Glioblastoma Cell-specific Expression of Fibroblast Growth Factor Receptor-1β Requires an Intronic Repressor of RNA Splicing¹

The FGFR-1³ is part of a family of tyrosine kinase receptors known to play key roles in the growth and differentiation of the several cell types. Alternative RNA processing of the FGFR-1 primary transcript creates functional diversity for this receptor by altering ligand specificity and affinity, subcellular localization, and activity of the tyrosine kinase (1, 2). Therefore, dysregulation of FGFR-1 alternative splicing has the potential to alter several processes involving normal cellular growth. In normal human glial cells, the predominantly expressed form of FGFR-1 (FGFR-1α) has three extracellular immunoglobulin-like domains involved in ligand binding. However, in glial cell tumors there is a switch to a form having two immunoglobulin-like domains (FGFR-1β; Ref. 3). The absence of the third domain has been shown to give FGFR-1β an affinity for acidic and basic FGFs 10-fold greater than that of the FGFR-1α (4, 5). Expression of this form of FGFR-1 in glial cells is believed to provide a cell-growth advantage and possibly to contribute to glial cell malignancy. The production of FGFR-1β results from skipping a single exon (α exon); therefore, alternative RNA processing plays a central role in glial cell malignancy. Using an RNA splicing cell culture model, we have previously identified an exon sequence that is required for α exon inclusion (6). In the present study, we have now identified a sequence located in the downstream intron that is required for α exon exclusion in the glioblastoma cell line SNB-19.

The human astrocytoma cell line SNB-19 and the human choriocarcinoma cell line JEG-3 were maintained as described previously (6, 7).

The plasmid constructs pFGFR-17, pFGFR-19, and pFGFR-20 have been described previously (7). All of the additional constructs used are derived from pFGFR-17 and maintain the same promoter (Rous sarcoma virus), flanking exons (hMT 2A gene), and vector backbone (pGEM 4; Fig. 1 A). Deletion constructs were created using a strategy in which two EcoRV sites were created by mutagenesis, followed by EcoRV digestion and self-ligation. All of the mutations were introduced using the QuikChange site-directed mutagenesis kit (Stratagene). For some reactions the protocol was modified to employ three primers instead of two primers as described previously (6). The EcoRV sites were introduced into seven separate locations using primers FP34, FP63, FP64, FP65, FP99, FP100, or FP101. For pFGFR-61, an insert encoding FGFR-1 exon 4 and flanking intron was derived by PCR amplification of a genomic P1 clone with primers FP11 and FP12. The PCR fragment was inserted into pFGFR-49, which was created from pFGFR-17 by the introduction of a HindIII site 97 bp upstream and an EcoRV site 8 bp downstream from the α exon by mutagenesis. Mutations were introduced into construct FGFR-17 using primers FP133 through FP138 to create plasmids pFGFR-90, pFGFR-91, pFGFR-92, pFGFR-93, pFGFR-94, and pFGFR-95. Detailed information on the creation of specific plasmids will be provided upon request. All of the mutations and deletions were confirmed by sequencing analysis.

All transfections were performed using Dosper liposomal reagent (Boehringer Mannheim) according to previously described methodology (6). For RT-PCR analysis, mRNA was isolated 72 h after transfection using the mRNA Capture kit (Boehringer Mannheim) with RT-PCR performed in the same tube that contained the captured RNA (6). In all of the cases, the forward primer DS8 was ³²P-end-labeled. PCR reactions included 18–20 amplification cycles, which was previously determined to be in the linear range of the reaction for these products (6). A minimum of three independent experiments was performed with quantification and statistical analysis performed using stored phorphor image analysis with a Molecular Imager (Model GS-363, Bio-Rad) as described previously (6).

The oligonucleotide primers used were as follows: (a) FP11: 5′-GGGGAAGCTTGCAAGACACCTCCAGGT-3′; (b) FP12: 5′-GGGGAGTACTACACGTACCTTGTAGCC-3′; (c) FP34: 5′-GAGGAGGTTCACCGATATCTGAAACTGGCAGAG-3′; (d) FP44: 5′-CCCGCATCACAGGGGAGGAG-3′; (e) FP63: 5′-CAGAGAGGTGACTAGATA-TCGTGGGCCCCAGG-3′; (f) FP64: 5′-GAGATCTGGGCAAGATATCG-TGGAAAAAAATCACAGGG-3′; (g) FP65: 5′-GGAGAGGAACCCAGG-ATATCCGCAAGGGAGCTG-3′; (h) FP99: 5′-CTCCTCCCCCTCCGA-TATCTGCCCCCACTCTGC-3′; (i) FP100: 5′-CAGGGAGCAATGTGATA-TCGCAGCAGTTTCTG-3′; (j) FP101: 5′-CCTCTCTGAGAGCCAGATAT-CGCGGCAGGCAGGG-3′; (k) FP133: 5′-CCCCCTCCGTGATCAAACC-CCACTCAAATTCAGAAACTGCTGC-3′; (l) FP134: 5′-CACTCTGCTT-CAGAAACAAAAAACCACTAACATTGCTCCC-3′; (m) FP135: 5′-CTG-CCCACTAACATAAATCCCTGCCTGCCGC-3′; (n) FP136: 5′-CCAC-TAACATTGCTCCCAAAAAAAAGCGTGGCTTGGCTCTC-3′; (o) FP137: 5′-CCCCCTCCGTGATCAAACCCCACTCAAATTCAGAAACAAAAAA-C-3′; and (p) FP138: 5′-CCACTAACATAAATCCCAAAAAAAA-GC-3′.

A failure to recognize the α exon during RNA processing of the FGFR-1 primary transcript has been correlated with glial cell malignancy (3). Sequence analysis of a region encompassing the α exon has failed to find splice site mutations in glioblastoma cell DNA, which suggests that the aberrant RNA results from changes in trans-acting factors. To begin to identify these regulatory factors, we previously developed an RNA splicing reporter/transfection cell culture model system that mimics the pathways observed in normal brain glial cells and glioblastoma (7). The reporter construct (pFGFR-17) contains a 3.8-kb fragment of the FGFR-1 genomic sequence inserted into hMT 2A gene intron 1 (Fig. 1,A). RNA transcripts derived from this reporter construct demonstrate cell-specific recognition of the α exon; the exon is included predominantly in JEG-3 cells (86%) and excluded in the glioblastoma cell line SNB-19 (27%; Fig. 1 B). Previous studies identified a sequence within the α exon that was required for exon inclusion in JEG-3 cells but had little effect on splicing in SNB-19 cells (6). To identify regulatory sequences involved in the glioblastoma cell-specific splicing event, we have continued our analysis, focusing on the region located downstream from the α exon.

Three constructs containing progressive deletions beginning at the 3′ end of the FGFR-1 insert were examined for α-exon inclusion in transfected SNB-19 and JEG-3 cells (Fig. 1,B). For two of these constructs, pFGFR-19 and pFGFR-20, there was no significant increase in the level of α exon inclusion in either SNB-19 or JEG-3 cells. This is consistent with previous findings for these constructs in which intron size rather than the specific sequence was required to maintain α exon exclusion in SNB-19 cells (7). However, the removal of an additional 133 nucleotides of intron (pFGFR-80), resulted in a dramatic increase in α exon inclusion in the SNB-19 cells. The inclusion level of the α exon went from 38 to 87% in SNB-19 cells, whereas it increased only slightly in JEG-3 cells (Fig. 1 B; compare pFGFR-20 with pFGFR-80).

The data above localize a potential SNB-19 cell-specific inhibitor of α-exon inclusion to an intron region between 71 and 204 nucleotides downstream from the exon. However, because these constructs contain large intron deletions, we cannot definitively assign an inhibitory element to this region. The downstream intron in pFGFR-80 is reduced to only 201 nucleotides. Because a small intron size is correlated with enhanced splicing efficiency, it is possible that the elevated inclusion of α exon results solely from this reduction (8). Therefore, we tested constructs containing smaller internal deletions targeted to this region (Fig. 2). The pFGFR-81 construct contains a 115-bp internal deletion mapped to the region identified in Fig. 1. In transfected SNB-19 cells, the pFGFR-81 construct showed the same dramatic increase in α-exon inclusion seen for deletion construct pFGFR-80 (from 27 to 72%; Fig. 2). Similarly sized deletions, one immediately downstream (pFGFR-82), and a deletion removing the γ exon (pFGFR-83) had no effect on the level of α-exon inclusion in transfected SNB-19 cells (pFGFR-56 and pFGFR-58; Fig. 2). The level of α-exon inclusion in transfected JEG-3 cells was not increased for any of the constructs compared with pFGFR-17, with a slight decrease in inclusion seen for pFGFR-82 and pFGFR-83 (Fig. 2). These data support the findings of the previous deletion constructs and suggest a role for this intronic sequence, rather than intron size, as a cell-specific inhibitor of α-exon inclusion.

To narrow the inhibitory region further, constructs with smaller internal deletions spanning the 115 nucleotides were made. Two constructs, pFGFR-85 and pFGFR-86, had enhanced levels of α exon inclusion (both were 71%) which were nearly identical to the level observed for pFGFR-81 (72%; Fig. 2). Two additional deletion constructs, pFGFR-84 and pFGFR-87, both displayed only a moderate increase in α-exon inclusion compared to pFGFR-17. There was no apparent increase in exon inclusion for any of the constructs in JEG-3 cells (Fig. 2). The failure to localize enhanced α-exon to a single construct suggests that the deletions of pFGFR-85 and pFGFR-86 may split the regulatory element. Rather than performing additional smaller deletions, we chose to focus on the entire 62-nucleotide region mapped. This region was deleted in pFGFR-88 (Fig. 3,A). As expected, we observed a significant increase in α-exon inclusion relative to pFGFR-17 that was specific for the SNB-19 glioblastoma cell line (Fig. 3 A).

The 62-nucleotide sequence is relatively distant from the α exon and flanking splice sites. We questioned whether the inhibitory action of the element was specific to the α exon. Restriction sites were engineered into pFGFR-17 to allow replacement of the α exon and its flanking splice sites with exon 4 of the FGFR-1 gene. This exon codes for the second extracellular domain of FGFR-1 and is constitutively included in all of the endogenous FGFR-1 gene transcripts (data not shown). When exon 4 was flanked by α-exon intron sequence (pFGFR-61), cell-specific recognition of the exon was observed (36% inclusion in SNB-19 cells versus 83% in JEG-3 cells; Fig. 3,B). To confirm that down-regulation required the intron element, the same 62-bp deletion was introduced into pFGFR-61. As expected, the inclusion of exon 4 was increased significantly in SNB-19 cells (72 versus 36%), with a modest increase in JEG-3 cells (98 versus 83%; Fig. 3 B, pFGFR-89). Therefore, the activity of this intronic repressor element is not limited to the α exon, which suggests that it may function to disruption recognition of constitutive splice sites rather than targeting specific α exon sequences such as the exon enhancer (6).

It is clear that the 62-nucleotide region functions as a cell-specific repressor of RNA splicing. A sequence comparison found that the region was also conserved in the rat gene with a higher degree of homology (87%) than the surrounding intron (data not shown). This suggests that the entire 62-nucleotide region may serve as a functional element. The region shares two features in common with other intronic regulatory elements; the sequence is pyrimidine-rich and contains multiple copies of a UGC motif (Fig. 4). The UGC motif is repeated seven times within the 62-nucleotide region but found in only four additional places in the remaining 229 nucleotides of the downstream intron (data not shown). The UGC motif is reported to play a role in the regulated splicing of several genes (9, 10, 11, 12, 13, 14, 15, 16). However, for most of these examples, the UGC motif serves as a part of an intronic activator of splicing. The exception is the α-tropomyosin gene: the UGC motifs are found in two intron repressor elements that flank exon 3. They function to ensure the skipping of exon 3 in smooth muscle cell (15, 16, 17). To test whether the UGC motifs play a similar role in glioblastoma cell-specific RNA splicing, we generated constructs containing mutations of the 62-bp region in pFGFR-17. Constructs pFGFR-90, pFGFR-91, pFGFR-92, and pFGFR-93 each contain mutations specifically targeting one or two UGC motifs in the 62-nucleotide region (Fig. 4). Although a moderate increase in α exon inclusion was seen for pFGFR-90 and pFGFR-91, none of the constructs obtained the level of exon inclusion observed for construct pFGFR-88 (Fig. 4). This data suggested that either the UGC motif was not a critical component of the element, or that there was functional redundancy. To examine this further, we mutated all of the UGC motifs simultaneously (Fig. 4, pFGFR-94). For this construct, the level of α exon inclusion was clearly elevated in SNB-19 cells without altering RNA splicing in JEG-3 cells. Although this observation clearly implicates this region as an important regulator of α exon splicing, the high number of nucleotides mutated makes it difficult to draw conclusions regarding the role of the UGC motif. The fact that constructs pFGFR-85 and pFGFR-86 both show elevated α exon inclusion suggests that the functional element may comprise motifs mapping to both regions. In addition, the key UGC motifs involved in α-tropomyosin gene splicing exist as direct repeats (15). Therefore, we decided to mutate the two UGCUGC motifs simultaneously (Fig. 4, pFGFR-95). When this construct was tested in both of the cell types, the level of α exon inclusion was similar to that observed for pFGFR-88 and pFGFR-94 (Fig. 4). Thus, it would appear that these two elements play a key role in the function of the splicing inhibitory element; however, we cannot rule out a role for additional nucleotides.

The 62-nucleotide repressor identified by this study is clearly a critical element required for α exon skipping. What is somewhat less clear is the mechanism by which this element acts to suppress exon recognition in glioblastoma cells. For the α-tropomyosin gene, the UGC motifs are thought to modulate the binding of PTB to adjacent polypyrimidine tracts and ultimately to suppress normal recognition of the upstream 5′ splice site (16, 17). An additional repressor element functions in a similar manner upstream from the exon to prevent recognition of the branch point. We have evidence of a similar intronic repressor located in the intron upstream from the α exon.⁴ However, although the FGFR-1 gene repressor identified here contains multiple UGC motifs, it lacks a PTB-binding motif (UCUU in a pyrimidine-rich environment; Refs. 16, 17). This suggests that PTB does not interact with the FGFR-1 downstream sequence to inhibit α exon inclusion and that an alternative mechanism mediates glioblastoma-specific splicing of the FGFR-1 transcript. Whether this mechanism is unique to FGFR-1 or plays a role in the aberrant RNA splicing of other genes affected by transformation remains to be addressed.