Neuronal Nsun2 deficiency produces tRNA epitranscriptomic alterations and proteomic shifts impacting synaptic signaling and behavior

Epitranscriptomic mechanisms linking tRNA function and the brain proteome to cognition and complex behaviors are not well described. Here, we report bi-directional changes in depression-related behaviors after genetic disruption of neuronal tRNA cytosine methylation, including conditional ablation and transgene-derived overexpression of Nsun2 in the mouse prefrontal cortex (PFC). Neuronal Nsun2-deficiency was associated with a decrease in tRNA m5C levels, resulting in deficits in expression of 70% of tRNAGly isodecoders. Altogether, 1488/5820 proteins changed upon neuronal Nsun2-deficiency, in conjunction with glycine codon-specific defects in translational efficiencies. Loss of Gly-rich proteins critical for glutamatergic neurotransmission was associated with impaired synaptic signaling at PFC pyramidal neurons and defective contextual fear memory. Changes in the neuronal translatome were also associated with a 146% increase in glycine biosynthesis. These findings highlight the methylation sensitivity of glycinergic tRNAs in the adult PFC. Furthermore, they link synaptic plasticity and complex behaviors to epitranscriptomic modifications of cognate tRNAs and the proteomic homeostasis associated with specific amino acids.


7.
No potential functional links between the proteome alteration or the e-phys results and behavior were provided or attempted.
8. NSUN2 targets a variety of tRNAs including glycine isodecoders. The quantitative results on other known NSUN2 target tRNAs need to be carefully investigated and described (doi: 10.3390/genes10020102).
9. In addition to tRNA, NSUN2 also targets vault RNA, an abundant lncRNA, and some mRNAs in neurons. The impact to these RNAs in the KO and OE mice should be described and explained (doi: 10.3390/genes10020102).
10. Authors claimed "ketamine…significant decrease in tRNAGlyGCC methylation 24 hours postinjection" (Line 144-145), but no actual values were mentioned in the main text. The quantification graph in Fig 3g and data in Supplementary table 9 showed highly similar methylation level between saline and ketamine treated animals. If there is a reduction, the reduction is in a range of 0.0-0.3%. Is this level of reduction sufficient for linking the antidepressant effect of ketamine and behavior in KO mice? This reviewer can not comfortably conclude the same with the authors.
11. Line 135-139. The conclusion on translational-control of glycine-rich proteins as the 3rd glycinergic regulatory pathway of central nervous system is not convincing.
12. There is no "Discussion" section.
13. To address the role of NSUN2 in PFC, a functional rescue using AAV8-Nsun2-GFP in KO background is preferred.
Minor points: 1. Please provide primer sequences used for RT-PCR in this study. 5. There is no sufficient data suggesting Fig 3b   6. I could not find information to describe the ages of the mice used for each experiment. 7. P value as 0.06 (Fig 3) was considered significant in this study. This is not conventional, please provide rationale for "significant". Reviewer #3: Remarks to the Author: In their manuscript Blaze and colleagues demonstrate that NSUN2-mediated tRNA methylation is required for the efficient translation of glycine-rich mRNAs in the postnatal mouse brain. The codon-specific shift of translation affects the global neural proteome and alters synaptic signalling and behaviour.

What does the Y-axis in
tRNA is the most extensively modified RNA, and many of these modifications have been linked to human diseases including neurological disorders. Mutations in the tRNA methyltransferase NSUN2 has been widely linked to neurological deficits in mouse and human. However, all previous studies focused on the role of NSUN2-specific tRNA methylation during development. Hence, the functional importance of NSUN2 in the postnatal brain is currently unknown. Moreover, the precise underlying molecular mechanisms of how NSUN2-mediated methylation controls mRNA translation rates are unclear. Here, the authors show for the first time that NSUN2-specific metylation of tRNAs determines codon usage. Given that dysfunctional RNA processing and translation in neuronal tissue is often observed in neurological diseases, this study is a timely and important contribution to the field. Overall, this is a very mature manuscript but is very dense, which makes it at times difficult to follow and the novel findings are at times a little lost.
1.Page 2; line 49: Only when reading the methods section, it becomes clear that the generated lines deplete NSUN2 postnatally using the CamK-Cre mice. This is very different from published mouse lines where only the developing brain was analysed. The function of NSUN2 in the postnatal brain is entirely unknown. Moreover, over-expression experiments have also not been performed before. I recommend highlighting the two facts better and early on in the text. 2.Page 3; line 83: Can the authors comment on why glycine is specifically affected? Could it be that this is simply due to the fact that in the brain all 3 cytosines in the variable loop are methylated by NSUN2. So, tRNAGly would be hypermethylated when compared to other tissues? 3.Page 4; line 92: Are neural proteins in general enriched in Gly when compared to all proteins? 4.I believe figure 3b is not referred to in the text. 5.If glycine-rich mRNAs are less translated, does this cause an increase of the amino acid glycine in the cells?
Dear Reviewers, We very much appreciate the highly constructive comments of each of the three Reviewers and provide below a detailed point-by-point response, and an overview table summarizing the newly added experiments and data analyses. We are pleased to report that we were able to address each of the Reviewers' comments. Generally, we find this manuscript to be well-written, informative, and compelling. We appreciate the rigorous controls conducted by the authors, for example in quantification of tRNA levels by three different approaches.

Reviewer requested New experiments/analysis Figure Result
Ribosome profiling 1 Ribosome profiling (RiboSeq) Fig. 2e; Extended Data Fig. 5b-d Slower decoding (increased ribosome dwell time) for the GGA codon, which corresponds to tRNA GlyTCC. Also, a significant association between translation of GGA codons and logTERs using RiboSeq data.
Fear conditioning in AAV-Nsun2 vs. AAV-GFP mice Extended Data Fig. 6d; Suppl. Table 12 No effect of PFC Nsun2 overexpression on fear memory acquisition, contextual fear memory, or cued fear memory Y-maze in AAV-Nsun2 vs. AAV-GFP for working memory Extended Data Fig. 6d; Suppl. Table 12 No effect of PFC Nsun2 overexpression on Y-maze working memory performance PFC-specific Nsun2 KO using AAVCre injection Extended Data Fig. 1e; Suppl.  Table 7 In our set of downregulated proteins, we identified 50 involved in synaptic transmission, which are even more enriched for glycine than the total downregulated population.

Nsun2 staining in neurons 2
In situ hybridization for Nsun2 exon 6 expression Extended Data Fig. 1b Detected a decrease in Nsun2 exon 6 mRNA in the KO cortex.  Table 10 Four glycine biosynthesis-related proteins significantly upregulated after Nsun2 KO and 1 downregulated Unbiased metabolomics screen in Nsun2 KO Fig. 3b; Suppl. Table 11 Of 19 amino acids, glycine is increased in Nsun2 KO while no other amino acids show significant alterations Analysis of data from Blanco et al. (2014) to identify mouse tRNAs with most Nsun2-methylation sites Extended Data Fig. 3b In mouse skin cells, three tRNA families have 3 Nsun2methylated cytosines (Gly, Glu, Pro)

Identification of significantly altered proteins involved in glycine biosynthesis
Comparison of wild-type cortex cytosine methylation in 3 tRNA with most Nsun2-methylation sites Extended Data Fig. 3c In our cortex methylation data of the three (Gly, Glu, Pro) tRNAs that have most Nsun2-methylated sites, we identify Gly tRNAs as having the most Nsun2-methylated sites in brain (4 sites) in addition to a DNMT2 site not present in other tissue types.
Explanation for glycine-selective deficit in tRNAs after widespread m5C loss 3 Glycine amino acid measurement 2,3 Fear conditioning after Nsun2 overexpression 2 Mechanistic link between PFC Nsun2 and bidirectional changes in behavior 2 General health of Nsun2 KO mice 1. One aspect that might need a better clarification in the text pertains to defining the boundaries between the novel findings in this manuscript, versus ones that are known. The fact that tRNA-Gly is a target of nsun2 is well-established. Also the fact that tRNA-Gly stability is nsun2 dependent has been characterized by studies by the Suzuki and Frye groups. In our view there is importance in replicating these findings, and in performing them in neurons, but precisely (textually) defining the parts that replicate previous findings versus ones that go beyond is in our view important.

Response:
We appreciate this comment and in response: (i) added a comprehensive introductory paragraph (Line 66-76) describing the previous work on Nsun2induced tRNA methylation: Evidence for the importance of solely NSUN2-mediated m 5 C in human health and disease was first provided by individuals with lossof-function mutation in NSUN2, displaying intellectual disability (ID), facial dysmorphism and distal myopathy [10][11][12][13] . The idea that NSUN2 deficiency causes neurological abnormalities was further delineated by studies in Drosophila (d) and mouse 9,10 , including dNsun2 knockdown with impaired short-term memory after aversive olfactory conditioning that was rescued by pan-neuronal expression of dNsun2 10 . Furthermore, mice with Nsun2 germline deletion demonstrated various impairments in locomotor activity and behavior together with reduced brain size due to excessive cell death in prenatal brain 9 . The mechanism thus far suggested for these deficits is impaired translation induced by increased tRNA fragmentation after Nsun2 ablation 14,15 . However, although tRNA cytosine methylation (m 5 C) has recently been profiled in various tissues including embryonic brain 9,16-22 the m 5 C RNA methylome and its functional relevance in the mature adult mammalian brain is still unexplored.
(ii) We then added (Line 76-88) added a comprehensive paragraph describing how our work differs from previous studies to show new functions of Nsun2 in adult brain function: In the current study, we use three different genetic approaches selectively targeting NSUN2 function in differentiated neurons of the postnatal and adult brain. These include neuron-specific conditional Nsun2 ablations and transgene-mediated increase in Nsun2 expression and methylation activity in adult prefrontal cortex (PFC), thereby focusing for the first time on the effect of NSun2 enzymatic activity on specific neuronal subpopulations in the mature brain. We report that Nsun2 shapes complex behaviors with neuronal function highly sensitive to bi-directional changes in tRNA methyltransferase activity. We show that the underlying mechanisms include alterations in tRNAs defined by high cytosine methylation specifically in the variable loop region, resulting in prominent deficits of glycine-specific tRNA isodecoders with corresponding shifts in the neuronal proteome due to decreased translational efficiency of glycine-rich neuronal proteins. Ultimately, these distortions in the glycinergic neuronal translatome lead to a striking 2.46-fold increase in PFC glycine levels associated with multi-fold increases in several key enzymes of the glycine biosynthetic pathway, further illustrating that in the adult brain, proteomic and metabolic homeostasis associated with specific amino acids is linked to epitranscriptomic regulation of cognate tRNAs.

2.
One aspect that I do believe the authors are establishing, for the first time, is the impact on translation of tRNA-Gly harboring proteins. The authors establish this by comparing mass-spectrometry data to RNAseq. Given that this is, in my view, one of the central findings in this paper, I would strongly encourage the authors to apply ribo-seq for this question. Ribo-seq is not only the state of the art for this kind of analysis, but would also provide an ability to analyze the data at codon-resolution, and to assess whether ribosomes are stalled at tRNA-Gly codons, which would strongly support their results.

Response:
We thank the Reviewer for this suggestion and have now conducted an additional experiment, performing RiboSeq on n=2 KO/2 WT cortex samples. Results show a dramatic increase in ribosome dwell time for GGA codons (corresponding to TCC anticodon for Gly compared to other Gly codons and other non-Gly codons, which corresponds to a drastic decrease in GlyTCC tRNA expression (Fig. 2a). We have added this RiboSeq data as an additional panel in Fig. 2e and in Extended Data Fig. 5b-d. Additionally, we used LogTERs from our RiboSeq data and identified a strong association between GGA codon-enrichment in transcripts and low LogTER mRNAs. We have presented this data in the Results section line 170-184.
Given our lack of expertise in neurology, we will not comment on the last section of the manuscript in which various nsun2-mediated phenotypes are documented. Finally, we have the three minor comments below: 1. Legends should clearly explain the figures, for example in figure 1b, the scheme is not on the right as stated, in extended figure 2f, the panels are not top and bottom to each other.

Response:
We have adjusted the legends to reflect the figures and appreciate the Reviewer's attention to this detail. figure 1C, the circles in one of the darker colors (green and blue) should be in a lighter shade, since when printing the figure or the legend in grey scale the colors of the circles seem identical.

Response:
We have also changed the color of the blue and green circles in Fig. 1 and Extended Data Fig. 3 to a lighter shade as requested by the Reviewer.

We found figure 1D a bit confusing. The combination of colors, images, and indicators of significance all combine into a somewhat unusual plot that is difficult to interpret.'
Response: We agree that Figure 1D has a lot of data and may be difficult for the reader to interpret, so we have made changes to make this figure clearer. First, we have significantly enlarged the figure panel and included lines to separate the 2 conditions (KO and AAV Nsun2 overexpression), each replicate (2 biological replicates each), and each cytosine. We have also added more description in the heat map legend denoting it represents logFC of the methylation percentage for each subject relative to the proper control.
Reviewer #2: Loss of function of NSUN2 in mouse and human led to motor, neurodevelopmental and cognitive defects, suggesting the neurological function of 5mC in RNA (Abbasi-Moheb et al. 2012;Khan et al. 2012;Martinez et al. 2012;Fahiminiya et al. 2014 Response: We agree with the Reviewer that the manuscript would benefit from a substantial increase in description and interpretation. We have now: (i) elongated the manuscript with a more thorough Introduction (newly added text includes all of lines 60-88) and Discussion (newly added text includes all of lines 246-323). This newly added text in the discussion starts with a lead-in sentence 'Which molecular mechanisms link Nsun2 to these robust behavioral and physiological phenotypes'? Lines 252-278 connect high Gly tRNA methylation to the fact that these tRNAs are selectively decreased after Nsun2 knock-out in mature neurons, then draw a link the decreased translational efficiencies of Gly-rich proteins including synaptic proteins, thereby drawing a link to the electrophysiological deficits and behavioral changes. To more directly address the Reviewer's comments, we now also included a new Table  (Suppl. Table 7) and Figure (Extended Data Fig. 4f) listing 50 key synaptic proteins that are downregulated in mutant, for which we now show data for glycine composition. Extended Data Fig. 4f demonstrates that these 50 key synaptic proteins (downregulated in Nsun2 mutant) show an even higher mean glycine percentage. In addition, we conducted for this resubmission additional metabolomic studies, showing a dramatic 146% increase in cortical glycine level, an alteration that was highly specific to this particular canonical amino acid. We now discuss the molecular alterations in the glycine biosynthesis pathway in our mutant mice in lines 279-304.  Fig 2b).

Response:
We thank the Reviewer for pointing this out. As requested by the Reviewer, we have now thoroughly revised the abstract and also: (i) conducted two additional experiments assessing fear memory and Y-maze working memory tests on Nsun2 overexpressing mice vs. controls. These additional data are shown in the newly added Extended Data Fig. 6d and in the text Lines 232-233.
(ii) included a new genetic manipulation of Nsun2 to produce a PFC-specific knockdown of Nsun2, for which we 1) confirmed a significant decrease in Nsun2 expression and GlyGCC cytosine methylation and 2) performed tests of behavioral despair and confirmed an anti-depressant phenotype similar to the Camk2a-Cre forebrain Nsun2 KO. The validation of this third genetic manipulation is in Extended Data Fig. 1e and Suppl. Table 13, and behavioral data is presented in Fig. 5e and Suppl. Table 12. For our second claim, we have clarified our reference to Suppl. Table 2 at line 99 that this refers to "tRNA expression levels independent of epitranscriptomic modification" in the text to make it clearer to the reader. Table 1 points to 162 tRNA isodecoders of which 12 tRNAs with significantly altered expression levels, with 7 of those being Gly tRNA isodecoders, which we have validated with two other techniques (YAMAT-seq and qPCR).
While hundreds of proteins were up and downregulate, a strength of our study was the advanced bioinformatic analysis using the FIRE heatmap (Fig. 2d) to correlate glycine codon content with translation efficiencies. When added to our previous analysis of amino acid content of proteins downregulated vs. upregulated (Fig. 2c), the data shown in Fig.2d further strengthens the link between deficits in Gly-rich tRNAs in the cell and the downregulation of glycine-rich proteins. To draw a more precise parallel between synaptic functioning and our proteomic outcomes, (as discussed in comment 1) we have also included a new Table (Suppl. Table 7) and Figure (Extended Data Fig. 4f) listing 50 key synaptic proteins that are downregulated in mutant, for which we now show data for glycine composition. Extended Data Fig. 4f demonstrates that these 50 key synaptic proteins (downregulated in Nsun2 mutant) show an even higher mean glycine percentage. We have also revised the abstract to reflect these findings.

Response:
We agree with the Reviewer that because our Nsun2 KO was specific to Camk2a-expressing excitatory glutamatergic neurons, it is likely that the downregulated proteins are in that cell type, and we have now noted in the discussion that a limitation of our proteomic experiments is the lack of cell specificity due to input restrictions. We have now added an additional analysis to our proteomics data to attempt to answer this question. To identify whether the repression of glycine-rich proteins was due to the neuronal proteome, and further glutamatergic neuron proteins, being enriched for glycine, we conducted additional analyses and have specified in the Results section lines 151-156 the following text as well as adding the new data to new Extended Data Fig. 4f: "These alterations were highly specific because glycine content of the entire set of n=434 neuron-specific proteins (UniProtKB) in our cortical proteome (comprised of neuronal and glial proteins) was very similar to the glycine content of the total proteome in our cortex homogenates (Mann-Whitney test, p=0.412; Extended Data Fig. 4f), confirming that the observed enrichment of high glycine content proteins in the fraction of down-regulated proteins in the Nsun2 mutant cortex do not reflect a non-specific decline of the neuronal proteome overall. "

The characterization of NSUN2-cKO mice is unclear. This reviewer is not sure what CamK-Cre line the authors were referring to. Assuming the authors meant CamK2a-cre mice lines, which has multiple lines with different Cre expression patterns, it is important to know which specific line was used in this study. For details, please see https://doi.org/10.1016/S0092-8674(00)81826-7. Without further information and description, the identity of the cells where KO is occurring in the mice is unclear.
Response: We now specify in the Methods section the timing of the KO using the Camk2a-Cre line and cite previous papers from our lab using the same line: "Nsun2 2lox/2lox mice were crossed with a CamK-Cre+ line to produce CamK-Cre+,Nsun2 2lox/2lox mice for knockout of Nsun2 in excitatory forebrain neurons. The calmodulin-kinase II (CamK)-Cre transgenic line results in widespread neuronal Cre-mediated deletion in forebrain before postnatal day 18 as previously described (refs)"

What is the general health of the cKO mice like?
Response: We have added new data in the text (lines 92-96) about the general health of the cKO mice, including: (i) size of conditional CK-Cre KO mice showing significant decrease in body and brain weight of Nsun2 KO mice.
(ii) that we observed the expected Mendelian ratios of offspring after breeding as stated below, with expected ratios being no different from actual ratios of mice and included this data in Extended Data Fig. 1c: "Nsun2 mutant mice were born and survived into adulthood at expected Mendelian ratios (Extended Data Fig. 1d) without overt health abnormalities and showed ~10% reductions in body and brain weight compared to WT littermates 14-16 weeks after birth (ANOVA; body weight, p<0.001; brain weight, p<0.001), with female but not male mutants exhibiting decreased brain/body ratio (t-test; p<0.01; Extended Data Fig. 1d; Suppl. Table 1)." 6. In Extended figure 2b, 5'tRF was also decreased, as well as the mature tRNA. Thus an instability-related mal-production of tRF can not explain the decrease of specific tRNAs. What is authors explanation about the distinct results in embryonic brain (Blanco S et al., 2014)?

Response:
We have now included a comprehensive introduction where we mention the canonical role of Nsun2 in tRNA fragmentation, citing the Blanco et al, 2014 study in embryonic brain in lines 71-74. We then discuss in lines 76-88 comprehensively how our work differs from previous studies including Blanco et al, as we detect a mechanism of translational control independent of tRNA fragmentation that is specific to the adult brain:

In the current study, we use three different genetic approaches selectively targeting NSUN2 function in differentiated neurons of the postnatal and adult brain. These include neuron-specific conditional Nsun2 ablations and transgene-mediated increase in Nsun2 expression and methylation activity in adult prefrontal cortex (PFC), thereby focusing for the first time on the effect of NSun2 enzymatic activity on specific neuronal subpopulations in the mature brain. We report that Nsun2 shapes complex behaviors with neuronal function highly sensitive to bi-directional changes in tRNA methyltransferase activity. We show that the underlying mechanisms include alterations in tRNAs defined by high cytosine methylation specifically in the variable loop region, resulting in prominent deficits of glycine-specific tRNA isodecoders with corresponding shifts in the neuronal proteome due to decreased
translational efficiency of glycine-rich neuronal proteins. Ultimately, these distortions in the glycinergic neuronal translatome lead to a striking 2.46-fold increase in PFC glycine levels associated with multi-fold increases in several key enzymes of the glycine biosynthetic pathway, further illustrating that in the adult brain, proteomic and metabolic homeostasis associated with specific amino acids is linked to epitranscriptomic regulation of cognate tRNAs.

No potential functional links between the proteome alteration or the e-phys results and behavior were provided or attempted.
Response: We believe we addressed this comment in our response to comment #1 from this Reviewer as stated below: This newly added text in the discussion starts with a lead-in sentence 'Which molecular mechanisms link Nsun2 to these robust behavioral and physiological phenotypes'? Lines 252-278 connect high Gly tRNA methylation to the fact that these tRNAs are selectively decreased after Nsun2 knock-out in mature neurons, then draw a link the decreased translational efficiencies of Gly-rich proteins including synaptic proteins, thereby drawing a link to the electrophysiological deficits and behavioral changes. To more directly address the Reviewer's comments, we now also included a new Table (Suppl. Table 7) and Figure (Extended Data Fig. 4f) listing 50 key synaptic proteins that are downregulated in mutant, for which we now show data for glycine composition. Extended Data Fig. 4f demonstrates that these 50 key synaptic proteins (downregulated in Nsun2 mutant) show an even higher mean glycine percentage. In addition, we conducted for this resubmission additional metabolomic studies, showing a dramatic 146% increase in cortical glycine level, an alteration that was highly specific to this particular canonical amino acid. We now discuss the molecular alterations in the glycine biosynthesis pathway in our mutant mice in lines 279-304.
8. NSUN2 targets a variety of tRNAs including glycine isodecoders. The quantitative results on other known NSUN2 target tRNAs need to be carefully investigated and described (doi: 10.3390/genes10020102).

Response:
We agree with the Reviewer that other Nsun2 targets are important to characterize in order to identify specificity for Gly tRNAs. We note that previous publications, such as Blanco et al (2014) have thoroughly characterized tRNA methylation across all isodecoders in the embryonic brain and therefore we did not conduct methylation assays on every tRNA. We originally included methylation data for Gly, Glu, and Asp because Gly and Glu are 2 out of 3 tRNAs that contain 3 Nsun2-mediated methylation sites, and Asp serves as a control for a known DNMT2 methylation site remaining unchanged after Nsun2 KO, as discussed in lines 252-278. We have now: (i) added an additional experiment with methylation data for the third tRNA that contains 3 methylation sites (Pro) as well as another nuclear tRNA (ValAAC) to provide additional evidence for Nsun2 KO remaining consistent across Nsun2-targeted tRNAs (See Extended Data Fig. 3a).
(ii) created an additional two figure panels demonstrating the thorough data provided by Blanco and colleagues (2014) on number of Nsun2-methylation cytosines at each tRNA isoacceptor in mouse skin (Extended Data Fig. 3b) and compared this to our data from wild-type mouse cortex showing glycine as a tRNA that in brain has an additional Nsun2-methylation site (Extended Data Fig. 3c), which points to the specificity for Gly in our studies.
9. In addition to tRNA, NSUN2 also targets vault RNA, an abundant lncRNA, and some mRNAs in neurons. The impact to these RNAs in the KO and OE mice should be described and explained (doi: 10.3390/genes10020102).

Response:
We agree that the effects of Nsun2 KO on all these species of RNA is an important future direction, and we have added in the Discussion lines 318-321 this information including references to studies that have previously characterized alterations to these RNAs after global Nsun2 KO.

Response:
We have removed the ketamine experiment from the manuscript due to the low magnitude of change pointed out by the Reviewer.
11. Line 135-139. The conclusion on translational-control of glycine-rich proteins as the 3rd glycinergic regulatory pathway of central nervous system is not convincing.
Response: As stated in response to Reviewer 3 comment 5 below, we believe uncovering a drastic change in glycine amino acid in the Nsun2 KO brain and changes in glycine biosynthetic proteins gives further cause to believe that Gly epitransriptomic mechanisms are an additional glycinergic regulatory pathway in the mature brain: (i) First, we noticed that within our proteomics data, there were 5 differentially expressed enzymes involved in glycine/serine biosynthesis, suggesting that glycine amino acid concentrations may be altered in the Nsun2 KO. We have now included an additional main Figure (Figure 3a) showing the glycine/serine biosynthesis pathways and using arrows to denote which enzymes were altered in the Nsun2 KO.
(ii) We then conducted an additional experiment with new cortical tissue samples in conjunction with the Rockefeller Proteomics Core, performing metabolomics for all amino acids to assess abundance in Nsun2 KO vs. WT. We did indeed find a dramatic increase in glycine in the Nsun2 KO, while other amino acids were unchanged, and we have added this data into the manuscript as stated below and in new Figure 3b. Further, we have added a paragraph about this change in amino acid biosynthesis into the Discussion section lines 279-304.
12. There is no Discussion section.

Response:
We agree this manuscript would benefit greatly from a discussion section to interpret results and discuss limitations and future directions, and have now added a substantial discussion section in lines 246-323.

To address the role of NSUN2 in PFC, a functional rescue using AAV8-Nsun2-GFP in KO background is preferred.
Response: We thank the reviewer for making this point, which is essentially asking whether Nsun2 in the PFC is essential for complex behaviors. To address this we have: (i) Added a third type of genetic experiment by localized ablation of Nsun2 in PFC specifically. We have conducted 2 additional validation experiments using this PFC-specific ablation of Nsun2 via AAV-Cre injection to produce a more direct link of causality for Nsun2 function in PFC, including qPCR for Nsun2 decrease after AAV-Cre injection and bisulfite amplicon sequencing of Gly tRNA showing decreased methylation. We have added this additional experiment into the text and into Extended Data Fig. 1d (validation of Nsun2 expression and Gly tRNA methylation) (ii) Performed tests of behavioral despair (TST and FST) on AAV-Cre injected mice and identified an antidepressant phenotype that is presented in Fig. 5e and discuss these findings in lines 233-243.
We're not aware of an established approach for reversing gene expression deficits in a widespread forebrain KO model. Due to the many brain regions affected by the forebrain KO of Nsun2, it is likely there are other areas where Nsun2 function essential for certain behaviors, but we start here by isolating PFC and conducting the AAV-Cre-mediated KO and AAV-Nsun2 transgene overexpression in that region.
Minor points: 1. Please provide primer sequences used for RT-PCR in this study.

Response:
We have now included primers used for RT-PCR in bisulfite amplicon sequencing of tRNAs as a new table, Suppl. Table 14 and refer to this in the methods section. For qPCR of Nsun2, we used Taqman probes and have listed the AssayID for Nsun2 and Gapdh in the methods. Fig 1a represent for the RNA-seq profile. At what age were the mice sampled?

Response:
We have now added the Y axis height on the RNAseq track for Nsun2 vs. KO in Figure 1a. Mice were sampled for RNAseq at ~8 weeks of age, which was added to the Methods section.

In Fig 1b, no merged image is provided to show neuron-expression of NSUN2 in the cortex.
Response: We agree with the reviewer that it is important to show Nsun2 staining in the cortex, but Figure 1b detected Nsun2-GFP from our overexpression virus. We have added a merged image to figure 1b to show Nsun2-GFP fusion protein staining merged with NeuN. Additionally, due to lack of good Nsun2 antibodies for fluorescence IHC in brain, we conducted a new experiment to confirm Nsun2 staining in cortex in WT and KO mice. We used RNA Basescope to perform in situ hybridization on cortical slices from layers II/III of wild type and Nsun2 KO cortex. We have included this data as a new panel in Extended Data Fig. 1b. Fig 1c as heatmaps, but they are not by definition. They are tables representing methylation status on each read. The number of rows should be described in the legend.

Response:
We understand the Reviewer's thought that tables in which we describe methylation aren't by definition heatmaps, so we have revised all text in the manuscript to call these "methylation maps" instead of heatmaps. We have also added the number of rows (reads) to the y-axis for each methylation map in all methylation data-containing figures.

There is no sufficient data suggesting Fig 3b.
Response: We acknowledge that Fig. 3b (new Fig 4b) is not a concrete statement for the findings from our paper, but we now refer to this in the Figure legend and in the text as a "Working Model" that we think is important to convey to the reader the molecular findings and their relation to neurotransmission graphically.
6. I could not find information to describe the ages of the mice used for each experiment.

Response:
We apologize for overlooking this, we used mice 10-12 weeks old for AAV injections and 13-16 weeks old for behavior testing and most molecular assays. We have added this information in the methods section as stated below: "Mice were 13-16 weeks old for all behavioral experiments and sacrificed at 13-16 weeks for all molecular experiments except for tRNA bisulfite sequencing and RNA sequencing of CamK-Cre Nsun2 KO mice, which were ~8 weeks old at time of sacrifice." 7. P value as 0.06 (Fig 3) was considered significant in this study. This is not conventional, please provide rationale for significant.

Response:
We have removed the wording that this was significant as we realize this is not a conventionally significant result.  Table 4. To ensure the reader will not overlook these data, we have added text when we refer to these tables next to the figure reference as shown below:

In
" Fig. 2a, statistical analyses for all detected isodecoders in Suppl. Table 2)" and "(Extended Data Fig. 4a; statistical analyses for all detected isodecoders in Suppl. (i) added a comprehensive introductory paragraph (Line 66-76) describing the previous work on Nsun2induced tRNA methylation: Evidence for the importance of solely NSUN2-mediated m 5 C in human health and disease was first provided by individuals with lossof-function mutation in NSUN2, displaying intellectual disability (ID), facial dysmorphism and distal myopathy [10][11][12][13] . The idea that NSUN2 deficiency causes neurological abnormalities was further delineated by studies in Drosophila (d) and mouse 9,10 , including dNsun2 knockdown with impaired short-term memory after aversive olfactory conditioning that was rescued by pan-neuronal expression of dNsun2 10 . Furthermore, mice with Nsun2 germline deletion demonstrated various impairments in locomotor activity and behavior together with reduced brain size due to excessive cell death in prenatal brain 9 . The mechanism thus far suggested for these deficits is impaired translation induced by increased tRNA fragmentation after Nsun2 ablation 14,15 . However, although tRNA cytosine methylation (m 5 C) has recently been profiled in various tissues including embryonic brain 9,16-22 the m 5 C RNA methylome and its functional relevance in the mature adult mammalian brain is still unexplored.
(ii) We then added (Line 76-88) added a comprehensive paragraph describing how our work differs from previous studies to show new functions of Nsun2 in adult brain function: In the current study, we use three different genetic approaches selectively targeting NSUN2 function in differentiated neurons of the postnatal and adult brain. These include neuron-specific conditional Nsun2 ablations and transgene-mediated increase in Nsun2 expression and methylation activity in adult prefrontal cortex (PFC), thereby focusing for the first time on the effect of NSun2 enzymatic activity on specific neuronal subpopulations in the mature brain. We report that Nsun2 shapes complex behaviors with neuronal function highly sensitive to bi-directional changes in tRNA methyltransferase activity. We show that the underlying mechanisms include alterations in tRNAs defined by high cytosine methylation specifically in the variable loop region, resulting in prominent deficits of glycine-specific tRNA isodecoders with corresponding shifts in the neuronal proteome due to decreased translational efficiency of glycine-rich neuronal proteins. Ultimately, these distortions in the glycinergic neuronal translatome lead to a striking 2.46-fold increase in PFC glycine levels associated with multi-fold increases in several key enzymes of the glycine biosynthetic pathway, further illustrating that in the adult brain, proteomic and metabolic homeostasis associated with specific amino acids is linked to epitranscriptomic regulation of cognate tRNAs.

Page 3; line 83: Can the authors comment on why glycine is specifically affected? Could it be that this is
simply due to the fact that in the brain all 3 cytosines in the variable loop are methylated by NSUN2. So, tRNAGly would be hypermethylated when compared to other tissues?

Response:
We appreciate the Reviewer's suggestion and agree that likely the specificity of tRNAGly is due to the four Nsun2-methylated cytosines at and around the variable loop, whereas the other tRNAs with three Nsun2-methylated cytosines are not as highly methylated. As discussed in comment 10 from Reviewer 2: We note that previous publications, such as Blanco et al (2014) have thoroughly characterized tRNA methylation across all isodecoders in the embryonic brain and therefore we did not conduct methylation assays on every tRNA. We originally included methylation data for Gly, Glu, and Asp because Gly and Glu are 2 out of 3 tRNAs that contain 3 Nsun2-mediated methylation sites, and Asp serves as a control for a known DNMT2 methylation site remaining unchanged after Nsun2 KO, as discussed in lines 252-278. We have now: (i) added an additional experiment with methylation data for the third tRNA that contains 3 methylation sites (Pro) as well as another nuclear tRNA (ValAAC) to provide additional evidence for Nsun2 KO remaining consistent across Nsun2-targeted tRNAs (See Extended Data Fig. 3a).
(ii) created an additional two figure panels demonstrating the thorough data provided by Blanco and colleagues (2014) on number of Nsun2-methylation cytosines at each tRNA isoacceptor in mouse skin (Extended Data Fig. 3b) and compared this to our data from wild-type mouse cortex showing glycine as a tRNA that in brain has an additional Nsun2-methylation site (Extended Data Fig. 3c), which points to the specificity for Gly in our studies.
3. Page 4; line 92: Are neural proteins in general enriched in Gly when compared to all proteins?
Response: This comment is similar to a comment made by Reviewer 2, and we point to our response to that comment below: We agree with the Reviewer that because our Nsun2 KO was specific to Camk2a-expressing excitatory glutamatergic neurons, it is likely that the downregulated proteins are in that cell type, and we have now noted in the discussion that a limitation of our proteomic experiments is the lack of cell specificity due to input restrictions. We have now added an additional analysis to our proteomics data to attempt to answer this question. To identify whether the repression of glycine-rich proteins was due to the neuronal proteome, and further glutamatergic neuron proteins, being enriched for glycine, we conducted additional analyses and have specified in the Results section lines 151-156 the following text as well as adding the new data to new Extended Data Fig. 4f: "These alterations were highly specific because glycine content of the entire set of n=434 neuron-specific proteins (UniProtKB) in our cortical proteome (comprised of neuronal and glial proteins) was very similar to the glycine content of the total proteome in our cortex homogenates (Mann-Whitney test, p=0.412; Extended Data Fig. 4f), confirming that the observed enrichment of high glycine content proteins in the fraction of down-regulated proteins in the Nsun2 mutant cortex do not reflect a non-specific decline of the neuronal proteome overall." figure 3b is not referred to in the text.

Response:
We have added a direct reference to Figure 3b (now Figure 4b) in the text.

If glycine-rich mRNAs are less translated, does this cause an increase of the amino acid glycine in the cells?
Response: We agree that this was a crucial question to answer for this manuscript, so we used a twopronged approach to answer this question: (i) First, we noticed that within our proteomics data, there were 5 differentially expressed enzymes involved in glycine/serine biosynthesis, suggesting that glycine amino acid concentrations may be altered in the Nsun2 KO. We have now included an additional main Figure (Figure 3a) showing the glycine/serine biosynthesis pathways and using arrows to denote which enzymes were altered in the Nsun2 KO.
(ii) We then conducted an additional experiment with new cortical tissue samples in conjunction with the Rockefeller Proteomics Core, performing metabolomics for all amino acids to assess abundance in Nsun2 KO vs. WT. We did indeed find a dramatic increase in glycine in the Nsun2 KO, while other amino acids were unchanged, and we have added this data into the manuscript as stated below and in new Figure 3b. Further, we have added a paragraph about this change in amino acid biosynthesis into the Discussion section lines 279-304.