Transfer RNA modifications and genes for modifying enzymes in Arabidopsis thaliana

21 modified nucleosides and 4 novel nucleosides were detected in tRNAs of Arabidopsis thalianaand hybrid aspen

The model plants, Arabidopsis thaliana and hybrid aspen (Populus tremula × tremuloides) were chosen for tRNA isolation and HPLC analysis. Because of the low yield of tRNA from plant tissues from previous experience, we used young seedlings of Arabidopsis and young leaves and shoot apices from hybrid aspen due to higher abundance of RNA in tissues of early developmental stages. From 5 g frozen tissue we were able to obtain approximately 1 mg total RNA using Trizol reagent. After removal of rRNA and mRNA by LiCl method we routinely obtained about 200 μg small RNA. From the last step of DE52 column purification about 40-50 μg tRNA were used for degradation and subsequent HPLC analysis. Gradient buffers consisting of three buffers were used to separate modified nucleosides and the elution time and spectrum of each peak were used to identify different modified nucleosides.

Twenty-one modified nucleosides were detected in Arabidopsis and hybrid aspen, listed according to the order of elution time from C30 column of HPLC analysis in Table 1. HPLC chromatograms of the two species were very similar (Figure 2); all modified nucleosides present in Arabidopsis were also present in hybrid aspen, with only slight differences for the relative abundance of some peaks. Dihydrouridine (D) is difficult to detect because it elutes together with Ψ, however D is well conserved and is the second most widely distributed modified nucleoside, therefore it should be present in plant tRNAs. Q was not analyzed because it is destroyed during the procedure used for tRNA extraction and digestion in this study. Q is present in E. coli and mammalian tRNA but absent in yeast tRNA[23]. Because the TGT gene responsible for Q biosynthesis is found in P. trichocarpa but not in A. thaliana, Q should be present in tRNA from hybrid aspen but absent in Arabidopsis. We compared the chromatogram with that from S. cerevisiae, calf liver and E. coli (Figure 3), certain prokaryotic tRNA modifications (e.g. s²C, s⁴U, mnm⁵s²U) were not found in plants, however m²A and ms²io⁶A which is present in bacteria but not in yeast and calf liver, was found in Arabidopsis and hybrid aspen tRNAs (Table 1). We also observed differences in tRNA modifications between plants, S. cerevisiae and calf liver. m³C and i⁶A are present in yeast tRNA but were not found in Arabidopsis and hybrid aspen. Genes for m³C modification were not identified. Several modified nucleosides (mcm⁵U, ncm⁵Um and Ar(p)) that were detected using purified single tRNA species from S. cerevisiae were not detected in this study. It is difficult to conclude whether these modified nucleosides were absent or of extremely low abundance. Wybutosine (Y) derivatives were not detected either, however, Arabidopsis genes involved in Y synthesis (At4g04670 and At1g75200) have been proposed [24].

Nucleosidesa	Arabidopsis & Populus	S. cerevisiae	Calf liver	E. coli
Db	+b	+b	+b	+b
Ψ	+	+	+	+
C	+	+	+	+
cmo5U	-	-	-	+
ncm5U	+	+	+	+
U	+	+	+	+
m3C	-	+	+	-
s2C	-	-	-	+
m1A	+	+	+	-
m5C	+	+	+	-
mnm5s2U	-	-	-	+
Cm	+	+	+	+
m7G	+	+	+	+
m5U	+	+	+	+
I	+	+	+	+
G	+	+	+	+
Qg	n.a	-	+	+
acp3U	-	-	-	+
s4U	-	-	-	+
Um	+	+	+	+
m1I	+	+	+	-
mcm5Uc	-	+	n.a.	n.a.
Gm	+	+	+	+
m1G	+	+	+	+
m2G	+	+	+	-
ac4C	+	+	-	-
A	+	+	+	+
m22G	+	+	+	-
mcm5s2U	-	+	n.a.	n.a.
Am	+	+	+	-
t6A	+	+	+	+
m2A	+	-	-	+
m6A	+	+	+	+
m6t6A	+	-	+	+
YOH	-	+	+	-
io6A	-	-	+	+
ms2io6A	+	-	-	-f
Y	-	+	-	-
Ar(p)d	n.a.	+	n.a.	n.a.
ncm5Ume	n.a.	+	n.a.	n.a.

a. Nucleosides are listed in the order of retention time in HPLC chromatogram, threshold for detection is approximately 0.002% of total area in HPLC chromatogram, any modified nucleoside below this threshold is designated as "-".

b. D (dihydrouridine) is not easily detected in HPLC system because of the very early retention time and because of its close elution with Ψ.

c. mcm⁵U is detected in yeast single tRNA prep but not in bulk tRNA preps.

d. Ar(p) is present on initiator tRNA in S. cerevisiae (Aström SU, 1994), its position in HPLC is unknown.

e. ncm⁵Um is detected in yeast single tRNA prep (Glasser AL, 1992), its position in HPLC is unknown.

f. E. coli has ms²i⁶A but Salmonella enterica has ms²io⁶A.

g. Q is not being analyzed because it is destroyed during tRNA extraction and digestion.

n.a. Data not available.

To summarize, four U derivatives, nine A derivatives, three C derivatives and five G derivatives were detected in a total of 21 modified nucleosides from Arabidopsis and hybrid aspen tRNAs. Four novel modified nucleosides were detected (marked with black triangles in Figure 2) and the identity of these plant-specific modified nucleosides requires further experimentation.

Using bioinformatics to find tRNA modifying genes in plants

Dihydrouridine (D) and pseudouridine (Ψ) modification genes belong to dihydrouridine synthase superfamily or pseudouridine synthase superfamily, respectively. Dus1p-Dus4p are required for D modification at six different positions in yeast tRNA [26]. Dus3p homologs in plants were well grouped, less well for Dus1p homologs and no grouping was obvious for Dus2p and Dus4p (Additional file1). Pseudouridine is the most widely distributed modified nucleoside. It has been identified at 15 different positions on yeast tRNA [27]. In total, almost 100 homologous genes were found in plants, which code for modification enzymes responsible for Ψ at different locations on plant tRNA (Additional file1). To completely understand the differences between these gene homologs requires more phylogenetic and motif analyses and will not be investigated in this study.

Methylation is the most common RNA modification, many methylated modified nucleosides exist in plant tRNA (m¹G, m²G, m²₂G, m⁷G, m⁵U, m⁵C, m¹A, m¹I, Am, Cm, Um and Gm, etc.). m¹G is one of the most conserved modifications in tRNA. Trm5p and Trm10p are enzymes involved in the modification of m¹G at different positions in S. cerevisiae. Although carrying similar biochemical activity, these two proteins do not share homology and are likely unrelated [28]. The Trm5p enzyme for m¹G37 modification is an ancient protein. It is also involved in m¹I modification [29]. Three TRM5 gene homologs and one TRM10 gene homolog were found in Arabidopsis and gene homologs are widely distributed in other plant species. Trm11p and Trm112p are both required for m²G modification in yeast tRNA [30]. One TRM11 and two TRM112 gene homologs were found in Arabidopsis. Conserved residues D215 of motif I and D291 of motif IV which are crucial for Trm11p catalytic activity [30] were conserved in all plant TRM11 gene homologs (Figure 4A). TRM1 codes for tRNA(m²₂G) methyltransferase in S. cerevisiae [31]. Three Arabidopsis TRM1 gene homologs were found. Plant TRM1 gene homologs were divided into two groups (Additional file1). Trm8p and Trm82p form protein complexes required for m⁷G modification [32]. Two TRM8 gene homologs were found in Arabidopsis and plant TRM8 gene homologs can be divided into two groups. Plant TRM82 gene homologs are recognized as WD40-domain proteins (the same domain was found in Trm82p) which confer a wide variety of functions. m⁵U is one of the most conserved modified nucleosides, Trm2p protein contains tRNA(m⁵U) methyltransferase activity in S. cerevisiae [33]. In Arabidopsis, two TRM2 gene homologs were found. The yeast Trm4p protein catalyzes formation of m⁵C at positions 34, 40, 48 and 49 [34]. Eight TRM4 gene homologs were found in Arabidopsis belonging to the NOP1/NOP2/Sun protein family. Trm6 and Trm61 are essential genes coding for the two subunits of tRNA(m¹A58) methyltransferase in yeast. One homolog was found in Arabidopsis thaliana for Trm6 and Trm61, respectively. m¹I modification requires two gene products in yeast, Trm5p for methylation and Tad1p for deamination of A [35]. TRM5 homologs have been mentioned above. The Tad1p protein contains a deaminase domain and the conserved residue, E103, is maintained in all plant TAD1 gene homologs (Figure 4B).

In addition to base methylation, ribose methylation requires another group of methyl-transferases. Trm13p is responsible for Am and Cm modification at position 4 in S. cerevisiae [36]. Trm13p does not share obvious homology with other methyltransferases, plant TRM13 gene homologs all contain the TRM13 superfamily domain. One Arabidopsis TRM13 gene homolog was found, however, we failed to detect decreased amounts of Am in T-DNA knock-out mutants of this gene (data not shown). The Trm7p protein is responsible for both Cm32 and Gm34 modification in yeast [37]. Three TRM7 gene homologs were found in Arabidopsis. Trm44p was identified recently as tRNA(Um44) methyltransferase in S. cerevisiae [38]. Although Um was detected in Arabidopsis and Poplar tRNAs in this study, TRM44 gene homolog were not found. TRM3 gene is responsible for Gm18 modification [39]. One TRM3 gene homolog was found in Arabidopsis; however, once again we did not find any change of Gm content in a T-DNA knockout mutant carrying an insertion in an exon of this gene. This may be due to the presence of Gm at other positions.

At least 13 proteins have been shown to be involved in ncm⁵U modification in S. cerevisiae [4]. Elp1-6 are components of the elongator complex which are also involved in ncm⁵U modification by unknown mechanisms. Sit4p, Sap185p, Sap190p and Kti12p are a group of proteins that affect the phosphorylation status of Elp1 protein [40]. One Elp1 homolog was found in Arabidopsis and a few were identified in other plant species. Interestingly the Arabidopsis abo1(Elp1) mutant has been shown to be more resistant to drought and oxidative stress [19]. Sit4p belongs to the calcuneurin-like phosphoesterase protein family and 26 SIT4 gene homologs were found in Arabidopsis. Four Arabidopsis genes were found to be Sap185p and Sap190p homologs. Kti11-14 proteins are involved in resistance to K. lactis killer toxin of S. cerevisiae [41]: Kti13p belongs to the RCC1 family (regulator of chromosome condensation family) involved in regulating chromatin partitioning and cell division; Kti14p belongs to the Casein Kinase I-like protein family and physically interacts with the Elongator complex [4]. One Arabidopsis gene was found for Kti11p, one for Kti12p, six were found for Kti13p and around 90 homologs were found for Kti14p.

Inosine is a common modified nucleoside found in tRNAs. In S. cerevisiae Tad2p and Tad3p are subunits of adenosine deaminase for I34 formation [7]. Both proteins contain a deaminase domain and position E56 in Tad2p which is important for activity was retained in all plant TAD2 homologs (Figure 4C). Tan1p is responsible for ac⁴C modification in yeast [42] and one Arabidopsis TAN1 homolog was found. Plant TAN1 homologs can be divided into two groups (Additional file1). The SUA5 gene has been identified as a tRNA(t⁶A) synthase [25]. One SUA5 homolog was found in Arabidopsis and several were identified in other plants. ms²io⁶A modifications have two side chains: the ms²-group requires the MiaB protein in S. enterica and E. coli [43] and for i⁶- group modification, the MOD5 gene is required in S. cerevisiae [44]. The MiaE protein is required for modifying i⁶A to io⁶A in S. enterica. We found ms²io⁶A present in both Arabidopsis and hybrid aspen tRNAs. Two MiaB gene homologs were found in Arabidopsis and nine isopentenyl-transferases (ATIPT) have been identified in Arabidopsis, however, only two (ATIPT2 and ATIPT9) use tRNA as substrate [22]. No MiaE homologs were found in Arabidopsis.

Based on the tRNA modification candidate genes found in Arabidopsis (Table 3), we decide to use publically available T-DNA mutant lines to identify genes specific for each modified nucleoside in Arabidopsis. We have chosen the genes of small gene families for which less than three genes were potentially involved in a certain modifications. 21 T-DNA insertional mutant lines were ordered from the European Arabidopsis Stock Center (NASC, http://arabidopsis.info/) for 13 genes involved in nine different modified nucleosides. Homozygote lines were isolated and modified nucleosides in total tRNA were subsequently analyzed. Twelve homozygous T-DNA lines were isolated, among them six lines were defective in four specific modified nucleosides: m¹G, m²G, m⁷G and ncm⁵U (Table 4), T-DNA lines and their insertion sites are shown schematically in Figure 5. T-DNA lines in gene At5g47680 (Trm10p homolog) showed a 50% decrease in m¹G content compared to wild type plants (Figure 6A). We named this gene AtTRM10. No m⁷G could be detected in plants with a T-DNA insertion in gene At1g03110 (Figure 6B). This gene is homologous to Trm82p. At1g03110 was named AtTRM82. Mutant plants from T-DNA NASC lines N661341 and N658947 showed no detectable ncm⁵U (Figure 6C). The corresponding genes, At5g13680 and At1g13870 were named AtELP1 and AtKTI12. At3g26410 was homologous to Trm11pthe gene required for m²G modification in S. cerevisiae. Only 7.3% of m²G remained in mutant plants compared to wild type plants (Figure 6D), At3g26410 was named AtTRM11.

T-DNA line (NASC line)	Gene	m1G/Ψ	m7G/Ψ	ncm5U/Ψ	m2G/Ψ
wt		0.344	0.338	0.014	0.273
N653345	At5g47680 (AtTRM10)	0.185 (53%)
N665836	At5g47680 (AtTRM10)	0.208 (60%)
N658418	At1g03110 (AtTRM82)		< 0.001 (< 0.2%)
N661341	At5g13680 (AtELP1)			< 0.0002 (< 0.01%)
N658947	At1g13870 (AtKTI12)			< 0.0003 (< 0.02%)
N622158	At3g26410 (AtTRM11)				0.020 (7.3%)

Amount of certain modified nucleosides were shown using relative quantification to internal standard (Ψ), percentage of wild type level was shown in bracket.

Subcellular localization of tRNA modifying enzymes is an important issue, tRNA molecules are distributed in different subcellular compartments therefore modified nucleosides differ in mitochondria, chloroplast and cytoplasm. We performed prediction of protein subcellular localization using three programs: TargetP, WoLFPSORT and ESLpred (Additional File 2). The results from the three prediction program complement each other because different algorisms were used. Distribution of plant tRNA modification in different subcellular organelles is one of the future work to do, however we need to be cautious about cross-contamination to avoid false-positives because some of the modified nucleosides are present in low abundance.

tRNA modifications are involved in regulating organ growth, stress responses and flowering time in Arabidopsis thaliana

Among the five genes identified in this study, two genes showed phenotype in the knock-out mutants. AtTRM10 and AtTRM82 mutants which showed dramatic decrease of m¹G (Figure 6A) and m⁷G (Figure 6B) modified nucleosides respectively, did not show any phenotype under LD conditions. The AtKTI12 mutant, which carries a T-DNA inserted in an exon of At1g13870 similar to the previously isolated drl1 mutant [45] showed no detectable ncm⁵U (Figure 6C), however, narrow leaves and meristem defect phenotypes in drl1 mutant were not observed in the AtKTI12 mutant.

The AtELP1 mutant, which carries a T-DNA insertion in the third exon of At5g13680, similar to the previously identified elo2 mutant[46] showed no detectable ncm⁵U (Figure 6C). The elo2 mutant belongs to the elongata mutants that have pleitrophic phenotypes, generally identified as reduced organ growth: narrow leaf, reduced growth of primary roots, altered inflorescence architecture and reduced length, delayed seeding growth [46]. The elo2/abo1 mutant also showed increased resistance to drought and oxidative stress, hypersensitivity towards ABA and elevated expression of anthocyanin biosynthesis genes[19, 47]. The AtELP1 protein can complement the yeast Δelp1 mutant [19] and physically interacts with AtKIT12 [46]. The AtELP1 mutant in this study showed a narrow leaf shape (Figure 7A), and also reduced leaf numbers compare to wild type plants (Figure 7B) and serrated leaf edges of the third and fourth true leaves (Figure 7C). These phenotypes were also observed under short-day conditions (data not shown). AtELP1 mutant plants showed reduced root growth on MS medium plate (Figure 7E) compared to wild type plants (Figure 7D). AtELP1 mutants had reduced lateral shoot growth after the removal of the primary shoot. Lateral shoots had difficulties with remaining erect due to a defect in vascular tissues (Figure 7G). Finally, silique morphology was aberrant in this mutant (Figure 7I).

The AtTRM11 mutant, which carries T-DNA insertion in the third exon of At3g26410, showed a small amount (7.3% of wild type level) of m²G (Figure 6D). Under LD conditions the AtTRM11 mutant plant showed an early-flowering phenotype (Figure 7A) as well as reduced root growth on MS medium plates (Figure 7C). In S. cerevisiae, Trm112p is needed for m²G modification by regulating Trm11p activity [30]. No T-DNA lines are available for the two TRM112 homologs in Arabidopsis. Modifying enzymes for m²G at other positions have not been reported.

Analysis of gene expression and phylogeny

We have investigated the expression pattern of all the Arabidopsis tRNA modification candidate genes identified in this study using the AtGenExpress database (Figure 8). 62 tissue samples were included. The candidate genes were grouped according to predicted function in specific modified nucleosides and mean-normalized expression values from the AtGenExpress database were transformed into log values for heat map construction using MeV (MultiExperiment Viewer) software. Most genes had prominent expression in rosette leaves and apex tissues, except for the D and ncm⁵U modification genes. The AtTRM10, AtTRM11, AtTRM82, AtKTI12 and AtELP1 genes identified in this study are marked with an asterisk. From the Tilviz database, except for AtKTI12, their expression was highest in apex tissues, and the expression level was higher in inflorescence apices than in vegetative apices according to the tiling dataset (Figure 8). AtKTI12 is only expressed in late stages of seed development. The expression heat map is a guideline for developmental and tissue specific expression of the candidate genes. The expression profile is very important for functional study. If the phenotype of the transgenic plants is consistent with the expression pattern, the following-up experiments will be performed in the right tissues and at the right developmental stages.

We searched for homologs for the 21 modified nucleosides and dihydrouridine present in Arabidopsis and hybrid aspen. Arabidopsis tRNA modification candidate genes are listed in Table 3. An unrooted Neighbour-Joining tree was constructed showing the phylogeny between plant genes in relation to yeast genes (Figure 9). The result shows that most gene families in Arabidopsis also exist in other plants; however, the number of genes in each family varies. In this section we will discuss only the homologs for the five Arabidopsis tRNA modification genes identified in this study: i.e. the homologs for AtTRM10, AtTRM11, AtTRM82, AtKTI12 and AtELP1. Trm10p is a conserved methyl-transferase, but it does not contain a typical AdoMet-binding domain and shares no homology with other classical tRNA methyltransferases, e.g. Trm5p [28]. Fourteen plant TRM10homologs were found, including one Arabidopsis gene (AtTRM10), two from M. truncatula and three from P. trichocarpa (Figure 9A). Trm11p is required for m²G modification in yeast tRNA. The residues D215 and D291 that are essential for Trm11p catalytic activity are retained in all plant TRM11 gene homologs, including AtTRM11 (Figure 4A). Trm82p belongs to the WD40-domain protein family. Members of this protein family have different biological functions. AtTRM82 is clearly responsible for m⁷G modification in Arabidopsis thaliana. Elp1 and Kti12p are well conserved proteins and both are involved in ncm5U modification. Only a few ELP1 homologs were found in plants (Figure 9D), in contrast to the KTI12 homologs that are found from archea to human (Figure 9E).

The majority of tRNA modification enzymes are not essential, in bacteria only TrmA enzyme has been shown to be essential, however the lethality is not due to lack of m⁵U modification on tRNA but from its effect on ribosome assembly by association with rRNA. In yeast three tRNA modifying enzymes/complexes were found to be essential: Gcd10p/Gcd14p, Tad2p/Tad3p and Thg1p. At present we could not conclude which plant genes encoding tRNA modification enzymes are essential, however during the preliminary screening of the T-DNA lines for tRNA modification genes in Arabidopsis we were not able to isolate homozygote plant from some of the lines (n≥24, n represent number of plants used), which is probably due to essentiality of the corresponding genes. To confirm this, we will increase the amount of plants for screening in the next generation (n≥200), at the same time we will carry out more comprehensive identification and confirmation of the tRNA modification candidate genes.

Plant growth conditions

For total RNA preparation from young seedlings Arabidopsis thaliana ecotype Col.0 was grown as a lawn in soil and vermiculite (3:1) in a greenhouse at 22°C/18°C (day/night temperature), with light intensity of 150 μmol m^-2 s^-1 and 60% humidity under long-day conditions (16 h-light/8 h-dark cycle). 3-weeks seedlings were harvested and frozen in liquid nitrogen for subsequent RNA extraction. Hybrid aspen (Populus tremula × tremuloides; clone T89) were grown in soil in green house at 22/18°C (day/night temperature), with light intensity of maximum 400 μmol m^-2 s^-1 from natural daylight (controlled by curtains, supplemented when required with high-pressure sodium lamps) and 80% humidity under long-day conditions (16 h-light/8 h-dark cycle). Young leaves and apical shoot tips within 2 cm from the top of about 1.5 m high trees were collected and frozen in liquid nitrogen for RNA extraction.

For phenotypic study, seeds for all lines were stratified 3 days in +4°C before being sown; plants were grown in LD conditions (16 hr photoperiod, light density 150 μmol m^-2 s^-1, day temperature 22°C, night temperature 18°C, and 60% humidity). MS (Murashige-Skoog, Duchefa Biochemie) medium, supplemented with 0.8% plant agar (Duchefa Biochemie) and 1% sucrose (Sigma) was used for Arabidopsis root growth studies under LD conditions (same as above).

All T-DNA mutant lines were purchased from The European Arabidopsis Stock Center (NASC, http://arabidopsis.info/). Homozygote or heterozygote genotypes were determined using gene specific primers designed by T-DNA primer design tool (Salk Institute Genomic Analysis Laboratory, http://signal.salk.edu/tdnaprimers.2.html), LBa1 primer (5'-TGGTTCACGTAGTGGGCCATCG-3') was used as left border primer for T-DNA insertion.

RNA isolation and digestion

Total RNA was extracted using Trizol Reagent (Invitrogen), and RNA concentration was determined using NanoDrop ND-1000 spectrophotometer (Thermo Scientific). sRNAs (including tRNA, miRNA and snRNA) were separated from rRNA and mRNA using the LiCl method: rRNA and mRNA were precipitated with 2 M LiCl final concentration, sRNAs in supernatant were precipitated with 3 volumes of ethanol, washed once with 70% ethanol and dissolved in 0.1 M Tris pH7.4, 0.1 M NaCl. tRNA was further purified using DE52 anion exchange resin: RNA in binding buffer (0.1 M Tris pH7.4, 0.1 M NaCl) was loaded on DE52 column (bed volume 2 ml), washed three times with 5 ml binding buffer each time, eluted with 7 ml elution buffer (0.1 M Tris pH7.4, 1 M NaCl). tRNA was precipitated with isopropanol, washed with 70% ethanol and dissolved in MQ water. 50 μg tRNA from Arabidopsis or hybrid aspen were degraded to nucleosides with P1 nuclease (Yamasa Corporation, Japan) and bacterial alkaline phosphatase (Sigma) as following: to 50 μg tRNA (in 100 μl MQ) add 10 ul of 20 mM ZnSO₄, 10 ul nuclease P1 (1 mg/ml, 200 units/mg in 30 mM NaAc pH5.3) and digest at 37°C for at least 24 hrs; add 5 ul bacterial alkaline phosphatase (about 190 units/ml, 30 units/mg, diluted 1:100 with water) and 20 ul 0.5 M Tris pH8.3 and digest at 37°C for 2 hrs, the digested nucleosides are now ready for HPLC analysis.

HPLC analysis

Modified nucleosides were analyzed using Reverse-phase HPLC (Waters Alliance HPLC system and Waters Absorbance Detector 2996; Waters, http://www.waters.com) and C-30 column (Develosil C-30 reverse-phase column, 250 × 4.6 mm; Phenomex Ltd.). The buffer gradient was as follows: buffer A (0.01 M NH₄H₂PO₄+2.5% MeOH, pH5.3), buffer B (0.01 M NH₄H₂PO₄+20% MeOH, pH5.1), buffer C (0.01 M NH₄H₂PO₄+35% Acetonitril). 0-12 min, 100% buffer A; 12-20 min, 100% A; 20-25 min 90% buffer A, 10% buffer B; 25-32 min, 75% buffer A, 25% buffer B; 32-36 min 40% buffer A, 60% buffer B; 36-45 min 38% buffer A, 62% buffer B; 45-80 min 100% buffer B; 80-87 min 100% buffer C; 87-95 min, 100% buffer A. The threshold level for detection of modified nucleosides was approximately 0.002% of the total area. The abundance of each modification was calculated relative to two internal standards (Ψ and t⁶A), with similar results.

Bioinformatics analysis of plant tRNA modifying genes

The protein sequences for tRNA modifying genes from S. cerevisiae or E. coli were used to find homologous genes in Arabidopsis thaliana using blastp tool (on TAIR database: The Arabidopsis Information Resource, http://www.arabidopsis.org), cut-off value 1e-06 (except for KTI13 and SUA5 where cut-off value is 1e-05). Other plant homologs were identified using the Arabidopsis genes as query sequence, using the tblastn program and the NCBI nucleotide collection (nr/nt) database with default settings, the cut-off value was above 60% positives and e-value above 1e-60. All plant genes were aligned with multiple sequence alignment using the CLUSTAW program http://align.genome.jp/. The unrooted neighbour-joining tree was constructed using Geneious Basic 4.5.5 Tree Builder http://www.geneious.com. Data for Arabidopsis gene expression in different tissues and under different developmental stages was downloaded from AtGenExpress database http://jsp.weigelworld.org/expviz/expviz.jsp. The developmental and tiling datasets were downloaded from the Tileviz database http://jsp.weigelworld.org/tileviz/tileviz.jsp. Mean-normalized expression values were transformed into log values. The heat maps were constructed using the MeV (MultiExperiment Viewer) v.4.3.02 software using default parameters.

Subcellular localization of proteins encoded by Arabidopsis tRNA modification candidate genes were predicted using three different programs: TargetP http://www.cbs.dtu.dk/services/TargetP/; WoLFPSORT http://wolfpsort.org/ and ESLpred http://www.imtech.res.in/raghava/eslpred/ with default settings for plant organisms, hybrid approach were chosen for ESLpred.