GFAPδ Expression in Glia of the Developmental and Adolescent Mouse Brain

Glial fibrillary acidic protein (GFAP) is the major intermediate filament (IF) protein in astrocytes. In the human brain, GFAP isoforms have unique expression patterns, which indicate that they play distinct functional roles. One isoform, GFAPδ, is expressed by proliferative radial glia in the developing human brain. In the adult human, GFAPδ is a marker for neural stem cells. However, it is unknown whether GFAPδ marks the same population of radial glia and astrocytes in the developing mouse brain as it does in the developing human brain. This study characterizes the expression pattern of GFAPδ throughout mouse embryogenesis and into adolescence. Gfapδ transcripts are expressed from E12, but immunohistochemistry shows GFAPδ staining only from E18. This finding suggests a translational uncoupling. GFAPδ expression increases from E18 to P5 and then decreases until its expression plateaus around P25. During development, GFAPδ is expressed by radial glia, as denoted by the co-expression of markers like vimentin and nestin. GFAPδ is also expressed in other astrocytic populations during development. A similar pattern is observed in the adolescent mouse, where GFAPδ marks both neural stem cells and mature astrocytes. Interestingly, the Gfapδ/Gfapα transcript ratio remains stable throughout development as well as in primary astrocyte and neurosphere cultures. These data suggest that all astroglia cells in the developing and adolescent mouse brain express GFAPδ, regardless of their neurogenic capabilities. GFAPδ may be an integral component of all mouse astrocytes, but it is not a specific neural stem cell marker in mice as it is in humans.


Introduction
Glial fibrillary acidic protein (GFAP) is a type III intermediate filament protein (IF; for review see [1,2]). IFs play important roles in cytomechanics and cell signaling [3][4][5]. GFAP is one of the IFs expressed by radial glia, adult astrocytes, and neural stem cells [6][7][8][9]. GFAP has several splice variants. The canonical isoform, GFAPa, contains nine exons and is the most abundantly expressed isoform in the human and mouse central nervous system [2,10]. Another isoform, GFAPd, differs from GFAPa in its unique carboxy-terminus, created by the replacement of exons 8 and 9 with exon 7+/7a [11,12]. This structure renders the assembly of GFAPd compromised, in that it is unable to form filaments by itself. For proper filament formation, another type III IF protein, such as GFAPa, is required [13,14]. The ratio of GFAPa and GFAPd has shown to be important factor in IF network formation [10,14]. Transfection of IF free cells with a GFAPa/GFAPd ratio of 3:1 already results in an aberrant condensed IF network. Ratios such as 1:1 and 1:3 result in improper filament formation [10]. This aberrant network formation may have functional consequences. GFAPd itself has already been shown to be involved with the c-secretase complex via its specific interaction with presenilin [15]. The c-secretase complex is a crucial mediator of Notch signaling and therefor important for stem cell biology. It is via this pathway that GFAPd is thought to be linked with neurogenesis.
Humans begin to express GFAPd at the same time pan-GFAP immunoreactivity is observed, around gestational week 13. This GFAPd expression is specifically found in radial glia, as denoted by co-expression of radial glial markers such as vimentin and nestin [16]. Radial glia are a type of precursor cell located in the ventricular zone (VZ) and in the medial pallium (MPall), the developing hippocampal formation [17][18][19]. Later in development, the VZ becomes the adult subventricular zone (SVZ) and the MPall transforms into the adult hippocampus. Radial glia are a heterogeneous population of cells that are able to self-renew and produce neurons as well as glia [20,21]. The production of neurons and glia is temporally dependent, with the peak of neurogenesis being around embryonic day 15 (E15) and the peak of radial glia-dependent gliogenesis around postnatal day 0 (P0) in rodents [20,22,23]. Interestingly, a second wave of gliogenesis takes place locally in the cortex during the first postnatal week of life [24]. As embryonic stages progress into postnatal ages, radial glia undergo direct transformation into astrocytes [25,26] but a small population of these astrocytes is thought to reside as neural stem cells in the adult brain.
GFAP expression in the developing mouse brain follows the basic progression of developing radial glia and astrocytes. Gfap transcripts can first be detected in the mouse brain between E9.5 and E11 [27], corresponding with the appearance of the first radial glia [28]. To note, there is a difference in GFAP expression timing between human and mouse. GFAP is first seen much earlier in mouse (corresponding to around 4.3-6.1 human gestational weeks) than in humans (13 gestational weeks [29]). GFAP is first expressed by radial glia around the telencephalic VZ, in the MPall, and fimbria (fi) [7,30]. Levels of Gfap mRNA at these early timepoints are very low, but as development progresses and gliogenesis commences, Gfap transcripts become abundantly expressed. After gliogenesis, Gfap levels plateau in the adult brain [30,31].
Neurogenesis continues in the adult brain, however it is much more restricted [9]. There are two major neurogenic niches of the adult brain, the SVZ located along the lateral wall of the lateral ventricle, and the subgranular zone (SGZ) in the hippocampus [32][33][34][35][36]. These areas are present in both humans and rodents [37,38]. Interestingly, GFAPd is highly expressed in the adult human SVZ [14]. It co-labels with stem cells markers such as sexdetermining region Y-box 2 (Sox2) and nestin, as well as the cell division markers minichromosome maintenance complex component 2 (MCM2) and proliferating nuclear antigen (PCNA). Primary adult human neurosphere cultures also express GFAPd along with nestin and the cell division marker Ki67 [39]. This population of GFAPd cells in the SVZ has been shown to be the quiescent neural stem cells of the adult human brain [39,40]. Notably, GFAPd is expressed in other human brain regions such as the olfactory bulb, rostral migratory stream, and glia limitans. In the adult mouse brain, immunostainings have shown that GFAPd is expressed in most astrocytes throughout the brain that express detectable levels of GFAPa, including astrocytes in the SVZ. Moreover, transcript levels of Gfapd in relation to the canonical isoform Gfapa (Gfapd/Gfapa transcript ratio) remain constant amongst neurogenic and non-neurogenic areas [10], suggesting that GFAPd is not a specific neural stem cell marker in the adult mouse brain.
This study takes a closer look at mouse GFAPd in relation to development and early adulthood. The expression of GFAPd in the developing mouse brain was examined using immunohistochemistry (IHC) and quantitative real time PCR (qPCR). Surprisingly though Gfapd transcripts were detectable from E12, GFAPd protein was only found from E18. For IHC experiments, special attention was paid to the developing SVZ and hippocampus from E12 to P10, as these areas house radial glia during development and maintain their neurogenic capacity throughout life. Unlike the situation in the human brain, GFAPd did not demarcate a specific population of cells. All cells that expressed GFAP throughout the brain (indicated by a pan-GFAP antibody), also expressed GFAPd. In vitro, GFAPd was present in both primary mouse astrocyte and neurosphere cultures. These data indicate that GFAPd may hold a different function in the mouse, as it is expressed in similar levels throughout all types of astrocytes in the developing and adolescent mouse brain.

Animals and Tissue Preservation
All experiments were carried out under the approval of the Animal Experimentation Committee of the Royal Netherlands Academy of Arts and Sciences (KNAW) with accordance to the European Community Council directive of November 24, 1986 (86/609/EEC). All efforts were made in order to minimize both the number and suffering of the animals involved in the current study.
For primary cultures, C57BL/6 pups between P0 and P3 were cooled on ice. They were decapitated and their heads were kept in cold DMEM (Life Technologies) until astrocyte or neural stem cell isolation.
For GFAP isoform transcript profiling during development, two nests of E12, E15, E18, and postnatal day 0 (P0) were used. Plug date was defined as E0. Mothers were sacrificed with an overdose of pentobarbital (0.40 ml/100 g) followed by cervical dislocation. Embryos were removed from the uterus and decapitated. Their brains were rapidly dissected and put directly into TRIsure (Bioline) or TRIzol (Invitrogen) for RNA isolation. P0 pups were first cooled on ice and then decapitated. Their brains were also quickly removed from the skull and homogenized in TRIsure. For immunohistochemistry (IHC), embryos were first cooled on ice and snap frozen in their entirety. For P0 and P5, pups were cooled on ice, decapitated, and then their heads were snap frozen. For later ages (P10 and P25), mice were given an intraperitoneal (i.p.) overdose of pentobarbital (0.15 ml/100 g), then perfused with saline followed by 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS; pH 7.4), followed by rapid brain dissection. After cryoprotection with 20% sucrose-PBS, their brains were snap frozen. All IHC tissue was stored at 280uC until further use. A cryostat (Leica CM3050) was used to cut 10 mm sagittal sections. These sections were then mounted on Superfrost Plus slides (Menzel-Glä ser), dried and stored at 220uC.

Primary Cell Isolation
For primary cell isolation, pups between P0 and P3 were used. After removal of the olfactory bulbs and cerebellum, the brain was chopped into small pieces and incubated with 2.5% trypsin (Invitrogen) for 5 min at 37uC. Deoxyribonuclease I from bovine pancreas (8 ml/ml, DNaseI; Sigma-Aldrich) was then added and cells were incubated at 37uC for an additional 5 min. A Pasteur pipette was used for the final dissociation step. DMEM with 10% Fetal Bovine Serum (FBS; Invitrogen) was added to the tube, cells were then spun at 1200 rpm for 10 min. The pellet was washed with 0.25% fungizone (Invitrogen) in DMEM. Cells were spun again, resuspended and then plated. For neural stem cell cultures, cells were plated in 6 well dishes (Greiner bio-one) with DMEM:F10 (Invitrogen) plus 1% penicillin streptomycin (penstrep; Invitrogen), 1% glutamine (Lonza), 1% N2 (Invitrogen), 20 ng/ml EGF (Peprotech/Tebu-bio), and 10 ng/ml bFGF (Peprotech/Tebu-bio). Growth factors were added twice a week. Astrocytes were plated in poly-L-lysine (PLL; Sigma-Aldrich) coated T75 flasks in DMEM:F10 plus 1% penstrep, 10% FBS, 0.25% fungizone. Cultures were purified approximately one week after isolation. Astrocyte cultures were taped onto a Unimax 2010 shaker (Heidolph, Schwabach, Germany) inside an incubator and shaken at 240 rpm for 2 hours. The supernatant was removed and cells were given fresh medium [41]. This procedure resulted in a mixed glia culture (predominantly microglia and astrocytes) where astrocytes represented around 80% of all cells based on immunohistological analysis of astrocyte markers such as GFAP and Vimentin.

RNA Isolation and cDNA Synthesis
After homogenization in TRIsure or Trizol, chloroform was added and samples were centrifuged. The aqueous phase was removed and mixed with an equal amount of isopropanol and 1 ml glycogen (Roche). RNA was allowed to precipitate for at least 2 days at 220uC. Samples were then centrifuged, the pellet washed twice with 70% ethanol, and air-dried. RNA pellets were resuspended with sterile MilliQ. RNA concentrations were determined using a Nanodrop (ND-1000; ThermoScientific, Wilmington, DE, USA). A fixed amount of RNA (250 ng) was used for cDNA synthesis carried out under manufacturer's protocol using the Quantitect Reverse Transcription Kit (Qiagen).
For immunostaining of primary astrocyte cultures, astrocytes were plated on poly-L-lysine (PLL) coated glass coverslips. Cells were fixed with 4% PFA in PBS (pH 7.4) for 10 minutes and washed twice with PBS. Cells were blocked with supermix (50 mM Tris, 154 mM NaCl, 0.25% Gelatin, 0.5% Triton-X-100 in H 2 O, pH 7.4; sumi) and incubated with primary antibodies diluted in sumi overnight at 4uC. The next day, cells were washed three times with PBS (pH 7.4) and incubated with secondary antibody diluted in supermix for 1 hour at room temperature.

The GFAPd Protein is Detected Later in Development than other GFAP Isoforms
To localize where and when GFAPd is first expressed, immunohistochemistry was performed on embryonic (E) and early postnatal (P) mice. In order to study the contribution of GFAPd expression to the total GFAP expression level, the GFAPd staining patterns were compared to that of the DAKO GFAP antibody (pan-GFAP). Particular focus was paid to the ventricular zone (VZ) and developing hippocampus (medial pallium; MPall), as these regions contain radial glia. GFAPd localization throughout various brain regions over all ages studied is summarized in Table 2.
Pan-GFAP immunoreactivity was first observed at E12, while GFAPd immunoreactivity was completely absent (Fig. 1A,B). As the GFAPd antibody is specific [10], this lack of GFAPd immunoreactivity at E12 indicates that there is little to no GFAPd protein expression. Pan-GFAP expression was weak and confined to parenchyma directly surrounding ventricular areas. At E15, the pan-GFAP immunoreactivity spread and was observed mostly in the VZ and the MPall. Yet still, GFAPd could not be detected (Fig. 1C,D). At E18, the pan-GFAP immunoreactivity grew, infiltrating more area of the VZ and MPall as well as the glia limitans, and GFAPd was observed for the first time ( Fig. 1E,F). The GFAPd staining was mainly observed in the VZ, MPall, and along the pial surface near the hindbrain (Fig. 1G). At P0, pan-GFAP immunoreactivity continued to demarcate a greater GFAPd expression was followed in major brain regions from E18 to P45. Strikingly, GFAPd immunoreactivity increases in some brain regions, such as the cerebellum, cortex, olfactory bulb and striatum, as the animal ages. While other brain regions, such as the fimbria, gradually downregulate GFAPd expression. Legend: 2not expressed; + lowly expressed; ++ moderately expressed; +++ highly expressed; n.d. not determined. For P10 to P45, the eyes and optic nerve were not preserved, as only the brains of these animals were processed for analysis. doi:10.1371/journal.pone.0052659.t002 population of cells, while the expression of GFAPd did not appear to shift in any way (Fig. 1G,H). Pan-GFAP immunoreactivity was observed to surround the lateral ventricle (Fig. 2i). Though not present at E15, GFAPd marks a population of bipolar cells in the VZ at E18, but more clearly at P0 ( Fig. 2A,B,C). These cells extend their processes towards the lateral ventricle and stretch towards the pia, indicative of radial glial cells. Pan-GFAP immunoreactivity displays the same pattern as GFAPd in the VZ, but from an earlier developmental stage onward (Fig. 2D,E,F). Radial glia cells are also present in the MPall. The GFAP expression pattern observed here mimics that seen in the VZ. Again, pan-GFAP immunoreactivity is detected at E12, and intensifies as development progresses. By E18, the majority of cells expressing GFAP are located in the fimbria   Fig. 2ii) or the dentate gyrus (DG; Fig. 2ii). Though not present at E12 (Fig. 2G), GFAPd is strongly expressed in the DG E18 and P0 (Fig. 2H,I). GFAPd marks a cluster of bushy cells with thick, stubby processes in the DG at E18 and P0. Pan-GFAP immunoreactivity marks relatively the same population of cells as GFAPd in the DG, but from an earlier developmental stage (Fig. 2J,K,L).

GFAPd is Expressed by Embryonic Progenitors
GFAPd and pan-GFAP mark specific populations of cells in the VZ, MPall, and along the pial surface during development. Cells that express both pan-GFAP and GFAPd in the VZ have a bipolar morphology and are hypothesized to be radial glial cells. In order to investigate whether GFAPd is indeed expressed in radial glia or in other cell types during development, a series of triple stainings was performed.
At E18, GFAPd always colocalized with vimentin, a radial glia marker. However, all cells that expressed vimentin did not necessarily co-express GFAPd (Fig. 3A, Sup. 1A). Those cells that expressed vimentin and GFAPd were bipolar in their morphology. Focusing around the LV, both GFAPd and vimentin could be found in the VZ and the SVZ, indicating that GFAPd is expressed by both radial glia and basal progenitors. As these cell types divide, GFAPd colocalization with a division marker was investigated. GFAPd and/or vimentin positive cells were rarely seen to colocalize with minichromosome maintenance complex component 2 (MCM2), a marker for the initiation of cell replication (Fig. 3A, Sup. 1A). In the E18 VZ and SVZ, GFAPd immunoreactivity was always seen to colocalize with GFAP Cterminus immunoreactivity (Fig. 3B, Sup. 1C). The GFAP Cterminus antibody marks GFAPa specifically in mouse. Notably, all GFAPd positive cells were also positive for the stem cell marker nestin. However, nestin expression is far more widespread at E18 in the VZ and SVZ than that of GFAPd (Fig. 3B, Sup. 1C).
The expression profile of the VZ at P0 mirrored that seen at E18. Again, all cells that expressed GFAPd also expressed vimentin, but not necessarily the other way around. It was also rare to see a colocalization of MCM2 and GFAPd (Fig. 3C, Sup. 1E). GFAPa immunoreactivity (as marked by the GFAP Cterminus antibody) is always seen to colocalize with GFAPd immunoreactivity. Notably, nestin immunoreactivity has changed from P0 and now completely overlaps with that of GFAPd (Fig. 3D, Sup. 1G). P0 marks the beginning of radial glia's transformation into astrocytes as well as the start of the peak of astrogenesis. Here, cells that express GFAPd are losing their bipolar phenotype while increasing the thickness and number of their processes.
In the developing E18 hippocampus, GFAPd, like in the VZ, always colocalizes with vimentin. However here in the hippocampus, vimentin marks a much broader population of cells than GFAPd. The amount of cells expressing both GFAPd and MCM2 is also very low (Fig. 4A, Sup. 1B). In the DG, MCM2 marks a small population of actively dividing cells. Few of these MCM2 positive cells are clearly GFAPd positive (Fig. 4A, Sup. 1B). As expected, the expression of GFAPd and GFAPa entirely overlaps. GFAPd and GFAPa mark a very specific population of cells only located within the DG and fimbria (Fig. 4B). Nestin also completely colocalizes with GFAPd in both the DG and the fimbria (Fig. 4B, Sup. 1D). At P0, the same basic expression pattern seen at E18 continues. In the DG and fimbria, most vimentin positive cells also express GFAPd. Whereas in the surrounding hippocampal formation, there is no GFAPd expression and, consequently, all vimentin positive cells are GFAPd negative (Fig. 4C, Sup. 1F). Nestin and GFAPa completely colocalize with GFAPd in the P0 hippocampus (Fig. 4D, Sup. 1H).

Gfapd mRNA is Detectable from E12
In order to obtain more detailed quantitative information on Gfap isoform expression level during development, real time quantitative PCR (qPCR) was performed. Whole brains of E12 through P0 mice were subjected to RNA isolation and qPCR analysis. The samples were first investigated for pan-Gfap expression. Transcript levels of most Gfap isoforms, as discerned with pan-Gfap primers, were detectable from E12 and progressively increased during development (Fig. 5A). There was a 150fold increase of pan-Gfap expression from E12 to P0. Subsequently, the transcription levels of Gfapd and the canonical isoform Gfapa were then determined.
All samples expressed Gfapa throughout every stage studied. Moreover, as development progressed, the transcript levels of Gfapa increased. Although GFAPd protein was unable to be detected at E12 and E15, the Gfapd transcript was detected. At E12, some (3 out of 5) samples had detectable Gfapd transcript levels, albeit at low level. At E15, all samples displayed detectable levels of Gfapd. The ratio between Gfapd and Gfapa expression was investigated from E15 to P0, as all samples in these stages expressed both Gfapd and Gfapa. The Gfapd/Gfapa ratio did not significantly change at any point in development (Fig. 5B). This finding clearly shows that the total brain Gfapd/Gfapa transcript ratio remains stable throughout both neurogenesis and gliogenesis.

GFAPd is Expressed by All Cells that Express pan-GFAP in the Adolescent Mouse Brain
In order to track GFAPd positive cells through later development, brains from early postnatal ages into adolescence were profiled using a series of antibodies. Pan-GFAP immunoreactivity is seen throughout the entire adolescent mouse brain. Though most notable in the astrocytes of the DG and SVZ, GFAP expression is also observed in astrocytes of the cortex. Surprisingly, all astrocytes that were marked by the pan-GFAP antibody, also expressed GFAPd. The expression patterns of GFAPd and pan-GFAP immunoreactivity were highly similar, but differed among brain regions. Both GFAPd and pan-GFAP displayed strong immunoreactivity in subcortical areas. GFAPd displayed a weaker immunoreactivity in regions such as the cortex while pan-GFAP immunoreactivity was more readily detectable.
At P5, many cells in the DG show pan-GFAP immunoreactivity (Fig. 6A,B,C). These cells have a relatively bushy morphology and are clustered closely together. As the animal matures, these pan-GFAP positive cells seem to extend their processes in a more organized fashion, resulting in a striking bipolar morphology around P10 (Fig. 6D,E,F) and long-range fibers at P25 (Fig. 6G,H,I). This shift in morphology seems to coincide with a decrease in pan-GFAP immunoreactivity. All cells that express pan-GFAP immunoreactivity also express GFAPd. The staining pattern of GFAPd completely overlaps with pan-GFAP immunoreactivity. In fact, the DG displays ubiquitous GFAPd expression in the same cell compartments as other GFAP isoforms.
(asterisks, C). At P0, GFAPd expression is broader in the VZ and SVZ than observed at E18. At P0, GFAPd now encompasses all the GFAPa and nestin immunoreactivity (arrow, D). Abbreviation: LV: lateral ventricle. Scale bars = 20 mm. doi:10.1371/journal.pone.0052659.g003  (dashed line, B). This immunohistological pattern seen at E18 also carries through to the P0 hippocampus. Vimentin expression is more widespread than GFAPd expression (arrow head, C), though there are double-positive populations as well Around P5 in the SVZ, the cells that show pan-GFAP immunoreactivity, like in the DG, are heavily clustered together. These cells also have a bipolar morphology (Fig. 7A,B,C), as already clearly seen by E18. From P10 to P25, these processes begin to grow and infiltrate the surrounding parenchyma. From P10 (Fig. 7D,E,F), but most notable at P25, the processes extend and thin out, as denoted by the punctate staining patterns in the SVZ and surrounding parenchyma (Fig. 7G,H,I).
In the P10 SVZ, vimentin expression marks a population of ependymal cells lining the lateral ventricle as well as a restricted population of astrocytes in the SVZ parenchyma. GFAPd immunoreactivity and vimentin expression can overlap but there are also exclusive vimentin positive and GFAPd positive cell populations present. Just as in development, the colocalization between MCM2 and GFAPd is rare (Fig. 8A,B). As colocalization of GFAPd and MCM2 was quite rare, other proliferation markers were also profiled. However even using a broader marker of proliferation, such as phosphohistone-H3 (pHH3), still resulted in limited colocalization with GFAPd (Fig. 8C). Notably, both MCM2 and pHH3 positive cells were located within classical neurogenic niches like the SVZ and SGZ. Nestin, GFAPa, and GFAPd are found to reliably colocalize within the same cells (Fig. 8D). In the P10 DG, vimentin marks a greater population of cells than GFAPd. That said, all GFAPd positive cells are also vimentin positive. GFAPd immunoreactivity hardly overlaps with that of MCM2 (Fig. 8E,F). As in the P10 SVZ, GFAPa, GFAPd, and nestin expression completely overlap (Fig. 8G,H).

GFAPd Displays a Similar Profile in vitro as it does in vivo
Primary astrocyte and primary neurosphere cultures from P0-P3 mice were made to investigate mouse GFAPd more closely in vitro. Firstly, cultures were assessed by qPCR to determine the Gfapd/Gfapa transcript ratio. Both neurosphere and primary astrocyte cultures expressed Gfapd and Gfapa transcripts. Moreover, the Gfapd/Gfapa transcript ratio did not significantly differ between neurosphere and astrocyte cultures (neurospheres: 9.37860.4725; astrocytes: 7.61561.715; unpaired t test p = 0.2498). Interestingly, the Gfapd/Gfapa ratio shifts in vitro (7.6) and in vivo (2.5). Both primary neurosphere cultures (Fig. 9A) and primary astrocytes (Fig. 9B,C,D) expressed GFAPd. Primary astrocyte cultures showed a complex phenotype, where GFAPa was commonly observed alone in the tips of the IF network (Fig. 9B,C). GFAPd expression was usually localized around the nucleus and in the soma (Fig. 9B,C,D). Performing a dye-swap experiment, where the secondary fluorphores were switched, resulted in the same finding (Fig. 9D). However, the GFAPd distribution throughout the IF network was variable, and could even be observed in the edges of the IF network (Fig. 9D).

Discussion
From the investigation of GFAPd during mouse development, it appears that if GFAPd is detectable, it is not limited to neurogenic cells. GFAPd immunoreactivity is first seen at E18. Here, GFAPd is mainly localized within radial glia of the VZ and MPall. As the animal matures, GFAPd positive cells begin to lose their bipolar morphology and shift towards a more star-like, mature phenotype. This shift is most evident from P5 to P10 in both the SVZ and hippocampus. At P25, GFAPd positive cells show complex branching and extended processes, indicative of mature astrocytes. GFAPd positive cells are also present in the classical neurogenic niches, reflecting the different types of SVZ astrocytes, including quiescent neurogenic astrocytes.
Throughout all timepoints studied, the expression of vimentin and nestin are mostly more widespread than that of GFAPd. In early stages of development (E18 and P0) the colocalization of GFAPd with vimentin and nestin indicates that GFAPd is expressed by both radial glia and undifferentiated precursors [44][45][46]. Interestingly, GFAPd is not homogenously present within these niches, which may indicate that GFAPd marks a subpopulation at E18. More cells acquire detectable GFAPd expression at (arrow, C) and MCM2 hardly colocalizes with GFAPd (asterisk, C). However MCM2 expression is only seen in those distinct areas with GFAPd immunoreactivity (dashed line, C). GFAPd, GFAPa, and nestin mostly overlap in the P0 hippocampus (arrow, D), but nestin expression is far more widespread (arrow head, D). Again, GFAPd immunoreactivity remains within the DG (dashed line, D). Abbreviations: DG: dentate gyrus, fi: fimbria. Scale bars = 20 mm. doi:10.1371/journal.pone.0052659.g004 Figure 5. Gfap transcript expression in the developing mouse brain. Though the transcript levels of pan-Gfap increase throughout development (p = 0.001, Oneway ANOVA; A), the ratio between Gfapd and Gfapa remains unchanged (B). All data is normalized to Gapdh and B-actin. Normalization procedures are described extensively in [43]. Gfapa and Gfapd are detected with equal efficiencies. The same Rn threshold was used in all qPCRs. Ratios are calculated as Gfapd/Gfapa x 100. Data is displayed as mean 6 s.e.m. doi:10.1371/journal.pone.0052659.g005 P0. Taken together, these data could indicate that a subpopulation of radial glia cells is maturing, thereby amassing GFAP expression [47][48][49]. This process commences around E18, coinciding with the start of astrogenesis, and advances through P0 -where more colocalization between GFAPd and vimentin can be seen, due to the greater expression of GFAPd at this timepoint. In addition to radial glia, GFAPd is also observed around the developing anterior commissure. These astrocytes excrete growth factors and forming physical barriers, allowing for the proper formation of commissures and, in turn, proper development of neural circuitry [50,51]. Strong GFAPd expression is also observed from E18 demarcating the supragranular and fimbrial bundles. These structures are crucial for hippocampal morphogenesis. Without the supragranular bundle, the DG is unable to fully form [52]. The presence of GFAPd in both radial glia and other astrocyte populations suggest that GFAPd may play a critical role in general astrocyte biology and is an integral part of the GFAP intermediate filament cytoskeleton.
The maturation of GFAPd positive cells coincides with differential marker expression. From P10, GFAPd marker coexpression shifts towards a more mature phenotype. Here, GFAPd positive cells seem to have lost most of their vimentin expression. This observation fits well with the GFAPd/vimentin pattern seen in development. Moreover, the reduced colocalization between vimentin and GFAPd in the SVZ indicates that GFAPd is not expressed in ependymal cells; as vimentin is commonly reported as an ependymal marker in the mature SVZ [45,48,53]. Cells expressing GFAPd can divide, as marked by MCM2 and pHH3 co-expression. However, they do so rarely and only within the SVZ and SGZ. At P10, nestin expression is more restricted than at E18 and P0, and completely overlaps with GFAPd immunoreactivity. This refined nestin expression along with its co-expression with GFAPd indicates that neurogenic astrocytes in the SVZ and SGZ express GFAPd [54,55].
Contrary to vimentin and nestin expression, GFAPa and GFAPd always colocalize from E18 onwards. This finding is unsurprising as GFAPd, unlike the canonical isoform GFAPa, is unable to form a functional IF network by itself [13,14]. This observation is in accordance to what has been seen in the adult mouse brain [10], where GFAPd and other GFAP isoforms such as GFAPa mostly colocalize within the same cell, whether that cell has a neurogenic phenotype or not. Notably, pan-GFAP immunoreactivity is observed 6 days prior to the appearance of GFAPd. Pan-GFAP immunoreactivity was observed in radial glia cells, analogous to previous reports [20]. The discrepancy between the appearance of pan-GFAP and GFAPd immunoreactivity could be attributed to the lower abundance of GFAPd transcript level (7.9% of Gfapa in the adult mouse brain [10]) or to a delay in GFAPd translation.
Even though GFAPd was first seen at E18 using immunohistochemistry, Gfapd transcripts were first seen much earlier, around E12 to E15. Not all samples express Gfapd at E12 (3 out of 5 samples) and those that do, express very low levels of Gfapd. The discrepancy between Gfap transcript and GFAP protein detection has been described before in vivo and in vitro [56][57][58][59][60][61]. Most strikingly, when Zhou and colleagues (2000) studied the stratum radiatum of the CA1, a glial dense region in the hippocampus, they found that there was much more expression of the Gfap transcript (74% of all cells) than of the GFAP protein (1.5% of all cells). Furthermore, there are reports that GFAP protein expression is preceded by Gfap transcript expression [57,58]. Although this phenomenon could be attributed to the lack of sensitivity of a given GFAP antibody, it is more likely due to a translational mechanism [60]. Interestingly, an altered translational mechanism retarding GFAP protein production also implies a distinctive role for Gfap mRNA, itself.
If the Gfapd transcript would have a specific neurogenic or astrogenic role, the Gfapd/Gfapa ratio would be expected to shift throughout different developmental stages. However, there is no significant change in the Gfapd/Gfapa ratio from E15 to P0. The Gfapd/Gfapa transcript ratio is higher in vitro than in vivo. This finding may be due to the stress of an in vitro environment to the  cell, causing the cell to change the fundamental composition of its IF network. That said however, the Gfapd/Gfapa transcript ratio also does not differ between primary astrocytes and primary neural stem cell cultures. These data are in accordance with previous findings from our lab, which show that the Gfapd/Gfapa transcript ratio remains stable throughout adult neurogenic and non-neurogenic mouse brain regions, even though the amount of transcript expression is variable [10].
Taken together, these results highlight the divergence between human and mouse GFAPd. In the adult human SVZ, there is a higher Gfapd/Gfapa transcript ratio when compared to a nonneurogenic region [14]. Unlike the developing mouse brain, human GFAPd (hGFAPd) simultaneously emerges with other GFAP isoforms at 13 weeks of gestation in the developing human VZ. GFAPd is present in radial glia in both the developing human and mouse. However, hGFAPd stays confined to the VZ and the Figure 9. In vitro expression of GFAPd. Primary neurospheres (A) and astrocytes (B-D) express both GFAPd and GFAPa. GFAPa, not GFAPd, can be clearly seen in the outer edges of the IF network (arrow, B-C). GFAPd is commonly found to surround the nucleus and fill the soma (arrow heads, C-D), and can, in some instances, be seen in the outer edges of the IF network (arrow, D). To determine whether a dye-swap had an effect on localization detection, a different secondary antibody conjugated to a different fluorphore was used. This dye-swap rendered the same results (D, image is recolored for presentation purposes). B-D are recorded using identical settings. GFAPd is shown in red. GFAPa, as marked by an antibody against the GFAP C-terminus, is shown in green. DAPI is shown in blue. Scale bars = 20 mm. doi:10.1371/journal.pone.0052659.g009 SVZ during development. To that end, pan-GFAP immunoreactivity is far more widespread than that of hGFAPd, which is in stark contrast to the situation in mouse where pan-GFAP immunoreactivity always coincides with that of GFAPd. Marker co-expression also differs between species. Human vimentin, nestin, and hGFAPd always colocalize in development [16], whereas the vimentin and nestin patterns in the embryonic mouse are far more widespread. This finding also supports previous evidence that primate radial glia undergo the transition from vimentin to GFAP expression earlier than rodent counterparts [62]. Most hGFAPd positive cells co-express a proliferation marker [16]. Proliferation markers were extensively investigated in this study, however the presence of a dividing cell expressing GFAPd was a rarity. GFAPd seems to demarcate a more specialized celltype in human than mouse. hGFAPd positive cells show a more defined, homogenous phenotype during development.
The current study set out to explore the involvement of GFAPd during mouse developmental neurogenesis. The results of this study indicate that mouse GFAPd and hGFAPd have differential expression patterns. Moreover, mouse GFAPd is only detectable well after the peak of embryonic neurogenesis, around the commencement of astrogenesis. During adolescence, GFAPd does not demarcate a specific population of neural stem cells in the mouse brain as it does in the human brain. In vitro, GFAPd is present in both multipotent, self-renewing neural stem cells as well as astrocytes. The Gfapd/Gfapa transcript ratio neither shifts during development nor in primary astrocyte and neural stem cell cultures. During development, GFAPd is expressed by a population of radial glia cells, which seem to be acquiring a more mature phenotype. However, it is also expressed by astrocytes involved in commissure formation. In the adolescent hippocampal formation, GFAPd is expressed in the neurogenic SGZ but also in the nonneurogenic fimbria. Taken together these observations indicate that though GFAPd is not a specific neural stem cell marker in the developing mouse brain, it may be an integral part of the intermediate filament network of all developing and mature astrocytes. Figure S1 Confocal analysis of GFAPd colocolization in the developing VZ and MPall. In the E18 VZ and DG, GFAPd always colocalizes with vimentin and rarely with MCM2. Here, vimentin expression is far more widespread than that of GFAPd (A-B). At this developmental stage, GFAPd always colocalizes with GFAPa and nestin. However like vimentin expression, nestin marks a broader range of cells than GFAPd (C-D). At P0, the separation of vimentin and GFAPd expression becomes more evident, where vimentin is commonly seen in GFAPd ependymal cells with the VZ (E). Again, colocalization between GFAPd and MCM2 is rare in both the VZ and the DG (E-F). Nestin expression has also transitioned at P0. Now, all cells that express nestin also express both GFAPd and GFAPa within the VZ (G). However, nestin still marks a larger population of cells in the hippocampus than GFAPd (H). Abbreviations: LV: lateral ventricle, DG: dentate gyrus. Scale bars = 20 mm. (TIF)