The landscape of Arabidopsis tRNA aminoacylation

Abstract

The function of transfer RNAs (tRNAs) depends on enzymes that cleave primary transcript ends, add a 3' CCA tail, introduce post-transcriptional base modifications, and charge (aminoacylate) mature tRNAs with the correct amino acid. Maintaining an available pool of the resulting aminoacylated tRNAs is essential for protein synthesis. High-throughput sequencing techniques have recently been developed to provide a comprehensive view of aminoacylation state in a tRNA-specific fashion. However, these methods have never been applied to plants. Here, we treated Arabidopsis thaliana RNA samples with periodate and then performed tRNA-seq to distinguish between aminoacylated and uncharged tRNAs. This approach successfully captured every tRNA isodecoder family and detected expression of additional tRNA-like transcripts. We found that estimated aminoacylation rates and CCA tail integrity were significantly higher on average for organellar (mitochondrial and plastid) tRNAs than for nuclear/cytosolic tRNAs. Reanalysis of previously published human cell line data showed a similar pattern. Base modifications result in nucleotide misincorporations and truncations during reverse transcription, which we quantified and used to test for relationships with aminoacylation levels. We also determined that the Arabidopsis tRNA-like sequences (t-elements) that are cleaved from the ends of some mitochondrial messenger RNAs have post-transcriptionally modified bases and CCA-tail addition. However, these t-elements are not aminoacylated, indicating that they are only recognized by a subset of tRNA-interacting enzymes and do not play a role in translation. Overall, this work provides a characterization of the baseline landscape of plant tRNA aminoacylation rates and demonstrates an approach for investigating environmental and genetic perturbations to plant translation machinery.

Keywords: Arabidopsis thaliana; aminoacylation; mitochondrial t‐elements; post‐transcriptional modifications; tRNAs.

Figure 1

MSR‐seq data from Arabidopsis (this study) and human cell line tRNAs (Watkins et al., 2022). Periodate treatment prior to MSR‐seq library construction was used to estimate aminoacylation levels based on protection of the terminal nucleotide of the CCA by the amino acid. Reported values for CCA tails (left) represent the percentages of reads with intact CCA tails after excluding reads that lacked more than just a single 3′ nucleotide. In contrast, non‐CC/CCA percentages (right) represent the fraction of all tRNA reads that lack two or more nucleotides at their 3′ ends, so values between the two panels do not sum to 100%. Such reads represent a large percentage of the dataset, especially in periodate‐treated libraries. Three biological replicates were sequenced per treatment, with highly significant differences for both CCA tail (P = 6.3e‐7, one‐way anova) and non‐CC/CCA percentages (P = 5.1e‐8, one‐way anova). The (periodate‐treated) human samples are very similar to their Arabidopsis counterparts in terms of both CCA percentage and non‐CC/CCA percentage.

Figure 2

Variation among Arabidopsis tRNA isoacceptor families and genomic compartments in CCA tail integrity in response to periodate treatment. Reported values represent the percentages of reads with intact CCA tails after excluding reads that lacked more than just a single 3′ nucleotide. Biological replicates are indicated by different shapes. The average frequency of intact CCA tails differs significantly across genomic compartments (P = 0.0005; one‐way anova), with a lower rate for nuclear‐encoded tRNAs than organellar tRNAs, as indicated by horizontal lines (means) in each panel. Met(e) and Met(i) refer to elongator and initiator tRNA‐Met genes, respectively. We also found a similar difference between organellar and nuclear reads when retention of CCA tails was expressed as a percentage of total reads and not relative to just reads ending in CCA or CC (P = 1.5e‐7; one‐way anova). A similar analysis subdivided into isodecoder families is presented in Figure S2.

Figure 3

Comparison of CCA tail retention across treatments. (a) Correlation in percentage of Arabidopsis reads with intact CCA tails are shown for each biological replicate. Reported values represent the percentages of reads with intact CCA tails after excluding reads that lacked more than just a single 3′ nucleotide. The top row shows the comparison between the negative control libraries (no periodate) on the x‐axis and the periodate‐treated test samples on the y‐axis. The bottom row shows the comparison between the periodate‐treated test samples on the x‐axis and the positive control samples (pre‐deacylated prior to periodate treatment) on the y‐axis. Each point represents an individual tRNA gene (minimum of 100 reads per gene), with color indicating genome of origin and size proportional to the average number of reads mapping to that gene. A one‐to‐one line is plotted in each panel. (b) Values on y‐axis represent the ratio of CCA percentages for the pre‐deacylated treatment versus the standard periodate treatment by isoacceptor family. A value of 1 indicates that pre‐deacylation did not result in any reduction of CCA tail percentage, whereas a value of 0 indicates that pre‐deacylation completely eliminated CCA tails. Each point represents an isodecoder family averaged across genes with a minimum of 100 reads in that family and three biological replicates. Point size reflects average read abundance for that isodecoder family. Met(e) and Met(i) refer to elongator and initiator tRNA‐Met genes, respectively. (c) Comparison of CCA retention rates [unweighted average of values in panel (b)] with aminoacyl‐tRNA estimated for Escherichia coli tRNAs at 37°C, pH 8.6 (Hentzen et al., 1972). These half‐life estimates are strongly correlated with the observed retention of CCA tails in Arabidopsis after a 30‐min pre‐deacylation at 37°C, pH 9.0 followed by periodate treatment (P = 7.9e‐6, log‐transformed). Reported r values in panels (a) and (c) are Pearson correlation coefficients. ***Statistical significance at a P < 0.001 threshold.

Figure 4

Inclusion of a synthetic tRNA spike‐in control in MSR‐seq libraries. Reported values for CCA tails (top row) represent the percentages of reads with intact CCA tails after excluding reads that lacked more than just a single 3′ nucleotide. In contrast, non‐CC/CCA percentages (bottom row) represent the fraction of all tRNA reads that lack two or more nucleotides at their 3′ ends, so values between the panels do not sum to 100%. Values are shown for all Arabidopsis tRNAs (gray) and the synthetic spike‐in tRNAs (red). Two biological replicates were performed for each of two RNA extraction methods: acid‐phenol (left panels) and Trizol (right panels). For each biological replicate, control and periodate‐treated libraries were generated.

Figure 5

5′ Mapping position of MSR‐seq reads for all Arabidopsis tRNA isodecoder families found in the nuclear genome. Counts are represented per thousand reads that mapped to the nuclear tRNA gene set (averaged across the three biological replicates). Control libraries (black bars) are shown as positive values above the x‐axis, while periodate‐treated libraries (red bars) are shown as negative values below the x‐axis. Mapping positions are standardized based on the Sprinzl coordinate system (Sprinzl et al., 1998).

Figure 6

Canonical tRNA structures. (a) Secondary structure with bases numbered according to the Sprinzl coordinate system (Sprinzl et al., 1998). Black triangles indicate positions that were commonly associated with 5′ truncations. Note that because of size selection, short fragments resulting from truncations towards the end of molecule would not have been detectable. (b) Representative tRNA 3D structure (Thermus thermophilus tRNA‐Glu; PDB 1G59), using the same color scheme as in panel (a).

Figure 7

3′ Mapping position of MSR‐seq reads for all Arabidopsis tRNA isodecoder families found in the nuclear genome. Counts are represented per thousand reads that mapped to the nuclear tRNA gene set (averaged across the three biological replicates). Control libraries (black bars) are shown as positive values above the x‐axis, while periodate‐treated libraries (red bars) are shown as negative values below the x‐axis. Mapping positions are standardized based on the Sprinzl coordinate system (Sprinzl et al., 1998).

Figure 8

Heatmap representation of RT misincorporations across different tRNA isodecoder families and nucleotide positions in MSR‐seq control (no periodate) libraries. The color scale reflects the average percentage of nucleotide substitutions and deletions for three biological replicates. Data are only reported for replicates with a read count >50. Mapping positions are standardized based on the Sprinzl coordinate system (Sprinzl et al., 1998), and corresponding structural positions are shown on the x‐axis (see Figure 6 for definition of structural elements).

Figure 9

Arabidopsis mitochondrial t‐element end‐positions and misincorporations. The 5′ cox1 and 3′ ccmC t‐elements were both detected at substantial abundance in MSR‐seq dataset. (a) The 5′ end (top) and 3′ end (bottom) mapping positions of reads are depicted in the same fashion as in Figures 5 and 7, respectively. Both ccmC (left) and cox1 (right) t‐elements are predominantly full‐length on the 5′ end, whereas ccmC shows a high‐frequency internal breakpoint among the 3′ ends. Periodate treatment leads to almost complete removal of the terminal 3′ nucleotide, indicating that the t‐elements are not aminoacylated. (b) The mitochondrial ccmC t‐element (top) is homologous to plastid tRNA‐IleCAT (bottom). Nucleotide misincorporation patterns reveal an abundance deleted bases and nucleotide substitutions at G26 in both transcripts, indicating that the t‐element likely shares the typical base modification at this position, although it appears to act as a “hard‐stop” in tRNA‐IleCAT but not in the t‐element. We did not detect modified bases in the cox1 t‐element, which is not shown in panel (b).