Transfer RNA acetylation regulates in vivo mammalian stress signaling

tRNA acetylation can be manipulated in vivo via Thumpd1

To develop an in vivo model for studying mammalian tRNA acetylation, we constructed a heterozygous KO mouse (Thumpd1^+/−) by targeting sites in exons 1 and 4 of the Thumpd1 gene located on chromosome 7 using CRISPR-Cas9. Next-generation sequencing of the Thumpd1 locus revealed the formation of large [~880 base pair (bp)] frameshift deletions in several lines, which could be readily differentiated from the wild-type (WT) allele by polymerase chain reaction (PCR) (Fig. 1, B and C). To measure the expression of Thumpd1, we analyzed Thumpd1 mRNA levels in the liver isolated from WT, heterozygous (Thumpd1^+/−), and KO (Thumpd1^−/−) variants. Thumpd1 transcript levels directly correlated with the presence of an intact allele, with WT mice expressing the highest and KO mice the lowest amount (fig. S1A). Anti-ac⁴C immunoblotting confirmed loss of tRNA acetylation (lower band) but not rRNA acetylation (top band) in the KO line (Fig. 1D).

Next, we sought to better understand the spectrum of tRNAs marked by Thumpd1-dependent cytidine acetylation. In our previous studies—as well as those of others—ac⁴C has been identified exclusively at C12 within the D-arm of eukaryotic tRNA^Leu/Ser (7, 19). However, it is not known whether ac⁴C is a pervasive characteristic of tRNA^Leu/Ser or marks only a subset of isoacceptors and/or isodecoders. Furthermore, a few recent studies have suggested a broader role for ac⁴C within tRNAs (20–22). To characterize this, we determined the presence of cytidine acetylation in eukaryotic tRNAs using ac⁴C sequencing (ac⁴C-seq) (Fig. 1E). This nucleotide resolution sequencing reaction uses acidic NaCNBH₃ to alter the structure of ac⁴C, resulting in misincorporations at modification sites upon reverse transcription (RT) (23). These can then be identified as mutations by cDNA sequencing. The value of ac⁴C-seq in discovery applications has been validated across multiple settings (19, 24). Pooling three replicate ac⁴C-seq analyses of tRNA-enriched total RNA isolated from livers of WT and Thumpd1 KO mice, we sampled over 9603 tRNA positions at depths of >100 reads, including over 75% of known tRNA^Ser and tRNA^Leu genes. NaCNBH₃-dependent misincorporations were observed across all tRNA^Ser and tRNA^Leu genes sampled, which summed 26 in total and included all nine isoacceptors (Fig. 1F and data S1A). In all cases, modifications occurred at C12. Misincorporation signals were, overall, more penetrant for tRNA^Ser than tRNA^Leu. However, careful inspection of the sequence coverage identified a higher propensity for stops adjacent to C12 of tRNA^Leu (fig. S1B). This is an important illustration that, although ac⁴C-seq is quantitative for individual sites, it is not absolute; misincorporation signals across sites may vary due to both reduced ac⁴C stoichiometry and each individual RNA’s propensity for stop versus readthrough. Strong misincorporation signals were also observed at C35 of tRNA^Leu_CAA, a known 5-formylcytidine (f⁵C) site. We have previously defined the cross-reactivity of f⁵C with the ac⁴C-seq chemistry (25). Other sites producing significant C > T signals upon acidic NaCNBH₃ treatment were characterized by low (<2%) misincorporation ratios. In addition to their low stoichiometry, these sites did not occur at the 5′-CCG-3′ consensus sequence whose modification Nat10 is known to favor and were not sensitive to Thumpd1 KO (Fig. 1G and data S1B). To see whether we could identify additional penetrant cytidine acetylation sites in tRNA, we used ac⁴C-seq to analyze small RNA fractions isolated from additional yeast, mouse, and human models. Here again, we found modification sites conserved across models to be confined to tRNA^Leu/Ser C12 (fig. S1, C to E, and data S1C). We further corroborated that C12 tRNA^Ser_CGA acetylation is abolished in Thumpd1 KO mice using amplicon sequencing (fig. S1F). Taken as a whole, our results indicate that the most penetrant sites of tRNA acetylation are confined to the D-arm of tRNA^Ser and tRNA^Leu across eukaryotes, occur across all detected isoacceptors, and are highly dependent on the presence of Thumpd1.

Thumpd1 loss in mice is associated with small size and sterility

Heterozygous Thumpd1^+/− mice are viable and fertile and do not present any overt phenotype. However, mice completely deficient in Thumpd1 (Thumpd1^−/−) have significantly smaller body sizes, suggesting an important role for this gene in growth (Fig. 2A). Despite their small size, adult Thumpd1^−/− mice are viable as assessed by maintenance of body weight over time (Fig. 2B). Breeding studies found that Thumpd1^−/− mice are born at sub-Mendelian ratios indicative of decreased pre- or perinatal fitness (Fig. 2C). Ovarian atrophy in KO mice is characterized histologically by decreased ovarian oocytes, developing follicles, and corpora lutea and replacement with interstitial cell hyperplasia (Fig. 2D). Testicular atrophy is characterized histologically by disorganized seminiferous tubules with loss or vacuolation of germ cells and multinucleated germ cells; exfoliated germ cells are also observed within the epididymis (Fig. 2E). Terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) staining was significantly increased in testes but not the liver, aligning with the selective pathology (fig. S2). Further consistent with these observations, repeated matings of male Thumpd1^−/− mice (6 to 21 weeks of age) with WT females and female Thumpd1^−/− mice (6 to 21 weeks of age) with WT males failed to produce pregnancies. In contrast to the neurodevelopmental defects observed in human patients (15), no obvious changes in brain weight or morphology were observed in KO mice. These findings define overt phenotypes that accompany loss of Thumpd1-regulated tRNA acetylation in an animal model.

THUMPD1 regulates overall levels and ribosome occupancy of tRNA^Leu

As a model system for mechanistic studies, we used CRISPR-Cas9 to delete THUMPD1 in an immortalized human embryonic kidney (HEK) 293T cell line. Our rationale was that working in human cells might highlight commonalities shared across eukaryotes while, moreover, providing a model technically tractable to analysis by a wide range of experimental techniques. Similar to observations in mice, targeting of the THUMPD1 gene using CRISPR-Cas9 resulted in loss of THUMPD1 protein and specific abrogation of tRNA acetylation (Fig. 3, A and B). Previous studies have found that deletion of the S. cerevisiae Thumpd1 homolog decreased levels of ac⁴C-containing tRNAs (9, 26). However, Northern blotting analysis did not detect clear changes in tRNA^Leu/Ser levels THUMPD1 KO cells (fig. S3), consistent with prior results in a HeLa cell line (15). To increase the sensitivity of our analysis, we performed unbiased profiling of tRNAs in our WT and KO models using modification-induced misincorporation tRNA sequencing (mim-tRNA-seq) (27). A strength of this method is its ability to assay levels of sequence-related tRNA isodecoders that are difficult to differentiate by hybridization. This revealed that THUMPD1 KO specifically affects the levels of ac⁴C-containing tRNAs in HEK-293T cells, causing an ~4-fold decrease in levels of tRNA^Leu_UAA-2 and smaller reductions in tRNA^Ser_AGA-3, tRNA^Ser_UGA-4, tRNA^Leu_AAG-3, and tRNA^Leu_UAA-3 (Fig. 3C and data S2). To define whether these subtle perturbations affected translation, we performed ribosome profiling in THUMPD1 WT and KO cells. Our data showed that loss of THUMPD1 results in increased A-site occupancy at UUG, CUA, CUU, and UUA codons (Fig. 3D and data S3). Invoking wobble rules, all four of these can be decoded by tRNA^Leu_UAA and tRNA^Leu_AAG (more often found as tRNA^Leu_IAG) (28). Both of these species are detected as down-regulated in our mim-tRNA-seq data. To better understand why stalling is selective for Leu but not Ser codons, we aggregated our mim-tRNA-seq data and evaluated total tRNA abundance at the anticodon level, summing all isodecoders that decode each codon. This revealed significant reductions in multiple tRNA^Leu populations, whereas tRNA^Ser remained largely unchanged (fig. S4). Our data contrasts with that of Darnell and co-workers, who found ribosomes were recalcitrant to stalling at Leu codons even upon leucine deprivation (29) and better matches the observation of Chou and co-workers, who detected altered ribosome occupancy at Leu codons upon deletion of the S. cerevisiae Thumpd1 homolog (30). Global proteomics and immunoblotting indicated that genes with decreased translation efficiency (TE) in THUMPD1 KO cells were modestly down-regulated at the protein level (Fig. 3E and data S4). Surveys of transcript level data for mRNAs such as PSMB4 (Fig. 3F) and CDK1 (fig. S5) indicate that, even in transcripts where strong pauses at tRNA^Leu-decoded codons can be visualized, ribosome occupancy is still observed downstream. This suggests that, even when THUMPD1-dependent pausing occurs at Leu codons, protein production may be unburdened.

One way tRNA modifications can shape the proteome is by influencing codon-biased translation (31–36). To test whether a similar phenomenon acted in THUMPD1 KO cells, we performed comparisons using a reporter containing multiple copies of Leu codons that exhibited stalling (UUA and CUA) or that did not exhibit stalling (CUG/C and UCG) upstream of an enhanced green fluorescent protein (eGFP) gene. No significant difference in protein production was observed relative to WT cells (Fig. 3G). To explore this phenomenon further, we analyzed codon composition in genes showing altered TE in THUMPD1 KO cells (data S5). Comparing canonical transcripts found in the TE up and TE down datasets (data S5), we found the latter contained a significantly greater percentage of the U/A-rich Leu codons that showed stalling in our ribosome profiling studies (UUG, CUA, CUU, and UUA) relative to C/G-rich Leu codons that did not show stalling (CUC and CUG; Fig. 3H, y axis, and fig. S6A). Compared to codon composition of all human genes, U/A-rich Leu codons were also modestly enriched (Fig. 3H, x axis, and fig. S6, B and C). Analysis of amino acid pairs within TE down transcripts also found modest enrichment of diamino acids containing U/A-rich Leu (table S5) (37). Evaluating the skew between representations of individual codons (e.g., Leu-UUA) relative to the entirety of their amino acid family (e.g., all Leu) in TE down coding sequences revealed a clear bifurcation between U/A- and C/G-rich Leu/Ser (Fig. 3I). The former are more enriched TE down mRNAs, offering additional evidence of U/A-rich Leu/Ser bias. Whereas TE down sequences exhibit a shift in Leu codon composition favoring U/A-rich variants, other codons such as Lys-AAA and Arg-CU exhibit the highest absolute frequencies of enrichment (Fig. 3H). This could indicate that context, rather than frequency, determines the effect of Thumpd1 KO on protein translation. As one example, association of positively charged amino acids with the ribosomal exit tunnel can decrease translational velocity (38) and cause premature termination (39, 40), raising the possibility that THUMPD1 may collaborate with this or other transcript features. Alternatively, the apparent bias of arginine and lysine codons may reflect a pathway effect as many TE-affected genes are ribosomal proteins and translation factors known to be enriched in positively charged amino acids (41). Our results paint a picture that requires holding two opposing ideas in mind: First, sequence biases exist that imply—but do not demonstrate—a subtle, context-dependent impact of THUMPD1 on codon-specific translation. Second, the magnitude of this effect appears insufficient to disrupt reporter gene expression or obviously account for the extensive remodeling of HEK-293T proteomes caused by loss of tRNA ac⁴C.

THUMPD1 loss stimulates phosphorylation of eIF2α

As an alternative to codon-specific effects, we considered what signaling pathways may help cells maintain homeostasis in response to reduced tRNA^Leu/Ser levels. The integrated stress response (ISR) is major cellular mechanism used to adapt to changing environmental conditions including nutrient deprivation, pathogens, heme deficiency, and endoplasmic reticulum (ER) stress (42). Uncharged tRNAs and/or ribosomal collisions activate the sensor kinase Gcn2, which, in turn, phosphorylates the translation factor eIF2α at Ser⁵¹ (Fig. 4A). This phosphorylation event promotes formation of a tight eIF2α/eIF2β complex incapable of the nucleotide recycling necessary for canonical cap-dependent translation. Instead, proteins are produced primarily from mRNAs whose translation is recalcitrant to this event such as the stress-responsive transcription factor ATF4. As a previous study has found uncharged tRNAs to be more rapidly degraded than their charged counterparts (43), we hypothesized that the lower levels of tRNA^Ser and tRNA^Leu in THUMPD1 KO cells may reflect reduced aminoacylation. However, applying both RT quantitative PCR (RT-qPCR) and hybridization-based assays, we did not observe significant evidence for decreased charging of tRNA^Leu in THUMPD1 KO cells (fig. S7, A and B). This is consistent with a recent study that found that deletion of the yeast Thumpd1 homolog does not affect tRNA charging (44). To probe for evidence of ribosome collisions, we applied disome profiling to profile mRNA fragments differentially protected by endogenous stacked ribosomes in the WT and THUMPD1 KO cell lines. High-resolution footprinting revealed increased A-site occupancy across several codons in THUMPD1 KO cells (Fig. 4, B and D, and table S6). Although amino acid–selective occupancy changes were not as pronounced as with monosome analyses, the greatest differences were observed at Ser (UCA) and Leu (UUG, UUA, and CUC; Fig. 4, B and D) codons. Each of these is decoded by an ac⁴C-containing tRNA.

To determine whether ribosome collisions were associated with increased ISR kinase activation, we analyzed lysates derived from THUMPD1 KO cells with a site-specific Ser⁵¹ P-eIF2α antibody. A subtle but reproducible increase in P-eIF2α was observed in THUMPD1 KO cells (Fig. 4E), which became more apparent upon glutamine (Gln) deprivation or integrated stress response inhibitor (ISRIB) treatment. Phosphorylation of eIF2α was also apparent in the previously reported THUMPD1 KO HeLa cell line (fig. S7C) (15) and, in both cell lines, was sensitive to treatment with a Gcn2 inhibitor (fig. S7, D and E). THUMPD1 KO was associated with a growth defect in HeLa but not HEK-293T (fig. S7F). In both HEK-293T and HeLa cells, THUMPD1 KO by itself was not sufficient to activate ATF4 (Fig. 4E and fig. S7C). This is in line with our proteomic and transcriptomic data, which did not find up-regulation of canonical ATF4 targets such as asparagine synthetase (ASNS) in THUMPD1 KO cells (fig. S7G). A genome-wide screen for previously unidentified regulators of ATF4 activation also did not identify THUMPD1 (45). The threshold at which Gln deprivation activates ATF4 expression appeared similar in WT and THUMPD1 KO HEK-293T cells (fig. S7H). This suggests either THUMPD1 is not a sufficiently strong stimulus to activate ATF4 or that our cell lines harbor compensatory mechanisms that blunt ISR activation in response to permanent THUMPD1 loss.

As increased levels of P-eIF2α would be expected to impede translation, we assessed nascent protein synthesis by subjecting WT and KO cells to labeling by O-propargyl puromycin (OPP). Loss of THUMPD1 was associated with significantly lower levels of OPP incorporation, consistent with a translation defect (Fig. 4F and fig. S7, I to L). In addition to reduced translation caused by eIF2α phosphorylation, our proteomic data also observed a significant decrease in ribosomal proteins in THUMPD1 KO cells (fig. S7G and data S12). This is in line with the notion that serial passage of THUMPD1 KO cells may select for clones that adapt to increased ribosome collisions by down-regulating the translational machinery. Consistent with this view, genome-wide Perturb-seq recently found that the single-cell RNA sequencing (RNA-seq) signature of THUMPD1 KO was most similar to those elicited by disrupting of genes involved in ribosome biosynthesis, including POLR1A, POLR1B, POLR1C, and POLR1E (fig. S7M) (46). Examination of THUMPD1 KO cells did not reveal changes in mTOR activation, suggesting that the effects of THUMPD1on ribosomal protein levels are not mediated by this pathway (fig. S7N). Because of the aforementioned potential for compensatory changes to be selected for in KO cell lines, care must be taken not to overinterpret these findings. However, the identification of a link between THUMPD1, eIF2α phosphorylation, and translational down-regulation is consistent with recent studies in model organisms (10, 47). These results indicate that loss of THUMPD1-dependent ac⁴C can cause ribosome collisions associated with eIF2α phosphorylation and defective translation, spurring us to explore the significance of this mechanism in vivo.

Thumpd1 genetically interacts with Gcn2 in vivo

Next, we sought to understand whether our observations from HEK-293T cells were relevant in vivo. As an initial test, we isolated tissues from age-matched WT and Thumpd1 KO mice and assessed levels of P-eIF2α by immunohistochemical (IHC) staining (Fig. 5A). Whereas eIF2α levels were statistically identical across all samples, the percentage of P-eIF2α positive cells was significantly increased in cryosections isolated from the brain (hippocampus and cerebral cortex; Fig. 5B). This change was organ specific as P-eIF2α levels in the kidney and liver were not affected (Fig. 5B and fig. S8). Applying mim-tRNA-seq to small RNA fractions isolated from the cerebellum and liver, we again observed tRNA^Leu and tRNA^Ser to be the most down-regulated (Fig. 5, C and D, and data S7). Consistent with the increased eIF2α phosphorylation observed in brain tissues, more isodecoders were affected in the cerebellum (7) than liver (2). Cerebella also displayed up-regulation of several tRNA isodecoders (Fig. 5C, green). As these tRNAs do not contain Thumpd1-dependent ac⁴C and did not change in other tissues, it is possible that their alteration may represent technical noise or secondary effects. Ribosome profiling of cerebellum tissue samples indicated that loss of Thumpd1 causes the greatest increase in occupancy at the Leu codon UUA, consistent with the potential for loss of tRNA acetylation to stimulate ribosome collisions and eIF2α phosphorylation (Fig. 5E and data S8).

To further investigate the consequences of Thumpd1 KO in the cerebellum, we carried out RNA-seq analysis. A recent study found that exit from pluripotency is associated with eIF2α phosphorylation and used WT and eIF2α S51A mouse embryonic stem cells to define a set of transcripts dependent on eIF2α phosphorylation upon withdrawal of “2i” [two chemical inhibitors of MEK1/2 (mitogen-activated protein kinase kinase 1/2) and GSK3α/β (glycogen synthase kinase 3α/β)] (48). Overlaying these genes onto our dataset, we found most of them were up-regulated upon Thumpd1 KO, consistent with eIF2α phosphorylation (Fig. 5F, cyan, and data S9). To assess activation of the ISR, we further compared our differentially expressed gene set to Atf4 target genes previously identified in mouse embryonic fibroblasts (Fig. 5F, red) (12, 49). Here, we found a statistically significant up-regulation of several ISR genes including Atf4 itself (FC_{KO v. WT} = 1.25x, adjusted P value = 8.7 × 10⁻⁴). Evaluation of ribosome occupancy across the Atf4 transcript also revealed increased ribosome density in THUMPD1 KO samples, consistent with enhanced translation through relief of upstream open reading frame (uORF) repression (50), the canonical mechanism by which eIF2α phosphorylation promotes ATF4 expression (fig. S9). However, most eIF2α- and Atf4-dependent gene sets were unaffected, consistent with a modest response. Transcriptomic analysis of liver tissue did not show this signature (data S10). Notably, among many affected pathways, gene set enrichment analysis (GSEA) indicated up-regulation of inflammatory signaling in the cerebellum but down-regulation of the mTOR signaling in the liver, possibly indicating a selective compensatory response in the latter tissue (fig. S10, A and B, and tables S9 and S10). RNA-seq of testes, a tissue that exhibits obvious pathology upon Thumpd1 KO, provided additional evidence for tissue-selective impact with up-regulation of inflammatory and apoptotic gene expression pathways (fig. S10C and table S11). This correlated with reduced levels of tRNA^Leu, although similar to cerebella, we observed that mim-tRNA-seq data was noisier relative to cell lines or tissues possibly reflecting more widespread dysfunction (fig. S10D). These findings indicate a degree of concordance between our in vitro and in vivo results and prompted us to explore links between Thumpd1 and the stress response in further detail.

Gcn2 is the most well-characterized sensor kinase involved in the tRNA-dependent ISR. In budding yeast (S. cerevisiae), deletion of a Gcn2 homolog protects from growth defects caused by the loss of tRNA body modifications such as ac⁴C, whereas in fission yeast (S. pombe), the opposite occurs (9, 10, 26, 47). The Janus-like ability of Gcn2 to either propagate or mitigate tRNA-dependent phenotypes has also been observed in mammalian systems. Evidence indicates that an overactive ISR helps drive pathology in Charcot-Marie-Tooth (CMT) disease (51), a hereditary disorder that can be driven by dominant mutations in several tRNA synthetase genes. This has led to Gcn2 inhibition being explored as a therapeutic approach (52). Alternatively, the group of Ackerman has shown that activation of Gcn2 plays a protective role in animal models of cerebellar and retinal degeneration caused by dual mutations in the ribosome rescue factor GTPBP2 and central nervous system (CNS)–specific tRNA^Arg (12, 53). However, the in vivo genetic interaction of Gcn2 with loss of a nonessential tRNA body modification has never been explored.

To test for a potential genetic interaction between Thumpd1 and Gcn2, we generated and bred Thumpd1^+/−, Gcn2^+/− males and females. Gcn2^−/− (Thumpd1^+/+) mice fed a normal diet had no overt phenotype, consistent with a previous study (54). Genotyping of 262 offspring mice at day 14 (P14) after birth indicated that, on the Gcn2^WT and Gcn2^+/− backgrounds, ~13% of mice born were Thumpd1^−/− (Fig. 5G and fig. S11A). The sub-Mendelian production of Thumpd1^−/− mice is consistent with our initial breeding studies. However, of mice with a full KO (Gcn2^−/−) background, ~22% of mice born were Thumpd1^−/−. The increased frequency of Thumpd1^−/− mice born on the Gcn2^−/− background is statistically significant and suggests that suppression of Gcn2 activation in Thumpd1^−/− mice beneficially affects prenatal fitness. In contrast, in the postnatal setting Gcn2 loss appeared to be extremely deleterious. Gcn2^−/−, Thumpd1^−/− mice were extremely runted (fig. S11, B and C) and uniformly perished before P30 even when extra care measures such as delayed weaning and heat pads were provided. The early lethality of Gcn2^−/−, Thumpd1^−/− dual KO made it challenging to identify and prepare these mice for histopathological analysis prior to decomposition. However, analysis of a limited number of animals (n = 4) found evidence for exacerbated phenotypes across multiple tissues, including one instance of cataract formation, increased germ cell degeneration, and a higher incidence of hydrocephalus, albeit with incomplete penetrance (fig. S11, D and E). The levels of cleaved caspase-3, an established marker of apoptosis, were significantly greater in double KO (DKO) tissues (fig. S12). Together, these results demonstrate the existence of a severe synthetic lethal genetic interaction between Thumpd1 and Gcn2 in mice.

Limitations of the study

Our work characterizes the occurrence of ac⁴C at C12 across a variety of tRNA^Ser and tRNA^Leu isodecoders. However, we cannot rule out that additional sites of tRNA acetylation may occur in different cell types or tissues. Similarly, our in vitro mechanistic studies were confined to a human cell line and do dismiss the possibility that more profound codon-biased translation may be regulated by THUMPD1-dependent ac⁴C in other settings. Delineating these and other potential mechanistic inputs will be important for a holistic understanding of how tRNA acetylation influences translation.

Because the only characterized function of THUMPD1 is to assist deposition of ac⁴C in eukaryotic tRNA, a parsimonious model is that the phenotypes observed in our study reflect changes in ac⁴C-dependent tRNA function. This notion is supported by the fact that ectopic expression of tRNA^Ser can rescue S. cerevisiae growth defects caused by mutation of a Thumpd1 homolog (9, 26). However, we cannot rule out the possibility that THUMPD1-dependent phenotypes may stem from an uncharacterized chaperone role of this protein (78). Structural and biochemical analyses of the eukaryotic NAT10/ THUMPD1 complex will be important to constructing mutants that can more precisely modulate THUMPD1’s activity.

Our strategy for modulating ac⁴C contrasts with a study using NAT10 deletion (79), a more severe perturbation that is not known to cause eIF2α phosphorylation or activation of the ISR. We hypothesize that differences in these models may arise due to the centrality of NAT10 to additional processes besides tRNA modification such as ribosome biogenesis (6). Evaluation of a NAT10 knockdown HeLa cell clone (79) revealed a more profound effect on eIF2α phosphorylation and nascent translation than caused by THUMPD1 KO (fig. S14), suggesting overlapping but distinct effects. The development of strategies for temporal modulation of NAT10 and THUMPD1 activity such as small molecule inhibition and degron-tagged models will likely be useful in differentiating first-order effects from compensatory mechanisms.

Last, our studies provide an overview of the organismal consequences of THUMPD1 deletion but do not specify in vivo mechanism or report behavioral phenotypes. Although we observe increased apoptosis in tissues showing strong phenotypes, suggesting that cell death contributes to tissue dysfunction, the precise mechanisms linking translational stress to cellular mortality in Thumpd1/Gcn2 DKO animals remain to be fully defined. Although behavioral analysis of Thumpd1 KO mice was not the focus of the current study, the molecular changes we observe in brain tissue—including increased eIF2α phosphorylation and altered gene expression—align with findings from other organismal studies of tRNA modification defects that have identified specific cognitive deficits upon extensive phenotyping (59). Defining these attributes will undoubtedly be helpful in applying these models to study the role of THUMPD1/tRNA acetylation in neurodevelopment.

Animal models

Thumpd1^+/− mice were generated by introduction of Cas9 protein and two synthetically modified guide RNAs (Synthego) into C57BL/6N fertilized eggs by microinjection. The synthetic guide RNAs were designed using sgRNA Scorer 2.0 (80) to target two cuts, one in exon 1 and one in exon 4. Gcn2^+/− mouse was previously described (54) and obtained from the Jackson Laboratory (no. 008240). All mice were maintained and backcrossed on C57/BL6 background. Mice were housed at 25°C in a 12-hour light and 12-hour dark cycle. Mice were 6 to 12 weeks of age and of either sex, with age-matched littermates used as controls unless noted otherwise. Following euthanasia, tissues were isolated, flash frozen in RNAlater using liquid nitrogen, and stored at −80°C until experiments were conducted. Frederick National Laboratory for Cancer Research is accredited by AAALAC International (Association for Assessment and Accreditation of Laboratory Animal Care) and follows the NIH Public Health Service Policy for the Care and Use of Laboratory Animals. All animal experiments were approved by the NCI-Frederick Institutional Animal Care and Use Committee (approval ID: 24-437).

Genotyping

Tail DNA from Thumpd1 mice were isolated using NaOH extraction from 10- to 12-day-old mice and subjected to PCR. PCR primer pairs used for genotyping were D9 (GCTCGTTCATGTTGCAGGTG) and E9 (CAAACTGTCGCGTACGTGTG), which amplify a short segment spanning the start of exon 1; D7 (TGAGACCACCTTCCCACAGA) and E7 (GAGTGCAAAAGACTCGCAGC), which amplify a short segment spanning the end of exon 4; and D7/E9, which amplify a product only when the segment between the guides is deleted. Genotypes were analyzed by agarose gel electrophoresis for routine genotyping, whereas MiSeq analysis was used to sequence and identify breakpoints.

Construction of HEK-293T and HeLa Thumpd1 KO cells

HEK-293T Thumpd1 KO cells were generated using the CRISPR-Cas9 system. Briefly, single guide RNAs (sgRNAs) targeting protein-coding sequence of human Thumpd1 were designed using sgRNA Scorer 2.0 (80). Oligonucleotides containing the 20-nucleotide spacer sequence, along with appropriate 5′ overhangs, were annealed by mixing equal quantities (∼50 pmol) of the forward and reverse oligonucleotides, heating for 2 min at 95°C, and cooling in steps of 5°C for 2 min duration until a final temperature of 25°C. Each pair of annealed oligonucleotides was ligated into the Bbs I site of the pX458 plasmid (81). SpCas9(BB)-2A-GFP (pX458) was a gift from F. Zhang (48138 Addgene plasmid; http://n2t.net/addgene:48138). Ligated plasmids then were transformed into the Stbl3 Escherichia coli strain, and colonies were grown out for large-scale plasmid preparation. Purified plasmids expressing Thumpd1 guides 1a and 1b were cotransfected (∼1 μg) into HEK-293T (WT) cells using Lipofectamine LTX (Thermo Fisher Scientific, 15338030), and bulk GFP-positive cells were then sorted using a flow cytometer and grown for ~5 days. After assaying for protein expression by immunoblotting, candidate clones were further subjected to single-cell sorting. Confirmation of gene editing for Thumpd1 was determined by Sanger sequencing of DNA, RNA-seq, anti-ac⁴C immuno-Northern blotting, and protein expression in whole-cell extract.

HeLa Thumpd1 KO cell line used in this study was generated by the O’Connell lab, as previously described (15).

Thumpd1^+/− cross to the MMTV-PyMT metastasis model

Male FVB/N-Tg(MMTV-PyVT)634Mul/J (RRID: MGI:3032640) mice were bred to 6- to 8-week-old Thumpd1^+/− heterozygous female animals to generate F1 offspring. Genotyping of female animals was performed by PCR (82), using DNA isolated from tail biopsies taken at weaning. MMTV-PyMT positive female animals were aged until experimental end point at 120 days of age. Lungs and tumors were harvested, and tumor weight and pulmonary surface metastases were determined for each animal after euthanasia by a single investigator blinded to Thumpd1 status. Statistical significance between the two Thumpd1 genotypes was calculated using Mann-Whitney U tests in GraphPad Prism. All animal experiments were approved by the NCI-Bethesda Institutional Animal Care and Use Committee (approval ID: LPG-002).

Histopathology and blood work

A full necropsy was performed to determine the spectrum of pathology in Thumpd1 KO mice. Organ weights were recorded prior to fixation for the heart, kidney, liver, brain, lung, and spleen. All tissues were fixed in 10% Neutral buffered formalin (NBF) for 72 hours and routinely processed for hematoxylin and eosin (H&E) staining and microscopic examination by a board-certified pathologist. At necropsy, blood was taken via cardiac puncture for complete blood counts (CBC), blood smear preparation, and clinical chemistry using Genesis hematology and Abaxis VetScan VS2 (Zoetis) analyzers. Full histopathology was performed on male and female WT (n = 6), Thumpd1 KO (n = 6), and Thumpd1/Gcn2 DKO (n = 4) mice.

Growth conditions of human cell lines: HEK-293T and HeLa

HEK-293T cells for THUMPD1 WT, KO, or rescue were grown in Dulbecco’s modified Eagle’s medium (DMEM) (Quality Biological, 68101520) supplemented with 10% fetal bovine serum (FBS) (Avantor Seradigm, 97068-085), 2 mM l-glutamine (Thermo Fisher Scientific, 25030081), and 1% penicillin-streptomycin (Thermo Fisher Scientific, 15140122). HeLa cells for WT, THUMPD1 KO, or NAT10 KD (clone A1) (79) were grown under the same condition, but media were supplemented with 1 mM sodium pyruvate (Thermo Fisher Scientific, 11360-070). Cells were grown at 37°C under 5% CO₂ and passaged at 80 to 90% confluence. Cells tested negative for mycoplasma with the LookOut MycoPlasma PCR Detection Kit following the manufacturer’s instructions (Sigma-Aldrich, MP00035).

Extraction of total RNA from mammalian cells

HEK-293T cells were grown until ∼80 to 90% confluency before harvesting. Cells were harvested by either scraping or by the addition of trypsin-EDTA and centrifuging at 500 rcf, 2 min at room temperature. Pelleted cells were washed twice with cold phosphate-buffered saline (PBS), and pellets were stored frozen at −80°C. Total RNA from human cells was extracted using TRIzol according to the manufacturer’s protocol. One milliliter of TRIzol was used per 1 × 10⁷ cells. The RNA pellet was resuspended in water and stored at −80°C. Isolated total RNA was incubated with Turbo DNase (Invitrogen, AM2238) for 30 min at 37°C to remove any DNA contamination. Typical extractions were carried out with 1 × 10⁷ cells and yielded 400 μg of total RNA. The quality of the RNA was assessed by using the Agilent 2100 Bioanalyzer with the RNA 600 nano kit (Agilent, 5067-1511), and the concentration was determined by the Nanodrop or Qubit fluorimeter.

Extraction of total RNA from Thumpd1 mouse tissues

The respective mouse organs were extracted from WT, heterozygous (Thumpd1^+/−), and KO (Thumpd1^−/−) mice and were stored in RNAlater Stabilization Solution (Thermo Fisher Scientific, AM7020) at −80°C. On the basis of the tissue type, portions of 50 to 100 mg were cut from the whole organ for RNA isolation. The tissues were minced into small pieces while soaked in RNA stabilizing agent as this improves penetration of RNA stabilizing agent and retains integrity of RNA in the tissues. The minced tissues were then transferred to a 2.0-ml prefilled (1.5 mm) zirconium homogenizer bead tube (Stellar Scientific, BS-BEBU-215) as quickly as possible while maintaining cold conditions. The bead tubes were prefilled with 1 l of chilled TRI reagent from the Direct-zol RNA Miniprep kit (Zymo Research, R2051). Samples were homogenized by bead beating with a BeadBug Microtube Homogenizer at 400 rpm. The samples in the bead tubes were given pulses four times for 30 s, incubating on ice for 30 s between each beating. To remove particulate debris from homogenized samples, the samples were left on ice for 5 min and centrifuged at 21,000 rcf for 3 min. The supernatants were transferred to nuclease-free 1.7-ml Eppendorf tubes. Total RNA was purified using the Direct-zol RNA Miniprep kit following the manufacturer’s protocol (Zymo Research, R2051), and total RNA was eluted in 50 μl of nuclease-free water. Total RNA was quantified using a Nanodrop, and the quality was checked using Agilent bioanalyzer (Agilent RNA 6000 Nano Kit, 5067-1511). The total RNA was stored at −80°C until use.

Extraction of proteins from Thumpd1 mouse tissues

The respective mouse organs were extracted from WT, heterozygous (Thumpd1^+/−), and KO (Thumpd1^−/−) mice and were stored in RNAlater Stabilization Solution (Thermo Fisher Scientific, AM7020) at −80°C. Portions of 50 mg of each tissue were cut from the whole organ for protein extraction. The tissues were minced into small pieces as quickly as possible while maintaining cold conditions. The minced tissues were then transferred to a 2.0-ml prefilled (1.5 mm) zirconium homogenizer bead tube (Stellar Scientific, BS-BEBU-215). The bead tube was prefilled with 500 μl of chilled TPER tissue protein extraction reagent (Thermo Fisher Scientific, 78510) with freshly added x1 protease inhibitor cocktail (Cell Signaling Technology, 5871). Samples were homogenized by bead beating with a BeadBug Microtube Homogenizer at 400 rpm. The samples in the bead tubes were given pulses four times for 30 s, incubating on ice for 30 s between each beating. To remove particulate debris from homogenized samples, the samples were left on ice for 5 min and centrifuged at 10,000 rcf for 5 min. The supernatants were transferred to clean tubes and were sonicated on ice using 700-W QSonica Q125 sonicator with the 1/16-inch (0.15875 cm) microtip (2- × 5-s pulse, 20% amplitude, and 10-s resting on ice between pulses). Samples were centrifuged at 21,000 rcf for 30 min at 4°C. The supernatant containing the proteins was collected, leaving behind cell debris. Protein quantification was done using a Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, 23225) following the manufacturer’s protocol. The proteins were stored at −80°C until use.

Immuno-Northern blotting of ac⁴C

Total RNA was isolated from cells or mouse tissues as described above and quantified using the Qubit RNA BR assay kit (Thermo Fisher Scientific). Immuno-Northern blots were performed using Invitrogen Northern-Max reagents (Thermo Fisher Scientific). The same amount of RNA (15 μg) from each condition was aliquoted and mixed with one volume of NorthernMax-Gly Sample Loading Dye (Thermo Fisher Scientific, AM8551). These were then incubated at 65°C for 30 min and separated on a 1% agarose-1X Glyoxal Gel prepared using 10X NorthernMax-Gly Gel Prep/Running Buffer (Thermo Fisher Scientific, AM8678). Gels were run at 80 V for ~70 min or until the dye front had migrated about 3 inches (7.62 cm). Loading controls were analyzed by imaging of ethidium bromide before transfer. RNA was transferred onto Amersham Hybond-N+ membranes (Cytiva, RPN119B) using a downward capillary method as described previously (83). After transfer, membranes were cross-linked three times at 150 mJ/cm² in a UV_254nm Stratalinker 2400 (Stratagene). Membranes were then blocked with 5% nonfat milk in 0.1% tris-buffered saline with Tween 20 (TBST) for 30 min at room temperature and washed three times for 5 min each in 0.1% TBST. Membranes were then incubated overnight at 4°C with the anti-ac⁴C antibody [Abcam, ab252215 (RRID: AB_2827750); 1:2000 dilution] in blocking buffer (5% nonfat milk in 0.1% TBST). Membranes were washed three times for 5 min in 0.1% TBST and then incubated with horseradish peroxidase (HRP)–conjugated secondary anti-rabbit immunoglobulin G (IgG) [Cell Signaling Technology, 7074 (RRID: 2099233); 1:10,000 dilution] in 5% nonfat milk for 1 hour at room temperature. Membranes were washed three times for 10 min each in 0.1% TBST. SuperSignal ELISA Femto Maximum Sensitivity Substrate reagent (Thermo Fisher Scientific, 37075) was added directly to the membrane, and the signal was detected via chemiluminescent imaging using an Amersham ImageQuant 800 (Cytiva, 29399482).

Modification-induced misincorporation tRNA sequencing

Analysis of tRNA-seq libraries and misincorporation

tRNA abundance was quantified using the mim-tRNA-seq package (https://github.com/nedialkova-lab/mim-tRNAseq). Scripts are available at https://github.com/CCBR/TRANQUIL. GitHub links are provided as additional resources and are not new to this study.

Ribosome and disome profiling

HEK-293T WT and Thumpd1 KO cells were grown in 10-cm plates. Cells were washed with PBS once and collected with 1 ml of footprint lysis buffer [20 mM tris-HCl (pH 8.0), 150 mM KCl, 5 mM MgCl₂, 1 mM dithiothreitol (DTT), 1% (v/v) Triton X-100, and cycloheximide (0.1 mg/ml)] with vigorous scrapping. Lysates were digested with DNase I (2 U/ml) (Thermo Fisher Scientific, AM2222) for 15 min on ice and clarified by centrifugation at 15,000 rpm for 15 min at 4°C. Lysates containing 20 μg of total RNA were digested with 750 U of RNase I (Thermo Fisher Scientific, AM2295) at 25°C for 1 hour with gentle shaking at 500 rpm and quenched by adding 200 U of SUPERase-In (Thermo Fisher Scientific, AM2694; 20 U/μl). Nuclease-treated lysates were layered on 0.9 ml of sucrose cushion [20 mM tris-HCl (pH 8), 150 mM KCl, 5 mM MgCl₂, 1 mM DTT, and 1 M sucrose]. Ribosomes were pelleted by centrifugation in a TLA100.3 rotor at 100,000 rpm for 1 hour at 4°C, and RNA was extracted by the miRNeasy mini kit (QIAGEN). Footprints between 25 and 40 and 40 and 80 nt were size selected separately for monosome and disome library preparation, respectively. Procedures for library construction were as described previously (85).

For cerebellum tissue samples, lysates were prepared by homogenizing freshly harvested cerebella tissues in 3 volumes of lysis buffer [150 mM NaCl, 20 mM tris-HCl (pH 7.4), 5 mM DTT, cycloheximide (100 μg ml⁻¹), 1% Triton X-100, 0.5% sodium deoxycholate, complete EDTA-free protease inhibitor, and RNasin plus (40 U mL⁻¹)] using a BeadBug Microtube Homogenizer at 400 rpm. Lysates were incubated for 10 min on ice and centrifuged at 1000 rpm for 3 min at 4°C. The supernatant was removed and flash frozen in liquid nitrogen until usage. For ribosome profiling, lysates containing 40 μg of total RNA were digested with 1000 U of RNase I (Thermo Fisher Scientific, AM2295) at 25°C for 1 hour with gentle shaking at 500 rpm and quenched by adding 200 U of SUPERase-In (Thermo Fisher Scientific, AM2694; 20 U/μl). Subsequent steps were carried out as previously described, similar to the procedure for the cell lines.

Leucine and serine repeat GFP reporter assays

GFP reporter assays

WT pmaxGFP vector and pmaxGFP vectors containing upstream Leu or Ser codon repeats were transiently transfected into HEK-293T cell line and the corresponding THUMPD1 knockdown cell lines using jetPRIME transfection reagent (Polyplus, 101000046) as suggested by the manufacturer, in 96-well dishes in triplicate. Specifically, 100 ng of plasmid DNA samples in 10 μl of jetPRIME buffer was mixed with 0.2 μl of jetPRIME reagent, incubated 10 min at room temperature, and added to each well containing 1 × 10⁴ cells in standard DMEM (Corning). eGFP fluorescence# in live cells was measured at 48 hours posttransfection time in a SPARK microplate reader (Tecan), with excitation at 482 nm and emission at 512 nm. The GFP reporter expression was calculated as the ratio of eGFP fluorescence sample values (in triplicate) normalized to the corresponding values of the sample of the WT pmaxGFP vector transfected into HEK-293T (set to the value of 1) with the SD calculated for each sample.

Western blotting for analysis of eIF2α phosphorylation

HEK-293T and HeLa THUMPD1 WT/KO cells and HeLa NAT10 WT/KD cells were plated at 500,000 cells per well in 6-well plates in 2 ml of complete media and allowed to adhere overnight. For ISRIB, GCN2-IN-6 treatment, and complete deprivation of glutamine, media were aspirated 24 hours after plating cells, 1500 μl of complete media with glutamine was replaced, and cells were treated dropwise with 500 μl of media containing ISRIB (Cayman Chemical, 16258) or GCN2-IN-6 (MedChemExpress, HY-130240) for final concentrations of 1.1 μM ISRIB or 50 nM GCN2-IN-6 with 0.01% dimethyl sulfoxide (DMSO). Control wells were treated with 0.01% DMSO vehicle. Cells were returned to the incubator for 30 min, and media were then aspirated. For wells indicated for glutamine starvation, cells were washed twice with media lacking glutamine. Glutamine-containing or lacking media were replaced, and cells redosed with ISRIB/DMSO as described above, and cells were returned to the incubator for 6 hours. After 6 hours, cells were washed in 1 ml of 1x cold PBS, scraped in 500 μl of PBS, transferred to a prechilled Eppendorf tube, and centrifuged (500 rcf x 5 min), and pellets were snap frozen and stored at −80°C. Cells were lysed by sonication in 1x PBS supplemented with 1x protease/phosphatase inhibitor (Cell Signaling Technology, 5872). For concentration-dependent glutamine deprivation, media were aspirated 24 hours after plating, cells were washed 1x with PBS, and media were replaced with 2 ml of media containing 2000, 200, 20, 2, 0.2, or 0 μM l-glutamine. Cells were incubated for 3 hours before harvesting and lysing cells as described above.

Total protein quantification was performed using Precision Red Protein Assay (Cytoskeleton, ADV02). SDS–polyacrylamide gel electrophoresis (PAGE) was performed using 4 to 12% Bis-Tris NuPAGE gels (Invitrogen, NP0322 and NP0323), with XCell SureLock Mini-Cells (Invitrogen, EI0002) and MES running buffer (Invitrogen, NP0002) according to the manufacturer’s protocols. Total protein (10 μg) was loaded per well, and BenchMark Pre-stained Protein Ladder was loaded on all gels (Invitrogen, 10748010). Gels were transferred by iBlot dry transfer (Invitrogen, IB1001) using nitrocellulose transfer stacks (Invitrogen, IB301001) at 20 V for 1 min, 23 V for 4 min, and 25 V for 2 min. Total protein content was visualized using Ponceau stain after washing with 5% acetic acid in water. Membranes were blocked in StartingBlock (PBS) Blocking Buffer (Thermo Fisher Scientific, 37538) for 20 min at room temperature. Membranes were then probed overnight at 4°C with antibodies for Thumpd1 [Bethyl, A304-385A-M (RRID: AB_2620838); 1:5000 or 1:2000 dilution], eIF2α [Santa Cruz Biotechnology, sc-133227 (RRID: AB_2096505); 1:1000 dilution], P-eIF2α [Abcam, ab32157 (RRID: AB_732117); 1:10,000 dilution], ATF-4 [Cell Signaling Technology, 11815S (RRID: AB_2616025); 1:1000 dilution], or vinculin [Bethyl, A302-535A-T (RRID: AB_2728768); 1:10,000 dilution], with all dilutions made in StartingBlock Blocking Buffer. Secondary antibodies were either anti-rabbit IgG HRP-linked antibody [Cell Signaling Technology, 7074S (RRID: AB_2099233)] or anti-mouse IgG HRP-linked antibody [Cell Signaling Technology, 7076S (RRID: AB_330924)] and were both incubated at 1:1000 dilutions in 5% nonfat dry milk in 1x TBST for 1 hour at room temperature. For all but one membrane, separate gels were run for eIF2α and P-eIF2α antibodies. Some membranes were reprobed with antibodies at different molecular weights. Membranes were washed at least three times with 1x TBST between antibodies and before imaging. Imaging of colorimetric and chemiluminescent signals was performed using an Amersham ImageQuant 800 (Cytiva, 29399482) and, for the chemiluminescent signal, using Lumiglo (Cell Signaling Technology, no. 7003) or SuperSignal ELISA Femto Substrate (Thermo Fisher Scientific, 37074) according to the manufacturer’s protocols.

Procedure for cell proliferation analysis

HEK-293T and HeLa THUMPD1 WT/KO cells and HeLa NAT10 WT/KD cells were seeded at 2.5 × 10³ cells per well in 96-well white-walled, clear-bottomed plates. Following cell adherence, cell viability was assessed using CellTiter-Glo Cell Viability Assay (Promega, G7570). Luminescence was measured following the manufacturer’s protocol using the BioTek Cytation 5 Imaging Reader (Agilent). The time of this measurement, day 0, was recorded, and the assay was performed daily at the same time for 5 subsequent days. The assay was performed with biological replicates (n > 3).

Immunohistochemistry of eIF2a and eIF2a-P

Tissue sections were stained on Leica Biosystems’ BondMax autostainer with the following conditions: heat-induced epitope retrieval with EDTA for 20 min followed by phospho-eIF2a [Cell Signaling Technology, 3398 (RRID: AB_2096481), rabbit monoclonal, 1:12] or total eIF2a [Abcam, ab169528 (RRID: AB_2819002), rabbit monoclonal, 1:400]. Positive controls included human colon carcinoma and human lung carcinoma. Isotype negative controls involved replacing primary antibody with nonclonal, isotype-matched antibody from the same species as the primary antibody. Sections of the brain, liver, and kidney were evaluated from WT (n = 3), Thumpd1 KO (n = 3), and Thumpd1/Gcn2 DKO (n = 3). Slides were digitalized at 20× objective (0.5 × 0.5 μm per pixel) using an Aperio AT2 scanner (Leica Biosystems) and analyzed using HALO (Indica Labs, v3.6). Appropriateness of staining and region of interest annotation were completed by a board-certified pathologist to include the cerebral cortex, hippocampus, liver, and renal cortex. The percentage of positive pixels is reported.

In situ OPP labeling, cell culture maintenance, and proteome harvesting for gel-based fluorescence readout

HEK-293T and HeLa THUMPD1 WT/KO cells and HeLa NAT10 WT/KD cells were cultured as described above. For gel-based fluorescence detection of protein translation, three 6-cm diameter dishes (USA Scientific, CC7682-3359) were plated for each cell line with a plating density of 5.0 × 10⁵ cells. Each plate contained 5 ml of complete growth medium and returned to the cell incubator for propagation. Growth media were replaced every 2 days. Upon reaching 80% confluency, 12.5 μl of 0, 4, and 10 mM OPP stocks dissolved in DMSO (Millipore Sigma, 276855) was added in a dropwise manner to each plate and swirled in a clockwise motion to ensure even distribution of OPP. Final OPP concentrations were 0, 10, and 25 μM. To demonstrate a change in translation, a control condition was set up for each cell line that received a 15-min preincubation with 12.5 μl of 72 mM cycloheximide (EMD Millipore, 239764-10MG) stock dissolved in DMSO for a final concentration of 180 μM before the 1-hour incubation with 25 μM OPP. Plates were subsequently returned to the incubator for a 1-hour incubation. Dosing of each plate was staggered by 10-min increments to ensure that harvesting was started after the 1-hour incubation was completed. Following incubation, growth media were removed by aspiration and replaced with 1 ml of PBS. Cells were harvested by lifting cells from the plate’s surface area with a cell lifter (VWR International, 75799-938). Once cells were dislodged from the plate, the PBS solution was transferred to a 15-ml conical tube, and the cell lifting process was repeated for a total of two times. Cells were centrifuged for 5 min at 500 rcf at 4°C to form cell pellets. Following centrifugation, the supernatant was aspirated, and cells were resuspended in 1 ml of ice-cold PBS and transferred to a 1.5-ml microcentrifuge tube. Cells were then pelleted by centrifugation for 5 min at 700g at 4°C. The supernatant was subsequently aspirated, and cell pellets were resuspended in 100 μl of lysis buffer consisting of 1X PBS supplemented with 1X protease inhibitor cocktail (Cell Signaling Technology, 5871). Cell pellets were lysed by sonication using a 700-W QSonica Q700 sonicator (15- × 2-s pulse, amplitude 1, and 30-s resting on ice between pulses). Following sonication, lysates were centrifuged for 30 min at 4°C at 21,000 rcf. After centrifugation, supernatants were recovered, and the protein concentration was determined by Precision Red Advanced Protein Assay (Cytoskeleton, AVD02) using the manufacturer’s 96-well plate format. Lysates were subsequently diluted to a protein concentration of 1.3 mg/ml using lysis buffer and stored at −80°C.

In situ OPP labeling and proteome harvesting for LC-MS analysis of protein translation

For liquid chromatography–mass spectrometry (LC-MS) detection of protein translation, four 15-cm diameter dishes were plated for each cell line with a plating density of 5.0 × 10⁶ cells. Each plate contained 15 ml of complete growth media and returned to the cell culture incubator for propagation. Growth media were replaced every 2 days. Upon reaching 80% confluency, 37.5 μl of 10 mM OPP dissolved in DMSO was added in a dropwise manner to each plate and swirled in a clockwise motion to ensure even distribution of OPP. The final OPP concentration was 25 μM, and plates were returned to cell incubator for 1-hour incubation. Dosing of each plate was staggered by 10-min increments to ensure that harvesting was started after the 1-hour incubation was complete. Following incubation, growth media were removed by aspiration and replaced with 2 ml of PBS. Cells were harvested by lifting cells from the plate’s surface area with a cell lifter. Once cells were dislodged from the plate, the PBS solution was transferred to a 15-ml conical flask, and the cell lifting process was repeated for a total of two times. Cells were centrifuged for 5 min at 500 rcf at 4°C to form cell pellets. Following centrifugation, the supernatant was aspirated, and cells were resuspended in 10 ml of ice-cold PBS and pelleted as described above. PBS was removed by aspiration, and pellets were resuspended in 1 ml of PBS prior to being transferred to a 1.7-ml microcentrifuge tube. Cells were then pelleted by centrifugation for 5 min at 700 rcf at 4°C. The supernatant was subsequently aspirated, and cell pellets were resuspended in 200 μl of lysis buffer consisting of 1X PBS supplemented with 1X protease inhibitor cocktail (Cell Signaling Technology, 5871). Cell pellets were lysed by sonication using a 700-W QSonica Q700 sonicator (15 × 2 s pulse, amplitude 35%, and 30 s resting on ice between pulses). Following sonication, lysates were centrifuged for 30 min at 4°C at 21,000 rcf. After centrifugation, supernatants were recovered, and the protein concentration was determined by Precision Red Advanced Protein Assay (Cytoskeleton AVD02) using the manufacturer’s 96-well plate format. Lysates were subsequently diluted to a protein concentration of 2 mg/ml using lysis buffer and stored at −80°C.

RT-qPCR–based analysis of Thumpd1 mRNA

Total RNA from mouse liver tissue was isolated and quantified using a Nanodrop as described above. RT-qPCR was performed using the Luna Universal One-step RT-qPCR kit (New England Biolabs, E3005S) using a LightCycler 480 II PCR instrument (Roche). Real-time instrument was programmed with the following thermocycling protocol:

RT at 55°C for 10 min, initial denaturation at 95°C for 1 min, followed by 45 cycles of denaturation at 95°C for 10 s and extension at 60°C for 30 s.

The cycle threshold (Ct) values were obtained from the LightCycler 480 SW 1.5.1 software. Thumpd1 primers used for the RT-qPCR of the total RNA from Thumpd1 WT, KO, and heterozygous liver tissues are listed in table S2.

Melt curves were also performed to confirm the presence of a single amplicon by RT-PCR and the absence of primer dimer. The heterogeneous and Thumpd1 KO samples were normalized to the WT mouse using $2^{-∆∆{\mathrm{C}}_{\mathrm{t}}}$ relative abundance formula. The data are represented in technical triplicate (n = 3).

Sanger sequencing analysis of ac⁴C in tRNA^Ser

Ac⁴C-seq

HEK-293T (American Type Culture Collection, CRL-3216) and 3T3 cells (American Type Culture Collection, CRL-1658) were cultured in 10-cm plates in DMEM with 10% FBS, penicillin (100 U/ml), and streptomycin (100 g/ml) at 37°C as described above. S. cerevisiae strain BY4741 was harvested in mid-log phase after growth at 30°C in standard YPD medium (1% yeast extract, 2% Bacto Peptone, and 2% dextrose). Total RNA from HEK-293T and 3T3 cells was isolated using TRIzol as described above. Total RNA was isolated from yeast using the MasterPure Complete RNA Purification Kit (VWR, MC85200) according to the manufacturer’s protocol. RNA concentrations and purity were measured by using a Nanodrop, and RNA was stored in pure water at −80°C prior to library construction.

RNA sequencing

Total RNA from Thumpd1 WT and KO HEK-293T cells and mouse tissues (liver, cerebellum, and testes) were extracted as described above. All the samples (n = 4 for mouse tissues and n = 3 for HEK-293T) were pooled and sequenced on NovaSeq 6000 S1 using Illumina Stranded Total RNA Prep, Ligation with Ribo-Zero Plus and paired-end sequencing. Briefly, RNA-seq FASTQ files were aligned to the reference genome using STAR (86) and raw counts data produced using RSEM (97). HEK-293T samples were aligned to GRCh38 using the GENCODE_46 GTF annotation, whereas mouse tissue samples were aligned to mm10 using the GENCODE_M21 GTF annotation. Downstream analysis and visualization were performed within the NIH Integrated Analysis Platform (NIDAP) using R programs developed by a team of NCI bioinformaticians on the Foundry platform (Palantir Technologies). The RSEM count matrices were filtered for low counts (<1 cpm) and normalized by quantile normalization using the limma package (98). Differentially expressed genes were calculated using limma-Voom (99). GSEA was performed using fgsea package (100)^.

Proteomic analysis

Northern blotting–based analysis of tRNA charging

qPCR-based tRNA charging assay

The tRNA charging assay was performed as previously described by Pavlova and co-workers (103). Briefly, HEK-293T THUMPD1 WT and KO cells were grown in DMEM as described above and washed with cold PBS and lysed by adding 1 ml of TRIzol (Thermo Fisher Scientific, 15596026) to the plate and incubating for 5 min on ice. Lysates were then collected and mixed with 200 μl of chloroform in an Eppendorf tube by shaking vigorously for 20 s. After centrifuging for 15 min at 18,600 rcf at 4°C, the top fraction was collected, and the total RNA was precipitated with 2.7x volumes of cold ethanol in the presence of 2 μl of GlycoBlue Coprecipitant (Thermo Fisher Scientific, AM9515) overnight at 4°C. The pellet was resuspended in 300 μl of tRNA precipitation buffer [0.3 M acetate buffer (pH 4.5) and 10 mM EDTA] and pelleted again by adding 2.7x volumes of cold ethanol, incubating at −20°C overnight, and centrifuging for 30 min at 18,600 rcf at 4°C. The pellet was washed with 80% ethanol and resuspended in 32 μl of tRNA resuspension buffer [10 mM acetate buffer (pH 4.5) and 1 mM EDTA] before determining the concentration via a Nanodrop. For the oxidation treatment, 2 μg of RNA from each sample was used. Oxidization reactions were done by treating the RNA with 0.2 M NaIO₄ in sodium acetate buffer (pH 4.5) for 20 min at room temperature in the dark. For the controls, 0.2 M NaCl was used. Reactions were quenched by adding 2.2 μl of 2.5 M glucose and incubating for 15 min at room temperature in the dark. Into each glucose-quenched reaction, 1 μl of yeast tRNA^Phe (Sigma-Aldrich, R4018) solution (7 ng/μl stock in water) was added to serve as a spike-in control and the RNA was precipitated with ethanol as previously described. Next, to facilitate the deacylation, pelleted RNA was resuspended in 100 μl of 50 mM Tris (pH 9) and incubated at 37°C for 45 min. The reactions were quenched with 100 μl of tRNA quench buffer [50 mM Na acetate buffer (pH 4.5) and 100 mM NaCl], and RNA was pelleted by ethanol precipitation. The pellet was resuspended in 10 μl of water, and the concentration was measured using a Nanodrop. Concentrations in all the tubes were adjusted to the same level by adding water. Next, the adapter ligation step was carried out with a 5′-adenylated DNA adaptor (5′-/5rApp/TGGAATTCTCGGGTGCCAAGG/3ddC/-3′) and ~380 ng of RNA, using T4 RNA ligase 2, truncated KQ (NEB, M0373) at 18°C, overnight. Then, the CSQ_RT primer (5′-GCTGCCTTGGCACCCGA-GAATTCCA-3′) was annealed (30 s at 90°C, 5 min at 65°C, and immediately put on ice for 1 min) to the adapter region of tRNA before carrying out the RT reaction with SuperScript RT IV (Thermo Fisher Scientific, 18090050) according to the manufacturer’s protocol. The synthesized cDNA was diluted (1:10), and 2.5 μl of each reaction was used to set up each qPCR with tRNA isodecoder-specific primers using 2x SYBR Green mix (Life Technologies, 4368702). The forward primer matched the 5′ end of the tRNA, and the reverse primer covered the junction between the 3′ end of the tRNA and the ligated adaptor. The primer pairs used are listed in table S2.

LightCycler 480 II PCR (Roche) instrument was used for the qPCR step with the following thermocycling protocol:

Initial denaturation at 95°C for 1 min, followed by 45 cycles of denaturation at 95°C for 10 s and extension at 60°C for 30 s.

The Ct values were obtained from the LightCycler 480 SW 1.5.1 software. The average Ct value for each reaction and control was calculated from two technical qPCR replicates for each of the three biological replicates. Last, the charged tRNA fraction was calculated by normalizing the average ^tRNACt to the yeast-Phe tRNA using the following formula

Analysis of mTOR activation in THUMPD1 KO cells

HEK-293T THUMPD1 WT/KO cells were plated (3 × 10⁶ cells) into 100-mm tissue culture plates with 10 ml of complete medium (10% FBS, DMEM, and 1% penicillin-streptomycin) and allowed to adhere for 48 hours. Next, cells were starved with DMEM lacking amino acids (Thermo Fisher Scientific, ME120086L1) for 1 hour and/or refed with complete medium for 1 hour. For inhibitor treatments, the medium was aspirated out and 10 ml of DMEM complete medium with DMSO, DMSO + rapamycin (10 nM), or DMSO + AZD2014 (1 μM) was replaced for 1 hour. After incubation cells were collected with a cell scraper and lysed in cold NP-40 lysis buffer (50 mM Tris, 150 mM NaCl, and 1.0% NP-40 at pH 8.0). The protein lysate (25 μg) was separated in Novex WedgeWell 10 to 20% tris-glycine gels (Thermo Fisher Scientific, XP10202BOX) for proteins smaller than 100 kDa or 3 to 8% tris-acetate gels (Thermo Fisher Scientific, EA03785BOX) for proteins larger than 100 kDa. Proteins were transferred to polyvinylidene difluoride membranes (Bio-Rad, 1620177), incubated with primary antibodies overnight at 4°C, and then incubated with secondary antibodies at room temperature for 1 hour in 1x TBST + 5% blocking (Bio-Rad, 1706404). Membranes were incubated with SuperSignal West Pico Plus Chemiluminescent Substrate (Thermo Fisher Scientific, 34578). All antibodies from Bethyl Laboratories were as follows: Raptor [A300-553A (RRID: AB_2130793)] and Rictor [A300-459A (RRID: AB_2179967)]. All antibodies from CST were as follows: (Thr³⁸⁹) p-p70 S6 kinase [9234S (RRID: AB_2269803)], S6K1 [2708S (RRID: AB_390722)] (Ser⁴⁷³) p-AKT [4058S (RRID: AB_331168)], total AKT [4685S (RRID: AB_10698888)], (Ser^235/236) p-S6 ribosomal protein [4858S (RRID: AB_916156)], (Ser^240/244) p-S6 ribosomal protein [2215S (RRID: AB_331682)], S6 ribosomal protein [2217S (RRID: AB_331355)], (Thr^37/46) p–4E-BP1 [9459S (RRID: AB_330985)], (Ser⁶⁵) p–4E-BP1 [9451S (RRID: AB_330947)], 4E-BP1 [9452S (RRID: AB_331692)], and GAPDH [2118S (RRID: AB_561053)]. Antibodies p-S6, S6, and 4E-BP1 were used at 1:3000 dilution and the remainder at 1:1000 dilution in 5% BSA in 1× TBST buffer with 0.04% sodium azide. Secondary antibodies for immunoblot analysis: 1:10,000 dilution for an α-mouse IgG HRP conjugate [Promega, W4021 (RRID: AB_430834)] and 1:20,000 dilution for an α-rabbit IgG HRP conjugate [Promega, W4011 (RRID: AB_430833)].

Northern blotting analysis of tRNA^Leu/Ser in THUMPD1 WT/KO HEK-293T cells

Total RNA was extracted using TRIzol reagent (Ambion) according to the manufacturer’s instructions. For each sample, 1.5 μg of total RNA, along with a ³²P-labeled Decade marker (Ambion), was resolved on 20% (w/v) acrylamide/8 M urea gels via electrophoresis. The RNA was subsequently transferred onto a Hybond-N membrane (Amersham Pharmacia Biotech) using a semidry transfer system at 5 W for 30 min. Following the transfer, the membrane was UV cross-linked to ensure RNA immobilization and prehybridized with PerfectHyb Plus Hybridization Buffer (Sigma-Aldrich) at 37°C for 10 min. The membrane was then hybridized overnight at 37°C with ³²P-labeled probes (see table S2 for probe sequences). After hybridization, the membrane was washed three times for 15 min each with 2× SSC containing 0.1% SDS at 37°C. For detection, the membrane was exposed to an Imaging Screen-K (Bio-Rad) for 1 hour, and the signal was captured using the Typhoon Trio Imaging System (GE Healthcare). Quantification of the Northern blot results was performed using the Quantity One software (Bio-Rad).

Quantification of apoptosis using CC3 immunohistochemistry and TUNEL

Formalin-fixed testes, brain, and liver tissues were submitted for H&E, cleaved caspase-3 immunohistochemistry (CST 9661, 1:800, HIER Citrate), and TUNEL assay. Positive controls included mouse testes and thymus tissues. IHC staining was performed on Leica Biosystems’ BondMax autostainer with the following conditions using the Epitope Retrieval 1 (Citrate) and Bond Polymer Refine Detection Kit (Leica Biosystems, DS9800) with omission of the PostPrimary Reagent. Isotype control reagents were used in place of the primary caspase-3 antibody for negative controls. Slides were digitalized at 20× objective (0.5 × 0.5 μm per pixel) using an Aperio AT2 scanner (Leica Biosystems) and analyzed using HALO (Indica Labs, v3.6). Appropriateness of staining and region of interest annotation were completed by a board-certified pathologist to include target tissues and exclude necrosis, nontarget tissues, and histology artifacts. Quantification of positivity was performed using digital image analysis, which included identifying stain vectors, optimizing cell detection algorithms, and thresholding chromogenic IHC staining based on positive and negative controls. The percentage of positive cells is reported for each stain.

The PDF file includes:

Other Supplementary Material for this manuscript includes the following: