eIF4E-independent translation is largely eIF3d-dependent

Abstract

Translation initiation is a highly regulated step needed for protein synthesis. Most cell-based mechanistic work on translation initiation has been done using non-stressed cells growing in medium with sufficient nutrients and oxygen. This has yielded our current understanding of 'canonical' translation initiation, involving recognition of the mRNA cap by eIF4E1 followed by successive recruitment of initiation factors and the ribosome. Many cells, however, such as tumor cells, are exposed to stresses such as hypoxia, low nutrients or proteotoxic stress. This leads to inactivation of mTORC1 and thereby inactivation of eIF4E1. Hence the question arises how cells translate mRNAs under such stress conditions. We study here how mRNAs are translated in an eIF4E1-independent manner by blocking eIF4E1 using a constitutively active version of eIF4E-binding protein (4E-BP). Via ribosome profiling we identify a subset of mRNAs that are still efficiently translated when eIF4E1 is inactive. We find that these mRNAs preferentially release eIF4E1 when eIF4E1 is inactive and bind instead to eIF3d via its cap-binding pocket. eIF3d then enables these mRNAs to be efficiently translated due to its cap-binding activity. In sum, our work identifies eIF3d-dependent translation as a major mechanism enabling mRNA translation in an eIF4E-independent manner.

Fig. 1. Ribosome profiling identifies mRNAs resistant to eIF4E inhibition.

A Scatter plot of log2 (fold change of Translation Efficiency 4E-BP1-4A/E.V.) versus significance. Significant candidates with log2(fold change) > 1 are red, <−1 are blue. Candidates used for reporters and for validation of the dataset are labeled with gene names. Significance was estimated with the Wald test performed by DESeq2 package, p-values adjusted for multiple comparison. B, C Validation of the ribosome profiling from A by qRT-PCR for endogenous mRNAs in polysome gradients. B Representative example of polysome gradients from HeLa overexpressing either 4E-BP1-4A or empty vector (E.V.). The “80S” and “polysome” fractions were collected as indicated. An equal amount of RLuc mRNA was spiked into each fraction to enable comparison of transcript distribution between fractions. C RNA quantification by Q-RT-PCR from 3 biological replicates. Distribution of resistant CDKN2B, RPP25, DYNHC1H1 and sensitive GLO1, RPL13A, GAPDH transcripts is depicted. D, E Validation of the ribosome profiling by BONCAT. D One representative replica of a BONCAT de novo protein synthesis assay. E Quantification of three independent biological replicates. The transcripts are sorted in order of decreasing resistance to 4E-BP-induced translation inhibition as predicted by the footprinting. All panels: data are presented as mean values, error bars=std. dev., Significance by unpaired, two-sided, t-test adjusted for multiple testing. ns=not significant.

Fig. 2. Translation of resistant transcripts is 5′-end dependent and IRES-independent.

A The 5′UTRs of mRNAs resistant to eIF4E inhibition are often sufficient to impart resistance to a luciferase reporter. Reporters carrying the 5′UTRs of the indicated candidate genes were cloned upstream of Renilla Luciferase (RLuc) and co-transfected with a Firefly Luciferase (FLuc) normalization control. Both the negative control RLuc reporter and the FLuc normalization control carry the 5′UTR of beta globin (HBB). eIF4E was inhibited by co-transfecting with 4E-BP1-4A. (E.V.=empty vector). B An mRNA-based bicistronic assay reveals little or no IRES activity in the 5′UTRs of resistant mRNAs. Only the positive control EMCV IRES and the 5′UTR of CDKN1B showed IRES activity significantly above background. C Most reporters carrying the 5′UTRs of resistant mRNAs are translated in a 5′-end dependent manner, assayed by introducing a stable stem-loop at their 5′-end. The stable stem-loops blunts translation for the vast majority of reporters. The EMCV IRES serves as a positive control for a 5′-end independent reporter. All panels: n = 3 biological replicates. Data are presented as mean values, error bars=std. dev., significance by unpaired, two-sided, t-test adjusted for multiple testing (A, C) or by Tukey’s multiple comparison test ANOVA (B). ns not significant.

Fig. 3. Translation of resistant transcripts is cap-dependent.

A, B The vast majority of translation of the resistant reporters requires cap. (A) Capped and A-capped mRNA reporters were transfected into HeLa cells and co-treated either with DMSO or TORIN for 6 h. EMCV IRES serves as positive control for cap-independent translation. B Only the only A-capped reporters from A are shown, with the DMSO condition normalized to 1. C Translation of the RPP25 reporter requires cap. Cells were transfected either with the standard RPP25 reporter, or with a RPP25 reporter driven by the U6 promoter which does not generate capped transcripts. The raw luciferase counts are shown. D Resistance of the RPP25 reporter to eIF4E inhibition requires cap. Cells were transfected either with standard HBB or RPP25 reporters which generated capped transcripts in cells, or with the uncapped RPP25 reporter illustrated in C, together with either 4E-BP1-4A or empty vector (E.V.) as a control. All panels: data are presented as mean values, error bars=std. dev.. panel A and B n = 4 biological replicates, panel C n = 7 technical replicates, panel D n = 3 biological replicates. Significance by unpaired, two-sided, t-test adjusted for multiple testing (D) or by Dunnett’s multiple comparison test ANOVA (A, B). ns not significant.

Fig. 4. Resistant mRNAs are released from eIF4E upon eIF4E inhibition.

A Schematic illustration of the experiment workflow. B Scatter plot of log2(fold change of RNA co-immunoprecipitated with eIF4E, 4E-BP1-4A (full-length)/E.V.) versus significance. Significant candidates with log2(fold change) >1 are red, <−1 are blue. The 183 ‘resistant’ mRNAs obtained from ribosome profiling in Fig. 1A are shown in yellow. Transcripts used for the validation in panel D are labeled by gene name. Significance was estimated with the Wald test performed by DESeq2 package, p-values adjusted for multiple comparison. C Resistant mRNAs are more strongly released from eIF4E upon eIF4E inhibition than average. Histogram of log2(fold change of RNA co-immunoprecipitated with eIF4E, 4E-BP1-4A(full-length)/E.V.) for resistant mRNAs (yellow) versus all mRNAs (gray). Box plots show median with lower and upper quartile. Whiskers: full data range. Significance by unpaired, two-sided, t-test, n = 10,528 genes in 2 biological replicates. D Confirmation of RNA-seq readout for eIF4E binding by Q-RT-PCR. mRNAs binding eIF4E were immunoprecipitated from control cells (E.V.), cells expressing 4E-BP1-4A (full-length) or cells treated with Torin (250 nM, 2 h) and quantified by Q-RT-PCR. All values are normalized to total cellular mRNA. error bars=std. dev., significance by Dunnett’s multiple comparison test ANOVA. ns=not significant. n = 2 biological replicates and 3 technical replicates. E Genome wide anti-correlation resistance to eIF4E inhibition and change in eIF4E binding. r = Pearson correlation coefficient.

Fig. 5. eIF4E-independent translation requires cap-mediated binding by eIF3d.

A Translation of resistant reporters is highly dependent on eIF3d. The indicated reporters were transfected into cells that had been treated with siRNAs targeting eIF3d or a negative control non-targeting siRNA. B The resistance to eIF4E inhibition is strongly blunted upon eIF3d knockdown. C Resistance to eIF4E inhibition requires the cap-binding capacity of eIF3d. The indicated reporters were transfected into control cells, eIF3d knockdown cells, or eIF3d knockdown cells reconstituted to express either wildtype or cap-binding-mutant eIF3d. D, E Increased binding of endogenous, resistant mRNAs to eIF3d upon eIF4E inhibition requires the cap-binding capacity of eIF3d. Either wildtype or cap-binding-mutant eIF3d were immunoprecipitated from cells expressing 4E-BP1-4A or empty vector (neg. control) (D) and co-IPing mRNAs were quantified by qRT-PCR (E). All panels: error bars = std. dev., significance by unpaired, two-sided, t-test adjusted for multiple testing (A, E) or by Dunnett’s multiple comparison test ANOVA (B, C). ns not significant. Panels A–C n = 3 biological replicates, panel E: n = 3 for CDKN2B and RPL13A, n = 4 for COL12A1, TXNIP, n = 5 for the rest of transcripts.

Fig. 6. The eIF3d binding element potentiates translation also in non-stressed conditions in a cap-binding-independent manner.

A eIF3d is needed for expression of the resistant reporters in non-stressed conditions independently of its cap-binding activity. The indicated reporters were transfected into control cells, eIF3d knockdown cells, or eIF3d knockdown cells reconstituted to express either wildtype or cap-binding-mutant eIF3d. B, C Multimerization of the CDKN1B functional element strongly boosts reporter expression in non-stressed conditions. Either 1, 2, 4, 6, or 8 tandem copies of the 108nt long CDKN1B functional element were cloned into the middle of the HBB 5′UTR of an RLuc reporter. B Luciferase activity was measured for plasmids either containing a CMV promoter to drive expression of the reporter (“CMV”) or lacking a CMV promoter (“delta CMV”) to control for the presence of a cryptic promoter in the CDKN1B functional element. C Luciferase activity levels were normalized to reporter mRNA levels, quantified by qRT-PCR, to confirmation it is a translational effect. D Multimerization of the CDKN1B functional element renders a reporter more dependent on eIF3d for its expression in non-stressed conditions. The same reporters as in panel B were transfected into control cells or eIF3d knockdown cells. All panels: data are the mean values, error bars = std. dev, significance by unpaired, two-sided, t-test adjusted for multiple testing (B, D) or by Dunnett’s multiple comparison test ANOVA (A, C). ns not significant. n = 3 for panels A, B, and D, n = 4 for panel C.

Fig. 7. Schematic illustration of proposed model.

The data are consistent with a model whereby a subset of mRNAs shifts upon eIF4E inhibition from binding eIF4E via their caps to binding eIF3d via the cap, and this enables them to continue being translated. This depends on the presence of an eIF3d binding site in the 5′UTR. This site also helps recruit eIF3d and potentiate mRNA translation in non-stressed cells, although in this case the cap-binding capacity of eIF3d is not needed.