Neural stem cells are responsible for the remarkable diversity of cell types in the adult and developing brains: they generate the wide variety of neuronal cells, astrocytes, and oligodendrocytes (Goldman and Luskin, 1998; Gage, 2000; Alvarez-Buylla et al., 2001; Temple et al., 2001). Moreover, recent reports have raised the possibility that neural stem cells can demonstrate even more plasticity than originally envisaged and may contribute to a range of cell lineages (Bjornson et al., 1999; Brazelton et al., 2000; Clarke et al., 2000; Blau et al., 2001). This broad potential of neural stem cells, associated with their vast proliferative capacity, recently has attracted great interest, not least because of its therapeutic potential.
To better understand the biology of neural stem cells, the issues of identification, visualization, and isolation of neural stem cells in the developing and adult brains need to be addressed. Neural stem cells are typically identified after fixation and extensive processing and staining with specific antibodies or by functional assays, e.g., by testing their capacity to form neurospheres. We decided to generate strains of mice that would permit access to neural stem cells in the developing and adult brains in vitro and in vivo. We generated mice in which neural stem cells are marked by the expression of green fluorescent protein (GFP). This expression is controlled by the regulatory elements of the gene that encodes nestin, an intermediate filament protein selectively expressed in neural stem cells of developing and adult brains. Nestin expression is regulated by enhancer elements residing in the intronic regions of the gene. These elements can direct the expression of a reporter gene in the neuroepithelium of the developing embryonic brain (Zimmerman et al., 1994; Josephson et al., 1998; Yaworsky and Kappen, 1999). We used these elements to generate nestin-GFP mice; in this article we present a detailed analysis of such transgenic animals. These mice served as a useful reporter system for exploring neurogenesis and plasticity of stem cells in developing and adult organisms. We demonstrate the usefulness of this approach by reporting novel features of progenitor cells in the adult hippocampus and by identifying distinct classes of neuronal precursors in the adult brain.
MATERIALS AND METHODS
Generation and analysis of transgenic mice
Fragments of the nestin gene (gift from Drs. R. McKay and L. Zimmerman; Zimmerman et al., 1994; Josephson et al., 1998; Yaworsky and Kappen, 1999) were subcloned into the pBSM13+ vector. The 5.8-kb fragment of the promoter region and the 1.8-kb fragment containing the second intron were combined with the cDNA of the enhanced version of GFP (EGFP; Clontech, Palo Alto, CA) and polyadenylation sequences from simian virus 40 and cloned into the pBSM13+ vector, generating nestin-GFP plasmid. In the final construct, EGFP cDNA was placed between the promoter and the intron sequences of the nestin gene (Fig. 1A), thus matching the arrangement of the regulatory sequences in the nestin gene. The plasmid was isolated and purified through centrifugation in cesium chloride and digested with the SmaI restriction enzyme; this removed the entire vector backbone, leaving the nestin-GFP sequences intact. An 8.7-kb fragment was purified by electrophoresis through the agarose gel and used for the pronuclear injections of the fertilized oocytes from C57BL/6 × Balb/cBy hybrid mice. Use of animals in the present experiments was reviewed and approved by the Cold Spring Harbor Laboratory Animal Use and Care Committee.
Eighty-six animals born after injections were screened by polymerase chain reaction using the primers 5′GGAGCTGCACACAACCCATTGCC3′ (corresponding to the sequences from the second intron of nestin gene) and 5′GATCACTCTCGGCATGGACGAGC3′ (corresponding to the sequences from the EGFP cDNA). The expected fragment of 510 bp was detected in eight of the 86 F0 mice (three male and five female). The progeny of three mice was expanded and used for the subsequent analysis. All three lines produced identical patterns of GFP expression; most work was done with line 33.
To study the transgene expression during embryonic development, nestin-GFP transgenic males were mated with C57BL/6 females. Embryonic ages were determined as the time since the appearance of the copulative plug (the first day was determined as e0.5). The fluorescent signal from the transgene was so strong that transgenic embryos were routinely determined using a fluorescent lamp.
For sectioning, embryos were immediately placed into ice-cold 4% paraformaldehyde in phosphate buffered saline, pH 7.4 (PBS; Sigma, St. Louis, MO) and left at 4°C overnight. They were then placed into a series of sucrose-PBS baths from 5% to 30% of sucrose at 5% increments to minimize shrinking. Solutions were changed after the embryo had completely submerged. Sectioning was performed on a Leica Frigocut 2800E with a cutting temperature of −17°C onto gelatin-coated slides. Sections were allowed to dry for 30 minutes before proceeding with immunohistochemistry.
Adult animals were sedated with 400 μl of 15% chloryl hydrate and transcardially perfused with 20 ml of ice-cold PBS, followed by 100 ml of ice cold 4% paraformaldehyde in PBS. Brains were dissected, postfixed with paraformaldehyde for 4 hours, and sectioned with a Vibratome or transferred to 30% sucrose-PBS at 4°C for cryosectioning. Cryosectioning was performed as above at the cutting temperature of -30°C.
Free-floating Vibratome sections or cryosectioned slices on gelatin-coated slides were analyzed according to the following procedure. Sections were washed two times with PBS and incubated for 4 hours at room temperature in PBS containing 2% Triton (Sigma) and 10% normal goat serum (Sigma) to permeate the tissue and block nonspecific binding of the antibody. Sections were then incubated with primary antibodies in PBS containing 0.25% Triton and 0.02% NaN3 at 4°C for 48 hours. The following primary antibodies were used: anti-nestin (R401, Chemicon, Temecula, CA; 1:100 dilution), anti–βIII-tubulin (TuJ1, Promega, Madison, WI; 1:1,000 dilution), anti–glial fibrillary acidic protein (GFAP; Sigma; 1:400 dilution), anti-Rip (Iowa University Hybridoma Bank, Iowa City, IA; 1:10 dilution), anti-CNPase (SMI91, Sternberger, Lutherville, MD; 1:100 dilution), and anti-NeuN (Promega; 1:100 dilution). After the incubation with the primary antibody, slides were washed three times with PBS, 0.25% Triton, and 2% NaCl, pH 7.4 (PBS-T; 20 minutes per wash). Next, biotinylated secondary antibodies (Vector Laboratories, Burlingame, CA; 1:200 dilution) in PBS containing 0.25% Triton and 0.02% NaN3 were added for an overnight incubation at 4°C. Slides were washed with PBS-T and incubated with a conjugate between streptavidin and either Alexa-594 (red fluorescence) or Alexa-633 (far red fluorescence; Molecular Probes, Eugene, OR) at 1:400 dilution in PBS containing 0.25% Triton for 4 hours at room temperature. Slides were washed three times with PBS-T (20 minutes per wash) and then in PBS and postfixed in 4% paraformaldehyde-PBS for 30 min.
For visualizing GFP-positive cells in the dentate gyrus (DG), 200-μm-thick sagittal Vibratome sections were immunolabeled for GFAP as described above. Sections were directly cleared in FocusClear (Pacgen, Vancouver, Canada). The slice was mounted in MountClear (Pacgen) and coverslipped, with nail polish around the perimeter, and viewed in the confocal microscope.
Images were taken with a Zeiss LSM 510 confocal microscope. High-resolution images were taken under a 40× C-Apochromat water-immersion objective lens (numerical aperture = 1.2). The images were stored at a size of 1,024 × 1,024 pixels. The distance between successive images (z-axis distance) was adjusted for the refractive index mismatch of the air and mounting medium, as described previously (Chiang et al., 2001). For colocalization experiments, optical sections were 0.8 μm or thinner. Fluorescence intensity was determined using the LSM 2.8 image browser Profile program. Numbers used were the highest value given when profiling a particular cell. Images were assembled with the Adobe Photoshop and Adobe Illustrator programs.
Bromodeoxyuridine (BrdU; Sigma) was injected intraperitoneally at 300 mg/kg, and mice were killed 2 hours later. For these experiments, GFP was detected by using anti-GFP monoclonal antibodies (Roche, Basel, Switzerland), because subsequent acid treatment destroys the fluorescent GFP signal. The secondary goat anti-mouse antibody, conjugated to Alexa Fluor 488 (green fluorescence; Molecular Probes), was used at 1:400 dilution for an overnight incubation at 4°C. The sections were washed as described above and then postfixed for 30 minutes at room temperature. They were then washed two times with PBS and incubated with 2 M HCl for 30 minutes at 37°C. After the acid treatment, sections were washed three times with PBS and then incubated for 10 minutes in 0.1 M tetraborate, pH 8.5. At this point the treatment continued according to the protocol described above. The primary anti-BrdU antibody (monoclonal rat ascites; Harlan, Indianapolis, IN; 1:500 dilution) was allowed to incubate with the sections overnight at room temperature. The secondary antibody was preadsorbed against mouse serum and conjugated to Alexa Fluor 568 (red fluorescence; Molecular Probes; 1:400 dilution).
Flow cytometry was performed with a Beckman Coulter Epics Elite cytometer. Animals were sedated and perfused with ice-cold Hanks' buffer, pH 7.4 (calcium and magnesium free). The brain was removed and digested in 2.5% trypsin, or the nestin-GFP expressing areas were microdissected away with a dissecting microscope and digested with trypsin. Cell suspension was triturated until no obvious chunks of tissue were apparent and until the nestin-GFP–expressing cells appeared singly under the fluorescent microscope. The suspension was washed several times in PBS to remove residual trypsin, passed through a 40-μm mesh, and resuspended at 7 × 106 cells/ml. The flow rate was maintained at 400 data points/second. Fluorescence cutoff was maintained at 10 times greater than the highest fluorescence of the negative control.
Neurospheres were generated by dissociating the brain as described above and plating cells in Dulbecco's Minimum Essential Medium and F12 with 5 μg/ml insulin (Sigma), 20 nM progesterone (Sigma; water soluble), 100 μM putrescine (Sigma), 30 nM sodium selenite (Sigma), 25 μg/ml transferrin (ICN, Costa Mesa, CA), and 25 μg/ml pituitary extract (Gibco, Invitrogen, Carlsbad, CA). Epidermal growth factor (EGF; Sigma) and fibroblast growth factor (FGF; Sigma) were added every 3 days to the culture medium at 20 ng/ml. For clonal analysis, cells were plated at 50 cells/mm2 in 1.2% methylcellulose (Sigma) to prevent their movement and aggregation. Cells that had grown into spheres after 3 weeks were transferred to laminin- (Gibco) and poly-ornithine- (Sigma) coated slides. Acid-washed slides were coated by incubating with 10 μg/ml poly-ornithine in water overnight at room temperature, followed by 5 μg/ml laminin overnight in PBS at 37°C. To induce differentiation, the media above was used, without EGF and FGF, but with 5% fetal bovine serum (Gibco).
Generation of transgenic mice
Several reports have demonstrated that expression of nestin in the neuroepithelial cells of the developing embryo is dependent on the presence of a transcriptional enhancer that resides in the second intron of the gene (Zimmerman et al., 1994; Josephson et al., 1998; Yaworsky and Kappen, 1999). Although this nestin enhancer has been shown to function in the developing embryo even when combined with a heterologous promoter, we combined the nestin enhancer with the nestin promoter region in our construct to increase the likelihood of faithfully reproducing the pattern of nestin gene expression in the brain. Thus, we generated a plasmid in which GFP cDNA (EGFP) and a polyadenylation signal from simian virus 40 is placed between the nestin promoter (5.8 kb of the flanking region) and the nestin enhancer (1.8 kb of the second intron; Fig. 1A). After the introduction of this construct into mouse embryos, eight independent transgenic lines were obtained with GFP expression patterns similar to each other. Three lines were then chosen for subsequent analysis, and all demonstrated a virtually identical pattern of GFP expression in developing and adult brains. No obvious defects were apparent during the development and adulthood of the transgenic mice. In these animals the transgene was transmitted normally, and the GFP expression pattern in the developing and adult central nervous systems remained invariant over the course of at least five generations.
GFP expression in transgenic animals parallels nestin expression in the brain
Nestin mRNA and protein are abundantly expressed in the developing and adult nervous systems. Nestin is also strongly expressed in the myotomes of the embryo. In addition, nestin expression has been reported in several other tissues and cell types, e.g., pancreatic islets, and the developing testis, tongue, tooth, and heart (Kachinsky et al., 1995; Terling et al., 1995; Hunziker and Stein, 2000; Zulewski et al., 2001). This relatively wide spectrum of nestin expression reflects the presence of various regulatory elements in the nestin gene (e.g., regulatory elements in the first and third introns direct nestin expression to myotomes; Zimmerman et al., 1994; Yaworsky and Kappen, 1999). Because we used the regulatory elements from the second intron of the nestin gene, expression of the transgene was expected to be restricted to a limited range of cell types of neural origin.
To determine whether the presence of GFP fluorescence correlates with the sites of nestin gene expression in the nervous system, we compared the distribution of GFP-positive cells with the distribution of nestin-immunoreactive cells (using monoclonal antibody R401 to identify nestin) in the brains of adult transgenic animals and with numerous reports describing the distribution of nestin-positive cells in the developing and adult brains. In the majority of instances, there was a precise correspondence between the GFP signal and nestin immunoreactivity in the adult brain (Fig. 1C). The overlap between GFP and nestin expression was obvious in the areas of active neurogenesis such as the subventricular zone (SVZ), rostral migratory stream (RMS), and DG (Fig. 1Ca,b). GFP expression and nestin immunoreactivity also overlapped in the posterior wall of the lateral ventricle (Fig. 1Cc) and in the Bergmann glia cells in the cerebellum (Fig. 1Cd). In most cases, not only the presence but also the level of expression of the nestin-GFP transgene expression corresponded well to the strength of the nestin immunoreactivity signal. The only case in which we could detect incomplete overlap between the GFP and nestin signals in the brain was in the glomerular layer of the olfactory bulb (OB), where only some periglomerular cells were weakly immunopositive for the R401 antibody, although many of them expressed GFP (Fig. 1Ce). Expression of nestin, but not of GFP, was also seen in the myotomes of the embryo, consistent with the absence of the first and third introns in our construct; these introns have been shown to direct transgene expression to non-neural tissues (Zimmerman et al., 1994; Josephson et al., 1998; Yaworsky and Kappen, 1999). Thus, the overall pattern of nestin-GFP transgene expression closely corresponded to the distribution of nestin-positive cells in the developing and adult nervous systems, indicating that GFP fluorescence can be considered a reliable marker of nestin expression in the brain.
Nestin-GFP expression during development
In the developing transgenic nestin-GFP embryos, GFP expression was evident as early as day 7 of embryonic development (e7). At e8, the GFP signal was seen predominately in the neural plate, and by e10 intense GFP fluorescence was seen throughout the developing nervous system. Through further days of embryonic development, GFP expression was strong throughout the neuroepithelium. The overall intensity of the GFP signal in the developing nervous system peaked between e14 and e15 and gradually decreased through the period of embryonic and early postnatal development. At e10 to e12, GFP signals marked the entire thickness of the cerebral wall, but, starting from e12, GFP expression became weaker near the pial surface and stronger in the ventricular zone (Fig. 2a–d). During the further course of brain development, the distribution of the GFP signal became even more tightly localized to the lateral ventricles, which were seen as lined by GFP-positive cells (Figs. 2e–i). By e20 GFP signal was seen in a narrow area of the SVZ and resembled the pattern of GFP distribution in the adult brain (see below).
GFP expression was also prominent in the developing cerebellum and was very strong in early postnatal (day 4) cerebellum, reflecting continuous neurogenesis in this region of the postnatal brain (Fig. 2j). Eight days after birth, GFP expression went down dramatically and in the adult brain (>1 month) could be detected only in some Bergmann glia cells.
During embryonic development, neuroepithelial cells divide in the ventricular zone of the cerebral wall and then migrate to more superficial layers and differentiate into neurons. This process is expected to correlate with a decrease in nestin expression (and GFP presence) as cells proceed toward neuronal differentiation. This is clearly seen in Figure 2k and 2l, where coronal sections of e12 and e15 brains of nestin-GFP embryos were stained with an antibody to the neuronal form of βIII-tubulin, which recognizes late neuronal progenitors and differentiated neurons. There were distinct zones of transgene expression near the ventricle and of βIII-tubulin expression at the periphery; there was some overlap at the border of the two zones, and there were very few GFP-positive cells in the βIII-tubulin–positive areas near the pial surface, where the differentiated neurons become located. This indicates that, in the developing brain, nestin-GFP transgene expression is decreased in the progeny of neuroepithelial cells, which enter the path of neuronal differentiation, and is absent in differentiated neurons, further confirming the validity of our strategy of using nestin-GFP to identify undifferentiated neural stem cells.
Nestin-GFP expression in the adult brain
In the adult mammalian brain persistent generation of new neurons is limited to only a few areas, such as the lateral wall of the lateral ventricle, the RMS, the OB, and the DG. Each of these regions demonstrated highly specific and intense expression of the GFP transgene in the adult animal (Fig. 3).
In the lateral ventricle, intensely GFP-positive cells were present in the ependymal and subventricular cell layers (Fig. 3b). In the RMS, bright GFP-positive cells intermingled with less bright cells, most of which were involved in chain migration along the RMS (Fig. 3c; also see below). Although the appearance of GFP-positive cells in the SVZ and the RMS differed, none of these cells had neuronal morphology, thus supporting the notion that GFP is expressed in neuronal progenitors but not in the differentiated neurons. In the DG, GFP-positive cells were localized exclusively in the subgranular cell layer (Fig. 3d), where neuronal progenitors reside (Kuhn et al., 1996; Gage, 2000). They were not detected in the granule cell layer, where differentiating neurons migrate after being born in the subgranular cell layer (see also Figs. 5–8), thus supporting the notion that transgene is expressed in progenitor, but not in differentiated, cells. Thus, in the adult brain GFP is selectively expressed in the areas related to continuous neurogenesis, consistent with being a marker of adult stem and progenitor cells.
Bright and dim GFP-positive cells
GFP expression marked entire areas of neurogenesis in the adult brain of the nestin-GFP transgenics. Two classes of GFP-expressing cells, however, were apparent in the SVZ and the RMS of the adult brain. A fraction of GFP-positive cells showed high levels of fluorescence (GFP-bright cells); most of the remaining cells had a much weaker GFP signal (GFP-dim cells). Importantly, the difference in GFP expression levels seen in the sections did not appear as a continuum of expression levels; rather, it appeared to reflect the presence of two distinct groups. This notion was confirmed when we performed a quantitative analysis of the levels of fluorescence on the sections of brains of the nestin-GFP transgenics (Fig. 4A). A distinct population of GFP-positive cells in the RMS showed a threefold stronger fluorescent signal than did the other GFP-positive cells. To determine whether these cell populations represented different functional groups, we probed nestin-GFP cells with antibodies to GFAP, which mark astrocytes, and with antibodies to βIII-tubulin, which mark neuronal progenitors. Remarkably, GFAP was expressed predominantly in GFP-bright, but not in GFP-dim, cells; conversely, βIII-tubulin was expressed predominantly in GFP-dim, but not in GFP-bright, cells (Fig. 4B). This supports the observation that, among the GFP-expressing cells in the brain of adult transgenic animals, distinct GFP-bright and GFP-dim groups can be distinguished and suggests that these groups may represent distinct populations of neuronal precursor cells in the adult brain, perhaps reflecting progressive stages in maturation of neuronal precursors (see also below).
Coexpression with other markers
We next sought to characterize the status of GFP-positive cells by examining the expression of various markers with antibodies to βIII-tubulin, doublecortin, NeuN, Rip, and GFAP. Expression of βIII-tubulin marks neuronal progenitors, young neurons, and mature neurons; expression of doublecortin is typical of migrating neuroblasts; expression of NeuN is specific to mature neurons; expression of Rip is characteristic of mature oligodendrocytes; and expression of GFAP is characteristic of astrocytes.
In the lateral wall of the lateral ventricles, very high levels of GFP expression mark several layers of cells. These cells include the ependymal cell layer and the SVZ. In the posterior area of the lateral ventricle, only the ependymal cells were marked by strong GFP expression, and βIII-tubulin was expressed in the cells immediately adjacent to the ependymal cell layer; in no instance was colocalization of GFP with βIII-tubulin detected (Fig. 5a). In contrast, in the anterior aspects of the lateral ventricles, several layers of cells were GFP positive, with distinct populations of GFP-bright and GFP-dim cells among them. GFP-bright cells tended to localize closer to the ventricle, whereas most of GFP-dim cells were present in the distal layers farthest from the ventricle (Fig. 5b). The βIII-tubulin–positive cells were clearly prevalent among the GFP-dim population and were seen directly apposed to the ependymal cell layer.
In the RMS the presence of the populations of GFP-bright and GFP-dim cells was obvious in sagittal (Fig. 5c) and coronal (Fig. 5d) sections. Both groups of GFP-positive cells were present along the entire length of the RMS and appeared randomly distributed across the RMS. The strength of the GFP signal in the RMS was related to the pattern of βIII-tubulin immunoreactivity: most of the GFP-bright cells did not coexpress βIII-tubulin, whereas GFP-dim cells often were colocalized with the βIII-tubulin signal (see also higher magnification images in Fig. 4B). Importantly, many of the GFP-dim cells were present in groups of cells morphologically identical to groups of neuroblasts involved in chain migration along the RMS (Lois and Alvarez-Buylla, 1994). Together, these findings support the notion that GFP-dim cells correspond to more advanced neuronal precursors.
In the DG, high levels of GFP expression were seen exclusively in the subgranular cell layer (Fig. 5e; see also Figs. 6–8), and we did not detect colocalization of the GFP and βIII-tubulin signals. Mature NeuN-positive neurons were located in the granule cell layer of the DG and did not overlap with GFP-positive cells of the subgranular cell layer (Fig. 5f).
Doublecortin is a microtubule-associated protein whose presence is characteristic of migrating neuroblasts in the developing and adult brains (Feng and Walsh, 2001). Staining with anti-doublecortin antibody was particularly helpful in showing chains of migrating GFP-positive neuronal precursors in the RMS (Fig. 6a, b). It also demonstrated that most of the doublecortin-positive cells in the RMS were GFP dim, and most of the GFP-dim cells were doublecortin positive (Fig. 6a,b). This finding provided further support to the idea that GFP-bright and GFP-dim cells represent distinct stages of differentiation of neuronal precursors.
Interestingly, a fraction of doublecortin-positive cells was detected migrating away from the RMS, often in close contact with the processes of GFP-positive cells (Fig. 6b). In the DG, doublecortin-positive cells were mainly present in the subgranular cells layer often adjacent to the GFP-bright cells (Fig. 6c). As in the RMS, expression of doublecortin was seen in the GFP-dim but not in the GFP-bright cells.
When GFP-positive cells were probed with Rip, a marker for mature oligodendrocytes, there was no detectable colocalization of the Rip and GFP signals in all regions analyzed (SVZ, RMS, DG, OB, cerebellum, and cortex). Interestingly, there were sharp borders between the two markers in the corpus callosum, where small gaps in the areas of strong Rip expression were often filled with cell bodies and processes of GFP-positive cells projecting radially from the RMS toward the cortex (not shown).
Whereas βIII-tubulin and doublecortin immunoreactivities often were found to colocalize with nestin-GFP-dim cells in the neurogenic areas, the presence of the astrocytic marker GFAP was typical of the brightly fluorescent GFP. This was seen particularly well in the most anterior areas of the SVZ and in the RMS (Fig. 7a,b; see also Fig. 4B). In the RMS, intensely fluorescent (GFP-bright) cells were typically GFAP positive; they were scattered along the entire path of the RMS and intermingled with the GFP-dim cells (which were determined to be βIII-tubulin and doublecortin positive; Figs. 4, 5). In the mid-posterior dorsal surface of the lateral ventricle, occasional GFP-and GFAP-positive cells extended processes through the corpus callosum (Fig. 7c). GFAP was also weakly expressed in the nestin-GFP–positive ependymal cells lining the ventricle (Fig. 7d). In the DG, all of the nestin-GFP–expressing cells were also positive for GFAP immunoreactivity (Fig. 7e).
Expression of nestin-GFP reveals novel features of neural stem cells in the DG
We studied GFP-expressing cells in the DG in more detail. We found that GFP is expressed exclusively in a class of cells with unusual features (Fig. 8A): their cell bodies were localized exclusively in the subgranular layer of the DG, and they extended radial projections that terminated in the molecular layer just beyond the granule cell layer. These projections appeared as unipolar, branching to two to three processes within the granule cell layer and terminating in the molecular layer with massive arbor-like clusters of very short fine leaflike processes.
All of the GFP-expressing cells in the DG also were positive for GFAP. Most of the GFAP signal was present in the processes crossing the granule cells layer (Fig. 8B). At the same time, there was no detectable colocalization of the GFP signal with the strongly GFAP-positive, highly branched stellate-shape astrocytes that flanked either side of the RMS or were scattered throughout the cortex and mid- and hindbrain.
BrdU incorporation in GFP-positive cells
To examine the capacity of GFP-positive cells in the adult brain to proliferate, we labeled dividing cells in vivo with BrdU for 2 hours and performed double immunolabeling with anti-BrdU and anti-GFP antibodies. In the SVZ and the RMS, a fraction of GFP-positive cells was labeled with BrdU, indicating that they undergo DNA synthesis. Conversely, most of the BrdU-positive cells were also GFP positive. Interestingly, the distribution of BrdU-positive cells between GFP-bright and GFP-dim cells differed in different neurogenic areas; e.g., in the proximal part of the RMS almost one third of the BrdU-positive nuclei was found among the GFP-bright cells, whereas in more distal areas (close to the olfactory bulb), most of the BrdU-positive cells had low levels of GFP expression (GFP-dim cells; Fig. 9). Together these results support the notion that different levels of expression of the nestin-GFP transgene mark different classes of neuronal precursors in the adult brain.
GFP-positive cells are multipotential
Neural stem cells generate all major types of differentiated cells in the brain. When cultured in vitro, they are capable of forming clonally derived groups of cells (neurospheres; Reynolds and Weiss, 1992; Morshead et al., 1994). When plated onto the appropriate substrate, cells of the neurosphere can generate differentiated neurons, astrocytes, and oligodendrocytes. To examine whether GFP-positive cells from the transgenic animals could produce cells of different lineages, we grew neurospheres from these cells and tested their differentiation potential. Cells were isolated from the adult brain and placed at clonal density (<10,000 cells/ml) in 1.2% methylcellulose to prevent their aggregation and to ensure that the resulting neurospheres were derived from single cells. They were incubated in the presence of FGF and EGF until neurospheres were formed from some of the GFP-positive cells. These neurospheres were then plated onto a coated surface and incubated further. The neurospheres that were generated from individual GFP-positive cells contained a large number of bright GFP-expressing cells (Fig. 10; although some heterogeneity with respect to GFP expression was apparent). However, as soon as the neurospheres were plated on the laminin substrate, the fluorescence of the sphere started to decline dramatically; after 2 weeks, only very few cells, mostly in the center of the attached sphere, were GFP positive. At the same time, most of the cells within the neurosphere and those that emerged and migrated away from the neurosphere expressed markers of differentiated cells: βIII-tubulin, GFAP, or CNPase (Fig. 10).
Nestin-GFP-positive cells are strongly enriched in neurosphere-forming cells; most of the neurosphere-forming cells of the adult brain are nestin-GFP positive
We next explored the feasibility of isolating a purified population of GFP-positive cells from the brain of nestin-GFP mice. To that end, we performed fluorescent-activated sorting of cells from the subventricular region of the early postnatal brain. This region was surgically isolated to enrich for nestin-GFP–expressing cells. The starting preparation contained 22% GFP-positive cells; after one round of sorting, the population contained 93% GFP-positive cells. This population was subjected to another round of sorting, after which 99% of the sorted cells were GFP positive (Fig. 11A). Thus, it is possible to isolate a highly purified population of nestin-GFP cells directly from the brains of these transgenic animals.
We next investigated whether the population of nestin-GFP cells in the adult brain (>2 months) was enriched in stem cells and measured the fraction of stem cells present among the nestin-GFP cells. Because neural stem cell properties strongly correlate with the ability to form neurospheres, we separated GFP-positive and GFP-negative cells with a fluorescence-activated cell sorter and compared the number of the neurosphere-forming cells in each population. Nestin-GFP-positive cells, which represented approximately 5% of the sorted population, formed 316 ± 59 neurospheres per 100,000 plated cells, whereas the nestin-GFP-negative population produced 0.22 ± 0.09 neurospheres per 100,000 plated cells (Fig. 11B; most of the neurospheres from the sorted GFP-negative population were small and atypical and did not produce neurons). This method demonstrated a 1,440-fold higher efficiency in the ability of nestin-GFP–expressing cells to form neurospheres as compared with their GFP-negative counterparts. Moreover, these data indicated that most of the cells in the adult brain that are capable of forming neurospheres are contained within the nestin-GFP–expressing group (i.e., GFP-positive cells gave rise to ∼70 times more neurospheres than did all the GFP-negative cells).
Together, these results show that nestin-GFP–expressing cells from the brain of adult transgenic mice can form neurospheres and can generate differentiated cells of neuronal and astrocytic lineages, and that these cells contain the bulk of neurosphere-forming cells of the brain, suggesting that these cells correspond to multipotent stem cells in the adult brain.
Our study presents a detailed characterization of a transgenic mouse strain designed to allow rapid identification, direct visualization, and ready isolation of neural stem and progenitor cells from the developing and adult brains. It provides a simple animal model system to follow the development of the nervous system, to characterize and dissect the pathways of neurogenesis, and to examine the response of neural stem cells to external stimuli. We use these animals to demonstrate the presence of two distinct groups of neuronal precursors in the adult brain and to reveal a population of cells with an unusual morphology in the DG.
One of the central conclusions of this study is that expression of GFP marks neural stem cells in our nestin-GFP transgenic mice. This conclusion is based on several lines of evidence: First, the transgene is expressed in those areas of the developing embryo that correspond to the neuroepithelial cells of the developing nervous system. Second, it is expressed in those areas of the adult brain (olfactory subependyma and DG), which are marked by persistent production of new neurons. Third, GFP expression is absent in those cells that have already undergone differentiation and in those areas of the brain that contain fully differentiated cells. Fourth, the sites of the transgene expression in the developing and adult nervous systems overlap with the sites of expression of nestin, which has served as a reliable marker of neural stem cells. Fifth, the GFP-positive cells are capable of forming neurospheres and producing a variety of types of progeny in vitro. Sixth, these cells are strongly (∼1,400-fold) enriched with neurosphere-forming cells. Seventh, most of the neurosphere-forming cells of the adult brain are present within GFP-expressing cells. Together, these results indicate that GFP-positive cells in the nestin-GFP transgenics are an accurate representation of neural stem cells in the developing and adult nervous systems.
An emerging concern regarding the use of transgenic reporters to examine the pattern of gene expression relates to the turnover of surrogate reporter proteins, such as β-galactosidase and GFP, in the cell. Although the presence of mRNA encoding the GFP transgene may closely reflect the temporal profile of the endogenous nestin mRNA, it is conceivable that the GFP protein remains in the neural stem/progenitor cell longer than does the endogenous nestin protein. This may result in scoring as stem cells, those GFP-positive cells that have already progressed along the differentiation cascade, lost their multipotentiality, and started to express progeny-specific, e.g., neuronal, markers. Such issues did not arise with our transgenic animals, because the GFP signal was not detected in those areas of the nervous system that contain differentiated cells. Moreover, we generated transgenic animals that carry the same nestin regulatory sequences driving expression of a version of GFP modified to have a short half-life in mammalian cells (Li et al., 1998); the pattern of GFP expression in such animals is very similar to that described in this report (Mignone and Enikolopov, unpublished observations). Furthermore, there are two specific examples presented in this paper, which strongly argue against this possibility. First, in the developing brain, there are distinct zones of GFP-positive cells and cells staining for βIII-tubulin (Fig. 2). Because cell division and migration to the periventricular zone occur rapidly at this stage (Caviness et al., 2000), our results suggest that cells lose their GFP signal soon after they start the process of neuronal differentiation. Second, in the DG of the adult brain, GFP-positive cells are seen only in the subgranular cell layer (in which neuronal precursors reside), not in the granule cells layer (to which precursors migrate when they start to differentiate; Figs. 5–7). Together these results argue that potential difference in half-life time of nestin versus GFP does not confound the interpretation of our transgenic mouse experiments.
The nestin-GFP mouse strain that we generated allows easy and immediate identification of neural stem cells in vivo and helps to avoid the artifacts and difficulties attendant to the current methodologies (prolonged processing, loss of weak signals, false positive signals, etc.). In addition to using nestin-GFP mice to isolate and manipulate neural stem cells, we observed the fine cellular anatomy of these GFP-positive cells. For instance, we found that the subgranular cell layer of the DG contains an unusual type of cells with highly branched morphology that traverse the granule cell layer and terminate in the molecular layer with an elaborated arbor-like structure. Importantly, these cells express GFAP. These cells resemble a GFAP-expressing subclass of cells in the DG (Kosaka and Hama, 1986; Seri et al., 2001) whose somata reside in the subgranular layer and which send radial fibers to the molecular layer. They also bear resemblance to Bergmann glia cells of the cerebellum, Müller glia cells of the retina, and the neurogenic radial glia cells of the developing neocortex (Miyata et al., 2001; Campbell and Gotz, 2002).
The main regulatory elements that drive the expression of GFP transgene in neural stem cells reside in the second intron of the nestin gene. Members of the POU transcription factor family bind to the nestin neural enhancer and establish neuroepithelial cell specificity during embryonic development (Josephson et al., 1998). Although the endogenous nestin gene is expressed in a range of tissues in developing and adult mice, by using only the selected regulatory elements of the second intron and the upstream flanking area, we narrowed the spectrum of cells in which the transgene is expressed and directed the GFP fluorescence more specifically to neural stem cells.
At the same time, in addition to GFP-positive cells in the neurogenic areas of the brain, we noticed GFP-expressing cells in some other locations and tissues in the nestin-GFP transgenic animals (unpublished observations). The detection of these cells may simply reflect increased sensitivity of detection of nestin-positive cells through GFP fluorescence (as compared with immunodetection of nestin) or may reflect spurious expression of the transgene. However, they could also represent cells with stem-like features in the regions of the nervous system heretofore not considered to be neurogenic. Furthermore, green fluorescence in cells outside the nervous system may mark stem cells for other (nonneural) lineages (Li et al., 2003). It will be an interesting challenge to follow these GFP-expressing cells in neural and non-neural tissues and examine their potential to serve as stem cells for the tissues in which they reside.
An important conclusion from our experiments is that GFP-positive cells of the RMS appear to comprise two classes. One group of cells expresses high levels of GFP and is positive for the presence of GFAP. Another group expresses significantly lower level of GFP and is positive for βIII-tubulin. This suggests that these two groups of nestin-GFP cells correspond to different functional classes, perhaps representing precursor cells at different stages of the differentiation cascade. This explanation is consistent with the recent finding that neural precursors in the adult brain can be divided into different classes that may represent the degree of progression through the differentiation cascade (Doetsch et al., 1997, 1999). It has been suggested that neural stem cells have properties of astrocytes, including bundles of GFAP in their processes; these cells (type B cells) give rise to rapidly dividing precursors (type C cells) that in turn give rise to migrating neuroblasts (type A cells; Doetsch et al., 1997, 1999; Laywell et al., 2000). These advanced migrating neuroblasts (type A precursors), but not astrocyte-like type B stem cells, are positive for βIII-tubulin, which is an early indicator that they will later convert into fully differentiated neurons. All three classes of cells express nestin, although the expression level is highest in type B cells. This scheme corresponds well with our observation of GFP-bright cells (which stain for GFAP but not for βIII-tubulin) and GFP-dim cells (which stain for βIII-tubulin but not for GFAP). Thus, our data are compatible with the idea that GFP-bright cells correspond to type B cells (stem cells), whereas GFP-dim cells correspond to type C and A cells (more advanced precursors). If this assumption is correct, it may be possible to isolate distinct classes of neuronal precursors directly from the brains of nestin-GFP transgenic animals.
Another important conclusion from our study is that not only nestin-GFP cells are enriched in stem cells (as judged by their higher efficiency in forming neurospheres) but also most of the neurosphere-forming cells of the adult brain are nestin-GFP positive. The significance of this observation is that the properties of nestin-GFP cells will reflect the properties of the majority of stem cells of the adult brain.
While this work was in progress, reports were published on the generation of two transgenic mouse lines by using regulatory elements from the second intron of the nestin gene (Yamaguchi et al., 2000; Kawaguchi et al., 2001; Sawamoto et al., 2001). Limited data on the patterns of GFP expression and their relation to markers of neural differentiation are presented in these reports. Because the specifics and the arrangements of the various regulatory elements differ between these three transgenic models, it will be informative to carefully compare the patterns of transgene expression in different lines and to explore the stem cell potential of the GFP-expressing cells from these different mouse lines.
In summary, our transgenic reporter mouse line provides a novel and useful starting point for studying generation of new neurons in the developing nervous system and in the adult brain. These mice also can be used for examining the response to aging, injury, and therapeutic regimens. They also present an approach to study the emerging issue of plasticity of stem cells from various tissues. They also can be used to isolate neural stem and progenitor cells from the developing and adult organisms. Thus, they can be used as a ready source of neuronal precursors to study RNA and protein profiles of neural stem cells, identify surface markers, and explore transplantation opportunities, for which there are often no adequate methodologies. In addition, these transgenic mice can be used as a reporter line by crossbreeding them with other mutant mouse lines to examine the effects of genetic mutations on neural stem cells. Further, they can be used to follow the physiologic response of neural stem cells to various disorders of the nervous system (neurodegeneration, traumatic brain injury, and cancer) and to the action of therapeutic drugs.
We are grateful to Ron McKay and Lyle Zimmerman for their generous gift of nestin regulatory sequences. We thank Chris Walsh for his gift of antibodies to doublecortin. We thank Stephen Hearn for help with confocal microscopy. We thank Enikolopov laboratory members for useful discussions. We are grateful to Julian Banerji for critical reading of the article.