Structural and catalytic roles of the human 18 S rRNA methyltransferases DIMT1 in ribosome assembly and translation

,


6,6
A) methyltransferase, and results obtained with a catalytically inactive DIMT1 variant. We found that DIMT1 1/2 heterozygous HEK 293T cells have a significantly decreased 40S fraction and reduced protein synthesis but no major changes in m 2 6,6 A levels in 18S rRNA. Expression of a catalytically inactive variant, DIMT1-E85A, in WT and DIMT1 1/2 cells significantly decreased m 2 6,6 A levels in 18SrRNA,indicatingadominant-negativeeffectofthisvarianton m 2 6,6 A levels. However, expression of the DIMT1-E85A variant restored the defects in 40S levels. Of note, unlike WT DIMT1, DIMT1-E85A could not revert the defects in protein translation. We found that the differences between this variant and the WT enzyme extended to translation fidelity and gene expression patterns in DNA damage response pathways. These results suggest that the catalytic activity of DIMT1 is involved in protein translation and that the overall protein scaffold of DIMT1, regardless of the catalytic activity on m 2 6,6 A in 18S rRNA, is essential for 40S assembly.
The biological significance of the DIMT1-mediated m 2 6,6 A in 18S rRNA was previously explored by the knockout studies of DIMT1 in multiple species, including bacteria, budding yeast, and Arabidopsis thaliana. The DIMT1 knockout leads to increased sensitivity of these organisms to stress conditions (16)(17)(18). Interestingly, previous studies suggested that the expression, but not the catalytic activity, of DIMT1 is important for rRNA processing and ribosome biogenesis (19). Bacteria with catalytically dead KsgA are more sensitive to antibiotics (20); however, in yeast, the catalytically inactive DIMT1 variants do not lead to obvious growth defects compared with the wild type (WT) strain (17). Sporadic studies of human DIMT1 suggested that DIMT1 is important for the regulation of cell proliferation in cancer (21)(22)(23)(24). Elevated levels of DIMT1 correlate with the progression of multiple myeloma and colon cancer (21)(22)(23)(24); however, the molecular mechanism is not understood. The only study to investigate the mechanism by which human DIMT1 influences 18S rRNA processing showed that the catalytic activity of DIMT1 is not required for this process, leaving the function of the evolutionarily conserved rRNA modification m 2 6,6 A uncharacterized (19). Whereas the finding of a noncatalytic structural role for DIMT1 is highly credible, no evidence suggests that the catalytic role of DIMT1 is required for ribosome biogenesis or protein translation (10,(25)(26)(27)(28)(29). In addition, whereas the mechanisms regulating partial modification and the functions of specialized ribosomes are largely unknown, changes in the rRNA modification pattern have been observed in response to environmental changes, during development, and in disease. Most of the studies of rRNA modifications were carried in the bacterial and yeast systems. In human cells, the best-characterized rRNA modifications are 29-O-methylation and pseudouridylation (30)(31)(32). Dysregulation of these two rRNA modifications can affect ribosome ligand binding and translation fidelity (30,32). However, the biological functions of most of the rRNA modifications, especially in mammalian systems, remains elusive. Here, we characterize the structure and molecular functions of human DIMT1. The results reveal that residue Glu85 is important for DIMT1-mediated m 2 6,6 A installation. We studied and compared the effects of DIMT1 1/2 heterozygous cells and the catalytically inactive DIMT1 E85A variant on ribosome assembly and protein translation. Collectively, our results suggest the importance of the roles of both the noncatalytic structural scaffold and the catalytic activity of DIMT1 in these processes.

Structural characterization of WT and variant human DIMT1
To gain a molecular-level understanding of DIMT1, we determined the three-dimensional structure of DIMT1 through X-ray crystallography. We used a previously deposited but not published structure of human DIMT1 (PDB entry 1ZQ9) as the template in molecular replacement in our study. The crystal structure was finally refined to 2.38-Å resolution with 99.4% completeness. The structure of DIMT1 belongs to space group P2 1 2 1 2 1 . Data collection and refinement statistics are listed in Table S1. This structure reveals two DIMT1 copies in the asymmetric unit, whereas DIMT1 exists as a monomer in solution ( Fig. 1B and Fig. S1C). The N-terminal domain (NTD) exhibitstheSAM-dependentMTasefold,whichisreminiscentof the Rossmann fold, with a central seven-stranded b sheet flanked by three a helices on each side (Fig. 1C). The C-terminal domain (CTD), located on one side of the b sheet, is composed of five a helices, which are linked to NTD by a loop (212-219) (Fig. 1C). There is a cleft between NTD and CTD, lined with a helix D, E, and F in the NTD and G and I in the CTD. The cleft is positively charged, providing a potential platform to bind RNA (Fig. 1, D and E). The structure of KsgA (PDB entry 3FTE), a prokaryotic homolog of DIMT1, in complex with a dsRNA has already been determined (11). Although DIMT1 and KsgA have limited sequence identity (26.05%, calculated using Clustal2.1 [33]), DIMT1 and KsgA have similar overall structures with a root mean square deviation (r.m.s.d.) of 1.518 Å for 147 residues. Althoughthestructuralsimilarities oftheNTDsofDIMT1 and of KsgA are extremely close, there is an extra a helix (helix I) in the CTD of DIMT1 (Fig. 1D). This extra a helix provides multiple interactions with the RNA substrate. Moreover, sequence alignment showed that this a helix is conserved in several eukaryotic species although absent from prokaryotes (Fig. 1F). This implies that these dimethyladenosine methyltransferases have different substrate affinity or selectivity between eukaryotes and prokaryotes. The structure we determined also reveals that several positively charged side chains on both sides of the cleft in human DIMT1 are potentially important for binding of the substrate RNA. We observe the residues Arg162, Arg174, and Lys201 in the NTD and Arg228, Lys253, and Arg256 in the CTD of DIMT1. Of note, both Lys253 and Arg256 are in the a helix I of DIMT1 (Fig. 1E).

DIMT1 deficiency leads to decreased protein translation
We next studied whether the deficiency in DIMT1 leads to defects in 40S assembly. Thus, we attempted to generate DIMT1 KO by employing CRISPR/Cas9. We screened for DIMT1 2/2 homozygous knockout single colonies by using several pairs of designed PCR primers covering the CRISPR cut sites ( Fig. 2A). However, all the single colonies we obtained were heterozygous with cut and uncut DIMT1 alleles (DIMT1 1/2 ) (Fig. S2B). These resultssuggestthatDIMT1isessentialforHEK293Tcellviability. Thus, we selected two DIMT1 1/2 single colonies from the CRISPR/Cas9 screening and confirmed their genotype using Sanger sequencing experiments (Fig. S2, C and D). Western blotting results revealed that the protein levels of DIMT1 significantly decrease in the DIMT1 1/2 single colonies compared with those in the WT cells ( Fig. 2A).
Next, we quantified the m 2 6,6 A levels in 18S rRNA extracted from WT and DIMT1 1/2 heterozygous cell lines using liquid chromatography triple-quadruple MS (LC-MS/MS). Unsurprisingly, the results showed that heterozygous DIMT1 1/2 does not lead to obvious changes in the m 2 6,6 A levels in 18S rRNA ( Fig.  2B and Fig. S3), which was seen in other RNA-modifying enzyme heterozygous cells. We also performed transient knockdown of DIMT1and a knockdown control for48 h,reachinga knockdown efficiency of ;95%. However, the LC-MS/MS results showed that DIMT1 transient knockdown does not change m 2 6,6 A levels in 18S rRNA (Fig. S4, A and B). This is likely because of the long lifetime of 18S rRNA (34). Strikingly, heterozygous DIMT1 1/2 displays a significantly decreased 40S level in polysome profiling when comparing the ratio of 40S peaks to 60S peaks ( Fig. 2C and Fig. S4C), which is likely through the decreased protein level of DIMT1. We observed decreased cell proliferation in DIMT1 1/2 heterozygous compared with WT cells (Fig. 2D). These results are consistent with the fact that DIMT1 is important for 18S rRNA processing and the biogenesis of the small subunit in the ribosome (6,7,19).
To understand whether DIMT1 deficiency leads to defects in protein translation, we performed pulse-chase labeling experiments using the unnatural amino acid homopropargylglycine (HPG) for metabolic labeling of newly synthesized proteins. WT and DIMT1 1/2 cells were incubated with HPG, which was further fluorescently labeled. The results showed that DIMT1 1/2 cells have a significant decrease in global protein translation compared withtheWTcells (Fig.2E).Wealsoemployedaninvitrotranslation system to quantify the protein synthesis in WT and DIMT1 1/2 cells. In detail, cells were first lysed under hypotonic conditions to keep the integrity and activity of the translation machinery of WT and DIMT1 1/2 cells. A firefly luciferase mRNA and necessary cofactors (i.e. ATP and GTP) were added to allow in vitro translation to occur. This system can mimic the cellular translation (35). As shown in Fig. 2F, DIMT1 1/2 led to an ;55% decrease in protein translation. Furthermore, we performed in vivo protein synthesis assays by incubating live cells with puromycin to label nascent peptides and traced peptide production via Western blotting, and GADPH was used as a loading control. As shown in Fig. 2G, the results are consistent with the results from the HPG assays and the in vitro translation experiments, indicating that DIMT1 1/2 exhibits significantly decreased protein synthesis. The same experiment was carried out in DIMT1 knockdown and knockdown control cells. The results showed that DIMT1 knockdownleadstoan;45% reductionofpuromycinsignalcomparedto that of the knockdown control cells (Fig. S4D). These results collectively show that the DIMT1 deficiency significantly impacts 40S assembly and global protein translation.

DIMT1 E85A is catalytically inactive
The catalytic activity of DIMT1 was suggested to not be required for 18S rRNA processing in yeast and human cells in a previous report (15,17,19,26). To explore the catalytic role of DIMT1, we intended to first identify a catalytic-inactive DIMT1 variant. Thus, we performed protein sequence alignment ( Fig.  S4E) of DIMT1 in multiple species. Based on the sequence alignment and previous studies of DIMT1 in prokaryotes and yeast (19,36), we selected Glu85 to mutate. An additional reason Glu85 was chosen for mutational analysis is that mutation at this residuewasshown inpatients withglioblastomamultiforme (37). We cloned, expressed, and purified full-length recombinant human WT and E85A variant DIMT1 proteins (Fig. S4F). We then performed in vitro methylation assays using the DIMT1 E85A variant and WT DIMT1 with a synthetic RNA probe bearing the same local structure as the two m 2 6,6 A sites in the 18S rRNA 45 helix (Fig. 3A). The LC-MS/MS results showed that WT DIMT1 effectively installs m 2 6,6 A in this RNA probe (Fig. 3, A and B). In contrast,theE58ADIMT1variantfailedtoinstallm 2 6,6 AintheRNA probe (Fig. 3, A and B and Fig. S5). Furthermore, we determined the structure of the E58A DIMT1 variant. Glu85 in WT DIMT1 and the mutated Ala85 in DIMT1 E85A are clearly modeled in the electron density map (Fig. 3C). The overall structures of WT DIMT1 and A in18S extracted from WT and DIMT1 1/2 cells. C, polysome profiles of WT and DIMT1 1/2 cells. The ratios of 40S to 60S and 60S to 80S were determined from two biological replicates (the second set is in Fig. S4C). D, cell proliferation assays performed in the WT and DIMT1 1/2 cells. E, imaging and quantification of fluorescence-labeled HPG signals in WT and DIMT1 1/2 cells. Nuclei were stained with NuclearMask. F, in vitro translation performed using cell extract from WT and DIMT1 1/2 cells. G, Western blotting of puromycin in WT and DIMT1 1/2 cells. The error bars show the integrated signals from each Western blotting sample in the image quantifiedby ImageJ. p values were determined using a two-tailed Student's t test forunpairedsamples. Error bars represent mean 6 S.D. **, p , 0. 01; ****, p , 0.0001; n.s., not significant. DIMT1 E85A variant does not have a major impact on 18S rRNA processing and ribosome assembly but impairs protein synthesis We further studied the catalytic role of DIMT1 on 18S processing, ribosome assembly, and global protein synthesis. As shown in Fig. S7A, the E85A DIMT1 variant did not lead to noticeable changes in the ratio of 18S to 28S rRNA. The results from qRT-PCR quantification of 18S rRNA extracted from WT and E85A DIMT1 variants are consistent with the results from the agarose gel images (Fig. S7, B and C). To analyze whether the DIMT1 E85A variant impairs 40S assembly, we conducted polysome profiling using cell lysate from DIMT1 1/2 1 WT, DIMT1 1/2 1 E85A, and DIMT1 1/2 1 emptyvector cell lines.As shown in Fig. 4A and Fig. S7D, DIMT1 1/2 1 E85A variant cells did not show obvious changes in the polysome profiles compared with that of the DIMT1 1/2 1 WT cells, whereas DIMT1 1/2 1 empty vector showed a major decrease specifically in the 40S assembly. These results suggest that the protein scaffold but not the catalytic role of DIMT1 is required for ribosome assembly. We further carried out EMSA (electrophoretic mobility shift assay/gel shift assay) to compare the RNA binding affinity of WT and E85A DIMT1. As shown in Fig. S7, E and F, E85A only displays a slight decrease of affinity to the RNA probes (the same as that used in m 2 6,6 A LC-MS/MS quantification) compared to WT DIMT1. In contrast, both DIMT1 1/2 1 E85A variant cells and DIMT1 1/2 1 empty vector cells present decreased protein synthesis, whereas the DIMT1 1/2 1 WT cells showed protein synthesis comparable to that of the knockout control cells, as revealed in the HPG assays and the in vitro translation assays  . Catalytically inactive E85A DIMT1 leads to decreased 40S and impaired protein synthesis. A, polysome profiles of DIMT1 1/2 1 E85A, DIMT1 1/2 1 WT, and DIMT1 1/2 1 empty vector cells. The ratios of 40S to 60S and 60S to 80S were determined from two biological replicates (the second set is in Fig. S7D). B, imaging and quantification of fluorescence-labeled HPG signals in DIMT1 1/2 1 E85A, DIMT1 1/2 1 WT, and DIMT1 1/2 1 empty vector cells. Nuclei were stained with NuclearMask.C, invitro translationperformed usingcellextract fromDIMT1 1/2 1 E85A,DIMT1 1/2 1 WT,and DIMT1 1/2 1 emptyvector cells.Thesignalswere normalized to the knockout control cells. D, volcano plots showing gene expression in DIMT1 1/2 1 E85A and DIMT1 1/2 1 WT cells. Differentially expressed genes are shown in red and blue (adjusted p , 0.05, log 2 fold change of .0.5, or log 2 fold change of ,20.5). GO enrichment analysis of the downregulated (E) and upregulated (F) genes between DIMT1 1/2 1 E85A and DIMT1 1/2 1 WT cells. p values were determined using a two-tailed Student's t test for unpaired samples. Error bars represent mean 6 S.D. *, p , 0.05; **, p , 0.01; ****, p , 0.0001; n.s., p . 0.05. (Fig. 2, E and F, and 4, B and C). The HPG results obtained using WT cells expressing empty vector, E85A variant DIMT1, or WT DIMT1 are consistent with the results obtained from the cells with the DIMT1 1/2 genetic background (Fig. S7G). These data have lent support to the conclusion that the catalytic role of DIMT1, although not required for 40S assembly, is important for protein translation.
The catalytic efficiency of human DIMT1 is important for the expression of genes involved in DNA damage response To better understand the role of DIMT1 catalysis in gene regulation and cellular function, we performed RNA-seq with purified poly(A)-RNA from DIMT1 1/2 1 WT cells and 1E85A variant cells. The biological replicates of the RNA-seq showed strong concordance between replicates, and the RNA-seq was validatedby usingqRT-PCR (Fig.S8, A and B). No enrichment for nonspecific RNA fragments, such as rRNA fragments, was observed in the sequencing results. The RNA-seq identified 1464 codinggenesthatwereupregulated(p , 0.05andlog 2 foldchange of .0.5) and 1350 coding genes that were downregulated (p , 0.05 and log 2 fold change of ,20.5) in DIMT1 variant cell lines compared with the WT cell line (Fig. 4D). Gene ontology analysis of the upregulated genes in DIMT1 variant cell lines revealed a strong enrichment for DNA damage repair and cell cycle arrest (Fig. 4E). The downregulated genes revealed a strong enrichment for energy metabolism regulation, which is expected because DIMT1 is a ribosome assembly factor and important for ribosome functions (Fig. 4F).
Defects in 18S rRNA m 2 6,6 A impair internal ribosomal entry site (IRES)-dependent translation and translation fidelity in mammalian cells One important function of the ribosome is to faithfully maintain the translational reading frame. In this process, cisacting mRNA elements (21 PRF signals) direct translating ribosomes to slip on an mRNA by one base in the 59 direction, thereby establishing a new reading frame (30). Viral mRNA signals that abrogate this function by programming ribosomes to shift frames (programmed ribosomal frameshifting, or PRF) have proven to be of tremendous utility as readouts of translation fidelity (38). This phenomenon has been previously reported in mammalian cells, especially under ribosomal defects and disease conditions. As shown in Fig. S1, DIMT1-mediated m 2 6,6 A sites in 18S rRNA are at the interface between 40S and 60S subunits. These sites have contacts with both mRNA and tRNA on the ribosome and may be important for preventing frameshifting. Thus,tostudywhetherDIMT1-mediatedm 2 6,6 Aisimportantfor preventing frameshift, we used Dual-Luciferase reporters to quantify changes in PRF signals guided by the HIV-1 PRF. As shown in Fig. 5A, DIMT1 1/2 1 E85A variant DIMT1 promotes a  A levels in 18S rRNA (Fig. 3F). Thus, the results suggest that DIMT1-mediated m 2 6,6 A sites in 18S rRNA participate in the regulation of translation fidelity in mammalian cells, which is consistent with the previous findings that Dim1 inactivation produces fidelity defects in yeast (39).
Furthermore, it has been reported that defects in rRNA modifications affect IRES-dependent translation (30). To investigate whether m 2 6,6 A sites in 18S rRNA impact IRES association, we employed an encephalomyocarditis virus (EMCV) IRES reporter (40) to investigate whether DIMT1mediatedm 2 6,6 Ain18SrRNAplaysaroleinribosomeassociation with IRES sequences. The EMCV IRES directly binds the eIF4G subunit of the eIF4 complex and bypasses the requirement of the cap and cap-binding factor eIF4E. We transfected the EMCV IRES reporters in DIMT1 1/2 1 WT, 1 E85A variant DIMT1, or 1 empty vector cell lines. As shown in Fig. 5B, only cells expressingtheE85Avariantbutnotcellsexpressingemptyvector present decreased luciferase signals. Of note, E85A-expressing cells but not empty vector-expressing cells show decreased levels of m 2 6,6 A in 18S rRNAs. Instead of using DIMT1 1/2 cells, WT HEK 293T cells 1 E85A variant DIMT1, but not cells expressing empty vector, showed decreased luciferase signals (Fig. S9A). In addition, we obtained similar results in cells under transient knockdown of DIMT1 compared with the knockdown control (Fig. S9B). Thus, these results suggest that DIMT1-mediated 18S m 2 6,6 A can influence ribosome association with the EMCV IRES sequence.

Discussion
Previous studies have suggested the overall scaffold of DIMT1 is important for 18S processing and ribosome assembly. However, the function of DIMT1-mediated m 2 6,6 A in 18S rRNA remains elusive. In this work, we show that lack of such a catalytic activity of human DIMT1 significantly decreases protein synthesis and translation fidelity, although this catalytic role of DIMT1 is not required for ribosome small subunit assembly. We found that DIMT1 ablation impairs 40S assembly and global translation. Furthermore, we have also studied the biological significance of the overall protein scaffold of DIMT1, including both the catalytic and noncatalytic domains, in ribosome assembly and translation. We also identified and characterized a catalytically inactive DIMT1 variant E85A that has a dominantnegative effect on decreasing m 2 6,6 A levels in 18S rRNA. Of note, expressing E85A DIMT1 leads to significantly decreased 40S assembly and global protein synthesis. These results, together with those of previous studies, support a model in which DIMT1 has a structural role in 18S rRNA processing and ribosome biogenesis and a catalytic role in protein synthesis and the regulation of translation fidelity, conferring a more comprehensive picture (Fig. 6).
Because previous studies show the high prevalence and progressively increased levels of DIMT1 in myeloma and acute leukemia (22,23), small molecules inhibiting DIMT1 catalytic efficiencies should be pursued. DIMT1-mediated m 2 6,6 A dimethylation sites are in the interface between the 40S and 60S ribosomes, which makes these sites accessible to the smallmolecule inhibitors. We determined the crystal structures of WT DIMT1 that will facilitate rational inhibitor design. There is a deposited but not published structure of human DIMT1 in complex with SAH (PDB entry 1ZQ9). The ligand-free structure of human DIMT1solved in this study superimposed well with the cofactor-bound structure (Fig. S10), which suggests that binding SAH does not cause major structural changes of DIMT1. Furthermore, we solved the structure of a catalytically inactive variant E85A DIMT1, which shows no major structural changes from WT DIMT1. E85G mutation of DIMT1 was frequently seen in patients with colon adenocarcinoma (41). We predicted that the mutation at the Glu85 site (either E85A or E85G) also slightly impairs the stability of the protein using bioinformatic algorithms (42). These studies suggest the catalytic residues of DIMT1 may contribute to the pathological mechanism.
The finding of the biologic consequence of m 2 6,6 A modifications in 18S rRNA by DIMT1 is important and potentially inspiring, because there are several other similar rRNAmodifying enzymes (e.g. Emg1/EMG1 and Bud23/WBSCR22) that are suggested to function as a scaffold in preribosomal complexes, and whether their catalytic activities are not required for pre-rRNA processing is still unknown (14,28). Given that most of the mechanistic studies were carried out in prokaryotes and yeast, our understanding of the molecular functions of rRNA modifications in mammals is limited. Thus, the study of these rRNA-modifying enzymes and their functions will likely reveal previously uncharacterized regulatory mechanisms.
The rRNA species are extensively modified. The downstream effects of most of the modifications remain elusive. DIMT1mediated m 2 6,6 A dimethylation in the two adjacent sites in 18S rRNA is conserved from prokaryote to mammals. However, only the structural role of DIMT1 was thought to be important for 18S rRNA processing and ribosome assembly in yeast. One previous studycarried out inhumancells using siRNA to transiently knock down DIMT1 drew essentially the same conclusion that the catalytic role of DIMT1 is nonessential (19). The t 1/2 of rRNAs range from 3-8 days among different human cell lines (43). Notably, we constructed stable cell lines expressing DIMT1 inactive variants that have a dominant-negative effect in decreasing the level of m 2 6,6 A in 18S rRNA. Our system removes the possibility that the nondecayed 18S rRNA still functions. Our results are consistent with the previous discoveries and suggest that the catalytic-inactive DIMT1 variants do not alter 18S rRNA processing. However, the catalytic-inactive DIMT1 variants showed significant defects in 40S assembly and global protein synthesis.Furtheranalysesrevealedapreviouslyuncharacterized function of the catalytic role of DIMT1 in the regulation of IRES association, translation fidelity, and the expression of genes involved in DNA repair. The expression of the catalytically inactive E85A variant DIMT1 leads to a modest decrease of EMCV IRES signals compared with the cells expressing WT DIMT1; however, this modest change cannot fully explain the drastic defects in protein synthesis. The transcriptional change shown in the RNA-seq data and other factors may also contribute to the difference in protein synthesis between E85A-and WT DIMT1-expressing cells.
In summary, this study began to uncover both the catalytic, i.e. m 2 6,6 A in 18S rRNA, and noncatalytic biological consequences of human DIMT1. The results highlighted the significant impact of DIMT1 on ribosomal small subunit assembly and translation. It is possible that DIMT1 possesses other undiscovered activities that need to be further investigated in future studies. Ribosome assembly and translation are complex processes involving many components and regulated by many factors. The findings described in this report will open a new avenue to study these critical biological processes.

Construction, expression, and purification
Human DIMT1 gene (gene ID 27292) was cloned from a human cDNA library and ligated into pET-28a vector for protein expression, and pPB vector was used for overexpression in mammalian cells. On the basis of this WT plasmid, the E85A mutant was constructed using fusion PCR and validated by Sanger sequencing. The primers used are listed in Table S2.pET-28a plasmids were transformed into Escherichia coli BL21(DE3) for purification. Expression of all recombinant proteins was induced with 0.5 mM IPTG when the cell density reached an optical density at 600 nm between 0.6 and 0.8. After growth for ;20 h at 16°C, the cells were collected and lysed in the lysis buffer (25 mM Tris, pH 7.5, 500 mM NaCl). After centrifugation at 4°C for 30 min, the supernatant was loaded onto a Ni 21 -affinity chromatography column (GE Healthcare), which was preequilibrated using the lysis buffer. The column then was washed three times with wash buffer (50 mM imidazole in lysis buffer), and the bound target protein was eluted with elution buffer (500 mM imidazole in lysis buffer). The eluted proteins were further purified with a HiTrap SPFF (GE Healthcare) column using a gradient elution formed by low-salt buffer (25 mM Tris, pH 7.5, 50 mM NaCl)and high-saltbuffer(25mM Tris,pH7.5,1 M NaCl)and a HiLoadSuperdex 200(GEHealthcare) columnwith S200 buffer (25 mM Tris, pH 7.5, 200 mM NaCl).

Crystallization, data collection, and structure determination
Crystals of both the WT and E85A mutant DIMT1 were grown at 12 mg/ml using the hanging-drop vapor diffusion method at 290 K and grew to maximum size in ;1 day in the buffer containing0.2Mammoniumsulfate,0.1MTris-HCl,pH8.5,20% PEG 3350. Crystals were transferred to cryoprotectant solution consistingoftherespectivereservoirsolutionsupplementedwith 25% (v/v) glycerol and then flash-cooled in liquid nitrogen. Data sets for all crystals were collected at the ID-17 beamline of the National Synchrotron Light Source II (NSLS-II) of the Brookhaven National Laboratory at a wavelength of 0.920 Å and at 100 K. The data sets were processed and scaled with HKL-3000. The structures were determined by the molecular-replacement method using MOLREP (44) as implemented in the CCP4 (45) package. All of the initial models were refined using the maximum-likelihood method implemented in REFMAC5 (46) as part of the CCP4 program suite and rebuilt interactively using Coot (47). The final models were evaluated with MolProbity (48) and PROCHECK (49). Thecrystallographic parametersarelisted in Table S1. The structure figures shown in this work were prepared with PyMOL. For the structures of WT and E85A DIMT1, the resolution cutoff was decided based on the mean I/ s(I) of the highest-resolution shell of .2 and an overall R merge of ,20%.

In vitro methylation assay
The in vitro methylation assays were performed in a 30-ml reaction mixture containing the following components: 4 mg biotinylated RNA probes, 24 mg protein (WT DIMT1 or E85A mutant), 1 mM SAM, 50 mM Tris, pH 7.5, 5 mM MgCl 2 , and 1 mM DTT. The reaction mixture was incubated at 16°C overnight. After incubation, streptavidin beads (Thermo Scientific) were used to purify the RNA probes, following the instructions from the manufacturer, and eluted with RNase-free water at 75°C for 5 min. The purified RNA probes then were digested and dephosphorylated to single nucleosides using nucleoside digestion mix (NEB, M0649S) for LC-MS/MS quantification. The sequence of the RNA probe is listed in Table S2.

EMSA
The RNA probe was purchased from IDT with the sequence of 59-rUrUrCrCrGrUrArGrGrUrGrArArCrCrUrGrCrGrGrArA-39, which stemmed from human 18S rRNA sequence. The RNA probewasdissolved byRNase-freewater at4mManddilutedto40 mM in EMSA binding buffer (25 mM Tris, pH 7.2, 150 mM NaCl, and 40 U/ml RNasin). The RNA probe then was heated at 75°C for 5 min to denature it. WT and E85A DIMT1 were diluted to a concentration series of 5 mM, 10 mM, 20 mM, 40 mM, and 80 mM in EMSA assaybinding buffer. 1 ml RNA probe and 1 ml proteinwere mixed with 3 ml binding buffer, and the mixture was incubated on ice for 30 min. The entire 5-ml RNA-protein mixture was loaded to the 10% TBE gel with 1 ml loading buffer (1% bromphenol blue and 50% glycerol) and then subjected to electrophoresis for 50 min at 120 V in a 4°C cold room. The gel was stained by SYBR Gold (Invitrogen, S11494) at room temperature for 5 min. Quantification was carried out using ImageJ to quantify intensity of the bottom free RNA band. The K D (dissociation constant) was calculated with nonlinear curve fitting (function Hyperbl) of Origin 8 software with y = (P13 x)/(P21x), where y is the ratio of [RNA-protein]/[total RNA], x is the concentration of the protein, P1 is set to 1, and P2 is K D .
Quantitative analysis of the m 2 6,6 A level using LC-MS/MS For transient knockdown, cells were seeded in 6-well plate at 40% confluence, and 2 ml of 20 nM siDIMT1 or control RNA was transfected to each corresponding well using Lipofectamine RNAiMax by following the manufacturer's instructions on the next morning. After 48 h of culturing in a 37°C, 5% CO 2 incubator, cells were collected with cell lifter and the 18S RNA was extracted as indicated in the RNA isolation section below. Purified RNA or RNA probes were digested and dephosphorylated to single nucleosides using nucleoside digestion mix (NEB, M0649S) at 37°C for 1 h. The detailed procedure was as previously described (50). The nucleosides were quantified using retention time and nucleoside-to-base ion mass transitions of 268.0!136.0 (A), 284.0!152.0 (G), and 296.0!164.1 (m 2 6,6 A). All quantifications were performed by converting the peak area from the LC-MS/MS to moles using the standard curve obtained from pure nucleoside standards. The percentage ratio of m 2 6,6 A to G then was used to compare the different modification levels.

Construction of knockout cell lines using CRISPR
Two pairs of single guide RNAs (sgRNAs) for DIMT1 were designed by using a website tool. These sgRNAs were then individually cloned into a pX459 vector (51). The plasmids containing these sgRNAs were subsequently transfected to HEK 293T cells using Lipofectamine 2000 (Invitrogen). After 24 h, 2 mg/ml puromycin was administered for another 48 h. The surviving cells were cultured as single colonies into 96-well plates to screen for the positive DIMT1 monoclonal knockout. 2 weeks later, single colonies were screened by PCR. Further validation was conducted using Western blotting with an antibody from Abcam (ab184978) and Sanger sequencing. The colonies showing as WT DIMT1 in Sanger sequencing are considered control knockout cells. The sequences of sgRNAs and primers employed used in the PCR screening are listed in Table S2.

Construction of stable overexpression cell lines
FLAG-HA-tagged WT DIMT1, FLAG-HA-tagged E85A mutant, or empty overexpression vector (pPB backbone) were transfected into WT HEK 293T cells or DIMT1 1/2 heterozygous HEK 293T cells. The cells were selected under 2 mg/ml puromycin for 2 weeks. During the selection period, cells were resuspended every 2 days with fresh DMEM supplemented with 10% FBS and 2 mg/ml puromycin. After 7 days, the surviving cells were separated into single cells in a 96-well plate and subjected to puromycin selection for another 7 days. The stable overexpression was confirmed by Western blotting using an anti-FLAG antibody (Thermo, MA1-91878-HRP).

Cell proliferation assay
The cells were first trypsinized and counted using a cell counting chamber slide (Invitrogen, 100078809). 1000 cells then were seeded into every well in a 96-well plate in 100 ml cell culture medium. The next day, 20 ml of CellTiter 96 ® AQueous One solution (Promega) was added to each well at 37°C, 5% CO 2 for 2 h. Absorbance at 490 nm was then measured using a GloMax plate reader (Promega).

Polysome profiling
Cells were seeded in a 10-cm plate 1 day before at 70% confluence. Before collecting cells, cycloheximide (CHX) was added to the cell culture medium at 100 mg/ml for 7 min. The medium then was discarded, and the cells were washed once with ice-cold 13 PBS containing CHX (100 mg/ml). The cells were collected by a cell lifter with 5 ml cold PBS containing CHX (100 mg/ml). Cells were pelleted at 500 3 g for 3 min at 4°C and resuspended with 500 ml lysis buffer (10 mM Tris, pH 7.4, 150 mM KCl, 5 mM MgCl 2 , 100 mg/ml CHX, 0.5% Triton-X-100, freshly added protease inhibitor, 40 U/ml SUPERasin). After lysing on ice for 15 min, the supernatant was collected by centrifugation at 15,000 3 g for 15 min. The cell lysate was then layered on top of a linear 10-50% sucrose gradient (10 mM Tris, pH 7.4, 150 mM KCl, 5 mM MgCl 2 , 100 mg/ml CHX, 40 U/ml SUPERasin) and centrifuged at 4°C for 150 min at 35,000 rpm (Beckman, rotor SW-40Ti). The samples then were fractioned and analyzed by a Gradient Station (BioCamp) equipped with a TRIAX flow cell (BioCamp). Theratios of the40S peak to 60S peak and 80S peak to 60S peak were calculated for each polysome profile. These ratios were normalized to knockout control cells for DIMT1 1/2 cells ( Fig. 2C and Fig. S4C) or normalized to DIMT1 1/2 1 vector for DIMT1 1/2 1 wtandDIMT1 1/2 1 E85A (Fig.4AandFig.S7D)to remove batch effect for each experiment.

HPG assay
The HPG assay was performed using the Click-iT HPG Alexa Fluor 594 protein synthesis assay kits (Life Technologies, C10429) by following the manufacturer's instructions. Briefly, cells were cultured in 6-well plates, with one coverslip in each well. On the day of the experiment, the regular cell culture medium was replaced by 1 ml of L-methionine-free RPMI 1640 medium containing 1 ml Click-iT reagent for 45 min. The cells then were washed once with 13PBS and fixed with 3.6% formaldehyde in 13PBS at room temperature for 15 min. After washing twice with 3% BSA, the cells were permeabilized using 0.5% Triton X-100 at room temperature for 20 min. The Click-iT reaction was carried out at room temperature for 30 min, followed by a quenching step. The DNA then was stained by HCS NuclearMask blue stain reagent for 3 min. Finally, the coverslips were mounted with antifade reagent (Invitrogen, P36970), and the images were captured using a Leica DM6000 motorized upright microscope with the same settings for all the images (Alexa Fluor 594 exposure time for all images was 500 ms, and NuclearMask Blue exposure time for all images was 2.5 ms). For quantification, six cells were unbiasedly selected in each HPG figure to intensify quantification of the integrated area using ImageJ. The regions next to the cells without fluorescence were similarly selected and quantified as the background. The following formula then was used to calculate the corrected cell fluorescence intensity: integrated intensity -(area of selected cell 3 mean integrated intensity of backgrounds). Finally, the fluorescence intensities of DIMT1 1/2 , DIMT1 1/2 1 wt, DIMT1 1/2 1 vector, and DIMT1 1/2 1 E85A cells were all normalized to the one in knockout control cells.

In vitro translation assay
In vitrotranslationwasperformedusinga previously described protocol (35). Briefly, cells were seeded in a 10-cm plate at 70% confluence 1 day before, and cells were then trypsinized and collected by centrifugation for 5 min at 1,000 3 g at 4°C and washed once with ice-cold 13PBS. The cells then were resuspended with an equal volume of freshly made ice-cold lysis buffer [10 mM HEPES-KOH, pH 7.6, 10 mM KOAc, 0.5 mM Mg (OAc) 2 , 5 mM DTT, and 1 tablet of complete EDTA-free proteinase inhibitor mixture (Roche) per 10 ml of buffer]. After hypotonic-induced swelling for 45 min on ice, cells were homogenized by forcing the cell suspension through a 27-G needleabout10-15timesuntil;95%ofcellsburst.Thecelllysate was centrifuged at 14,000 3 g for 1 min at 4°C, and the supernatant was collected. The protein concentration in the extract was measured using a Bradford assay. An equal amount of cell lysate was used in the in vitro translation assays. Each translation reaction contained 50% cell lysate, 0.84 mM ATP, 0.21 mM GTP, 21 mM creatine phosphate (Roche), 45 U/ml creatine phosphokinase(Roche), 10 mM HEPES-KOH, pH7.6, 2 mM DTT, 2 mM Mg(OAc) 2 , 50 mM KOAc, 8 mM amino acids (Promega), 255 mM spermidine, and1 U/ml RNase inhibitor. Translation reaction mixtures were incubated for 90 min at 30°C, after which luciferaseactivitywas measuredusing a Dual-Luciferase reporter assay (Promega) with a GloMax plate reader (Promega). All measurements were normalized to the knockout control cells.

Protein quantitation and Western blotting
Protein concentration for samples was calculated using the Bradford assay (5000006, Bio-Rad). Protein samples were boiled at 95°C with Laemmli sample buffer for 10 min. After brief centrifugation, samples were loaded onto SDS-PAGE gels. After running at 180 V for 1 h, the gel was transferred to PVDF membranes by semidry transfer apparatus at 20 V for 50 min. The PVDF membranes then were blocked with 5% milk in 13 PBST for 30 min at room temperature and incubated with 3% milk in 13PBST containing the corresponding antibody overnight at 4°C. After washing three times with 13PBST, horseradish peroxidase (HRP)-conjugated secondary antibody (1:20,000) in 1% milk was applied and incubate at room temperature for 1 h. After washing three times with 13 PBST, the membrane was visualized using an ECL Western blotting detection kit (Thermo Fisher).

Protein synthesis assay
The rate of global protein synthesis was determined using puromycin to label nascent peptides as described previously (52). Briefly, cells were split into a 6-well plate 1 day before the experiment. After culturing overnight, 1 mM puromycin was added to the medium for 1 h. The cells then were washed twice with 13 PBS and collected using a cell lifter. Samples were then analyzed by SDS-PAGE followed by Western blotting using an anti-puromycin antibody (Sigma-Aldrich, MABE343). The intensity for each sample was quantified by ImageJ and normalized by the intensity of GAPDH as the loading control. The final signals were normalized to the control.

Dual-Luciferase reporter assay
For testing translation fidelity, we constructed an in-frame control, encoding a Renilla-firefly luciferase fusion protein, and a 5921-reading-frame reporter, with HIV 5921 frameshift signal (listed in Table S2) between Renilla and firefly luciferases. 5921 frameshift percentages are calculated by calculating the ratio of firefly to Renilla luciferase reads and then normalized to the inframe control.
For IRES reporter assay, EMCV IRES Dual-Luciferase reporter (30) was used. Cells were split into a 6-well plate 1 day before, and then 1 mg reporter plasmid was used to transfect cells. 24 h later, the Dual-Luciferase reporter assay system kit (Promega) was used to examine the expression of the luciferases. The relative luciferase activity was calculated by dividing Fluc by Rluc and normalized to the individual control.

RNA isolation
For total RNA extraction, TRIzol reagent (Invitrogen) was usedbyfollowingthemanufacturer'sinstructions.For18SrRNA, 5 mg total RNA was subjected to further separation on a 1.5% agarose gel. 18S rRNA in the gel slices were then extracted using a Zymoclean gel RNA recovery kit (R1011). Poly(A)-RNA was extracted from the total RNA by using a Dynabeads mRNA purification kit (Ambion) by following the manufacturer's instructions.

RNA-seq
Purified poly(A)-RNA was fragmented by sonication at 30 s on and 30 s off per cycle, for 30 cycles, using a Bioruptor Pico (Diagenode). The RNA-seq library then was constructed using a TruSeq stranded mRNA kit (Illumina). The samples were sequenced by Illumina NextSeq 550 with single-end 75-bp read length. Raw reads were mapped to the reference genome (hg38) using Hisat2 (53). Parameters used were -no-unal (not report the unaligned reads), -known-splicesite-file (generated from the UCSC hg38 annotation file), and -k 1 (report one alignment). Afterward, genomic alignments were counted in R using the GenomicAlignments package (54). Gene expression levels were normalized using DEseq2 (55). A significance threshold of adjusted p value of ,0.05 was classified as statistically significant differential expression. Gene ontology analysis was carried out with the Database for Annotation Visualization and Integrated Discovery (DAVID) tool (56,57).

Data availability
The sequence data have been deposited in the NCBI GEO database under the accession code GSE152811. All other data are available in the online Supporting Information. Atomic coordinates and structure factors for the reported crystal structures have been deposited with the Protein Data bank under accession numbers 6W6C and 6W6F. Funding and additional information-This work was supported by the National Institutes of Health (NIH) R35 to K. F. L. (GM133721). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.