be a non-embryonic or non-foetal source of stem cells suitable for cell replacement strategies in the treatment of CNS disorders(Kopen et al., 1999; Akiyama et al., 2002a; Akiyama et al., 2002b; Brazelton et al., 2000). An understanding of the molecular regulation of such a `mesenchymal-neural'transition may be very important when considering the use of SCs in the treatment of CNS disorders.

In our study, rSCs have been isolated from adult rat bone marrow using a differential adhesion procedure and selected by their capacity of rapid proliferation. We demonstrate that these stromal cells are able to express nestin. This expression is dependent on two factors. The first one is the absence of serum-derived components in the culture medium. Indeed, in our study using western blot analysis and cell counting, higher levels of nestin expression were found in serum-free conditions, although RT-PCR and western blot demonstrate a faint signal for nestin even in the presence of serum. The second factor is the number of cell passages. A minimum of ten passages is required for nestin expression by 75% of rSCs. However, the capacity of rMSC to differentiate into adipocytes, osteocytes or chondrocytes does not change as a function of the number of passages. This requirement of 10 passages(which correspond to 25 doubling populations) in vitro was reproducible in 4 independent experiments. Recently, Jiang et al.(Jiang et al., 2002),demonstrated that rare cells within murine bone marrow MSC cultures can differentiate not only into the mesenchymal lineage cells but also into endothelium, ectoderm and endoderm. These rare cells, which have been named as multipotent adult progenitor cells (or MAPCs), can be expanded for more than 100 population doublings, bringing about an enrichment in MAPCs in MSCs culture. The increase in the percentage of nestin-positive MSCs as a function of the number of passages could possibly be explained by the presence of these MAPCs in our cultures and their increase in number with additional passages.

We show that only rSCs from nestin-positive cultures are able to form clusters or aggregates in the non-adherent conditions used to cultivate NSCs. Nestin expression by rSCs and their ability to grow in suspension in such defined culture conditions bring them nearer to the NSCs phenotype. However,when nestin-positive rSCs spheres were plated onto an adherent surface, no glial and/or neuronal differentiation was observed. It seems that the complete neural differentiation of MSCs observed in vivo, may require the involvement of several induction signals which have not been reproduced in vitro. It is for this reason that rSCs and NSCs were co-cultivated. It was hoped that such an experiment would reproduce in vitro some of the complex molecular interactions that are required to induce a full neural differentiation of mesenchymal cells in vivo.

Under these co-culture conditions, nestin-positive rSCs were able to express GFAP, the astroglia-specific intermediate filament protein, but also GLAST, another marker of astroglia(Shibata et al., 1997). When nestin-negative rSCs were co-cultivated with NSCs under identical conditions,a low percentage of GFAP-positive cells of mesenchymal origin was observed,suggesting that nestin expression is a prerequisite for rSCs differentiation into a GFAP- or a GLAST-positive cell type. This observation is in agreement with the notion that an ordered succession of stimuli is needed to promote such a differentiation.

Given the fact that it has been reported that 0.2-1 of ES cells per 10⁵ bone marrow cells can fuse with an other cell-type and thus mimic a differentiation and/or plasticity(Terada and al., 2002), we have excluded this possibility by two methods: first, we could not find any mouse-specific antigenic labeling on GFAP-positive cells of rSCs origin, and we exclude a tetraploidy in our co-culture by FACS analysis. So,GFAP-expression by rSCs are independent of cell-fusion events or GFP downregulation in mNSC, but well a neural phenotypic plasticity of rSCs.

In conclusion, nestin expression by rSCs should be regarded as a first step in their progression to the neural lineage. A better knowledge of the regulation of their differentiation into astrocytes and the definition of appropriate culture conditions to obtain their differentiation into neurons and/or oligodendrocytes is still needed before considering MSCs as an appropriate cellular material to be used for cell replacement therapies in CNS disorders.

This work was supported by the grant of the Fonds National de la Recherche Scientifique (FNRS) of Belgium, by the Fondation Medical Reine Elisabeth(FMRE), by the Fonds Charcot, and by the Belgian League against Multiple Sclerosis. S.W. is a research fellow of the Télévie (FNRS). B.R. and P.L. are senior research associate and research associate of the FNRS,respectively. We thank Dr André Gothot (Department of Haematology,University of Liège, C.H.U. Sart Tilman, Belgium) and Bernard Coomans(Center for Cellular and Molecular Neurobiology, University of Liège,Belgium) for help with the FACS analysis, Dr Gary Brook (Department of Neurology, RWTH, Aachen, Germany) for his comments during the preparation of the manuscript, and Dr Tanaka (Department of Neurological Diseases, Kodaira,Japan) for the gift of anti-GLAST antibody.