PTBP1-dependent regulation of USP5 alternative RNA splicing plays a role in glioblastoma tumorigenesis

Abstract

Aberrant RNA splicing is thought to play a key role in tumorigenesis. The assessment of its specific contributions is limited by the complexity of information derived from genome-wide array-based approaches. We describe how performing splicing factor-specific comparisons using both tumor and cell line data sets may more readily identify physiologically relevant tumor-specific splicing events. Affymetrix exon array data derived from glioblastoma (GBM) tumor samples with defined polypyrimidine tract-binding protein 1 (PTBP1) levels were compared with data from U251 GBM cells with and without PTBP1 knockdown. This comparison yielded overlapping gene sets that comprised only a minor fraction of each data set. The identification of a novel GBM-specific splicing event involving the USP5 gene led us to further examine its role in tumorigenesis. In GBM, USP5 generates a shorter isoform 2 through recognition of a 5' splice site within exon 15. Production of the USP5 isoform 2 was strongly correlated with PTBP1 expression in GBM tumor samples and cell lines. Splicing regulation was consistent with the presence of an intronic PTBP1 binding site and could be modulated through antisense targeting of the isoform 2 splice site to force expression of isoform 1 in GBM cells. The forced expression of USP5 isoform 1 in two GBM cell lines inhibited cell growth and migration, implying an important role for USP5 splicing in gliomagenesis. These results support a role for aberrant RNA splicing in tumorigenesis and suggest that changes in relatively few genes may be sufficient to drive the process.

Figure 1

Identification of PTBP1-mediated RNA splicing targets. (A) PTBP1 expression levels derived from Affymetrix exon array data. Left graph displays values for U251 GBM cells treated with control (PTB Nrl) or knockdown (PTB KD) morpholinos. Right graph shows the PTBP1 expression values for the 10 GBM samples and 10 nontumor samples used in this study. (B) and (C) Venn diagrams depicting the statistical derivation of genes with PTBP1-mediated changes in splicing. Datasets defined in (A) were used to derive PTBP1 target genes as defined by the indicated GeneSpring GX statistical parameters.

Figure 2

Identification of individual exon splicing events regulated by PTBP1. Probeset hybridization plots were generated for the four genes, RTN4, ATP2B4, PALM2-AKAP2, and USP5, identified in Figure 1C. For each gene, the upper U251 dataset graph provides individual Affymetrix probeset identification numbers with exons schematically illustrated. Alternative splicing events are denoted by unfilled exons. The alternative event examined in the PALM2-AKAP2 gene involves the choice between a 5′ splice site (ss) and a polyadenylation site (A) in exon 7. The probesets denoting the PTBP1-regulated events are boxed. The lower graphs provide data derived from the patient sample arrays.

Figure 3

Splicing of USP5 correlates with PTBP1 expression levels. (A) Schematic representation of the USP5 mRNA isoforms. Alternative use of a 5′ splice site within exon 15 removes 69 bases of coding sequence to generate isoform 2. The location of PCR primer sets and Affymetrix probeset 3402325 are indicated. (B) Plot of dataset values for USP5 splicing versus PTBP1 expression levels. (C) Validation of USP5 splicing was performed by quantitative RT-PCR as described in Materials and Methods. Samples included normal human astrocytes (NHA), normal human cortical neurons (NHCN), U251 and LN229 GBM cell lines treated with control morpholino (Nrl) or PTBP1 knockdown morpholino (KD), two nontumor brain samples (NT) and three GBM samples (GBM).

Figure 4

Identification of PTBP1-binding sites and modulation of USP5 RNA splicing. (A) Schematic representation of SELEX-derived PTBP1-binding sites using HeLa cell nuclear extract within the USP5 alternatively regulated region [38]. The graph displays sequences found enriched (dark grey) or underrepresented (light grey) when examined for PTBP1 binding. The peak PTBP1-binding region maps from 6842820 to 6842829. PTBP1 binding is distinctly absent from the USP5 isoform 2 5′ splice site region. The UCSC Browser download for the genomic region indicated was modified to include only relevant information. The hybridization sites for the morpholino oligonucleotides and the sequence used to generate the UV cross-linking probe used in (B) are indicated. (B) UV cross-linking of U251 cell lysate with USP5 intron 15 IR800-endlabeled probe. Cross-linked complexes were immunoprecipitated (IP) with PTBP1 antibody (PTB) or antimouse immunoglobulin antibody (Ig), separated by SDS polyacrylamide gel electrophoresis and visualized by IR scanning. The PTBP1-specific band is noted. (C) Schematic illustration of the predicted outcome derived from the hybridization of 15 ss morpholino to USP5 precursor RNA. (D) Quantitative and semiquantitative (inset) RT-PCR analysis of U251 and LN229 GBM cells treated with control (C) or 15 ss morpholino.

Figure 5

Modulation of USP5 splicing to generate isoform 1 is associated with growth and migration reductions in GBM cells. U251 and LN229 GBM cells were treated with control or 15 ss morpholino as described in Materials and Methods. Cells were monitored for growth and migration as indicated in individual graphs. *P<0.05