Functional significance of U2AF1 S34F mutations in lung adenocarcinomas

Abstract

The functional role of U2AF1 mutations in lung adenocarcinomas (LUADs) remains incompletely understood. Here, we report a significant co-occurrence of U2AF1 S34F mutations with ROS1 translocations in LUADs. To characterize this interaction, we profiled effects of S34F on the transcriptome-wide distribution of RNA binding and alternative splicing in cells harboring the ROS1 translocation. Compared to its wild-type counterpart, U2AF1 S34F preferentially binds and modulates splicing of introns containing CAG trinucleotides at their 3' splice junctions. The presence of S34F caused a shift in cross-linking at 3' splice sites, which was significantly associated with alternative splicing of skipped exons. U2AF1 S34F induced expression of genes involved in the epithelial-mesenchymal transition (EMT) and increased tumor cell invasion. Finally, S34F increased splicing of the long over the short SLC34A2-ROS1 isoform, which was also associated with enhanced invasiveness. Taken together, our results suggest a mechanistic interaction between mutant U2AF1 and ROS1 in LUAD.

Fig. 1. Recurrence and co-association of U2AF1 S34F mutations.

a Recurrence of U2AF1 S34F mutations in multiple data sets including data summarized from cBioPortal, TRACERx, Stanford patients, and cancer cell lines including LUAD. b The approximate phylogeny tree of the two samples collected from a Stanford LUAD patient, illustrating U2AF1 S34F mutation and CD74-ROS1 fusion as truncal events. c Co-occurrence of SLC34A2-ROS1 fusions and U2AF1 S34F mutations in LUAD subset of cohort in a: A total of 2112 cases were analyzed, with highlighted cases of S34F mutations and ROS1 fusions shown in red. A Fisher’s exact test was performed to determine statistical significance.

Fig. 2. Genome-wide mapping of RNA interactions for wild-type and S34F mutant U2AF1.

a Schematic illustration of U2AF1 iCLIP. IP, immunoprecipitation; RNase, ribonuclease. Doxycycline-inducible FLAG-HA-tagged wild-type and S34F-mutant U2AF1 plasmids were transfected into HCC78 cells. Induced lysates were purified on anti-Flag-M2 agarose beads followed by a series of wash steps to specifically elute FLAG peptide-containing complexes and recaptured with anti-HA agarose. Standard iCLIP steps were subsequently performed to generate deep sequencing libraries. b Western blots were performed to confirm doxycycline induction of FLAG-HA-U2AF1 wild-type and S34F mutant constructs utilizing anti-HA and anti-U2AF1 monoclonal antibodies. Percent of HA-tagged U2AF1 of total U2AF1 is shown. c Autoradiogram of ³²P-labeled RNA crosslinked to Flag-HA-U2AF1 trimmed with two different concentrations of RNase A. RNA-protein complexes are seen in the purifications from Flag-HA-U2AF1 cells but not from the parental control cell line, HCC78. d Genomic distribution of U2AF1 iCLIP reads. lncRNA, long noncoding RNA. Wild-type and mutant U2AF1 iCLIP reads were annotated to known repetitive and non-repetitive regions of the human genome with percentage of total iCLIP reads shown. e Binding distribution of wild-type and S34F mutant U2AF1 iCLIP reads. UTR untranslated region, CDS coding sequence; 3′ SS.

Fig. 3. Determining binding specificities of wild-type and mutant U2AF1.

a U2AF1 binds a subset of 3’ SSs. Maximum-likelihood analysis was utilized to determine the 3′ SS occupancy of wild-type and S34F mutant U2AF1. Each dot represents an average occupancy of a group of 40 genes, in relation to average CLIP density per 3′ SS. b Metagene analysis of wild-type and S34F mutant U2AF1 binding interactions to pre-mRNA 3′ SSs. Normalized RT-stop density is shown across 3′ SS positions on the x-axis. c RT stops >10 were assigned to clusters defined from a window of 3′ SSs of −6 to + 2 nucleotides utilizing 3′ SS annotation files. The cluster assignments for wild-type and mutant samples are shown in left and right, respectively. d Z-scores were generated based on random hexamer nucleotide motif frequencies bound to wild-type and S34F mutant U2AF1. A subset of Z-scores showing enrichment of 50% of the hexamers bound to mutant U2AF1 for CAG and its reverse complement CTG trinucleotides. e WebLogos depicting binding preferences of wild-type and mutant U2AF1 for the main peaks in b. Position 0 is the intron exon junction depicted by the figure. f Generation of a preferential binding score to wild-type and mutant U2AF1. The top 10 scoring hexamers and its common trinucleotide sequences preferentially bound to mutant U2AF1 and wild-type U2AF1 are shown in red and blue, respectively.

Fig. 4. Global binding and expression differences in wild-type and mutant U2AF1.

a U2AF1 S34F globally alters binding in HCC78. Scatter plots comparing CLIP reads of individual genes bound to wild-type and mutant U2AF1. Genes that were preferentially bound (fold change >1.5) in favor of mutant (red) and wild-type (blue) are shown. b U2AF1 S34F globally alters expression in HCC78. Scatter plots comparing FPKM values of individual genes bound to wild-type and mutant U2AF1. Genes that were preferentially expressed (fold change >1.5) in favor of mutant (red) and wild-type (blue) are shown. c Comparison of mutant to wild-type ratios U2AF1 CLIP reads and FPKM values of individual genes. 87 genes for mutant and 23 genes for wild-type are shown to be differentially bound and expressed in favor of the mutant and wild-type, respectively. d Cumulative frequencies of alternative splicing events. A statistically significant threshold for alternative splicing events comparing wild-type and mutant U2AF1 (Bayes factor >5) is shown. A3′SS, alternative 3′ SSs; A5’SS, alternative 5’ splice sites; SE, skipped exons; RI, retained introns. e Alternative splicing events preferentially spliced in favor of mutant or wild-type U2AF1 at greater than 10% frequency (Bayes factor >5). ∆PSI = difference in “percentage spliced in”. f Scatter plot comparison of expected alternative splicing events and observed events as a percentage of all events. Gray area represents statistically significant region with P < 0.01. g The frequency of each trinucleotide at 3′ splice sites of alternatively skipped exons in U2AF1 S34F mutant and wild-type transduced cells based on their change in inclusion. h Fraction of skipped exon events with TAG or CAG at the 3′ splice site in wild-type and U2AF1 S34F mutant TCGA LUAD cases (n = 175 for wild-type and 125 for mutant). i Overlap of differentially skipped exon events in HCC78 (Z-score of ψ greater/less than ±1.64, P = 0.05) and TCGA U2AF1 S34F mutant LUAD cases (P = 0.05).

Fig. 5. Functional significance of U2AF1 S34F in relation to SLC34A2-ROS1 fusions.

a Gene set enrichment analysis (GSEA) plot of differential expression data. The most enriched gene set is the hallmark epithelial mesenchymal transition. b Relative protein expression of E-cadherin and fibronectin 1 (FN1) in HCC78 cells. Experiments were performed in triplicate. c Overexpression of U2AF1 S34F mutant increases invasion potential in HCC78. Wild-type and mutant U2AF1 were overexpressed in a doxycycline-inducible manner and plated for 48 h. d Schematic of SLC34A2-ROS1 splicing in HCC78. Exon 4 of SLC34A2 is fused to exon 32 or exon 34 of ROS1, creating long and short isoforms, respectively. e Relative expression of long:short SLC34A2-ROS1 isoform ratios in mutant and wild-type U2AF1 overexpressed cell lines as measured by RT-qPCR using isoform specific primers. The long and short isoforms in the mutant and wild-type overexpressed U2AF1 cell lines were normalized to the parental cell line, HCC78. f Relative protein expression of long to short SLC34A2-ROS1 isoforms in mutant and wild-type U2AF1 transduced cell lines as measured by Western blots, where HCC78 isogenic cells were exposed by 0.2 uµg/ml of doxicycline for 0, 1 day and 5 day to induce the U2AF1 expression. The long and short isoforms in the mutant and wild-type transduced U2AF1 cell lines were normalized to the GAPDH, with densitometry data from 5d summarized in bar plot (right). The band intensity was measured by Image J software. g The long ROS1 isoform increases invasion potential in the NIH-3T3 cell line after 48 h. The short and long SLC34A2-ROS1 isoforms were transduced in NIH-3T3 fibroblasts using lentivirus. h Sensitivity of wild-type versus S34F mutant U2AF1 transduced HCC78 cells to the ROS1 inhibitor crizotinib. Experiments were done with 6 technical replicates where cells were exposed to the drug for 24 h and cell survival rate was measured by manufacturer’s standard protocol (CellTiter-Glo® Luminescent Cell Viability Assay, Promega). i Proposed mechanism of U2AF1 S34F in LUADs with ROS1 fusions. U2AF1 S34F alters short and long SLC34A2-ROS1 isoform ratios. Error bars are used to illustrate the standard deviations around the means.