Identification of tRNA nucleoside modification genes critical for stress response and development in rice and Arabidopsis

Background Modification of nucleosides on transfer RNA (tRNA) is important either for correct mRNA decoding process or for tRNA structural stabilization. Nucleoside methylations catalyzed by MTase (methyltransferase) are the most common type among all tRNA nucleoside modifications. Although tRNA modified nucleosides and modification enzymes have been extensively studied in prokaryotic systems, similar research remains preliminary in higher plants, especially in crop species, such as rice (Oryza sativa). Rice is a monocot model plant as well as an important cereal crop, and stress tolerance and yield are of great importance for rice breeding. Results In this study, we investigated how the composition and abundance of tRNA modified nucleosides could change in response to drought, salt and cold stress, as well as in different tissues during the whole growth season in two model plants–O. sativa and Arabidopsis thaliana. Twenty two and 20 MTase candidate genes were identified in rice and Arabidopsis, respectively, by protein sequence homology and conserved domain analysis. Four methylated nucleosides, Am, Cm, m1A and m7G, were found to be very important in stress response both in rice and Arabidopsis. Additionally, three nucleosides,Gm, m5U and m5C, were involved in plant development. Hierarchical clustering analysis revealed consistency on Am, Cm, m1A and m7G MTase candidate genes, and the abundance of the corresponding nucleoside under stress conditions. The same is true for Gm, m5U and m5C modifications and corresponding methylation genes in different tissues during different developmental stages. Conclusions We identified candidate genes for various tRNA modified nucleosides in rice and Arabidopsis, especially on MTases for methylated nucleosides. Based on bioinformatics analysis, nucleoside abundance assessments and gene expression profiling, we propose four methylated nucleosides (Am, Cm, m1A and m7G) that are critical for stress response in rice and Arabidopsis, and three methylated nucleosides (Gm, m5U and m5C) that might be important during development. Electronic supplementary material The online version of this article (10.1186/s12870-017-1206-0) contains supplementary material, which is available to authorized users.


Background
As indispensable participants in protein synthesis, tRNAs are highly modified in both prokaryotes and eukaryotes. The genes that encode tRNA modification enzymes constitute far more than the tRNA coding genes, which further highlights the importance of tRNA modification [1,2]. Besides the canonical functions in protein synthesis, emerging evidence emphasized critical role for tRNA nucleoside modification in regulation of cellular response to stimuli and developmental signals [2][3][4][5][6][7]. Previous studies in bacteria, yeast and animal systems illustrated various tRNA modified nucleosides affecting bacterial virulence, yeast exocytosis, C.elegans embryo development and human mitochondrial diseases such as MELAS and MERRF [7][8][9][10][11][12][13].
However, we understand quite poorly about the function of tRNA nucleoside modifications on either development or stress response in higher plants. Plants are particularly prone to environmental changes such as drought, cold and high salinity stresses. Research of tRNA nucleoside modification changes upon various stresses or during development is scarce in higher plants. In Arabidopsis, a few components of the Elongator complex (ELP1-6, KTI11-14) were shown to regulate cell proliferation, anthocyanin biosynthesis, drought stress tolerance and immune response [14][15][16][17][18]. These genes participated in ncm 5 U (5-carbamoylmethyluridine) modification at position 34 on certain tRNA species [19][20][21]. In addition to the elp mutants, certain Arabidopsis mutants of tRNA nucleoside modification genes also showed mild phenotypes, such as slow growth of smo2 [22] and reduced respiration rate and less fresh weight of seedlings in tad1 [23]. It has been long postulated that modified nucleosides on tRNA molecules may function as "biosensor" for environmental and physiological changes [24,25], as a fast module to regulate gene expression at translational level. In agreement with this hypothesis, the abundance of tRNA modified nucleosides do change in response to various stresses [26]. The mutants of AtTRM4B, m 5 C (5-methylcytidine) methylation modification gene, had shorter primary roots and increased sensitivity to oxidative stress [27]. Our recent study showed that Am nucleoside (2'-Omethyladenosine) was induced by salt stress in a variety of plants including Rice, Poplar, Arabidopsis and Brachypodium [28]. Overexpression of OsTRM13, the corresponding gene for Am and Cm modification at position 4, improved salt stress tolerance in rice [28].
Among the hundreds of modified nucleosides present on tRNA molecules, methylation is the most ubiquitous and abundant type [29][30][31]. Nucleoside methylation are catalyzed by methyltransferases (MTases) [30,31]. tRNA MTases in S. cerevisiae are mostly AdoMet-dependent methyltransferase, either in RFM superfamily or SPOUT superfamily [32][33][34][35][36]. Another method of classification of AdoMet-dependent MTases was suggested by Schubert HL et al. by using the catalytic domain as criteria for function annotation [37]. MTases for tRNA nucleoside methylations in S. cerevisiae were considered as good references for the study of tRNA MTases in higher plants. Although a few tRNA MTase in Arabidopsis have been reported [22,27,38], little is known in cereal crops such as rice. The PlantRNA database (http://plantrna.ibmp.cnrs.fr/plantrna/) provides decent amount of information on tRNA genomic sequences and processing enzymes (such as the 5′ /3′ processing and intronsplicing enzymes), but not information for the exact location of modified nucleosides in tRNA sequences, either in Arabidopsis or in rice. As for the function of tRNA nucleoside modification enzymes, very few were supported with experimental data in Arabidopsis [22,23,38,39], and none in O. sativa.
The detection and quantification of tRNA modified nucleosides provide a basis for studying the function of modifying enzymes. In comparison with the HPLC method used before for nucleoside quantification, Liquid chromatography-coupled tandem quadrupole mass spectrometry (LC-MS/MS) is a powerful tool with higher sensitivity for modified nucleosides quantification [6,40]. In this study, LC-MS/MS was used to quantify changes of tRNA modified nucleosides under stress conditions as well as in different tissues along plant development in rice and Arabidopsis. 25 known tRNA nucleosides can be clearly separated and detected (Additional file 5: Figure  S1), among which 12 were methylated nucleosides catalyzed by different TRM enzymes as MTases or associated partners (Additional file 5: Figure S2). With protein sequence homology, we identified candidate genes for a subset of methylated nucleosides. Hierarchical clustering was performed on abundance of modified nucleoside and transcript level of modification candidate genes from developmental or stress dataset. The results indicated that, besides Am nucleoside, other methylated nucleosides such as Cm, m 1 A (1-methyladenosine), and m 7 G (7-methylguanosine) might participate in regulation of stress response; whereas Gm (2'-O-methylguanosine), m 5 U (5-methyluridine) and m 5 C (5-methylcytidine) most likely participated in regulation of pant development.

Identification of tRNA MTase candidate genes in rice and Arabidopsis
O. sativa and A. thaliana were selected for the monocot or the dicot model organism. tRNA MTase candidate genes were identified based on protein sequence homology with yeast tRNA MTases. 13 Trm proteins involved in methylated nucleosides in S. cerevisiae were used as query sequences, for blastp search of the rice and Arabidopsis candidates (Table 1, Additional file 4). Figure S2 (Additional file 5: Figure S2) showed the name of the yeast TRM genes, name and structure of the corresponding nucleosides, and the location of these methylated nucleosides on tRNA molecules. Trm1 for N2,N2-dimethylguanosine (m 2 2 G) modification; Trm9 for the last step of 5-methyloxy carbonyl-methyluridine (mcm 5 U) and 5-methyl-amino-methyluridine (mnm 5 U); Trm3 and Trm7 for 2'-O-methylguanosine (Gm) and 2'-O-methylcytosine (Cm) respectively; Trm13 for 2'-Omethyladenosine (Am); Trm2 for 5-methyluridine (m 5 U), Trm11 and Trm112 form a complex for 2methylguanosine (m 2 G); Trm5 and Trm10 for 1methylguanosine (m 1 G) at position 37 and position 9, respectively; Trm8 and Trm82 from a complex for 7methylguanosine (m 7 G); Trm6 and Trm61 for 1methyladenosine (m 1 A); Trm4 for 5-methylcytidine (m 5 C), and Trm140/ABP140 for 3-methylcytidine (m 3 C) modification at position 32 of yeast tRNA [41] We did not quantify m 3 C in this study, since we only got one unique peak (Additional file 5: Figure S1) under 258/126 m/s (Q1/Q3) which overlapped with the m 5 C nucleoside standard. Trm6-Trm61 is a two-subunit complex responsible for m 1 A58 in tRNA i Met , tRNA Asn GUU, tRNA Arg ACU and a few other isoacceptors [42,43]. However, TRM61 is a direct MTase with conserved AdoMet-binding domain. This binding domain was not found in TRM6 protein, indicating TRM6 is accessory protein that may play a role in maintaining the stability of the complex [44,45]. Similarly, TRM82 is a noncatalytic subunit of the heterodimeric complex TRM8-TRM82, which catalyzes m 7 G formation at position 46 [46]. TRM112 is a 15-KDa zinc-finger protein that assisted TRM9 and TRM11 for mcm 5 U and m 2 G methylation, respectively [47,48].
As shown in Table 1, 33 candidate genes were identified in both O. sativa and A. thaliana, respectively, with a blastp cutoff value at 1.0E-06 (Table 1). A further analysis by PFAM database confirmed the presence of conserved domain of the TRM genes, in both rice and Arabidopsis candidate genes ( Table 2). Twenty two candidate genes were selected from rice and 20 from Arabidopsis, after a manual screen based on the domain structure similarity between the query sequences and the candidate genes ( Table 2).
The tRNA nucleoside MTase candidate genes show an uneven distribution both in Arabidopsis and rice genomes The chromosomal location of the rice or Arabidopsis tRNA nucleoside MTase candidate genes were shown in Fig. 1. Chromosome sizes and the position of each gene can be estimated by the scale on the left. The chromosomal distribution patterns showed that some chromosomes and chromosomal regions had a relatively high distribution of MTase genes. For example, the 22 rice TRM candidate genes were mapped to 10 out of 12 rice chromosomes and 54.5% of them were anchored on chr1, chr2 and chr3, whereas one only on chr.5, chr.7, chr.8 and chr.10, respectively, and no MTase gene was located on chr.11 or chr.12 (Fig. 1a). LOC_Os04g25360 and LOC_Os04g25990 located on chr.4 (marked out by the wireframe) were tandem duplicated gene pairs homologous to Trm61 (Fig. 1a). In Arabidopsis, TRM candidate genes were found to be located more frequently on chr.4 and chr.5 (12 of 20), especially within a ca.10 Mb segment region on chr.5 (Fig. 1b). On the contrary, only two TRM candidates were mapped to chr.1 or chr.2 (Fig. 1b). No tandem duplicated gene pairs were found among Arabidopsis MTase candidate genes.

Phylogenetic and conserved motif analysis of MTase in rice and Arabidopsis
Multiple sequence alignment was employed to construct a phylogenic tree of the tRNA nucleoside MTase candidates in rice and Arabidopsis, together with the 13 Trm query proteins in yeast (Fig. 2).
As shown in Fig. 2, MTases that modify analogous substrates were closely clustered: the three MTase groups related to cytidine modification (Trm4p, Trm140p for m 5 C and m 3 C respectively, and Trm7p for Cm modification) were clustered in group I; also found in group I was Trm9p which participated in the last step of mcm 5 U methylation. However, Sequence-structurefunction analysis of tRNA m 5 C methyl-transferase Trm4p and RNA m 5 U methyltransferase revealed that RNA:m 5 C MTases shared a number of common features with the RNA:m 5 U MTases [49], suggesting a tentative evolution route of the two groups of 5-methylpyrimidine MTases. On the contrary, group II and group III MTases were responsible for methylpurine modifications: Trm3p, Trm8p and Trm11p corresponding to Gm, m 7 G and m 2 G in group II, and Trm5p, Trm10p and Trm1p corresponding to m 2 2 G and m 1 G methylations in group III. In between group II and III we found Trm61p and Trm13p relatives for Am and m 1 A methylations (Fig. 2).
Conserved motif analysis by MEME program [50] revealed the presence of conserved residues within the catalytic domain of group I, group II and group III MTases (Fig. 3). All four methyl-pyrimidine MTases (Trm4p, Trm7p, Trm9p, Trm140) exhibited a striking conservation of a Glycine (G) residue, as well as a consensus "VLD(L/M)CAA(P/N)" motif as its neighbor (Fig.  3a, Additional file 5: Figure S3). The strong conservation suggested that the Glycine residue and its surrounding motif most likely are involved in a structure that is required for substrate binding and/or catalysis. Similarly, a conserved pattern, the GxE/D motif was found in group II, from the Trm3p, Trm8p and Trm11p homologs (Fig.  3b). Previously, a crystal structure of the Trm8p revealed that E126 (as part of the conserved motif ) is involved in a bidentate hydrogen bond with the ribose hydroxyl group [46]. Finally, a conserved DLD motif was found from Trm1p and Trm10p homologs from group III (Fig.  3c). In supporting of the functional relevance for this DLD motif, D211 from Trm1p of S. cerevisiae and D132 from Trmp1 of Aquifex aeolicus have been shown as the catalytic center for m 2 2 G methylation [51,52].
Am, cm, m 1 A and m 7 G methylated nucleosides respond to stress treatment in rice and Arabidopsis Hierarchical clustering analysis about tRNA nucleoside methylated modification and the candidate genes expression under stress tolerance was constructed ( Fig. 4 and Fig. 5, Additional file 6). The amounts of methylated nucleosides under biotic and abiotic stress was quantified by LC-MS/MS method, in total 15 methylated nucleosides were quantified (marked in red color, Additional file 5: Figure S1). The gene expression profiles  Fig. 4 and Fig. 5 Additional file 6). An obviously decrease of m 1 A and m 7 G nucleosides was observed under cold stress or salt stress, the level of Cm nucleoside was found to be decreasing under drought stress, whereas the level of Am nucleoside increased dramatically under salt stress in both rice and Arabidopsis (Fig. 4, Fig. 5, Fig. 7a, and b). In a recent study, we have shown that abscisic acid (ABA) treatment also induced a significant increase of Am nucleosides in rice seedlings [28].
Three nucleosides gm, m 5 U and m 5 C are important in plant development Nine different tissues in rice and six representative tissues of Arabidopsis were used to investigate tRNA nucleosides levels during the whole development period. Similar as in the stress dataset, the expression profile of methylated nucleoside candidate genes were obtained from the CREP database (http://crep.ncpgr.cn) for rice genes, and the TileViz database (http://jsp. Weigelworld.Org/tilevi-z/tileviz.jsp) for Arabidopsis genes (Additional file 1). A heat map was constructed based on the expression data and the levels of tRNA nucleosides in     Fig. 4 Hierarchical clustering analysis between tRNA nucleoside modification and the Arabidopsis candidate genes expression under stress conditions. Heat map was generated using raw data of nucleosides abundance or MTase candidate genes expression level under stress conditions, respectively, horizontally normalized and logarithmically transformed. LN2 value was presented in color key, which corresponded with ratio of change on either nucleoside level of the gene expression level under various stress conditions. A hierarchy-clustering was shown on the left, showing similar patterns between nucleosides and MTase candidate genes. Solid boxes indicated nucleosides and corresponding genes clustered together  Fig. 5 Hierarchical clustering analysis between tRNA nucleoside modification and the rice candidate genes expression under stress conditions. Heat map was generated using raw data of nucleosides abundance or MTase candidate genes expression level under stress conditions, respectively, horizontally normalized and logarithmically transformed. LN2 value was presented in color key, which corresponded with ratio of change on either nucleoside level of the gene expression level under various stress conditions. A hierarchy-clustering was shown on the left, showing similar patterns between nucleosides and MTase candidate genes. Solid boxes indicated nucleosides and corresponding genes clustered together different tissues under various stages of development (Fig. 6, Additional file 6). Interestingly, three nucleosides Gm, m 5 U and m 5 C were clustered together with their corresponding candidate genes (Trm3p for Gm, Trm2p for m 5 U and Trm4p for m 5 C, respectively), both in rice (Fig. 6a) and in Arabidopsis (Fig. 6b). Many nucleoside modification genes in rice were found to be highly expressed in callus sample, and generally low levels in stamen tissues (Fig. 6A). In Arabidopsis, a high expression level of tRNA nucleoside modification candidate genes was found in root samples (Fig.  6b), a good consistency between m 5 C nucleoside and the corresponding candidate genes AtTRM4a (At4g40000) and AtTRM4b (At2g22400) were found in roots (Fig. 6b). The high level of m 5 C nucleoside and its corresponding methylation genes suggested a possible role in root development in Arabidopsis, which is consistent with recent findings of short root phenotype in attrm4b mutant [27]. Figure 8 summarized methylated nucleosides associated with stress tolerance or development, with corresponding MTase candidate genes illustrated in yellow circles (Os_ for rice genes, At_ for Arabidopsis genes, Table 1). As mentioned above, Am, Cm, m 1 A and m 7 G nucleosides were regulated by stress (marked by green circle), in particular the Am nucleoside for salt stress and ABA treatment. The abundance of these nucleosides under stress conditions were clustered together (indicated by green lines in Fig. 8) with the expression level of the MTase candidate genes, i.e. TRM13, TRM7, TRM61 and TRM8 homologs in both rice and Arabidopsis. Another five nucleosides, m 5 U, Gm, m 2 G, Um and m 2 2 G, also changed under stress conditions but showed no consistency with the expression level of the MTase candidate genes (gray circle, green dotted lines in Fig. 8). As for development, we found that three nucleosides Gm, m 5 C and m 5 U that might be important to plant development, these three nucleosides were clustered together with their MTase candidate genes (red circle, red solid lines in Fig.  8). m 5 C nucleoside abundance was high in young tissues, especially in roots. The abundance of Gm nucleoside increased in old leaves ( Fig. 7c and d), indicating its potential role in leave senescence. Several other nucleosides such as m 2 G, m 1 A, m 1 G, ncm 5 U also showed tissue/organ specificity during development, but no consistency was found with the expression of their corresponding MTase candidate genes (gray circle, red dotted lines in Fig. 8).

Discussion
In this study, we investigated the profile of tRNA modified nucleosides in different tissues and under various abiotic stress conditions in both rice and Arabidopsis seedlings. Rice is an important cereal crop and a monocot model plant. We identified 22 rice tRNA MTase candidate genes and 20 Arabidopsis genes by protein sequence homology to S. cerevisiea. Analysis of conserved motifs from the candidate MTases showed several types of conserved motifs including residues that might be involved in catalytic activity (Fig. 3). Also from motif analysis, three groups of MTases responsible for cytidine methylations, Trm4p (for m 5 C), Trm140(for m 3 C) and Trm7p (for Cm modification), were clustered together. A high conserved Glycine residue was identified from these groups of MTases (Fig. 3a), possibly indicating a role for Gly in structure required for substrate binding. In supporting of this, an isosteric mutation of K179 in Trm4p (neighbor to the conserved G177) was completely inactive [50]. G177 and K179 in Trm4p are located in motif I, which is regarded as AdoMet-binding site rather than the catalytic center. Likewise, G55 from Trm7p is also located in motif I predicted as AdoMetand ribose-binding sites, while another conserved residue predicted by the motif analysis, D49 (Fig. 3a), is associated with binding to ribose and phosphate group of the nucleotide to be methylated [53]. The conserved G in Trm140p (G549) was also reported critical for AdoMet-binding. As for the second conserved D residue, point mutation of D547 in Trm140p resulted in decreased activity in m 3 C methylation [41]. Taken together, the high conserved domain in group I (Fig. 3a) most likely constitute for an AdoMet-binding site and indispensable for catalytic activity.
With protein sequence homology, we identified candidate genes for a subset of modified nucleosides from rice and Arabidopsis genomes. However, in half of the branches (Fig. 2) more than one candidate genes were present. This could be anticipated because gene duplication event is rather common in plant kingdom, additionally tissue-specific and subcellular organelles may demand for different set-up of tRNA nucleoside modification machinery. With hierarchical clustering analysis between gene expression and abundance of modified nucleosides in tissues and stress conditions, we found that not every candidate gene can cluster together with the modified nucleosides level. This could be explained in the following ways: 1) the unequal preference between redundant candidate genes, i.e. one of the redundant candidate genes might bear the primary modification activity whereas the other might be inactive. An example of this is the Trm4 homologs in Arabidopsis: AtTRM4a was shown to be inactive while AtTRM4b participated in the methylation at tRNA positions C48, C49, and C50 [27]. This is consistent with our result from hierarchical clustering analysis, that AtTRM4b(At2g22400) expression and the corresponding modified nucleosides m 5 C were clustered together under development (Fig. 6). 2) Insignificant or irregular changes in tRNA modified nucleosides and the expression of corresponding candidate genes. In other word, these nucleosides may hardly be regulated by drought, cold or salt stress.
From hierarchical clustering, we propose that certain methylated nucleosides, namely Am (2′-methyladenosine), Cm (2'-O-methylcytidine), m 1 A (1methyladenosine), and m 7 G(7-methylguanosine) were involved in plant stress responses, while Gm(2'-Omethylguanosine), m 5 U (5-methyluridine), m 5 C (5methylcytidine) in plant development. With the corresponding MTase genes identified in rice and Arabidopsis, it is possible to verify the function of these candidate genes in regulation of certain type of abiotic stress, or in general physiology for plant growth and development. Actually, our preliminary result of several Arabidopsis candidate genes suggested these candidate genes do have a role in stress response or root growth, our recent finding of OsTRM13 in salt stress tolerance in rice also suggests that the function of tRNA nucleoside modification and modification genes is yet to be illustrated.

Conclusion
We showed in this study that both physiological conditions and environmental stimuli led to changes in tRNA modified nucleosides, these changes were accompanied by regulation on the expression of corresponding modification genes. By quantification of both the level of methylated nucleosides and the candidate MTase genes, under drought, cold and high salinity stresses in rice and Arabidopsis, we found four methylated nucleosides Hierarchical clustering analysis between tRNA nucleoside methylated modification and candidate genes expression level during development stage. Heatmap profiles about the abundance of methylated nucleosides and the expression level of the rice (a) or Arabidopsis (b) MTase candidate genes from developmental dataset. Heat map was generated from raw data of nucleoside levels and gene expression levels, normalized horizontally. LN2 values was shown above in color key. Solid boxes indicated nucleosides and the corresponding genes clustered into the same group. In panel B, green boxes indicated m 5 C nucleoside and the corresponding candidate genes in Arabidopsis. Solid boxes indicated nucleosides and corresponding genes clustered together (Am, Cm, m 1 A, and m 7 G) critical for stress response and similarly three nucleosides (Gm, m 5 U and m 5 C) for plant development. All results showed good consistency in rice and Arabidopsis. This highlighted future work for transgenic manipulation and functional study of the corresponding modification enzymes. Phylogeny and motif analysis also suggested conserved residues within each group of MTases, those residues may be important for catalytic activity which awaits to be tested by future work.

Plant material and growth conditions
Nipponbare rice (O. sativa cv. Japonica) seeds were germinated in sterilized water for 2 days and then grown hydroponically in a climate chamber with a 16 h-light /8 h-dark cycle at 28°C. At 3-leaves stage (2-3 weeks after germination), seedlings were transferred to the field located at HuaZhong Agricultural University, Wuhan, HuBei Province (latitude N31°, longitude E115°, average altitude 27 m). The growth season starts from April-May and ends around August-September. Arabidopsis in Colombian-0 background was used in this study. The sterilized seeds were sowed on 1/2 MS medium and grown in a culture room at 22°C after vernalization, with 16 h light/8 h dark photo period at 4°C. 10 days old seedlings were transplanted in soil, growing in a growth chamber with similar conditions.

Sampling of tissues during the whole growth season
Mature rice seeds were sterilized with 70% ethanol followed by 2.5% sodium hydrochloride, washed and soaked in distilled water for 2 days. Plumule and radicle were sampled after seed germination in the dark for 48 h. Seedlings were grown hydroponically in the climate chamber with the settings abovementioned. The following tissue samples were taken during the cultivation in the climate chamber: 3d-seedling and 3d-root taken 3 days after germination, 3-leaf seedling and corresponding root taken 2 weeks after germination. Samples of leaf, stem and root during the shooting stage (about 70 days after sowing), heading stage (about 90 days after sowing) and flowering stage (85-95 days after sowing) were taken after transferring seedling to the field. In addition, panicle, spikelet and anther samples were also taken during flowering stage. A callus sample was taken from MS medium after induction and two rounds of sub-culturing. All samples were taken with three biological duplicates, flash frozen and stored in −80°C for further use.
For Arabidopsis sampling, roots were sampled 9-10 days after seed germination. Other samples were taken through the whole development stage as following: apex (seed germination for 1-2 day), seedling (germination for 7 day), young leaf (4-leaves stage), old leaf (rosette leaves in flowering stage), stem (base stems in flowering stage). All samples were taken with three biological replicates, flash frozen and stored in −80°C for further use.

Stress treatment
For rice stress treatment, 9-d-old seedlings from climate chamber were used as starting materials. Cold stress was applied by putting rice seedlings into pre-equilibrated 4°C cold water, samples were taken 0 d, 1 d, 3 d, 5 d, and 7 d after the treatment. A control set was taken with normal growth conditions. Drought (air dry) stress was applied by transferring seedlings onto 3MM filter paper under room temperature; samples were taken at the same time points as above. For high salinity stress, distilled water was changed into 200 mM NaCl for hydroponic cultivation and samples taken 0d, 1d, 3d, 5d, and 7d after salt stress. Samples were taken in triplicates, each sample with 15-20 seedlings.
Twenty one days old Arabidopsis Col.0 seedlings were used as starting materials for stress treatment. Drought stress was applied by water withholding and samples were taken at 0 d, 3 d, 7 d after the treatment. Cold stress and salt stress were applied in similar ways as for rice seedlings described above, a control set was taken with normal growth conditions. All samples were taken in triplicates, each sample with 5-10 seedlings.

tRNA isolation and nucleoside analysis by LC-MS/MS
Small RNAs (< 200 bp) were extracted using microRNA Extraction Kit (Omega Bio-tek Inc.), digested into nucleosides and analyzed by LC-MS/MS with an LC-20A HPLC system and a diode array UV detector (190-400 nm). Details for LC-MS/MS analysis was described previously [28]. The abundance of each nucleoside under stress conditions was normalized to that in control set, the ratio (fold of change) of each nucleoside was calculated and used for heatmap generation.
Database search for rice and Arabidopsis tRNA nucleoside modification candidate genes and bioinformatic analysis The protein sequences of known tRNA nucleoside MTases from S. cerevisiae were used as query sequences, for retrieval of rice gene homologs with Blastp search tool on RGAP database (http://rice.plantbiology.msu.edu/index.shtml), or for Arabidopsis gene on TAIR database (http://www.arabidopsis.org/). A cut-off value was set as 1.0E-6 for initial identification of candidate genes. Protein sequences were manually verified by protein domain analysis on pfam (http://pfam.xfam.org/). The logos and conserved motifs were identified by MEME online searching engine (http://meme-suite.org/ tools/meme) with parameters setup as the following: motif site distribution, any number of repetitions; maximum number of motifs, 5; motif width, from 8 to 50. Conserved motifs were verified on ESPript 3.0 (http:// espript.ibcp.fr/ESPript/cgi-bin/ESPript.cgi) with an ALN file constructed by ClustalX 2.0.