Perseverance of protein homeostasis despite mistranslation of glycine codons with alanine

Abstract

By linking amino acids to their codon assignments, transfer RNAs (tRNAs) are essential for protein synthesis and translation fidelity. Some human tRNA variants cause amino acid mis-incorporation at a codon or set of codons. We recently found that a naturally occurring tRNA^Ser variant decodes phenylalanine codons with serine and inhibits protein synthesis. Here, we hypothesized that human tRNA variants that misread glycine (Gly) codons with alanine (Ala) will also disrupt protein homeostasis. The A3G mutation occurs naturally in tRNA^Gly variants (tRNA^Gly_CCC, tRNA^Gly_GCC) and creates an alanyl-tRNA synthetase (AlaRS) identity element (G3 : U70). Because AlaRS does not recognize the anticodon, the human tRNA^Ala_AGC G35C (tRNA^Ala_ACC) variant may function similarly to mis-incorporate Ala at Gly codons. The tRNA^Gly and tRNA^Ala variants had no effect on protein synthesis in mammalian cells under normal growth conditions; however, tRNA^Gly_GCC A3G depressed protein synthesis in the context of proteasome inhibition. Mass spectrometry confirmed Ala mistranslation at multiple Gly codons caused by the tRNA^Gly_GCC A3G and tRNA^Ala_AGC G35C mutants, and in some cases, we observed multiple mistranslation events in the same peptide. The data reveal mistranslation of Ala at Gly codons and defects in protein homeostasis generated by natural human tRNA variants that are tolerated under normal conditions.

This article is part of the theme issue ‘Reactivity and mechanism in chemical and synthetic biology’.

1. Background

Transfer RNAs (tRNAs) are integral to normal cell function as they are essential components of the protein synthesis machinery. In all cells, the 20 proteinogenic amino acids are incorporated into proteins following their specific ligation to a family of tRNAs [1]. The existence of tRNAs was first proposed in 1955 as a set of adaptor molecules that link nucleic acid sequences to the protein sequences they encode [2]. Shortly thereafter, tRNAs were discovered that functioned as the hypothesized adaptors to carry amino acids to the ribosome for protein production [3]. The discovery of tRNAs was essential to deciphering the genetic code [4,5].

In the standard genetic code, the 64 nucleotide triplets or codons are used to encode 20 amino acids and three stop signals that terminate protein synthesis. There is redundancy to the genetic code that allows for tRNA isoacceptors with different anticodons to accept the same amino acid and decode multiple codons [6]. The tRNA genes are transcribed by RNA polymerase III, and tRNA transcripts undergo a series of cleavages and post-transcriptional modifications to become mature tRNAs (reviewed in [7]). Mature tRNAs are composed of 73–90 nucleotides that are arranged into 4 or 5 base-pairing stems and connecting loops [8]. In three dimensions, tRNAs fold into an L-shaped structure with the anticodon on one end and the amino acid accepting site on the other. The acceptor stems contain a conserved 3′ CCA sequence that is necessary for ligation of the amino acid to 3′-terminal A76 on each tRNA [9] (figure 1). The anticodon loop of each tRNA contains a central three nucleotide anticodon sequence that reads a codon or set of codons found in each messenger RNA (mRNA) sequence.

Figure 1.

Schematic of mechanisms to tRNA-dependent mistranslation of Ala at Gly codons. (a) The naturally occurring tRNA^Ala G35C (red circle) (tRNA^Ala_ACC) variant acquires a Gly anticodon and decodes Gly codons, including GGU as indicated, with Ala. Because the A34 base is converted to inosine, owing to superwobble, tRNA^Ala_ACC will also decode GGC and GGA Gly codons [10]. (b) The natural human tRNA^Gly_C/GCC A3G (red circle) variants acquired the G3 : U70 major identity element for AlaRS. The Ala accepting tRNA^Gly also decodes the indicated Gly codons with Ala.

To become substrates for protein synthesis, tRNAs are recognized by their cognate aminoacyl-tRNA synthetases (aaRSs) and undergo a two-step, ATP-dependent reaction that ligates an amino acid to its cognate tRNA (reviewed in [11]). Identity elements are nucleotides present in each tRNA required for recognition by the cognate aaRS. Although many tRNAs are distinguished through anticodon recognition by their cognate aaRS enzyme, most tRNAs require additional or sometimes different identity elements for recognition [12]. For example, the G3 : U70 base pair found in the acceptor stem of alanine tRNA (tRNA^Ala) is the major determinant for AlaRS, and the enzyme does not recognize the tRNA^Ala anticodon [13]. Seryl-tRNA synthetase (SerRS) [14] and pyrrolysyl-tRNA synthetase [15] also do not recognize the tRNA anticodon, while leucyl-tRNA synthetase only recognizes the middle base (A35) of the anticodon as an important recognition element in some species [16].

Because some aaRSs do not recognize the complete anticodon sequence, among the plethora of natural human tRNA variants [17] are mutants that will not affect aminoacylation but will or may alter the codon(s) decoded by a particular tRNA, resulting in mistranslations across the proteome. Further, tRNAs recognized by their anticodon could sustain a mutation elsewhere in their sequences that may result in the creation of an identity element for a different aaRS [18]. Similar to other nucleic acids, tRNAs are not exempt from mutation, and some mutations in tRNAs can cause loss or gain of function [19]. Initially, data from the 1000 Genomes Project suggested that individuals carry one or two tRNA gene variants compared to the reference genome [17,20]. Targeted and deep sequencing efforts increased coverage of all human tRNA genes to show that individuals have 60–70 tRNA variants compared to reference [21], including common as well as rare tRNA variants with strong potential to cause mistranslation in the cell. The existence of natural tRNA variants that cause mistranslation is emerging rapidly [22], yet nearly all human tRNA variants remain uncharacterized. We recently found a tRNA^Ser_AGA G35A variant that occurs in 2% of the population caused serine mis-incorporation at phenylalanine codons in mammalian cells. Our data showed that a single tRNA variant inhibited protein synthesis, increased cytotoxicity, and inhibited degradation of huntingtin protein aggregates in a cellular model of neurodegenerative disease [22]. The study demonstrated that natural tRNA variants can cause amino acid mis-incorporation and phenotypic defects in cells.

Here, we investigated three tRNA variants identified in publicly available sequencing databases [17,21]. The tRNA^Ala_AGC, tRNA^Gly_CCC and tRNA^Gly_GCC variants (table 1) each have the potential to mistranslate Ala at glycine (Gly) codons through either anticodon or identity element mutations. In mammalian cell lines expressing wild-type or mutant tRNAs, we characterized protein synthesis levels under normal growth and conditions of proteasome inhibition. We used mass spectrometry to confirm that the naturally occurring tRNA^Ala_AGC G35C and tRNA^Gly_GCC A3G variants both induced mis-incorporation of Ala at multiple Gly codons despite showing limited impact on protein homeostasis and cell toxicity.

Table 1.

Naturally occurring tRNA variants characterized in this study. (n.d.—not detected.)

tRNA gene [23]	tRNA variant	codons read	AA mis-incorporated	allele frequency [21] (%)	allele frequency [17,20] (%)
Ala-AGC-6-1	G35C	Gly GGU/C/A	Ala	8.5	6.5
Gly-CCC-1-1	A3G	Gly GGG	Ala	5.2	n.d.
Gly-GCC-1-5	A3G	Gly GGC/U	Ala	1.2	1.2
Ser-AGA-2-3	G35A	Phe UUU/C	Ser	3.0	1.8

2. Methods

(a) . Cloning and plasmid purification

Genomic DNA from human embryonic kidney 293 (HEK 293) cells was used to amplify the tRNA genes of interest via polymerase chain reaction (PCR) following the Pfu Ultra II polymerase protocol (Agilent, CA, USA). Primers were designed to amplify 500 base pairs (bps) up and downstream of tRNA genes of interest. PCR products were separated on a 1.5% agarose gel and visualized with ethidium bromide using a ChemiDoc MP (BioRad, CA, USA). Product bands were excised, purified using the Geneaid Gel Extraction Kit (Geneaid, Taiwan, Taiwan, Republic of China), and then used as templates for subsequent PCR reactions. For wild-type tRNAs, a second round of PCR was performed using primers designed to capture 300 bp up and downstream of the tRNA gene. PciI or NcoI restriction sites were added to the 5′-end of each primer to allow for PciI or NcoI (New England Biolabs, MA, USA) digestion and subsequent ligation into the PciI site on pPANcherry using T4 DNA ligase (NEB). We constructed pPANcherry by removing the EGFP gene (using SpeI and NheI restriction sites and re-ligation) from the EGFP-mCherry construct in WT-PAN (Addgene no. 99638, MA, USA). Integrating the tRNA variants into pPANcherry, thus, enables mCherry to serve as a marker for transfection and for protein synthesis levels in individual cells. Mutant tRNAs were synthesized by overlap extension PCR using the first-round products as a template as before [22]. The mutant tRNA genes were amplified as half molecules, where the mutation was incorporated in overlapping 5′ ends of the primers (electronic supplementary material, table S1). The two overlapping tRNA halves were then used as templates in a second round of PCR that amplified the full-length mutant tRNA including 300 bp of up and downstream native sequence. These full-length mutant tRNA alleles were then cloned into the PciI site of pPANCherry as above. All plasmids were verified by DNA sequencing (Azenta US, IN, USA).

(b) . Mammalian cell culture, transfection and fluorescence microscopy

Cell culture experiments were performed in HEK 293T (ATCC no. CRL-3216) or mouse neuroblastoma Neuro2a (N2a; ATCC no. CCL-131) cells as indicated. Cells were grown and maintained at 37°C, with 5% CO₂ and humidity. Cells were cultured in Dulbecco's modified Eagle medium (4.5 g l⁻¹; Gibco) with 10% fetal bovine serum (Gibco by Life Technologies) and 1% penicillin-streptomycin (P: 100 IU ml⁻¹; S: 100 µg ml⁻¹; Wisent Bioproducts, Quebec, Canada). Plasmid transfections were performed using Lipofectamine 3000 (Invitrogen, ThermoFisher Scientific, MA, USA) according to the manufacturer's instructions for the appropriate plate size. All fluorescent images were taken using the EVOS FL Auto Imaging System (ThermoFisher). mCherry fluorescence was visualized using the EVOS LED RFP light cube (excitation 531 ± 40 nm; emission 593 ± 40 nm). After plating and transfection, cells were incubated at 37°C, with 5% CO₂ and humidity. Images were captured 24 h post-transfection and the level of mCherry fluorescence per cell was determined using a custom ImageJ script as before [22].

(c) . Cytotoxicity and proteasome inhibition assays

All cytotoxicity and proteasome inhibition assays were done with at least n = 3 biological replicates for each cell line and condition. At 24 h post-transfection with plasmids containing a tRNA variant and mCherry, cell media was replaced with media containing 0 or 10 µM proteasome inhibitor (MG132, Sigma-Aldrich), and cells were incubated at 37°C for 4 h before images were captured using the EVOS microscope. Cytotoxicity was measured with the Cytotox-Glo assay (Promega, WI, USA). Following the manufacturer's instructions, 50 µl of assay reagent were added to each well and an initial luminescence reading was taken using the Synergy H1 Plate Reader (BioTek, VT, USA) before 50 µl of assay buffer including digitonin was added to each well and a final luminescence reading was taken. The initial luminescence reading quantifies the population of dead cells in each well as a result of wild-type or mutant tRNA expression, and the addition of digitonin allowed a determination of the total number of cells present in each well. Details for cell harvesting, lysis, and western blotting are in the electronic supplementary material.

(d) . Protein purification and mass spectrometry

N2a cells were plated and transfected in 10 cm plates as described above. At 24 h post-transfection, cells were harvested and stored in −80°C prior to use. Following the RFP-Trap Agarose kit protocol (Chromotek, Munich, Germany), cells were lysed and mCherry was purified from cell lysates. Purified mCherry was diluted with 2 × sodium dodecyl sulfate (SDS)-running buffer and separated using SDS–polyacrylamide gel electrophoresis with 10% acrylamide. Gels were then stained with Coomassie blue dye. The bands corresponding to mCherry were identified at 37 kDa, excised from the gel, placed in sterile 1.5 ml centrifuge tubes containing 5% acetic acid, and submitted for mass spectrometry analysis at the Biological Mass Spectrometry Laboratory (The University of Western Ontario, London, Canada). Trypsin was used to digest mCherry and the Orbitrap Elite Velos Pro mass spectrometer (ThermoFisher) was used in FT/IT/CID configuration to identify amino acids mis-incorporated in mCherry as described before [24].

3. Results

(a) . Identifying mistranslating transfer RNAs

The human tRNA genes tRNA^Ala_AGC (Ala-AGC-6-1 gene), tRNA^Gly_CCC (Gly-CCC-1-1) and tRNA^Gly_GCC (Gly-GCC-1-5) each have naturally occurring variants in the population with the potential to mistranslate Gly codons with Ala (table 1). Alanyl-tRNA synthetase (AlaRS) recognizes the G3 : U70 base pair of tRNA^Ala as the major identity determinant. Since AlaRS does not ‘read’ the tRNA^Ala anticodon [13], mutations to the tRNA^Ala anticodons are likely to cause mistranslation. We identified a common G35C variant in the human tRNA^Ala_AGC 6-1 gene (tRNA^Ala_ACC) that occurs in more than 6% of individuals. Such a mutation is expected to retain Ala accepting activity while the ACC anticodon will decode Gly instead of Ala codons.

Conversely, glycyl-tRNA synthetase recognizes the tRNA^Gly anticodon and normally requires the C35 and C36 bases of the anticodon in addition to the A73 discriminator base and the C2 : G71 base pair in the acceptor stem for glycylation of the tRNA [25]. Thus, while tRNA^Gly anticodon mutations are more likely to lead to loss of function, mutations at other sites can create another tRNA's identity element in their sequence [12]. Some tRNA^Gly sequences naturally have the U70 base, and we identified two different tRNA^Gly variants with an A3G mutation. The mutation produces a tRNA^Gly with a G3 : U70 base pair and is expected to be an efficient Ala acceptor that retains its ability to decode Gly codons. The A3G mutation occurs commonly (greater than 5% allele frequency) in the tRNA^Gly_CCC 1–1 gene and in a smaller but still substantial population (approx. 1% of individuals) in the tRNA^Gly_GCC 1–5 gene (table 1). Each of these tRNA^Ala and tRNA^Gly mutants, thus, have potential to cause Ala mis-incorporation at Gly codons.

(b) . Protein synthesis in mistranslating cells

Inhibited protein synthesis is a hallmark response to mistranslation in mammalian cells [22,26,27]. We hypothesized that the human tRNA^Gly and tRNA^Ala variants will cause mistranslation of Gly codons with Ala and lead to defects in protein synthesis. To test this hypothesis, mCherry was co-expressed as a marker for protein production from a plasmid also bearing a wild-type or mutant tRNA allele. We recently showed that the human tRNA^Ser_AGA (Ser-AGA-2–3 gene) G35A variant (tRNA^Ser_AAA) caused serine mis-incorporation at phenylalanine codons and inhibited protein synthesis in N2a cells [22]. Here, we employed a G35A variant derived from the Ser-AGA-2–5 gene as a control that we have found to cause a greater level of mistranslation in cells. We co-expressed mCherry with tRNA^Ser_AGA or tRNA^Ser_AAA in N2a cells and measured mCherry fluorescence (figure 2a,b). N2a cells expressing the mutant tRNA^Ser showed a substantial and significant decrease in mCherry fluorescence per cell compared to cells expressing the wild-type tRNA^Ser allele. The data are in agreement with our previous result with the tRNA Ser-AGA-2–3 G35A gene variant and confirm a significant tRNA^Ser_AAA-dependent reduction in protein synthesis in N2a cells.

Figure 2.

Protein synthesis levels in normal and mistranslating cells. (a) Brightfield and fluorescent images showing mCherry production in cells co-expressing wild-type or the indicated tRNA variants; (b) quantitation of mean mCherry fluorescence per cell. (c) Western blot of mCherry and a GAPDH loading control and (d) quantification of fold change of mCherry protein levels normalized by GAPDH. Independent sample t-tests were computed (n.s.—not significant, ***p ≤ 0.001, #p ≤ 0.0001) based on at least n= 3 biological replicates.

We used the same approach to evaluate each of the wild-type and mutant tRNA^Gly or tRNA^Ala pairs (table 1) for their impact on mCherry production. Thus, mCherry fluorescence per cell was recoded in cells expressing the wild-type tRNA (tRNA^Ala_AGC or tRNA^Gly_G/CCC) or their respective variants (tRNA^Ala_ACC or tRNA^Gly_G/CCC A3G : U70) with the potential to decode Gly codons with Ala. The mutant tRNA^Ala_ACC, tRNA^Gly_GCC A3G and tRNA^Gly_CCC A3G each showed no significant change in mCherry fluorescence compared to cells expressing the wild-type tRNA counterpart. Western blotting confirmed these observations (figure 2c,d). While tRNA^Ser_AAA caused a significant decrease in mCherry protein levels relative to the GAPDH loading control, none of the tRNA^Ala or tRNA^Gly variants lead to a significant change in mCherry protein levels. We note that our mCherry construct contains a total of 27 Gly codons. Nearly all instances are GGC Gly codons and one is a GGU Gly codon. While these Gly codons can be misread by both tRNA^Gly_GCC A3G and by the mutant tRNA^Ala_ACC, the tRNA^Gly_CCC A3G variant should only be able to misread GGG Gly codons. Rather than engineer an mCherry protein that is sensitive to a particular kind of codon mis-reading, as we have done elsewhere for Ser mis-incorporation [24], here we used mCherry as a measure of protein production, which can be slowed generally as a hallmark of mistranslation in mammalian cells [22,26,27].

(c) . Cytotoxicity in mistranslating Neuro2a cells

Previous work established toxicity resulting from mistranslating tRNAs expressed in mammalian cells [26,28] including from a tRNA^Ser_AAA variant [22] as employed here. By contrast, other mistranslating tRNAs elicit no cytotoxicity and are well tolerated in human cells [18,26–28]. We confirmed a significant increase in cytotoxicity in cells expressing the mutant tRNA^Ser_AAA compared to the wild-type tRNA allele (tRNA^Ser_AGA) (figure 3) in N2a cells. By contrast, compared with the wild-type tRNA allele, none of the tRNA^Ala_ACC or tRNA^Gly_G/CCC A3G : U70 variants caused a significant change in cytotoxicity (figure 3). Interestingly, cells expressing any of these tRNAs or their potential mistranslating mutants showed relatively less cell death than cells expressing the tRNA^Ser_AAA variant, and cells expressing tRNA^Ala or tRNA^Gly_CCC or their respective mistranslating variants showed less cell death than cells expressing wild-type tRNA^Ser_AGA.

Figure 3.

Cytotoxicity in N2a cells expressing wild-type or mistranslating tRNA variants. Quantification of cytotoxicity (Cytotox-Glo) observed in N2a cells expressing wild-type or the indicated tRNA variant. The tRNA^Ser_AGA 2–5 gene G35A variant (tRNA^Ser_AAA) causes mis-incorporation of Ser at Phe codons [22] and significant toxicity in N2a cells. The tRNA^Ala and tRNA^Gly variants caused no significant changes in toxicity compared to cells expressing the wild-type counterpart. Independent sample t-tests were computed (n.s.—not significant, **p ≤ 0.01) based on at least n = 3 biological replicates.

(d) . Mass spectrometry identification of Ala mistranslation at Gly codons

Although under normal conditions the tRNA^Ala and tRNA^Gly variants did not show detectible changes in protein synthesis or cytotoxicity, we used mass spectrometry to identify mis-incorporation in mCherry caused by the tRNA^Ala_AGC G35C mutant and the tRNA^Gly_GCC A3G mutant. As a control, mCherry was isolated from cells expressing the wild-type tRNA^Ala_AGC. The liquid chromatography-tandem mass spectrometry (LC–MS/MS) analysis of mCherry proteins purified form cells expressing potential mistranslating tRNAs revealed abundant evidence of Ala mis-incorporation at multiple Gly codons.

In all cell lines, we identified peptides indicating Ala mis-incorporation at G73, G107 and G108. The peptide hits at G73 may represent a fragmentation product of the cyclized red fluorophore, which is composed of residues M71, Y72 and G73. We also identified a consistent signal for mis-incorporation at positions G107 and G108 in all cell lines. Finally, in the wild-type cell line, we found just one peptide each supporting mis-incorporation at positions G121 and G196 (figure 4a; electronic supplementary material, figure S1; table 2). Compared to peptides identified in cells expressing mutant tRNAs (figure 4; table 2), peptide quality scores [29] were generally lower in the wild-type cell line (−10logP = 30–46), indicating greater uncertainty in the match of peptide to spectrum. Thus, these spectra may represent natural Ala mis-incorporation at Gly codons, a methylation at a nearby site or a mis-identification of the peptide, which is also suggested by the low peptide quality scores.

Figure 4.

tRNA-dependent mis-incorporation of Ala at Gly codons detected by mass spectrometry. Tandem mass spectrometry analysis revealed alanine mis-incorporation in mCherry isolated from N2a cells expressing (a,b) the wild-type tRNA^Ala_AGC; (c,d) the tRNA^Ala_AGC G35C mutant, and (e,f) the tRNA^Gly_GCC A3G mutant. Mis-incorporation events in mCherry (red boxed g = Gly + 14.02 = Ala) isolated from each cell line are depicted alongside (b,d,f) example spectra from the identified peptides. Red boxed ‘g’ symbols above peptide sequence annotate mistranslated residues identified with minimal ion intensities of greater than or equal to 1%.

Table 2.

Summary of Ala mis-incorporation at Gly codons identified by MS/MS.

tRNA gene	tRNA variant	mCherry residue	maximal − 10logP	ion counts Ala/Gly
Ala-AGC-6-1	WT	G107A	69.37	3 A/7 G
	WT	G108A	64.26	3 A/7 G
	WT	G121A	46.99	1 A/14 G
	WT	G196A	35.12	1 A/17 G
Ala-AGC-6-1	G35C	G5A	55.78	2 A/9 G
	G35C	G73A	67.90	5 A/10 G
	G35C	G107A	63.70	3 A/10 G
	G35C	G108A	69.02	3 A/11 G
	G35C	G138A	48.41	1 A/6 G
	G35C	G160A	42.33	1 A/8 G
	G35C	G196A	58.42	2 A/18 G
Gly-GCC-1-5	A3G	G5A	61.30	4 A/11 G
	A3G	G29A	52.94	2 A/16 G
	A3G	G95A	56.59	5 A/19 G
	A3G	G107A	75.15	4 A/15 G
	A3G	G108A	69.41	5 A/15 G
	A3G	G121A	44.13	2 A/29 G
	A3G	G138A	58.43	1 A/17 G
	A3G	G147A	48.34	3 A/7 G
	A3G	G160A	66.97	3 A/18 G
	A3G	G164A	69.48	3 A/15 G
	A3G	G175A	54.35	4 A/18 G
	A3G	G176A	57.70	5 A/18 G
	A3G	G196A	72.08	2 A/28 G
	A3G	G224A	64.70	2 A/3 G
	A3G	G229A	54.54	1 A/8 G

Cells expressing the tRNA^Ala or tRNA^Gly mistranslating mutants showed ample evidence of Ala mis-incorporation at multiple Gly codons in mCherry. Gly sites that showed a greater level of mis-incorporation (greater than 1% ion intensity) were annotated with a small and red boxed ‘g’ symbol above the sequence (figure 4; electronic supplementary material, figures S1–S3). Nearly all mistranslated peptide hits in the wild-type cell line were less than 1% of the maximal ion intensity (electronic supplementary material, figure S1; figure 4a,b), indicating a low background level of potential Ala mis-incorporation. By contrast, cells expressing the tRNA^Ala (figure 4c,d; electronic supplementary material, figure S2) or tRNA^Gly (figure 4e,f; electronic supplementary material, figure S3) mutants demonstrated evidence of mis-incorporation at several different Gly residues, supported by multiple and high-quality peptide hits observed at greater than 1% ion intensity. While we saw some evidence of mistranslation at four of the 25 Gly sites in wild-type cells, cells expressing the tRNA^Ala mutant showed strong evidence of mistranslation at 7 out of 25 Gly sites, and in cells expressing the tRNA^Gly mutant we found Ala at 15 out of 25 Gly residues (figure 4; electronic supplementary material, figures S1–S3; table 2). The peptide to spectra match quality scores were higher in cells expressing tRNA^Ala and tRNA^Gly variants compared to wild-type tRNA (table 2). The higher ion intensity and higher quality scores of peptides from cells expressing the mutant tRNAs indicate confident identification of mistranslated peptides.

The mCherry construct contains nearly all GGC Gly codons and only one other Gly codon (GGU). In addition to ample evidence of mistranslation at GGC, we also observed Ala mis-incorporation at Gly56 (GGU). We detected peptides at Gly56 suggesting Ala mis-incorporation in the mistranslating cell lines (electronic supplementary material, figure S4); however, the ion intensities at these sites were less than 1% of maximal (electronic supplementary material, figures S2 and S3), suggesting relatively lower abundance of the Gly56Ala peptide. Cells expressing tRNA^Ala_ACC showed two independent peptide hits that identify Ala at Gly56 (electronic supplementary material, figures S2 and S4A) albeit with a lower quality score (−10logP = 38.8). Cells expressing tRNA^Gly_GCC A3G showed stronger evidence (−10logP = 58.5) of mis-incorporation at Gly56 (GGU) (electronic supplementary material, figure S4B), and the same peptide indicated a double mis-incorporation of Ala at both Gly56 and Gly57 (electronic supplementary material, figure S4B).

Interestingly, mistranslating cells expressing the tRNA^Gly_GCC A3G variant showed additional evidence of double (electronic supplementary material, figure S5A,B) and also triple mis-incorporation (electronic supplementary material, figure S5C) events in individual peptides from mCherry. In the wild-type cell line, we recorded only a single peptide suggesting dual mis-incorporation at G107, G108 with an ion intensity of less than 1%. Cells expressing tRNA^Gly_GCC A3G produced peptides representing multiple dual mis-incorporation events at the same site (electronic supplementary material, figure S3). We identified peptide hits displaying dual mis-incorporation at G106, G164 (electronic supplementary material, figure S5A) and at G175, G176 (electronic supplementary material, figure S5B).

Together the data demonstrate that both naturally occurring tRNA^Ala_AGC G35C and tRNA^Gly_GCC A3G : U70 variants induce mistranslation of Gly codons with Ala. Thus, naturally occurring anticodon or identity element mutations are viable routes to amino acid mis-incorporation that is tolerated under normal conditions.

(e) . Protein synthesis in HEK 293T cells expressing wild-type or mistranslating transfer RNAs

Because previous studies suggest that tRNA gene expression profiles show both cell-type and tissue-specific dependence [30], we reasoned that a particular mistranslating tRNA variant may have a greater or lesser impact in different cell lines resulting from differential expression of the tRNA or of other tRNAs that the mutant tRNA competes with for aminoacylation or decoding. Thus, we assayed protein synthesis in HEK 293T cells expressing wild-type or mistranslating tRNA variants. The mCherry fluorescence per cell in HEK 293T cells revealed an expected defect in protein synthesis in cells expressing the tRNA^Ser_AAA variant (figure 5a,c). In comparison to wild-type tRNA alleles; however, the tRNA^Ala_ACC or tRNA^Gly_G/CCC A3G : U70 variants showed no change in mCherry fluorescence per cell in HEK 293T cells (figure 5a,c), similar to our observations in N2a cells.

Figure 5.

Proteasome inhibition reveals defective protein synthesis in cells expressing tRNA^Gly_GCC A3G. Brightfield and fluorescence images showing mCherry fluorescence in cells expressing wild-type or mistranslating tRNA variants in (a) 0 µM or (b) 10 µM MG132. Quantitation of mean fluorescence per cell was plotted for cells treated with (c) 0 µM or (d) 10 µM MG132. Independent sample t-tests were computed (n.s.—not significant, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001) based on at least n = 3 biological replicates.

(f) . Cell-specific cytotoxicity of transfer RNA-dependent mistranslation

In N2a cells expressing tRNA^Ser_AAA, we observed significant cytotoxicity induced by Ser mis-incorporation at Phe codons. By contrast, in HEK 293T cells we found no significant change in toxicity in cells expressing tRNA^Ser_AGA versus tRNA^Ser_AAA. The fact that protein synthesis is repressed in HEK 293T cells expressing tRNA^Ser_AAA (figure 5a,c) suggests that mistranslation occurs at a level that does not cause increased levels of cell death (figure 6a). The data indicate that the level of tRNA-dependent mistranslation or its associated phenotypic defects may be cell specific. For each of the tRNA^Ala and tRNA^Gly variants we recorded no changes in cytotoxicity in HEK 293T cells (figure 6a), similar to our findings in N2a cells above.

Figure 6.

Proteasome inhibition reveals toxicity from Ser mis-incorporation at Phe codons but not from Ala mis-incorporation at Gly codons in HEK 293T cells. Quantification of cytotoxicity (Cytotox-Glo) observed in HEK 293T cells in (a) standard conditions or (b) following treatment with 10 µM MG132 proteasome inhibitor. Independent sample t-tests were computed (n.s.—not significant, *p ≤ 0.05) based on at least n = 3 biological replicates.

(g) . Mistranslation in the context of proteasome inhibition

Since mass spectrometry confirmed Ala mis-incorporation is caused by natural human tRNA^Ala and tRNA^Gly variants, we hypothesized that additional stress on the cell could reveal a phenotypic defect associated with Ala mis-incorporation at Gly codons. MG132 is a potent inhibitor of the proteasome [31] that we previously used to measure defects in protein synthesis and protein degradation in cells expressing the mistranslating tRNA^Ser_AAA variant [22]. We anticipated that inhibition of a major protein degradation pathway would cause mis-made and mis-folded proteins to accumulate and disrupt protein homeostasis in mistranslating cells.

HEK 293T cells expressing wild-type or mistranslating tRNAs were exposed to MG132 and mCherry fluorescence was measured as a marker for protein synthesis (figure 5b,d). Cytotoxicity was monitored to identify toxic genetic interactions between proteosome inhibition and mistranslation (figure 6). In the presence of 10 µM MG132, both tRNA^Ser_AGA G35A and tRNA^Gly_GCC A3G mutants lead to significantly decreased fluorescence of mCherry per cell compared to cells expressing their wild-type tRNA counterparts. Interestingly, the tRNA^Gly_GCC A3G and not the tRNA^Gly_CCC A3G variant caused a reduction of protein synthesis in the context of proteasome inhibition. The tRNA^Ala_ACC variant also showed no further evidence of disrupted protein homeostasis in cells treated with MG132. Both the Phe-decoding tRNA^Ser_AAA and the Ala mis-incorporating tRNA^Gly_GCC A3G showed a more significant inhibition of protein synthesis under proteasome inhibition compared their wild-type tRNA counterparts.

Application of MG132 increased cell death in each HEK 293T cell line tested (figure 6a,b); however, we observed no apparent increase in cytotoxicity in cells expressing wild-type versus mistranslating tRNA^Ala or tRNA^Gly variants in the context of proteasome inhibition. Even tRNA^Gly_GCC A3G, which showed defective protein synthesis in MG132 treated cells, was well tolerated in terms of cell toxicity. Interestingly, we observed a small but significant reduction in cytotoxicity in HEK 293T cells expressing tRNA^Ser_AAA variant compared to the wild-type tRNA^Ser under proteaomse inhibition (figure 6b). This finding contrasts the situation in N2a cells, which showed a strong synthetic toxic interaction between tRNA^Ser_AAA expression and proteasome inhibition [22]. These data again indicate cell-specific phenotypic defects associated with different mistranslating tRNAs.

4. Discussion

(a) . Mistranslation from natural transfer RNA^Ala and transfer RNA^Gly variants in human cells

The human genome contains a diverse and large family of tRNAs. Humans encode more than 600 tRNA genes [20]. Although some tRNA genes may be pseudogenes, estimates are that at least 400 tRNA genes are ‘high confidence’ or probably functional tRNAs [32]. Evidence from human serum samples suggests expression of 411 different tRNA genes, and the human liver alone expresses at least 224 different tRNA genes [33]. A recent approach based on experimental data as well as sequence properties of the tRNA genomic loci predicted that at least 314 tRNAs are functional and expressed [34]. Different human cells and tissues show differential regulation of tRNA genes [30,35,36], which include different tRNAs for each proteinogenic amino acid as well as many copies of each tRNA isoacceptor. Even the tRNA isodecoders that share the same anticodon are found in multiple copies. For example, the human tRNA^Gly genes include 28 high confidence genes which use one of three anticodon sequences (5′-G/C/UCC-3′) to decode four (GGN) glycine codons. The human tRNAome contrast that in yeast (275 tRNAs) or Escherichia coli (88 tRNAs), which have much smaller copy numbers of tRNAs. Human tRNAs are also more diverse than their counterparts in yeast [17]. The diversity of human tRNA genes is not confined to the differences between tRNA isodecoders in a single genome, rather each individual harbours approximately 60–70 single nucleotide polymorphisms in their tRNA genes compared to the reference genome [21].

Some of these variants produce tRNAs that are likely to mis-read the genetic code and cause mistranslation that affects the entire proteome. While some mistranslating tRNA variants are found in exceeding rarity (1–2 examples per 100 000 genomes), other tRNAs that cause or may cause mistranslation are found as common variants in the population, with mean allele frequencies in excess of 5%. We previously characterized and provided additional data here regarding a tRNA^Ser_AGA G35A variant that mis-incorporates Ser at Phe codons. The tRNA^Ser G35A variant occurs in 2–3% of the population [17,21], yet some potential tRNA mistranslators are also found in up to 8.5% of the population, like the tRNA^Ala and tRNA^Gly variants we characterized here (table 1). These variants represent two different mechanisms by which a tRNA can gain a function to mis-incorporate amino acids. In the case of tRNA^Ala, the anticodon variant is ignored by AlaRS, and the Ala-tRNA^Ala_ACC is produced which decodes Gly codons. The tRNA^Gly A3G variants instead acquired an AlaRS identity element G3 : U70 [13], which leads to Ala-tRNA^Gly synthesis and decoding of Gly codons with Ala. Our observations here, together with our previous study of a human tRNA^Pro G3 : U70 variant [18], demonstrate that AlaRS can mis-aminoacylate different kinds of tRNA variants that create a G3 : U70 base pair and cause Ala mis-incorporation in mammalian cells.

A recent study focused a similar kind of naturally occurring tRNA variant that creates a G4 : U69 base pair in different tRNAs, including tRNA^Thr. The G4 : U69 base pair is also sufficient to confer Ala accepting activity on a different tRNA [37]. Although the human AlaRS produces Ala-tRNA^Thr with a G4 : U69 variant, mammalian cells and murine tissues possess an activity in their threonyl-tRNA synthetase that de-acylates the mis-made Ala-tRNA^Thr, apparently preventing mistranslation [38]. We can confirm that no similar activity is found in mammalian (N2a) cells to proofread Ala- on tRNAs that read Gly codons. Overall, our findings demonstrate that common tRNA variants can cause mistranslation of a kind or a level that cells are able to tolerate under normal conditions.

(b) . Amino acid similarity may alleviate mistranslation toxicity in cells

Here we compared a tRNA^Ser variant that mistranslates Phe codons, as well as a series of tRNA variants that mis-read Gly codons with Ala. In terms of their chemical similarity in solution or in the context of a folded protein, Gly and Ala are much more similar amino acids than Phe and Ser. In agreement with this fact, we observed significant cytotoxicity only in cells expressing the Phe-decoding tRNA^Ser variant, while the tRNA^Ala and tRNA^Gly mutants lead to no changes in cytotoxicity despite showing strong evidence of amino acid mis-incorporation.

Several examples in the literature support the idea that the similarity between an intended and mis-incorporated amino acid is an important factor that will impact the degree and severity of the cellular response to mistranslation or resulting phenotypic defects in mistranslating cells. In studies in yeast, tRNA^Ser variants were assayed for the ability to mis-read Arg codons and tRNA^Pro variants that mis-read Pro codons with Ala were also characterized [39]. Ser mistranslation at Arg codons induced production of heat shock proteins and showed a greater decrease in growth rate compared to yeast cells that substituted Ala at Pro codons. Thus, the nature or similarity of mis-incorporated amino acids will impact the cellular response and the level of toxicity associated with mis-made proteins.

Previous work in mammalian cells addressed similarity of ambiguously encoded amino acids in the context of Ser mis-incorporation. A series of chimeric or synthetic tRNA^Ser variants with several different anticodons were expressed in mammalian cells. Each variant caused Ser mis-incorporation at codons for several different amino acids [26]. Key evidence on the impact of mistranslation was found by surveying green fluorescent protein (GFP) production and fluorescence in HEK 293 cells, similar to our approach using mCherry as a marker of protein synthesis. Cells that mis-incorporated Ser at Lys codons showed only a minimal defect, while Ser mis-incorporation at Ile codons displayed the greatest defect in protein synthesis. In terms of severity of the phenotypic defect, cells mistranslating Ser for a group of amino acids showed increasing reductions in protein synthesis (Lys < Gln < Asn < Arg < Tyr < Pro < Glu < His < Asp) between 95% and 60% of wild-type levels. A tRNA^Ser with an Ile anticodon reduced GFP levels to 20% that of cells expressing a wild-type tRNA^Ser. We found that the tRNA^Ser_AAA variant reduced mCherry levels per cell to 70% of that in N2a cells expressing the wild-type tRNA or to 84% of the level in HEK 293 cells expressing wild-type tRNA. Under normal conditions, we found that tRNA^Ala and tRNA^Gly variants had no effect on mCherry levels, while in the context of proteasome inhibition in HEK 293T cells we recorded a reduction in mCherry levels in cells expressing tRNA^Gly_CCC A3G to 85% of that in cells expressing the wild-type tRNA^Gly. Thus, our data suggest that Ala mis-incorporation at Gly codons is less disruptive to protein homeostasis than to Ser mis-incorporation at Phe codons.

The rate or level of mistranslation can also contribute to toxicity or growth defects associated with mistranslation [24,40]. Two synthetic tRNA variants, one that causes more frequent errors with Ser mis-incorporation at Ala codons and another that causes less frequent Ser mistranslation at Leu codons, were characterized in murine NIH3T3 cells [28]. Fascinatingly, although both tRNA mutants were well tolerated in cells, the Ala-tRNA^Ser mistranslator stimulated the protein kinase B (AKT) cell survival pathway and increased the growth rate of tumour cells injected into mice. The lower frequency mistranslator and less conservative Ser at Leu substitution was indistinguishable from a control lacking additional tRNA.

Another factor that can influence the level of mistranslation is related to how many codons are effectively read by a mistranslating tRNA. The tRNA^Ala_ACC variant we characterized has an A34, which is normally converted to inosine (I34) [41]. The I34 base is capable of superwobble and can read codons ending in U, C or A [10]. The fact that we showed this tRNA reads multiple GGC codons with Ala, suggests that the I34 modification is intact. The mCherry construct contains only one other Gly codon (GGU), and we found evidence that both tRNA^Ala and the tRNA^Gly variants direct Ala mis-incorporation at Gly56 (GGU) (electronic supplementary material, figure S4). In on-going studies, we are working towards quantitative methods relying on mistranslation sensitive fluorescent reporters [24] as well as absolute quantitation approaches with mass spectrometry [42] to accurately establish the level of Ala mis-incorporation Gly codons. In our future work, these approaches will also allow us to quantify the level of Ala mis-incorporation at each of the four Gly codons.

The number of codons mis-read by each tRNA mutant will certainly impact the phenotype that we observed. The tRNA^Gly_GCC can decode two Gly codons (GGC, GGU), while the tRNA^Gly_CCC can only decode the GGG codon. By contrast, the tRNA^Ala_ACC will be able to decode three Gly codons as a result of superwobble (GGU/C/A). In the human genome, the Gly codon usage is as follows: GGU (215 544 codons, 16% of Gly codons), GGC (453 917 codons, 34% of Gly codons), GGA (325 243 codons, 25% of Gly codons) and GGG (326 879 codons, 25% of Gly codons). This pattern of codon usage is highly conserved in the mouse genome as well. Thus, tRNA^Ala_ACC will mis-read 976 704 codons (75% of Gly codons), tRNA^Gly_GCC (A3G) will mis-read 651 461 codons (59% of Gly codons) and lastly tRNA^Gly_CCC (A3G) will mis-read the fewest number by far at 326 879 codons (25% of Gly codons). While we realize the numbers of the codons decoded in the transcriptome will differ, the overall codon usage provides an approximate view of the burden of codon mis-reading resulting from each tRNA mutant. Together this information supports our observation that tRNA^Gly_GCC A3G elicits a significant phenotypic defect compared to tRNA^Gly_CCC A3G. It is also possible that the phenotype we observe under proteasome inhibition is owing to Ala mis-incorporated by tRNA^Gly_GCC A3G at the highly used GGC/U Gly codons that may starve the cell of Ala needed to translate Ala codons.

The fact that tRNA^Gly_GCC A3G leads to a greater defect than the mutant tRNA^Ala_ACC may be related to tRNA expression levels. We are conducting an on-going tRNA sequencing project, which is beyond the scope of the current work, to characterize the expression level of the tRNA^Gly and tRNA^Ala mutants investigated here. Although those studies are not yet complete, currently available data suggests that the tRNA Gly-GCC-1-5 gene is more active than the tRNA Ala-AGC-6-1 gene [34]. The properties of the local sequence context in the approximately 300 base pairs up and down stream, which are included in our expression constructs of each of the mutant tRNA alleles, contains important determinants of tRNA expression. The tRNA Gly-GCC-1-5 gene has a greater density of CpG dinucleotides as well greater CpG island density compared to the tRNA Ala-AGC-6-1 gene. Moreover, the phylogenetic conservation of Gly-GCC-1-5 across mammals is greater than that of Ala-AGC-6-1. Each of these factors were shown to positively correlate with higher activity or expression of the Gly-GCC-1-5 compared to the Ala-AGC-6-1 gene [34]. These predictions of tRNA gene activity agree with both our finding of a significant phenotypic defect resulting from tRNA^Gly_GCC A3G compared to the mutant tRNA^Ala_ACC, and with our MS/MS data showing greater Ala mis-incorporation from tRNA^Gly_GCC A3G compared to tRNA^Ala_ACC.

(c) . Cell viability and protein homeostasis despite low fidelity translation

We are only at the beginning of our understanding of human tRNA variants and their ability to cause mistranslation or impact health and disease [17]. Mutations that cause tRNAs to mistranslate are quite distinct from mis-sense mutations in individual protein coding genes. There are many examples of mis-sense mutations in human protein coding genes that cause disease [43], and in these cases, all of the resulting protein produced from the defective gene will contain the mistaken amino acid according to proper translation of the mis-sense mutation. By contrast, mistranslating mutations in tRNA genes affect all instances of the codons that the mutant tRNA is able to decode. Thus, unlike a single mutation in a protein coding gene, mistranslating tRNAs may affect the translation of all protein coding genes. Of course, because the mutant tRNA competes with many gene copies of native tRNAs that accurately decode the same codons, the resulting proteome contains a fraction of mistakes resulting from codons that are mis-read some proportion of the time.

Indeed, the relative abundance of some tRNA variants in the population, such as the common tRNA^Gly and tRNA^Ala mutants characterized here, suggests that conservative amino acid substitutions are less harmful than other kinds of mistakes in protein synthesis and that our cells are robust to significantly elevated levels of error in translation. Initial proposals, including Crick's frozen accident [44] and Orgel's error catastrophe hypothesis [45], envisioned the proteome as a fragile entity that is intolerant to errors. Crick's theory considered changes to standard genetic code or ambiguous decoding of codons as a phenotype that would most certainly be selected against during evolution. Since that time, many changes to the genetic code have been identified in natural species, and changes to codon assignments most often require evolution of tRNA variants [46]. Indeed, the yeast Candida albicans decodes CUG codons ambiguously as Ser (97%) and Leu (3%) as a result of an atypical tRNA^Ser_CAG that is recognized by both leucyl-tRNA synthetase and SerRS [47].

Orgel's theory [45] suggested that low fidelity translation would be fatal to cells because the protein synthesis machinery produced from a low fidelity system would increasingly accumulate errors to the point where the protein components of the translation apparatus themselves would no longer function to produce the proteins needed to support life. In cells expressing a natural tRNA^Ser_AAA variant mistranslation can be toxic [22] or, in the extreme limit, high levels of mis-incorporation can prohibit cell viability [48,49]. Cells, however, from across the diversity of life can maintain viability even in the context of dramatically increased levels of amino acid mis-incorporation [18,50]. In E. coli, several mis-sense suppression systems, including those that substitute Cys at Pro codons, Ser at Thr codons, Glu at Gln codons, and Asp at Asn codons all showed negligible impact on cell growth. Protease deficient E. coli strains, however, showed strong and differential sensitivity to each kind of mistranslation [51], demonstrating the importance of protein quality control mechanisms to mitigate the impact of low fidelity translation. In other contexts, including under conditions of oxidative or chemical stress, amino acid mis-incorporation can be beneficial or provide a selective advantage to bacterial [52], yeast [50] and mammalian cells [53]. In the case of Ala mis-incorporation at Gly codons in human cells, our data affirm the robust nature of cells to mistakes in protein synthesis.

5. Conclusion

We found that naturally occurring human tRNA^Ala anticodon and tRNA^Gly identity element variants are both viable routes to Ala mis-incorporation at Gly codons. Some of these variants occur commonly in the human population and while they cause mistranslation as we verified by mass spectrometry, their impact on cells under normal conditions is minimal. Thus, human cells are robust to Ala mis-incorporation at Gly codons, suggesting that tolerance to mis-made proteins is a mechanism that enabled these variants to become so common in the population. One mutant showed defective protein synthesis only in the context of inhibited protein degradation, suggesting that the accumulation of mis-made protein caused defects in protein synthesis. Our data demonstrate that natural tRNAs that mistranslate Gly codons with Ala have the potential to impact cell health under conditions of stress.

Data accessibility

The mass spectrometry data have been uploaded to the PRIDE database: PXD034180 (10.6019/PXD034180).

The data are provided in the electronic supplementary material [54].

Authors' contributions

F.H.: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, writing—original draft, writing—review and editing; J.T.L.: conceptualization, methodology, validation, visualization, writing—review and editing; P.O'D.: conceptualization, data curation, funding acquisition, investigation, methodology, project administration, resources, supervision, visualization, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was supported by the Natural Sciences and Engineering Research Council of Canada (grant no. 04282 to P.O'D.); Canada Research Chairs (grant no. 232341 to P.O'D.) and the Canadian Institutes of Health Research (grant no. 165985 to P.O'D.).