Regulation of expression of glial filament acidic protein

In higher eucaryotes, the cytoskeleton is composed of three filamentous networks consisting of microtubules, actin filaments or intermediate filaments. Early in the developmental lineage leading to mature astrocytes, there is a switch from the synthesis of vimentin to the synthesis of glial filament acidic protein (GFAP) as the principal intermediate filament protein. The restricted expression of GFAP to astrocytes has made it a convenient marker for these cells, as well as for the study of gliomas. In addition, GFAP is a major component of the glial scars that occur following physical or biological trauma in the CNS. GFAP is therefore a major cell type-specific protein whose expression changes with development and responds to external cues.

Several years ago, we cloned a GFAP-encoding cDNA from mouse brain and determined the complete sequence of the corresponding single copy gene (Balcarek and Cowan, 1985). We have used a variety of constructs (described herein) to define two major cis-acting regulatory elements that contribute to the cell type-specific expression of this gene. One is a positive upstream regulatory element that resides about 1.5 kb 5’ to the transcriptional initiation site (nucleotides —1631 to-1479). The second is a negative regulatory element that lies within the first intron of the gene.

The constructs used in this work are described in the figure legends. Details of the assembly of these constructs has been published elsewhere (Sarkar and Cowan, 1991).

Three antisense ³²P-labelled probes were generated for quantitative analysis of expression in transfected cells (Fig. 1). The first extended from nucleotide-316 to +232 of the GFAP gene; the second spanned the cap site of the human β-globin gene, and was used to detect expression of GFAP/human β-globin chimeras; and the third extended from nucleotides-316 to +227 of a mouse β-tubulin cDNA, M/31 (Wang et al. 1986) and was used for the analysis of transcripts from GFAP/M/31 chimeras (see text).

HeLa cells, rat C6 gliomas and U251 human astrocytoma cells were grown in Dulbecco’s modified Eagle’s medium containing 10 % calf serum. Cells were cotransfected with the various GFAP promoter constructs and the pSVneo plasmid (Mulligan and Berg, 1981) as described by Graham and Van der Eb (1973). 48h after the addition of DNA, cells were split and the aminoglycoside G418 was added to the medium (0.4-0.5mgml^#1 final concentration) so as to select for cells that had acquired the pSVneo plasmid. 2-3 weeks later, G418-resistant cells were pooled, amplified and used for the isolation of RNA (Chirgwin et al. 1979).

Two cell lines that express GFAP at high levels were chosen for these studies: rat C6 cells (Benda et al. 1971; Brissel et al. 1975) and human U251 cells (Westermark, 1973). HeLa cells were used as a model of a non-glial cell that expresses no detectable GFAP.

In early experiments, we attempted to use the chloramphenicol acetyl transferase (CAT) gene as a reporter function to assay for the activity of a variety of GFAP-derived constructs. However, we found that the CAT plasmid without any GFAP promoter sequences was expressed at a high level in U251 cells, and it was impossible to distinguish background CAT activity from activity due to cloned GFAP-derived sequences. Consequently, we adopted an RNAase protection assay to define cis-acting elements in the mouse GFAP gene. This assay depends on the hybridization of a labelled DNA probe with specific RNA transcripts such that hybridised RNA molecules become resistant to digestion with ribonucleases. Differences in DNA sequence allowed us readily to distinguish between transcripts from the endogenous rat (C6) or human (U251) GFAP genes. There are two advantages to this approach. First, transcripts that initiate artifactually on vector sequences can be readily distinguished from those that are correctly initiated. Second, GFAP encoding sequences are retained. This turned out to be important, since sequences within the GFAP gene itself do indeed affect its level of expression.

We first established that the GFAP gene is regulated in a cell type-specific fashion by transfecting the entire cloned GFAP gene (including 2 kb each of upstream and downstream flanking DNA) into U251, C6 and HeLa cells. Analysis by RNAase protection of RNA prepared from pooled G418-resistant transfected cells (see Materials and methods) showed the presence of a diagnostic 232-nucleotide protected fragment derived from correctly initiated transcripts in U251 and C6 cells; no corresponding fragment was detectable in parallel experiments with HeLa cells (data not shown). Thus, the GFAP gene fragments used in these experiments contain sequences that are required in cis for the expression of the mouse GFAP gene in glial cells.

We first examined the consequences of removing elements of the 5’ flanking sequence by deleting a 501 bp segment extending from nucleotide-1980 to-1479. This led to the virtual abolition of expression in transfected U251 cells (Fig. 2) leaving the very low level of expression in HeLa cells unaffected. Therefore, a significant positive regulatory element (which we refer to as the distal element) resides within this 501 bp sequence.

To see whether the distal element could confer cell type-specific expression on a heterologous promoter, we assembled constructs (termed G.Hβ l and G’.H/3gl, see Fig. 3) in which the distal element was coupled in either orientation in front of a human β-globin gene fragment containing upstream sequences including the TATA box, but excluding the natural enhancer. Control experiments showed that this latter fragment did not express on its own in either U251 or HeLa cells (Fig. 3, tracks 1 and 6). A third chimeric plasmid was also constructed in which the distal element was located downstream from the human β-globin promoter. Analysis of U251 cells by RNAase protection assays yielded two bands (Fig. 3): a 140 bp band that is derived from correctly initiated GFAP transcripts and a 213 bp fragment representing transcripts initiated both at the authentic GFAP cap site and at the second exon of the human f>-globin gene. The construct G.Hβ gl is strongly expressed in U251 cells, in contrast to G’.Hβ gl or GOP.Hβgl (Fig. 3, tracks 3-5). None of these constructs are expressed in HeLa cells (Fig. 3, tracks 8-10); a control construct in gene lacking the distal element. Upper arrowhead shows location of the protected fragment corresponding to correctly initiated transcripts from the transfected mouse GFAP gene. Lower arrowhead shows protected fragment resulting from the endogenous expression of human GFAP.

which the SV40 enhancer was linked to the human β-globin gene yielded high levels of correctly initiated transcripts in both U251 and HeLa cells (Fig. 3, tracks 2 and 7). Two conclusions may be drawn from these data: (1) the distal element is capable of driving a heterologous gene in a cell type-specific fashion when it is positioned upstream in the 5-3’ orientation; (2) the distal element cannot function in either a distance or orientation independent manner, in contrast to the behavior of classical enhancers.

To define critical regulatory regions within the distal element in more detail, various subfragments were made and fused to the 5’ end of the human β-globin promoter. Each construct was assayed for its ability to express in U251 cells (see Materials and methods). The results of these experiments (Fig. 4) show that the positive regulatory element conferring cell type specificity on the human /S-globin gene is abolished by the deletion of sequences between nucleotides-1631 and-1569, suggesting that the binding of positive regulatory factor(s) depends on sequences in this region, either directly or indirectly, and that the upstream regulatory element is contained within the region-1631 to —1479.

To assess the contribution of intragenic sequences to the regulation of the GFAP gene, experiments were conducted in which these sequences were substituted by an unrelated cDNA (encoding a mouse /(-tubulin) which served as a reporter. Transfection of this construct (Gβ 1) into U251 and HeLa cells resulted in expression in both cell types, though at a higher level in U251 cells (Fig. 5, tracks 1 and 3). This result contrasts with parallel experiments using the entire GFAP gene cloned in the same vector; in this case, the level of expression in U251 cells is vastly greater than in HeLa cells (Fig. 5, tracks 2 and 4). These data imply the existence of one or more regulatory elements that lie 3’ to the transcriptional start site of the GFAP gene.

Note that whereas the entire upstream fragment leads to the expression of a chimeric gene (G/β 1) in both HeLa and glioma cells (Fig. 5), if only a portion of this fragment corresponding to the distal element is included, expression of another chimeric gene (G.Hβ gl) is restricted to glioma cells (Fig. 3). The difference between these two results is unlikely to be due to the different reporter sequences used; rather it must be due to sequences between the distal element and the transcription start site (nucleotides — 1479 to —1): sequences in this region direct the expression of some constructs in HeLa cells, but only in the absence of the GFAP gene itself. These data suggest a negative interaction between sequences within the GFAP gene and upstream promoter elements (see Discussion).

To further define the intragenic regulatory sequences, we made a series of constructs all containing the same 2 kb of upstream sequences, but lacking varying portions of the 3’ end of the GFAP gene (Fig. 6A-E). The transcriptional activity of each of these constructs was measured by RNAase protection following transfection into U251 and HeLa cells. Deletion constructs up to and including sequences downstream from nucleotide +2836 (located in the 4th intron) had little discernable effect on the level of expression in U251 cells and were expressed at a very low level (if at all) in HeLa cells (Fig. 6B, tracks 1-4 and 6-9). In contrast, however, removal of sequences downstream from nucleotide +897 (located in the 1st intron) resulted in a dramatic increase in the expression of correctly initiated transcripts in HeLa cells (Fig. 6B, tracks 5 and 10); this construct had therefore lost cell type specificity. This result implies the existence of a negative regulatory element that lies within the GFAP gene itself and is responsible, at least in part, for conferring cell typespecific expression by inhibiting expression in non-glial cells.

We made several additional deletion constructs in an attempt to further delineate this negative intragenic element (Fig. 6C). There is a conspicuous increase in expression in HeLa cells when the 3’ deletion of sequences progresses from nucleotides +1264 to +1084 (labelled 6 and 7 in Fig. 6C and D). This result suggests the existence of a negative regulatory element within the first intron of the GFAP gene.

Our data, together with that of Miura et al. (1990), demonstrate that a number of elements act together to direct the regulated expression of the gene encoding GFAP. Miura et al. (1990) identified three trans-acting factor binding sites in the GFAP gene by DNAase I footprinting between nucleotides —163 and —82. These binding sites, termed GFIII, GFII and GFI, show homology with the cAMP response element, the NF1 binding site motif and the AP2 binding site motif, respectively. By site-directed mutagenesis, NF1 was shown to be a positive regulator required for efficient expression of GFAP, while the two flanking sites are negative regulators of GFAP expression.

In contrast, our work has uncovered two more distant regulatory regions, a strong, glial-specific positive regulatory element (the distal element) contained between nucleotides —1631 and —1479, and a negative regulatory element located in the first intron of the GFAP gene (the intragenic element). The distal element alone can drive the expression of a heterologous gene in glioma cells (but not in HeLa cells), although not in the manner of a classical enhancer (Fig. 3). On the other hand, deletion of the intragenic element leads to the activation of GFAP expression in HeLa cells (Fig. 6).

Whereas our results are for the most part compatible with those of Miura et al. there are two points of apparent conflict. (1) Muirá et al. found that many of their constructs, all of which lack intragenic sequences, are expressed specifically in glioma cells, and not control cells, whereas we show that removal of intragenic sequences activates transcription in HeLa cells (Figs 5, 6). This difference can probably be ascribed to the different control cells used in the two studies. (2) Miura et al. found that what we call the distal element is not necessary for expression of their constructs in glioma cells, whereas we find that it is essential for expression in glioma cells (Fig. 2). This difference may be ascribed to the presence of intragenic sequences in our constructs: our data suggests that there is a negative interaction between intragenic sequences and proximal GFAP upstream promoter el-ements (see above, and Sarkar and Cowan, 1991). This negative interaction might lead to the dominance of the distal element we observe in our study. Since the constructs of Miura et al. lack intragenic sequences, this may allow the more proximal GFAP promoter elements they identify to predominate over the effects of the distal element in their study. The implied interaction of the upstream and intragenic elements is intriguing and could play a role in the modulation of GFAP expression during development and in response to injury.