Differential Regulation of the Melanoma Proteome by eIF4A1 and eIF4E

Abstract

Small molecules and antisense oligonucleotides that inhibit the translation initiation factors eIF4A1 and eIF4E have been explored as broad-based therapeutic agents for cancer treatment, based on the frequent upregulation of these two subunits of the eIF4F cap-binding complex in many cancer cells. Here, we provide support for these therapeutic approaches with mechanistic studies of eIF4F-driven tumor progression in a preclinical model of melanoma. Silencing eIF4A1 or eIF4E decreases melanoma proliferation and invasion. There were common effects on the level of cell-cycle proteins that could explain the antiproliferative effects in vitro Using clinical specimens, we correlate the common cell-cycle targets of eIF4A1 and eIF4E with patient survival. Finally, comparative proteomic and transcriptomic analyses reveal extensive mechanistic divergence in response to eIF4A1 or eIF4E silencing. Current models indicate that eIF4A1 and eIF4E function together through the 5'UTR to increase translation of oncogenes. In contrast, our data demonstrate that the common effects of eIF4A1 and eIF4E on translation are mediated by the coding region and 3'UTR. Moreover, their divergent effects occur through the 5'UTR. Overall, our work shows that it will be important to evaluate subunit-specific inhibitors of eIF4F in different disease contexts to fully understand their anticancer actions. Cancer Res; 77(3); 613-22. ©2016 AACR.

Figure 1. eIF4A1 and eIF4E are positive regulators of melanoma proliferation and invasion

Proliferation rates (left) and invasion rates (center) with representative images (right) for the (A) WM858 MSTC treated with siRNAs against eIF4A1 (si-eIF4A1 #1), eIF4E (si-eIF4E #1) or scrambled sequence (si-Scr), (B) WM46 MSTC treated with mammalian expression vectors encoding eIF4A1 (oe-eIF4A1), eIF4E (oe-eIF4E) or empty vector (oe-Empty) and (C) A375 melanoma cell line treated with si-Scr, or one of two independent siRNAs against eIF4A1 and eIF4E. Proliferation values were normalized to day one, and plotted relative to si-Scr. Percent invasion is calculated as the number of invasive cells on a matrigel-coated filter relative to the number of migratory cells on an uncoated control filter seeded at the same density. Error bars = standard error of the mean (SEM), n≥3 replicates/treatment, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Figure 2. eIF4A1 and eIF4E commonly regulate cell cycle and oxidative phosphorylation

(A) Heatmap, hierarchically clustered by sample and gene, showing log2 fold-change relative to si-Scr for differentially expressed proteins in A375 treated with si-eIF4A1 #1 or si-eIF4E #1 (log2 fold-change ≥|1.25|, FDR<0.1). (B) Overlap between decreased (blue) or increased (red) proteins with each siRNA treatment. (C) Enriched Reactome pathways among decreased or increased proteins in the si-eIF4A1 #1 or si-eIF4E #1 treatments. In cases where multiple enriched pathways were under the same parent pathway, only the parent pathway has been shown. For complete enrichment results, see Table S1. (D) Enrichment plots for the commonly regulated Reactome categories, Cell Cycle and Tricarboxylic Acid Cycle/Eelectron Transport Chain (TCA/ETC). Black lines indicate the rank positions of category members based on average fold-change in all knockdown treatments. NMD=nonsense mediated decay, EJC=exon junction complex.

Figure 3. The eIF4F-dependent cell cycle network influences melanoma patient survival

(A) Commonly regulated proteins that are annotated to the Reactome Cell Cycle pathway mapped onto phases of the cell cycle based on their Reactome subcategory annotations. Proteins that mediate transitions or function in multiple phases are shown in the gaps. Blue font denotes decreased protein expression level. (B) The frequency of each cell cycle phase in A375 cells treated with si-Scr, si-eIF4A1 or si-eIF4E (error bars = SEM, n=3 replicates). (C) The frequency of cells expressing the apoptotic marker, annexin 5 (ANXA5), or dead cell marker, propidium iodide (PI), in A375 cells treated with si-Scr, si-eIF4A1 or si-eIF4E (error bars = SEM, n=3 replicates). (D) Kaplan-Meier survival curves generated from 328 cutaneous melanoma patients in TCGA. Patients used to generate the eIF4E curve were in the first or fourth quartiles based on normalized expression of the indicated protein (n=82 per group). Patients used to generate the CDK1/CCNB1/CCND1 curve had above-median and below-median expression of all three proteins. The less stringent cutoff was applied to obtain a roughly equivalent sample size. For (B–D), *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Figure 4. Integrated RNA and protein analysis defines direct translational targets of eIF4A1 and eIF4E

(A) Overlap in transcripts detected in A375 by RNA-seq (FPKM≥1, n=2 biological replicates per treatment) and MS3 (peptides≥1, n=3 biological replicates per treatment). (B) Average log2 protein versus average log2 mRNA fold-change for differentially expressed proteins in si-eIF4A1 #1 or si-eIF4E #1 treatments. Red and blue points denote proteins that do not have a concordant change in mRNA level (mRNA fold-change in 2 biological replicates ≤1.25 and/or FDR<0.1 for increased proteins, ≥ −1.25 and/or FDR<0.1 for decreased proteins). (C) Manually-clustered heatmap showing log2 fold-change at the mRNA and protein level for differentially expressed proteins that are annotated to the commonly enriched Cell Cycle or TCA/ETC Reactome pathways. ‘Factor-specific’ refers to proteins that are directly regulated by one factor and indirectly regulated by the other.

Figure 5. CDS and 3′UTR features predict direct regulation by eIF4A1- and eIF4E

(A) Manually-clustered heatmap of effect sizes (Cliff’s D) for thirteen mRNA features among direct targets of eIF4A1, eIF4E, or both factors. Dotted lines demarcate 5′UTR, CDS, 3′UTR and whole transcript features. p-values are shown for features where p<0.05 and effect size is greater than 20%. (B) Fold-repression of a miRNA-targeted Renilla luciferase reporter relative to a non-targeted firefly luciferase reporter. *p<0.05, **p<0.01.