DENR promotes translation reinitiation via ribosome recycling to drive expression of oncogenes including ATF4

Abstract

Translation efficiency varies considerably between different mRNAs, thereby impacting protein expression. Translation of the stress response master-regulator ATF4 increases upon stress, but the molecular mechanisms are not well understood. We discover here that translation factors DENR, MCTS1 and eIF2D are required to induce ATF4 translation upon stress by promoting translation reinitiation in the ATF4 5'UTR. We find DENR and MCTS1 are only needed for reinitiation after upstream Open Reading Frames (uORFs) containing certain penultimate codons, perhaps because DENR•MCTS1 are needed to evict only certain tRNAs from post-termination 40S ribosomes. This provides a model for how DENR and MCTS1 promote translation reinitiation. Cancer cells, which are exposed to many stresses, require ATF4 for survival and proliferation. We find a strong correlation between DENR•MCTS1 expression and ATF4 activity across cancers. Furthermore, additional oncogenes including a-Raf, c-Raf and Cdk4 have long uORFs and are translated in a DENR•MCTS1 dependent manner.

Fig. 1. DENR promotes reinitiation after short and long uORFs.

a, b Metagene profiles for 80S ribosome footprinting of control and DENR knockout (“KO1”, “KO2”) HeLa cells showing the position of the 5′end of ribosome footprints relative to a start and b stop codons. Read counts were normalized to sequencing depth. “Smooth” indicates the curve was smoothened with a 3 nt sliding window. c Observed Z-score vs. theoretical Z-score plot for the log₂(fold change in translation efficiency) in control versus DENR knockout HeLa cells of 7788 detected transcripts. 517 translationally downregulated transcripts (red). No upregulated transcripts were identified by this analysis. d Ribosome occupancy on the DROSHA transcript and 5′UTR (inset). Read counts were normalized to sequencing depth and scaled to mRNA abundance (one value for each cell line), graphs were smoothened with a sliding window of 60 nt (main plot) or 5 nt (inset). 5′UTR features: 1 amino acid uORFs (red), longer uORFs (gray). e Change in translational efficiency of groups of transcripts containing uORFs of indicated lengths. Transcripts in each group can also contain uORFs from other groups, except transcripts with 1 AA uORFs, which were excluded from all other groups. ****p < 10⁻⁶ nonparametric Kruskal–Wallis test corrected for multiple comparisons. Box middle = median, hinges = quartiles, whiskers = deciles. f Ribosome occupancy on the ATF4 transcript and 5′UTR (inset). Read counts were normalized to sequencing depth and scaled to mRNA abundance (one value for each cell line), and graphs were smoothened with a sliding window of 30 nt (main plot) or 5 nt (inset). 5′UTR features: 1 amino acid uORFs (red), longer uORFs (gray). Source data are provided as a Source Data file.

Fig. 2. DENR•MCTS1 and eIF2D promote ATF4, c-Raf, and a-Raf translation.

a DENR and MCTS1 knockout cells have defective ATF4 induction upon ER-stress. Cells were treated with 1 µg/ml tunicamycin for 16 h. Results are representative of three biological replicates. b Defect in ATF4 induction is rescued by reconstitution with DENR. Control or DENR^KO cells were transfected with either GFP or DENR expression plasmid, reseeded and treated with 1 µg/ml tunicamycin for 16 h. Results are representative of three biological replicates. c Cells lacking DENR and eIF2D show little induction of ATF4 upon stress. Control or DENR^KO HeLa cells were transfected with siRNA targeting either GFP or eIF2D mRNA. Western blot showing reduced ATF4, a-Raf, and c-Raf protein levels. Results are representative of three biological replicates. d Translation of the ATF4 5′UTR reporter is DENR dependent both during unstressed and stress conditions. Treatment for 16 h with either 1 µg/ml tunicamycin or low amino acids (2.5% of DMEM). Results are representative of three biological replicates. Three technical replicates are shown. Unpaired, two-sided, nonparametric t-test: *p < 0.05, **p < 0.005, ***p < 0.0005. p values from top to bottom: 0.39, 0.0025, 0.015, 0.00049, 0.029. e Validation of DENR-dependent 5′UTRs predicted from ribosome footprinting. Graph shows 2–4 biological replicates for each reporter. (n = 4 for Lamin B1, Lamin B1 + stuORF, and c-Raf; n = 3 for a-Raf, PIK3R1, and PIK3R2; n = 2 for MAP2K6 and DROSHA) Unpaired, two-sided, nonparametric t-test: *p < 0.05, **p < 0.005, ***p < 0.0005, ****p < 0.00005. p values from top to bottom: 0.85, 0.000041, 0.0097, 0.0038, 0.00072, 0.0043, 0.31, 0.0081. f DENR^KO and eIF2D knockdown synergistically decrease translation of the DENR-target reporters. Treatment with 1 µg/ml tunicamycin or low amino acids (2.5% of DMEM) for 16 h. Results are representative of two biological replicates. Three technical replicates are shown. Unpaired, two-sided, nonparametric t-test: *p < 0.05, **p < 0.005. p values from top to bottom: 0.0081, 0.00087, 0.032, 0.016, 0.98, 0.019, 0.00054, 0.00073, 0.54. g c-Raf and a-Raf protein levels are DENR dependent. Control or DENR^KO cells were transfected with either GFP or DENR expression plasmid. Results are representative of three biological replicates. Source data, including uncropped western blots with molecular weight marker positions, are provided in the Source Data file.

Fig. 3. Penultimate codon identity determines DENR-dependent reinitiation after longer uORFs.

a uORF1 and uORF2 of the ATF4 5′UTR are DENR-dependent elements. Cells were treated with 1 µg/ml tunicamycin for 16 h. uORFs were mutated by changing the start codon to TAC. uORFs were extended by inserting three TCC^Ser codons directly after the uORF start codon. Results are representative of three biological replicates. Three technical replicates are shown. Unpaired, two-sided, nonparametric t-test: *p < 0.05, **p < 0.005, ***p < 0.0005. p values from top to bottom: 0.00023, 0.0088, 0.31, 0.92, 0.28, 0.70, 0.00018, 0.00020, 0.15, 0.89, 0.16, 0.53, 0.00013, 0.00016, 0.0022, 0.0026. b The two long uORFs in the a-Raf 5′UTR impart DENR dependence. uORFs were mutated by changing the start codon sequence to TAC. Results are representative of three biological replicates. Three technical replicates are shown. Unpaired, two-sided nonparametric t-test: *p < 0.05, ***p < 0.0005, ****p < 0.00005. p values from top to bottom: 0.19, 0.00060, 0.0077, 0.0000053, 0.025. c The last two codons of a-Raf uORF1 determine DENR dependence. Results are representative of three biological replicates. Three technical replicates are shown. Unpaired, two-sided, nonparametric t-test: *p < 0.05, ***p < 0.0005. p values from top to bottom: 0.71, 0.00013, 0.30, 0.010, 0.0052, 0.00027, 0.00036, 0.0067, 0.13. d Introduction of GCG^Ala-Codon in a 2 amino acid uORF causes DENR-dependent reinitiation. Graph shows 3–6 biological replicates per reporter (n = 6 for TCC; n = 5 for GCG and TAT; n = 4 for CTG; n = 3 for Lamin B1 and Lamin B1 + stuORF). Unpaired, two-sided nonparametric t-test: *p < 0.05, **p < 0.005. p values from top to bottom: 0.30, 0.014, 0.30, 0.13. e DENR dependence of predicted DENR-target 5′UTRs as shown by translation luciferase reporters. Results are representative of three biological replicates. Three technical replicates are shown. Unpaired, two-sided nonparametric t-test: **p < 0.005, ****p < 0.00005. p values from top to bottom: 0.077, 0.00003, 0.00067, 0.0020, 0.48. f CDK4 protein levels are decreased by DENR^KO and rescued by reconstitution with DENR. Control or DENR^KO cells were transfected with either GFP or DENR expression plasmid, reseeded and harvested after 16 h. g DENR knockout and eIF2D knockdown decrease CDK4 protein levels. Control or DENR^KO HeLa cells were transfected with siRNA pools targeting either GFP or eIF2D mRNA. Results are representative of two biological replicates. Source data, including uncropped western blots with molecular weight marker positions, are provided in the Source Data file.

Fig. 4. DENR-dependent 40S ribosome recycling is affected by the penultimate codon transcriptome-wide.

a, b DENR facilitates 40S ribosome recycling after translation of a main ORFs (mORF) and b uORFs, as seen by an accumulation of 40S recycling intermediates on stop codons in the absence of DENR. Metagene profiles for 40S ribosome footprinting of control and DENR knockout HeLa cells showing the position of the 5′end of ribosome footprints relative to stop codons of a all protein coding ORFs or b all uORFs. Read counts were normalized to sequencing depth. c 40S ribosome recycling on the TUBA1B stop codon is not DENR-dependent. 40S (top panel) and 80S (bottom panel) ribosome occupancy on the TUBA1B transcript. Read counts were normalized to sequencing depth, and graphs were smoothened with 10 nt sliding window. 5′UTR features: 1 amino acid uORFs (red), longer uORFs (gray). d 40S ribosome recycling on the RAN mRNA is DENR-dependent. 40S (top panel) and 80S (bottom panel) ribosome occupancy on the RAN transcript. Read counts were normalized to sequencing depth, and graphs were smoothened with a sliding window of 10 nt. 5′UTR features: 1 amino acid uORFs (red), longer uORFs (gray). e Main ORFs that show an accumulation of 40S recycling intermediates in DENR^KO cells are enriched for specific codons in the penultimate position. Penultimate codon enrichments for mORFs that show 40S accumulation on the stop codon (n = 1222) relative to all detected mORF stop codons (n = 6566). Significance was assessed using binomial tests, adjusted for multiple testing. f uORFs that show an accumulation of 40S recycling intermediates in DENR^KO cells are enriched for specific codons in the penultimate position. Penultimate codon enrichments for uORFs that show 40S accumulation on the stop codon (n = 891) relative to all detected uORF stop codons (n = 9982) Statistical significance was assessed using binomial tests, adjusted for multiple testing. Source data are provided as a Source Data file.

Fig. 5. DENR/MCTS1 and eIF2D promote ATF4 activity in cancer.

a DENR is required for optimal proliferation of HeLa cells. Control and DENR^KO HeLa cells were transfected with GFP or DENR expression plasmids. Proliferation was assayed using CellTiter-Glo. Results are representative of three biological replicates. b DENR and eIF2D inhibit proliferation in an additive manner. Control and DENR^KO HeLa cells were transfected with siRNAs targeting GFP or eIF2D, proliferation was assayed using CellTiter-Glo. Results are representative of three biological replicates. c ATF4 is required for optimal proliferation of HeLa cells. ATF4 protein was depleted using four independent siRNAs targeting ATF4. Cell proliferation was assayed using CellTiter-Glo. Results are representative of three biological replicates. d DENR promotes proliferation in part through ATF4. Control and DENR^KO HeLa cells were transfected with GFP, DENR, or ATF4 expression plasmids. Proliferation was assayed using CellTiter-Glo. Results are representative of three biological replicates. e, f DENR is required for efficient induction of the ATF4 target gene ASNS. e Q-RT-PCR of ASNS and ATF4 mRNA levels normalized to Actin B mRNA (n = 3 technical replicates). f Western blot analysis of control or DENR^KO HeLa cells treated with 1 µg/ml tunicamycin for 16 h. Cells were additionally transfected with GFP or DENR expression plasmids as indicated. Quantification of ATF4 and ASNS band intensity is normalized to tubulin. Cells from the same experiment as in e. Data are presented as mean values ± SD of three technical replicates. Unpaired, two-sided, nonparametric t-test: *p < 0.05, ***p < 0.0005. p values from left to right: 0.79, 0.016, 0.000096. Results are representative of three biological replicates. g ATF4 and ASNS expression is DENR dependent in Jurkat cells. Cells were transduced with virus encoding shRNAs targeting DENR or MCTS1 and after 7 days harvested for western blotting, tunicamycin treatment at 1 µg/ml for 16 h. Results are representative of three biological replicates. h ASNS mRNA levels, as a readout for ATF4 activity, correlate to DENR + MCTS1 mRNA levels across all cancers. Correlation, using Xena, of TCGA mRNA expression data. Shown is spearman’s r and p value. n = 11060. p value <2.2 × 10⁻¹⁶. i An 11-gene transcriptional signature for ATF4 activity correlates to DENR + MCTS1 mRNA levels across all cancers. Correlation, using Xena, of TCGA mRNA expression data. Shown is spearman’s r and p value. n = 11060. p value <2.2 × 10⁻¹⁶. Source data, including uncropped western blots with molecular weight marker positions, are provided in the Source Data file.