Jackson et al. (2006) have reported that adult glial fibrillary acid protein (GFAP)-expressing neural stem cells (NSCs) also express platelet-derived growth factor (PDGF) receptor-α (PDGFRα), and that their stimulation by PDGF induced the formation of a glioma-like mass. Here, we reexamined the relationship between PDGFRα and GFAP expression within the three-dimensional organization of the adult periventricular area. Using four independent PDGFRα antibodies, we found that adult mouse GFAP-expressing NSCs and PDGFRα-expressing cells represent two distinct populations of neural precursors. Examination of the adult periventricular area in a mouse line that expresses nuclear-localized enhanced green fluorescent protein under the control of the PDGFRα promoter confirmed that GFAP-expressing NSCs do not express PDGFRα. Furthermore, PDGF-responsive neural precursors were found at least one cell layer subjacent to the ependymal layer, and were evenly distributed across the lateral ventricular wall, which contrasts with the reported patchy and often ependymal localization of adult GFAP-expressing NSCs. Adult human PDGFRα-expressing neural precursors were also found not to express GFAP. PDGF-responsive neural precursors, but not GFAP-expressing NSCs, responded to infusions of PDGF by generating glioma-like masses. Our results do not support the view that GFAP-expressing NSCs are the origin of glioma-like masses that form after intraventricular PDGF infusion.
Adult neural stem cells (NSCs) are targets of strategies for CNS repair after injury or disease and may contribute to brain tumor formation. NSCs in the adult rodent brain reside in the periventricular area of the lateral ventricles (Reynolds and Weiss, 1992) and the hippocampus (Palmer et al., 1997). Immunological and cytoarchitectural analyses have revealed that adult periventricular NSCs express GFAP, are located within the ependymal and subependymal layers, and are concentrated in pockets along the medial and lateral walls of the lateral ventricles (Doetsch et al., 1999; Merkle et al., 2007; Mirzadeh et al., 2008; Shen et al., 2008). NSCs also populate the adult human hippocampus (Eriksson et al., 1998; Kukekov et al., 1999; Arsenijevic et al., 2001) and periventricular area (Kirschenbaum et al., 1994; Pincus et al., 1997; Sanai et al., 2004; Curtis et al., 2007).
Several distinct populations of neural precursors that express platelet-derived growth factor receptor-α (PDGFRα) reside throughout the mouse and human brain (Rakic and Zecevic, 2003; Kessaris et al., 2006; Parras et al., 2007). Termed oligodendrocyte progenitor cells (OPCs) and PDGF-responsive neural precursors (PRPs) (Chojnacki and Weiss, 2004), they are slowly dividing and primarily generate oligodendrocytes (ffrench-Constant and Raff, 1986; Wolswijk and Noble, 1989; Gregg et al., 2007). OPCs/PRPs and adult periventricular NSCs are also thought to be two distinct neural precursor populations.
Expression of PDGFRα and the lack of GFAP expression by OPCs/PRPs used to distinguish them from GFAP-expressing periventricular NSCs. Adult periventricular NSCs were initially isolated by their dependence on epidermal growth factor (EGF) to proliferate and form spherical clusters of undifferentiated cells, termed neurospheres (Reynolds and Weiss, 1992). It was later shown that GFAP-expressing NSCs activated by the cytosine arabinoside-induced death of their progeny, the transit-amplifying progenitors, express the EGF receptor (EGFR) (Pastrana et al., 2009). Furthermore, EGFR and PDGFRα label distinct neural precursor populations during embryonic development (Chojnacki and Weiss, 2004) and in the adult periventricular area (Jackson et al., 2006). However, postnatal and adult OPCs/PRPs have now been reported to express EGFR (Aguirre et al., 2004). Similarly, NG2 expression, previously described as specific for OPCs/PRPs (Nishiyama et al., 1996), has been found in EGFR-expressing transit-amplifying progenitors (Aguirre et al., 2004), although this has been disputed (Cesetti et al., 2009; Komitova et al., 2009). Recently, PDGFRα has been reported to be expressed by a subset of adult periventricular GFAP-expressing NSCs (Jackson et al., 2006), suggesting that it is not specific to OPCs/PRPs. This observation and the formation of glioma-like masses after intraventricular infusions of PDGF led Jackson et al. (2006) to conclude that adult periventricular GFAP-expressing NSCs are a potential source for PDGFRα-overexpressing brain tumors. However, it has now been reported that the adult ventricular surface contains multiple distinct populations of NSCs (Merkle et al., 2007), but whether there are differences in PDGFRα expression between these populations has not been examined.
To determine whether there are differences in PDGFRα expression among the multiple adult periventricular NSC populations, we reexamined the expression of PDGFRα in the adult periventricular area. Using both immune- and transgenic mouse-based techniques, we found that neither adult mouse nor human GFAP-expressing NSCs expressed PDGFRα. Furthermore, PDGFRα expression was found to be evenly distributed along the ventricular wall, and at least one cell layer subjacent to the ependymal layer, in contrast to the reported distribution of GFAP-expressing NSCs (Mirzadeh et al., 2008; Shen et al., 2008). In addition, only parenchymal infusion of PDGF generated glioma-like masses, whereas intraventricular infusions had no noticeable effect on proliferation. Our results suggest that PDGFRα expression distinguishes between OPCs/PRPs and adult GFAP-expressing NSCs.
Materials and Methods
CD-1 mice stocks were maintained in the University of Calgary Animal Resources Center. B6.129S4-Pdgfratm11(EGFP)Sor/J mice were obtained from The Jackson Laboratory and stocks were maintained in the University of Calgary Animal Resources Center.
Human brain tissue.
Human fetal [21, 21, and 17 gestational weeks (g.w.)] brain tissue was obtained from therapeutic abortions. Adult brain tissue [corpus callosum (CC) from a 20 year old and a 16 year old, temporal lobe periventricular area from a 35 year old] was obtained at the time of resection from the Foothills Hospital, Calgary, Alberta.
Embryonic day 14 (E14) mouse brains were fixed for 4 h with ice-cold 4% paraformaldehyde. Twelve-micrometer sections were obtained after the fixed brains were incubated for 24 h in 10% sucrose followed by 24 h in 25% sucrose, and mounted in OCT (Tissue Tek). Adult male CD1 mice were killed by an intraperitoneal injection of sodium pentobarbital (11 mg) followed by transcardial perfusion with 20 ml of ice-cold PBS, 20 ml of ice-cold 4% paraformaldehyde, and an 8 h postfix in ice-cold 4% paraformaldehyde. For periventricular area staining, sections (30 μm) were obtained using a Leica VT100S vibratome. Human fetal and adult brain tissue was fixed for 24 h in ice-cold 4% paraformaldehyde and processed for sectioning as described for E14 mouse brains. Some sections processed for staining with the Spring Bioscience antibodies received a 10 min boiling citrate buffer wash. E14 mouse brain sections were then incubated with the primary antibodies for 2 h at 37°C diluted in PBS containing 0.3% Triton X-100 (if diluted) at the following dilutions: R&D Systems goat anti-mouse PDGFRα (1:10), Santa Cruz rabbit anti-PDGFRα (1:10), Spring Bioscience mouse-anti PDGFRα (neat), Spring Bioscience rabbit anti-PDGFRα (neat; lot #s 61212 and 70629), Stallcup rabbit anti-PDGFRα (1:100; kind gift from Dr. Bill Stallcup, Sanford-Burnham Medical Research Institute, La Jolla, CA), and Santa Cruz goat anti-phospho PDGFRα (SC-12910: 1:50). Sections were then incubated for 1 h at room temperature in PBS containing 10% normal donkey serum (NDS) followed by a 1 h incubation at 37°C with the appropriate donkey secondary antibodies (all from Jackson ImmunoResearch). Adult mouse brain floating vibratome sections were incubated for 24 h at room temperature with the primary antibodies in PBS containing 0.3% Triton X-100 (PBST: if diluted) at the following dilutions: Millipore Bioscience Research Reagents (MBRR) mouse anti-GFAP (1:1000), BTI rabbit anti-GFAP (1:400), R&D Systems goat anti-mouse PDGFRα (1:50), Santa Cruz rabbit anti-PDGFRα (1:100), Spring Bioscience mouse-anti PDGFRα (neat), Spring Bioscience rabbit anti-PDGFRα (1:10; lot # 61212), and Stallcup rabbit anti-PDGFRα (1:200). Sections were then incubated for 1 h at room temperature in PBS containing 10% NDS followed by a 1 h incubation at 37°C with the appropriate donkey secondary antibodies. For the infusion experiments, once fixed the brains were incubated for 24 h in 10% sucrose followed by 24 h in 25% sucrose, and mounted in OCT (Tissue Tek). Staining of the sections followed the same procedure used for the embryonic mouse brain sections using the primary antibodies at the following dilutions: R&D Systems goat anti-mouse PDGFRα (1:10), MBRR mouse anti-GFAP (1:1000), and MBRR rabbit anti-NG2 (1:200). For BrdU (Serotec rat anti-BrdU; 1:100) and PDGFRα (R&D System; 1:10), double-labeling sections were incubated in acetone for 30 s at room temperature, 20 min in 1N HCl at 60°C, 24 h incubation at room temperature with primary antibodies, 1 h in PBS containing 10% NDS, and 2 h at room temperature in the appropriate secondary antibodies. This was followed by staining with the appropriate secondary antibodies for 24 h at 4°C in PBS containing 0.3% Triton X-100 as well as Hoechst 33258 added for 1 h for the first wash. Staining of human brain tissue followed the same procedure used for E14 mouse brain sections except using the primary antibodies at the following dilutions: MBRR mouse anti-GFAP (1:500), R&D Systems goat anti-human PDGFRα (1:20), Spring Bioscience rabbit anti-PDGFRα (neat). Fluorescence-labeled slides were mounted with Fluorsave (Calbiochem). Slides processed for hematoxylin and eosin staining were mounted with Cytoseal XYL (Richard-Allan Scientific). Postnatal day 0 mouse cortices were dissociated and plated on poly-l-ornithine-coated glass coverslips and cultured in a serum-free media hormone mix (MHM) (Chojnacki and Weiss, 2008). After 24 h, the cells were either left in MHM or stimulated with PDGF-AA (100 ng/ml) for 3 h. The cells were then fixed for 20 min in ice-cold 4% paraformaldehyde and stained for PDGFRα expression using the R&D systems goat-anti-mouse PDGFRα antibody (1:20). Images were captured on an Axiocam camera (Zeiss) mounted on a Zeiss Axioplan2 microscope (Zeiss) using Axiovision software, and figures were composed in Adobe Photoshop 6.0 (Adobe Systems). Confocal images where captured on an Olympus Optical Fluoview BX-50 laser scanning confocal microscope using Fluoview software.
Whole mounts of the adult lateral ventricular wall from B6.129S4-Pdgfratm11(EGFP)Sor/J mice were dissected out as described by Mirzadeh et al. (2008) and then fixed for 30 min in 4% ice-cold paraformaldehyde as described by Shen et al. (2008). Whole mounts were then rinsed four times for 15 min per wash with PBS followed by a 1 h block at room temperature in PBST containing 10% NDS. This was followed by a 24 h incubation at 4°C in PBST containing 10% NDS with rabbit anti-GFAP (BTI; 1:400), mouse anti-β-catenin (BD Biosciences; 1:500), or Santa Cruz goat anti-phospho PDGFRα (SC-12910: 1:50). Afterward, whole mounts were rinsed four times with PBST (15 min per wash) and then incubated for 2 h at room temperature in the appropriate secondary antibodies. This was followed by four 15 min washes in PBST and a 1 h incubation with Hoechst 33258. Whole mounts were mounted in Fluorsave. Six-hundred-nanometer confocal sections of whole mounts were captured on an Olympus Optical Fluoview 1000 laser scanning confocal microscope using Fluoview software.
PDGF infusions and BrdU administration.
Male CD1 mice (8–9 weeks) were anesthetized with a 1:10 dilution of sodium pentobarbital in saline (120 mg per kg of body weight, i.p.) and implanted with an osmotic pump (Alzet 1007D, Alza Corp.), placed dorsal subcutaneously. PDGF-AA (human recombinant; R&D Systems and Peprotech) was dissolved in 0.9% saline containing 1 mg/ml bovine serum albumin (Sigma) at a concentration of 3.3 ng/μl. Mice received an infusion of 40 ng of PDGF-AA per day, as the manufacturer-specified flow rate is 0.5 μl/h. Control mice were implanted with an osmotic pump containing 1 mg/ml bovine serum albumin (Sigma). The contents of the osmotic pump were channeled though tubing attached to a cannula that was implanted into either the lateral ventricles of the brain (+0.2 mm anterior/posterior, +0.8 mm medial/lateral lateral, and dorsoventral −2.5 mm below the dura with the skull leveled between lambda and bregma) or the parenchyma lateral to the lateral ventricle (+0.2 mm anterior/posterior, +2.0 mm medial/lateral lateral, and dorsoventral −2.5 mm below the dura with the skull leveled between lambda and bregma). Each animal was infused for 7 consecutive days. A minimum of three mice were infused per group. One hour before the animals were killed, 1 injection of BrdU (Sigma, 120 mg per kg, dissolved in 0.007% NaOH in phosphate buffer) was administered intraperitoneally.
Identification of bona fide PDGFRα expression in the embryonic mouse forebrain
To determine whether PDGFRα expression can be used to distinguish between OPCs/PRPs and adult periventricular GFAP-expressing NSCs, we tested the specificity of five different antibodies directed against the protein. We tested an R&D Systems goat anti-mouse PDGFRα and a Santa Cruz rabbit anti-PDGFRα antibody (Chojnacki and Weiss, 2004), an anti-PDGFRα from Dr. Bill Stallcup (rabbit; ectodomain) (Nishiyama et al., 1996), and two Spring Biosciences antibodies (mouse and rabbit derived) used by Jackson et al. (2006). With the antibodies, we stained frozen sections of the E14 ventral forebrain wherein PDGFRα mRNA has been demonstrated to be present (Pringle and Richardson, 1993). Both the R&D Systems (Fig. 1A, red) and Santa Cruz (Fig. 1B, green) antibodies detected cells within the E14 ventral forebrain, and dual labeling revealed that they labeled the same population of cells (Fig. 1C,D, orange). This suggested that true PDGFRα expression was being detected by the R&D Systems and Santa Cruz antibodies. Both the Stallcup (Fig. 1F) and Spring Biosciences rabbit (Fig. 1J) antibodies colabeled with the R&D Systems PDGFRα antibody (Fig. 1E–L), suggesting that these antibodies also detected bona fide PDGFRα expression. In contrast, the Spring Bioscience mouse-anti PDGFRα antibody failed to strongly label cells in the ventral forebrain (Fig. 1M–P), regardless of whether the tissue was unmasked using a manufacturer-recommended citrate buffer wash (Fig. 1Q–T). The faint staining that was observed did not colabel with R&D Systems' goat anti-mouse PDGFRα antibody (Fig. 1Q–T).
We further confirmed that the R&D Systems goat anti-mouse PDGFRα antibody was specifically detecting PDGFRα expression by using the B6.129S4-Pdgfratm11(EGFP)Sor/J mouse strain. In this strain of mice, the endogenous PDGFRα promoter drives the expression of an EGFP-H2B fusion protein, which targets EGFP-H2B to the nucleus of PDGFRα-expressing cells. In E14 brain sections of B6.129S4-Pdgfratm11(EGFP)Sor/J heterozygous mice, confocal microscopy revealed that all cells that were labeled by the R&D Systems goat anti-mouse PDGFRα antibody also expressed EGFP-H2B (Fig. 1U–X). In addition, when postnatal day zero mouse cortex was dissociated, plated, and after 24 h cultured for a further 3 h in either MHM or PDGF (100 ng/ml), the localization of PDGFRα as detected by the R&D Systems antibody changed. In the absence of PDGF, PDGFRα staining was membrane bound (Fig. 1Y,Y1), whereas after 3 h in the presence of PDGF, PDGFRα expression was internalized (Fig. 1Z,Z1). This agrees well with the known internalization of PDGF receptors after ligand binding (Heldin et al., 1982; Coats et al., 1994), and provides further evidence that the R&D Systems anti-mouse PDGFRα antibody specifically detects the protein. The data suggest that four of the tested antibodies were suitable for determining whether expression of PDGFRα could distinguish between adult periventricular GFAP-expressing NSCs and OPCs/PRPs.
In addition, we examined the specificity of SC-12910, an antibody directed against phosphorylated PDGFRα that was also used to demonstrate that GFAP-expressing periventricular NSCs express the PDGFRα receptor (Jackson et al., 2006). In E14 brain sections of B6.129S4-Pdgfratm11(EGFP)Sor/J heterozygous mice, confocal microscopy revealed that the majority of EGFP-H2B-expressing cells were not labeled by SC-12910 (Fig. 2A--D). In addition, most of the cells labeled by SC-12910 were not EGFP-H2B positive (Fig. 2C, arrows), demonstrating the nonspecificity of the antibody. SC-12910 also nonspecifically labeled cells in whole mounts of the adult mouse periventricular area (Fig. 2E--G, and arrows in H). Together, the data suggest that SC-12910 is not suitable for the detection of phosphorylated PDGFRα in immunohistochemistry applications.
PDGFRα is not expressed by adult mouse periventricular GFAP-expressing NSCs
We reexamined the expression of PDGFRα in the adult mouse brain periventricular area. We first tested two different GFAP antibodies to ensure that we could detect bona fide GFAP expression. Confocal microscopy of sections of the adult mouse periventricular area stained with the MBRR mouse anti-GFAP and BTI rabbit-anti-GFAP antibodies revealed identical staining patterns (Fig. 3A–C). This suggested that the two antibodies were indeed specific for GFAP. Next, we examined whether any population of cells colabeled for GFAP and PDGFRα expression in 30 μm coronal vibratome sections of the adult mouse periventricular area. All of the antibodies that detected PDGFRα expression in the E14 ventral forebrain, labeled cells that had multiple processes in the corpus callosum and within the periventricular area, reminiscent of OPCs/PRPs (Fig. 3D,G,J,M). Both the R&D Systems and Stallcup antibodies strongly labeled these cells, whereas the Santa Cruz and Spring Bioscience rabbit antibody showed weaker staining. None of the PDGFRα antibodies colabeled GFAP-expressing cells (Fig. 3D–O). In contrast, the Spring Bioscience mouse anti-PDGFRα antibody labeled GFAP-expressing periventricular astrocytes (Fig. 3P–R). Use of the recommended citrate buffer unmasking wash resulted in no detectable staining (data not shown). Furthermore, the Spring Bioscience mouse anti-PDGFRα antibody failed to costain cells labeled with the R&D Systems anti-mouse PDGFRα antibody in sections of the adult mouse brain (Fig. 3S–U).
Our observations revealed that adult periventricular GFAP-expressing NSCs within the lateral ventricular wall never colabeled for PDGFRα in 30 μm coronal sections of the adult mouse brain. However, several different adult periventricular NSCs populations are localized in concentrated pockets along the ventricular wall (Merkle et al., 2007). Therefore, it remained a possibility that the coronal sections we examined missed a subpopulation of GFAP-expressing cells that express PDGFRα. In addition, the dense network of GFAP-positive fibers intermixed with PDGFRα-positive processes also made it difficult to ensure that we had not missed a minor population of cells expressing both proteins. Therefore, we processed whole mounts of the adult lateral ventricular wall obtained from B6.129S4-Pdgfratm11(EGFP)Sor/J heterozygous mice for immunohistochemistry against GFAP and β-catenin (to label cell membranes and identify the ventricular surface). We used these animals specifically because the nuclear localization of EGFP-H2B, indicative of PDGFRα expression, would make it easier to colocalize PDGFRα expression within clearly GFAP-expressing NSCs, should such double-labeled cells exist. Thus, the effect of the dense meshwork of GFAP-positive fibers could be neutralized. Confocal analysis (Fig. 4) revealed that within the posterior dorsal wall of the adult lateral ventricle, an area enriched for GFAP-expressing NSCs (Mirzadeh et al., 2008), no cells expressing EGFP-H2B colabeled for GFAP (three whole-mount ventricular walls from different animals; #1: 191 EGFP-H2B cells in fifteen 60× fields; #2: 100 EGFP-H2B cells in ten 60× fields; #3. 71 EGFP-H2B cells in five 60× fields). Furthermore, lack of PDGFRα and GFAP colabeling was not due to the absence of ventricular surface-contacting GFAP-expressing NSCs within the selected fields. Within whole mounts 2 and 3, we identified 207 GFAP-expressing NSCs (average of 13.8 per field). These cells were either fully intercalated between ependymal cells as described by Shen et al. (2008) or extended a thin process toward the ependymal layer as described by Mirzadeh et al. (2008) (Fig. 4B,E,F). None of the 207 GFAP-expressing NSCs coexpressed PDGFRα-driven EGFP-H2B (Fig. 4D,H,K). Furthermore, EGFP-H2B-expressing nuclei were always at least one cell layer subjacent to the ependymal layer (Fig. 4I–K). The data suggest that adult GFAP-expressing cells do not express PDGFRα, and that PDGFRα expression distinguishes between adult mouse periventricular NSCs and OPCs/PRPs.
PDGFRα-expressing cells in the human adult periventricular area do not express GFAP
It is possible that adult periventricular GFAP-expressing NSCs of the adult human brain may have antigen profiles distinct from those of the adult mouse brain. Therefore, we first sought to determine whether we could also detect bona fide PDGFRα expression in sections of the fetal human brain. We used the Spring Bioscience rabbit anti-PDGFRα antibody in conjunction with an R&D Systems goat anti-human PDGFRα antibody to stain for PDGFRα expression in sections of adult human brain. Confocal microscopy revealed that both antibodies labeled the same population of cells in three different samples of sectioned fetal human brain tissue (Fig. 5A–C). Furthermore, both antibodies labeled cells with multiple processes in two different samples of human corpus callosum (Fig. 5D–F); however, use of the R&D Systems antibody resulted in stronger and more specific staining. Next, we used the R&D Systems goat anti-human PDGFRα antibody together with an anti-GFAP antibody to examine the expression patterns of the proteins in the human temporal lobe periventricular area. Confocal microscopy revealed that PDGFRα-expressing cells within the adult human periventricular area did not express GFAP (Fig. 5G–J). Furthermore, we examined the periventricular white matter in the same section, which contained an abundance of PDGFRα-expressing cells and cells expressing GFAP, and found that none of the cells expressed both antigens (Fig. 5K). The data suggest that adult human periventricular GFAP-expressing cells do not express PDGFRα, and that adult human NSCs and OPCs/PRPs can be distinguished from each other by the expression of PDGFRα.
Infusion of PDGF-AA into the parenchyma, but not the lateral ventricle, of the adult mouse brain generates a cellular mass
Jackson et al. (2006) reported that infusion of PDGF into the lateral ventricles led to the generation of a glioma-like mass on either the lateral or medial side of the infused ventricle. This was attributed to the expansion of PDGFRα-expressing periventricular GFAP-expressing NSCs. Our examination of PDGFRα and GFAP expression in cells of the adult mouse periventricular area revealed that the two proteins were not coexpressed (Figs. 3, 4). Therefore, it was possible that the generation of a glioma-like mass after intraventricular infusion of PDGF-AA was actually the result of OPC/PRP proliferation. Like Jackson et al. (2006), we infused 40 ng of PDGF-AA per day or vehicle control for 7 d into the lateral ventricles, and examined the infused brains using hematoxylin and eosin staining. We sectioned rostrally and caudally to the infusion site, and found that we had correctly targeted the lateral ventricle as evidenced by the needle tract (Fig. 6A–F). However, we did not observe the formation of a cellular mass on either the lateral or medial side of the ventricle regardless of whether we infused PDGF-AA procured from R&D Systems (Fig. 6A–F) (n = 3) or Peprotech (n = 4, data not shown). Interestingly, the formation of a cellular mass that could be mistaken as originating from the lateral ventricle was only observed when the cannula missed the lateral ventricle and was infusing PDGF into the parenchyma immediately medial or lateral to it (Fig. 6G–H1). This suggested that OPCs/PRPs, present throughout the parenchyma of the adult brain, were responsible for the generation of cellular masses after PDGF infusion. To test this, we infused PDGF-AA or vehicle control, from either R&D Systems or Peprotech, into the parenchyma lateral to the lateral ventricle for 7 d, and examined the sections of the infused brains with hematoxylin and eosin staining. Vehicle infusion did not result in the generation of a cellular mass at the infusion site or rostral or caudal to it (Fig. 6I–K). However, PDGF-AA infused into the parenchyma resulted in a cellular mass that extended at least 0.5 mm rostral and caudal to the site of infusion (n = 8) (Fig. 6L–N). The core of the cellular mass was composed of some autofluorescent background (Fig. 7, asterisks) as well as PDGFRα-expressing cells that also labeled for BrdU (Fig. 7A–D), which was administered 1 h before the mice were killed. Despite the abundant GFAP expression observed in the overlying corpus callosum, these dividing PDGFRα-expressing cells lacked GFAP expression (Fig. 7E–H). However, the PDGFRα-expressing cells also labeled for the OPC/PRP antigen NG2 (Nishiyama et al., 1996) (Fig. 7I–L). Together, the data suggest that the cellular mass generated after PDGF infusion into the adult mouse brain is generated by OPCs/PRPs resident in the parenchyma and not by adult periventricular NSCs.
Jackson et al. (2006) reported that GFAP-expressing NSCs in the adult periventricular area express PDGFRα. Furthermore, they found that infusion of PDGF into the mouse lateral ventricles promoted the expansion of the periventricular area into glioma-like masses, which suggested a link between periventricular GFAP-expressing NSCs and gliomas. In contrast, our findings unambiguously demonstrate that neither mouse nor human periventricular area GFAP-expressing NSCs express PDGFRα. Furthermore, our results strongly suggest that it is PDGFRα-expressing OPCs/PRPs populating the parenchyma of the adult brain that generate cellular masses after PDGF infusions.
In our study several lines of evidence suggest that PDGFRα expression is a feature that distinguishes OPCs/PRPs from adult periventricular GFAP-expressing NSCs. First, examination of GFAP and PDGFRα labeling in the adult periventricular area demonstrated mutually exclusive staining patterns. We were also able to reproduce the double labeling of PDGFRα and GFAP observed by Jackson et al. (2006), but only by using a mouse antibody from Spring Bioscience directed against PDGFRα, which we determined does not detect the protein in vivo (Figs. 1, 3). Second, PDGFRα expression, as revealed by EGFP-H2B expression, was evenly distributed along the ventricular wall and was found at least one layer subjacent to ependymal cells. This is in contrast to the reported localized concentrations of adult periventricular GFAP-expressing NSCs along the ventricular surface (Merkle et al., 2007), and the observation that a subpopulation are found intercalated within the ependymal layer (Shen et al., 2008).
Third, presumed expression of PDGFRα by adult periventricular GFAP-expressing NSCs would have predicted their expansion after intraventricular PDGF infusion. In our study, we found that PDGF-AA infused into the lateral ventricle never resulted in the formation of a cellular mass on either the lateral or medial sides of the ventricle. This may in part be because we found that PDGFRα-expressing cells were situated at least one cell layer subjacent to the ependymal layer (Fig. 4), and may therefore not have access to signals reaching the ventricular surface. In contrast, parenchymal infusions just lateral or medial to the lateral ventricle gave the appearance of glioma-like masses expanding outward from either the medial or lateral walls of the lateral ventricle. Similarly, the glioma-like masses formed after PDGF infusion originated from either the medial and lateral walls, but not both simultaneously (Jackson et al., 2006). In contrast, intraventricular EGF infusion results in the simultaneous expansion of neural precursors on both the lateral and medial aspects of the ventricle (Craig et al., 1996). Our findings are in agreement with two recently published studies that found that NG2-expressing precursors that presumptively express PDGFRα do not express GFAP in the adult periventricular area (Cesetti et al., 2009; Komitova et al., 2009). However, these studies did not directly address whether a population of GFAP-expressing NSCs expresses PDGFRα. Together, these data suggest that the adult OPCs/PRPs, and not adult periventricular NSCs, are the precursors to the formation of cellular masses after PDGF infusion.
Other lines of evidence also support the hypothesis that OPCs/PRPs and not periventricular GFAP-expressing NSCs generate cellular masses in response to PDGF infusion. Jackson et al. (2006) found that PDGF promoted the generation of primary neurospheres from the adult periventricular area and that it also increased the generation of secondary neurospheres. This led them to conclude that PDGF acted directly on periventricular GFAP-expressing NSCs, to promote their self-renewal. We (Chojnacki and Weiss, 2004) and others (Bögler et al., 1990; McKinnon et al., 1990) have previously reported that FGF can promote the expansion of OPCs/PRPs. Dissections of the adult periventricular area ultimately contain at least two distinct precursor cells: GFAP-expressing NSCs, and OPCs/PRPs. It would be impossible to dissect the subventricular zone free of the OPCs/PRPs that are located throughout the parenchyma and immediately subjacent to the ependymal layer. Therefore, cultures of adult periventricular area in the presence of PDGF-AA and FGF-2 likely stimulate the formation of neurospheres by OPCs/PRPs and GFAP-expressing NSCs, resulting in an increased number of neurospheres generated as compared to FGF2 alone. Furthermore, primary neurospheres passaged into FGF2 and PDGF would also generate more neurospheres, since FGF2-responsive NSCs can also generate PDGFRα-expressing neural precursors (Menn et al., 2006) (A. Chojnacki and S. Weiss, unpublished observations). In sum, the results of our study and careful examination of those of Jackson et al. (2006) strongly suggest that only OPCs/PRPs proliferate in vivo in response to PDGF.
Our observations that OPCs/PRPs, but not periventricular GFAP-expressing NSCs, proliferate to generate cellular masses after infusion of PDGF into the brain parenchyma may have implications on the origins of brain tumors. Our work is consistent with that of Assanah et al. (2006), wherein retroviral overexpression of PDGFB, in rat brain white matter, led to the generation of tumors that closely resembled human glioblastomas. They also found that the majority of the cells that populated the tumor expressed OLIG2, NG2, and PDGFRα, all markers of OPCs/PRPs. However, <3% of the infected cells expressed GFAP. Most glioma cell lines secrete PDGF-A and/or PDGF B, at levels of up to 50 ng/ml (Nistér et al., 1986; Betsholtz et al., 1989; Assanah et al., 2006), which would be sufficient to drive the proliferation of not only the tumor cells but surrounding genotypically normal OPCs/PRPs in an autocrine/paracrine fashion. Additionally, PDGFRα is overexpressed by human glial tumors (Fleming et al., 1992; Guha et al., 1995). Together with these studies, our results suggest that OPCs/PRPs and not periventricular GFAP-expressing NSCs may be the cell of origin for a subset of tumors wherein an autocrine/paracrine PDGF signaling loop may initially drive their proliferation. We do not exclude the possibility that periventricular GFAP-expressing NSCs or their EGFR-expressing progeny could be cells of origin for other brain tumor subtypes. Nevertheless, our study strongly suggests that introducing mutations identified in secondary glioblastomas into mouse or human PRPs could provide a powerful model for identifying pathways required for their transformation into brain tumor-generating cells.
This work was supported by the Canadian Institutes of Health Research and studentship (G.M.) and scientist (S.W.) awards from the Alberta Heritage Foundation for Medical Research. We thank M. Gotz and D. van der Kooy for comments on earlier versions of this manuscript.
- Correspondence should be addressed to Samuel Weiss, 1A10 Health Research Innovation Centre, 3330 Hospital Drive NW, Calgary, Alberta T2N 4N1, Canada.