Background

In eukaryotes, small non-protein-coding RNAs have emerged as key guidance molecules that fulfill important, vital functions in many cellular processes, such as transcription, translation, splicing, DNA replication and RNA processing [1]. It is thought that small RNAs are produced by Dicer-like proteins (DCLs) from their precursors, which can be stem-loop RNA transcripts or long double-stranded RNAs (dsRNAs). The small RNAs directly interact with proteins from the Piwi/Argonaute (AGO) family to form the core of the RNA induced silencing complex (RISC) [2, 3]. However, data collected from animals suggest a correlation between small RNA diversity and morphological complexity [4, 5]. Recent progress in the understanding of the non-canonical mechanisms of small RNA biogenesis has been achieved in mammalian systems [6]. Massive amounts of data produced by next generation sequencing technologies also revealed subclasses of small non-coding RNA species that were derived by alternative biogenesis pathways and only partially met classical definitions, such as small RNAs derived from genomic loci containing repeat sequences [28], Drosophila, chicken and humans [8].

Heat response of csRNAs

On a subcellular level, heat stress has serious consequences for chloroplast activity and structure [29], for example, the loss of chloroplast Cu/Zn superoxide dismutase and increased damage to photosystem II [30]. In the seedlings exposed to 46°C for 1 hour (HT), csRNA abundance declined significantly compared to that of MT seedlings, accounting for approximately 25% of unique sequences and 36% of the csRNA abundance reads in the MT dataset (Figure 5). Additional File 7 lists the down-regulated csRNAs in HT seedlings (less than one-fifth), Additional File 8 shows up-regulated ones (more than 5-fold), and Additional File 9 shows the miRNA microarray analysis of the csRNAs that were remarkably changed after heat treatment.

Figure 5
figure 5

Fold change of unique sequences and the abundance of Chinese cabbage csRNAs in the HT seedlings.

Full size image

Among heat-responsive csRNAs, the rRNA-derived csRNAs (rRNA csRNAs) were the most affected. Under heat stress, the abundance of rRNA csRNAs was reduced to approximately 49%, while the number of the unique csRNAs was reduced to approximately 29% (Figure 5 and Additional File 3). Specifically, the abundance of the csRNAs derived from each rRNA were drastically reduced in HT seedlings, and the rRNA-csRNAs of various length were affected (Additional File 10), showing the broad sensitivity to heat stress.

We chose two families of the most abundant heat-reduced rRNA csRNAs, csR-5sr-1 and csR-23sr-1, for further analyses. csR-23sr-1 and csR-5sr-1 are derived from 23S and 5S rRNAs, respectively, (sequences in Additional File 9). In Chinese cabbage, csR-23sr-1 are predominantly 26 and 21 nt, while csR-5sr-1 are typically 26 nt in length, and they were reduced after heat treatment (Figure 6A and 6C). To test how heat stress affects accumulation of csR-5sr-1 and csR-23sr-1 csRNAs, we performed Northern hybridization of small RNAs. In the HT seedlings, 26 nt csR-23sr-1 and 26 nt csR-5sr-1 csRNAs accumulated much less than in the MT seedlings (Figure 6B and 6D). These data suggest that accumulation of rRNA-derived csRNAs is affected by heat stress.

Figure 6
figure 6

Identification of csR-23sr-1 and csR-5sr-1. (A) and (C) Size distribution analysis of the csR-23sr-1 family and the csR-5sr-1 family in sequencedatasets with core sequences (underlined in Additional File 9). (B) and (D) Northern blotting analysis of csR-23sr-1 and csR-5sr-1 in RNA samples from heat-treated and control seedlings of Chinese cabbage. U6 spliceosomal RNA was used as an RNA loading control.

Full size image

As a whole, the tRNA-derived csRNAs only slightly declined in the HT seedlings compared to those in the MT seedlings (Figure 5). However, the length of these csRNAs was related to heat response (Additional File 4C and 4D). In the HT seedlings, the abundance of the longer csRNAs (29-32nt) decreased, but that of the shorter ones (16-25nt) increased compared to those in the MT seedlings (Table 2). For each tRNA, abundance of the short csRNAs increased while that of the longer csRNAs declined except for tRNA-Gly. csR-trnA-1 and csR-trnA-2 are two most predominant tRNA-derived csRNA families, and originated from chloroplast tRNA-Ala (Figure 7A). The former showed two peaks (17 and 29 nt) while the latter displayed only one peak (23 nt) (Figure 7B and 7C) (sequences in Additional File 9). Northern blotting showed that abundance of 29 nt csR-trnA-1 csRNAs under heat stress declined while those of 17 nt csRNAs increased (Figure 7D), and abundance of 23 nt csR-trnA-2 csRNAs remarkably increased (Figure 7E), consistent with the deep sequencing result. Like rRNA-derived csRNAs, tRNA-derived csRNAs are affected by heat stress.

Table 2 Heat response of tRNA-derived csRNAs
Full size table
Figure 7
figure 7

Identification of csR-trnA-1 and csR-trnA-2. (A) An diagram showing the location of csR-trnA-1 and csR-trnA-2 in chloroplast tRNA-Ala. Two exons of tRNA-A are indicated by boxes. (B) and (C) Size distribution of the csR-trnA-1 family and the csR-trnA-2 family in sequence datasets with core sequences (underlined in Additional File 10). (D) and (E) Northern blotting of csR-trnA-1 and csR-trnA-2 in the HT seedlings of Chinese cabbage. U6 spliceosomal RNA was used as an RNA loading control.

Full size image

In addition to the above-mentioned origins, small RNAs of 21~24 nt from mRNAs and igRNAs increased in HT seedlings compared to those in the control lines (Additional File 4E-H), and those showing relatively high abundances were verified by a microRNA microarray analysis (Additional File 9).

Validation of csRNAs in chloroplasts

To investigate whether the csRNAs are located in the chloroplast, we isolated the highly purified chloroplasts of Chinese cabbage and Arabidopsis and performed Northern blotting. For comparison, the RNA samples from the intact and broken chloroplasts were separated. The lack of nuclear RNA contamination was confirmed by PCR after reverse transcription, using primers designed for (i) At4g10760 (the nuclear gene encoding the chloroplast MTA protein) and (ii) At4g15030 (encoding the nuclear TCP4 protein) gene sequences (Figure 8A). csRNA-4.5sr-1 (sequence in Additional File 9) was the most abundant csRNA family, and its csRNAs were derived from the 3' end of the 4.5S rRNA in chloroplasts with its 3' termini located -1 to +1 nt relative to the last nucleotide of the rRNA. In the csRNA datasets of Chinese cabbage, three abundance peaks at ~18, 22 and 25 nt were displayed consistently for the HT and MT seedlings (Figure 8B). Using Northern blotting analyses, the 18 nt csRNAs and the other csRNAs belonging to csRNA-4.5sr-1 were identified in the intact and broken chloroplast RNAs. Nevertheless, the csRNA-4.5sr-1 csRNAs in the broken chloroplast were less abundant than in the intact chloroplasts (Figure 8C). Similarly, the csR-23sr-1 csRNAs in the broken chloroplasts of Arabidopsis were less abundant than in the intact chloroplasts (Figure 8E). In addition, amount of ~16 nt csR-23sr-1 csRNAs is in agreement with our Arabidopsis deep sequencing results (Figure 8D). We noticed several RNA molecules of 80-200 nt in size that might be the multiple precursors (Figure 8C and 8E).

Figure 8
figure 8

Validation of csRNAs in chloroplasts. (A) RT-PCR analysis of ATCG00020 (chloroplast gene encoding chloroplast psbA protein), At4g10760 (nuclear gene encoding chloroplast MTA protein) and At4g15030 (nuclear gene encoding the nuclear TCP4 protein). (B) and (C) Size distribution and Northern blotting of csR-4.5sr-1 in Chinese cabbage. (D) and (E) Size distribution and Northern blotting of csR-23sr-1 in Arabidopsis. Buffer: chloroplast isolation buffer mentioned in [47]; Br_IC: intact chloroplasts of Chinese cabbage. Br_BC: broken chloroplasts of Chinese cabbage; At_IC: intact chloroplasts of Arabidopsis; At_BC: broken chloroplasts of Arabidopsis; At_total: total RNA samples extracted from Arabidopsis.

Full size image

Apart from the observations above, one small RNA core sequence, csRNA-ig-1, was represented by a large number of reads at an average size of 22 nt in each of the Chinese cabbage csRNA datasets and a specific 24 nt size in the Arabidopsis chloroplast dataset, which mapped to one intergenic RNA locus but nowhere else. This small RNA sequence is identical to the Ntc-2 small non-coding RNA reported and verified by Lung and colleagues in their tobacco chloroplast study [10]. Interestingly, csRNA-ig-1 was induced under heat stress (Additional Files 8 and 9).

Putative structure of csRNAs and their precursors

To associate small RNAs with the folded conformations of RNA, we analyzed the secondary structure of some rRNA and tRNA molecules. According to the secondary structures of 5S and 4.5S rRNA from E. coli[31, 32] and 16S rRNA from Haloferax volcanii[33], we simulated the secondary structure of some rRNAs-csRNAs of Chinese cabbage with high abundance. csR-5sr-1 and csR-5sr-2, two of the major csRNA sets derived from the Chinese cabbage chloroplast 5S rRNA (Additional File 9), matched a "single-stem-with-bulge" frame (type I) and a "self-hairpin" frame (type II) of 5S rRNA (Figure 9A). The parallel analysis of the csRNAs from 4.5S RNA of E. coli indicated that two csRNAs derived from 4.5S rRNA and three csRNAs derived from 16S RNA were located in "single-stem-with-bulge" frame (Additional File 11A). Meanwhile, three of the other csRNAs from 16S rRNA present a potential secondary structure of "self-hairpin" (Additional File 9). csR-16sr-1 is the most abundant csRNA from 16S rRNA. This csRNA and csR-16sr-2 exhibited "two-stem-with-hinge" frame (Additional File 11B).

Figure 9
figure 9

The predicted secondary structures and putative precursors of csR-5sr-1 and csR-5sr-2. (A) The prediction of csR-5sr-1 and csR-5sr-2 on the secondary structure of 5S rRNA from E. coli, according to their location numbers on Chinese cabbage chloroplast 5S rRNA. (B) The sequenced transcripts of putative precursor of csR-5sr-1. The dashed line indicates the internal deletion. (C) Alignment of the putative precursors of csR-5sr-1 with 5S rDNA. The arrow and round-head arrow directed the first and last nucleotides of 5S rRNA, respectively. csR-5sr-1 sequence is underlined in red color. The dotted lines indicate the position of PCR primer 2.

Full size image

tRNA-derived small RNAs were mostly identified by starting from the first nucleotide of the mature tRNA sequences. Compared to the cloverleaf structures of the tRNA, this position-specificity resulted in a 25 nt csRNA that had the exact same sequence as the acceptor arm from the 5' end, including the D-arm and D-loop of the tRNA; a few of nucleotides of csRNAs of 27~31 nt were matched to the stem of anticodon arm; and even a 36 nt csRNA was paired with the complete anticodon (Additional File 11C). Zhang et al. [27] reported a class of ~58-60 nt non-coding RNAs in the phloem sap of pumpkin, whose cleavage sites fell in the anticodon and D arm of tRNA. csRNAs of Chinese cabbage are much shorter than these non-coding RNAs, but both are generated in anticodon or D loop of tRNA.

To investigate the conceivable precursors of the csRNA sequences, circular RT-PCR was carried out. Two different RNA fragments of csR-5sr-1 precursor were isolated and sequenced (Figure 9B). Pre-1 was 141 nt in length, 20 nt longer than mature sequence of 5S rRNA, with a 3 nt extension to the 5' ends, and a 15 nt extension to 3' ends (Figure 9C). Pre-2 was 77 nt in length, with a 13 nt extension outer of 3' ends, but with a 58 nt internal deletion. According to the secondary structure deduced from E. coli (Figure 9A), Pre-1 and Pre-2 are the putative precursors of csR-5sr-1.

Relationship between csRNAs and the putative targets

To detect the functions of csRNAs, we searched for the transcripts of the coding genes that are complemented by the csRNAs responsive to heat stress. csR-mYCF-8, a csRNA family of Chinese cabbage (Additional File 9) that declined under heat treatment (Figure 10A), perfectly matched to the chloroplast ycf1 gene. Using adjacent sequences as primers, the real-time PCR showed that expression of gene dramatically increased in the HT seedlings compared with those in the MT seedlings (Figure 10B), implying a silencing role of csR-mYCF-8 csRNAs in expression of ycf1 gene. Similarly, atpE was a putative target of csR-mATP-2, which was up-regulated in the HT seedlings, just opposite to the reduced abundance of the csR-mATP-2 csRNAs (Figure 10). Other than ycf1 and atpE, rpoA and psbM were other two possible targets paired with csR-RPO-1 and csR-mPSB-6 csRNAs, respectively (Additional File 9). As shown by small RNA deep-sequencing and microRNA microarray, these two csRNAs were remarkably increased by heat stress in the HT seedlings (Figure 10A). Opposite to this, the expression of the rpoA and psbM gene was down-regulated in the HT seedlings (Figure 10B). These consistent results revealed that csRNAs play certain roles in regulation of gene expression.

Figure 10
figure 10

Expression analysis of the four putative target genes of csRNAs. (A) A csRNA abundance comparison between HT and MT samples from Chinese cabbage deep-sequencing datasets. (B) Fold changes of putative target gene expression.

Full size image

Discussion

Chloroplast RNA generates a novel subset of small RNAs

Although studies in recent years have reported the existence of small non-coding RNAs in DNA-containing cellular organelles [10–12], a genome-wide chloroplast profiling study is not reported yet, and therefore, the expression patterns, biological functions and biogenesis pathways of stable small RNA species remain largely unknown. From small RNA libraries of both Chinese cabbage and Arabidopsis, we selected two classes of the chloroplast-related small RNAs (csRNAs). By comparing chloroplast-specific small RNAs with csRNAs in Arabidopsis, we showed that most of the small RNAs in the csRNA dataset were derived exclusively from the chloroplast genome rather than the nuclear or mitochondrial genomes; thus, csRNA dataset may represent the features of chloroplast-derived small RNAs. Using highly purified chloroplast RNA samples, we confirmed by Northern blotting that most of the csRNAs were produced and accumulated inside chloroplasts.

Among csRNAs from the ~150 kb sequence of the chloroplast genome, about 90% were mapped to chloroplast rRNA and tRNA loci. Although fragments from rRNA and tRNA are regularly considered to be degradation contaminants, we present several clues suggesting a different viewpoint: a) Most of the tRNA/rRNA-derived csRNAs were associated with the 5' ends of tRNAs or 3' ends of rRNAs, respectively, and were even precisely aligned to the first nucleotides of mature tRNA, representing specific positions in the chloroplast genome. b) Despite wide varieties in size, the csRNA families exhibited distinct size distributions. A csRNA family is usually predominated by small RNAs of one or two determinate sizes, although it is classified as a group of small RNAs, constant in their core sequence and varying in the 5' or 3' extender length. c) Some extra nucleotides ahead and behind the matured RNA sequences were revealed from the putative csRNA precursors (such as Pre1 and Pre2 of csR-5sr-1). Considering the high abundance and end-specificity of some csRNA families such as csR-5sr-1, the csRNAs may not be the degradation products. d) csRNAs not only originated from the region corresponding to the mature RNAs but also from the externally and internally transcribed spacer regions, indicating that their biogenesis may require un-transcribed precursors or intermediate products. Collectively, our data suggest an undiscovered biogenesis pathway of small RNAs from RNA processing that differs from the existing knowledge about the single nucleotide decay from the end of a DNA or RNA chain by exonucleases.

csRNAs are highly sensitive to high temperature

Comparing the csRNAs from heat-treated samples with those from the non-stressed control group, the most remarkable change was that the rRNA-derived csRNAs declined significantly in the HT dataset to only ~50% of their original abundance shown in the MT dataset. However, the reduction in the 5S rRNA-derived csRNAs may due to the decease of chloroplast rRNA/rDNA by heat stress, as reported in maize chloroplast rRNA [34] and black mustard (Brassica nigra) [35]. The rRNA transcripts could be the sources of not only mature rRNA, but also csRNA, as most chloroplast small RNA sequences were identical to the sense strand of chloroplast DNA. Under these circumstances, csRNA biogenesis from rRNA may be passively controlled by the quantity of total transcripts. However, other possibilities still exist: the csRNAs may decrease to a certain extent in an initiative manner compared to the changes in transcripts. It is challenging to compare the changes in transcripts, mature rRNAs and csRNAs directly. Since the relations between two of the three are still indeterminable, it is unknown whether there is another factor involved in their processing pathways. One possible alternative model for the regulation of RNA levels by csRNA is as follows: polycistronic precursors produce both csRNAs and mature RNAs; as a result, fewer csRNAs lead to more mature rRNAs processed from the precursor, and less interference caused by the base-pairing competition between fragments with the same nucleotide sequences. In addition, a previous study involving 5S rDNA-derived siRNAs in DDM1 and MET1-deficient mutants suggested a role of siRNAs in 5S rDNA chromatin organisation [36], which may hint to a similar role for the 5S rRNA-originated csRNAs in the chloroplast.

Among the tRNA-derived csRNAs, the number of shorter csRNAs (15-25nt) was reduced, while the number of longer csRNAs (29-32nt) was increased. Interestingly, the 3' termini of the shorter and longer csRNAs mapped within the D arm and anticodon arm, respectively, indicating that the tRNA cleavage sites were located in the middle. tRNA halves, corresponding to the 5' or 3' sequences of the tRNAs, derived from mature tRNAs cleaved at their anticodon loops by an unidentified RNase, were able to inhibit translation and potentially transmit long-distance signals in plant phloem sap [27]. Processed tRNA halves were also detected in prokaryotes, fungi, yeast, Arabidopsis, and human cells, under various stresses, such as early amino acid starvation or oxidative stress [37–40]. Although we do not yet know whether those csRNAs associated with tRNA 5' ends in our datasets are functional 5' halves or not, there is a possibility that they are generated by a similar pathway in the chloroplast, reflecting an ancient mechanism centrally associated with stress conditions.

In contrast to small RNAs from rRNA, tRNA and mRNA, the igRNA-derived csRNAs increased by 30% after heat treatment.

csRNA may play roles in plant heat resistance

Leaf etiolation is one of the major indicators of plant sensitivity or tolerance to high temperature. When the intracellular environment is exposed to moderate heat stress, the structure of the thylakoid membrane is changed and cyclic electron flow around photo system I is increased, resulting in massive photo-oxidative stress and the concomitant release of highly cyto-toxic free radicals and reactive oxygen species [41, 42]. When the subcellular structures of the chloroplast are damaged by heat stress, photosynthetic capacity is greatly reduced [41]. This change is usually concomitant with leaf etiolation. For the heat-resistant genotypes of Chinese cabbage, leaf etiolation appears later or only slightly compared with the heat-sensitive genotypes (data not shown). In the HT seedlings, most of the csRNAs showed reduced abundances 1 hour after the 46°C heat treatment, the time period before the appearance of leaf etiolation. One hypothesis is that the heat-responsive csRNAs play roles in the maintenance of subcellular structures and photosynthetic capacity of the chloroplast. The csRNAs derived from tRNAs are possibly play therole similar to the reported tRNA halves under various stress conditions in prokaryotes, fungi, yeast, Arabidopsis and human cells [27, 38–40]. When seedlings are exposed to high temperature, the biogenesis of the heat-responsive csRNA is affected. The change in abundance of some csRNAs affects the putative target genes, which may play an important role in plant resistance to heat stress.

Plant chloroplasts are closely related to leaf development, while the latter is regulated by many miRNAs and small RNAs [43–46]. However, it remains to be elucidated how the novel csRNAs are cooperating with the canonical miRNAs in leaf development after exposure to high temperatures and if they are affecting each other. To address the roles of these heat-responsive csRNAs in plant heat resistance, a study has been planned to compare the gene expression patterns of heat-resistant cultivars of Chinese cabbage with the heat-sensitive genotypes. This study could provide insight into understanding the molecular mechanism behind csRNA-mediated heat resistance and a fast and efficient way to improve the heat resistance of Chinese cabbage and other important crops.

Conclusion

The chloroplast is an important organelle that contains a plenty of small RNAs. Our results show that many members of csRNA families were highly sensitive to heat stress. Some csRNAs respond to heat stress by silencing target genes. We suggest that proper temperature is important for production of chloroplast small RNA, which are associated with plant resistance to heat stress.

Methods

Plant materials and growth conditions

Wu-11, a heat-sensitive inbred line of non-heading Chinese cabbage (B. rapa ssp. chinensis), was used in this study, unless indicated otherwise. The seeds were sown in pots and germinated in an incubator. After one week, the seedlings were transplanted to the growth chambers in Phytotron of SIPPE (Shanghai Institute of Plant Physiology and Ecology) and grown at 22°C with 16 h of light per day. Three-week-old seedlings were autoclaved at 46°C (HT, high temperature) and 22°C (MT, moderate temperature) in two water bath kettles, respectively, for one hour. After heat treatment, some of the seedlings were used for RNA isolation while the rest were transferred to growth chambers for further growth.

The seeds of Arabidopsis thaliana were surface-sterilized in 70% ethanol for 1 min, followed by 0.1% HgCl2 for 10 min, washed four times in sterile distilled water and plated in molten 0.1% water agar on top of solid 1% sugar MS0 medium. Plates were sealed with Parafilm, incubated at 4°C in darkness for 2-3 days and then moved to a growth chamber at 22°C with 16 h of light per day.

RNA extraction and quantity detection

After heat treatment, the above-ground portions of the seedlings were harvested, quick-frozen immediately with liquid nitrogen and stored at -80°C. Each treatment comprised 10 seedling samples that were divided into two replicates. Total RNAs were extracted using TRIzol Reagent (Invitrogen). RNA concentration was measured by Eppendorf Biophotometer 6121. DNA was removed by digestion using TURBO DNA-free kit (Ambion).

Chloroplast isolation

Chloroplasts were isolated from seedlings using a Percoll (GE, 17089101) gradient-based method [47]. The intactness of chloroplast was estimated by microscopic examination (Olympus BX51 wide-field microscope with differential interference contrast) and Hill reaction [48].

Small RNA deep sequencing and sequence analysis

Chinese cabbage and Arabidopsis total RNA samples were sent to Keygene N.V., Wageningen, the Netherlands and Zhejiang-California Nanosystems Institute, China respectively, where the samples were prepared and sequenced on Illumina GAII sequencer, according to the manufacturer's protocol. The adaptor sequences in Illumina sequencing reads were removed using "vectorship" in the EMBOSS package. The small RNA with length of 9~36 nt were mapped to chloroplast genomes (GenBank accession number: DQ231548) and nuclear genomic sequences (BRAD, Brassica Database, http://brassicadb.org/brad/) of Chinese cabbage (Brassica rapa ssp. chinensis) and nuclear, chloroplast, mitochondrial genomes of Arabidopsis (http://www.arabidopsis.org). All data from different samples were normalized before comparison analysis.

Northern blotting analysis

RNA (20 μg) was fractionated on a 19% denaturing poly-acrylamide gel (PAGE), transferred to a Hybond-XL membrane (Amersham biosciences-GE healthcare) by capillary transfer using 20 × SSC buffer, and fixed by UV-crosslinking. Pre-hybridization was carried out at 42°C for 2 × 30 min using PrefectHyb Hybridization Buffer (Sigma). Oligo-nucleotide DNA probes, were labeled with γ32P ATP (5000Ci/mmol) using a T4 polynucleotide kinase (Roche). Hybridization was performed overnight at 42°C. Sizes of csRNA molecules were estimated using oligo-nucleotide DNA/RNA sense-probes with sequences same with csRNAs. In addition to equivalent RNA loading (20 μg detected by Eppendorf Biophotometer 6121), U6 spliceosomal RNA was used as an RNA loading control for total RNA sample.

Quantitative real-time reverse transcription PCR

10 mg of each Chinese cabbage total RNA sample were treated twice with 4 U of TURBO DNase (Ambion) for 30 min at 37°C. For cDNA synthesis, 200 ng RNA were reverse transcribed using SuperScript™ First-Strand Synthesis System with random hexamers (Invitrogen) according to manufacturer's recommendations. A MyiQ2 real-time PCR detection system (Bio-rad) was used for qPCR (2 min at 50°C, 10 min at 95°C, 40 × (15 s at 95 °C, 1 min at 58°C)) with the Power SYBR Green kit (Applied Biosystems; 10 μl master mix, 5 μl cDNA (diluted 1: 10 for chloroplast atpE, psbM, rpoA and ycf1; and 1: 5 for Chinese cabbage UBQ5), 1 μl of forward and reverse primer (1 μM each), 3 μl H2O; atpE primers: S-tttgctgagcttcttgtgga, A- cgaatcgaattgtttgggat; psbM primer: S-ggaacgagaatgaagagtgc, A-tcccgagatattccaaagaa; rpoA primer: S-agattctgggaggcaattct, A-aagcacttcatcaagcctcc; ycf1 primers: S-accaatccaacccatttcat, A-ccaagttcaatgttagccaga; UBQ5 primers: S-tccgtccaccttgtagaactg, A-tgaaaaccctaacggggaaa). At least four biological replications were detected for each gene. All reactions were done in triplicate, and for each reaction a 'no RT' control was included. Data analysis was performed with the iCycler IQ real-time detection system (Bio-Rad), with automatic Ct and baseline setting and UBQ5 mRNA concentrations as internal control. Relative transcript quantities were calculated with the ΔΔCt method.

Circular reverse transcription PCR (cRT-PCR)

The precise 5' and 3' ends of the csRNA precursors were determined by circular RT-PCR. The DNase treated chloroplast RNAs were incubated with 40U of T4 RNA ligase (New England Biolabs) according to manufacturer's protocol, which could generate circular RNAs with the 5' and 3' ends joined together. After phenol-chloroform extraction, single-strand cDNAs were synthesized using Revertra Ace (TRT-101, Toyobo) and specific reverse primer. The sequences of the reverse transcript products were complementary to the small RNA precursors, with breakpoints at the ends. PCR amplification were then carried out using a forward primer (cRT Primer 1: gctcgacgccaggatg) as same sequence as csR-5sr-1, and a reverse primer (cRT Primer 2: gtgttaagcttttcat) which complement to the DNA sequence following the binding site of csRNA (indicated in Figure 9). PCR amplification consisted of 30 cycles of 30s duration at 94°C, 30s at 48°C and 30s at 72°C. All cRT-PCR products were purified by MagExtractor (NPK-600, Toyobo), and then ligased into pMD18-T vector (D101A, Takara) for sequencing.