Nuclear factor-I regulates glial fibrillary acidic protein gene expression in astrocytes differentiated from cortical precursor cells

Authors

  • Beatriz Cebolla,

    1. Instituto de Investigaciones Biomédicas Alberto Sols, Consejo Superior de Investigaciones Científicas/Universidad Autónoma de Madrid, Madrid, Spain
    Search for more papers by this author
  • Mario Vallejo

    1. Instituto de Investigaciones Biomédicas Alberto Sols, Consejo Superior de Investigaciones Científicas/Universidad Autónoma de Madrid, Madrid, Spain
    Search for more papers by this author

Address correspondence and reprint requests to Mario Vallejo MD PhD, Instituto de Investigaciones Biomédicas ‘Alberto Sols’, Calle Arturo Duperier 4, 28029 Madrid, Spain. E-mail: mvallejo@iib.uam.es

Abstract

The elucidation of the transcriptional mechanisms that regulate glial fibrillary acidic protein (GFAP) gene expression is important for the understanding of the molecular mechanisms that control astrocyte differentiation during brain development. We investigated regulatory elements located in a proximal region of the GFAP promoter, important for expression in cortical precursor cells differentiating into astrocytes. One of these elements recognizes transcription factors of the nuclear factor-I family (NFI). We found that, in primary cultures of cortical cells, NFI occupies the GFAP promoter prior to the induction of astrocyte differentiation. In the developing cerebral cortex, the onset of expression of NFI coincides chronologically with the beginning of astrocytogenesis. Mutational analysis of the GFAP gene and transfections in primary cortical precursors show that inhibition of binding of NFI to the GFAP promoter results in decreased levels of transcriptional activity and is required for the synergistic stimulation of the GFAP promoter by the astrogenic agents, pituitary adenylate cyclase-activating polypeptide and ciliary neurotrophic factor, which in combination enhance astrocyte differentiation to generate astrocytes with longer processes. Thus, NFI appears to be an important factor for the integration of astrogenic stimuli in the developing central nervous system.

Abbreviations used
bFGF

basic fibroblast growth factor

ChIP

chromatin immunoprecipitation

CNTF

ciliary neurotrophic factor

E17

embryonic day 17

EMSA

electrophoretic mobility shift assay

GFAP

glial fibrillary acidic protein

MutA/B

mutation A/B

NFI

nuclear factor-I

PACAP

pituitary adenylate cyclase-activating polypeptide

TGF-β1

transforming growth factor β1

WT

wild type

Neural progenitor cells that proliferate in the embryonic central nervous system respond to specific extracellular signals that determine their differentiation into neurons or glial cells (Gross et al. 1996; Johe et al. 1996; Bonni et al. 1997; Park et al. 1999). In the case of astrocytes, expression of glial fibrillary acidic protein (GFAP) constitutes a cell-specific phenotypic hallmark of their identity. The restrictive expression of GFAP in astrocytes indicates the requirement of specific combinations of transcription factors assembled on the GFAP gene promoter that occur only in this type of cells. Therefore, the elucidation of the transcriptional mechanisms that stimulate GFAP gene expression in a cell-specific manner is important for the understanding of how phenotypic transitions take place during the differentiation process that generates astrocytes in the developing nervous system.

At least three major signalling pathways that promote the generation of astrocytes in the central nervous system have been identified. One relies on the phosphorylation of intracellular JAK and STAT proteins to activate GFAP gene expression in response to stimulation by ciliary neurotrophic factor (CNTF) (Johe et al. 1996; Bonni et al. 1997; Koblar et al. 1998; Rajan and McKay 1998). Another one is activated by some bone morphogenetic proteins (Gross et al. 1996; Nakashima et al. 2001), that act via intracellular activation of Smad proteins (Ebendal et al. 1998; Zhang et al. 1998). A third mechanism relies on the synthesis of cAMP elicited in fetal cerebrocortical precursor cells by pituitary adenylate cyclase-activating polypeptide (PACAP), thus triggering GFAP gene expression and promoting astrocyte differentiation (McManus et al. 1999; Vallejo and Vallejo 2002).

Stable expression of GFAP in astrocytes requires permanent chromatin remodelling to allow the formation of cell-specific combinations of transcription factors assembled on the GFAP promoter (Song and Ghosh 2004). As the onset of astrocyte differentiation is accompanied by the stimulation of GFAP gene expression, a number of studies have been carried out to investigate how transcriptional mechanisms that induce expression of this gene, silent in neurons and in undifferentiated neural precursors, are activated. These studies have led to the identification of several DNA cis-regulatory elements present in the promoter regions of the human, mouse and rat GFAP genes. Transcription factors that activate GFAP gene expression by acting on these sites include AP-1 (Masood et al. 1993; Barnett et al. 1995), STAT and Smad proteins (Bonni et al. 1997; Kahn et al. 1997; Rajan and McKay 1998; Nakashima et al. 1999; Sun et al. 2001; Rajan et al. 2003), steroid hormone receptors (Stone et al. 1998), and proteins activated by transforming growth factor (TGF)-β1 (Krohn et al. 1999). In addition, activation of GFAP gene expression requires cell-specific CpG site demethylation (Takizawa et al. 2001).

Despite these advances, a detailed knowledge of the molecular mechanisms that regulate the stable expression of GFAP in astrocytes is still missing. In the present study, we demonstrate that binding of nuclear factor-I (NFI) to a specific site in the rat GFAP gene promoter is required for optimal GFAP gene expression in cortical precursor cells induced to differentiate into astrocytes by exposure to PACAP and CNTF.

Materials and methods

Reagents

Radioactive compounds were obtained from Amersham Biosciences (Little Chalfont, UK). Nucleotides were purchased from Promega (Madison, WI, USA). Tissue culture medium and reagents were obtained from Life Technologies (Rockville, MD, USA). Basic fibroblast growth factor (bFGF) and CNTF were from PeproTech EC Ltd. (London, UK). PACAP-38 and all other reagents were obtained from Sigma (Madrid, Spain) unless otherwise specified.

Plasmids

Luciferase reporter plasmids bearing 5′-flanking sequences of the rat GFAP gene promoter (Bonni et al. 1997; Krohn et al. 1999) were kindly provided by Dr Irina Rozovsky (University of Southern California, Los Angeles). The longest fragment used (GFAP-A7Luc) corresponds to nucleotides −1546 to +13, relative to the transcription initiation site (Condorelli et al. 1994), and the rest of them (GFAP-A6Luc to GFAP-A2Luc) correspond to smaller fragments generated by sequential 5′ deletions (Fig. 1). An additional plasmid, GFAP-A9Luc was constructed. For this purpose, a synthetic double-stranded oligonucleotide with KpnI and BglII sites at the 5′ and 3′ ends, respectively, corresponding to a GFAP minimal promoter (nucleotides −35 to +13), was inserted into the plasmid pGL2-Basic (Promega).

Figure 1.

 Deletional analysis of the promoter region of the rat GFAP gene. (a) Schematic depiction of the 5′ deletion constructs of the rat GFAP-luciferase fusion gene used in the transfection studies. (b) Relative luciferase activities elicited after transfections of RC2.E10 cells with luciferase reporter plasmids bearing 5′ deleted fragments of the GFAP promoter as indicated above. Values are expressed as percentages of the activities elicited by GFAP-A7Luc, and represent the mean ± SEM of at least three independent experiments carried out in duplicate.

The plasmids GFAP-A7MutBLuc and GFAP-A7MutCLuc were generated by oligonucleotide directed mutagenesis using GFAP-A7Luc as a template. For this purpose, we used Pfu Turbo DNA polymerase from a QuikChange (Stratagene, La Jolla, CA, USA) site-directed mutagenesis kit following the instructions provided by the manufacturer. Template DNA was eliminated by digestion with DpnI. GFAP-A7MutBLuc incorporates a 4-bp mutation between nucleotides −71 and −68, and GFAP-A7MutCLuc incorporates a 4-bp mutation between nucleotides −65 and −62. The sequences of the oligonucleotide primers containing the four base mismatches are as follows: mutation B (MutB) 5′-TGGGGTGCCATTAGGAAGTCAGGG-3′ and mutation C (MutC) 5′-GGGGTGCTGCCAGAGGATCAGGGGCAGA-3′.

Cell culture and transfections

RC2.E10 cells derived from rat E16 fetal cortex express GFAP when exposed to 8Br-cAMP (Schwartz and Vallejo 1998; McManus et al. 1999). These cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum at a temperature of 33°C and were transfected with Lipofectin (Life Technologies) as described previously (Schwartz and Vallejo 1998). After transfection, cells were cultured in the presence of 1 mm 8Br-cAMP.

Primary cortical cell cultures from the cerebral cortex from fetal brains of embryonic day 17 (E17) Wistar rats were prepared as described (Vallejo and Vallejo 2002). Briefly, cell suspensions prepared by trituration of the dissected fetal cortex were washed and re-suspended in serum-free Dulbecco's modified Eagle's medium containing N1 supplement (Sigma) and 1 mm sodium pyruvate (defined medium; Bottenstein and Sato 1979), to which bFGF (20 ng/mL) was added. Cells were seeded into poly ornithine-coated 10-cm dishes at a density of 2–4 × 104 cells/cm2, and were expanded at 37°C for 3–4 days. Medium was replaced every 2 days.

For differentiation experiments, cells were plated into 35-mm dishes at a density of 4 × 104 cells/cm2. After incubation at 37°C for 24 h, bFGF-containing medium was replaced with bFGF-free defined medium, and PACAP (1–100 nm) and/or CNTF (1–50 ng/mL) were immediately added. After the addition of these compounds, cells were incubated for 2 days, at the end of which they were processed for immunocytochemistry.

For transfections, 106 cells/plate were seeded into poly ornithine-coated 60-mm dishes in defined medium in the presence of bFGF (20 ng/mL). After an overnight incubation, the medium was removed and 6 μg of reporter plasmid DNA mixed with 10 μL FuGENE 6 transfection reagent (Roche Molecular Biochemicals, Indianapolis, IN, USA) were added in 4 mL defined medium and incubated for 4 h. After this, the medium-DNA mix was removed and substituted with fresh defined medium, or with defined medium containing PACAP or CNTF at the indicated concentrations.

Luciferase activity was measured using a commercial assay system (Promega) 48 h after transfection. Luciferase activity elicited after transfection of RSV-Luc was used as an independent standard for normalization, and efficiencies were corrected by using the Renilla luciferase assay system (Promega). All the values are expressed as mean ± SEM of at least three independent experiments carried out in duplicate.

Electrophoretic mobility shift assays (EMSA)

EMSA were carried out with nuclear extracts (Schreiber et al. 1989) in the presence of the protease inhibitors pepstatin A (1 mg/mL), leupeptin (10 mg/mL), aprotinin (10 mg/mL) and p-aminobenzamidine (0.1 mm). Protein concentrations were determined by the Bio-Rad protein assay (Bio-Rad Laboratories, Hercules, CA, USA) with bovine serum albumin as a standard. Synthetic complementary oligonucleotides with 5′-GATC overhangs were annealed and labelled by a fill-in reaction using α-32P–dATP and Klenow enzyme. Binding reactions were carried out in the presence of 2 μg of poly(dI.dC) and competitor oligonucleotides, as indicated, using nuclear extracts (10 μg of protein) incubated with 20 000 cpm of radiolabelled probe (approximately 6–10 fmol) in a total volume of 20 μL containing 20 mm potassium phosphate (pH 7.9), 70 mm KCl, 1 mm DTT, 0.3 mm EDTA, and 10% glycerol. The sequences of the oligonucleotides used were (sense strand): 85/−61 wild type (WT): 5′-GATCCATTCAATGGGGTGCTGCCAGGAAGTA-3′; −85/−61MutA: 5′-GATCCATCTGGTGGGGTGCTGCCAGGAAGTA-3′; −85/−61MutB: 5′-GATCCATTCAATGGGGTGCCATTAGGAAGTA-3′; −85/−61MutC: 5′-GATCCATTCAATGGGGTGCTGCCAGAGGATA-3′.

Western blot

Nuclear extracts (Schreiber et al. 1989) from cerebral cortex of developing rats or whole-cell extracts from cultured primary cells were prepared and proteins (20 μg) were resolved by sodium dodecyl sulfate – polyacrylamide gel electrophoresis and blotted onto a nitrocellulose membrane. NFI or CREB immunoreactivities were detected with rabbit polyclonal primary antisera (Santa Cruz Biotechnology, Santa Cruz, CA, USA) used at 1 : 5000 or 1 : 2000 dilution, respectively, followed by incubation with a goat anti-rabbit peroxidase-conjugated secondary antibody (1 : 10 000 dilution; Bio-Rad Laboratories). GFAP or β-actin immunoreactivities were detected with specific monoclonal antibodies (clone G-A-5, 1 : 10 000 dilution; and clone AC-15, 1 : 5000 dilution, respectively; Sigma) followed by incubation with a horse anti-mouse peroxidase-conjugated secondary antibody (1 : 5000 dilution) (Bio-Rad Laboratories). Immunoreactive bands were visualized using an enhanced chemiluminescence detection system (Amersham Biosciences).

Chromatin immunoprecipitation (ChIP) assays

ChIP assays were carried out basically as descried by Gerrish et al. (2001). Subconfluent cultures of E17 cortical cells were treated with 1% formaldehyde for 10 min at 22°C and the cross-linked protein–DNA complexes were isolated. Chromatin was sonicated in a volume of 200 μL, and diluted in buffer containing 1% Triton X-100, 2 mm EDTA, 20 mm Tris-HCl (pH 8.0), 150 mm NaCl and protease inhibitors. Immunoprecipitations were carried out using an anti-NFI antiserum (N-20; Santa Cruz Biotechnology), or control normal rabbit IgG (sc-2025, Santa Cruz Biotechnology), and antibody–protein–DNA complexes were isolated by incubation with protein A–Sepharose. After extensive washing, the DNA was eluted and detected by PCR using oligonucleotide primers that amplify a fragment of the GFAP gene spanning nucleotides −220 to +51. PCR conditions were: 95°C for 5 min, followed by 30 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 30 s, after which a 5-min incubation at 72°C followed. The sequence of the oligonucleotide primers are: forward 5′-CCCTCTCCTGACCCATTTACCAGAA-3′ and reverse 5′-GCCCCTGACCATCGTCTCGGAGGAG-3′. As a negative control, PCR was also carried out using oligonucleotide primers that amplify a fragment of the somatostatin gene spanning nucleotides −550 to −120. PCR conditions were: 95°C for 5 min, followed by 30 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 30 s, after which a 5-min incubation at 72°C followed. The sequence of the oligonucleotide primers are: forward 5′-GATTGGACAAAGTGATGCTC-3′ and reverse 5′-AGTGAGGGGAGGCGACAC-3′. As an additional control, GFAP ChIP assays were carried out in non-GFAP-expressing Rat-1 fibroblasts. PCR products were run on a 1% agarose gel, stained with ethidium bromide and photographed.

Immunocytochemistry and digital image analysis

GFAP immunocytochemistry was carried out with cells plated into poly ornithine-coated 35-mm tissue culture dishes, using a specific monoclonal antibody (1 : 300 dilution; clone G-A-5; Sigma) exactly as described (Vallejo and Vallejo 2002). For the detection of NFI immunoreactivity, a specific anti-NFI antiserum (Santa Cruz Biotechnology) was used at a 1 : 1000 dilution. In control experiments, the antiserum was pre-incubated for 1 h with a specific NFI blocking peptide (Santa Cruz Biotechnology) used in a 5-fold molar excess. Immunodetection was carried out using immunoperoxidase staining with a Vectastain ABC kit (Vector Laboratories, Burlingame, CA, USA).

For the determination of the length of astrocyte processes, GFAP-immunostained cells were visualized on a Nikon microscope equipped with an Olympus DP50 digital camera using a 20 × objective. Measurement of the total length of the processes in each GFAP-positive cell was carried out on digital images generated at the image analysis core facility in our institute using calibrated AnalySIS imaging software (Soft Imaging Systems Gmbh, Münster, Germany). Three or four randomly selected images per plate, taken from three independent experiments carried out in duplicate, were analysed for each condition. The total number of cells analysed range between approximately 500 and close to 700, as indicated at the lower section of Fig. 8(g).

Figure 8.

 (a–f) Immunocytochemical staining of GFAP in astrocytes generated from primary precursor cells prepared from the cerebral cortex of E17 rat fetuses. To induce astrocyte differentiation, cells were treated for 2 days with PACAP (10 nm), CNTF (10 or 30 ng/mL), or both agents simultaneously used at the indicated concentrations. Cells exposed to PACAP and CNTF simultaneously (c, f) exhibit processes that are longer than those exhibited by cells treated with only one of them separately (a, b, d, e). (g) Quantitative representation of the average process length per cell in astrocytes generated by treatment with PACAP, CNTF, or both together over a period of 2 days, used at the indicated concentrations. Values represent mean ± SEM of data gathered from three independent experiments. The total number of cells analysed in each condition (n) is indicated.

Animals

Experimental protocols involving Wistar rats for the preparation of primary cultures or protein extracts from nervous tissue were approved by the institutional committee on research animal care, and meet the requirements of current Spanish and European Community legislation.

Results

Evaluation of transcriptional control regions by deletional analysis of the GFAP promoter in neural cells

RC2.E10 cells constitute a phenotypically homogeneous and stable monoclonal cell line that express GFAP and that have been used extensively in transfection studies in our laboratory (Schwartz and Vallejo 1998; McManus et al. 1999; Schwartz et al. 2000; Pérez-Villamil et al. 2004). Thus, we initially used this cell line to determine the approximate boundaries of transcriptional control regions that regulate cell-specific GFAP gene expression.

Cells were transfected with the luciferase reporter plasmid GFAP-A7Luc, spanning approximately 1.5 kb of the GFAP promoter, or luciferase reporter plasmids generated by sequential 5′ deletions of the fragment in GFAP-A7Luc (Bonni et al. 1997). GFAP-A7Luc reporter activity was readily detectable in transfected RC2.E10 cells. As shown in Fig. 1, a 5′ deletion of the GFAP promoter to nucleotide −1342 (GFAP-A6Luc) resulted in an increased luciferase activity by approximately twofold. In contrast, luciferase activity was significantly reduced when a plasmid with a deletion to nucleotide −1049 (GFAP-A5Luc) was used (Fig. 1). However, more extensive deletions to nucleotides −384 (GFAP-A3Luc) and −106 (GFAP-A2Luc) resulted in partial recovery and slight enhancement, respectively, of the transcriptional activity relative to that observed with the full-length GFAP-A7Luc plasmid. Finally, the luciferase activity observed in RC2.E10 cells transfected with a plasmid bearing a deletion to nucleotide −35 (GFAP-A9Luc) was low and close to background levels (Fig. 1), indicating that the region spanning nucleotides −106 to −35 contains regulatory elements that are sufficient to activate transcription of a similar magnitude to that observed with the full-length promoter.

NFI recognizes the proximal promoter region of the GFAP gene

The proximal promoter region identified in the transfection experiments includes a segment that contains a sequence motif that coincides with the consensus binding site 5′-TGG(N6)GCCA-3′ for NFI (Liu et al. 1997; Krohn et al. 1999; Rafty et al. 2002). This site, spanning nucleotides −79 to −67, was found to be important for induction of GFAP gene transcription by TGF-β1 and repression by interleukin-1β in primary astrocytes, although binding of NFI was not demonstrated (Krohn et al. 1999).

To test for DNA–protein interactions in this region by EMSA, a double-stranded oligonucleotide spanning nucleotides −85 to −61 was synthesized. Incubation of this oligonucleotide with nuclear extracts prepared from RC2.E10 cells resulted in the generation of DNA–protein complexes whose specificity was determined by competition with unlabelled oligonucleotide added in excess to the binding reaction, or with an oligonucleotide of unrelated sequence, that failed to compete (Fig. 2a).

Figure 2.

 NFI present in RC2.E10 cells recognizes a specific target site in the rat GFAP gene promoter. (a, b, c) Electrophoretic mobility shift assays showing the binding of proteins present in nuclear extracts of neural RC2.E10 cells to an oligonucleotide probe corresponding to a region of the rat GFAP promoter that contains a consensus NFI binding site (nucleotides −85 to −61). (a) Nuclear extracts were incubated in the absence (–) or presence of a competing oligonucleotide of identical probe sequence (Comp) or in the presence of a non-specific competing (NSC) oligonucleotide of unrelated sequence, each used in a 100-fold molar excess. (b) Differential competitions carried out with the addition of a 100-fold molar excess of wild-type (WT) or mutated (Mut) oligonucleotides. The sequence of the competing oligonucleotides is shown in (d). (c) Binding reactions carried out in the absence (–) or presence of either normal rabbit serum (NRS) or anti-NFI α8199 or α2902 antisera. Arrows indicate bands corresponding to protein–DNA complexes containing NFI and to the supershifted (SS) complexes. (d) Sequence of the WT or mutated oligonucleotides used in the electrophoretic mobility shift assays. The consensus NFI binding site is indicated within a rectangle, and the nucleotides mutated are underlined. The nucleotide number relative to the transcription initiation site is indicated on top. (e) Relative luciferase activities elicited after transfections of RC2.E10 cells with either WT or mutated (Mut) GFAP-A7Luc reporter plasmids. As indicated in (d), the plasmid GFAP-A7MutBLuc (MutB) incorporates a 4-bp mutation within the NFI site, whereas the plasmid GFAP-A7MutCLuc (MutC) incorporates a 4-bp mutation immediately downstream from the NFI site (see Fig. 2d). Values are expressed as percentages of the activities elicited by GFAP-A7Luc, and represent the mean ± SEM of at least three independent experiments carried out in duplicate.

These complexes were also competed when an oligonucleotide bearing a 4-bp mutation located immediately 5′ to the NFI consensus binding site (MutA) was used (Fig. 2b). In contrast, an oligonucleotide incorporating a 4-bp mutation within the NFI binding site (MutB) failed to compete these complexes efficiently (Fig. 2b). Finally, we also used an oligonucleotide incorporating a 4-bp mutation located immediately 3′ to the NFI consensus binding site (MutC). Addition of this oligonucleotide resulted in partial competition, allowing the identification of a slower migrating complex corresponding to proteins that bind immediately downstream from the NFI consensus site (Fig. 2b).

To determine whether the identified complexes contain NFI proteins, we carried out EMSA in the presence of two different anti-NFI antisera. One of these antisera (anti-NFI/8199) recognizes an N-terminal region that is conserved within the DNA binding and dimerization domains of all NFI isoforms (Gronostajski 2000), whereas the other one (anti-NFI/2902) recognizes a segment of residues located within the C-terminal region of the NFI-C protein (Ortiz et al. 1999). Addition of the 8199 antiserum resulted in the disappearance of the lower complex, indicating the presence of bound NFI proteins (Fig. 2c). In turn, the addition of the 2902 antiserum resulted in the disappearance of the same complex and the appearance of a supershifted band, indicating that one of the NFI proteins recognized by the oligonucleotide in RC2.E10 cells corresponds to NFI-C (Fig. 2c). Thus, these experiments indicate that the nucleotides defined by MutB are essential for binding of NFI proteins to this site.

To investigate the functional consequences of lack of binding of NFI to the GFAP gene, we introduced mutations corresponding to the B or C sites in the context of the full-length promoter within the GFAP-A7Luc plasmid. Mutation of nucleotides located within the NFI site resulted in a significant decrease in GFAP-A7Luc activity (Fig. 2e, MutB). In contrast, mutation of nucleotides located downstream from the consensus NFI binding site (MutC) did not alter significantly the luciferase activity elicited by GFAP-A7Luc in transfected RC2.E10 cells (Fig. 2e). Thus, these experiments support the notion that NFI is important for expression of the GFAP gene.

NFI is expressed by primary neural precursor cells from the developing cortex

As RC2.E10 cells are a clonal cell line originally immortalized with a viral oncogene, they may not be necessarily representative of precursor cells that differentiate into astrocytes in the developing brain (McManus et al. 1999). Therefore, we sought to determine whether NFI is expressed in primary precursor cells from the developing cortex, and whether it is important for the expression of GFAP in astrocytes differentiated from these cells. Based on our previous work, we used PACAP as an stimulus for the differentiation of astrocytes from cortical cells (McManus et al. 1999; Vallejo and Vallejo 2002).

First, we investigated by immunocytochemistry the presence of NFI in cortical precursor cells that differentiate into astrocytes. As shown in Fig. 3, we found NFI immunoreactivity in the nuclei of primary cortical cells prepared from E17 rat brains and cultured in defined medium in the presence of bFGF. We have previously determined that these cells are proliferating, nestin-positive neuroepithelial precursors (McManus et al. 1999; Vallejo and Vallejo 2002). Figure 3 also shows that NFI is present in astrocytes differentiated from these neural precursors in response to exposure to PACAP (100 nm) over 2 days. In both cases, comparison with images taken by phase contrast micros‘copy showed that all cells were immunoreactive for NFI (Fig. 3).

Figure 3.

 Expression of NFI in primary cortical neuroepithelial precursor cells induced to differentiate into astrocytes by exposure to PACAP. Cells were cultured in defined medium in the presence of bFGF (20 ng/mL) for 24 h. After this, the medium was replaced with fresh identical medium (left panels), or with medium lacking bFGF but containing PACAP (100 nm; right panels). Two days after this, cells were fixed and processed for immunocytochemistry for GFAP or NFI. Note that NFI immunoreactivity is detected in cells both before and after astrocyte differentiation. The lower panels represent cells processed for NFI immunocytochemistry in the presence of an NFI-specific blocking peptide.

Next, we sought to determine whether NFI present in primary neuroepithelial precursor cells is functionally related to the stimulation of GFAP gene expression that occurs when astrocytes differentiate. Thus, to determine whether NFI binds specifically to cognate elements of the endogenous GFAP gene in the context of native chromatin in vivo, we used ChIP assays. We found that the anti-NFI antiserum, but not control rabbit IgG, immunoprecipitates from primary cortical precursor cells a fragment of formaldehyde cross-linked chromatin that contains the NFI binding site within the proximal promoter region of the GFAP gene defined in our previous experiments (Fig. 4a). This immunoprecipitated fragment was detected both in primary cortical precursor cells from E17 rat brains cultured in defined medium in the presence of bFGF, and in astrocytes whose differentiation from these cells had been induced by withdrawal of bFGF and treatment with PACAP (100 nm) for 2 days (Fig. 4a). In contrast, no DNA amplification was obtained in samples immunoprecipitated from Rat-1 fibroblasts, which were used as non-GFAP-expressing control cells (Fig. 4b).

Figure 4.

 Chromatin immunoprecipitation assays indicating the occupation of the endogenous GFAP gene promoter by NFI present in primary cortical neuroepithelial cells. Immunoprecipitations were carried out in the presence of anti-NFI antiserum or control normal rabbit IgG, using undifferentiated primary cortical precursor cells growing in the absence of PACAP (–PACAP) or astrocytes generated after exposure of those cells to 100 mm PACAP (+PACAP). PCR products corresponding to the region of the GFAP gene promoter that contains the NFI binding site are shown. Specificity of the immunoprecipitated chromatin fragment corresponding to the GFAP gene was tested by carrying out PCR amplifications corresponding to a fragment of the promoter of the somatostatin gene (SMS). In this case, no DNA fragment was amplified in samples obtained after immunoprecipitation with the anti-NFI antiserum. In addition, no DNA was amplified when Rat-1 fibroblasts were used.

The onset of expression of NFI in the developing cortex coincides with the beginning of astrogenesis

In the developing rat brain, cortical astrocytogenesis starts at about E17 and proceeds post-natally (Jacobson 1993; Qian et al. 2000). To investigate the existence of a possible relationship between the onset of astrogenesis and NFI expression, we carried out western immunoblots using nuclear extracts prepared from developing cerebral cortex obtained from E14 to E19 fetuses or from newborn rat pups. We found that NFI immunoreactivity was practically undetectable in cortical cells from E14 rats. In contrast, at E17 an immunoreactive band corresponding to NFI was clearly detectable (Fig. 5a). NFI immunoreactivity was also detected in nuclear extracts prepared from cerebral cortex of E19 and newborn rats (Fig. 5a). For comparison, GFAP immunoreactivity was also determined in protein extracts from developing cerebral cortex. As expected, GFAP was practically undetectable at E14, but was clearly detectable with increasing intensity from E17 (Fig. 5a). Thus, expression of NFI in the developing rat brain coincides chronologically with the onset of astrocyte differentiation, and remains elevated during the time in which astrocytogenesis occurs.

Figure 5.

 NFI expressed during late brain development binds to the consensus NFI site present in the rat GFAP gene promoter. (a) Western immunoblot showing the expression of NFI and GFAP in nuclear or whole cell extracts, respectively, prepared from developing cerebral cortex of rat embryos of 14 (E14), 17 (E17) and 19 (E19) gestation days, or from the cortex of rat pups of 1 day of post-natal age (P1). Note that NFI and GFAP exhibit similar patterns of expression. CREB immunoreactivity was used as a control for even loading. (b, c, d) Electrophoretic mobility shift assays showing the binding of proteins present in the nuclear extracts indicated in (a) to an oligonucleotide probe corresponding to a region of the rat GFAP promoter that contains a consensus NFI binding site (nucleotides −85 to −61). In (c) and (d), only nuclear extracts from the cortex of E17 rat fetuses were used. (c) Competitions were carried out by adding a 100-fold molar excess of WT or mutated (Mut) oligonucleotides. The sequences of those oligonucleotides are indicated in Fig. 2(d). The arrow indicates the presence of protein–DNA complexes that bind to the NFI consensus site. (d) Binding reactions were carried out in the presence of either normal rabbit serum (NRS) or anti-NFI α8199 or α2902 antisera. The arrow indicates protein–DNA complexes containing NFI that disappear in the presence of the α8199 antiserum, and the arrowhead indicates supershifted complexes generated in the presence of the α2902 antiserum.

Nuclear extracts prepared directly from developing cerebral cortex were also used in EMSA to assess the binding of nuclear proteins to the NFI site in the GFAP gene promoter. Incubation of a 32P-labelled oligonucleotide including the NFI site of the GFAP gene (nucleotides −85 to −61) with nuclear extracts prepared from E14 fetuses yielded only a weak retarded band, indicating that, at this time of development, this site is not efficiently recognized by nuclear proteins of the fetal cortex (Fig. 5b). In marked contrast, incubation of E17 nuclear extracts resulted in the generation of DNA–protein complexes that yield intensely labelled retarded bands. These complexes could also be detected when nuclear extracts from cerebral cortex of E19 or newborn rats were used (Fig. 5b). Thus, binding of nuclear proteins to the NFI site coincides with the increase in NFI expression observed in developing cerebral cortex starting at around E17 and extending to at least post-natal day 1.

Competition experiments using nuclear extracts from E17 rat cerebral cortex showed that the integrity of the NFI consensus sequence is essential for binding of nuclear proteins to this site. Thus, an oligonucleotide containing mutated nucleotides outside and upstream from the NFI site (MutA) was able to compete as efficiently as the WT oligonucleotide, an oligonucleotide containing mutated nucleotides within the NFI site (MutB) failed to compete completely, and an oligonucleotide containing mutated nucleotides outside and downstream from the NFI site (MutC) could only compete partially (Fig. 5c).

Finally, the addition of anti-NFI/8199 or anti-NFI/2902 antisera to the binding reaction indicated that NFI proteins expressed in cells from the developing cortex are present within the DNA–protein complexes assembled on the probe (Fig. 5d).

The NFI binding site is important for optimal GFAP promoter activity in differentiating astrocytes

In order to investigate whether the NFI site is important for GFAP gene expression during astrocyte differentiation, we carried out transient transfections with GFAP–luciferase reporter plasmids, using primary cortical cells obtained from E17 rat fetuses. GFAP gene expression was induced after transfections by treatment with PACAP (100 nm) (Vallejo and Vallejo 2002) or with CNTF (30 ng/mL) (Johe et al. 1996; Bonni et al. 1997; Rajan and McKay 1998; Park et al. 1999) for 48 h.

GFAP-A7Luc basal activity was clearly detected above background levels and represented approximately 1% of the activity elicited by RSV-Luc. Treatment of transfected precursor cells with PACAP or with CNTF resulted in a 4–5-fold stimulation of GFAP-A7Luc activity (Fig. 6), which reflects the observed astrocyte differentiation induced by both agents in separate plates of cells, as monitored by immunocytochemistry (not shown). When both PACAP and CNTF were administered simultaneously on the same cells, GFAP-A7Luc activity was increased to higher levels than those observed when each one of these factors was used alone (Fig. 6). This combined effect was observed both with relatively low concentrations of PACAP (10 nm) and CNTF (10 ng/mL), and with the higher concentrations used (100 nm and 30 ng/mL, respectively; Fig. 6). In this latter case, the effect of the combined treatment on luciferase activity was not higher than the one observed when a low concentration of PACAP (10 nm) was administered with a higher concentration of CNTF (30 ng/mL) (Fig. 6).

Figure 6.

 Relative luciferase activities elicited after transfections of primary cortical neuroepithelial precursor cells from the brains of E17 rat fetuses with either WT GFAP-A7Luc reporter plasmid, or a with similar plasmid that incorporates a 4-bp mutation either within (MutB) or outside (MutC) the NFI binding site (these mutations correspond to those indicated in Fig. 2d). Transfected cells were left untreated or were treated with PACAP, CNTF, or both. Values are expressed as percentages of the activities elicited by GFAP-A7Luc in untreated cells, and represent the mean ± SEM of at least three independent experiments carried out in duplicate. *p < 0.005; **p < 0.001 (anova). Differences in the activities elicited by PACAP and/or CNTF on the MutB plasmid were not statistically significant.

Similar transfections were carried out using the GFAP-A7MutBLuc reporter, which contains a 4-bp mutation within the NFI site in the GFAP promoter. The presence of this mutation in the NFI site resulted in significantly decreased basal levels of luciferase activity (Fig. 6). However, treatment of transfected cells with PACAP stimulated the transcriptional activity of the mutated GFAP promoter, although the luciferase activity elicited by this treatment only reached levels similar to those exhibited by the unstimulated WT promoter. Similarly, treatment with CNTF did not result in levels of luciferase activity above those observed in basal conditions with the WT promoter (Fig. 6). Notably, when both agents were used simultaneously on the transfected cells, the observed stimulation of luciferase activity from GFAP-A7MutBLuc was not higher that those observed when either PACAP or CNTF were added independently (Fig. 6).

Finally, transfection of the GFAP-A7MutCLuc reporter plasmid incorporating a 4-bp mutation immediately 3′ to the NFI binding site yielded luciferase levels that were similar to those observed after transfections with the WT promoter construct (Fig. 6). In this case, treatment of transfected precursor cells with PACAP or with CNTF resulted in a 2–3-fold stimulation of GFAP-A7MutCLuc activity (Fig. 6) and, when both PACAP and CNTF were administered together, this activity was further increased to higher levels than those observed when each one of these factors was used alone (Fig. 6).

PACAP and CNTF cooperate to stimulate GFAP expression synergistically in differentiating astrocytes

The experiments described in Fig. 6 suggest that NFI is important for the integration of the signalling mechanisms elicited by PACAP and CNTF at the level of the GFAP promoter, resulting in enhancement of GFAP gene expression when both factors are present simultaneously. To test more directly the hypothesis that PACAP and CNTF can act together to enhance GFAP expression synergistically during astrocytogenesis, we carried out astrocyte differentiation experiments using cortical precursor cells exposed to both factors.

Primary cortical precursor cells from E17 rat brains were cultured in defined medium in the presence of bFGF for at least 24 h after plating, and bFGF was withdrawn prior to treatment with CNTF or with PACAP. Exposure of cells to CNTF alone resulted in a concentration-dependent increase in the number of GFAP-positive cells generated over a period of 2 days in culture (Fig. 7). The highest level of differentiation response in terms of percentage of GFAP-positive cells relative to the total number of cells was reached at a concentration of CNTF of 30 ng/mL (Fig. 7). As in previous studies (Vallejo and Vallejo 2002), exposure of cells to PACAP resulted in a concentration-dependent generation of astrocytes. The optimum concentration of PACAP to trigger the differentiation of most cortical precursor cells is 100 nm. Therefore, we used suboptimal concentrations of PACAP (10 nm) and CNTF (10 ng/mL) to assess the possible existence of a cooperative effect between the two agents in promoting astrocyte differentiation.

Figure 7.

 Generation of GFAP-expressing cells from primary cortical precursor cells induced by PACAP and CNTF. Cortical cells from E17 rat brains were cultured in serum-free defined medium containing 20 ng/mL bFGF. After withdrawal of bFGF, cells were left untreated (–) or were treated for 2 days with PACAP or CNTF at the indicated concentrations. One group of cells was exposed to 10 nm PACAP and 10 ng/mL CNTF simultaneously. Cells were then fixed and processed for immunocytochemistry with a monoclonal anti-GFAP antibody. The percentage of GFAP-positive cells relative to the total number of cells per field of vision is depicted. Values represent mean ± SEM of data gathered from three experiments carried out in duplicate. At least 8–10 non-overlapping fields of vision per dish were examined.

As shown in Fig. 7, the exposure of cells to relatively low concentrations of PACAP and CNTF combined resulted in an additive increase in the percentage of GFAP-positive cells, as compared with the number of GFAP-positive cells obtained with each agent given independently. Considering the number of differentiated cells as the only parameter of the response, these results do not necessarily support the conclusion that CNTF and PACAP act synergistically. However, we noticed that the presence of both PACAP and CNTF simultaneously resulted in the generation of GFAP-positive cells with a more elaborate morphology than cells treated with only one of these agents independently (Fig. 8). To analyse this in detail, we treated cortical precursor cells with different concentrations of PACAP and CNTF, alone or in combination, and quantified the total length of processes generated in each cell after performing GFAP immunocytochemistry. As shown in Fig. 8, cells treated with either PACAP or CNTF alone generate processes of approximately similar lengths that do not increase in a concentration-dependent manner (Fig. 8g). In contrast, when PACAP and CNTF are present simultaneously at relatively low concentrations (10 nm and 10 ng/mL, respectively), astrocytes generated extend cellular processes that are significantly longer that those developed by cells treated with either PACAP or CNTF alone, even when these agents are used at higher concentrations (Figs 8a–c, quantified in Fig. 8g). Even longer processes were observed when PACAP, used at low concentration (10 nm), was present simultaneously with CNTF at an intermediate concentration (30 ng/mL) (Figs 8d–f, quantified in Fig. 8g), or when PACAP, used at a high concentration (100 nm) was present simultaneously with CNTF at a low concentration (10 ng/mL; image not shown, quantified in Fig. 9g). Thus, in these cases the length of the processes when both agents are present together is higher than additive, relative to that observed with each single factor, indicating the existence of synergism between PACAP and CNTF to increase GFAP expression. Western immunoblot confirmed that the content of GFAP in cells exposed to the combined treatment is higher than those observed in cells treated with a single agent (Fig. 9).

Figure 9.

 Western immunoblot showing expression of GFAP in extracts of primary cortical cells treated for 2 days with PACAP and/or CNTF at the concentrations (nm for PACAP, ng/mL for CNTF) indicated on top of each lane. β-Actin immunoreactivity was used as a control to monitor even loading.

Discussion

Positive and negative regulation of GFAP gene expression

Astrocytes arise from the same precursor cells that give rise to neurons, but relative to the generation of neurons, the generation of astrocytes is delayed in time during brain development (Qian et al. 2000; Gorski et al. 2002; Anthony et al. 2004). This suggests the existence of silencing mechanisms that prevent the inappropriate expression of GFAP before the onset of astrogenesis. Indeed, it has been shown that stimuli that induce neuronal differentiation trigger simultaneously the activation of inhibitory mechanisms that prevent GFAP gene transcription (Park et al. 1999; Sun et al. 2001). Astrocytes themselves have been recently suggested to act as progenitor cells to generate neurons in the adult central nervous system (Doetsch et al. 1999; Heins et al. 2002), and it is likely that during that process active repression of GFAP gene expression at the transcriptional level occurs. In addition, it has been shown that astrocytes generated from radial glial precursors can repress GFAP expression and revert to a radial glial phenotype (Schmid et al. 2003).

Our results show that the GFAP gene promoter contains regions that behave as transcriptional silencers. Thus, a significant elevation in the basal activity of the GFAP promoter was observed with a 5′ deletion to nucleotide −1342, suggesting the existence of negative-regulatory elements upstream from that position.

A deletion to nucleotides −1049 significantly decreased transcriptional activity, indicating the existence of important positive-regulatory elements located between nucleotides −1342 and −1049. However, the progressive increases in transcriptional activity observed with further deletions to nucleotides −384 and −106 suggest the presence of transcriptional repressor elements located between nucleotides −1049 and −384, as well as between −384 and −106. Finally, we found that regulatory elements located between nucleotides −106 and −35 are sufficient to yield levels of transcriptional activity approximately similar to those observed with the full-length GFAP promoter used in these studies.

We found that NFI binds to a regulatory element located within this region, and that integrity of this element is important for the activity of the GFAP promoter. NFI is known to be required for cAMP-dependent inducibility of the myelin basic protein gene (Clarck et al. 2002), and to be directly regulated by cAMP in thyroid cells (Ortiz et al. 1999). However, although mutation of the NFI binding site significantly reduced basal promoter activity, it did not affect induction by PACAP, which signals via cAMP (Vallejo and Vallejo 2002), in cortical precursor cells, indicating that elements regulated by PACAP/cAMP are independent of the NFI site. Work related to the detailed characterization of these elements is in progress in our laboratory and will be reported elsewhere.

Binding of NF1 to the GFAP promoter

Our data indicate the presence of the NFI-C isoform complexed on the proximal promoter region of the GFAP gene, but other isoforms may also be present because complexes visualized by EMSA were inhibited in the presence of the 8199 antiserum, which recognizes all related NFI proteins, and the antiserum used in ChIP assays does not discriminate among the different NFI isoforms. Indeed, NFI-A has previously been detected in the mouse cortex at E15 (Shu et al. 2003), a developmental age that corresponds approximately to rat E16.5, in agreement with our finding of increased binding to the NFI site in nuclear extracts from rat E17 cortex. The possible presence of several NFI isoforms, as well as the relative heterogeneity of the intact fetal brain tissue used to prepare the nuclear extracts, may have contributed to the relatively low band resolution observed in these experiments as compared with those carried out with the RC2.E10 cell line.

The NFI family of transcription factors, encoded by at least four different genes, comprises a complex group of related proteins that dimerize with one another (Gronostajski 2000). Some of these isoforms have been shown to be present in the central nervous system (Sumner et al. 1996; Chaudhry et al. 1997; Shu et al. 2003; Gray et al. 2004) and to regulate the expression of neural genes (Bedford et al. 1998; Clarck et al. 2002; Wang et al. 2004). We found that the expression of NFI proteins appears to be practically undetectable in the cortex of E14 rat embryos. However, previous studies have shown that several NFI gene transcripts are present in the mouse cortex from at least E11.5 (Chaudhry et al. 1997; Shu et al. 2003). This apparent discrepancy may be as a result of poor translation of NFI mRNAs early in development or to relative limitations of the antiserum to detect specific NFI protein isoforms.

Krohn et al. (1999) reported that treatment of astrocytes with TGF-β1 results in increased binding of nuclear proteins to the NFI site. However, binding of NFI proteins themselves could not be demonstrated. We could not detect the presence of NFI using the same commercial antiserum used by Krohn et al. (1999), and therefore limitations of this reagent cannot be ruled out. Nevertheless, these authors used fully differentiated mature astrocytes, and it is possible that in these cells treated with TGF-β1 the NFI site may be preferentially occupied by other proteins, whereas NFI may play a more important role in the stimulation of GFAP gene expression during astrocyte differentiation from cortical precursor cells. In support of this concept, it should be noted that disruption of the genes encoding NFI-A or NFI-B impairs GFAP gene expression and causes glial-cell related developmental anomalies in mice (Das Neves et al. 1999; Shu et al. 2003; Steele-Perkins et al. 2005). Furthermore, it is noteworthy that NFI expression increases, coinciding with the onset of astrocytogenesis at around E17, and that at this time it occupies the GFAP promoter in vivo.

At E14, cortical progenitor cells treated with PACAP (our unpublished observations), CNTF or bone morphogenetic proteins, do not give rise to astrocytes, because competence to respond to astrogenic signals is only gained late in development (Mabie et al. 1999; Molnéet al. 2000; Pachinsion et al. 2001; Takizawa et al. 2001). The acquisition of competence is gained, in part, by mechanisms that involve chromatin remodelling induced by bFGF (Song and Ghosh 2004). It has been found that, during the process of chromatin remodelling that facilitates tissue-specific gene activation, NFI acts early to preset the chromatin structure and stabilize the position of nucleosomes to maintain an open chromatin conformation compatible with transcriptional activity (Lefevre et al. 2003; Belikov et al. 2004; Vicent et al. 2004). This type of chromatin transition induced by NFI could represent a general mechanism for determining tissue-specific gene expression (Belikov et al. 2004).

Taken together, several of our observations are in agreement with a possible role of NFI in participating in the promotion of an active chromatin state of the GFAP gene prior to astrocyte differentiation. First, NFI appears to be undetectable at E14, a developmental time in which cortical precursor cells are not competent to generate astrocytes; second, NFI expression increases and is readily detectable at E17, a developmental time that coincides with the onset of astrocytogenesis; third, NFI occupies the GFAP promoter in cortical precursor cells from E17 embryos prior to the induction of astrocyte differentiation; and fourth, prevention of binding of NFI to the GFAP gene by mutation of the NFI site significantly impairs full activation of the GFAP promoter by PACAP or CNTF, even although the signalling pathways that mediate CNTF or PACAP transcriptional transactivation effects appear to be intact.

Synergism of PACAP and CNTF during astrocytogenesis

Data from transfection experiments in primary cortical cells were interpreted, bearing in mind that modest changes in GFAP promoter activity could generate relatively large amounts of GFAP protein that accumulate in the cells (Bonni et al. 1997). Thus, the enhanced stimulation of the GFAP promoter by simultaneous exposure of cells to PACAP and CNTF cannot simply be explained as an additive effect, because the introduction of a mutation that prevents binding of NFI to the GFAP promoter, although reducing the basal activity, does not prevent transcriptional transactivation by PACAP alone (about 8-fold) or CNTF alone (about 6-fold), but prevents the additional enhancement observed in the presence of both simultaneously. The notion that CNTF and PACAP cooperate to facilitate astrocytogenesis is further supported by our finding that astrocytes generated in the presence of both factors acquire a more elaborate morphology than those generated by exposure to only one of them, with significantly longer processes than expected with an additive effect, thus reflecting a higher production of GFAP as a consequence of a synergistic interaction. Other astrogenic factors have also been found to cooperate to facilitate astrocyte differentiation (Nakashima et al. 1999; Rajan et al. 2003).

The molecular basis for the observed synergism remains to be determined. Ongoing experiments in our laboratory have identified a cAMP response element that mediates PACAP-induced GFAP transactivation (Vallejo and Vallejo 2002) located between nucleotides −106 and −35, that is distinct from the NFI binding site described here. In contrast, CNTF promotes astrocyte differentiation via activation of STAT proteins that bind specific GFAP promoter elements located at a distance (Bonni et al. 1997; Kahn et al. 1997; Rajan and McKay 1998). Therefore, it is possible that NFI is important for the integration of independent signalling mechanisms that synergize at a transcriptional level to facilitate GFAP gene transcription during astrocyte differentiation. How NFI interacts with other components of this transcriptional mechanism at the level of the GFAP promoter remains the subject of further investigations.

Acknowledgements

We thank Irina Rozovsky for providing GFAP–luciferase reporter plasmids, Rocio Ajo for help with immunocytochemistry, Ricardo Uña for assistance with quantitative image analysis, Ana Aranda and Pilar Santisteban for comments and antisera, and members of our laboratory for critical reading of the manuscript. This work was funded by grants from the Community of Madrid (08.5/0028/2001 and GR/SAL/0774/2004) and the Spanish Ministry of Education and Science (BMC2002-00870). BC was supported by a predoctoral fellowship from the Community of Madrid.

Ancillary