To isolate human neural stem cells using adherent condition as used in 10 , human spinal cord was rapidly processed as previously described 13 and seeded with neural stem cell medium containing serum and EGF/FGF. Under these conditions, cells with a mesenchymal morphology (Figure 1A) grew rapidly and could be propagated for approximately 3 months (approximately 25 divisions and 9-10 passages) before undergoing proliferation decrease, probably as a consequence of cellular senescence (Figure 1B). The cells were characterized at passage 4 by QPCR and immunofluorescence. We first measured the expression of genes typically expressed by cells of the neural lineage, i.e astrocytes (Aldh1l1, Gfap), oligodendrocytes (Cnp, Mbp, Plp1, Sox10, Ugt8), oligodendrocyte progenitor cells (Ascl1, Nkx2.2, Nkx6.1, Olig1/2, Pdgfrα/β), neurons (Dcx, Nefl) and stem cells (Fabp7/Blbp, Nestin, Prominin/CD133, Sox2). Surprisingly, although these cells were derived from the CNS, we found that the expression of these genes was either low or undetectable (Figure 1C, D) and only Nestin, Cnp and PdgfRα/β could be detected with less than 25 cycles of QPCR. We confirmed by immunofluorescence the absence of detectable expression of Sox2, GFAP and Olig2 which are well documented markers for neural stem cells and glial cells (Figure 2A). We then questioned the presence of markers frequently expressed in mesenchymal cells (Acta2, CD44, Col1A1, Fn1, Vim, Tgfβ1), bearing in mind that these can also be found expressed in cells of the neural lineage, and mesenchymal transcription factors (FoxC2, Snai1/2, Twist1, Zeb1/2). In sharp contrast with the neural lineage genes, the expression of these genes was readily detected (Figure 1C, D). We confirmed by immunofluorescence and western blot that Snai2 and Twist1 were expressed at the protein level (Figure 2A, B). These two transcription factors were detected in the nuclei of most cells (>80%) (Figure 2A).

The possibility existed that the isolated cells were derived from the in vitro differentiation of endogenous spinal neural precursor cells which would adopt a mesenchymal phenotype when cultured in serum-containing medium. To explore this hypothesis, we cultured human fetal brain neural stem cells in this medium for three passages and then explored the phenotype of these cells by QPCR and immunofluorescence (Figure 1C, D, 2A). In contrast with cells isolated from the adult spinal cord, our immunofluorescence analysis revealed that these cells expressed a high level of several typical neural markers such as GFAP, Sox2, Olig2 whereas Snai2 and Twist1 were not detectable (Figure 2A). QPCR confirmed these results and further indicated the maintenance of several typical markers for neural cells (for instance CD133/Prom1, Dcx, PLP1 respectively for stem, neuronal and oligodendrocytic cells) even when cultured in serum-containing media (Figure 1C, D).

The high expression of transcripts for smooth muscle actin (Acta2) in spinal cord cells suggested that the culture contained a majority of vascular muscular cells, an assumption consistent with the detection of Nestin and PdgfRα/β (Figure 1C), two markers also expressed by vascular cells 14 15 . In order to confirm this possibility, we explored by immunofluorescence the expression of Acta2 and Nestin, together with Calponin, Caldesmon and NG2, three other markers of vascular muscular cells. The vast majority of the cells (>80%) were positive for these five proteins, confirming that these cultures mostly contained vascular muscular cells (Figure 3A-E).

One unexpected transcription factor detected in the culture by QPCR was Nkx6.1 (Figure 1D). This homeodomain protein is implicated in the ontogenesis of oligodendrocytes and of pancreatic cells 16 17 but its expression in vascular cells has not been reported in adults. The presence of Nkx6.1 protein was confirmed by immunofluorescence and western blot in these cultures (Figure 2A, B). The vast majority of the cells (>80%) showed a clear nuclear staining and these cells coexpress Calponin, Caldmesmon and NG2 (Figure 3C, D, E). In contrast, Nkx6.1 was not detected in human fetal neural stem cells cultured in the same conditions for three passages (Figure 2A). The expression of Nkx6.1 by vascular cells was then confirmed in vivo on adult human spinal cord sections. Nkx6.1⁺ cells were detected around large vessels in the meninges and in the parenchyma (Figure 4A, B). As previously reported 13 , Nkx6.1 was also expressed by ependymal cell of the central canal (data not shown). We concluded from these data that adherent cultures derived from adult human spinal cord led to the isolation of a cellular population mainly composed of vascular muscular cells that express the homeodomain protein Nkx6.1 and the mesenchymal transcription factors Snai2 and Twist1.

It has recently been reported that adult stem cells with mesenchymal features located in various tissues could behave as multipotential stem cells 11 18 19 20 21 . These cells could give rise to cells of several lineages including neurons and myelinating cells. We thus explored the multipotentiality of Nkx6.1⁺ cells by placing them in different media and culture conditions. First, by culturing them in non adherent conditions in defined neural stem cell medium, these cells were not able to form neurospheres, a hallmark of neural stem cells. Second, we further examined the possibility of a neural differentiation by labelling the cells with a GFP-expressing lentivirus, then by coculturing them for 2 weeks with human young neurons, obtained by differentiation of human fetal brain neural stem cells. In these conditions, we could not detect staining of GFP-labelled cells (100 individual cells examined) with neuronal (Map2ab), astrocytic (GFAP) or oligodendrocytic (Olig2, Nkx2.2) markers, strongly indicating the absence of neural differentiation of these cells (Figure 5). Third, we treated the cells with BMP2, BMP4, BMP7, CNTF, CNTF/BMP4, LIF, LIF/BMP4 or T3 to promote an astrocytic 22 23 or an oligodendrocytic fate 24 respectively, but again no differentiation was observed (data not shown).

Considering the mesenchymal features of these cells and as vascular muscular cells have been reported to be endowed with stem cell properties 18 20 we next examined the capacity of Nkx6.1⁺ cells to generate chondrocytes, adipocytes and osteoblasts, three lineages commonly produced by mesenchymal stem cells. The cells were placed in adequate media, and the differentiation was monitored by staining with Oil red O (adipogenesis), Alizarine red S (osteogenesis) and Alcian blue (chondrogenesis) and by QPCR for markers for adipocytes (Lpl, Fabp4, PPARg), chondrocytes (Col2b, Col10, Aggregan, Mmp13, Comp, Sox9) and osteoblasts (SPARC, Osteocalcin, Runx2, phosphatase alkaline). As a control, the differentiation of human mesenchymal stem cells isolated from bone marrow was monitored in parallel. Under these conditions, no differentiation towards chondrocytes or adipocytes was observed (not shown). Only deposition of calcium phosphate precipitates, as observed with Alizarine red S staining (Figure 6A), could convincingly be detected. This was accompanied by a sharp increase of alkaline phosphatase mRNA (Figure 6B) and activity which was detected with Naphthol ASBI phosphate and Fast Red Trisodium staining (Figure 6C). No increase of the other osteoblast markers compared to that obtained with bone marrow mesenchymal stem cells could be demonstrated (not shown) suggesting absence of complete differentiation towards osteoblasts. Equally, we could not detect the expression of Nkx6.1 by western blot (Figure 2B) and immunofluorescence (Figure 3F) in bone marrow mesenchymal stem cells, suggesting that CNS Nkx6.1⁺ vascular cells are weakly related to these cells. We also examined the capacity of human fetal neural stem cells cultured in the presence of serum to mineralize in the same conditions. Figure 6A, C shows however that these cells were unable to do so.

Finally, to ascertain that mineralization was produced by Nkx6.1⁺ cells and not by other contaminating cell types, we performed single cell cloning in 96 wells to isolate Nkx6.1⁺ cell clones which were then placed in mineralizating conditions. Among 34 isolated Nkx6.1⁺ clones, 95% of them were able to mineralize (Figure 6D).

We concluded from these studies that Nkx6.1⁺ vascular muscular cells isolated from the adult CNS do not behave as neural or mesenchymal stem cells but can readily mineralize in vitro.

Human spinal cord was obtained from one organ-donor patient (40 years old, accidental death) in strict agreement with the French bioethics laws (articles L1232-1 to L1232-6) and after approval by the French institution for organ transplantation. An informed consent from the family was obtained by the organ procurement organization (OPO) for this study. A sample of the lumbar spinal cord was removed and rapidly processed for cell culture as previously described 13 . Cells were cultured at approximately 4 × 10³ per cm² cell culture flasks in the presence of DMEM/F12 with L-glutamine complemented with N2 supplement (Fischer Scientific, http://www.fishersci.com), 35 μg/ml of bovine pituitary extract (Fischer Scientific), 5% of certified fetal calf serum (Fischer Scientific), 20 ng/ml of human EGF and FGF2 (Peprotech, http://www.peprotech.com), heparin (2 μg/ml), ciprofloxaxine (2 μg/ml), gentamycin (10 μg/ml) (Fischer Scientific), fungizone (0.25 μg/ml) (Fischer Scientific), fungin (10 μg/ml) (Invivogen, http://www.invivogen.com). Media was replaced twice a week and cells were passaged at 80-90% confluency using a 3 min dissociation step at 37°C with Trypsin/EDTA (Fischer Scientific) containing 0.25% Trypsin. Cells were frozen in complete media containing 10% DMSO. For cloning experiments (Figure 6D), the cell culture was dissociated and single cell seeding in 96 well plates was performed using an Aria cytometer (BD, http://www.bdbiosciences.com) equipped with an automated single cell deposition device. After 3 weeks, each clone was split in two and expanded to perform Nkx6.1 staining and osteogenic differentiation.

To assess the capacity of human adult spinal cord cells to differentiate into oligodendrocytic, astrocytic or neuronal cells, cells were seeded on differentiated human fetal brain neural stem cells. The latter were obtained by growing human fetal brain neural stem cells (16 weeks of gestation, Lonza, http://www.lonza.com) as neurospheres then by differentiating them on laminin-coated coverslips to generate young neurons and glial cells according to the manufacturer's protocols. Spinal cord human cells were pre-labelled using an infection with a GFP-expressing lentivirus to distinguish them from fetal brain cells. Cells were co-cultured for 1-2 weeks before assessing their differentiation into glial and neuronal cells using neural lineage markers (Map2ab, Gfap, Olig2, Nkx2.2). To assess the capacity of the spinal cord cells to differentiate into astrocytes, cells were treated by BMP2 (10 ng/ml), BMP4 (10 ng/ml), BMP7 (10 ng/ml), LIF (10 ng/ml), LIF + BMP4, CNTF (10 ng/ml) or CNTF+BMP4 for 4 days before staining for GFAP. For oligodendrocytic differentiation, cells were treated for one week with 15 nM Triiodo-L-Thyronine (T3) (Sigma, http://www.sigmaaldrich.com) + N2 supplement (Fischer Scientific, http://www.fishersci.com) or Neurobasal media with B27 and vitamin A, before staining for GalC and O4 antigens.

As a control, the differentiation of human mesenchymal stem cells isolated from bone marrow as described in 35 was monitored in parallel with spinal cord cells and human fetal neural stem cells. For adipogenic differentiation, cells were plated at a density of 8 × 10³ cells/cm² in DMEM-F12 containing 5% newborn calf serum, 1 μM dexamethasone, 50 μM isobutyl-methylxanthine and 60 μM indomethacin. Stimulation was carried out for 3 weeks with the medium changed every 2-3 days. Adipocyte differentiation was confirmed on day 21 by visualization of lipid droplets by oil red O staining and QPCR using primers as described in 35 . For chondrogenic differentiation, cells were cultured in 15 ml polypropylene conical tubes with 2.5 × 10⁵ cells, after centrifuging for 5 min at 600 g. The resulting pellets were cultured in 500 μl DMEM supplemented with 0.1 μM dexamethasone, 0.17 mM ascorbate-2 phosphate, 1% insulin-transferrin-sodium selenite supplement and 10 ng/ml of recombinant TGFβ3 (R&D Systems). Media were changed every 2-3 days. Differentiation was assessed on samples on day 21 by QPCR using specific primers 35 . Osteogenesis was induced in spinal cord cells and human fetal brain neural stem cells by culturing at low density (2.5 x10³ cells/cm²) in osteogenic medium consisting of DMEM (high glucose) with 10% fetal calf serum, 10 mM β-glycerophosphate, 0.1 μM dexamethasone, 0.05 mM ascorbic acid (Sigma) and NaH₂PO₄ 3 mM. Medium was changed every three days for 2 weeks. Osteoblast differentiation was assessed on day 14 by visualization of matrix calcification by Alizarin red S staining and QPCR as previously described 35 . Phosphatase alkaline activity detection by naphthol ASBI phosphate and Fast Red Trisodium staining was performed as previously described 36 . In this condition, cells were cultured for 2 weeks in osteogenic medium but without NaH₂PO₄.

For paraffin sections, human lumbar spinal cord samples were fixed in formaldehyde-acetic acid-methanol (10%-10%-40% v/v) for 72 h, embedded in paraffin, then processed for antigen demasking by boiling for 40 min in EDTA 10 mM pH 7.2. Immunoperoxydase detection (diaminobenzidine, Dako) was performed as previously described 13 on 4 μm paraffin sections. Immunodetection was performed on cell cultures as previously described in 37 . Nuclei were stained with 2 μg/ml Hoechst 33242. The following primary antibodies were used: Acta2 (1:100, mouse monoclonal, 1A4, Sigma), Caldesmon (1:250, rabbit monoclonal, Abcam AB32330), Calponin (1:250, rabbit monoclonal, Abcam AB46794), GFAP (1:1000, rabbit polyclonal, Dako Z0334), Map2ab (1:500, mouse monoclonal HM-2, Sigma), Nestin (1:1000, rabbit polyclonal, Chemicon AB5922), Nkx2.2 (1:40, mouse monoclonal 74.5A5, Developmental Biology Hybridoma Bank, University of Iowa), Nkx6.1 (1:100, mouse monoclonal F55A10, Developmental Biology Hybridoma Bank), Olig2 (1:500, rabbit polyclonal, IBL 18953), Snai2 (1:50, rabbit monoclonal, Cell Signalling C19G7), Twist1 (1:2, mouse monoclonal, Abcam AB50887). Cy3 or Cy5 conjugated-secondary antibodies (1:500, Invitrogen) were used.

RNA extractions from spinal cord cells (passage 4) and human fetal brain neural stem cells (cultured in spinal cord cell medium for three passages) were performed using kits from Qiagen http://www.qiagen.com. Quantitative real-time RT-PCR was performed as described previously 38 . PCR was performed using the SYBR Green PCR Core Reagents Kit (Perkin-Elmer Applied Biosystems). The thermal cycling conditions comprised an initial denaturation step at 95°C for 10 min and 50 cycles at 95°C for 15 s and 65°C for 1 min. Experiments were performed with duplicates for each data point. Primer sequences are provided as additional file 1, Table S1. We quantified transcripts of the TBP gene (TATA binding protein) (Figure 1) or GAPDH (Figure 6) as the endogenous reference RNA control. Results, expressed as n-fold differences in target gene expression relative to the reference gene (termed Ntarget), were determined with the following formula: Ntarget = 2^ΔCt sample, where the ΔCt value of the sample was determined by subtracting the average Ct value of the target gene from the average Ct value of the reference gene.

Protein preparation and Western blot procedure were described in 39 . Precision Plus Protein WesternC Standards were used (Bio-Rad).