Isolation and characterization of neural precursor cells from the Sox1–GFP reporter mouse

Animals

Neonatal Sprague-Dawley rats were used as graft recipients in the transplantation experiments. Timed pregnant [day of vaginal plug = embryonic day 0.5 (E0.5)] wild-type (WT) or Sox1–GFP transgenic mice were generated by crossing WT NMRI females with WT or heterozygous (+/–) Sox1–GFP (Aubert et al., 2003) males. Following lethal exposure to CO₂ the embryos were removed and the WT and Sox1–GFP^+/– embryos were identified under a fluorescence microscope before being taken for dissection or immersion-fixed overnight in 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS; 0.1 m) at 8 °C. Fixed embryos were cryoprotected in 15% sucrose and cryosectioned at a thickness of 12 µm. All animal-related procedures were conducted in accordance with local ethical guidelines and approved animal care protocols.

Tissue dissection and preparation of single-cell suspensions

Whole telencephalon or subdissections of various brain regions [brainstem, cortex, lateral ganglionic eminence (LGE) and medial ganglionic eminence (MGE)] were dissected from E12.5–13.5 Sox1–GFP or WT mice and mechanically dissociated into single-cell suspensions in PBS without Ca²⁺ or Mg²⁺ (PBS^–Ca2+/Mg2+ Gibco). Tissue from 20 Sox1–GFP^+/– embryos was pooled for the transplantation experiments. In order to obtain GFP+ve cells from adults, three adult Sox1–GFP mice received lethal injections of pentobarbitone before the brains were removed and placed on ice for 5 min. A brain block (Kopf, Germany) was used to collect 1-mm coronal sections which were placed in L-15 dissection medium (Gibco) and the lateral aspects of the ependymal zone and SVZ were precisely removed under a Leica dissection microscope. The tissue pieces were incubated at 37 °C for 15 min in 10 mL of dissociation solution (HBSS^–Ca2+/Mg2+ Gibco) supplemented with HEPES, 0.015 m (Gibco); d-glucose, 5.4 mg/mL, 1 × trypsin, DNase, 80 U/mL; hyaluronidase, 0.7 mg/mL; and kynerenic acid, 2 mg/mL (all from Sigma), followed by gentle trituration, and incubation in the same solution at 37 °C for another 10 min. Excess solution was removed and the tissue further mechanically dissociated to a single-cell suspension.

Bulk neurosphere cultures

Single cells were plated at a density of 5 × 10⁵ cells/mL in a serum-free expansion medium composed of a 1 : 1 mixture of Dulbecco's modified Eagle's medium (DMEM) and F12 (both Gibco), supplemented with epidermal growth factor (20 ng/mL; human recombinant; R and D Systems, Oxon, UK) and basic fibroblastic growth factor (10 ng/mL; human recombinant; R and D Systems), a defined hormone and salt mixture (Reynolds and Weiss, 1992) including insulin, 20 µg/mL; transferrin, 100 µg/mL; progesterone, 20 nm; putresceine, 60 µm; sodium selenite, 30 nm (all from Sigma); and penicillin–streptomycin, 1% (Gibco). Cultures were maintained at 37 °C in a humid atmosphere with 95% O₂ and 5% CO₂. Fresh medium was added to the cultures every 2nd or 3rd day. After 7 days, neurospheres were either processed for FACS analysis or passaged by mechanical dissociation to a single-cell suspension and then re-suspended at a density of 10⁵ cells/mL. For the transplantation experiments, cultures passaged 5 or 6 times were prepared for FACS on day 7 after the last passage.

Cell preparation for FACS

Cells were mechanically dissociated to obtain a single-cell suspension in a solution containing 2 mm EDTA (Sigma) and 0.5% bovine serum albumin (Sigma) in PBS^–Ca2+/Mg2+. Cells were harvested by centrifugation (500 g, 5 min, 4 °C), filtered through a 70-µm cup filter (Falcon) and kept on ice until FACS analysis. Cell sorting was performed on a FACS Vantage TS flow cytometer (Becton Dickinson, San Jose, CA, USA) equipped with 488-nm argon and 633-nm He-Ne lasers. In all sorting procedures, an initial gating was performed in order to exclude cell debris and cell doublets based on forward- and side-scatter information, dead cells were excluded based on labelling with 7-aminoactionomycin-D (7-AAD; Sigma; 10 µL/mL) and the system was first calibrated with WT tissue in order to determine the background signal for GFP fluorescence. For in vitro experiments, three subpopulations were sorted according to GFP intensity: GFP^Negative, GFP^High and GFP^Low populations. The GFP^Low and GFP^High populations were defined arbitrarily by gating approximately half-way between the lower and upper limits for the GFP signal. When sorting from neurospheres or acutely dissociated foetal or adult brain for the transplantation experiments, all GFP+ve cells were taken.

Clonal analysis and in vitro differentiation

Cells were plated at clonal density (10 cells/µL) with an automated cell deposition unit (Becton Dickinson) into a 96-well plate containing a mixture (1 : 4) of conditioned medium and expansion medium (see experimental design in Fig. 4D). After 7 days in vitro, the number of first-generation spheres was counted. Individual first-generation spheres were then transferred into new wells and the single isolated spheres were dissociated into a single-cell suspension to study the formation of second-generation spheres. This procedure was repeated to study third-generation sphere formation. For in vitro differentiation, 10-day-old third-generation spheres were plated onto poly l-lysine-coated chamber slides in the same medium, with the exception that growth factors were replaced with 1% foetal bovine serum. After 7 days in vitro cells were fixed for 15 min with 4% PFA at room temperature (RT), rinsed three times in PBS and processed for immunocytochemistry.

Transplantation procedure

Sorted GFP+ve cells were resuspended in HBSS^–Ca2+/Mg2+ at 1 × 10⁵ cells/µL (foetal tissue), 1 × 10⁴ cells/µL (neurospheres) or 0.5 × 10⁴ cells/µL (adult tissue), with the low yield of neurosphere and adult sorted cells dictating the lower density of these cell suspensions. Under deep hypothermic anaesthesia, 1 µL of cell suspension was delivered over 2 min to the striatum of postnatal day 3 rats using a glass capillary attached to the tip of a 5 µL Hamilton syringe as previously described (Nikkhah et al., 2000). Injection co-ordinates (mm) were anterior, 0.7 and lateral, 1.9 to bregma, and 2.9 below the dural surface. A minimum of four animals were grafted with each cell preparation. Six weeks after transplantation animals received lethal doses of pentobarbitone and were transcardially perfused with 0.9% saline followed by 4% PFA. The brains were postfixed for 2 h, cryoprotected in 25% sucrose and sectioned on a freezing microtome (30-µm coronal sections).

Immunohistochemistry and immunocytochemistry

Slide-mounted or free floating sections, or plated cells, were preincubated in potassium-buffered PBS (KPBS) with 5% normal serum and 0.025% Triton-X (Amresco, Solon, Ohio, USA) for 1 h at RT, followed by incubation overnight at RT with primary antibodies diluted in the incubation solution. For peroxidase-based detection of the M2M6 antibody, the tissue was quenched of endogenous peroxidase activity for 15 min with 3% H₂O₂ in KPBS prior to incubation with primary antibody. The primary antibodies used were: mouse anti-adenomatous polyposis coli (APC; Calbiochem, 1 : 200), mouse anti-polysialylated neural cell adhesion molecule (PSA-NCAM) (Sigma, 1 : 500), mouse anti-Nestin (BD Biosciences, 1 : 500), rat anti-Musashi1 (kindly provided by Dr H. Okano, 1 : 1000), mouse anti-neuronal nuclei (NeuN; Chemicon, 1 : 100), chicken anti-GFP (Chemicon, 1 : 5000), rat anti-M2 and -M6 (courtesy of Dr C. Lagenaur, available from Developmental Studies Hybridoma Bank; both 1 : 50), rabbit anti-dopamine- and cAMP-regulated phosphoprotein of 32 kDa (DARPP-32; Chemicon, 1 : 1000), rabbit anti-choline acetyltransferase (ChAT; Chemicon, 1 : 500), mouse antiparvalbumin (Sigma, 1 : 2000), rabbit anti-SOX1 and rabbit anti-SOX2 (both kindly provided by Dr R. Lovell-Badge, 1 : 1000), mouse anti-2′3′-cyclic nucleotide-3′-phosphodiesterase (CNPase) (1 : 100, Sigma), rabbit anti-glial fibrillary acidic protein (GFAP; 1 : 1000, DAKO), and rabbit antiβ-III-tubulin (1 : 1000, Covance, Berkeley, California, USA). After three rinses in kPBS, sections or plated cells were incubated with fluorescein isothionocyanate–Cy2- and/or Cy3-conjugated secondary antibodies (Jackson Immunoresearch, 1 : 200) for 2 h at RT for fluorescent labelling of the primary antibodies. For nuclei staining, plated cells were incubated for 5 min in 4,6-diamidino-2-phenylindole (DAPI; Sigma, 1 : 1000). For peroxidase-based labelling of the M2M6 antibody, a peroxidase-conjugated donkey antirat secondary antibody (Jackson Immunoresearch, 1 : 200) was used (2 h, RT), before the material was incubated with 3,3-diaminobenzidine (0.5 mg/mL, Sigma) and precipitation of the chromophore was catalysed by the addition of 1% H₂O₂. Material labelled with fluorescent markers was coverslipped using PVA-DABCO (Peterson, 1999), while the 3,3-diaminobenzidine-labelled material was dehydrated in alcohol and xylene before being coverslipped with DEPEX™ mounting media (BDH).

Quantification

Single-cell suspensions generated from E12.5 cortex, LGE and MGE were re-suspended at a density of 10⁵ cells/mL in DMEM solution, and plated on poly l-lysine-coated chamber slides. Less than two hours after plating, cells were fixed for 15 min in ice-cold 4% PFA and immediately processed for immunostaining as described above. Total cell numbers in random visual fields were quantified by counting all DAPI-stained nuclei, and the proportion of cells expressing Sox1, Sox2, Nestin, Musashi1 or PSA-NCAM was also determined. Quantifications were performed in triplicate and a minimum of 1000 cells were counted. All results are reported as mean ± 1 SEM.

Imaging and schematic drawings

The macroscopic fluorescence image of an embryonic Sox1–GFP brain illustrated in Fig. 1 was acquired using a dissection microscope (Leica) equipped with a digital camera (ProgRes; Jenoptik, Germany). Confocal images (5-9) were acquired using a Leica DMRE laser-scanning microscope equipped with GreNe, HeNe and Argon lasers.

In order to provide low-magnification overviews of graft-derived fibre innervation throughout the host brain, fibre patterns were accurately traced (Canvas v9.0.4) over digital photomontages from coronal sections immunohistochemicaly labelled with M2M6.

GFP was expressed in germinal zones of the developing brain

In this study, we have made use of a transgenic mouse that has GFP knocked into the Sox1 locus (Aubert et al., 2003). In these mice, the GFP reporter is widely expressed throughout the developing nervous system (Fig. 1A–C) and also in the developing lens, in a pattern closely mimicking that of the endogenous Sox1 protein (Aubert et al., 2003). Notably, however, GFP expression is completely absent from the developing ventral mesencephalon. This is consistent with immunohistochemistry for the Sox1 protein, which is also not expressed in the dopamine precursor domain of the ventral mesencephalon (not shown). However, we have observed that the related Sox family member, Sox2, is expressed in this region (E. Andersson, M. Parmar and L. Thompson, unpublished observations).

Double immunohistochemistry using antibodies raised against the GFP and Sox1 proteins demonstrates that virtually all the Sox1-positive cells in the developing forebrain are also GFP-positive (Fig. 1D–I). At both E12.5 (Fig. 1D–F) and E16.5 (Fig. 1G–I), Sox1 and GFP are highly expressed in the VZ of the ventral telencephalon. Low levels of Sox1 and GFP can also be detected in the developing cortex (see high magnification images from Fig. 1F). As GFP is a relatively stable protein, its expression can also be detected in cells that have down-regulated Sox1. Thus, GFP was expressed at relatively high levels in Sox1+ve cells within the VZ but also at lower levels in some Sox1–ve cells in the SVZ. This is reflected in cell counting from acutely dissociated E12.5 forebrain, which shows that the percentage of GFP+ve cells is slightly greater than that of Sox1+ve cells (Fig. 1J). High levels of GFP expression were also detected in the region of the developing hippocampus (Fig. 1D). In the adult brain we observed GFP expression in a number of areas including the subgranular layer of the dentate gyrus, consistent with a previous report (Aubert et al., 2003), as well as in the walls of the lateral ventricles, the other major neurogenic area of the adult brain (Fig. 1K–M). As for the foetal brain, the GFP expression domain surrounding the lateral ventricles appeared somewhat broader than that of Sox1 protein expression.

To determine the extent of overlap between Sox1 and GFP in different areas of the developing forebrain, we subdissected the proliferative regions (which included both the VZ and SVZ) from cortex, LGE and MGE of E12.5 Sox1–GFP embryos. The cells were dissociated, plated and, 2 h later, stained with anti-GFP and anti-Sox1 antibodies. From all three areas, the majority of the cells expressed both GFP and Sox1 (Fig. 1J): 91.4 ± 2.1% of the cortical cells were GFP+ve and 84.7 ± 2.3% of these also expressed Sox1; 83.1 ± 1.8% of the LGE cells expressed GFP and out of those 76.1 ± 2.4% also expressed Sox1; 80.3 ± 3.3% of the MGE cells were GFP+ve, and 70.0 ± 2.4% of the MGE cells were both GFP- and Sox1+ve.

GFP expression overlapped with other known markers for neural stem and progenitor cells

To further characterize the telencephalic cells expressing GFP, we performed immunohistochemistry for GFP in combination with other known markers of neural stem and progenitor cells: the family member Sox2 (Zappone et al., 2000), the intermediate filament protein Nestin (Lendahl et al., 1990) and the RNA binding protein Musashi1 (Sakakibara et al., 1996; Sakakibara and Okano, 1997). We found that Sox2 is expressed in the forebrain VZ in a dorsal to ventral gradient opposite to that of Sox1 and Sox1–GFP, i.e. with the highest expression dorsally, in the developing cortex (Fig. 2A). Cell counting in acutely dissociated tissue showed that the majority of GFP-expressing cells also expressed Sox2, but that there was a small (< 10%) population of cells that expressed Sox2 but were GFP–ve (Fig. 2B). There was also considerable overlap between GFP and Nestin (Fig. 2C) and between GFP and Musashi1 (Fig. 2E). However, in both cases, a small population of GFP-expressing cells that did not express Nestin or Musashi1 was also present (Fig. 2D and F). Further, we analysed to what extent GFP expression overlapped with that of PSA-NCAM. At this gestational stage, PSA-NCAM was only present in the mantle region of the LGE and the MGE where GFP is no longer expressed (Fig. 2G). Quantification confirmed that virtually all GFP-expressing cells were PSA-NCAM–ve (Fig. 2H).

The Sox1–GFP population was expandable in vitro and contained all self-renewing and multipotent cells

Neurosphere culture is a commonly used method for expanding neural stem and progenitor cells in vitro (Reynolds and Weiss, 1992). Cells isolated from E12.5 forebrain of Sox1–GFP mice and plated in serum-free medium containing epidermal growth factor and basic fibroblastic growth factor formed neurospheres at the same frequency as from forebrain dissections of age-matched WT mice (M. Parmar, unpublished data). In the neurosphere cultures established from the forebrain of the Sox1–GFP mice, virtually all spheres contained GFP-expressing cells and flow cytometric analysis showed that primary neurosphere cultures contained 34.8 ± 2.2% (n = 3) GFP+ve cells (Fig. 3A). After passaging, all spheres examined still contained GFP-expressing cells but the proportion of GFP-expressing cells decreased to ≈ 5%, where it remained constant throughout the culture period: 5.2 ± 0.2% at passage 2 (Fig. 3B; n = 3) and 5.1 ± 0.8% at passage 5 (Fig. 3C; n = 3).

From secondary spheres we defined three subpopulations of cells based on their relative GFP intensity (see Fig. 4A and B). The resulting GFP^High,GFP^Low and GFP^Negative populations contained ≈ 1.5, 3.5, and 80% of the total cell population, respectively (Fig. 4B). Seven days after sorting and plating at clonal density (Hulspas et al., 1997), the number of spheres formed from each subpopulation was quantified (Fig. 4C). The GFP^High population represented only 1.5% of the total cells, yet the majority of the neurosphere-forming cells (1.7 ± 0.1% of the GFP^High cells) were found in the GFP^High subpopulation. Some neurosphere-forming cells (0.5 ± 0.1%) were also present in the GFP^Low subpopulation that made up 3.5% of the total cells, whereas no spheres were generated from the GFP^Negative subpopulation. In a control population (all live cells gated, i.e a mixture of all three subpopulations) < 0.04 ± 0.02% of the cells formed neurospheres. Thus, the number of spheres formed from the GFP^High population represented a > 40-fold enrichment compared to the control cells. Because preparation and handling of foetal neural stem and progenitor cells for FACS reduces the sphere-forming ability of treated cells by at least 6-fold (M. Parmar, unpublished observation; also see Kawaguchi et al., 2001), it can be estimated that ≈ 10% of the GFP^High expressers are neurosphere-forming stem cells.

To confirm the stem cell characteristics of the spheres generated from the GFP-expressing cells, their self-renewal capacity and multilineage potential were tested in a clonal in vitro stem cell assay (outlined in Fig. 4D). Thirty-three individual spheres formed after FACS sorting (21 from the GFP^High and 12 from the GFP^Low population) were picked under a light microscope. The self-renewal properties of the GFP^High and the GFP^Low sorted cells were analysed by subcloning of the individual spheres. We found that all (21/21) spheres formed from the GFP^High population, but only 66% (8/12) of the spheres formed from the GFP^Low population, formed new second-generation spheres. When these second-generation spheres again were individually picked and subjected to the same sphere-forming assay, more spheres from the GFP^High than from the GFP^Low population formed new tertiary spheres (70% vs. 47%, n = 108). The individual third-generation spheres (originating from the GFP^High and GFP^Low populations) were picked and plated under differentiation conditions. After 7 days, all spheres had generated cells with morphological characteristics of astrocytes and many clones also contained neurons and oligodendrocytes, as detected by expression of β-III-tubulin and CNPase, respectively (Fig. 4E). Of the 24 third-generation spheres generated from the GFP^High population, all but two generated both neurons and glia: 13 clones (54.2%) generated all three cell types (neurons, astrocytes and oligodendrocytes), nine clones (37.5%) generated both astrocytes and neurons and two clones (8.3%) contained only astrocytes (Fig. 4F). In contrast, only one of the 11 third-generation spheres from the GFP^Low population generated both neurons and glia, and the other 10 generated only astrocytes after differentiation in vitro (Fig. 4F).

Sox1–GFP cells isolated from the embryonic and adult brain gave rise to neurons and glia in vivo

We further analysed the Sox1–GFP+ve cells for their capacity to generate different cell types in vivo, following intracerebral transplantation. Thus we FACS-isolated the GFP-expressing cells from the E13.5 LGE, MGE, cortex or brainstem of Sox1–GFP mice and transplanted them into the striatum of neonatal rats. The FACS analysis showed that the proportion of GFP+ve cells in these regions was ≈ 93% (brainstem), 80% (cortex), 88% (MGE) and 75% (LGE). Notably, the vast majority of the GFP–ve cells did not survive the dissociation and FACS procedure as evidenced by 7AAD incorporation (LGE, Fig. 3D; other regions not shown). Six weeks after transplantation, native GFP fluorescence was no longer detectable in any of the grafted animals. Grafted cells were identified in the rat hosts through immunohistochemistry for the mouse-specific marker M2M6. Cells with both neuronal (Fig. 5A) and glial (Fig. 5B) morphology could be identified and a number of the glial cells appeared closely associated with striatal fibre bundles (Fig. 5C). These cells expressed phenotypic markers characteristic of oligodendrocytes, including APC (Fig. 5D–F) and CNPase (not shown). Labeling for other phenotypic markers confirmed the presence of graft-derived astrocytes (GFAP; Fig. 5G–I) and neurons (NeuN; Fig. 6J–L). The presence of multiple cell types from different neural lineages was a consistent feature of all grafts regardless of the embryonic brain region from which the cells were isolated.

Compared to the high proportion of GFP+ve cells in foetal tissue, the SVZ dissected from adult brains of Sox1–GFP^+/– mice contained relatively few GFP+ve cells (< 15%; Fig. 3E). The majority of low-expressing GFP+ve cells as well as GFP–ve cells did not survive the FACS procedure, as indicated by 7AAD incorporation (Fig. 3E). The cells expressing relatively high levels of GFP survived the dissociation and FACS procedure (representing ≈ 2.5% of the total cell population) and were taken for transplantation. Six weeks following transplantation, the grafted M2M6+ve cells were distributed in the striatal parenchyma, often found to be intimately associated with host blood vessels (Fig. 5M), and were almost exclusively of a glial phenotype with astrocytic morphology. A number of these cells were also seen to express GFAP (Fig. 5N). Notably, the grafts were almost completely devoid of neurons, although single examples of cells with neuronal morphology were occasionally seen in all animals analysed (Fig. 5O). Furthermore, no white matter-associated oligodendroglial cells were noted. As stem cells in the adult SVZ are known to generate neurons destined for the olfactory bulb, we also examined the host olfactory bulbs for the presence of M2M6+ve cells; however, none were observed.

Sox1–GFP cells from the developing brain were regionally specified

Although acutely dissociated Sox1–GFP cells from all four embryonic regions analysed were capable of generating neurons, astrocytes and oligodendrocytes after transplantation, certain other features of the grafts were dependent on where the cells were originally derived from. For example, grafts of Sox1–GFP cells derived from LGE or cortex were much larger than those derived from MGE or brainstem, with the cortical grafts being the largest and MGE grafts the smallest. Furthermore, the fibre outgrowth and neuronal phenotypes found in the grafts also differed depending on the cell source.

Cortex-derived grafts

Cortical Sox1–GFP+ve cells gave rise to the largest grafts, with many intensely M2M6+ve glial cells identifiable in and around the graft core (Fig. 7A). Although it was difficult to identify M2M6+ve neuronal cell bodies against the background of M2M6+ve glial cells, double staining with M2M6 and NeuN revealed that the graft cores did in fact contain many neurons (data not shown). The pattern of M2M6+ve fibre outgrowth typically associated with these grafts is illustrated in schematic coronal sections (Fig. 7E, i-vi). Graft-derived fibres could be found throughout the striatum, including the most anterior aspects rostral to the graft core (Fig. 7B) and in a particularly dense pattern in the dorsomedial striatum (Fig. 7C). Fibres were also seen traversing through the corpus callosum and in the dorsomedial striatum of the contralateral hemisphere (Fig. 7D). Caudal to the graft core, there was a robust innervation of the host thalamus, particularly ventral thalamic nuclei (Fig. 7F), and scattered M2M6+ve fibres could also be seen as far back as the subthalamic nucleus (Fig. 7G).

Neurosphere expanded Sox1–GFP cells had a more limited differentiation potential in vivo

Compared to the Sox1–GFP cells from acutely dissected foetal tissue, Sox1–GFP cells isolated from expanded neurosphere cultures were shown to have a more limited differentiation potential in vivo. Specifically, these cells lose the ability to generate neurons after transplantation. GFP-positive cells isolated from LGE derived neurosphere cultures survived the dissociation and FACS comparatively well (Fig. 3F), but the sorted cells only gave rise to glial cells following transplantation (Fig. 9A). The glial population included GFAP-expressing astrocytes (Fig. 9B) and APC-expressing oligodendrocytes (Fig. 9C). The oligodendrocytic phenotype was associated with the positioning of the transplanted cells in host white matter tracts such as the internal capsule and corpus callosum (Fig. 9D). Similar results were obtained after grafting Sox1–GFP cells isolated from expanded E13.5 cortical tissue (not shown).