Determination of tRNA aminoacylation levels by high-throughput sequencing

Abstract

Transfer RNA (tRNA) decodes mRNA codons when aminoacylated (charged) with an amino acid at its 3' end. Charged tRNAs turn over rapidly in cells, and variations in charged tRNA fractions are known to be a useful parameter in cellular responses to stress. tRNA charging fractions can be measured for individual tRNA species using acid denaturing gels, or comparatively at the genome level using microarrays. These hybridization-based approaches cannot be used for high resolution analysis of mammalian tRNAs due to their large sequence diversity. Here we develop a high-throughput sequencing method that enables accurate determination of charged tRNA fractions at single-base resolution (Charged DM-tRNA-seq). Our method takes advantage of the recently developed DM-tRNA-seq method, but includes additional chemical steps that specifically remove the 3'A residue in uncharged tRNA. Charging fraction is obtained by counting the fraction of A-ending reads versus A+C-ending reads for each tRNA species in the same sequencing reaction. In HEK293T cells, most cytosolic tRNAs are charged at >80% levels, whereas tRNASer and tRNAThr are charged at lower levels. These low charging levels were validated using acid denaturing gels. Our method should be widely applicable for investigations of tRNA charging as a parameter in biological regulation.

Figure 1.

Charged DM-tRNA-seq method. (A and B) Periodate oxidation and β-elimination can differentiate between charged and uncharged tRNAs before sequencing. Periodate selectively oxidizes the 3′ end of uncharged tRNA (A), while the 3′ end of charged tRNA is protected (B). Treatment with high pH causes β-elimination of the oxidized nucleotide and deacylation of charged tRNA. Thus, after end repair with T4 PNK, charged tRNAs will end in –CCA while uncharged tRNAs will end in –CC. (C) Modified DM-tRNA-seq to determine charging fractions. tRNA is first treated with demethylase to remove common tRNA modifications (m¹G, m¹A, m³C) that impair reverse transcription. Then, a DNA/RNA hybrid with a 1 nt DNA overhang (T/G) is added to act as a primer for the TGIRT template switching reaction. This primer contains binding sites for Illumina PCR/library prep (light blue). For this method, we extended the DNA/RNA hybrid (green) to prevent incorrect assignment of –C and –A ending tRNAs due to in silico trimming. After reverse transcription and cDNA purification and circularization, the cDNA is PCR amplified to create a barcoded library for Illumina sequencing.

Figure 2.

Charged tRNA-seq optimization. (A) β-elimination of the final nucleotide is 100%. Using a 5′-radiolabeled model oligo, β-elimination was shown to be complete. After mock treatment with NaCl, the length of the RNA oligo remains the same. However, after treatment with NaIO₄, the RNA is 1 nt shorter, even in the presence of 1 μg/μl total RNA. (B) Phosphate removal from the 3′ nt is 100%. By labeling the final bridging phosphate, the status of the 3′ end can be monitored. After mock treatment with NaCl, the bridging phosphate remains intact, even with PNK treatment. To remove this phosphate, both NaIO₄ and PNK treatment is required. A radiolabeled DNA oligo is included as a reference. (C) There is no cross-hybridization of T- and G-ending primers with 3′A- and 3′C-ending oligos. Model oligos were used in order to assess extension of 3′A- and 3′C-ending oligos from T- and G-ending primers. TGIRT will reverse transcribe only when the primers are complementary to the last nucleotide of the oligo (3′C-ending oligo, G-ending primer and 3′A-ending oligo, T-ending primer). The expected cDNA size is 101 nt for the 3′A-ending oligo and 100 nt for the 3′C-ending oligo. The minor product visible in all lanes including no RNA template added (∼85 nt) is likely derived from aberrant RT extension of the RNA and DNA primers in the reaction mixture. (D) cDNA from TGIRT reactions that were purified for Illumina sequencing. Treatment with demethylases removes m¹A58, m¹G37 and m³C32 which are major roadblocks for TGIRT reaction. This allows for sequencing of longer cDNA transcripts, including both full-length tRNA reads and other longer abortive cDNAs caused by TGIRT stops at other modifications and/or the low processivity of TGIRT.

Figure 3.

Charged tRNA-seq results. (A) Spike-in model oligos and tRNA standards show that charged-tRNA-seq is quantitative. Two pairs of model oligos (black) were added in different molar ratios to the three biological replicates (see ‘Materials and Methods’ section). Additionally two uncharged tRNAs and one in vitro charged tRNA were also added. The ‘charged fraction’ measured by sequencing was plotted against the input ‘charged fraction’. This calibration curve shows that the measured charged fraction is linear across different charged fractions. (B) The measured charged fraction is independent of the abundance of tRNA. The abundance of nuclear and mitochondrial-encoded tRNA isodecoders was normalized to the most abundant isodecoder. tRNA isodecoders that were within 1000-fold of the most abundant tRNA were plotted. The charged fraction is independent of the abundance, suggesting that the measured charged fraction should be accurate even for low abundance tRNA isodecoders. tRNA^Ser and tRNA^Thr isodecoders are highlighted. (C) Most abundant tRNA isodecoders are highly charged. The charged fraction of the top abundance isodecoder for each isoacceptor was plotted. Most abundant isodecoders are highly charged (>80%) while tRNA^Ser and tRNA^Thr isodecoders have lower charging fraction (60–80%).

Figure 4.

Validation of low charging fraction for tRNA^Ser/Thr. Total RNA was loaded onto a 6.5% acid denaturing gel to separate charged from uncharged tRNAs. Deacylated RNA was also run as a control. Trp (A) and mt-Trp (B) are highly charged while tRNA^Ser(GCU) (C) and tRNA^Thr(UCU) (D) have a lower charged fraction. In (C and D), five biological replicates were performed and only two are shown for simplicity.