Mistranslation from an endogenous tRNA variant in human pan-genome cell lines

Abstract

Human transfer RNA (tRNA) anticodon variants are a source of translation error. The tRNA^Ser_AGA-2–3 variant (G35A) occurs in 2% of the human population and causes mis-incorporation of serine at phenylalanine codons. Here, we developed a dual fluorescent reporter to quantify mis-incorporation levels in live human and murine cells and validated mistranslation by mass spectrometry. In β-lymphocytes from individuals in the 1000 genomes project, we confirmed the anticipated genotype of cells with A35 minor alleles, and tRNA sequencing demonstrated expression, C32 hypo-modification, and partial 5′-fragmentation of the endogenous mutant tRNA^Ser_AAA. Nanoparticle delivery of the fluorescent reporter confirmed serine mis-incorporation in the pan-genome cell lines. The data demonstrate that a natural genome-encoded human tRNA mutant causes mistranslation in cells derived from healthy individuals. Our findings have important implications for translation fidelity in humans and the application of missense suppressor tRNAs to medicine.

Graphical Abstract

Graphical Abstract.

Introduction

In the flow of biological information from DNA to RNA to protein, translation of the messenger RNAs is the most prone to error [1]. Mistakes in protein synthesis can result from multiple sources, including mutations in ribosomes [2], aminoacyl-transfer RNA (tRNA) synthetases (AARSs) [3–5], tRNAs [6–8], and components of the protein quality control machinery [8]. Translation fidelity depends critically on the accurate and specific aminoacylation of the cellular tRNA pool. Functional tRNAs and specific aminoacylation of each tRNA isoacceptor family with the appropriate cognate amino acid are critical to accurate and efficient protein synthesis in all cells. To ensure accurate translation, AARSs have specificity for recognizing the correct amino acid and catalysing its ligation to the cognate tRNA. On the ribosome, the aminoacyl–tRNA products of this reaction must then faithfully read the correct codon or set of codons in each mRNA corresponding to the cognate amino acid carried by the tRNA. Mutations to the anticodon or to identity element nucleotides [9] in tRNAs that establish specificity with the cognate AARS can result in nonfunctional tRNAs [10] or tRNAs that promote amino acid mis-incorporation across the proteome [11–15] as a consequence of the fact that the ribosome does not proofread mis-aminoacylated tRNAs.

While mistranslation occurs at a basal level in all cells, estimates of the error rate vary greatly. Some studies suggest 1 in every 10 000 codons are mistranslated by the ribosome [16, 17]. Data based on sensitive luminescent reporters, however, show the basal error rate is much lower and in the range of 1 mistake in every 10⁶–10⁸ codons [18]. Fascinatingly, diverse cells can tolerate significantly elevated levels of mistranslation of ∼1%–10% per codon [13, 19–21], and certain mutations that reduce translation fidelity are associated with disease. Mistranslation caused by editing defective alanyl–tRNA synthetase variants were associated with neurodegenerative disease in mouse models [22] and microcephaly in humans [23]. Mutations in histidyl–tRNA synthetase that reduce translation fidelity at histidine codons are causative for Charcot–Marie–Tooth disease (CMT) [5].

Given these findings, we were surprised to identify several functional tRNAs with anticodon or identity element mutations in human genomes. These tRNA variants represent a range from exceedingly rare (<<1%) to more common alleles (2%–8%) in the human population [6, 24]. In subsequent studies, we expressed several of these natural human tRNA variants in multiple mammalian cell lines and demonstrated their functionality in generating mis-incorporation of serine (Ser) at phenylalanine (Phe) codons [15], alanine (Ala) at glycine (Gly) codons [12], and leucine (Leu) at Phe and tryptophan (Trp) codons [11]. Each of the ectopically expressed human tRNA variants caused mistranslation in cells according to mass spectrometry, and we found that mistranslating cells showed a range of defects in their ability to efficiently produce a fluorescent protein as a marker for protein synthesis.

Although many of the mistranslating tRNA variants were well-tolerated in human cells, according to cytotoxicity and cell viability assays, we found the tRNA that mis-incorporates Ser at Phe codons caused significant phenotypic impacts or defects in cells where protein homeostasis was also under stress. The G35A mutation in tRNA^Ser_AGA 2–3 produces a Ser-accepting tRNA with a Phe-decoding AAA anticodon, and the minor allele is found in 2% of the human population [6, 24]. Inhibition of proteasomal protein degradation is synthetic toxic with this tRNA mutant in human cells [15] as was co-expression of an ALS-causative allele of the fused in sarcoma protein [14]. In cellular models of Huntington’s disease, the tRNA^Ser_AAA slowed poly-glutamine aggregation kinetics and inhibited their degradation in neuroblastoma cells [15]. Although tRNA mutations are normally not included in association studies, perhaps because the tRNA^Ser_AGA 2–3 gene occurs in a region close to several protein coding genes, this particular G35A allele (rs147439337) was mapped to the nearby flanking genes (ZNF184, HNRNPA1P1; previous annotation XXbac-BPGBPG34I8.1) and found to be significantly underrepresented in individuals with depression according to a genome-wide association study based on data from 1.9 million individuals in the United Kingdom biobank [25].

Mass spectrometry is an important method to identify diverse peptides, including mistranslated peptides, but the approach is only quantitative in assessing the relative abundance of identical peptides in different samples. The problem of quantifying mistranslation events requires comparing the levels of two different peptides, one properly translated and one with a mis-incorporated residue. A single amino acid change can dramatically alter the retention time and intensity of the observed peptide even when peptides differing by a single residue are supplied in the same amount to the mass spectrometer [26]. Several studies have used fluorescent [27] or luminescent [28] reporters to quantitate mistranslation events in cells. These reporters generally engineer a mutant fluorescent protein that is dark when properly translated and fluorescence is restored when a critical residue is mistranslated [29]. The level of fluorescence restoration is, thus, proportional to the level of mistakes made at the mistranslation sensitive site. Reporters of this type have been developed to study threonine [30], valine [31], methionine [32], and proline (Pro) [29] mistranslation. Mistranslation can be enhanced under cellular stress conditions, such as oxidative stress [32] or nutrient deprivation [13], and error-sensitive fluorescent reporters have been instrumental in documenting and measuring error levels in protein synthesis under stress [33].

Here we addressed the problem to measure mistranslation levels from a natural human tRNA variant in live cells by developing a dual fluorescent reporter (GFP-mCherry Ser151Phe) that is sensitive to Ser mis-incorporation at Phe codons (Fig. 1). After establishing the functionality of the mistranslation reporter by measuring error levels from the G35A variant in two different tRNA^Ser genes in human and murine cell lines, we then used the reporter to show that cells derived from individuals with the A35 minor allele indeed mistranslate the genetic code. To assess the function of endogenous tRNA^Ser_AAA variants and sample the diversity of the human pan genome, we characterized multiple β-lymphocyte cell lines derived from healthy individuals in the 1000 genomes project that contain genomically encoded tRNA^Ser_AAA (Table 1). DNA sequencing confirmed the presence of the genome encoded G35A variants. We then used tRNA sequencing to reveal expression, hypomodification, and fragmentation of the mutant tRNA, and polymer nanoparticle-based delivery of our live cell reporter confirmed mistranslation of Ser at Phe codons in the β-lymphocytes. Our data show that cells derived from healthy individuals mistranslate the genetic code with a single endogenous tRNA gene variant.

Figure 1.

Design of dual fluorescent reporter sensitive to serine mistranslation. The fluorescent reporter consists of a GFP-mCherry fusion protein, where mCherry is fluorescently inactive due to a point mutation (Ser151Phe). GFP is always active and acts as a control to measure protein production in individual live cells and to normalize mCherry fluorescence levels. High-fidelity protein synthesis results in no restoration of mCherry Ser151Phe fluorescence. Here, we established the ability of the dual reporter to detect and quantify mistranslation events resulting from a naturally occurring human serine tRNA G35A variant that causes a mistranslation at the Phe codon (UUU) at position 151, which leads to a revival in mCherry fluorescence that is proportional to the level of errors at the UUU151 codon.

Table 1.

Human β-lymphocyte lines with wildtype tRNA^Ser_AGA and G35A variants

Coriell Institute biobank cell line accession	Cell line ID	Donor sex	Ethnic group	tRNASerAGA gene	tRNASer N35 genotype
HG00096	096	Male	Scottish	2–3	G/G
HG00097	097	Female	Scottish	2–3	G/G
GM20508	508	Female	Toscani	2–3	G/G
GM20845	845	Male	Gujarati Indian	2–3	G/G
GM20847	847	Female	Gujarati Indian	2–3	G/G
GM20518	518	Male	Toscani	2–3	G/A
GM20800	800	Female	Toscani	2–3	G/Aa
GM20821	821	Female	Toscani	2–3	G/A
GM20876	876	Female	Gujarati Indian	2–3	G/Aa
GM20878	878	Female	Gujarati Indian	2–3	G/A
GM20896	896	Female	Gujarati Indian	2–3	G/A
HG01468	468	Female	Colombian	2–2	G/Ab

^aFor the G35A genotype at tRNA^Ser_AGA 2–3 gene (rsid: rs147439337), the cell ID 800 and 876 genotype is listed a A/A homozygous at the Ensembl database [39], yet our genotype (Supplementary Fig. S10) and tRNA sequencing data (Fig. 7) confirm these cell lines are heterozygous.

^bThe HG01468 cell line has the G35A genotype in the tRNA^Ser_AGA 2–2 gene (rsid: rs571240633).

Materials and methods

Cell lines, plasmids, cloning

Mammalian cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA, United States), and cell culture work was conducted in the human embryonic kidney (HEK) 293T cells (CRL-3216, ATCC), in murine neuroblastoma Neuro-2a (N2a) cells (CCL-131, ATCC), or in human K562 lymphoblast cells (CCL-243, ATCC) as indicated. Lymphoblastoid cell lines (β-lymphocytes) from individuals in the 1000 genomes project were obtained from the Coriell Institute biobank (see accession numbers in Table 1). These cell lines were originally from peripheral blood mononuclear cells collected from healthy individuals and transformed with the Epstein–Barr virus at the Coriell Institute. The genotype of several β-lymphocyte lines was confirmed by amplifying a region including the tRNA^Ser AGA-2–3 gene using polymerase chain reaction (PCR) and the primers listed in Supplementary Table S1.

The WT-PAN [32] (kanamycin resistant) plasmid was obtained from Addgene (plasmid #99638) and contains a constitutively expressed eGFP-mCherry dual fluorescent fusion protein. Additional mCherry-only plasmids (pPanCherry [14]) were made by removing the EGFP sequence via digestion and ligation of the isoschizomeric SpeI/NheI sites (New England Biolabs, Ipswich, MA, United States) flanking the eGFP. Plasmid manipulations were done in Escherichia coli DH5α. The S151F (TTT) mutation in mCherry was introduced using site-directed mutagenesis (see primers in Supplementary Table S1).

The WT-PAN plasmids were then further engineered by introducing tRNA expression cassettes. The tRNA^Ser AGA-2–2, 2–3, 2–4, and 2–5 genes with ∼300 base pairs (bp) of flanking sequence were amplified via PCR from genomic DNA purified from HEK 293T cells. A second round of PCR was done to add PciI restriction enzyme site flanking the PCR product. The wildtype tRNA expression cassettes were digested with PciI, gel purified and ligated with T4 DNA ligase (New England Biolabs) into a PciI digested and gel-purified WT-PAN vector. Successful clones were verified by DNA sequencing at the London Regional Genomics Centre (Western University, London, Ontario). The anticodon variant G35A for tRNA^Ser AGA-2–2, 2–3, 2–4, and 2–5 were introduced in PCR fragments using overlap extension PCR with a mutant primer (Supplementary Table S1). The final PCR product also includes flanking PciI restriction sites and the mutant tRNA alleles were digested and ligated into WT-PAN as above and confirmed with DNA sequencing.

Plasmid transfection

HEK 293T, K562, and N2a mammalian cells were cultured in Dulbecco’s modified Eagle medium (DMEM 4.5 g/l; Gibco) with 10% fetal bovine serum (FBS; Gibco, Thermo Fisher Scientific, Waltham, MA, United States) and 1% streptomycin–penicillin (S/P) (Wisent Bioproducts, St. Bruno, Quebec, Canada) and incubated at 37°C with 5% CO₂ and humidity. Transfections were done in a 6-well, 24-well, or 96-well plates (Corning, Inc., Corning, NY, United States) using 2.5, 1.0, or 0.1 μg of plasmid DNA, respectively. The GFP-mCherry and tRNA expressing plasmids were purified using a Geneaid MIDI-Prep kit (GeneAid, New Taipei City, Taiwan). At ∼75% confluency, lipofectamine 3000 (Invitrogen, Thermo Fisher Scientific) was used to transfect cells following manufacturer’s protocols. Briefly, a master mix of purified plasmid was made by diluting DNA with DMEM (4.5 g/l; Gibco) and thorough mixing with the P3000 reagent at [DNA] = 2 μl/μg. A lipofectamine 3000 mix was made in the same final volume of DMEM in a separate tube as the DNA master mix, adding 1.5 μl of lipofectamine reagent for every 2 μl of P3000 reagent followed by vigorous shaking for 30 s. The plasmid DNA solution was then combined to this lipofectamine 3000 mixture at a 1:1 ratio briefly shaken and incubated for 10 min at room temperature. An aliquot of 10, 50, or 250 μl of this transfection mixture was added to each well on either a 96-, 24-, or 6-well plate, respectively. The plates were gently shaken to ensure even distribution of transfection mixture and incubated at 37°C with 5% CO₂ and humidity for 24 h.

Human β-lymphocytes (Coriell Institute, New Jersey, United States; Table 1) were grown as suspension cultures in T-25 flasks and in 250 ml spinner flasks (CELLSPIN, Integra Bioscience, Zizers, Switzerland) with magnetic rods rotating at ∼40 revolutions per minute (rpm). The cells were grown in Roswell Park Memorial Institute (RPMI 1640) media with 15% FBS (Gibco) and 1% S/P (Wisent Bioproducts) and incubated at 37°C with 5% CO₂ and humidity. For transfection of the WT-PAN plasmids with wildtype GFP-mCherry or the GFP-mCherry Ser151Phe mutant plasmid and lacking any additional tRNA allele, ∼5 × 10⁴ β-lymphocyte cells were transferred per well onto a 96-well plate (Corning) and transfected with a TransIT-X2 dynamic delivery system (Mirus Bio, Madison, WI, United States). Following the manufacturer’s protocol and at suggested volumes, a 1:3 molar ratio of DNA to polymer nanoparticle transfection reagent was used, in which purified DNA was added to a concentration of 1 μg/μl. The DNA and TransIT-X2 nanoparticles were mixed and incubated at room temperature for 20 min prior to the addition of 5 μl of transfection mixture to each well in a 96-well plate. Following gentle shaking of the plate to ensure even distribution of nanoparticle/DNA complex, and the cells were incubated at 37°C with 5% CO₂ and humidity for 24 h.

Fluorescence microscopy

Fluorescent microscopy imaging was performed using a Cytation C10 Imaging Reader (Agilent, Santa Clara, CA, United States) on cells 24 h post transfection. Images of cells were taken using a 10 × Olympus Plan Fluorite phase objective with brightfield and with the GFP (excitation 469 ± 37 nm; emission 525 ± 39 nm) and TexasRed filter cubes (excitation 586 ± 15 nm; emission 647 ± 57 nm) to detect GFP and mCherry fluorescence, respectively. To account for a diversity of expression levels in different cells, images were acquired at long [Illumination intensity = 10, Integration time (ms) = 200] and shorter exposure times [Illumination intensity = 10, Integration time (ms) = 5]. We found that GFP and mCherry ratios were not significantly different in long as compared to short exposure images, so we plotted data from the short exposure images in the figures and included long exposure images in the Supplementary Information for ease of viewing mCherry recovery. Images were further processed and analyzed using an automated ImageJ [34] macro (Supplemental Information Appendix) to annotate cells and measure GFP and mCherry intensity within individual cells defined as regions of interest (ROIs). Images were imported to ImageJ [34], background fluorescence was subtracted using a rolling circle function, and then images were converted to mask, and an ROI was defined by cells with detectable GFP fluorescence in the GFP channel. Afterwards, both GFP and mCherry intensities were measured and annotated in each ROI, which is automatically compiled into an exported Microsoft Excel document. Thresholds for ROI cutoffs based on the brightness of the image pixels were adjusted based on cell type due to morphology; with either a lower and upper bound of (10, 200) for adherent cells or (15, 80) for cells in suspension. The ImageJ [34] macro source code and comments are included the Supplemental Information Appendix.

Flow cytometry

Approximately 1.5 × 10⁶ transfected HEK293T or N2a cells were harvested from a six-well plate and centrifuged at 300 × g for 5 min; excess media was aspirated. HEK 293T or N2a cells were resuspended in 300 μl of sorting buffer [phosphate buffered saline (PBS) with 3% FBS) and stained with either 1μl of Sytox Blue (S11348, Invitrogen) or Sytox Orange (S11368, Invitrogen), respectively. Cell sorting was done on a BD FACS Aria III cell sorter (BD Life Sciences, Oregan, United States), using either a Blue 488 nm laser with B 530/30 filter and 502 LP mirror or 561 nm laser with a YG 610/20 filter and 600 LP mirror for GFP and mCherry fluorescence, respectively. Detectors were chosen based on optimized protocols for detection of common fluorescent proteins. A preliminary sort was done on nontransfected cells stained with and without the indicated cytotoxicity dye to optimize flow settings for cell morphology and Forward Scatter (FSC) gates to remove nonviable cells. A secondary preliminary sort was done with transfected cells expressing the GFP-mCherry S151F and wildtype tRNA^Ser_AGA plasmid to establish FSC gates to distinguish cells transfected with the negative control and establish the background threshold above which actual mCherry fluorescence was detectable. Subsequent sorted cells were categorized as falling into GFP−, GFP+/mCherry−, or GFP+/mCherry+ gates (see Supplemental Information for specific FSC gate placement) and either discarded if GFP− or placed in recovery media (DMEM with 20% FBS, 1% S/P) if GFP+. The raw GFP and mCherry fluorescence data and ratios from individual cells that were GFP+ were analyzed using FlowJo software (BD Life Sciences) and Microsoft Excel. Heatmaps to plot the numbers of cells with different ratios of mCherry/GFP fluorescence (i.e. mistranslation levels) as a function of GFP fluorescence were generated with MATLAB Version 24.2.0.2863752 (R2024b) Update 5 (MATLAB, Natick, Massachusetts, United States) using the Heatscatter plugin (Heatscatter plot for variables x and y Version 1.1.0.0, MATLAB Central File Exchange).

Cytotoxicity assay

HEK 293T and N2a cells were transfected with pPANCherry plasmids also containing wildtype tRNA^Ser_AGA or tRNA^Ser_AAA variants with different flanking sequence contexts from human tRNA^Ser_AGA 2–2, 2–3, 2–4, and 2–5 genes. Cytotoxicity was quantified 24 h post transfection using the CytoTox-Glo luminescent assay (Promega, Madison, WI, United States). Following the manufacturer’s protocol, CytoTox-Glo Cytotoxicity Assay Reagent was prepared by transferring the entire content of one bottle of Assay Buffer provided to an AAF-Glo Substrate bottle and mixing well. A 50 μl aliquot of CytoTox-Glo Cytotoxicity Assay Reagent was added to each well on a 96-well plate, with each well containing ∼1 × 10⁴ cells and incubated at room temperature for 15 min. Luminescence was measured to detect the cellular proteases released into the medium from dead cells. Lysis buffer was prepared following manufacturer’s protocol by transferring 33 μl of digitonin provided to a fresh 5 ml bottle of CytoTox-Glo assay buffer and mixing well. Cells were then lysed by adding 50 μl of lysis buffer to each well at room temperature with orbital shaking for 15 min. The luminescent signal was then measured again to determine the total signal from each well when all cells are lysed. The luminescent signal from before lysis was normalized to after lysis to determine relative cytotoxicity.

Cell viability assay

HEK 293T and N2a cells were grown to ∼75% confluency in a 96-well plate in DMEM media (4.5 g/l; Gibco, 10% FBS, 1% S/P) and transfected with the GFP-mCherry plasmids expressing either tRNA^Ser_AGA or tRNA^Ser_AAA variant. At 24 h post transfection, viability was measured using the cell counting kit 8 (CCK8) (Sigma–Aldrich, St. Louis, MO, United States). The assay uses a tetrazolium compound (WST-8) that produces an orange formazan dye upon reaction with cellular dehydrogenases. The level of dehydrogenase activity is a proxy for cell viability and proportional to the number of viable cells. Following manufacturer’s protocol, 10 μl of CCK8 assay reagent was added to each well containing 100 μl of media and incubated for 4 h at 37°C with 5% CO₂ and humidity. A BioTek Synergy H1 microplate reader was used to measure the absorbance at 450 nm in each well. The data were exported and analyzed in Microsoft Excel. For β-lymphocytes, ∼5 × 10⁴ cells were transferred in a 96-well plate with RPMI 1640 media (15% FBS; 1% S/P) with additional Ser concentrations ranging from 0 to 500 μg/ml as indicated and incubated at 37°C with 5% CO₂ and humidity. The CCK8 viability assays were conducted at 24, 48, and 72 h after the addition of serine supplemented media. Plates were gently agitated during incubation on a plate shaker (40 rpm).

Western blotting

HEK 293T and N2a cells grown to 75% confluency in a six-well plate were transfected. At 24 h post transfection, growth media was aspirated and cells were lifted in phosphate buffered saline (1 × PBS, pH 7.4; Corning Cellgro, Corning, NY, United States) supplemented with 1 mM ethylenediaminetetraacetic acid (EDTA), harvested by pipetting, and centrifuged in 1.5 ml microcentrifuge tubes at 900 × g for 3 min at 4°C. Supernatant was removed and cells washed with ice cold PBS (Corning Cellgro) and centrifuged again. Supernatant was removed and cells were suspended in 200 μl of mammalian cell lysis buffer (50 mM Na₂HPO₄, 1 mM Na₄P₂O₇, 20 mM NaF, 2 mM EDTA, 2 mM ethylene glycol bis(2-aminoethyl)tetraacetic acid (EGTA), 1 mM Triton X-100, 1 mM dithiothrieitol, 0.3 mM phenylmethylsulfonyl fluoride), incubated on ice for 10 min, and centrifuged at 21 000 × g for 10 min at 4°C. Supernatants were collected in 1.5 ml microcentrifuge tubes and kept on ice. Protein concentrations in the lysates were determined in triplicate using a Bradford Assay. Lysates of 20 μg of protein were separated via 12% sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE), including a protein size standard (Bio-Rad, Hercules, CA, United States). Proteins were transferred to methanol-activated polyvinylidene difluoride (PDVF) membranes using a Trans-Blot Turbo transfer system (Bio-Rad, 20 V and 1.3 A for 15 min). The PDVF membranes were then submerged in blocking solution [5% bovine serum albumin (BSA), 0.1% Tween-20, 1 × tris buffered saline (TBS)] for 1 h. Membranes were then incubated with anti-mCherry (ab213511, Abcam, Cambridge, United Kingdom) or anti-GAPDH (MAB374m, Sigma–Aldrich) diluted to 1:2000 in blocking solution and incubated overnight at 4°C. Membranes were washed 3 × 10 min in wash buffer (1% BSA, 0.1% Tween-20, 1 × TBS) followed by a 1 h incubation with goat anti-mouse secondary antibody (Li-cor Biosciences, Lincoln, NB, United States) or donkey anti-rabbit secondary antibody (Li-cor Biosciences) diluted to 1:5000. Membranes were washed with secondary wash buffer (2 × 10 min, in 1 × TBS-Tween, and once in 1 × TBS for 10 min). Membranes were imaged using a ChemiDoc MP imager (Bio-Rad) using the IRDye 680RD (Ex Red Epi Illumination, Em 715/30 Filter) for anti-GAPDH or IRDye 800 (Ex Epi Illumination, Em 835/50 filter) for anti-mCherry, using built in auto-exposure settings.

RFP-Trap purification of mCherry variants

At 24 h post transfection, HEK 293T and N2a cells grown in six-well plates were harvested, and washed as described in the western botting section and then re-suspended in 200 μl lysis buffer (50 mM Na₂HPO₄, 1 mM Na₄P₂O₇, 20 mM NaF, 2 mM EDTA, 2 mM EGTA, 1 mM Triton X-100, 1 mM dithiothrieitol, 0.3 mM phenylmethylsulfonyl fluoride), and centrifuged at 21 000 × g for 10 min at 4°C. Supernatants were collected in 1.5 ml microcentrifuge tubes and kept on ice. The mCherry protein was extracted using RFP-Trap Agarose kit for immunoprecipitation (Chromotek, Munich, Germany) according to manufacturer’s instructions. Cell lysates were mixed with 300 μl of dilution buffer (10 mM Tris–HCl, pH 7.5, 150 mM NaCl, 0.5 mM EDTA). Aliquots of 25 μl of RFP-Beads slurry (binding capacity of 22.5 μg) were equilibrated (washed 3 × 500 μl dilution buffer, centrifugation at 2500 × g for 5 min) and mixed with diluted lysates, and rotated end-over-end for 1 h at 4°C. Beads were washed 3× with 500 μl of wash buffer (10 mM Tris–HCl, pH 7.5, 150 mM NaCl, 0.05% Nonidet P40 Substitute, 0.5 mM EDTA). After the final wash, the beads were centrifuged at 21 000 × g for 10 min. The protein on the mCherry loaded beads were then separated by 12% SDS–PAGE and visualized with Coomassie blue dye. The purified GFP-mCherry protein was then analyzed by mass spectrometry. Additional methods for tandem mass spectrometry, ordered two-template relay sequencing (OTTR-seq) for tRNAs, and surface sensing of translation (SUnSET) assay are provided in the Supplemental Methods.

Statistical analyses

The data are presented as means of at least N = 3 biological replicates or more as annotated. Error bars represent the standard deviation of the mean. Statistical analyses were conducted in Microsoft Excel and included Tukey’s honestly significant difference (HSD) tests, Mann–Whitney U tests, single factor analysis of variance (ANOVA), or Student’s t-tests as indicated in each figure caption.

Results

Mistranslating tRNAs cause defects in protein production

The G35A variant occurs in natural human genomes in the tRNA^Ser_AGA 2–3 gene (TRS-AGA-2–3) with a 2% allele frequency and less frequently in the tRNA^Ser_AGA 2–2 gene (TRS-AGA-2–2) (<0.2% allele frequency). There are no examples of the G35A mutation in the other tRNA^Ser_AGA 2 genes. The six tRNA^Ser_AGA 2 genes (2–1 to 2–6) produce identical mature tRNAs, but these genes differ substantially in their upstream and downstream flanking sequences and genomic loci. To gain insight into the impact of the anticodon mutation in different tRNA^Ser_AGA-2 genes, we selected four of these tRNA^Ser genes (AGA-2–2, 2–3, 2–4, 2–5) and cloned each with ±300 bp of native flanking sequencing into a plasmid with a single mCherry fluorescent reporter (Supplementary Fig. S1). We [11, 12, 15] and others [21] have used the production of a fluorescent protein to measure the ability of different mistranslating tRNAs to reduce the level of over-produced fluorescent protein in mammalian cells. At 24 h post-transfection and using fluorescence microscopy, we captured images of individual K562 (Supplementary Fig. S1A), HEK 293T (Supplementary Fig. S1B), and N2a (Supplementary Fig. S1C) cells and used the images to determine the level of mCherry fluorescence per cell (Supplementary Fig. S1D–F).

In agreement with our previous findings [14], cells expressing the tRNA^Ser_AAA variants showed reduced levels of mCherry fluorescence compared to cells expressing the counterpart wildtype tRNA^Ser_AGA gene or cells expressing no additional tRNA from the mCherry plasmid. Western blotting for the mCherry protein (Supplementary Fig. S1G) confirmed reduced mCherry protein levels in cells expressing tRNA^Ser_AAA. The data show that the tRNA^Ser_AGA 2–3 mutant and even more so the 2–5 anticodon mutant displayed the most significant reductions in mCherry levels compared to the AAA anticodon in the 2–2 and 2–4 tRNA^Ser genes. The tRNA mutants also showed a greater impact on mCherry levels in HEK 293T and N2a cells relative to K562 cells. Together, the data suggest the cell type and gene in which an anticodon mutant is located will affect the phenotypic impact of the mutant tRNA.

Because certain anticodon mutant tRNAs can slow protein synthesis [21] by inhibiting translation initiation via the integrated stress response, we probed HEK 293T cells transfected with wildtype or mutant tRNA for eIF2a and phospho-eIF2a levels (Supplementary Fig. S2). We observed no changes in eIF2a levels or phosphorylation in cells expressing wild type compared to mutant tRNA. Thus, the data suggest the integrated stress response is not activated by the tRNA^Ser_AAA mutants. To further validate this result, we also measured global protein synthesis levels using a puromycin-based labeling assay (Supplementary Fig. S3) and found that cells expressing the tRNA^Ser_AAA 2–3 or 2–5 anticodon mutants had no significant impact on global protein levels relative to un-transfected cells. As an additional control, we applied the proteosome inhibitor MG132 to un-transfected cells and observed an expected reduction in protein levels [15]. We did observe a small but significant reduction in protein levels in cells transfected with the mCherry S151F mutant and the wildtype tRNA^Ser_AGA 2–5 (Supplementary Fig. S3D). The data are in agreement with our previous findings on tRNA^Ala anticodon mutants [35], and re-affirm that the fluorescent protein marker (as in Supplementary Fig. S1) is a sensitive assay for tRNA-induced defects in protein production in mammalian cells.

Cytotoxicity and viability in cells expressing mistranslating tRNA^Ser variants

HEK 293T and murine N2a cells were transfected with a plasmid co-expressing mCherry and no additional tRNA or the indicated tRNA^Ser_AGA or G35A variant. We observed no significant cytotoxicity compared to the control in HEK 293T cells (Fig. 2A), yet each of the mutant tRNAs caused significant increases in cytotoxicity in murine N2a cells (Fig. 2B) compared to cells transfected with wildtype tRNA or with no additional tRNA. To determine if the tRNA variants caused defects in cell proliferation, we also conducted cell viability assays in HEK 293T cells (Fig. 2C) and in N2a cells (Fig. 2D) transfected with wildtype or AAA anticodon mutants in the 2–3 and 2–5 tRNA^Ser gene contexts. Both mutant tRNAs led to a similar and significant reduction in cell viability in HEK 293T cells (Fig. 2C), yet we observed no significant difference in murine N2a cells expressing the anticodon variants. Thus, the expression of the tRNA^Ser_AAA variants led to measurable but distinct phenotypic defects in human as opposed to murine cells. We did not observe a correlation between the level of defects in toxicity or viability and the level of impact on mCherry protein production noted above. For example, although the 2–5 tRNA mutant caused a greater reduction in mCherry production than the 2–3 tRNA variant, the impact on toxicity in N2a cells and viability in HEK 293T cells were similar for the G35A mutant in two different tRNA gene contexts.

Figure 2.

Cytotoxicity and cell viability in mistranslating human and murine cells. Plasmids encoding mCherry and wild-type human tRNA^Ser_AGA or variant tRNA^Ser_AAA with different flanking sequence contexts from human tRNA^Ser_AGA 2–2, 2–3, 2–4, and 2–5 genes were transfected into (A) human HEK 293T and (B) murine N2a cells; (A, B) cytotoxicity was determined at 24 h post-transfection using CytotoxGlo. (C, D) Cell viability (CCK8 assay) was also measured in (C) HEK 293T cells and in (D) N2a cells at 24 h after transfection with a plasmid expressing the wildtype GFP-mCherry reporter and the indicated wildtype tRNA^Ser_AGA or AAA anticodon mutant. Error bars represent ±1 standard deviation of the mean and statistical analysis was calculated using single factor ANOVA and was based on at least N = 3 biological replicates (n.s., not significant; *P <.05).

Design and characterization of a dual-fluorescent reporter for Ser mis-incorporation in live cells

To measure mistranslation levels for the tRNA variants in different cell types, we designed a dual fluorescent reporter including a mCherry mutant that is sensitive to Ser mis-incorporation. We had previously found that mCherry mutant Ser151Pro provided a quantitative readout for tRNAs that mis-incorporated Ser at Pro codons in E. coli [29]. We had also established that mCherry Ser151Phe was produced in E. coli but lacked any detectable red fluorescence [29]. Here, we encoded the UUU Phe codon at position 151 in mCherry in a pCDNA3.1-derived plasmid with a cytomegalovirus major immediate-early gene (CMV) promoter driving expression of the GFP-mCherry fusion protein (WT-PAN) [32]. The eGFP acts as a control to measure protein production levels, which can vary between individual cells and is essential to control for our observation that mistranslating cells have a reduced ability to over-produce protein (Supplementary Fig. S1). Fluorescence restoration of the mutant mCherry gene will report on serine mis-incorporation at the Phe151 codon. The GFP-mCherry fusion protein contains 22 additional Phe codons that are all UUC. Although mis-incorporation at these sites could in principle impact fluorescence of mCherry or GFP, our studies below confirm that mistranslating cells produce GFP-mCherry at reduced levels, but the fluorescence ratio of mCherry/GFP for the wildtype fusion protein are not substantially different between normal and mistranslating cells.

Although we had shown that the mCherry mutant Ser151Phe inactivated red fluorescence [14], we had not yet tested the ability of a tRNA variant to revive the mCherry mutant fluorescence before. Thus, we cloned human tRNA^Ser alleles containing the wild-type (tRNA^Ser_AGA 2–3 or tRNA^Ser_AGA 2–5) or corresponding G35A variant encoding tRNA^Ser_AAA into a plasmid that co-expresses GFP-mCherry or the GFP-mCherry Ser151Phe mutant. We then used transfection experiments in HEK 293T and N2a cells to test our hypothesis that tRNA^Ser_AAA will revive the fluorescence of the mCherry Ser151Phe mutant in proportion to the level of mistakes at that site.

We transfected the resulting plasmids into HEK 293T (Fig. 3) and N2a (Supplementary Fig. S4) cells to determine the level of mistranslation produced by the 2–3 and 2–5 tRNA variants. At 24 h after transfection, fluorescence microscopy was used to image cells for GFP and mCherry fluorescence and bright field images were also acquired. Images were taken from three different fields of view for each of N = 3 biological replicates for each condition, and the data are based on GFP and mCherry levels from a total of ∼500–1000 individual cells per condition. To establish a scale to measure mistranslation levels, we normalized the average mCherry/GFP ratio in cells expressing the wildtype reporter and the mutant tRNA to 1.0. We reasoned that because mistranslating cells produce less GFP-mCherry than cells transfected with wildtype tRNA, the maximal amount of mCherry (wildtype, WT) that can be produced in mistranslating cells is the appropriate reference for the relative level of mCherry (S151F) recovery that we observed in mistranslating cells.

Figure 3.

Quantifying mistranslation levels in live human cells. HEK 293T cells were transfected with a plasmid co-expressing GFP fused to mCherry (WT) or to mCherry S151F and wildtype tRNA^Ser_AGA or mutant tRNA^Ser_AAA. (A) Representative images of brightfield, GFP, and mCherry fluorescence are shown alongside (B) zoomed-in views (white boxes in A) showing GFP, mCherry, and merged fluorescent images of the indicated cell lines. Western blotting of the reporter with mCherry and GAPDH (loading control) specific antibodies from cells expressing wildtype or anticodon variants of (C) tRNA^Ser_AGA 2–5 or (D) tRNA^Ser_AGA 2–3 show the full length GFP-mCherry protein (54 kDa) and a lower molecular weight proteolytic product that occurs in mCherry [29]. (E, F) The mistranslation level is determined by plotting the ratio of mCherry/GFP fluorescence per cell. The images in (A) and in Supplementary Fig. S5 were quantified and normalized to the mCherry/GFP fluorescence ratio observed in mistranslated cells expressing the wildtype reporter. Error bars represent ±1 standard deviation of the mean and statistical analysis was calculated using single factor ANOVA and was based on N = 3 biological replicates (n.s., not significant; *P <.05; **P <.01).

Since we found no substantial difference between the mCherry/GFP ratio in cells expression wildtype versus mutant tRNA, our data suggest that normalization to wildtype or mistranslating cells will lead to equivalent resulting mistranslation levels. Although cells expressing mutant tRNA produce less GFP-mCherry protein per cell relative to cells expressing wildtype tRNA (Supplementary Figs S1 and S4, and Fig. 3), we found that the mCherry/GFP ratios were not significantly different in comparing human cells expressing the wildtype GFP-mCherry reporter and either of the wildtype or mutant tRNAs according to microscopy (Fig. 3) and flow cytometry (Fig. 4). In murine N2a cells expressing the mutant tRNA, we observed a small but significant increase in the mCherry/GFP ratio of the wildtype reporter relative the same reporter observed in cells expressing the wildtype tRNA (Supplementary Fig. S4). We note that this same difference was not significant according to more accurate measurements made using flow cytometry and described below (Fig. 4).

Figure 4.

Quantifying mistranslation levels in live human and murine cells by flow cytometry. Flow-cytometry was used to measure GFP and mCherry fluorescence in 2.5 × 10⁶ individual viable HEK 293T (above) or murine N2a cells (below) transfected with the wildtype GFP-mCherry or GFP-mCherry S151F mutant and the indicated wildtype tRNA^Ser_AGA or tRNA^Ser_AAA variant. (A) The fluorescence measurements were converted into heatmaps to depict the population of cells observed across mCherry and GFP fluorescence levels. (B) To determine the mistranslation level resulting from each tRNA mutant in the indicated cell lines, the mCherry/GFP fluorescence ratio from individual cells were calculated and plotted. The data were normalized to the wildtype mCherry/GFP fluorescence ratio observed in mistranslating cells. Error bars represent ±1 standard deviation of the mean and statistical analysis was calculated using single factor ANOVA and was based on N = 3 biological replicates (n.s., not significant; **P <.01; ***P <.001; ^****P <.0001).

Because we observed a range of GFP fluorescence intensity across individual cells and conditions, we used both long and short exposure settings (Supplementary Fig. S5) to capture images. Analysis of the data obtained under both settings led to indistinguishable mCherry/GFP ratios, thus, we used short exposure images to avoid saturation from highly expressing cells.

Following cell imaging, western blotting of protein harvested from the same cells confirmed expression of the GFP-mCherry reporter in each cell line and the anticipated reduced production of the reporter in mistranslating cells (Fig. 3C and D, and Supplementary Fig. S4C and D). As a result of a well-known proteolytic product of mCherry, both full length and cleaved GFP-mCherry protein products were observed as before [29]. In cells expressing the same tRNA variant, the wildtype and Ser151Phe mutant reporters were produced at similar levels.

In cells expressing the wildtype tRNA and the mutant GFP-mCherry Ser151Phe reporter, we observed similar GFP levels as observed for the wildtype reporter, but there was no detectable red fluorescence above the background level (Fig. 3, and Supplementary Figs S4 and S5). In contrast, we recorded significant levels of red fluorescence restoration of the Ser151Phe reporter only in cells also expressing the tRNA^Ser_AAA mutants. We then quantified the mistranslation levels and found a similar level of ∼10% mistranslation at the Phe151 codon resulted from the 2–3 and 2–5 tRNA mutants in both murine and human cells (Fig. 3E and F, and Supplementary Fig. S4E and F). In HEK 293T cells, we found the tRNA^Ser_AAA 2–3 (9.2 ± 1.6% mis-incorporation of Ser at Phe151) and 2–5 (10.6 ± 1.2%) variants produced similar and statistically indistinguishable mistranslation levels. These levels were also consistent with our observations in murine cells for the 2–3 (10.9 ± 1.0%) and 2–5 (10.0 ± 0.7%) tRNA variants (Table 2). The data demonstrate tRNA-dependent restoration of the mCherry Ser151Phe mutant, and the dual fluorescent reporter enabled quantitative measurements of mistranslation levels resulting from two different tRNA variants in human and murine cells.

Table 2.

Mistranslation levels in human and murine cells

Detection method/tRNA variant	Mistranslation level in HEK 293T cells	Mistranslation level in N2a cells
Fluorescence microscopy		​
tRNASerAAA 2–5	10.6 ± 1.2%	10.0 ± 0.7%
tRNASerAAA 2–3	9.2 ± 1.6%	10.9 ± 1.0%
Flow cytometry
tRNASerAAA 2–5	12.5 ± 0.2%	7.6 ± 0.1%
tRNASerAAA 2–3	11.4 ± 0.1%	9.8 ± 0.1%

Measuring mistranslation in live mammalian cells with flow cytometry

Fluorescence microscopy is an excellent approach to visualize mistranslation levels in live cells, yet the number of cells analyzed (∼10³) represents a fraction of a larger cell population. Moreover, the fluorescence microscopy imaging cubes have a limited dynamic range [36], and different individual cells express the reporter over a wide range of levels. To address these issues, we analyzed the same cell lines as investigated above by flow cytometry and recorded GFP and mCherry levels from 2.5 × 10⁶ individual control and mistranslating cells. HEK 293T and N2a cells were transfected with plasmids co-expressing the GFP-mCherry or GFP-mCherry Ser151Phe reporters with wildtype tRNA^Ser 2–3 or 2–5 or the respective AAA anticodon mutants. At 24 h post transfection, cells were sorted and separated based on flow cytometry standard gates (Supplementary Fig. S6) to isolate GFP positive cells. The data are plotted using heat maps to show the population of cells across a broad range of GFP and mCherry fluorescence intensities (Fig. 4). HEK 293T cells expressing the wildtype reporter and either the wild type or mutant tRNAs, all show a clear pattern in which most of the cells have a 1:1 ratio of mCherry to GFP fluorescence (Fig. 4). In HEK 293T cells, we observed a linear trendline between mCherry and GFP fluorescence across the entire range of fluorescence intensity with a minority of cells showing some dispersion from linearity. While the same linear trend holds on average in murine cells, the population of lower fluorescing cells did not strictly conform to a linear relationship perhaps due to greater morphological diversity of N2a cells, which can spontaneously differentiate into neurons [37], relative to the morphologically homogenous HEK 293T cells. Nevertheless, the average mCherry/GFP ratio for the wildtype reporter in normal and mistranslating cells was not significantly different from 1.0 in any case according to flow cytometry (Fig. 4).

In cells expressing the mutant reporter and wildtype tRNA, we recorded significant GFP fluorescence with no evidence of mCherry fluorescence signal above background. Significant restoration of the mCherry signal was evident in mistranslating cell lines that also expressed the mutant GFP-mCherry reporter. Analysis of the mCherry/GFP ratios from the flow cytometry data revealed more accurate measurements of mistranslation (0.1%–0.2% error) relative to microscopy (0.7%–1.6% error) (Table 2), and the mistranslation levels were similar to the 10% level we recorded using microscopy. In HEK 293T cells, we recorded mistranslation levels for tRNA^Ser_AAA 2–3 (12.5 ± 0.2%​) and 2–5 (11.4 ± 0.1%​) variants based on the flow cytometry data (Fig. 4B). For N2a cells, the reported mistranslation level for the 2–3 (9.8 ± 0.1%) and 2–5 (7.6 ± 0.1%)​ variants was reduced compared that observed in human cells. There was no significant difference between the mistranslation level produced by the tRNA^Ser 2–3 versus the 2–5 derived anticodon mutant.

Given the large range of GFP-mCherry reporter expression levels observed in the transfected cells, we wondered if there was a relationship between mistranslation level and the ability of individual cells to express the reporter. We then plotted the level of mis-incorporation observed in individual cells as a function of the total GFP level observed in the same cell (Fig. 5). Fascinatingly, we found that the higher expressing cells showed dramatically lower levels of mCherry recovery, while low expressing cells showed higher levels of mis-incorporation. The relationship between mistranslation level and reporter expression level follows an inverse power law relationship that is nearly identical comparing cells expressing the tRNA^Ser_AAA derived from the 2–3 and 2–5 tRNA^Ser genes. In murine N2a cells, we observed the same inverse power law relationship (Supplementary Fig. S7). The data indicate that high levels of reporter expression correlate with reduced levels of detectable translation error (∼0.1%) compared to low expressing cells that showed higher levels of error, including cells far exceeding the 10% average error level. The data suggest that the aminoacylated mutant tRNA is a limiting factor that is able to generate many mistakes at low reporter expression levels, but the mutant tRNA may be more effectively outcompeted by cognate phenylalanyl–tRNA^Phe in cells expressing the reporter at high levels. Our future studies will investigate the nature of the tRNA pool in these different cell populations.

Figure 5.

Mistranslation levels in individual human cells as a function of dual reporter expression. Based on the data from the flow cytometry analysis (Fig. 4), the logarithm of the mistranslation level observed in individual HEK 293T cells expressing the tRNA^Ser_AAA (A) 2–3 or (B) 2–5 variant was plotted as a function of the GFP fluorescence level observed in the same cells. (C) The plots were overlaid to highlight reduced GFP fluorescence levels in the cells expressing the 2–5 variant (as in Supplementary Fig. S1) and reveal a nearly identical inverse power-law relationship (tRNA^Ser_AAA 2–3 blue solid line; tRNA^Ser_AAA 2–5 dark orange dashed line) between mistranslation level and the expression level of the reporter for both tRNA variants.

Mass spectrometry confirms mistranslation in the GPF-mCherry reporter

To validate Ser mis-incorporation in the mCherry Ser151Phe protein, the dual reporter was isolated from normal and mistranslating HEK 293T cells using an RFP-trap affinity column. In cells expressing GFP-mCherry Ser151Phe and the tRNA^Ser 2–3 (Fig. 6A) and 2–5 (Fig. 6B) anticodon mutants, we readily observed matching peptides containing either Ser because of mistranslation or Phe resulting from accurate decoding of Phe151 (Fig. 6C).

Figure 6.

Mass spectrometry of GFP-mCherry S151F isolated from human cells. The mCherry S151F protein was isolated from HEK 293T cells expressing the indicated wildtype tRNA^Ser_AGA or G35A anticodon variant. Y and b ion spectra of mistranslated peptides (above, f = Ser incorporated at Phe151) are compared to matching properly translated peptides (below, F = Phe incorporated at Phe151) derived from cells expressing (A) tRNA^Ser_AAA 2–3 or (B) tRNA^Ser_AAA 2–5. (C) Peptide coverage maps are shown for the peptide surrounding the Phe151 mutation site. Blue lines indicate the expected peptide sequence listed at the top, while the boxed f annotates Ser mis-incorporation at F151. The apparent level of Ser mis-incorporation at (D) Phe151 and (E) at all Phe codons in the GFP-mCherry protein was estimated from the mass spectrometry data based on the relative area under the monoisotopic peak for a mistranslated relative to a properly translated peptide found in the protein samples extracted from the indicated HEK 293T cell lines. (F) We also used spectral counts, i.e. the number of mistranslated relative to properly translated peptides identified in each sample, as an alternate approach to estimate mis-incorporation level at the Phe151 codon. Error bars represent ±1 standard deviation of the mean and statistical analysis was calculated using single factor ANOVA and was based on N = 3 biological replicates (n.s., not significant; *P <.05).

To estimate mis-incorporation levels from the mass spectrometry data, we compared the area under the curve of the monoisotopic peak of mistranslated peptides relative to that of matching and properly translated peptides (as in Fig. 6A and B). Based on this comparison, we estimated a level of ∼0.05% mis-incorporation at the Phe151 site (Fig. 6D). In addition to mis-incorporation at Phe151, we identified mistranslation from the 2–3 and 2–5 tRNA mutants with Ser observed at all but 6 of the 22 Phe codons in the GFP-mCherry reporter. We then estimated the level of Ser mis-incorporation at each Phe codon in GFP-mCherry Ser151Phe and found a greater range of mistranslation from 0.01%–10% with the average level of ∼0.5%–1% Ser mis-incorporation at Phe codons (Fig. 6E). Finally, we used spectral counting as an alternative approach to estimate mis-incorporation. In this case, we calculated the ratio of the number of peptides (ion counts) observed with a mistranslated Phe151 (Ser151) to the number of peptides observed with Phe151. Spectral counting suggested a level of 0.3% mis-incorporation of Ser at Phe151 (Fig. 6F). In all cases the MS/MS analysis showed significantly more evidence of mistranslation in cells expressing the mutant tRNA compared to cells expressing wildtype tRNA, in which we found no evidence of high confidence peptide hits representing mistranslation. Although the MS/MS analysis clearly and unambiguously identified mistranslation at the Phe151 codon and at most of the Phe codons in the reporter, the estimation of mistranslation level using either area under the isotopic peak or spectral counting led to a drastic underestimation of the actual level of ∼10% we measured using the reporter in live HEK 293T cells.

As noted in the section below, the A34 base in tRNA^Ser_AAA is modified to inosine. The I34 base can in principle read codons ending in U, C, or A. Thus, explaining our finding ample evidence at the UUC codons and at UUU151 in the GFP-mCherry reporter. Although the tRNA variant could potentially also decode Leu UUA codons as Ser, our reporter contains no UUA codons.

tRNA sequencing of mistranslating HEK 293T cells quantifies tRNA^Ser_AAA levels and hypomodification

To characterize the expression level of the mutant tRNA and its potential impact on the tRNAome, we conducted OTTR-Seq [38] (see Supplemental Methods) to quantitate all tRNA transcripts in the HEK 293T cell line transfected with wildtype or mutant tRNA^Ser variants (Supplementary Fig. S8, Supplementary Data File 1). In humans, nine tRNA^Ser_AGA genes contribute to produce tRNA^Ser_AGA transcripts referred as the tRNA^Ser_AGA pool. Our plasmid expressed tRNA^Ser_AAA variants account for ∼35% of the tRNA^Ser_AGA pool, and the AAA anticodon mutant was not detected in cells transfected with wildtype tRNA^Ser (Supplementary Fig. S8A). The tRNA^Ser 2–3 and 2–5 derived anticodon mutants were expressed at similar levels that were not significantly different (Supplementary Fig. S8A and B). When considering all 25 tRNA^Ser genes and their transcripts representing four tRNA^Ser isodecoder families, the mutant tRNA^Ser accounts for ∼15% of the total cellular tRNA^Ser pool. The tRNA^Ser_AAA is aminoacylated by seryl-tRNA synthetase like the endogenous tRNA^Ser species, but on the ribosome the mutant tRNA competes for decoding Phe codons with the population of tRNA^Phe species in the cell. Interestingly, the 10 tRNA^Phe genes are expressed at a significantly lower level than tRNA^Ser, thus, we estimate the mutant tRNA^Ser_AAA comprises ∼75%–82% of the pool of Phe-decoders in the cell (Supplementary Fig. S8B).

In addition to identifying the A35 nucleotide variants in our mutant tRNAs, OTTR-seq also records nucleotide mismatches that are the result of post-transcriptionally modified bases in the tRNAs. The tRNA sequencing data, thus, allowed us to quantify base modification levels for 3-methyl-C32 (m³C32) (Supplementary Fig. S8C), inosine 34 (I34) (Supplementary Fig. S8D), and methylations of bases G26, Ce2, and A58 (Supplementary Fig. S8E). We observed that in the mutant tRNA, methylation of C32 was substantially reduced to 28% modified in the mutant compared to 82%–95% modified in wildtype tRNA^Ser_AGA (Supplementary Fig. S8C). Inosine at position 34 in the mutant tRNA, which is important for wobble decoding, shows a small but significant 4% reduction in I34 compared to wildtype tRNA, which is close to 100% modified. A smaller and nonsignificant reduction in I34 was observed for the 2–3 mutant (Supplementary Fig. S8D).

We then analyzed the tRNA sequencing data to identify changes in the relative expression of all tRNA transcripts in comparing HEK 293T cells transfected with wildtype or mutant tRNA^Ser (Supplementary Fig. S9). The 2–3 tRNA mutant caused significant and more than two-fold upregulation of several tRNAs, including Arg-TCG-1 and -5, Arg-TCT-1, Cys-GCA-16, Ala-AGC-4 and -8, Gly-TCC-3 and -4, while just one tRNA (Leu-CAA-2–1) was downregulated in cells expressing the 2–3 mutant (Supplementary Fig. S9A). In contrast, compared to cells expressing wildtype tRNA, cells expressing the 2–5 anticodon mutant had several tRNAs downregulated more than two-fold, including Lys-TTT-3–1, Leu-CAA-2–1, Thr-AGT-1–3, Ile-AAT-5–3, and Arg-TCG-1–1. (Supplementary Fig. S9B). Comparison of the tRNA levels between the two mutant lines (Supplementary Fig. S9C) and the two wildtype cell lines (Supplementary Fig. S9D) showed only 2 or 3 significantly changed tRNAs, indicating minor changes relative to the comparison of cells transfected with wildtype versus mutant tRNA. Although the mechanisms by which tRNA mutants impact the levels of other tRNAs are not yet known, the fact that several tRNAs are downregulated in cells transfected with tRNA^Ser_AAA 2–5 may explain the greater defect in protein over-production caused by the 2–5 mutant relative to the other tRNA^Ser mutants we investigated (Supplementary Fig. S1).

tRNA sequencing confirms expression of the tRNA^Ser G35A variant in pan-genome cell lines

Because the human tRNA^Ser_AGA 2–3 G35A variant (rsid: rs147439337) is so frequently found in human genomes, we readily identified 89 β-lymphocyte lines in the Ensembl database [39] derived from individuals that participated in the 1000 genomes project that are available in the Coriell Institute biobank and that have the A35 genotype. We acquired cells derived from males and females, including 5 cell lines from individuals with the wildtype G35 genotype and 7 lines with the heterozygous A35 minor allele. Six of these A35 lines have the G35A variant in the tRNA^Ser_AGA 2–3 gene, while a single variant (HG01468), and the only one available in the database, has the G35A variant in the tRNA^Ser_AGA 2–2 gene (rsid: rs571240633) (Table 1). Although the number and composition of cell lines is not sufficient at this stage of research to make conclusions on the basis of sex as a biological variable, in future we will expand our studies to examine potential differences between cells derived from males and females. To confirm the genotype of each cell line, we PCR amplified the genomic region ±300 base pairs surrounding the tRNA^Ser_AGA-2–3 gene. In each case, DNA sequencing confirmed the presence of the expected wildtype allele as well as the heterozygous nature of individuals with the A35 variant. Peaks corresponding to both G and A at the N35 site were observed in the variant cell lines (Supplementary Fig. S10).

To establish whether the mutant tRNAs were expressed in these cells and measure the levels of the A35 variant, we conducted OTTR-seq on β-lymphocytes with the G/G35 (Supplementary Fig. S11) and G/A35 genotypes (Supplementary Fig. S12, Supplementary Data File 1). In each of the variant cell lines, we unambiguously detected the A35 variant in the mature tRNA^Ser_AGA-2 pool (Fig. 7 and Supplementary Fig. S12). In comparing each of the variant cell lines, we found varying levels of tRNA^Ser_AAA expression. Cell ID 821 (see Table 1) showed minimal expression, while more substantial levels were observed in the 518, 800, and 876 cell lines. We found that each of the A35 variant cell lines expressed the tRNA^Ser_AAA mutant at a level of 0.5%–4% as a fraction of the tRNA^Ser_AGA pool (Fig. 7A). Compared to the total tRNA^Ser pool in the cells, tRNA^Ser_AAA accounts for 0.1%–1%. As noted above, the mutant tRNA competes with the Phe-decoding tRNA pool for translation of UUU/C Phe codons. The mature tRNA^Ser_AAA in the variant cell lines represent 1%–6% of the tRNA^Phe pool (Fig. 7B). This is the first report of expression of an endogenous, genome encoded, and naturally occurring tRNA anticodon variant in human cells.

Figure 7.

tRNA sequencing quantifies expression and modification levels of genome-encoded wildtype and anticodon variant tRNA^Ser in human β-lymphocytes. Ordered Two-Template Relay (OTTR) sequencing performed on RNA purified from human β-lymphocytes confirms expression of a G35A tRNA^Ser variant in each cell line with the G/A genotype. In cells with the G/G genotype, only tRNA-Ser-AGA-2 reads containing G35 (blue bars) were identified, whereas in cells with the G/A genotype, both G35 (grey) and A35 (orange stripes) containing reads were found. (A) Quantification of the tRNA^Ser_AAA level as a percentage of the endogenous tRNA^Ser_AGA pool (expressed from 9 genes) and (B) compared to the level of the other Phe-decoding tRNAs (endogenous tRNA^Phe are expressed from 10 genes). (C) Normalized tRNA sequencing coverage profiles from all cell lines expressing G35 tRNA^Ser_AGA-2 reads (grey) and A35 tRNA^Ser_AAA reads (orange stripes) illustrates the appearance of a 5′-fragment at residue 56 in the A35-contining tRNA reads. (D) Analysis of the reads show the G35A variant is hypo-modified at m³C32, while wildtype tRNA^Ser was mostly modified. (E) Quantification of inosine (I₃₄) levels show that two of the A35 cell lines have significantly lower inosine modification levels. (F) We observed generation of a 5′-fragment of tRNA^Ser_AGA, which occurred less frequently in G35 containing reads and was found in generally higher and varying levels in the A35 reads in different cell lines. Error bars represent ±1 standard deviation of the mean and statistical analysis was calculated using Student’s t-tests and was based on N = 3 biological replicates for (A–C) modification and fragment levels, or a (D, E) Mann–Whitney U test for tRNA levels (n.s., not significant; *P <.05; **P <.01; ***P <.001; ^****P <.0001).

We then analyzed the tRNA-seq data for potential defects in tRNA^Ser_AAA modification and processing. Fascinatingly, we identified defects both in anticodon loop modifications and in tRNA fragmentation. Just as we found in HEK 293T cells, each of the endogenous tRNA^Ser_AAA  variants was almost or completely deficient in methylation at base C32 (Fig. 7D). The variant cell line 821 showed ∼25% modification at C32, while the other variant cells lines were <10% modified at this site in the mutant tRNA. In contrast, reads with the G35 allele, either from wildtype or variant cell lines, showed high levels (>70%) of m³C32 modification. In two cell lines with the A35 minor allele, we also identified defects in A34 to I34 modification (Fig. 7E). The remaining variant cells showed no significant defects in I34 levels, and similarly as above for C32 modification, the wildtype tRNAs were fully modified at I34.

Unlike our findings in HEK 293T cells, we observed substantial 5′-fragment generation from the tRNA^Ser_AAA variant (Fig. 7F). The 5′-fragment includes bases 1–56 (Supplementary Fig. S12) and lacks the remaining 3′ segment of the tRNA. Such a fragment cannot participate in translation but may have other functions in regulating gene expression (reviewed in [40]). In each cell line, there was significantly more fragmentation of the mutant relative to the wildtype tRNA, and the level of fragmentation varied from 5% to > 75% of tRNA^Ser_AAA lacking the 3′-end. In two variant cell lines (800, 876) and in one of the wildtype cell lines (508), we observed a low level (20%–25%) of fragmentation of the wildtype tRNA.

tRNA variant cell lines are associated with slow growth

To assess potential phenotypic defects in cells expressing the endogenous tRNA^Ser_AAA variant, we conducted cell viability studies over a 72-h time course (Supplementary Fig. S13). We used a CCK8 assay that measures the activity of cellular dehydrogenases, which is a proxy for cellular viability. For β-lymphocytes representing the wildtype G/G35 genotype (cell ID 0096, 845, 847) and the G/A35 variant genotype (cell ID 876, 878, 896), ∼5 × 10⁴ cells were seeded in RPMI 1640 media without or with added Ser in a final concentration ranging from 10 to 500 μg/ml (0.9–47.6 mM), and cell viability was measured at 24, 48, and 72 h (Supplementary Fig. S13). In normal media conditions (containing 0.29 mM Ser) and without additional Ser, we observed significantly lower levels of cell viability in each of the variant cell lines relative to cells with the G/G35 genotype.

To determine if increased Ser levels would exacerbate the phenotypic defect that we observed in the G35A cell lines, we assessed viability at increasing Ser concentrations. The viability defect observed in normal media was evident at each Ser concentration (Supplementary Fig. S13A–C) and in total growth (Supplementary Fig. S13D), which we estimated as the area under the cell viability curves. We observed no significant impact on Ser until the maximal concentration (500 μg/ml), which compromised the viability of all cell lines. In agreement with our findings in HEK 293T expressing tRNA^Ser_AAA variants (Fig. 2C), the mutant cell lines showed reduced viability compared to cells with the G35 genotype, however, the phenotypic defect was not further exacerbated by the addition of Ser.

Visualization and quantitation of mistranslation in human pan-genome cell lines

While the expression levels of genome encoded tRNA^Ser_AAA were substantially lower than the ectopically expressed tRNA mutant in our studies in HEK 293T and N2a cells, we have shown before that tRNAs that represent a small proportion of the tRNA pool can participate in translation [41]. Thus, we transfected the β-lymphocyte lines with our dual fluorescent reporter. For the transfection experiments, we used a plasmid expressing the wildtype GFP-mCherry or mutant GFP-mCherry Ser151Phe reporters with no additional tRNA cassette. The β-lymphocytes grow in suspension culture, and they are difficult to transfect [42]. Indeed, our first attempts to transfect the reporter plasmid using lipofectamine failed. Thus, we used a TransIT-X2 no-liposomal polymer nanoparticle engineered to improve transfection efficiency in many cell types. This approach improved transfection, however, the transfection efficiency and the level of GFP-mCherry reporter expression was greatly reduced in β-lymphocytes compared to our observations in HEK 293T cells. Nevertheless, we readily detected the GFP-mCherry signals in these cells at 24 h after transfection (Fig. 8, and Supplementary Figs S14–S16).

Figure 8.

Human β-lymphocytes transfected with the GFP-mCherry S151F reporter show mistranslation from genomically encoded tRNA^Ser anticodon variants. Human β-lymphocytes were grown in suspension culture and transfected with a plasmid only expressing GFP-mCherry or GFP-mCherry S151F and lacking any ectopically expressed tRNA cassette. At 24 h after transfection, brightfield, GFP, and mCherry fluorescence images were captured (Supplementary Figs S14 and S15) and a custom ImageJ Script (Supplementary Appendix) was used to quantify the GFP and mCherry fluorescence intensity from individual transfected cells (Supplementary Fig. S16). Representative brightfield merge, GFP, and mChery images are shown for the (A) wildtype genotype (G/G) and (B) the A35 variant genotypes (G/A). (C) The ratios of mCherry/GFP fluorescence per cell area were then plotted from the indicated cell lines transfected with the wildtype (green striped bars) or with the GFP-mCherry S151F mutant (orange striped bars) reporter. Contrast and brightness of images were adjusted equally in all panels post-analysis for ease of visualization. Error bars represent ±1 standard deviation of the mean and statistical analysis was calculated using ANOVA single factor test and was based on at least N = 3 biological replicates (n.s., not significant; ^****P <.0001). Significant differences between mCherry/GFP and mCherryS151/GFP ratios are indicated by horizontal lines below the asterisks, while the differences between mCherry S151/GFP restoration in cells with the G/A35 genotype versus the G/G35 genotype are indicated by asterisks above the orange striped bars.

The β-lymphocytes in suspension culture grow as spheroids or clusters of various sizes. We observed GFP and mCherry expression from the reporter varied within and between the cell clusters (Fig. 8A and B, and Supplementary Figs S14–16). In imaging analysis, we treated each cluster as a ROI and normalized the GFP and mCherry signals by the area of the ROI. Like our studies in HEK 293T and N2a cells, we observed a consistent ∼1:1 ratio of GFP to wildtype mCherry fluorescence in each cell line (Fig. 8C and Supplementary Fig. S16A). For the mutant reporter (GFP-mCherry Ser151Phe), while the GFP signal was clearly visible in each cell line, we observed no fluorescence recovery in cells with the wildtype G35 genotype (Fig. 8A, and Supplementary Figs S14 and S16B). We did, however, observe significant rescue of the mCherry Ser151Phe fluorescence in each of the cell lines tested that expressed the endogenous tRNA^Ser_AAA (Fig. 8B, and Supplementary Figs S15 and S16B). After normalizing to GFP, we calculated mistranslation levels for the A35 variant cell lines 878 (35.8 ± 21.5%), 876 (11.0 ± 8.5%), and 896 (33.0 ± 22.0%)​ (Fig. 8C and Table 3). We note that because the reporter was expressed at much lower levels in β-lymphocytes than in HEK 293T or N2a cells coupled with our observations that lower reporter expression levels strongly correlate with higher mistranslation levels (Fig. 5 and Supplementary Fig. S7), may together help explain the relatively high levels of mistranslation observed in some individual β-lymphocytes. Although the average level of mistranslation observed in the β-lymphocytes appeared higher than in HEK 293T and N2a cells, the variance in the measurement showed that these levels, while significantly above background (Fig. 8C), were not significantly different from the 10% level we recorded above. While we will refine this approach in future studies, perhaps by genomically integrating the reporter in these cell lines, our studies here conclusively demonstrate elevated levels of Ser mis-incorporation at Phe codons in cells that we showed express the anticodon variant tRNA relative to cells that express only wildtype tRNA^Ser. Our experiments show that endogenous human tRNA anticodon variants cause mistranslation.

Table 3.

Mistranslation levels in human β-lymphocytes with genome encoded tRNA^Ser_AAA

Cell line ID	tRNASer N35 genotype	Mistranslation level
876	G/A	11.0 ± 8.5%
878	G/A	35.8 ± 21.5%
896	G/A	33.0 ± 22.0%

Mass spectrometry identifies mistranslated peptides in β-lymphocytes

We next used mass spectrometry to identify Ser mis-incorporation events at Phe codons. We attempted to isolate the rescued mCherry Ser151Phe reporter (as in Fig. 6), however, this led to low levels of protein that were identified by mass spectrometry as GFP-mCherry but without sufficient coverage of the protein to identify mistranslation. We then examined the proteome of β-lymphocytes with the wildtype or A35 variant genotypes and we found some evidence of Ser mis-incorporation in each cell line (Supplementary Fig. S17A and Supplementary Table S2). We were able to identify moderately high confidence peptide hits representing Ser mistranslation at Phe codons in the proteome of three of the cell lines with the A35 variant genotype (876, 878, 896), and we found only very few and substantially lower confidence peptide hits in the wildtype cell line 096 (Supplementary Fig. S17B and Supplementary Table S2). The cell line 847, which has the G35 genotype, also included a number of mistranslated peptides, including 3 peptides with spectral match quality scores (−10LogP > 30) similar to the peptides we identified in the mutant lines. In general, peptides found in 876 and 896 variant lines showed higher quality scores for mistranslated peptides than did the wildtype lines (Supplementary Fig. S17B). In addition, according to spectral counting the 878 variant cell line had a significant 2-fold increase in Ser mis-incorporation relative to the 096 wildtype cell line (Supplementary Fig. S17C). As noted above, since it is possible that tRNA^Ser_AAA could decode Leu UUA codons with Ser, we searched the MS/MS data for spectra matching peptides with Ser at Leu sites. We found no significant evidence of peptides containing Ser and Leu positions where MS2 spectra passed our scoring criteria of relative fragment ion intensities >2% and a minimum of 3 consecutive ions at the mis-incorporation site.

One limitation of this aspect of our study is that it is challenging to grow the β-lymphocytes to sufficient densities to produce ample protein for MS/MS that is needed to identify minor protein species, such as mistranslated peptides. Overall, while some evidence of mistranslation was identified, the MS/MS approach was unable to definitively show the increased mistranslation that we observed clearly in the anticodon variant cell lines using the dual fluorescent reporter.

Discussion

The human tRNAome, tRNA expression, and anticodon variants

Humans encode >600 tRNA genes, and the human tRNAome is the product of >400 active tRNA genes that are regulated in expression [43, 44]. Expanding on this diversity, single nucleotide polymorphisms (SNPs) occur in all human tRNA genes across our population [6, 24, 44, 45]. The impacts of tRNA gene polymorphisms on tRNA function, health, and disease are open and active areas for research. Given the central role of tRNA in accurate translation of the genetic code, we [6, 24] and others [44, 45] were fascinated to discover the striking genetic diversity of SNPs in human tRNAs, including those that impact the anticodon [11, 35] and AARS identity elements [12]. Some anticodon variants are exceedingly rare (tRNA^Leu < 0.001% allele frequency) [11], while other tRNA^Ser (2%) and tRNA^Ala (6%–8%) [12] anticodon variants are more common in the population. The tRNA^Ser_AGA-2–3 gene is most expressed in development, stem cells, and cell lines [43] and is predicted to be an active tRNA [46], suggesting that for individuals with the G35A genotype in the tRNA^Ser_AGA 2–3 gene, the tRNA^Ser_AAA is likely to be differentially expressed in human cells and tissues. Fascinatingly, our tRNAseq experiment also confirmed expression of a G35A mutant from the tRNA^Ser_AGA 2–2 gene, which is constitutively expressed in all human cells and tissues [47]. This finding may explain why the G35A variant is found in 2% of individuals in the tRNA^Ser_AGA 2–3 gene but only in <0.2% of human genomes in the 2–2 gene.

Our previous studies showed that tRNA^Ser_AAA variant can be ectopically expressed and does mistranslate Phe codons with Ser [15], thus, showing the sequence elements of the tRNA^Ser_AAA-2–3 gene are suitable for expression and participation in mistranslation in human cells. Our data here confirm that the mutant tRNA is expressed from the endogenous gene in cells with the A35 allele, and we found the mutant tRNA is deficient in methylation at C32 and fragmented in β-lymphocytes. Although the role of C32 methylation in tRNA function is not completely defined [48], recent work shows the C32 methylation is important for efficient decoding of AGU (Ser) rich mRNAs [49]. In agreement with our findings in human cells, pervious work showed the G35 residue is important for function of the yeast C32 methylation enzyme (Trm140) [50]. Our observations of fragmentation of the tRNA anticodon variant in human pan-genome cell lines indicates a novel strategy for the cell to mitigate the impact of mistranslating tRNAs.

While our studies have focused on germline tRNA variants, recent work has identified tRNA anticodon mutants, including the mistranslating tRNAs and same tRNA^Ser_AAA investigated here, in cancer cells as a result of somatic mutations in human tumours [51]. The impact of these tRNA variants on health, cancer [52], and disease progression is an expanding area of investigation.

Diverse cells tolerate elevated levels of mistranslation

While it was originally thought that any substantial deviation from the genetic code would result in lethality, it is now known that organisms from bacteria to mammals can tolerate elevated levels of mistranslation. Mistranslating tRNAs were initially discovered in genetic experiments in E. coli [53–55]. These studies isolated strains with a nonsense, missense, or frame-shift mutation in an essential gene, e.g. tryptophan synthetase (trpA) [56]. Among the genetic factors identified as the trpA mutant suppressors were a variety of tRNAs [56]. Mutant tRNAs that acquired the ability to read premature stop codon mutations in trpA were called nonsense suppressors, while missense suppressor and frame-shift suppressor tRNAs could restore function of the trpA protein by mis-reading the mutant mRNA (reviewed in [57]). These pioneering studies showed that living cells tolerate substantially elevated levels of mistakes in protein synthesis, and that under certain conditions making mistakes in protein synthesis can be essential for cell viability [58]. Several studies have discovered that different kinds of mistranslation can be beneficial to cells under stress [33] and translation fidelity can be regulated in response to stress [59]. For example, we discovered a tRNA^Pro mutant that mis-incorporated Ala at Pro codons ∼6% of the time and rescued a stress sensitive allele in yeast [20]. Another study showed proteome-wide methionine mis-incorporation in response to oxidative stress, perhaps to help proteins resist oxidative damage [60, 61].

Mutations in AARSs [5, 62, 63], tRNA genes [11, 21], or defects in tRNA modification enzymes [64, 65] can cause mistranslation and widespread mis-incorporation of amino acids in proteins. The impact of mistranslation depends on the level of amino acid mis-incorporation, the strength of the codon:anticodon pairing [11, 66], the nature of the substituted amino acids [11], and the cell in which the source of mistranslation is expressed. In human cells, a series of tRNA^Ser anticodon variants that mis-incorporated Ser at codons representing 10 of the 20 proteinogenic amino acids were transfected into HEK 293 cells [21]. Although the Phe anticodon was not assessed, tRNAs that mis-incorporated Ser at Lys, Gln, and Asn showed limited impacts compared to Ser mis-incorporation at Glu, His, Asp, and most significantly Ile codons. The Ile-decoding tRNA^Ser caused dramatic reductions in cell viability and GFP protein production with increased cytotoxicity and induction of the integrated stress response; these phenotypes were more modest for the Lys and His-decoding mutant tRNA^Ser, similar to our observations with the Phe-decoding tRNA^Ser. Despite the observed phenotypic defects, mammalian cells grew and remained viable in the context of different kinds of mistranslation as a result of ectopically expressed synthetic [21] and natural [11, 12, 15, 35] tRNA variants.

Our studies here add human cells with endogenous tRNA variants to a growing diversity of life that can tolerate elevated levels of mistranslation, well above the basal level of error in the 10⁻⁴–10⁻⁸ range [16–18, 21]. Although cells tolerate higher than anticipated error levels, part of the reason this is possible is that cells encode mechanisms to process and degrade mis-made and misfolded proteins. For example, in our previous studies we found that co-expression of tRNA^Ser_AAA together with inhibition of the proteosome was synthetic toxic to human cells [15]. These studies all highlight the fact that the perfectly accurate production of the proteome is not a pre-requisite for life and that living organisms, including humans, rely on protein quality control mechanisms to tolerate errors in translation.

Live-cell reporters that convert errors in protein synthesis into light

Several methods exist to study mistranslation and detect mis-incorporation of amino acids in proteins and across the proteome. Proteomics and mass spectrometry are indispensable methods for detecting amino acid mis-incorporation [67, 68]; however, this approach faces technical limitations in complex proteome samples. Even in light of these challenges and advances in mass spectrometry, a recent review article stressed the need for future refinement in detecting amino acid mis-incorporation with mass spectrometry, since low abundance of the minor species that represent mistranslated peptides is a key limitation [67]. Indeed, we have found greater success in confidently identifying mis-incorporation in purified proteins [11, 12, 15, 35] compared to complex proteomes using mass spectrometry. Our studies above of β-lymphocytes, in which reaching protein levels sufficient for confident identification of mistranslated peptides was also a limiting factor, attest to this fact and provide rationale for the need for more sensitive approaches.

An alternate and complementary method to mass spectrometry involves engineering proteins as sensors for amino acid mis-incorporation. Although a sensor for Ser incorporation at Phe codons was not previously developed, several other fluorescent or luminescent reporters were engineered to detect nonsense suppression [41, 69] as well as different kinds of missense suppression [70, 71]. Several mistranslation sensors were derived from enhanced green fluorescent protein (eGFP) including T65V [31], Y66K [72], D129P [20], and E222Q [73]. Other fluorescent proteins, including mCherry M72E [32], have been employed as sensors of Met mis-incorporation at Glu codons. Our reporter, mCherry Ser151Phe, functions like these examples, whereby the mutant protein is produced but nonfluorescent, while mistranslation events that revert the mutant to the wildtype protein will restore fluorescence. Inspired by previous studies [32], we also used a fused GFP to act as an internal control for changes in protein production levels between cells and across conditions. In contrast to these approaches, a dimming reporter was developed to measure mis-incorporation at threonine (Thr) codons [30]. Finally, dual-luciferase luminescent reporters have been applied to measure levels of mis-incorporation of Glu at Gln codons and Asp at Asn codon that occur in mycobacterial translation systems, and the same study concluded that a GFP-based sensor is both more sensitive and more appropriate for high-throughput applications [28].

Implications for tRNAs in disease and medicine

In some cases, mistranslation may contribute to disease. The impact of mistranslation on cellular function is highly dependent on the level and kind of mistranslation [11, 66], the cell type, growth conditions, and the ability for cells to degrade mis-made or mis-folded proteins [15, 19]. In neurons, mistranslation generated by editing defective alanyl–tRNA synthetases can lead to misfolded proteins that contribute to neurodegenerative disease in the mouse [74]. In this case, the mutant AARS impacts all tRNA^Ala in the cell. In the case of a histidyl–tRNA synthetase mutant that causes CMT, mutations in the active site of the tRNA synthetase led to mistranslation of the proteome, which was rescued by overexpression of the cognate tRNA^His [5].

Although variants outside of the anticodon in tRNA^Arg and the selenocysteine tRNA (tRNA^Sec) have been linked to disease (reviewed in [6]) to the best of our knowledge, no anticodon mutations in cytosolic tRNAs have been found to cause disease. Pathogenic mutations in the anticodon of mitochondrial tRNAs are rare (reviewed in [75]), however, some anticodon variants in mitochondrial tRNAs are linked to disease. In mt-tRNA^Phe, the G34A synonymous anticodon variant is associated with myoclonic epilepsy [76], while the mt-tRNA^Pro G36A variant [77], which is associated with ragged red muscle fibers and mitochondrial myopathy, generates a Ser anticodon in a Pro accepting tRNA.

Our on-going studies have revealed several examples of tRNA anticodon variants in human genomes and found that these are functional when expressed ectopically in human cells [11, 12, 15, 35]. As the lessons of suppressor genetics indicated above, cells can tolerate some level of mistakes if a defective gene product can be translated into a functional protein. Indeed, the application of tRNAs to correct or suppress genetic defects that cause disease is central to the emerging field of tRNA therapeutics [71]. The applications of tRNA therapeutics include cognate tRNAs, nonsense suppressor tRNAs, and missense suppressor tRNAs. As noted for the HARS variant above [5] and also shown for glycyl–tRNA synthetases variants that cause CMT [78], cognate tRNA supplementation can rescue defects or malfunction of AARS variants that cause disease. Nonsense suppressor tRNAs are directed to correct premature stop codons that account for 11% of genetic disease alleles in humans. Not only do nonsense suppressor tRNAs provide efficient and high-fidelity translation [41] of premature stop codons, but they can also antagonize nonsense mediated decay [79], which acts to reduce the level of mRNAs containing premature stop codons. Nonsense tRNA therapeutics were successfully delivered to the mouse [80], and they can be delivered to human cells as tRNA genes [41], synthetic RNAs [81], or by prime-editing [82]. Since missense mutations account for ∼50% of disease-causing alleles in humans, a new frontier in tRNA medicine involves using missense suppressor tRNAs as therapeutics. In one case, a disease-causing CAPN3 Arg-to-Gln mutation was partially reverted [70]. In cellular models of Huntington’s disease, we found that mis-incorporation of Ser or Ala at Gln codons ameliorated aggregation of the disease-causing poly-glutamine protein [83]. Thus, missense suppressor tRNAs have the potential to revert missense mutations and to suppress dominant impacts of disease-causing genes.

tRNA therapeutics will require a therapeutic window in which sufficient wildtype protein can be produced from a disease-causing allele without introducing more mistakes than can be absorbed by the protein quality control machinery of the cell. The fact that endogenous anticodon variants already exist in the human population provides support to the idea that humans can tolerate tRNAs that make mistakes. These findings coupled with the fact that tRNA anticodon mutants are found in human genomes, suggest that missense suppressor tRNAs could be deployed as therapeutics.

Conclusion

Here we describe the engineering of a dual GFP-mCherry fluorescent reporter that is sensitive to Ser mis-incorporation at Phe codons. The reporter can be used to elucidate the amount of mistranslation in diverse cell types and in future to measure the impact of genetic or chemical interactions that increase or decrease the level of mistranslation. The mCherry protein can be further engineered to measure missense suppression with Ser at other codons. Although expression of the tRNA anticodon variant in human cells was associated with growth defects, the cells remained viable in the context of 10% mis-incorporation of Ser at Phe codons. Ser mistranslation caused defects in protein over-production, but not in global protein synthesis. We found that the β-lymphocyte lines derived from individuals in the 1000 genomes project express the endogenous tRNA anticodon variant and generated mistranslation according to our fluorescent reporter. Our tRNA sequencing analysis also revealed hypomodification of m³C32 in tRNA^Ser_AAA and in the β-lymphocytes, varying and significant levels of 5′ fragmentation of the endogenous anticodon variant. While any functional roles of the 5′ fragment are not yet known, the cell’s ability to fragment the tRNA variant reduces the level of mature mistranslating tRNA that can participate in translation and may protect cells from a higher error levels. This fact may also allow the G35A variant to persist in our population. The fact that the mutant tRNAs are expressed and generate Ser mis-incorporation suggests that human cells are more tolerant to translation errors than previously recognized, supporting the feasibility of applying missenses suppressor tRNAs to medicine.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

Ontario Ministry of Research and Innovation (ER18-14-183); Canadian Institutes of Health Research (165985, 185961, 197789); Natural Sciences and Engineering Research Council of Canada (04282, 04776, 04839, 580214, 580241, 596289); Canada Research Chairs (232341); Rare Disease Models and Mechanism Network; Huntington Society of Canada Research Chair. Funding to pay the Open Access publication charges for this article was provided by CIHR [197789 to P.O.].

Data availability

All mass spectrometry data are available in the PRIDE database: Project Name: Mistranslation from endogenous tRNA variants in human pan-genome cell lines; Project accession: PXD064886; Project DOI: 10.6019/PXD064886.

The tRNA sequencing data files for our studies in HEK 293T cells and β-lymphocyte lines are available at the NCBI BioProject database, under accession number PRJNA1282223.

The tRAX output, including read counts, are also listed in Supplementary Data File 1 (available at figshare: https://doi.org/10.6084/m9.figshare.30611138).