mRNA circularization by METTL3-eIF3h enhances translation and promotes oncogenesis

Abstract

N⁶-methyladenosine (m⁶A) modification of mRNA is emerging as an important regulator of gene expression that affects different developmental and biological processes, and altered m⁶A homeostasis is linked to cancer^1-5. m⁶A modification is catalysed by METTL3 and enriched in the 3' untranslated region of a large subset of mRNAs at sites close to the stop codon⁵. METTL3 can promote translation but the mechanism and relevance of this process remain unknown¹. Here we show that METTL3 enhances translation only when tethered to reporter mRNA at sites close to the stop codon, supporting a mechanism of mRNA looping for ribosome recycling and translational control. Electron microscopy reveals the topology of individual polyribosomes with single METTL3 foci in close proximity to 5' cap-binding proteins. We identify a direct physical and functional interaction between METTL3 and the eukaryotic translation initiation factor 3 subunit h (eIF3h). METTL3 promotes translation of a large subset of oncogenic mRNAs-including bromodomain-containing protein 4-that is also m⁶A-modified in human primary lung tumours. The METTL3-eIF3h interaction is required for enhanced translation, formation of densely packed polyribosomes and oncogenic transformation. METTL3 depletion inhibits tumorigenicity and sensitizes lung cancer cells to BRD4 inhibition. These findings uncover a mechanism of translation control that is based on mRNA looping and identify METTL3-eIF3h as a potential therapeutic target for patients with cancer.

Extended Data Fig. 1 |. METTL3 binding close to the stop codon enhances translation.

a, Schematic diagram of reporter plasmids containing Firefly luciferase cDNA and different positions of MS2 binding sites. b, Western blotting with indicated antibodies. Two independently performed experiments show similar results. c, qRT-PCR analysis of reporter mRNAs. Each tested reporter mRNAs were normalized to RLuc mRNAs. The FLuc:RLuc ratio for each construct with FLAG-MS2 expression was set to 1. Error bars represent mean ± SD; n = 3 biologically independent samples. d, Tethering assay to measure translation efficiency as described in (Fig. 1h). Error bars represent mean ± SD; n = 3 biologically independent samples; two-sided t-test. e, Colloidal Coomassie blue staining of recombinant protein His-FLAG-MS2, His-FLAG-MS2-METTL3, or His-FLAG-MS2-METTL3 (1-200). Two independently performed experiments show similar results. f, Ethidium bromide-stained agarose gel electrophoresis of the indicated in vitro transcribed reporter mRNAs; FLuc-MS2bs without poly (A) tail (Poly (A) -) or FLuc-MS2bs with 30nt poly (A) tail (Poly (A) +). Two independently performed experiments show similar results. g, In vitro translation of reporter mRNAs using either H1299 cell extracts or Rabbit reticulocyte lysate (RRL). The levels of in vitro-translated FLuc protein were analyzed using luciferase assays. Value of FLuc activity in the presence of His-FLAG-MS2 recombinant protein was set to 1.0. Error bars represent mean ± SD; n = 6 independent experiments. Two-sided t-test, *** denotes multiple comparison for the p-values showing P < 0.001.

Extended Data Fig. 2 |. N-terminal region of METTL3 promotes translation.

a, Schematic diagram of METTL3 deletion mutants or mutation in METTL3 catalytic domain. b, Western blotting with indicated antibodies. Two independently performed experiments show similar results. c, qRT-PCR analysis of reporter mRNAs. FLuc-MS2bs mRNA levels were normalized to RLuc mRNAs. The FLuc:RLuc ratio obtained in FLAG-MS2 (control) was set to 1. Error bars represent mean ± SD; n = 3 biologically independent samples. d, Tethering assay to measure translation efficiency of reporter mRNAs as described in (Fig. 1 h). Error bars represent mean ± SD; n = 3 biologically independent samples. Two-sided t-test, ** denotes multiple comparison for the p-values showing P < 0.01.

Extended Data Fig. 3 |. METTL3 associates with translation initiation factors.

a, Deletion mutants of METTL3 were expressed in HeLa cell. The total-cell extracts (Input) and the cap-associated protein samples were analyzed by Western blotting using the indicated antibodies. Two independently performed experiments show similar results. b, Cap-association assay with METTL3 depletion. The total-cell extracts (Input) and the cap-bound protein samples were analyzed by Western blotting using the indicated antibodies. m⁷GpppG cap analogue was used for antagonizing cap-associating proteins binding to m⁷GTP-Agarose. Two independently performed experiments show similar results. c, Same as (b) except HeLa cells were transfected with CTIF, eIF3b or eIF4GI siRNA. Two independently performed experiments show similar results. d-f, Mass spectrometry of FLAG-METTL3 interacting proteins. d, Proteins that were co-immunopurified with FLAG-METTL3 subjected to 4-12% Tris-Glycine SDS-PAGE. Colloidal Coomassie blue staining was performed. n=1 independent experiment. e, Gene ontology analysis of the identified proteins from Mass spectrometry. n=1 independent experiment. Hypergeometric distribution (one-tail) with Bonferroni adjustment was used to determine enrichment statistical significance. f, Table showing the translation involving factors identified from Mass spectrometry.

Extended Data Fig. 4 |. N-terminal region of METTL3 directly interacts with MPN domain of eIF3h.

a, EM images of polyribosome with METTL3-gold particle labeling. Red arrows indicate METTL3 with immuno-gold particle (6 nm). Three independently performed experiments show similar results. b, Counting of METTL3 with gold particle labeling in each polyribosome. c, EM images of polyribosome with METTL3 and eIF4E. Red arrows indicate METTL3 with immuno-gold particle (6 nm) and yellow arrows indicate eIF4E with immuno-gold particle (10 nm). Four independently performed experiments show similar results. d, Average distance between immuno-gold particles was measured. n = 6 biologically independent samples from at least three independent experiments. Error bars represent mean ± SD. e, Colloidal Coomassie blue staining of recombinant protein His-METTL3 or His-METTL3 1-200 amino acid fragments (1-200). Two independently performed experiments show similar results. f, Colloidal Coomassie blue staining of recombinant GST-tagged protein eIF3g, eIF3h, eIF3i, eIF3j or eIF3m. Two independently performed experiments show similar results. g, GST-eIF3h was co-purified with His-METTL3 in the presence of either rabbit IgG (rIgG) or α-METTL3 antibody. Levels of co-purified His-METTL3 were analyzed by Western blotting. Two independently performed experiments show similar results. h, Schematic diagram of human eIF3h deletion mutants. i, Colloidal Coomassie blue staining of recombinant GST-eIF3h, -eIF3h (1-222) or -eIF3h (29-222). n = 1 independent experiments. j, GST pull-down of indicated eIF3h deletion mutants. Co-purified His-METTL3 was analyzed by Western blotting. n = 1 independent experiments. k, Western blotting demonstrates efficient knockdown of eIF3h protein. Three independently performed experiments show similar results. l, qRT-PCR analysis demonstrates efficient down regulation of eIF3h mRNA. Error bars represent mean ± SD; n = 3 biologically independent samples; two-sided t-test. m, qRT-PCR analysis of reporter mRNAs. FLuc-MS2bs reporter mRNAs were normalized to RLuc mRNAs. The FLuc:RLuc ratio obtained in FLAG-MS2 was set to 1. Error bars represent mean ± SD; n = 3 biologically independent samples.

Extended Data Fig. 5 |. METTL3 has no significant effect on mRNA stability.

a, Western blotting with indicated antibodies. Three independently performed experiments show similar results. b, Gene ontology analysis of the overlapping mRNAs (n=809) in (Fig. 2d). Hypergeometric distribution (one-tail) with Bonferroni adjustment was used to determine enrichment statistical significance. c, qRT-PCR analysis using indicated primers. Error bars represent mean ± SD; n = 3 technical replicates. d-e, Half-life of endogenous mRNAs was analyzed by qRT-PCR using indicated primers. Error bars represent mean ± SEM; n = 6 independent experiments. For (d), two-sided t-test; multiple comparison for the p-values showed that there were no significant differences between the samples for all the tested mRNAs, P > 0.05.

Extended Data Fig. 6 |. Widespread role of METTL3 in oncogene translation.

a, IP of endogenous METTL3 and Western blotting analysis using indicated antibodies. Two independently performed experiments show similar results. b, Density plot reflects the distribution of changes in percent spliced In (ΔPSI) values and according p-values for alternative splicing events detected by rMATs v3.2.5 (rMATs is developed based on a hierarchical framework and likelihood-ratio test was used to detect differential splicing). Splicing events at a FDR < 5% and deltaPSI > 0.1 are considered as significant. Black dots indicate total mRNAs. Red dots (4,276 mRNAs) indicate more than 2-fold less translating mRNAs in METTL3 depleted cells. c, Western blot using indicated antibodies in control-, METTL3- or YTHDF1-knockdwon cells. Two independently performed experiments show similar results. d, qRT-PCR analysis of endogenous BRD4 mRNAs. Error bars represent mean ± SEM; n = 3 biologically independent samples. e, Annexin V/PI staining of METTL3 knockdown and control A549 cells upon JQ1 treatment that was analyzed by FACS. n = 3 independent experiments.

Extended Data Fig. 7 |. Identification of a conserved Alanine residue in the N-terminal region of METTL3 required for its interaction with eIF3h.

a, Secondary structure prediction of the N-terminal (1-200) region of METTL3 protein showing putative alpha helices (blue lines). b, Evolutionary conservation of the N-terminal (1-200) region METTL3 protein. c, Computational modeling of the 3D structure of the N-terminal (77-163) region METTL3 protein, based on the coordinates of PDB: 3HHH. d, Western blotting analysis using indicated antibodies. Two independently performed experiments show similar results. e, qRT-PCR analysis of reporter mRNAs. FLuc-MS2bs mRNA levels were normalized to RLuc mRNAs. The FLuc:RLuc ratio obtained in FLAG-MS2 (control) was set to 1. Error bars represent mean ± SD; n = 6 independent experiments. f, IP of FLAG-METTL3 WT or A155P and Western blotting analysis using indicated antibodies. Two independently performed experiments show similar results. g, Staining of recombinant protein His-FLAG-MS2-METTL3 WT or His-FLAG-MS2-METTL3 A155P. Two independently performed experiments show similar results.

Extended Data Fig. 8 |. METTL3 expression correlates with lung tumor stage and promotes tumorigenicity.

a, Representative staining image of control and different stages lung cancer samples (n=75). Lower panels show the enlarged sections of the upper ones. Scale bar=30μM. b, Western blotting analysis using indicated antibodies. n = 1 independent experiments. c-d, Tumor images (c) and plot of tumor weight (d) at the endpoint in the xenograft experiment. Error bars represent mean ± SEM; n = 5 independent mice; two-sided t-test. e-h, Western blotting analysis using indicated antibodies. For (e), two independently performed experiments show similar results. For (f-h), n = 1 independent experiments. i, Tumor images at the endpoint in the xenograft experiment. Scale bar, 20mm. j, Overlapping of m⁶A containing genes identified in four lung cancer patient samples. k, Distribution of m⁶A sites.

Extended Data Fig. 9 |. Polysome conformation is affected by METTL3 and m⁶A modification in primary human lung tumors.

a, EM images of polyribosomes. Images were taken from the samples in (Fig. 3e). Scale bar, 50 nm. Six independently performed experiments show similar results. b, Gene ontology analysis. Common methylated genes refers to the methylated genes in all four patient samples. Not methylated genes refers to the genes not methylated in any of the four patient samples. Hypergeometric distribution (one-tail) with Bonferroni adjustment was used to determine enrichment statistical significance. c, Venn Diagram showing m⁶A peak overlap between patient tumor samples and cells (H1299 and A549).

Extended Data Fig. 10 |. Expression of METTL3 and eIF3h is positively correlated in many tumor types.

a, METTL3 gene expression among TCGA tumors. Box plots display the full range of variation based on the five number summaries (minimum, first quartile, median, third quartile, and maximum). TP= primary solid tumor, NT= solid tissue normal. Two-sided Wilcoxon signed-rank test was used for statistical significance. b, eIF3h gene expression among TCGA tumors. Box plots display the full range of variation based on the five number summaries (minimum, first quartile, median, third quartile, and maximum). TP= primary solid tumor, NT= solid tissue normal. Two-sided Wilcoxon signed-rank test was used for statistical significance. c, Plot illustrating the Pearson’s correlations of expression level between METTL3 and eIF3h in eight TCGA tumors, in which both METTL3 and eIF3h are significantly changed compared with normal tissues.

Fig. 1 |. METTL3 enhances translation of target mRNAs by interacting with eIF3h.

a, Electron microscopy (EM) procedure. b, EM images of polyribosomes. Red arrows; METTL3 with immuno-gold particle (6 nm), yellow arrows; CBP80 with immuno-gold particle (10 nm). Three independently performed experiments show similar results. c-d, Far Western (FW). c, Staining of eIF3 complex. A breakdown product is denoted (ΔeIF3a). Two independently performed experiments show similar results. d, FW of purified eIF3 complex. Two independently performed experiments show similar results. e, GST-tagged eIF3 subunits and co-purified His-METTL3 or 1-200 aa analyzed by Western blotting. Two independently performed experiments show similar results. f, Proximity ligation assay (PLA). Two independently performed experiments show similar results. g, Co-IPs from control or eIF3h knockdown cells. Two independently performed experiments show similar results. h, Tethering assays. Error bars = mean ± SD; n = 3 biologically independent samples, two-sided t-test. i, Model.

Fig. 2 |. METTL3 promotes translation of a large subset of mRNAs.

a, Polysome profile. Two independently performed experiments show similar results. b, Scatter plot of RNA-Seq data. Average read number from two METTL3 knockdowns is plotted. c, Scatter plot of translation efficiency (TE). Average read number from two shMETTL3 samples. d, Venn diagram showing mRNAs with >2-fold change in TE and METTL3 PAR-CLIP data. e, Features of overlapping mRNAs (n=809) from (d) was compared with all (18,115) expressed genes. Error bars = mean ± SEM; unpaired t-test (two-sided). f, qRT-PCR analysis. Error bars = mean ± SD; n = 3 technical replicates. g, Box plot represents global mRNA (12,479 mRNAs) stability profiling from two biological replicates. Two-sided Wilcoxon rank sum test; the results are statistically not significant. h, qRT-PCR analysis of METTL3-associated mRNAs using two different α-METTL3 antibodies. Error bars =mean ± SD; n = 2 independent experiments. i-l, Western blot. At least two independently performed experiments show similar results. m, MTS assay of A549 cellular proliferation upon JQ1 treatment. Error bars =mean ± SD; n = 3 independent experiments. Two-sided t-test, multiple comparison for the p-values; *** P < 0.001, ** P < 0.01. n, Quantification of (sum of early and late) apoptotic cells. Error bars +mean ± SD; n = 3 independent experiments. Two-sided t-test, multiple comparison for the p-values; * p<0.05, **p<0.01.

Fig. 3 |. METTL3-eIF3h interaction is crucial for enhanced mRNA translation and polysome conformation.

a, Co-IPs. Two independently performed experiments show similar results. b, Tethering assay. Error bars =mean ± SD; n = 6 independent experiments; two-sided t-test. c, Western blot. Two independently performed experiments show similar results. d, qRT-PCR. Error bars =mean ± SD; n = 2 independent experiments. e, In vitro translation with Rabbit reticulocyte lysate. Error bars =mean ± SD; n = 6 independent experiments; two-sided t-test. f, Analysis of 20 images from each sample in (e). g, Peak analysis of polysome profiling coupled with in vitro translation. Two independently performed experiments show similar results. h, EM images of polyribosomes. Images were taken from the samples in (g). Scale bar, 50 nm. Two independently performed experiments show similar results.

Fig. 4 |. Role of METTL3 and m⁶A in lung cancer cells and primary human tumors.

a-b, METTL3 IHC in primary lung adenocarcinoma and adjacent normal tissue. a, Error bars =mean ± SD; n = 75; two-sided Wilcoxon signed-rank test. b, . Error bars =mean ± SD; Stage I, n =37; Stage II, n =18; Stage III, n =20; two-sided Wilcoxon signed-rank test. c, Tumor-growth of xenografts from A549 cells stably expressing indicated shRNAs. Error bars =mean ± SEM; n=5 independent mice. d-e, Quantification of invasive BJ cells. Cells were transiently transfected with indicated siRNAs (d) or plasmids (e). Error bars =mean ± SEM; n = 5 for (d), n = 9 for (e) independent experiments; two-sided t-test. f, Quantification of NIH-3T3 cells colony formation. Error bars =mean ± SEM; n = 3 independent experiments; two-sided t-test; ns, not significant. g-h, Quantification of MEFs (g) or MB352 (h) cells colony formation. Error bars =mean ± SEM; n = 3 independent experiments; two-sided t-test; ns, not significant. i, Tumor weight of xenografts derived from NIH-3T3 cells stably expressing indicated proteins. There was no tumor formation in empty vector group during the observed period. Error bars =mean ± SEM; n = 8 independent mice; two-sided t-test. j-l, Global profiling of m⁶A targets in primary lung cancer samples. j, Sequence motif identified in m⁶A MeRIP-seq. k, Metagene analysis of m⁶A peaks. l, Integrative genomics viewer plots of representative m⁶A containing genes. Four lung tumors show similar results.