MCTS2 and distinct eIF2D roles in uORF-dependent translation regulation revealed by in vitro re-initiation assays

Abstract

Ribosomes scanning from the mRNA 5' cap to the start codon may initiate at upstream open reading frames (uORFs), decreasing protein biosynthesis. Termination at a uORF can lead to re-initiation, where 40S subunits resume scanning and initiate another translation event downstream. The noncanonical translation factors MCTS1-DENR participate in re-initiation at specific uORFs, but knowledge of other trans-acting factors or uORF features influencing re-initiation is limited. Here, we establish a cell-free re-initiation assay using HeLa lysates to address this question. Comparing in vivo and in vitro re-initiation on uORF-containing reporters, we validate MCTS1-DENR-dependent re-initiation in vitro. Using this system and ribosome profiling in cells, we found that knockdown of the MCTS1-DENR homolog eIF2D causes widespread gene deregulation unrelated to uORF translation, and thus distinct to MCTS1-DENR-dependent re-initiation regulation. Additionally, we identified MCTS2, encoded by an Mcts1 retrogene, as a DENR partner promoting re-initiation in vitro, providing a plausible explanation for clinical differences associated with DENR vs. MCTS1 mutations in humans.

Keywords: DENR-MCTS1; In Vitro Translation; Re-Initiation; eIF2D; uORF.

Figure 1. DENR is a re-initiation factor that promotes the translation of hundreds of uORF-containing transcripts, including Klhdc8a and Asb8.

(A) Immunoblot analysis of NIH/3T3 cells confirms DENR and eIF2D depletion upon lentiviral transduction with Denr or Eif2d shRNAs as compared to con1 or con2 shRNAs. β-tubulin serves as a loading control. (B) Scatter plot of changes between Denr shRNA- and control shRNA-treated cells for RNA abundances vs. translation efficiencies (RPF counts/RNA counts), based on ribosome profiling and RNA-seq experiments. Evaluation is based on Denr shRNA (triplicates) and two independent control shRNAs (triplicates each). mRNAs with significant change for TE are represented in green, and mRNAs with significant change for RNA abundance are represented in olive (adjusted p values <0.10, Wald test followed by FDR adjustment; see Dataset EV1). Positions of Asb8 and Klhdc8a are indicated. (C) Violin plots comparing 5′ UTR lengths of transcripts with lower TE upon Denr depletion (n = 221, green) vs. all expressed transcripts (n = 9203, grey) or the subset of all expressed transcripts that contain at least one translated uORF (n = 3273). Denr-responsive mRNAs have longer 5′ UTRs (median = 260 nt) than overall expressed transcripts (median = 153 nt) (p = 1.12e-14, Kolmogorov–Smirnov test), even when a comparison is restricted to the overall transcripts that also contain translated uORFs (n = 3273, purple; median = 220 nt; p = 0.0036). Of note, in this analysis, we detected for Denr-responsive uORF-containing transcripts on average 3.36 uORFs per UTR, and for all uORF-containing transcripts 2.28 uORFs per UTR. (D) Analysis of the proportion of transcripts with at least one translated uORF. The transcripts with lower TE upon Denr depletion are enriched for translated uORFs. For all expressed transcripts (n = 9266), 3430 carry a translated uORF, whereas 3370 have a uORF sequence that is non-translated, and 2465 have no uORF at all. In Denr shRNA cells, transcripts with significantly reduced TE (n = 179 after removal of transcripts with ambiguous/several expressed 5′ UTRs), 156 have translated uORFs vs. 13 non-translated uORFs and 10 no uORF (p value <1e-5, Fisher’s exact test). (E) The uORF length distribution of transcripts with reduced TE in Denr shRNA-transduced cells (n = 522) is similar to that of all expressed transcripts (n = 7811) (p value = 0.37, Kolmogorov–Smirnov test). (F) Quantification of translation efficiencies on the coding sequence for Klhdc8a and Asb8 in cells treated with shRNAs targeting Denr (green, n = 3) or controls (con1 shRNA in dark grey, n = 3; con2 shRNA in light grey, n = 3); Klhdc8a: p = 0.01 and Asb8: p = 2.81e-3, two-tailed unpaired t-test. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (G) Mapped footprint A-sites of Klhdc8a and Asb8 transcripts in control and Denr knock-down cells. Read counts were normalised to library depth by subsampling, and replicates were merged for increased coverage. (H) Western Blot analysis of WT, Denr KO and Eif2d KO HeLa cells validates depletion of KLHDC8A in the absence of DENR. U2AF65 serves as loading control. (I) Schematic of the Klhdc8a 5′ UTR reporters used for the cellular and the in vitro translation assays. The ‘no uORF reporter’ serves to evaluate the regulation by the 5′ UTR in the absence of the uORF (100% signal). The ‘overlapping uORF reporter’ can only produce a luciferase signal through leaky scanning. Comparison of the WT reporter, containing the relevant uORF, with the two former reporters allows to specifically calculate re-initiation. Not depicted: all plasmids also express Renilla luciferase for internal normalisation. (J) Normalised luminescence signal (firefly/Renilla) of the Klhdc8a reporters after transduction in WT and Denr KO HeLa cells. The canonical initiation signal of the ‘no uORF reporter’ was set to 100% in WT and Denr KO cells individually. The shaded boxes represent the signal that can be calculated for uORF inhibition (grey; ‘no uORF’ signal minus ‘WT uORF’ signal), re-initiation (blue; ‘WT uORF’ signal minus ‘overlapping uORF’ signal) and leaky scanning (pink; ‘overlapping uORF’ signal) (significance calculated using two-tailed unpaired t-test). WT, no uORF: n = 3; Denr KO, no uORF: n = 5: WT, uORF: n = 7; Denr KO, uORF: n = 7; WT, overlapping uORF: n = 4; Denr KO, overlapping uORF: n = 7. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (K) Schematic representation of ribosomal fluxes on Klhdc8a 5′ UTR estimated from the results shown in panel (J). Source data are available online for this figure.

Figure 2. In vitro translation of Klhdc8a and Asb8 reporters recapitulates regulation in vivo.

(A) Outline of the preparation of HeLa translation-competent extracts and of the capped RNA reporters for the in vitro translation assays. The Renilla reporter mRNA was used throughout in vitro assays for normalisation purposes. (B) Immunoblot analysis of proteins extracted from HeLa cells, i.e. whole cell lysate, supernatant and pellet post-dual-centrifugation and translation-competent extract post-translation. MCTS1, DENR, eIF2D and RPL23 are consistently recovered in the translation-competent extracts. During the in vitro translation reaction, eIF2α-Ser51 becomes phosphorylated. (C) Immunoblot analysis of whole HeLa cell lysate and translation-competent extracts after in vitro translation reveals that the increase in eIF2α-p51 occurring during the in vitro translation reaction can be prevented by the addition of recombinant GADD34Δ1-240. β-tubulin serves as a loading control. (D) Raw luminescence signals of firefly and Renilla luciferase reporters after in vitro translation in HeLa translation-competent extracts complemented with or without GADD34Δ1-240 validates reduced translational output upon eIF2α phosphorylation. n = 17 for each condition. (E) Normalised luminescence signal (firefly/Renilla) of the Klhdc8a reporters after in vitro translation in WT and Denr KO HeLa lysates (left) and rescue assay with 0.5 µM recombinant MCTS1-DENR (right); ‘no uORF reporter’ signal was set to 100% and uORF inhibition, re-initiation and leaky scanning were calculated as indicated by the shaded boxes and as in Fig. 1J (significance between ‘WT uORF’ signal in WT HeLa lysate and ‘WT uORF’ signal in Denr KO lysate calculated using two-tailed unpaired t-test). All non-rescue conditions (left part of plot) n = 6–7; all rescue conditions (right part of plot): n = 5. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (F) Schematic representation of ribosomal fluxes on the Klhdc8a 5′ UTR reporter estimated from the results shown in panel (E). (G) As in (E), but for the Asb8 reporters. All non-rescue conditions (left side of plot): n = 6; all rescue conditions (right side of plot): n = 5. (H) As in (F) for the Asb8 5′ UTR in vitro data. (I) Normalised luminescence signal (firefly/Renilla) of the Klhdc8a reporters with mutated uORF penultimate codon (GTG to GTC or GCG) reveals the effect of penultimate codon identity on DENR-dependence. Significance was calculated using a two-tailed unpaired t-test. All wild-type constructs (left part of plot): n = 5–6. All mutant constructs (middle and right part of plot): n = 7–9. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (J) Normalised luminescence signal (firefly/Renilla) of the Asb8 reporters with an inserted codon (TGT or GCG) between the uORF start and stop codons indicates that lengthening of the uORF decreases DENR-dependence. Pairwise significances were calculated using a two-tailed unpaired t-test. All wild-type and GCG mutant conditions: n = 8; all TGT mutant conditions: n = 7. Two-way-ANOVA revealed a significant interaction between genotype and reporter length for the 1-aa vs. TGT reporter (p = 0.022), indicating that the difference between wild-type and Denr KO lysates was significantly different between the 1-aa and 2-aa uORF constructs. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. Source data are available online for this figure.

Figure 3. eIF2α phosphorylation status moderately affects re-initiation rate of start-stop uORF and longer uORF reporters.

(A) Normalised luminescence signal of Klhdc8a reporters after in vitro translation in WT and Denr KO HeLa extracts complemented with 16 ng/μl recombinant GADD34Δ1-240 (left) or without GADD34Δ1-240 (right) (significance calculated using two-tailed unpaired t-test). All ‘with GADD34’ conditions (left part of plot): n = 7–8. All ‘no GADD34’ conditions (right part of plot): n = 4–5. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (B) Schematic representation of ribosomal fluxes on Klhdc8a 5′ UTR in vitro estimated from the results shown in panel (A). Quantifications in the presence of GADD34Δ1-240 are shown in black and without GADD34Δ1-240 in yellow. (C) As in (A) for Asb8 reporters. For all conditions n = 9–10. (D) As in (B) for Asb8 reporter data shown in (C). Source data are available online for this figure.

Figure 4. No evidence for eIF2D uORF re-initiation activity in vivo but activity in vitro.

(A) Schematic representation of human MCTS1-DENR and eIF2D proteins, with domains indicated. The relevant amino acid positions defining the domains are specified. (B) Distribution of ribosomes on 5′ UTRs relative to CDS in Denr knock-down (in green) or Eif2d knock-down (in blue) vs. control cells (n = 4242). Denr knock-down cells show a strong redistribution of ribosomes from CDS to 5′ UTR, whereas the effect is milder but still significant for Eif2d knock-down cells (p values calculated using Wilcoxon signed rank test). (C) Scatter plot of changes between Eif2d shRNA and control cells for RNA abundances vs. translation efficiencies (RPF counts/RNA counts), based on ribosome profiling and RNA-seq experiments. Evaluation is based on two independent Eif2d shRNAs (triplicates each) and two independent control shRNAs (triplicates each). mRNAs with a significant change in TE were represented in blue and mRNAs with a significant change in RNA abundance were represented in olive green (adjusted p values <0.10, Wald test followed by FDR adjustment; see Dataset EV2). Positions of Asb8, Klhdc8a, Hoxa3, Med23, Rps20, Ndc80 and Cenpa are indicated. (D) Analysis of the proportion of transcripts with at least one translated uORF. The transcripts with lower TE upon Eif2d depletion are not enriched for translated uORFs. For all expressed transcripts (n = 9266), 3430 carry a translated uORF, whereas 3370 have a uORF sequence that is non-translated, and 2465 have no uORF at all. In Eif2d shRNA cells, transcripts with significantly reduced TE (n = 58 after removal of transcripts with ambiguous/several expressed 5′ UTRs), 19 have translated uORFs vs. 19 non-translated uORFs and 20 no uORF (p value = 0.59, Fisher’s exact test). (E) Quantification of translation efficiencies on the coding sequence of selected transcripts with lower TE upon Eif2d depletion in control and Eif2d-depleted cells (p values calculated using two-tailed unpaired t-test). For each of the conditions, con1 shRNA, con2 shRNA, Eif2d shRNA1, Eif2d shRNA2: n = 3. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (F) Normalised luminescence signal (firefly/Renilla) of lentivirally transduced reporters with 5′ UTRs of selected transcripts with lower TE upon Eif2d depletion (significance calculated using two-tailed unpaired t-test). Vector 5′ UTR: n = 12 for both control and Eif2d shRNA2. Cenpa, Ndc80m Rps20, Klhdc8a 5′ UTRs: n = 6 per shRNA condition. Med23 5′ UTR: n = 4 for control shRNA and n = 3 for Eif2d shRNA2. The signal of individual reporter expressed in WT cells was set to 100%. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (G) Normalised luminescence signal (firefly/Renilla) of the Asb8 reporters after in vitro translation in WT and Eif2d KO HeLa extracts (p value calculated using two-tailed unpaired t-test). n = 6 for each condition. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (H) Normalised luminescence signal (firefly/Renilla) of the Klhdc8a reporters after in vitro translation in WT and Eif2d KO HeLa extracts (left), supplemented with 0.5 µM recombinant eIF2D (middle) or supplemented with 0.5 µM of recombinant MCTS1-DENR (right); p values calculated using two-tailed unpaired t-test. For non-rescue conditions (left part of plot): n = 14 for the WT conditions and n = 9 for the Eif2d KO conditions. For eIF2D rescue conditions (middle part of plot): n = 4–7. For MCTS1-DENR rescue conditions (right part of plot): n = 5. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (I) Normalised luminescence signal (firefly/Renilla) of the Klhdc8a reporters after in vitro translation in WT and Denr KO HeLa extracts (left) or supplemented with 0.5 µM recombinant eIF2D (right); p value calculated using two-tailed unpaired t-test. For all WT conditions n = 6; for all Denr KO conditions: n = 4. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (J) Normalised luminescence signal (firefly/Renilla) of the Klhdc8a reporters after in vitro translation in WT and Denr / Eif2d double KO HeLa extracts (left), supplemented with 0.5 µM recombinant eIF2D (middle) or supplemented with 0.5 µM recombinant MCTS1-DENR (right); p values calculated using two-tailed unpaired t-test. For non-rescue conditions (left part of plot): n = 16 for WT and n = 9 for double KO conditions. For eIF2D rescue conditions (middle part of plot): n = 6–7 for WT and n = 4 for double KO conditions. For MCTS1-DENR rescue conditions (right part of plot): n = 5 for WT and n = 4 for double KO conditions. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (K) Normalised luminescence signal (firefly/Renilla) of the Klhdc8a reporters after in vitro translation in WT and Eif2d KO HeLa extracts supplemented with GADD34Δ1-240 (left), without GADD34Δ1-240 (middle), or without GADD34Δ1-240 and complemented with 0.5 µM eIF2D (right); p values calculated using two-tailed unpaired t-test. Left and middle part of plot: n = 8 for WT and n = 6 for Eif2d KO conditions. For the right part of the plot: n = 5 for all conditions. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. Source data are available online for this figure.

Figure 5. MCTS2 interacts with DENR in vivo and promotes re-initiation in vitro.

(A) Schematic representation of endogenous DENR tagged with 3×FLAG-NeonGreen-dTAG. Treatment with the dTAG-13 ligand targets DENR-3×FLAG-NeonGreen-dTAG for proteasomal degradation. (B) Genotyping PCR analysis of Denr-tagged clones validates the insertion of the 3×FLAG-NeonGreen-dTAG cassette upstream of Denr stop codon. The tagged Denr amplicon has a length of 2067 bp and a WT Denr of 885 bp. (C) Western Blot analysis of NIH/3T3 DENR-3×FLAG-NeonGreen-dTAG cells treated for 7 or 14 h with 500 nM dTAG-13 shows efficient depletion of DENR upon treatment. An asterisk indicates a non-specific band that migrates at a similar height as the tagged protein. (D) Mass spectrometry analysis of proteins co-immunoprecipitated with endogenously tagged DENR in non-treated vs. dTAG-13-treated cells reveals the interaction of MCTS2 with DENR. Only DENR and significant hits enriched in non-treated cells are shown (p values calculated using a two-tailed unpaired t-test followed by correction for multiple testing). (E) Normalised RNA abundances and ribosome footprints of Denr, Mcts1 and Mcts2 in control and DENR-depleted NIH/3T3 cells. For con1, con2 and Denr shRNA: n = 3. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (F) Alignment of human, rhesus macaque, mouse and rat MCTS1 and MCTS2 amino acid sequences. The amino acids changing relative to mouse MCTS1 are highlighted with different colours according to their side-chain chemistry. (G) Structural model of MCTS1-DENR (turquoise/green) in interaction with the 40S ribosomal subunit (grey), Met-tRNA^Met_i (magenta) and an mRNA that is shown schematically on its predicted trajectory (orange). The amino acids that undergo changes in MCTS2 are indicated in red in the structure and labelled. PDB accession numbers used for depiction: 5vyc (40S and MCTS1-DENR), 5oa3 (initiator tRNA), 6ms4 (MCTS1-interacting domain of DENR). (H) Normalised luminescence signal (firefly/Renilla) of the Klhdc8a reporters after in vitro translation in WT and Denr KO HeLa lysates (left), supplemented with 0.5 µM of recombinant DENR, MCTS1-DENR or MCTS2-DENR as indicated (p values calculated using two-tailed unpaired t-test). No rescue conditions (left): n = 6 for WT and n = 4 for Denr KO. DENR and DENR-MCTS1 rescue conditions (middle): n = 3 throughout. DENR-MCTS2 rescue conditions (right): n = 4 throughout. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. (I) As in (H), for the Asb8 reporters. No rescue conditions (left): n = 5 throughout. DENR rescue conditions: n = 6–8. DENR-MCTS1 rescue conditions: n = 5 throughout. DENR-MCTS2 rescue conditions (right): n = 5–7. The lower and upper boundaries of the boxes correspond to the first and third quartiles, while the middle line represents the median. The upper and lower whiskers extend to the largest and lowest values, limited to values that are not more distant than 1.5× the distance between the first and third quartiles. Source data are available online for this figure.