CFIm25 links alternative polyadenylation to glioblastoma tumour suppression - PubMed (original) (raw)

. 2014 Jun 19;510(7505):412-6.

doi: 10.1038/nature13261. Epub 2014 May 11.

Affiliations

Chioniso P Masamha et al. Nature. 2014.

Erratum in

Abstract

The global shortening of messenger RNAs through alternative polyadenylation (APA) that occurs during enhanced cellular proliferation represents an important, yet poorly understood mechanism of regulated gene expression. The 3' untranslated region (UTR) truncation of growth-promoting mRNA transcripts that relieves intrinsic microRNA- and AU-rich-element-mediated repression has been observed to correlate with cellular transformation; however, the importance to tumorigenicity of RNA 3'-end-processing factors that potentially govern APA is unknown. Here we identify CFIm25 as a broad repressor of proximal poly(A) site usage that, when depleted, increases cell proliferation. Applying a regression model on standard RNA-sequencing data for novel APA events, we identified at least 1,450 genes with shortened 3' UTRs after CFIm25 knockdown, representing 11% of significantly expressed mRNAs in human cells. Marked increases in the expression of several known oncogenes, including cyclin D1, are observed as a consequence of CFIm25 depletion. Importantly, we identified a subset of CFIm25-regulated APA genes with shortened 3' UTRs in glioblastoma tumours that have reduced CFIm25 expression. Downregulation of CFIm25 expression in glioblastoma cells enhances their tumorigenic properties and increases tumour size, whereas CFIm25 overexpression reduces these properties and inhibits tumour growth. These findings identify a pivotal role of CFIm25 in governing APA and reveal a previously unknown connection between CFIm25 and glioblastoma tumorigenicity.

PubMed Disclaimer

Conflict of interest statement

The authors declare no competing interests as defined by Nature Publishing Group, or other interests that might be perceived to influence the results and discussion reported in this paper.

Figures

Extended Data Figure 1

Extended Data Figure 1. Design and optimization of the qRT-PCR assay to monitor APA of three test genes

(a) Schematic denotes the relative location of the common and distal primer annealing sites in each test gene and the approximate locations of the annotated proximal and distal poly(A) sites depicted as pPAS and dPAS, respectively. The numbers demarcate where the 3′UTR starts and ends according to ENSEMBL. (b) Ethidium stained agarose gel of RT-PCR products of equal cycle number from the different amplicons using HeLa cell mRNA. (c) Both the common and distal Cyclin D1 amplicons were cloned into the same pcDNA3 plasmid in tandem. Three dilutions of each plasmid were made and amplified individually with each amplicon in triplicate. The 2 lines on the graph depict the amplification curve for the common and distal amplicons. The expectation is that identical Ct values should be attained for each, given that the PCR reactions were conducted using identical amounts of starting material. The average of 3 individual experiments is shown for each dilution and the average Ct deviation of either amplicon at all of the dilutions was calculated as a correction factor. (d) The experiment shown in panel c was repeated for Dicer-1 and Timp-2 to determine their respective correction factors, which was then applied to experiments shown in Figure 1.

Extended Data Figure 2

Extended Data Figure 2. Summary of RNA-Seq alignment and reproducibility of PDUI and CFIm25 depletion-induced 3′UTR shortening

(a) RNA-seq read statistics of the four biologically independent experiments where HeLa cells with treated with either control siRNA (Control) or CFIm25 siRNA (CFIm25kD). Pie chart on the right represents genomic distribution of reads that were mapped to human genome hg19. The percentage was calculated by averaging all samples. CDS: coding region. (b) Histogram of gene expression of RefSeq genes with FPKM (Fragments per Kilobase of transcript sequence per million mapped paired-end reads) no less than 1. (c) Scatterplot of the two biological replicates for each condition with high Pearson correlation (r ≥ 0.9) demonstrating a high level of reproducibility between sample PDUI scores. Each dot represents the PDUI of a transcript. (d) Genome browser screen images from four independent RNA-seq experiments. Each represents an independent biological sample where HeLa cells were transfected with either the control siRNA (Con.) or an siRNA that knocked down CFIm25. Both Vma21 and SPCS3 were found to undergo 3′UTR shortening after CFIm25 knockdown while FHL1 was found not to change.

Extended Data Figure 3

Extended Data Figure 3. Shortened transcripts have more “UGUA” CFIm25 binding motifs than unaltered transcripts

(a) CFIm25 is known to bind to the UGUA motif. The number of UGUA motifs within the 3′UTRs of genes with 3′UTR shortening after CFIm25 knockdown relative to genes with unaltered 3′UTRs was calculated and compared. Here we selected the genes without 3′UTR change according to |ΔPDUI|≤0.05. (b) iCLIP tags from Martin et al (GEO accession number GSE37398) was superimposed onto data derived from PDUI analysis of CFIm25 knockdown cells. The box plot demonstrates the enrichment of CFIm25 binding within 3′UTRs that are altered after CFIm25 knockdown (P = 6.1e-107, _t_-test).

Extended Data Figure 4

Extended Data Figure 4. Gene expression changes of genes with shortened 3′UTRs

Pie graph was calculated from the list of 1,450 genes exhibiting shortened 3′UTRs due to CFIm25 knockdown. The differentially expressed gene analysis was performed using edgeR with FDR ≤ 0.05 (see Methods)

Extended Data Figure 5

Extended Data Figure 5. The Pearson correlation between gene expression fold change and the number of lost negative regulatory elements

Left panel, the number of lost AREs due to 3′UTR shortening was calculated using the ARE database and plotted against change in gene expression levels after CFIm25 knockdown. Right panel, similar to the left except the number of lost patented miRNA target sites (Targetscan 6.2) was plotted.

Extended Data Figure 6

Extended Data Figure 6. Overlap between shortening events in GBM with low CFIm25 and shortening events in HeLa cells after CFIm25 Knockdown

Left Y-axis (red) represents the percentage of shortening events in low CFIm25 GBM which are also shortened in HeLa cells following CFIm25 knockdown. Right Y-axis (blue) represents the number of shortening events in low CFIm25 GBM against different |ΔPDUI| cutoffs.

Extended Data Figure 7

Extended Data Figure 7. Overexpression of CFIm25 reduces invasion and colony formation while CFIm25 depletion increases invasion and colony formation

(a) U251 cells were transfected with either GFP or CFIm25. Top left panel, Cells were replated in soft agar and the number of colonies/clusters formed were determined. Lower left panel, Matrigel invasion assay for cells overexpressing CFIm25 or GFP. (b) Top right panel, LN229 cells were transfected with either control or two different lentiviral plasmids targeting CFIm25 (KD1 and KD2). Stably transfected cells were plated on soft agar and the resulting colonies were counted for KD1 and KD2 respectively. Lower right panel, LN229 cells were transfected with either control or two different siRNAs (KD1 and KD2) directed against CFIm25 and were replated for a matrigel invasion assay. All the experiments were done in triplicate and shown is the mean ± standard deviation. All p-values were from the 2-tailed student _t_-test of the control vs. sample. *P<0.1, **P<0.01, ***P<0.001.

Extended Data Figure 8

Extended Data Figure 8. Overexpression of CFIm25 in U251 tumors reduces their size and weight

U251 xenograft tumors were isolated from nude mice on day 84 post implantation and measured for volume (a) and weight (b) (n=10). U251-GFP represents control U251 cells expressing GFP while U251-CFIm25 represents cells transduced with a lentivirus that over-expresses CFIm25.

Extended Data Figure 9

Extended Data Figure 9. Reduction in CFIm25 expression levels enhances LN229 tumor size and weight

LN229 xenograft tumors were isolated from nude mice on day 40 post implantation and measured for volume (a) and weight (b) (n=10). LN-229-shCon. represents control lentiviral transduced cells while LN229-shCFIm25 represents cells transduced with a lentivirus that expresses shRNA targeting CFIm25.

Figure 1

Figure 1. CFIm25 depletion leads to consistent and robust 3′UTR shortening of test genes

(a-c) Western blot analysis of HeLa cell lysates treated with control siRNA (Con.) and siRNAs individually targeting each of members of the CPA machinery. In all cases, tubulin was used as a loading control. (d-f) Quantified results of three biologically independent qRT-PCR experiments on RNA isolated from cells represented in panels A-C with the factors presented in the same order as shown in Westerns A-C. See methods for quantification details.

Figure 2

Figure 2. The DaPars algorithm identifies broad targets of CFIm25 in standard RNA-seq data

(a) RNA-seq read density for 3′UTR, terminal exon and upstream exon(s) after the control (Con.) siRNA treatment and CFIm25 knockdown (CFIm25KD) in HeLa cells. (b) Diagram depicts how the differential alternative 3′UTR usage was identified based on DaPars. Y-axis is the fitting value of the DaPars regression model and the locus with minimum fitting value (red point) is the predicted alternative pPAS for the RNA-seq data (below). (c) Scatterplot of PDUIs in control and CFIm25KD where mRNAs significantly shortened (n=1,450) or lengthened (n=3) after CFIm25KD (FDR ≤0.05, |ΔPDUI| ≥ 0.2 and at least two fold-change of PDUIs between CFIm25KD and control) are colored. The shifting towards proximal PAS is significant (P < 2.2e−16, binomial test). (d) Correlation between distal PAS site usage and gene expression levels of control and CFIm25KD. The x-axis denotes |ΔPDUI|; a negative value indicates that proximal PAS is prone to be used in CFIm25KD. The y-axis denotes the logarithm of the expression level of genes from the CFIm25KD relative to the control sample. (e) Representative RNA-seq density plots along with ΔPDUI values for genes whose 3′UTR is shortened in response to CFIm25KD. (f) Representative RNA-seq density plots along with ΔPDUI values of genes whose 3′UTR is unchanged by CFIm25KD.

Figure 3

Figure 3. Increased pPAS usage after CFIm25 depletion results in increased protein translation and enhanced cell proliferation

(a) qRT-PCR results of select genes shown as fold change in dPAS usage after CFIm25 depletion. Experiments were performed in triplicate with standard error shown. The inset shows Western blot analysis demonstrating effective knockdown of CFIm25 using two distinct siRNAs. (b) Results of Western blot analysis of cell lysates after knockdown of CFIm25 using. (c) Growth of HeLa cells was measured after RNAi of CFIm25 compared to cells transfected with control or the siRNA to CFIImPcf11 (Unr.). Results shown are mean± s.d (_n_=3). (d) Graph representing luciferase activity from cells transfected with a luciferase reporter containing the 3′UTR of either GAPDH or of Smoc1 after being transfected with either control or CFIm25 siRNA. Data is the average of three independent experiments and error bars are their standard deviation.

Figure 4

Figure 4. Altered expression of CFIm25 modulates GBM tumor growth

(a) The global analysis of 3′UTR changes in GBM patient samples with either high or low levels of CFIm25. Scatterplot of PDUIs from both datasets using the same cutoffs as in Figure 2c. The shifting to proximal PAS in the low CFIm25 group is significant (P<2.2e-16; binomial test). (b) Representative UCSC Genome Browser images of RNA-seq data demonstrating 3′UTR shortening after CFIm25 knockdown in HeLa and in GBM patient samples having high (blue) or low CFIm25 expression (red). (c) Western blot analysis of lysates from two GBM cell lines. Note the overexpressed myc-CFIm25 also increases endogenous CFIm25 levels in U251 cells. (d) Growth comparison of U251 tumors overexpressing either GFP (control) or CFIm25 and data represents the average of 10 mice per group. Right panel is representative H/E and Ki67 staining of U251-GFP tumors (upper) or U251-CFIm25 tumors (lower). (e) Growth comparison of LN229 tumors derived from cells transduced with lentiviruses expressing a scrambled shRNA (control) or with lentiviruses expressing shRNA targeting CFIm25. The data represents the average of 10 mice per group. Right panel is representative H/E and Ki67 staining of LN229 tumors expressing shRNA targeting CFIm25 (upper) or LN229 tumors expressing scrambled shRNA (lower).

Comment in

References

    1. Elkon R, et al. E2F mediates enhanced alternative polyadenylation in proliferation. Genome Biol. 2012;13:R59. - PMC - PubMed
    1. Sandberg R, Neilson JR, Sarma A, Sharp PA, Burge CB. Proliferating cells express mRNAs with shortened 3′ untranslated regions and fewer microRNA target sites. Science. 2008;320:1643–1647. - PMC - PubMed
    1. Mayr C, Bartel DP. Widespread shortening of 3′UTRs by alternative cleavage and polyadenylation activates oncogenes in cancer cells. Cell. 2009;138:673–684. - PMC - PubMed
    1. Ji Z, Lee JY, Pan Z, Jiang B, Tian B. Progressive lengthening of 3′ untranslated regions of mRNAs by alternative polyadenylation during mouse embryonic development. Proceedings of the National Academy of Sciences. 2009;106:7028–7033. - PMC - PubMed
    1. Mangone M, et al. The landscape of C. elegans 3′UTRs. Science. 2010;329:432–435. - PMC - PubMed

Publication types

MeSH terms

Substances

Grants and funding

LinkOut - more resources