Increased levels of eIF2A inhibit translation by sequestering 40S ribosomal subunits

Abstract eIF2A was the first eukaryotic initiator tRNA carrier discovered but its exact function has remained enigmatic. Uncharacteristic of translation initiation factors, eIF2A is reported to be non-cytosolic in multiple human cancer cell lines. Attempts to study eIF2A mechanistically have been limited by the inability to achieve high yield of soluble recombinant protein. Here, we developed a purification paradigm that yields ∼360-fold and ∼6000-fold more recombinant human eIF2A from Escherichia coli and insect cells, respectively, than previous reports. Using a mammalian in vitro translation system, we found that increased levels of recombinant human eIF2A inhibit translation of multiple reporter mRNAs, including those that are translated by cognate and near-cognate start codons, and does so prior to start codon recognition. eIF2A also inhibited translation directed by all four types of cap-independent viral IRESs, including the CrPV IGR IRES that does not require initiation factors or initiator tRNA, suggesting excess eIF2A sequesters 40S subunits. Supplementation with additional 40S subunits prevented eIF2A-mediated inhibition and pull-down assays demonstrated direct binding between recombinant eIF2A and purified 40S subunits. These data support a model that eIF2A must be kept away from the translation machinery to avoid sequestering 40S ribosomal subunits.

SUPPLEMENTARY DATA -Grove et al.  Figure S1. eIF2(α,β,γ) and eIF2A are expressed at similar levels in various human cell types and eIF2A-mediated inhibition remains with TEV protease cleavage. A) Comparison of endogenous eIF2(α, β, γ) and eIF2A protein levels in various human cell types.

Supplementary
Data was obtained from pax-db.org using the Geiger, MCP, 2012 dataset. Ppm = parts per million. It should be noted that pax-db.org also contains data sets calculated from spectral counts (these data sets have "SC" in their name) that are less confident due to the lower protein coverage. B-D) Calculation of eIF2A (B), RPS6 (C), and RPL7 (D) concentrations in RRL by Western blot. Recombinant tag-less eIF2A and purified 80S ribosomes from RRL were titrated as a standard curve; quantification showed signal was in the linear dynamic range. The intensity of the 4% RRL sample was used for eIF2Α and the 1% RRL sample was used for RPS6 and RPL7 in the line equation and then multiplied by a dilution factor of 5 or 20, respectively, to determine the concentration of each protein in 20% RRL. For RPS6 and RPL7, these concentrations were averaged since 40S and 60S ribosomal subunits are equimolar. E) SDS-PAGE and Coomassie stain of MBP-eIF2A-His6 without and with TEV protease treatment. F) Response of in vitro translation reactions programmed with nLuc mRNA supplemented with 1.68 μM mock-cleaved and 1.68 μM TEV protease-cleaved MBP-eIF2A-His6. Bars represent the mean. n=3 biological replicates. Comparisons were made using a two-tailed unpaired t-test with Welch's correction.
Supplementary Figure S2. Recombinant eIF2A inhibits translation of mRNAs with different 5ʹ UTRs and coding sequences. A-C) In vitro translation of different reporter mRNAs with a titration (0-3.12 μM) of His6-mEGFP-FLAG or MBP-eIF2A-His6-FLAG. mRNAs tested were β-globin 5ʹ UTR nLuc mRNA (A), ATF4 5ʹ UTR nLuc mRNA (B), and AUG-3XFLAG-RLuc mRNA (C). n=3 biological replicates. A non-linear regression was used to calculate the IC50 and is shown as the line with the 95% CI included as a watermark.
Supplementary Figure S3. Human cell derived eIF2A-FLAG is inhibitory and recombinant eIF2A does not alter reporter mRNA levels during translation. A) SDS-PAGE and Coomassie stain of recombinant eIF2A-FLAG expressed and purified from HEK293T cells (obtained from OriGene). 1 µg was loaded. B) Response of in vitro translation reactions programmed with nLuc reporter mRNAs in the presence of 1.2 μM insect cell derived His6-mEGFP-FLAG, 1.2 μM insect cell derived MBP-eIF2A-His6-FLAG, or 1.2 μM human cell derived eIF2A-FLAG. Bars represent the mean. n=3 biological replicates. Comparisons were made using a two-tailed unpaired t-test with Welch's correction. C) Relative levels of nLuc reporter mRNA before and after translation with Protein Storage Buffer, 1.68 μM E. coli derived His6-MBP, 1.68 μM E. coli derived MBP-eIF2A-His6, 1.68 μM insect cell derived His6-mEGFP-FLAG, or 1.68 μM insect cell derived MBP-eIF2A-His6-FLAG. Separate identical reactions were either left on ice (0°C) or translated at 30°C for 30 min. nLuc mRNA abundance for each condition was relative to 0°C samples kept on ice. Reactions were spiked with 0.2 ng control FFLuc mRNA before total RNA was extracted using TRIzol. cDNA was subsequently synthesized and analyzed by RT-qPCR. Bars represent the mean. n=3 biological replicates. Comparisons were made using a two-tailed unpaired t-test with Welch's correction.
Supplementary Figure S4. Optimization of RLuc reporter mRNAs. A) In vitro translation of nLuc reporter mRNA set 1 (caccNNNguc-3XFLAG-RLuc), set 2 (ccccNNNguc-3XFLAG-RLuc), or set 3 (ccccNNNucc-3XFLAG-RLuc) in RRL. The Kozak sequence surrounding the AUG or non-AUG start codon is perfect (set 1) or mutated (sets 2 and 3). Wei et al. has shown that imperfect Kozak sequences cause more drastic efficiencies between AUG and near-cognate start codons in RRL (49). Raw luciferase values (relative luciferase units; RLU) are reported. Bars represent the mean. n=3 biological replicates. B) Same as in A, but each reporter set is relative to the respective AUG-encoded reporter. Bars represent the mean. n=3 biological replicates. Set 3 reporters are used in Figure 2 and  Figure S4 were used. Luciferase levels are normalized to AUG-3XFLAG-RLuc. Bars represent the mean. n=3 biological replicates. Comparisons were made using a two-tailed unpaired t-test with Welch's correction. B) Response of in vitro translation reactions programmed with AUG-and non-AUG-3XFLAG-RLuc reporter mRNAs in the presence of 1.68 μM His6-mEGFP-FLAG or 1.68 μM MBP-eIF2A-His6-FLAG. Bars represent the mean. n=3 biological replicates. Comparisons were made using a two-tailed unpaired t-test with Welch's correction.
Supplementary Figure S6. Single domain mutants of eIF2A are not translationally repressive. A) AlphaFold structural prediction of full-length human eIF2A with confidence coloring, which uses the predicted local distance difference test (pLDDT). B) SDS-PAGE and Coomassie stain of recombinant MBP-eIF2A-His6 mutants. 2 μg of protein was loaded. C) Response of in vitro translation reactions programmed with nLuc mRNA in the presence of 1.68 μM of the indicated recombinant protein. Bars represent the mean. n=3 biological replicates. Comparisons were made using a two-tailed unpaired t-test with Welch's correction.
Supplementary Figure S7. eIF2A represses translation initiation before 48S initiation complex formation. Replicate gradients as shown in Figure 4. A) nLuc mRNA distribution along a 5-30% (w/v) buffered sucrose gradient. In vitro translation reactions were supplemented with 50 µM lactimidomycin (LTM) to stall 80S ribosomes before the first translocation cycle and with either 1.68 μM His6-MBP or 1.68 μM MBP-eIF2A-His6, then diluted and separated on buffered sucrose gradients. B) Same as in A, but instead supplemented with 5 mM GMPPNP to capture 48S initiation complexes at the start codon.
Supplementary Figure S8. Control experiments with PV, HCV, and CrPV IGR IRES nLuc mRNA reporters. A) Schematic of PV IRES nLuc reporters with the nLuc ORF harboring an AUG or AAA start codon. B) Comparison of in vitro translation reactions programmed with either PV IRES nLuc mRNA or PV IRES (AUG to AAA) nLuc mRNA. The AUG to AAA mutation dramatically reduced the signal produced from the reporter, providing evidence that the desired AUG start codon of nLuc is primarily being used in RRL. Bars represent the mean. n=3 biological replicates. Comparisons were made using a two-tailed unpaired t-test with Welch's correction. C-D) The ability of nLuc mRNA, HCV IRES nLuc mRNA, and CrPV IGR IRES nLuc mRNA to interact with 40S ribosomal subunits and 80S ribosomes in RRL was assessed by 5-30% (w/v) sucrose gradient ultracentrifugation. Translation reactions were assembled identically as in Figure 5B with control (no IRES) nLuc mRNA, HCV IRES nLuc mRNA, or CrPV IGR IRES nLuc mRNA in the presence of 1.68 μM His6-mEGFP-FLAG or 1.68 μM MBP-eIF2A-His6-FLAG but incubated on ice for 30 min to allow interactions between the mRNAs and translation machinery (samples were NOT incubated at 30°C as in Figure 5B). Reactions were then diluted and separated on buffered sucrose gradients, and reporter mRNA abundance was measured across the gradient as described in the Materials and Methods. Duplicates are shown side-byside. The A260 nm trace of an untranslated RRL reaction (assembled and kept on ice, NOT incubated at 30°C) is shown in C. Control (no IRES) nLuc mRNA did not co-sediment with 40S subunits. A significant proportion of the HCV IRES nLuc mRNA and CrPV IGR IRES nLuc mRNA did sediment in 40S-containing fractions, with a higher proportion of CrPV IGR IRES nLuc mRNA being found in the 40S-containing fractions than HCV IRES nLuc mRNA. These data demonstrate that these assay conditions are favorable for both IRESs to stably interact with the 40S subunit. Despite the CrPV IGR IRES nLuc mRNA being able to bind pre-formed vacant 80S ribosomes from salt-washed 40S and 60S subunits (64), little to no CrPV IGR IRES nLuc mRNA sedimented in the 80S-containing fractions. Interestingly, although inhibiting translation of the HCV IRES and CrPV IGR IRES nLuc reporter mRNAs ( Figure 5B), addition of recombinant eIF2A did not robustly prevent both IRES nLuc mRNAs from co-sedimenting with 40S subunits. Nor did eIF2A cause more IRES nLuc mRNAs to accumulate at the top of the gradient as would be expected if eIF2A was interfering with 40S subunits interacting with either IRES. To speculate, these data overall may suggest that eIF2A is preventing mRNA from being stably inserted into the mRNA channel in the 40S subunit. Unpublished eCLIP data for eIF2A from Wei et al. provides some evidence that eIF2A crosslinks to rRNA in the vicinity of the mRNA exit channel (87). The positioning/structure of eIF2A on the 40S is not empirically determined in Wei et al. when our manuscript was in preparation or in any other manuscript that we know of. Even for IRESs, which bind to the E, P, and/or A sites of the ribosome first, the mRNA must still be inserted into the mRNA channel after initiation to efficiently allow the ribosome to decode and translocate along the coding sequence. In canonical cap-and scanning-dependent translation, eIF4F bound to the mRNA recruits the 43S PIC and the mRNA is then loaded into the mRNA channel of the 40S subunit; however, this complex is not stable like IRES•40S complexes until start codon recognition occurs and the canonical 48S initiation complex is formed.