As the precise location of NSCs in the adult brain remains controversial9, 10, 11, cells were gathered from the two areas with the reported highest NSC content: the ependymal and subventricular zones of the lateral ventricular walls. To identify the putative NSCs within these populations, we examined single-cell suspensions for the expression of a variety of cell surface markers and characteristics with a fluorescence-activated cell sorter (FACS). The frequency of NSCs in sorted sub-populations was determined by the ability of individual cells to generate multipotent neurospheres1, 12, 13.
Initial NSC enrichment was accomplished by sorting viable cells on the basis of size as assessed by forward light scatter (Fig. 1a). Approximately 80% of the total NSC activity in the unsorted population was found in cells >12 
m in diameter (>90 units forward scatter; Table 1). The next enrichment step was achieved by sorting for differential peanut agglutinin (PNA) binding, which has previously been used to fractionate murine haematopoietic stem cells14. From the FACS profiles, three populations were examined: PNAhi, PNAlo and PNAmid (Fig. 1a). The >12-m PNAlo fraction, representing 1.73 
0.75% (mean s.e.m.; n = 3) of the unsorted population, contained the greatest NSC activity with about 1 in 7 cells giving rise to a neurosphere (Table 1). Whereas numerous antigens were detected on the PNAlo population, including CD30, CD34 and CD90.2, heat-stable antigen (HSA, mCD24a) was found to most effectively enrich sub-populations of PNAlo cells with NSCs. The PNAlo HSAlo sub-population (Fig. 1b, c), a discrete population representing 0.27 0.07% of the unsorted population, comprised NSCs almost exclusively. When cultured under clonal conditions, where one PNAlo HSAlo cell was plated per well, about 80% of the cells (1:1.28) gave rise to neurospheres (Table 1). Furthermore, the PNAlo HSAlo population contained 63.2% of the NSC activity present in the unsorted population, suggesting that the great majority of NSCs in the periventricular region are of this phenotype.
Figure 1: Flow cytometric purification of NSCs.
a, Viable CNS cells sorted on forward light scatter (<7, 7–12, >12 m) and PNA expression (PNAlo, PNAmid and PNAhi); NSC activity was predominantly found in the >12 m PNAlo cells (boxed). b, Cells >12 m were next sorted on the basis of HSA immunoreactivity; NSC activity was predominant09ly found in the PNAlo HSAlo cells (boxed). c, Contour plot of boxed region in b shows antigenic expression of this population in more detail.
To determine whether all PNAlo HSAlo cells had similar biological potential, clonally derived spheres from PNAlo HSAlo cells (Fig. 2a) were differentiated and assessed by immunocytochemistry for the presence of neurons and glia. All spheres (501 examined) contained astrocytes positive for glial fibrillary acidic protein (GFAP+), neurons positive for 
-tubulin type III, and oligodendrocytes positive for O4 (Fig. 2b). Indeed, clones derived from PNAlo HSAlo cells possessed all the characteristic in vitro properties, such as the ability to self renew (data not shown), of previously described NSCs1, 6, 13.
Figure 2: PNAlo HSAlo adult NSCs give rise to neural and muscle cell types.
A single PNAlo HSAlo cell generates a neurosphere (a) that, when transferred to differentiating conditions, has progeny (b) including neurons (red, -tubulin type III), astrocytes (blue, GFAP) and oligodendrocytes (green, O4). When co-cultured with C2C12 cells, freshly isolated PNAlo HSAlo GFP+ NSCs assume a spindle-like morphology (c) often with multiple nuclei (DAPI+, arrow, d), suggestive of muscle cells. Myogenic conversion of NSCs confirmed by anti-fast MyHC type II (red, e) and anti-
-actinin-2 (red, f) immunoreactive cells containing GFP+ nuclei. In many cases, immunostaining procedures reduced the intensity of cytoplasmic GFP to below detectable levels, thus GFP expression was found predominantly in the nucleus. When injected into C57BL/6J pups in utero, donor-derived (-gal+, green) PNAlo HSAlo NSCs generated NeuN+ (red, g) neurons and GFAP+ (red, h) astrocytes. Scale bars: a, 80 m; b, c, e, 40 m; d, g, h, 20 m; f, 15 m.
To further characterize the PNAlo HSAlo NSCs, we examined the expression of other neural markers. As expected, all PNAlo HSAlo NSCs examined (4,836 cells) expressed nestin, a marker of putative NSCs15, and were negative for other differentiated cell markers (-tubulin type III, O4 and GFAP). The lack of GFAP was of interest, as it has been shown that GFAP-expressing cells in the subventricular zone gave rise to neurospheres in vitro10. GFAP expression, however, was determined using an adenoviral vector expressing green fluorescent protein (GFP) under the control of a GFAP promoter, which may be far more sensitive than immunocytochemistry.
PNAlo HSAlo NSCs were found to lack cilia (data not shown) and also lacked the expression of the ependymal marker HSA10, 16, 17, which suggests that the PNAlo HSAlo NSCs reside in the subventricular zone. However, when the ependymal cell layer was labelled with DiI (ref. 11), FACS analysis revealed that 32.1 4.3% of PNAlo HSAlo cells were DiI positive. Furthermore, the PNAlo HSAlo DiI+ population contained equal activity of stem cells to that of the PNAlo HSAlo DiI-, whereas other DiI+ populations contained extremely low stem cell activity (<0.001%). Thus, the most parsimonious interpretation is that there is a common NSC that resides, or has access to, both the subventricular and ependymal zones, thereby resolving the contentious issue concerning the location of the NSCs9, 10, 11.
Although the in vitro data strongly indicate that PNAlo HSAlo is the predominant NSC type, the possibility remains that another type of NSC may account for a considerable proportion of in vivo activity. As the primary neuronal output of NSCs in the ventricular region in the adult animal is olfactory neurons10, we examined the size of the PNAlo HSAlo stem cell population in a mutant strain of mice, querkopf, which generates greatly reduced numbers of olfactory neurons in the adult animal and has been shown to have a stem cell deficiency during cortical development8. The FACS profile of querkopf mice was identical to wild-type littermates and indistinguishable from that in Fig. 1b, except for a sixfold reduction (6.17 1.36) in the percentage of cells in the PNAlo HSAlo compartment. No significant change in the size of any other population was observed, and the stem cell activity was still restricted to the PNAlo HSAlo compartment in the mutant. Thus, the only detectable change in the cellular content of the periventricular region associated with this loss of neuronal production is a specific depletion of the PNAlo HSAlo NSCs, strongly supporting a central functional role for these cells in situ.
To further examine their in vivo potential, freshly isolated PNAlo HSAlo NSCs were injected into the lateral ventricles of E12.5 mice. Donor PNAlo HSAlo NSCs were isolated from BU5X transgenic mice, which express both lacZ18 and enhanced GFP19 but show identical stem cell characteristics to those of CBA mice (data not shown). When examined 4 weeks after birth, GFP+ neurons and glia were detected in the cortex (Fig. 2g, h), demonstrating that the purified populations of stem cells retain the capacity to differentiate in vivo and do not require accessory cell stimulation.
Recently, the possibility has been raised that bone marrow cells have the potential to give rise to neurons in vivo20, 21. FACS analysis of PNAlo HSAlo cells, however, revealed an absence of haematopoietic cell surface antigens CD34, CD90.2, CD117 and CD135 and the endothelial marker CD31. Furthermore, most PNAlo HSAlo cells were brightly labelled with the DNA-binding dye Hoechst 33342 (data not shown), whereas activity of haematopoietic stem cells is unambiguously found in the Hoechst 'low'-staining compartment22. Taken together, these data strongly suggest that the PNAlo HSAlo NSC is distinct from the bone-marrow-derived stem cell and that the latter is unlikely to be important in neuronal production in the adult CNS.
To examine whether NSCs had the innate potential to give rise to non-neural cell types, we co-cultured freshly isolated PNAlo HSAlo GFP+ NSCs with C2C12 cells under conditions previously described to give rise to myogenic cells23. Many GFP-expressing cells assumed a spindle-like, myocyte-like shape after 4 d in vitro (Fig. 2c), often containing multiple nuclei, which is characteristic of myotubes (arrow, Fig. 2d). Importantly, GFP was co-expressed in cells expressing the myogenic markers -actinin-2 (Fig. 2e) and fast MyHC (Fig. 2f): in the case of myotubes, often only one bright GFP+ nucleus among several GFP- nuclei; in the case of myocytes, a single GFP+ nucleus (arrow, Fig. 2e; Fig. 2f). Other populations of sorted cells did not exhibit myogenic potential (data not shown). As clonal analysis and direct observation revealed no evidence of significant cell division occurring after plating (data not shown), the frequency of NSCs with myogenic potential could be determined directly. We found that 57 5.67% of the plated PNAlo HSAlo GFP+ NSCs differentiated into spindle-shaped MyHC+ myocytes, or myotubes. This demonstrates that the potential for muscle differentiation is intrinsic to most PNAlo HSAlo NSCs, and that the same cells have the potential to generate neural and muscle cells. This is an unequivocal demonstration of pluripotential activity in a freshly isolated stem cell.
Overall, the data strongly indicate that the predominant type of NSC in the subventricular zone of the forebrain has the phenotype PNAlo HSAlo nestin+, and the intrinsic potential to differentiate into both neural and non-neural cells. The identification and purification of NSCs provide a means to study directly the regulation of NSC differentiation both in vitro and in vivo, with the ultimate aim of stimulating endogenous neuronal production to replace damaged neuronal tissue. These methods permit the NSCs to be compared directly to stem cells from other tissues, including embryonic stem cells, to establish whether they have similar biological and molecular properties.
Methods
Tissue preparation and FACS analysis
Adult CBA mice (8–10 weeks old) were killed by cervical dislocation, and periventricular tissue was dissected (essentially as described24) into HEPES-buffered Eagle's medium (HEM). CNS tissue was diced then transferred to Mg2+/Ca2+-free HBSS (containing 10 mM HEPES at pH 7.6, 200 g ml-1 EDTA, 0.5 mM trypsin, 0.001% DNase), for 20 min at 37 °C. Six millilitres of HEM with 5% fetal calf serum (FCS) was added and tissue collected by centrifugation (7 min at 100g). The supernatant was removed and the pellet triturated in 200 l PBS (pH 7.4), producing a single-cell suspension that was subsequently passed through a 70-m cell strainer (Falcon) to remove debris. For immunostaining, this suspension was incubated for 20 min at 4 °C with FITC-conjugated PNA (1:200; Vecta), or PNA-FITC combined with PE-conjugated mCD24a (1:200; clone M1/69, PharMingen), then rinsed twice with PBS by centrifugation, and finally with PBS, 1% FCS, propidium iodide (100 g ml-1; Molecular Probes) to label dead cells before sorting on a FACS II (Becton-Dickinson) flow cytometer. Cell viability was typically >95% and all FACS gates were set using unlabelled cells. Antibodies against CD31, 34, 90.2, 117 and 135 (1:200; PharMingen) were directly conjugated to either FITC or PE where appropriate.
Cell culture
Neurospheres were generated as previously described13. Briefly, sorted cells were plated with bFGF (bovine recombinant, 10 ng ml-1; Boehringer) and EGF (receptor grade, 20 ng ml-1; Collaborative Research) in NS-A basal serum free media (basal media; Euroclone) containing 2 mM l-glutamine, 0.6% glucose, 60 M putrescine, 20 nM progesterone, 30 nM sodium selenite, 25 g ml-1 insulin and 100 g ml-1 apo-transferrin (Sigma). Initially, cells were seeded at 10 viable cells cm-2, final enrichment steps at one cell per well. The ratio of number of spheres formed after 7 d in vitro to number of cells plated is the NSC frequency. Subsequent passaging of primary spheres was performed by mechanically dissociating collected spheres (centrifuged for 10 min at 100g) into a single-cell suspension and replating in basal media containing EGF and FGF-2 (complete medium) as described1, 12. Spheres were differentiated by transfer to glass coverslips coated with poly-l-ornithine (one sphere per coverslip) in complete medium for 1 d, basal medium for 1 d, then basal medium with 1% FCS for 4–5 d, then assessed by immunocytochemistry for neurons and glia.
Freshly isolated NSCs from BU5X mice were co-cultured with C2C12 myogenic cells (5 
103 cells cm-2) for 2 d in DME medium (Gibco) containing 20% heat-inactivated FCS, then in DME supplemented with 1% normal horse serum (CSL) for an additional 2–4 d in vitro. Cultures were fixed for 5 min with 4% paraformaldehyde then processed for immunocytochemistry.
Immunocytochemistry
Double-antigen immunocytochemistry on neurospheres was performed as described previously1, 12, using monoclonal antibody to -tubulin type III (Sigma), GFAP antisera (Dako) and monoclonal antibody to O4 (immunoglobulin-, IgM, Boehringer). Neural-derived muscle cells were identified by simultaneous detection of endogenous GFP and muscle cell types identified by mouse monoclonal antibodies to fast MyHC (NCL-MHCf, 1:10, Novocastra) or -actinin-2 (1:750; ref. 25). These antigens were detected by appropriate TRITC-conjugated IgG secondary antibody (1:200; Southern Biotech). The proportion of neural cells that differentiated into muscle cells was determined by counting the number of GFP+ nuclei expressing myogenic markers divided by the number of NSCs plated (counted 6 h after plating). To detect the progeny of freshly isolated BU5X NSCs injected into (C57BL/6 DBA/2) F1 hybrid pups (E12.5), mice were perfused with 4% paraformaldehyde, their brain sectioned (10-m slices) and immunocytochemistry performed to double-label, donor-derived cells immunoreactive to -gal (anti--gal, Chemicon) with known markers for astrocytes (anti-GFAP, Chemicon) or neurons (anti-NeuN, Chemicon). Images were captured on a Nikon Diaphot microscope using a KX-85 (Apogee) camera.
-actinin (