The METTL5-TRMT112 N6-methyladenosine methyltransferase complex regulates metabolism and development via translation

Ribosomal RNAs (rRNAs) have long been known to carry modifications, including numerous sites of 2’O-methylation and pseudouridylation, as well as N6-methyladenosine (m6A), and N6,6-dimethyladenosine. While the functions of many of these modifications are unclear, some are highly conserved and occur in regions of the ribosome critical for mRNA decoding. Both 28S rRNA and 18S rRNA carry m6A, and while ZCCHC4 has been identified as the methyltransferase responsible for the 28S rRNA m6A site, the methyltransferase responsible for the 18S rRNA m6A site has remained uncharacterized until recently. Here, we show that the METTL5-TRMT112 complex is the methyltransferase responsible for installing m6A at position 1832 of human 18S rRNA. TRMT112 is required for the metabolic stability of METTL5, and human METTL5 mutations associated with microcephaly and intellectual disability disrupt this interaction. Loss of METTL5 in human cancer lines alters the translation of transcripts associated with mitochondrial biogenesis and function. Mettl5 knockout mice display reduced body size and evidence of metabolic defects. This m6A site is located on the 3’ end of 18S rRNA, which may become surface-exposed under some circumstances and thus may play a regulatory role in translation of specific transcripts. While recent work has focused heavily on m6A modifications in mRNA and its roles in mRNA processing and translation, deorphanizing putative methyltransferase enzymes is revealing previously unappreciated regulatory roles for m6A in noncoding RNAs.


INTRODUCTION
Chemical modifications on RNA are a critical facet of gene expression regulation.
Historically, modifications on tRNA and rRNA have been thought to have high-stoichiometry and be relatively static, while work in the last decade on mRNA modifications suggests they are often substoichiometric and more dynamic (1). rRNA is heavily modified with numerous chemical marks, including pseudouridine, 2'O-methylation (2'OMe), N 7 -methylguanosine (m 7 G), N 1 -methyladenosine (m 1 A), N 6 -methyladenosine (m 6 A), and N 6,6 -methyladenosine (m 6,6 A) (2)(3)(4). While these modifications are thought to play critical structural roles and many of the regulatory enzymes have been identified, it is often difficult to assign specific functions to individual modifications due to the numerous interactions between the three rRNAs and 80+ protein components that form the ribosome.
While some modifications play structural roles in ribosome assembly, others may regulate translation of specific transcripts. Disruption of rRNA modification processes has been implicated in a class of developmental disorders called ribosomopathies (5)(6)(7)(8). Interestingly, though the ribosome is ubiquitously essential for translating protein, ribosomopathies often manifest as tissue-specific disorders, the molecular mechanisms of which we do not understand in many cases (9).
The METTL protein family is a class of S-adenosyl-methionine-dependent methyltransferases, with over thirty family members that methylate DNA (10), RNA (11)(12)(13)(14), and protein (15,16) substrates. Some, such as METTL3, METTL14, and METTL16, have wellcharacterized functions as RNA m 6 A methyltransferases (11,12), but others remain poorly understood. Notably, mutations in many of these enzymes, including those whose functions are poorly understood, have been implicated in human diseases such as developmental abnormalities and cancers (17)(18)(19). Revealing METTL protein substrate specificity, activity, and function are critical first steps towards understanding how mutations in these enzymes cause human disease. More specifically, mutations in METTL5 have been implicated in developmental abnormalities including microcephaly, intellectual disabilities, and attention deficit hyperactivity disorder (ADHD), but until recently very little was known about METTL5 function (18,20). We were intrigued by this connection when we came across METTL5 via proteomics experiments aimed at identifying novel proteins with methyltransferase activity.
elegans also carries out this function, but not in the context of a complex with TRMT112, as C.
elegans lack a TRMT112 homologue (25,26). Interestingly, the role of METTL5 in protein translation differs across the different model organisms studied. Consistent with these reports, we find that METTL5 forms a complex with TRMT112 to m 6 A methylate 18S rRNA in human cell lines. We further find that TRMT112 is critical for stabilizing METTL5 at the protein level, and that depletion of TRMT112 is sufficient to reduce METTL5 protein levels. Catalytically inactive METTL5 mutants can retain this critical association with TRMT112, but METTL5 mutations derived from human patients dramatically reduce this interaction. While we did not see global changes in protein translation upon METTL5 depletion, we did find evidence of dysregulated translation of specific genes. To complement our cellular studies, we generated Mettl5 knockout mice, and validated the loss of 18S rRNA m 6 A1832 in tissues from these mice. Consistent with Ignatova et al.
and Wang et al., we observe smaller body size in our Mettl5 KO mouse model (22,27). We do not observe significant behavioral changes that have been previously described, but through visual observation and RNA sequencing of mouse tissues, we do find evidence of metabolic defects that have not yet been reported. We propose that METTL5 may regulate the translation of transcripts associated with lipid metabolism, resulting in metabolic dysregulation that may play a role in the developmental phenotypes seen in human patients.

METTL5 is an m 6 A methyltransferase stabilized by TRMT112
METTL5 is a member of the METTL family of S-adenosyl-methionine-dependent methyltransferases, which is not found in yeast but conserved in higher eukaryotes ranging from C. elegans, to mice, to humans. While the activities and functions of some METTL proteins have been elucidated over the last decade (11)(12)(13)(14), many remain poorly understood. RNA methyltransferases, particularly the METTL3-METTL14 complex, have been demonstrated to have numerous cellular functions through their methyltransferase activity, which raises the question as to how less wellcharacterized METTL proteins might regulate cellular processes. METTL5 drew our attention in the course of a biochemical screen which was aimed at identifying novel RNA methyltransferase activity through biochemical fractionation followed by a liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS)-based assay (data not shown). We surveyed commonly used cancer cell culture lines by western blot and found that HeLa, HEL, and K562 cells showed relatively higher METTL5 expression ( Figure 1A). HeLa cells are the only adherent line of these three and are easily transfected, making biochemical and imaging studies straightforward. Thus, we generated METTL5 knockout (KO) HeLa lines using CRISPR-Cas9 and two sets of two guide RNAs to generate deletions across two different exons (Supp. Fig. 1A). After puromycin selection, single clones were isolated, expanded, and tested for METTL5 expression by western blot ( Figure 1B Through this process, we isolated both KO lines and clones in which METTL5 expression was unaffected (e.g., clones 4 and 6), which serve as controls throughout this work.
Based on the presence of an NPPF motif, similar to other motifs found in m 6 A RNA methyltransferases (e.g., DPPW in METTL3), we hypothesized that METTL5 may be an m 6 A methyltransferase. However, through our efforts to use in vitro methylation assays with tagged, overexpressed METTL5 protein to validate its activity, we noted lower-than-typical yields of purified METTL5 protein from both bacterial and mammalian expression systems, and low modification fractions in our in vitro assays (data not shown). Since methyltransferases such as METTL3-METTL14 function as a complex, this prompted us to perform proteomic analysis of METTL5 binding proteins using FLAG-METTL5 expressed in FreeStyle 293-F cells. TRMT112 was a particularly intriguing candidate binding protein (Supp. Fig. 1B), because it is known to bind and regulate other methyltransferases such as AlkBH8 (a tRNA methyltransferase) (28) and HemK2 (a protein methyltransferase) (29). Indeed, we found a direct interaction between METTL5 and TRMT112 ( Figure 1C) (21). Mutation of the METTL5 NPP to AAA (amino acids 127-129, abbreviated METTL5-3A), predicted to abrogate its m 6 A methyltransferase activity, slightly weakened but did not completely disrupt this interaction. Mutations in METTL5 have also been found in human patients with intellectual disability and microcephaly (18,20). We introduced two of these human variants, G61D and K191Vfs*10, into our FLAG-METTL5 construct. While FLAG-METTL5-G61D expressed at similar levels to METTL5-WT, the interaction with TRMT112 was significantly compromised, potentially explained by structural changes introduced by disrupting flexibility of a loop region ( Figure 1C, Supp. Fig. 1C). FLAG-METTL5-K191Vfs*10 expressed at much lower levels, suggesting that this truncated protein may not fold properly ( Figure 1C). Based on a previous report, the third variant, METTL5-R115Nfs*19, expresses only at very low levels, so we did not test its expression or interaction with TRMT112 (20).
The interaction with TRMT112 stabilizes METTL5 protein, as siRNA knockdown of TRMT112 reduced expression of METTL5 to approximately half, relative to negative controls ( Figure 1D), though not to the same extent as siRNA knockdown of METTL5 directly. The stabilization effect occurs at the protein level, as METTL5 transcript levels are unaffected by TRMT112 knockdown (Supp. Fig. 1D). Conversely, coexpression of FLAG-METTL5 with FLAG-TRMT112 in HeLa cells significantly increased FLAG-METTL5 protein expression relative to solely expressing FLAG-METTL5 ( Figure 1E). Coexpression of FLAG-TRMT112 with FLAG-METTL5 substantially increased protein expression yields from FreeStyle 293-F cells, allowing us to purify FLAG-METTL5-FLAG-TRMT112 complex for in vitro methyltransferase assays. Using deuterated S-adenosyl methionine (d3SAM), we performed in vitro methyltransferase assays with co-purified FLAG-METTL5-FLAG-TRMT112, using total RNA isolated from METTL5-WT and METTL5-KO HeLa cells. Analysis of deuterated, methylated nucleosides by LC-MS/MS revealed the appearance of d3m 6 A in all four samples ( Figure 1F). RNA from METTL5-KO cells accumulated approximately double the amount of d3m 6 A as RNA isolated from METTL5-WT cells, consistent with the idea that total RNA from METTL5-KO cells has less m 6 A to begin with. Parallel measurements in these same samples showed 10-fold lower levels of d3m 1 A that remained consistent across all samples. Taken together, these results suggest that METTL5 is an m 6 A RNA methyltransferase that requires TRMT112 for stability, as also corroborated by van Tran et al. (21).

18S rRNA is a major substrate of the METTL5-TRMT112 complex
To begin to understand the function of the METTL5-TRMT112 complex in cells, we first sought to identify its subcellular distribution. Fluorescence microscopy using antibodies targeting endogenous METTL5 revealed nuclear puncta that colocalized with the nucleolar protein fibrillarin, but also had a more diffuse staining pattern in the cytoplasm (Figure 2A). This distribution was also verified using biochemical fractionation, which revealed that, while there was no detectable METTL5 associated with chromatin, there is a small pool of nuclear METTL5, with the majority of the protein localized to the cytoplasm ( Figure 2B). This subcellular localization is particularly intriguing given the large difference observed in d3m 6 A methylation of total RNA isolated from WT and KO cells ( Figure 1F), suggesting a relatively large pool of RNA to be methylated by METTL5-TRMT112. The nucleolus is a critical hub for the processing and assembly of ribosome components, and rRNAs are known to be m 6 A methylated. Both 28S and 18S rRNAs have an m 6 A site (3), and while ZCCHC4 has been identified as the methyltransferase for 28S rRNA A4220 (30), the methyltransferase for 18S rRNA was unknown when we initiated these studies, and was first reported in van Tran et al. (21).
To identify RNA substrates of METTL5, we performed crosslinking-assisted immunoprecipitation (CLIP) of FLAG-tagged METTL5 followed by high throughput sequencing of bound RNAs. To maximize the chances of success, we also incorporated 4-thiouridine to facilitate crosslinking (31) (see 'Materials and Methods' for a detailed protocol). High-throughput sequencing revealed a slight enrichment for 18S rRNA transcripts in the immunoprecipitate (IP) relative to input ( Figure 2C), as well as other enriched transcripts (Supp. Fig. 2A). m 6 A-seq in METTL5-WT and -KO HeLa cells showed loss of an m 6 A peak near adenosine 1832 in 18S rRNA while the adjacent m 6,6 A1850 and m 6,6 A1851 sites remained unchanged ( Figure 2D), supporting the hypothesis that 18S rRNA is a major substrate of the METTL5-TRMT112 complex. We applied biotinylated DNA probes complementary to the region of 18S rRNA containing both m 6 A1832 and the neighboring m 6,6 A1850,1851 sites to capture and purify these fragments from METTL5-WT and -KO cells for LC-MS/MS analysis. Results confirmed that METTL5-TRMT112 methylates 18S rRNA A1832, as this fragment showed dramatically lower levels of m 6 A in all KO lines tested, relative to WT ( Figure   2E, left panel). In contrast, the nearby m 6,6 A sites showed almost no change among these cell lines ( Figure 2E, right panel), demonstrating that 18S rRNA processing is not dramatically altered, and suggesting that these modifications are regulated independently.
To verify that METTL5-TRMT112 is not only necessary, but also sufficient for deposition of 18S rRNA m 6 A1832, we expressed and purified FLAG-METTL5-WT and catalytically inactive FLAG-METTL5-3A with FLAG-TRMT112 in FreeStyle 293-F cells and tested their activity on RNA probes containing 18S or 28S rRNA sequences, using LC-MS/MS to measure d3m 6 A levels.
Neither FLAG-METTL5-WT nor FLAG-METTL5-3A complexes could efficiently methylate a 28S rRNA 12-mer ( Figure 2F), suggesting specificity for 18S rRNA sequence. Only METTL5-WT-TRMT112 complexes could effectively methylate 18S rRNA 12-mer or 60-mer probes, while the 3A mutant showed nearly undetectable activity. With this highly effective 18S rRNA 12mer substrate in hand, we performed a more detailed assessment of the individual contributions of METTL5 and TRMT112 in methyltransferase activity. While copurified FLAG-METTL5 and FLAG-TRMT112 could effectively methylate 18S rRNA probes, this activity was undetectable when either protein was purified individually, or when d3SAM was left out of the reaction ( Figure 2G). The loss of METTL5 activity in the absence of TRMT112 is likely the result of poorly folded METTL5 protein, as suggested by increased stability of METTL5 in the presence of TRMT112 ( Figures 1D, 1E). Taken together, our results are consistent with recent reports (21,22,24) that the METTL5-TRMT112 complex m 6 A methylates 18S rRNA and that the NPPF motif is critical for its catalytic activity.
To assess whether METTL5-TRMT112 may have other RNA substrates, we delved more deeply into the other transcripts enriched in our METTL5 CLIP experiment (Supp. Fig. 2A,B), which included both coding and non-coding RNAs. Cross-referencing these METTL5-bound RNAs with differentially methylated transcripts from our m 6 A-seq experiment (Supp. Fig. 2C) revealed only 25 overlapping transcripts (Supp. Fig. 2D). While the most prominent motif in the m 6 A peaks overall was GGACU, suggesting that most m 6 A peaks were METTL3/METTL14 dependent (11), the 25 transcripts in common between the two datasets showed enrichment for UAA, the motif containing the m 6 A site in 18S rRNA. Thus, though it has been reported that 18S rRNA is the only METTL5-TRMT112 substrate (21), it remains possible that other targets may exist. Notably, the coding transcripts identified in our CLIP experiment are enriched for genes involved in mitochondrial biogenesis and function (Supp. Fig. 2B). RNA-seq analysis of differentially expressed transcripts in METTL5-WT and -KO HeLa cells also revealed enrichment for small molecule transport and lipid and cholesterol biosynthesis pathways (Supp. Fig. 2E). These preliminary connections to metabolism and lipid biosynthesis, both liver-based functions, led us to generate HepG2 METTL5-KO cell lines, which may reflect gene expression pathways in the liver (Supp. Fig. 2F, 2G) more closely.

METTL5 regulates translation of a subset of transcripts
Through the course of our experiments with both HeLa and HepG2 cells, we noted that METTL5-KO cells tended to grow more slowly than the corresponding METTL5-WT cells. To measure this difference directly, cell growth curves were monitored over the course of 96 hours using DNA-dye-based CyQuant cell proliferation assays. In both HeLa and HepG2 cells, METTL5 KO cell lines tended to grow more slowly than wild type cells, with a small but consistent difference in growth rate ( Figure 3A, Supp. Fig. 3A). rRNAs are heavily modified with numerous chemical groups which regulate ribosome biogenesis and function, and play diverse roles in gene expression regulation. The slight growth defect we observed suggested that METTL5 is likely not essential for ribosome biogenesis (21), but that it may regulate the translation of a subset of transcripts that collectively slows cell proliferation. Consistent with this, polysome profiling in both HeLa ( Figure   3B, 3C) and HepG2 cells (Supp. Fig. 3B) shows no notable changes in global translation. We note that the effect of METTL5 depletion on global translation has differed across recently published reports on METTL5 function, especially across different cell lines (21)(22)(23)(24)(25).
To identify specific transcripts whose regulation may be disrupted by loss of METTL5, we Sequencing of ribosome-protected fragments revealed several differentially translated transcripts in METTL5 KO cells ( Figure 3E), the two most striking being METTL5, which was expectedly significantly downregulated in METTL5-KO cells, and CALM1, a calcium binding protein known to regulate cell proliferation and growth. Upon inspecting ribosome-protected fragment and input reads aligned to CALM1 in IGV, we noted similar transcript levels in both WT and KO cells, but greatly increased ribosome occupancy in KO cells, especially in exon 4 (Supp. Fig.   3C). In addition, numerous genes involved in the biogenesis and regulation of mitochondria were also significantly upregulated in METTL5 KO cells ( Figure 3F). Overall, although our ribosome profiling data showed that a subset of transcripts were affected more than others, it did not point to a clear mechanism explaining the preference for certain transcripts. Annotation of translated ORFs revealed higher levels of internal and novel ORFs, suggesting that dysregulated translation may be the result of frameshifting or translation of ordinarily non-coding transcripts (Supp. Fig. 3D). While there was slight dysregulation across codons due to METTL5 depletion, no single codon or amino acid stood out as being particularly affected in terms of occupancy at the P-site, which is near 18S m 6 A1832 (Supp. Fig. 3E).

Mettl5 KO mice demonstrate growth and metabolic changes
To assess METTL5 function at the level of a whole organism, we generated Mettl5 knockout mice (Mettl5 -/-) by disrupting exon 2 of Mettl5 with CRISPR-Cas9 (Supp. Fig. 4A). Mettl5 expression was undetectable in these mice by qPCR and greatly diminished by RNA-seq (Supp. Fig. 4B,C).
Critically, m 6 A levels on 18S rRNA isolated directly from the brains and livers of these mice were abolished, while the neighboring m 6,6 A site remained stable, consistent with our findings in METTL5-KO HeLa and HepG2 cells ( Figure 4A). Consistent with Ignatova et al. (22), we observe that Mettl5 KO mice are subviable and saw fewer than expected Mettl5 -/mice from both HET/HET ( Figure 4B) and HET/KO (Supp. Fig. 4D) breeding pairs. It was also immediately, visibly apparent that Mettl5 -/mice were consistently smaller than wild type (WT, +/+) and heterozygous (HET, +/-) littermates ( Figure 4C). Monitoring mouse weight across several weeks revealed that this size difference persists over time in both male and female mice, with the Mettl5 -/mice consistently weighing less than heterozygotes ( Figure 4D). Though these measurements were taken from weeks 4 to 10, we noted that the difference was already present at the point of weaning (4 weeks), and persisted with similar magnitude, suggesting that this difference arose early in development, and possibly even prior to birth.
In dissecting mice to harvest tissues for gene expression analysis, we also noted that Mettl5 -/mice consistently had less body fat than Mettl5 +/or Mettl5 +/+ mice (Supp. Fig. 4E). Indeed, RNA-seq analysis of brain and liver tissues revealed gene expression patterns that suggest altered metabolism in these mice. In particular, gene ontology analysis of genes downregulated in Mettl5 -/mouse liver revealed changes in genes involved in lipid biosynthesis and storage, consistent with the reduced body fat observed in Mettl5 -/mice ( Figure 4E,F). Noting the dramatic downregulation of Thyroid Hormone Responsive Protein (Thrsp) ( Figure 4E), we measured T3 levels in blood collected from the mice used for these RNA-seq studies and found T3 levels may be elevated in the blood of Mettl5 -/mice relative to heterozygotes (Supp. Fig. 4F).
Given the intellectual disability noted in human patients with mutations in METTL5, we also wanted to determine whether we could recapitulate any of these cognitive or behavioral changes in our mouse model. Recent reports suggest that loss of Mettl5 results in reduced locomotor activity and exploratory activity, and defects in learning and memory (22,27). In our mouse model, we did not observe similar defects in locomotor and exploratory activity from rotarod performance and open field tests ( Figure 4G,H), in fear-based learning from a shuttle box test (Supp. Fig 4G), or in instrumental learning from FR1 acquisition (Supp. Fig 4H,I). However, we note that our experiments were done comparing knockout mice with heterozygotes, not wild type mice, which combined with small variations in experimental setup could explain the different outcomes.

DISCUSSION
In our study, we identified TRMT112 to be the primary binding partner of METTL5, a finding consistent with other recent reports (21)(22)(23). We found that protein-protein interactions between METTL5 and TRMT112 are important for METTL5 stability ( Figure 1D,E; Supp. Fig.   1C,D), and methyltransferase activity ( Figure 2G), though the latter is likely a consequence of the stabilizing effect of TRMT112 on METTL5. Of clinical significance, we found that the METTL5-TRMT112 interaction was severely abrogated by mutations known to cause intellectual disability in humans ( Figure 1C) (18,20). Structural analysis of the three major human variants reported by Richard et al. and Hu et al. suggests that all three mutations (R115Nfs*19, K191Vfs*10, G61D) would likely disrupt proper folding (Supp. Fig. 1C). Of these, the frameshift variants, R115Nfs*19 and K191Vfs*10, were demonstrated to have lower expression (18,20). Structural analysis suggests that the METTL5 G61D -TRMT112 complexes that do form may lack catalytic activity, since the G61D mutation creates a polar interaction with S-adenosyl methionine (Supp. Fig. 1C). Disruption of the METTL5-TRMT112 interaction likely also decreases the stability of these METTL5 variants, negatively impacting their activity and contributing to disease. To our knowledge, this is the first time that the METTL5-TRMT112 interaction in the context of human disease has been investigated.
Intriguingly, our results further suggest the possibility that there are other pathways dysregulated by disease-causing METTL5. For example, the availability of TRMT112 to its other binding partners, which include AlkBH8 and HemK2, could also be dysregulated in this context (28,29,32).
We characterized METTL5 m 6 A methyltransferase activity and investigated its substrates in vitro and in vivo, identifying 18S m 6 A1832 as a major METTL5 substrate in mammalian cells, a result corroborated by recent reports (21,22,33). Intriguingly, though the nucleolar localization of METTL5 pointed to its role in rRNA methylation, we also found significant cytoplasmic localization of METTL5 by fluorescence microscopy and biochemical fractionation (Figure 2A,B). Significant cytoplasmic localization has also been reported in neurons (20) and Drosophila melanogaster (23), raising questions about the cytoplasmic function of METTL5. One plausible cytoplasmic role for METTL5 is late-stage methylation of rRNA, in line with an analysis by van Tran et al. of published ribosome structures that found density for the METTL5-TRMT112 complex near helix 44 only at late stages of processing (21,34). METTL5 may also remain associated with the ribosome in the cytoplasm, where it could regulate translation by recruiting factors to the ribosome. We note, however, that western blotting for METTL5 across polysome profiling fractions did not reveal an interaction with translating ribosomes in HeLa cells ( Figure 3C). Lastly, cytoplasmic METTL5 localization could suggest the existence of other METTL5 methylation targets, and differences in abundance and localization of substrates in various cell lines could contribute to discrepancies in its localization. Our CLIP-seq and m 6 A-seq identified a small subset of transcripts that were both associated with METTL5 and whose levels changed in METTL5-KO cells, respectively, supporting this idea (Supp. Fig. 2A-D). While some of these may represent indirect interactions, it is interesting that motifs mimicking the 18S m 6 A1832 sequence context are enriched for differentially methylated m 6 A peaks on transcripts overrepresented in the METTL5 CLIP data (Supp.  Figure 3E) and greater effects at the translational than the transcriptional level ( Figure   3D), consistent with 18S rRNA being a major cellular substrate of METTL5. Interestingly, a more drastic effect on polysomes has been documented in mESCs (22,24,33). Clues to how METTL5 causes transcript-and context-specific effects may come from the location of the 18S m 6 A1832 site at the tip of helix 44 near the decoding center (4,21). Rong et al. proposed that the methyl group may fine-tune the conformation of the decoding center and suggested that the methylated adenine and its base-pairing partner are in closer proximity to mRNA in the human ribosome than in structures lacking the methylation from other organisms (24). While we and others have observed through ribosome profiling that certain transcripts and codons are more affected than others ( Figure 3E, Supp. Fig 3C,E), the specific codons and transcripts affected are not consistent across datasets (22,33). Loss of 18S m 6 A1832 may also affect the position of helix 44, which is located near binding sites for key initiation and re-initiation factors, including eIF1, eIF1A, DENR, and eIF2D (35)(36)(37). Indeed, Rong et al. reported altered binding of the initiation factors eIF3A and eIF4E to translating ribosomes and decreased phosphorylation of the translation initiation-related signaling protein RPS6 in METTL5-KO cells (24). It is thus possible that some of the effects of METTL5 may be mediated by changes in binding of initiation-related factors to the ribosome. Considering the larger context beyond the ribosome itself may also shed light on differences in the effects of METTL5 knockout on translation in different cell lines, as there are known differences in translation-related signaling pathways among cell lines, especially in stem cells (38).
Our findings, particularly from our Mettl5 knockout mouse model, provide insight into hereditary METTL5-related diseases caused by human variants. To our knowledge, this is the third Mettl5 knockout mouse model reported, but the first for which 18S m 6 A loss was validated (22,27).
Guided by the observations that Mettl5 -/mice weighed less ( Figure 4D) and had decreased body fat (Supp. Fig. 4E), we further investigated the metabolism of the mice through RNA-seq of their livers.
We found dysregulated lipid, lipoprotein, and fatty acid metabolic pathways in knockout mice ( Figure 4F) and changes in thyroid hormone signaling ( Figure 4E, Supp. Fig. 4F). Interestingly, metabolic dysregulation was also suggested in a report of metl-5 knockout in C. elegans (24). The mechanism(s) leading to these changes remain to be elucidated but could have clinical significance, as many patients with METTL5-associated microcephaly and intellectual disability reported by Richard et al. are reported to have reduced body weight (20). Unlike recent reports (22,27), however, tests with our mouse model for neurological and behavioral deficits failed to show statistically significant differences between heterozygous and knockout mice in learning ability, motor activity, or exploratory activity ( Figure 4G,H; Supp. Fig. 4G,H,I), potentially due to differences in experimental details or mouse model design (Supp. Fig. 4A). Another possibility is that the loss of activity in heterozygotes may be enough to affect neurological function, although the intellectual disability and microcephaly clinical phenotypes were reported to be autosomal recessive (20). We also note that the described patient METTL5 mutants are expressed to some degree, meaning our complete knockout cell lines and mouse model do not entirely mimic the physiological conditions. It also remains possible that the neurological and behavioral defects caused by METTL5 mutations may be more subtle than is easily detectable by the simple tasks we tested but significant enough in disrupting complex tasks in humans to lead to clinical intellectual disability.
Given that 18S m 6 A1832 is currently the only validated substrate of METTL5 and the effects of METTL5 knockout seem to be most significant at the translational level, questions arise about how losing a single methyl group in the context of the whole ribosome may lead to organism-level effects.
Notably, abnormalities in development and metabolism have been found in characterized clinical ribosomopathies (39). Furthermore, the differences seen in the effects of METTL5 in different cell types and tissues mirror other ribosomopathies that are very tissue-specific (9) Fig. 1A). These combinations of plasmids were transfected into HeLa or HepG2 cells using Lipofectamine 2000 according to manufacturer instructions. After 48 hours, media was changed to media containing 1μg/mL of puromycin, and cells were allowed to grow for approximately a week, with media changes as needed as cell death progressed. Remaining cells were then trypsinized and diluted such that they could be plated at approximately 1 cell per well. Single cells were allowed to grow over the course of 2-4 weeks and collected and expanded as needed. METTL5 KO cell lines were identified via PCR to verify the appropriate deletion and verified by western blot.

Cellular fractionation
Cellular fractionation was performed essentially as previously described (43). before resuspending the beads in 450μL of this buffer (25μL of these beads were transferred to a new tube for analysis by western blot). Proteinase K (Sigma) was pre-incubated at 37ºC for 30 minutes prior to use. 50μL of Proteinase K was added to beads resuspended in Proteinase K buffer. Inputs were similarly treated with Proteinase K by adding 225μL 2x Proteinase K buffer (100mM Tris pH 7.4, 150 mM NaCl, 12.5mM EDTA, 2% SDS) and 25μL Proteinase K to 250μL of input sample.
Proteinase K treatment of both input and bead samples was done by incubating samples 50ºC for 1 hour, shaking at 1000rpm. 3 sample volumes of Trizol were added to each, and RNA was purified Samples were sequenced in one lane of an SR50 flow cell on a HiSeq4000 (50bp single-end reads).
Data analysis: Analysis was performed very similarly to what would be done for an RNA immunoprecipitation (IP) experiment to assess which RNA transcripts are bound to a protein of interest. Three replicate experiments were performed as described above, with corresponding input and IP samples for each replicate. Quality of .fastq files was checked using FastQC v0.11.5 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Adapters were trimmed using Cutadapt (45) and files were then aligned to the human genome hg38 using hisat2 v2.1.0 (46) in splice-aware mode and resulting bam files were indexed and sorted with samtools version 1.7 (47). DESeq2 (48) and R version 4.0.3 (49) were used to analyze differentially expressed transcripts, thereby identifying transcripts that were enriched in the IP samples relative to input samples.
rRNA fragment purification A 40-nt rRNA fragment was purified similarly to previously described (30). Briefly, 3-4μg of a biotinylated DNA probe designed to bind to the target rRNA region was combined with 33μg of total RNA in 3.33x hybridization buffer (250mM HEPES pH 7, 500mM KCl) in a total volume of 150μL.
The mixture was incubated at 90ºC for 7 min and then allowed to slowly cool to room temperature were counted with the function CountReads, and peaks were called using Fisher's exact test with the function callPeakFisher. The R package QNB was used for inferential testing and differentially methylated peaks were called with an adjusted P-value < 0.1 (52). Motif searches were performed using HOMER v4.11 (53). For a background reference, sequences were extracted from random 200bp peaks that were sampled from an mRNA transcript (51).

Polysome and ribosome profiling
Polysome profiling was performed similarly to previous reports (54,55 7) were grown to confluency prior to lysis in Trizol (Invitrogen). Total RNA was purified as described above ('RNA purification') and sequencing libraries were generated using an HIV reverse transcriptase evolved for m 1 A detection, as previously described (59). Libraries were sequenced using an Illumina NovaSeq6000 on an S1 flowcell with 100bp paired end reads. For RNAseq analysis, only read 1 (R1) was used.
Gene ontology analysis was performed with MetaScape (60). Open field: Open field chambers were 40cm x 40cm (Med Associates) with lighting at 21 lux.

RNA-seq, mouse tissues
Infrared beams recorded the animals' locomotor activity and rearing movements (vertical activity).
Mice were put in the open filed chamber for 1 hour to record their activity.
Passive avoidance: The shuttle box used contained two chambers: one chamber illuminated and the other dark (Kinder Scientific). Mice were transported to the behavior room and were handled for three minutes for 3 days before the passive avoidance experiment. During tasks, the right chamber remained illuminated while the left chamber remained dark. Training began by placing the mouse into the illuminated chamber facing away from the shut guillotine door. The mouse was allowed to explore the illuminated chamber for 2 minutes. The door was then opened to let the mouse explore both the illuminated and dark chambers for 5 minutes. At the end of this exploration period, the door was shut after returning the mouse into the illuminated chamber. Two minutes later, the door was opened. Latency to step into dark chamber was recorded by the computer as the baseline. Upon entering the dark chamber, the door was closed and one foot shock (0.2 mA, 2 seconds) was delivered. Ten seconds later, the mouse was removed from the dark chamber and put back to the home cage. After 24 hours, the mouse was put into the light chamber for 2 minutes and then the latency to step into dark chamber was recorded as the 24 hour memory.   2G). When multiple comparisons were made between experimentally determined pairs, the Sidak method was used to correct for multiple comparisons ( Figure 4A, Supp. Fig. 4G). In experiments with multiple variables and parameters tested, a two-way ANOVA with the appropriate multiple comparison correction was used (Sidak test for Figure 2F, Tukey test for Figure 2G).

DATA AVAILABILITY
Raw and processed data files from all high throughput sequencing experiments have been deposited in the NCBI Gene Expression Omnibus (GEO) with the following accession numbers: GSE174435 (METTL5 CLIP-seq), GSE174503 (RNA-seq of HeLa cells), GSE174420 (m 6 A-seq), GSE174418 (RNA-seq of mouse tissues), GSE174419 (ribosome profiling).