Alignment of transcriptome data
There were a total of 12 RNA-seq samples that treated with different temperatures (25 °C, 35 °C, 40 °C, and 45 °C) in our study; each condition has three replicates. We aligned the transcriptome data to the organelle reference genome respectively. The size of mitochondrion genome is about 773,279 bp, encoding 158 genes, for each sample; there were about 80,000 reads mapped to reference with mapping rate about 0.45% (std = 0.185). For chloroplast, the size of its genome is about 160,928 bp, encoding 120 genes. There were about 400,000 reads mapped to reference with mapping rate about 2.13% (std = 0.7681). Generally speaking, more genes are located in mitochondrion compared with chloroplast, and then, the mapping reads of mitochondrion should be more than that of chloroplast; however, it turned out, an opposite result beyond our expectation, chloroplast has a higher mapping rate significantly, which may be due to sources of samples; more chloroplast mRNA in leaves were extracted and sequenced in this study. The statics of reference-guided mapping rate was shown in Fig. 1.
Identification of RNA editing sites
In order to increase the sequencing depth and reliability of editing sites, the resulting bam files of three replicates under one condition were merged for subsequent identification of RNA editing sites. Based on the SNP-calling results and organelle genome annotation file, a total of 749 RNA editing sites were identified in both organelles; however, a few editing sites only appeared under certain condition. Take samples at 25 °C temperatures in chloroplast as illustration, the attributes of RNA editing sites were shown in Fig. S1. For mitochondrion, there were 627 RNA editing sites identified, involving 53 genes; the number of RNA editing sites identified at different temperatures (25 °C, 35 °C, 40 °C, and 45 °C) correspond to 468, 509, 563, and 582, along with the increment of temperature, the number of sites increased obviously, as shown in Table 1. In contrast, there were only 122 editing sites identified in chloroplast, involving 43 genes; the number of sites did not appear to be rising along with temperature increment; 95 editing sites were identified under three conditions (25 °C, 35 °C, 40 °C); and only 82 sites were identified under 45 °C temperature. The statistics results showed that most of editing sites were C-to-U; for chloroplast, 97 out of 122 editing sites were C-to-U, the second-most type was G-to-A (5 out of 122); for mitochondrion, 602 out of 627 editing sites were C-to-U, the second-most type was G-to-A (25 out of 627). The detailed information of RNA editing sites in all chloroplast and mitochondrion samples were listed in Table S1 and Table S4. There were two possibilities for higher number of editing sites in mitochondrion; one is data bias in sequencing depth; the mapping rate under higher temperature is higher than that of lower temperature, which may give rise to generation of new editing sites; another is upregulation of several PPR proteins; detailed information can be found in the sixth part of results.
Table 1 The statistics of identified RNA editing sites in mitochondrion and chloroplast Characteristics of the statistics for RNA editing sites
We also found that RNA editing occurred in second codon position was mainly the largest in both organelles, followed by first codon position except three conditions (35 °C, 40 °C, and 45 °C) of chloroplast, as shown in Fig. 2. In mitochondrion, globally, 30%, 58%, and 12% of the 627 identified editing sites were found at first, second, and third codon positions, respectively. Similarly, in chloroplast, 14%, 76%, and 10% of the 122 identified editing sites were found at first, second, and third codon positions, respectively. Furthermore, the statistics of editing type showed that the majority (~ 95%) of the editing events resulted in non-synonymous codon changes. Interestingly, we found that the amino acid changes tend to be hydrophobic; the change from hydrophilic to hydrophobic was the highest, followed by the change from hydrophobic to hydrophobic; take condition of T25 for an example, the proportion of hydrophobic2hydrophobic: hydrophilic2hydrophilic: hydrophobic2hydrophilic: hydrophilic2hydrophobic was 114:49:36:206 in mitochondrion, and 13:9:2:62 in chloroplast. In addition, about ~ 55% of the amino acid changes were hydrophilic2hydrophobic produced by editing sites mainly at second codon positions. The most amino acid changes were Ser-to-Leu and Pro-to-Leu; serine is hydrophilic, whereas Leucine and Proline are both hydrophobic. The above results were in good agreement with previous studies (Takenaka et al. 2013b; Yan et al. 2018), which demonstrated that the RNA editing caused an overall increase in hydrophobicity of the resulting proteins.
Reduced RNA editing efficiency with the temperature rises
We also performed statistics and cluster analysis for the RNA editing efficiency. On the whole, the average efficiency of RNA editing sites was about 0.56. For chloroplast and mitochondrion, the average RNA editing efficiency was 0.59, 0.58, 0.48, and 0.42 and 0.64, 0.61, 0.58, and 0.57 respectively under four conditions (25 °C, 35 °C, 40 °C, and 45 °C). With the increase of temperature, the average editing efficiency both declined gradually, as shown in Fig. 3. Actually, since only a large part of editing site demonstrated strongly decreased RNA editing efficiency, the rest editing sites, however, have no significant changes, then no significant differences was detected on the whole level for both organelles. In addition, we separated out the RNA editing sites with “step-up” and “step-down” editing efficiency, where “step-up” denotes the editing efficiency increases as the temperature increases; conversely, “step-down” denotes the editing efficiency decreases as the temperature increases. Finally, a total of 244 sites editing sites were identified in both organelles, and most of these sites have “step-down” editing efficiency. There were 175 sites demonstrated the trend of decreasing (30 for chloroplast, 145 for mitochondrion), whereas 69 sites demonstrated the trend of increasing (11 for chloroplast, 58 for mitochondrion). For the “step-down” editing sites, two-tailed Wilcoxon rank-sum test was used to perform pairwise comparisons, a remarkable significance (p < 0.05) was detected in the comparison of RNA editing efficiency between each two neighboring conditions except for T40-T45 of chloroplast, as shown in Fig. 4. Furthermore, the cluster analysis results also showed that the clustering relationship among samples agreed with the changes of temperatures, and there was a large area demonstrated the trend of decreasing, as shown in Fig. 5. In total, our results suggested that RNA editing process was acutely sensitive to temperature; it is possible that differential RNA editing is one process that allows plants such as grape to rapidly adapt to varying environmental temperatures. Detailed information of RNA editing efficiency was listed in Table S2 and S5. Pairwise comparison of editing allele proportion in all samples was listed in Table S3 and S6.
Genes with changes of RNA editing efficiency
We annotated the involved genes with editing sites of “step-up” and “step-down” RNA editing efficiency. Hence, a total of 68 genes were annotated (25 for chloroplast, 43 for mitochondrion), as shown in Fig. 6 and Fig. S2. For chloroplast, several genes have more editing sites, especially maturase K (matK) and NADH dehydrogenase subunit 2 (ndhB) genes; both genes have four changed editing sites. All the sites of ndhB gene (Chl-100212, Chl-148651, Chl-101400, Chl-101409) were C-to-U editing type and demonstrated the trend of decreasing; the corresponding amino acid changes were His-to-Tyr, Pro-to-Leu, Ser-to-Phe, and Pro-to-Leu, the four amino acids all changed to be hydrophobic. Whereas three sites of matK gene demonstrated the trend of decreasing, one site showed a rising trend. For mitochondrion, more genes have editing efficiency changed sites, such as NADH dehydrogenase gene family (nad4/5/7), ATPase gene family (atp6/9), heme trafficking system membrane gene family (ccmB/C/FC/FC/FN), mitochondrial Cytochrome c oxidase gene family (cox1/2/3), and ribosomal gene family (rps4/7).
ndhB gene encodes components of the thylakoid ndh complex which purportedly acts as an electron feeding valve to adjust the redox level of the cyclic photosynthetic electron transporters. ndhB gene contains by far the higher number of editing sites, 10 in Arabidopsis, probably because the proofreading mechanism that ensures identical sequences of the two inverted repeated regions of plastid DNA makes improbable the fixation of C-to-U back mutations (Martin and Sabater 2010). Previous studies reported that ndh complex is related to stress resistance; transgenic tobaccos defective in the ndhB gene have impaired photosynthetic activity at actual but not at high atmospheric concentrations of CO2 (Horvath et al. 2000). Furthermore, positive selection in ndhB gene was detected in ferns and angiosperms; the adaptive evolution may affect the energy transformation and light resistant; notably, many ndh genes were lost or pseudogenes in gymnosperm.
matK gene, single copy with the length of 1500 bp, usually encodes in the trnK tRNA gene intron, probably assists in splicing its own and other chloroplast group II introns (Hao da et al. 2010), involving genes include the transcripts of trnK, trnA, trnI, rps12, rpl2, and atpF; tRNAs and proteins produced by these genes are essential for chloroplasts to function properly. Similarly, matK gene also suffers adaptive evolution in angiosperms, which means a lot for the transcription process of related genes (Hao da et al. 2010). Thus, the changes of amino acid sites resulting from evolution or RNA editing may fine-tunes maturase performance.
Expression analysis of RNA editing genes and PPR genes
Transcriptome analysis of RNA-seq data was also performed for measuring and comparing the levels of gene expression of RNA editing genes and PPR genes. However, for RNA editing genes, no expression difference was detected under different temperatures, suggesting that temperature stress only affect the RNA editing events and has no influence on expression level for those genes. In order to investigate the reason for the reduced RNA editing efficiency with the increasing of temperature, we also evaluated the expression of RNA editing genes and PPR proteins. After blast searching, a total of 419 proteins were identified as PPRs, and 414 PPR proteins were expressed. Interestingly, the expression level of most PPR proteins demonstrated a downregulated tendency along with the increasing of temperature, as shown in Fig. 7. Moreover, the PPR proteins expression pattern of conditions from 35 to 40 °C revealed a transition point of downregulation. Gene differential expression analysis between samples under two conditions (25 °C, 45 °C) was also performed, as shown in Table S7; a total of 31 PPRs were differently expressed (p value < 0.01, |FoldChange| > 2). Compared with 26 downregulated PPRs, there were still 5 upregulated PPR proteins, such as PPR proteins: GSVIVG01031345001 (PP284_ARATH), GSVIVG01012156001 (PP327_ARATH), GSVIVG01008664001 (PP425_ARATH), revealing different functions of PPR proteins. Hence, on a whole level, there is a positive correlation between the PPR proteins expression and RNA editing efficiency; it is reasoned that the reduced RNA editing efficiency may result from the dropped expression of most PPR proteins.