Oncogenic CDK13 mutations impede nuclear RNA surveillance

Abstract

RNA surveillance pathways detect and degrade defective transcripts to ensure RNA fidelity. We found that disrupted nuclear RNA surveillance is oncogenic. Cyclin-dependent kinase 13 (CDK13) is mutated in melanoma, and patient-mutated CDK13 accelerates zebrafish melanoma. CDK13 mutation causes aberrant RNA stabilization. CDK13 is required for ZC3H14 phosphorylation, which is necessary and sufficient to promote nuclear RNA degradation. Mutant CDK13 fails to activate nuclear RNA surveillance, causing aberrant protein-coding transcripts to be stabilized and translated. Forced aberrant RNA expression accelerates melanoma in zebrafish. We found recurrent mutations in genes encoding nuclear RNA surveillance components in many malignancies, establishing nuclear RNA surveillance as a tumor-suppressive pathway. Activating nuclear RNA surveillance is crucial to avoid accumulation of aberrant RNAs and their ensuing consequences in development and disease.

Figure 1:. CDK13 has properties consistent with a tumor suppressor.

A) CDK13 melanoma mutation plot. red *=phosphorylation site. B) Patient survival plot with CDK13 mutation or downregulation vs. remaining patients. p=0.0028 Log-rank. n= patients. C) Patient kinase-domain mutations mapped on the CDK13 crystal structure. pT871=phosphorylation site. D) In vitro kinase assay of wild-type and patient-mutated CDK13 using full-length CTD₅₂ as the substrate. One-way ANOVA with no kinase vs. all conditions; WT CDK13 ****=q=0.0001, all other comparisons non-significant. Mean +SD. E) Representative photos of mitfa:BRAF^V600E; p53−/−; mitfa−/− (Triples melanoma model zebrafish) with 1) control guide RNA (gRNA), 2) cdk13 gRNA, 3) human CDK13^WT, or cdk13 gRNA and human CDK13^WT at 4 weeks post-fertilization. F) Quantification of melanocytes at 3 days post-fertilization for Triples zebrafish injected with 1) control gRNA, 2) cdk13 gRNA, 3) human CDK13^WT, or cdk13 gRNA and human CDK13^WT. p=one-way ANOVA, multiple comparisons. Mean +/− SD. *= q=0.0186, **=q=0.0030, ****=q<0.0001. (N)=zebrafish. G) Phospho Histone H3 Serine 10 (PH3) antibody staining/mm² of melanomas from Triples zebrafish injected with cdk13 gRNA compared with control gRNA. p=0.014 (Mann Whitney two-tailed t test). (N)=melanomas. H) 9-week photos of melanocyte-specific expression of EGFP, human CDK13^WT, patient-mutant CDK13, or CDK13^K734R (catalytically dead control) in Triples zebrafish. Arrows=melanomas. I) % melanoma-free survival of Triples zebrafish with melanocyte-specific expression of EGFP, CDK13^R860Q (patient mutation), and CDK13^K734R (catalytically dead). ****=p<0.0001 (log-rank). (n)=zebrafish. J) PH3 antibody staining/mm² of melanoma from Triples zebrafish expressing EGFP vs. CDK13 mutant (W878L or P893L). p=0.0125 (Mann Whitney test, two-tailed). (n)=melanomas. K) Melanoma-free survival of zebrafish with melanocyte-specific expression of CDK13^W878L with melanocyte-specific CRISPR of either a control gene or ccnT1. p=0.0098 (log-rank). (n)=zebrafish. L) Melanoma-free survival of zebrafish with melanocyte-specific CRISPR of a control gene or ccnT1 alone. ns=non-significant (log-rank). (n)=zebrafish. M) Doubling time for human melanoma cells expressing CLOVER, CDK13^WT, CDK13^W878L, or CDK13^R860Q. One-way ANOVA with CLOVER vs. all conditions. CDK13^W878L q=0.0044, CDK13^R860Q q=0.0009. ns = non-significant. Mean +/−SD. n=3 biologic replicates. All box plots with box demonstrating the 25–75 percentile range, solid line represents median, and whiskers show max to min.

Figure 2:. CDK13 mutation results in accumulation of RNAs that prematurely terminate in introns.

A) log2-fold normalized exon expression in CDK13^R860Q-(n=5) vs. EGFP-(n=4) expressing zebrafish melanomas. ****=p<2.2×10−16, first vs. last exon. B) log2-fold normalized exon expression for CDK13^mut (n=3) vs. CDK13^WT (n=5) matched control patient melanomas. ****=p<2.2×10−16, F/AF vs. L/AL exon. C) log2-fold normalized exon expression in Cdk13−/− vs. Cdk13−/−; Cdk13+ (control) mouse embryonic stem (ES) cells. 48 and 72 hours = hours of Cdk13 depletion (n=4 for each). ****=p<2.2×10−16, F/AF vs. L/AL exon. D-E) log2-fold CDK13^mut/control cleavage site utilization (3’ seq) in UTRs, introns, and exons. D) Zebrafish melanomas and E) human melanoma cells. F) Integrative Genomics Viewer plot of 3’ sequencing showing increased SUV39H1 ptRNA in CDK13^mut-expressing vs. control human melanoma cells. Red box indicates the ptRNA 3’ end. G) Models for ptRNA accumulation in CDK13^mut cells. H) RNA expression in CDK13^mut/control human melanoma cells measured by digital droplet PCR for 4 genes with increased ptRNAs by 3’ sequencing and 2 control genes. F= first exon. L = last exon. 2-way ANOVA, multiple comparisons. adjusted p-values from left to right= **** <0.0001, ***=0.0006, * = 0.0280, ***=0.0004, ns=0.8626, ns=0.9158. +/− SD. In box plots, the black horizontal line indicates the median and whiskers extend to 1.5 × the interquartile range (IQR). p-values A-C = two-sided Wilcoxon rank sum tests.

Figure 3:. Prematurely terminated RNAs accumulate post-transcriptionally and are translated - including intronic sequences.

A) Models for ptRNA accumulation as measured by Transient Transcriptome Sequencing (TT-seq). TSS = transcriptional start site. TES = transcriptional end site. B) Metagene plots are shown for TT-seq coverage over exons within non-overlapping transcripts from expressed genes (n = 7452) from CDK13^mut vs. CLOVER-expressing human melanoma cells. TSS to TES regions are shown, flanked +/− 1kb genomic sequence. C) Schematic for analysis of TT-seq and RNA-seq coverage around intronic polyadenylation sites. D) Box plots of the ratio of upstream (−300 to −1 nt) to downstream (+1 to +300 nt) read coverage at the 3’ cleavage locations as schematized in C). P-values by Wilcoxon Signed Rank Test compared to the median ratio from all samples. E) Schema of tandem mass spectrometry data analysis from CDK13^mut- vs. EGFP-expressing zebrafish melanomas. F) Log2 peptide summed signal to noise (SSN) CDK13^mut- vs. EGFP-expressing zebrafish melanomas plotted across a normalized protein length. n=3 zebrafish melanomas each condition. Black lines = 95% confidence intervals. G) Log2 CDK13^mut- vs. EGFP-expressing zebrafish melanoma peptide measurements plotted by % protein length for significantly affected proteins for Ilk (upper) and Crk (lower). Gray squares = individual measurements. Error bars = SD. H) Log2 intronic peptides SSN from CDK13^mut- vs. EGFP-expressing zebrafish melanomas as predicted from 3’ sequencing on the horizontal axis plotted by −log p-value on the vertical axis (t-test, two tailed). Labels = enriched intronic peptides.

Figure 4:. Mutant CDK13 disrupts the polyA RNA exosome.

A) CDK13 IP-MS schema. B) Heatmap of average total peptides from anti-V5 IP-MS of CDK13 WT-V5 (n=3) or CLOVER-V5 (n=2). C) Log2 fold change CDK13^WT/control total peptides by (−) log p-value (two-tailed t test). Upper right quadrant indicates proteins with a (−) p-value <0.05 and a log fold change enrichment over control >3.3. D) Graphic depicting ZC3H14 phosphorylation sites identified from ZC3H14 IPed from either CDK13^WT - or CDK13^mut-expressing human melanoma cells. E) In vitro kinase assay of wild-type CDK13/CCNT1 using ZC3H14 full-length WT and S475A substrate. p= 0.00720098 2-tailed t-test WT vs. S475A, n=3 each condition. F) Total peptides as a % of bait (ZC3H14 total peptides) for proteins with binding most regulated by CDK13 kinase activity. Mean +SD. n=biologic replicates. Gray dots = individual values. G) Box plot showing log2 fold change of siZFC3H1, ZC3H14^S475A, or ZC3H14^S475E as compared to the relevant control for ptRNAs (left), last exons (middle), or internal exons (right). Significantly changed genes (q<0.1) plotted for all RNA types in each experiment. Z1 = ZFC3H1, Z14 = ZC3H14. ptRNA = RNAs generated by usage of intronic polyadenylation sites. Internal exon = constitutive internal exon (non-alternatively spliced) isoform. In box plots, the black horizontal line indicates the median and whiskers extend to 1.5 × the interquartile range (IQR). Statistics in Table S5 (medians and two-sided Wilcoxom rank sum p values). H) IGV plot of TP53 ptRNA which is regulated by CDK13, ZFC3H1, and ZC3H14 S475 phosphorylation. Scale = normalized RNA expression (note different scales). I) Doubling time of human melanoma cells expressing CDK13^mut, ZC3H14^S475A, ZC3H14^S475E, or CDK13^mut+ZC3H14^S475A. One-way ANOVA multiple comparisons.

Figure 5:. Prematurely Terminated RNA Accumulation is Oncogenic.

A) ptRNA quantification from somatic CDK13^mut (n=14) as compared to matched CDK13^WT (n=14) cancers (many types). ****=p<2.2e⁻¹⁶ (two-sided Wilcoxon rank sum) The black horizontal line indicates the median and whiskers extend to 1.5 × the interquartile range (IQR). B) IGV RNA-seq coverage plot of PAXT-target TP53 ptRNA from sample subset from A). C) TP53 ptRNA in TP53 locus. IPA = intronic polyadenylation site. CDS = coding sequence. D-E) Triples melanoma model zebrafish with melanocyte-specific expression of EGFP vs. human TP53 ptRNA. D) 7-week photos. E) % melanoma-free survival (log-rank). (n)=zebrafish. F) SUV39H1 ptRNA in SUV39H1 locus. G-H) Triples melanoma model zebrafish with melanocyte-specific expression of EGFP vs. human SUV39H1 ptRNA. G) 7-week photos. H) % melanoma-free survival (log-rank). I-J) Lollipop plots of ZFC3H1 (I) and ZC3H18 (J) mutations in non-redundant publicly available sequencing data from all cancers.