Innate Immune and Chemically Triggered Oxidative Stress Modifies Translational Fidelity

Due to the central importance of tRNA aminoacylation and translational accuracy in understanding the biology of mammalian cells under normal and pathological conditions, we devised a method to measure tRNA misacylation in cells. Our method combines pulse radiolabeling of cells with [³⁵S]-Met with microarrays developed for measuring tRNA abundance ⁸ (Figs. 1A, S1, S2). We hybridized total tRNA to arrays that detect the 274 distinct chromosomal human tRNA species as closely related members of 42 families and all 22 mitochondrial tRNAs, and phosphorimaged to visualize and quantitate ³⁵S-Met-tRNAs hybridized to the array.

For HeLa cells, we detected intense radioactive spots representing five different methionyl-tRNA (tRNA^Met) probes, as expected. Unexpectedly, we easily detected less intense radioactive spots representing several non-tRNA^Met probes. Further, when we infected HeLa cells with influenza A virus (“Flu”) or adenovirus 5 (“Adeno”) prior to pulse radiolabeling, the level of radioactive signals from non-tRNA^Met probes greatly increased (Fig. 1A), and could be exclusively detected by adding excess Met-oligo probes to block hybridization of tRNA^Met (Figs. 1B, S3). We employed multiple approaches to conclusively establish that radioactivity emanating from non-cognate tRNA probes derives from aminoacylated [³⁵S]-Met and not other [³⁵S]-containing material (detailed in Methods and Supplemental Information). We ruled out radiolabeling of tRNAs with thio-modifications via catabolism of [³⁵S]-Met (Figs. 1C, S4). We validated [³⁵S]-Met-misacylation using two non-array based methods for several tRNA species (Figs. 1D,E, S5). To distinguish aminoacyl- from peptidyl-tRNAs, we treated total RNA with aminopeptidase M prior to array hybridization to remove N-terminal ³⁵S-Met residues from peptidyl-tRNAs (Figs. 1F, S6). We also ruled out misacylation resulting from contaminants in the ³⁵S-Met preparation (Fig. S7).

In uninfected cells, the 8 cytosolic misacylated tRNA families totaled ~1.5% of the cumulative radioactivity of all five tRNA^Met families. Upon Flu infection, misacylation increased in 3 of the 8 species and appeared in 18 new tRNA families. Remarkably, the cumulative radioactivity on non-tRNA^Met species totaled ~13% of that of all tRNA^Met families (Figs. 1G, S8). Cells infected with Adeno or vaccinia virus (Vac) demonstrated a similar pattern and degree of misacylation. Increased Met-misacylation in virus-infected cells is not an artifact of increased tRNA expression, as increased misacylation does not correlate with the minor changes in tRNA abundance induced by viral infection (Fig. S9). Under all conditions tested, we failed to detect misacylation of any mitochondrial tRNA, demonstrating the selectivity of misacylation for cytosolic tRNAs (Fig. 1).

We next demonstrated tRNA-Met-misacylation in vitro (Figs. S10, S11). HeLa cell derived-methionyl-tRNA synthetase (MetRS) migrates in two major sucrose gradient fractions; one containing the 11-protein multisynthetase complex, the other containing the multi-synthetase complex associated with polysomes ⁹. Each sedimenting form of the multisynthetase complex demonstrated similar acylation activity with [³⁵S]-Met. The polysome-associated form clearly mediated misacylation among a subset of the misacylated tRNA families identified in vivo. The free form of the multisynthetase complex exhibited less misacylation activity, demonstrating that the fidelity of tRNA synthetases can depend on the higher order structure of the AARS. Further, we showed that while the multisynthetase complex misacylated tRNA^Lys isoacceptors, free LysRS did not (Fig. S11). This is consistent with misacylation being performed by MetRS within the multi-RS complex.

Since aminoacylated tRNAs can be used for non-translation functions ^{10, 11}, it was critical to establish whether Met-misacylated tRNAs are used in translation. A pulse-chase experiment revealed that cognate and non-cognate tRNAs demonstrate a similar off-rate for [³⁵S]-Met after a 3 min chase period with excess unlabelled Met (Fig. 2A). Blocking translation by incubating cells with cycloheximide during the cold Met chase prevented the loss of [³⁵S]-Met from cognate and non-cognate tRNAs in parallel (Figs. 2A, S12A,B). These findings strongly support global use of misacylated tRNAs in protein synthesis. Next, we quantitated the incorporation of [³⁵S]-Met into HA-epitope tagged ubiquitin (Ub-HA), selected as a reporter because it possesses a single Met residue. Infecting cells with Vac increased the specific activity of Ub (dpm/μg protein) by ~1.8 fold, consistent with Vac induced increase in tRNA misacylation (Fig. 2B). 2D gel electrophoresis demonstrated Vac induced alterations consistent with translational utilization of misacylated tRNAs (Fig. 2C). Mass spectrometry detected Ub peptides containing Lys-to-Met substitutions, confirming the translation of misacylated tRNA^Lys(CTT) as predicted from the array data (Fig. 2D,E).

We detected virus-induced increases in misacylation (Fig. S13) and global utilization of Met-misacylated tRNAs in protein synthesis (Fig. S14) even when viral infectivity was inactivated by UV irradiation. Exposing HeLa cells, which express all human TLRs ¹², to the TLR3 ligand poly-inosine-cytosine (poly-IC, which mimics double stranded viral RNA), or the TLR4 ligand lipopolysaccharide (LPS, derived from bacterial cell walls) also increased tRNA misacylation (Figs. 3A, S15A). LPS- and poly-IC-induced misacylation patterns overlapped significantly with each other and with virus-induced misacylation. We obtained similar results with CpG oligonucleotides, a TLR9 ligand (data not shown). Not all immune signaling events increase [³⁵S]-Met-misacylation in HeLa cells, however. Exposing cells to interferon-β or interferon-γ did not increase misacylation, although each cytokine altered the expression of cytoplasmic and mitochondrial tRNAs within 24h of initial exposure (data not shown).

We extended these findings to mouse bone marrow-derived dendritic cells (DCs, Figs. 3B-D, S15B), and liver cells in a living mouse by injecting [³⁵S]-Met into the portal vein (Fig. 3E). Both cell types demonstrated a level and pattern of tRNA misacylation similar to HeLa cells, firmly establishing the in vivo relevance of misacylation. We failed to detect misacylation after labeling cells with either [³⁵S]-Cys or ³H-labeled Ile, Phe, Val, or Tyr (Fig. S16, note that specific activities of other commercially available amino acids are too low to detect misacylation at greater than 0.5%). Thus, misacylation could well be limited to Met.

As viral and bacterial infections activate myriad stress response pathways in cells, we examined the ability of chemical or physical stressors to modulate misacylation. We incubated HeLa cells at 42°C, or exposed them to the Asn-linked glycosylation inhibitor tunicamycin or the proteasome inhibitor MG132, treatments that induce an unfolded protein response via distinct pathways ¹³. While tunicamycin and MG132 increased tRNA misacylation by ~2-fold, heat shock decreased tRNA misacylation by ~2-fold (Fig. S17). The pattern of misacylation induced by tunicamycin and MG132 was limited to subset of RNA families seen in response to viruses and TLR ligands. Allowing HeLa cells to grow past confluence, a condition known to induce stress related genes ¹⁴, also induced misacylation (Fig. S16B).

Heightened generation of reactive oxygen species (ROS) by activation of NADPH oxidases is a common downstream effect of many cellular stressors ¹³. It is well established that genetically encoded Met residues can act in cis to protect enzyme active sites against ROS-mediated damage ¹⁵, and Met protects E. coli against oxidative damage-induced death ⁷. ROS oxidize the highly reactive sulfur in Met which is restored to its reduced state by Met-sulfoxide reductases through NADPH oxidation ¹⁶. We hypothesized that tRNA Met-misacylation protects cells against oxidative stress by replacing the amino acids we identify in the arrays with Met. Because tRNA misacylation is induced rapidly, this mechanism allows immediate extra-genetic incorporation of Met residues in newly synthesized proteins that provide protection against increased ROS levels.

As predicted by this hypothesis, exposing HeLa cells to ROS-inducing agents (arsenite, telluride, or H₂O₂) induces Met-misacylation at high levels (Figs. 4A,B, S18A,B). Arsenite-induced misacylation did not require protein synthesis, as it was unaffected by the addition of CHX at the time of arsenite exposure (Fig. S12C). Oxidizing agents act at least in part by increasing cellular NADPH oxidase activity ¹⁷, and in each case, misacylation was significantly reduced by diphenyleneiodonium (DPI), a broad inhibitor of these oxidases. TLR activation is known to induce ROS in neutrophils and DCs ¹⁸. Remarkably, treating HeLa cells with DPI inhibited poly-IC induced misacylation, implicating ROS as the trigger for TLR-induced misacylation (Figs. 4B,C, S18C). DPI also inhibited LPS and poly-IC induced misacylation in DCs (Fig. 4D,E).

We propose that Met-misacylation is a protective response to cellular stressors that increase levels of ROS. This is consistent with the recent proposal that the ROS scavenging capacity of Met selects for mitochondrial genetic recoding of AUA from Ile to Met ¹⁹. Alternative explanations for Met-misacylation include the possibilities that Met-misacylation is a non-productive byproduct of oxidative stress that exacerbates stress by decreasing translational fidelity (particularly since replacement of charged surface residues with Met would be predicted to increase protein aggregation), and that misacylated Met tRNAs function in cellular methylation or amino acid transport pathways ²⁰.

It has been demonstrated that upon mutating aminoacyl-tRNA synthetases or introducing exogenous misacylating tRNAs, E. coli, yeast, and mice tolerate and adapt to increased errors in tRNA aminoacylation ^{3, 21, 22}. Theoretical considerations ²³ support higher error thresholds for translational fidelity than those observed for tRNA aminoacylation in vitro. We demonstrate that mammalian cells have an intrinsic ability to modify tRNA misacylation and translational fidelity. The extent to which this ability, currently limited to Met, extends to any of the 14 amino acids yet to be examined, is an open question.

In summary, we have shown that tRNA misacylation with Met is a common and regulated event in mammalian cells. Although the full implications of this phenomenon remain to be explored, there is a practical bottom line: decoding mRNA into protein in living cells is not as simple as generally believed. tRNA misacylation-based protein sequence diversity, like RNA splicing and post-translational modifications, may represent an evolutionary strategy for expanding and manipulating the information encoded by nucleic acids ²⁴.