MicroRNA-mediated regulation of gene expression in the response of rice plants to fungal elicitors

Abstract

MicroRNAs (miRNAs) are small non-coding RNAs that have important regulatory functions in plant growth, development, and response to abiotic stress. Increasing evidence also supports that plant miRNAs contribute to immune responses to pathogens. Here, we used deep sequencing of small RNA libraries for global identification of rice miRNAs that are regulated by fungal elicitors. We also describe 9 previously uncharacterized miRNAs in rice. Combined small RNA and degradome analyses revealed regulatory networks enriched in elicitor-regulated miRNAs supported by the identification of their corresponding target genes. Specifically, we identified an important number of miRNA/target gene pairs involved in small RNA pathways, including miRNA, heterochromatic and trans-acting siRNA pathways. We present evidence for miRNA/target gene pairs implicated in hormone signaling and cross-talk among hormone pathways having great potential in regulating rice immunity. Furthermore, we describe miRNA-mediated regulation of Conserved-Peptide upstream Open Reading Frame (CPuORF)-containing genes in rice, which suggests the existence of a novel regulatory network that integrates miRNA and CPuORF functions in plants. The knowledge gained in this study will help in understanding the underlying regulatory mechanisms of miRNAs in rice immunity and develop appropriate strategies for rice protection.

Introduction

Small RNAs (sRNAs) are short non-coding RNAs that guide gene silencing in most eukaryotes.^1,2 Plants have 2 main classes of sRNAs, microRNAs (miRNAs) and small interfering RNAs (siRNAs),^3-5 with a major difference being their mode of biogenesis and/or function. MiRNAs are produced from single-stranded RNA precursors with unique stem-loop structures processed in 2 steps by a RNase III DICER-like (typically DCL1) to result in an miRNA-miRNA* duplex. After transfer to the cytoplasm, miRNA duplexes are loaded into the RNA-induced silencing complex (RISC), where ARGONAUTE 1 (AGO1) is the core component. However, siRNAs result from processing long perfect double-stranded RNA (dsRNA) precursors that are synthesized by the activity of a RNA-dependent RNA polymerase (RDR). Plant siRNAs can be further divided into heterochromatic siRNAs (hc-siRNAs), natural antisense siRNAs (nat-siRNAs), trans-acting siRNAs (ta-siRNAs), long small interfering RNAs (lsiRNAs) and phased secondary siRNAs (phasiRNAs).⁵ The production and function of each class of sRNAs have consistent requirements for specific members of the DCL, RDR and AGO gene families. DCL1 and AGO1 are mainly involved in the miRNA pathway.⁶

MiRNAs regulate gene expression by triggering sequence-specific cleavage or translation repression of the target transcripts.^7,8 In plants, miRNAs have a pivotal role in the regulation of gene expression during development^9,10 and adaptation to a variety of abiotic stresses, such as drought, cold, salinity, and nutrient deficiency.^11-13 Evidence is also accumulating for a role of miRNAs in the plant response to pathogen infection.^14-19 The first indication of a miRNA playing a role in plant immunity came with the finding that treating Arabidopsis plants with the flagellin-derived elicitor peptide flg22 induced accumulation of miR393, which negatively regulates transcripts for the F-box auxin receptors (i.e., transport inhibitor response, TIR). MiR393-mediated repression of auxin signaling results in bacterial resistance.²⁰ However, our understanding of the functional roles of miRNAs in biotic stress responses is far less than that in plant development and abiotic stress responses.

Most of the miRNAs reported in early articles are highly conserved throughout the plant kingdom and target transcription factors.²¹ Plants also express species-specific miRNAs that function in distinct biological processes and/or response to environmental stress.^22,23 Studies on pathogen-associated miRNAs have been conducted mainly in the interaction of Arabidopsis plants with the bacterial pathogen Pseudomonas syringae or treatment with the flg22 elicitor from P. syringae, but less is known about miRNAs mediating defense against fungal pathogens.

Plants possess a potent immune system for defense against most pathogens. Pathogenic organisms are recognized by conserved pathogen-associated molecular patterns (PAMPs, also known as elicitors) through pattern-recognition receptors (PRRs) located on the cell surface. Perception of PAMPs triggers basal defense, also known as PAMP-triggered immunity (PTI), which encompasses the immune responses against most pathogens.²⁴ PTI components include deposition of callose, production of reactive oxygen species (ROS), protein phosphorylation processes, and induction of pathogenesis-related (PR) gene expression, among others. Moreover, the essential role of the phytohormones salicylic acid (SA), ethylene (ET) and jasmonic acid (JA) in resistance to pathogens is well established in plants.^25,26 Successful pathogens have evolved mechanisms to counteract the basal defense, by delivering effector proteins into the plant cells that interfere with PTI.²⁷ In turn, many plants have evolved another layer of immunity, the so-called effector-triggered immunity (ETI) consisting of proteins encoded by resistance (R) genes that specifically identify the corresponding effector proteins produced by the pathogen. Most studies of plant immunity have focused on the transcriptional regulation of protein-coding genes, and much less is known about miRNA-mediated regulatory processes. We have some examples of miRNAs that guide cleavage of disease resistance genes in Solanaceae and Leguminosae species.^16,28,29 More recently, an apple miRNA, Md-miRLn11, was reported to target a nucleotide-binding site leucine-rich repeat (NBS–LRR) protein that mediates resistance against apple leaf spot disease.³⁰ The current scenario is that miRNAs are implicated in both PTI and ETI responses.

Rice (O. sativa L) is one of the most important cereal crops in the world and the main staple food crop for more than 50% of the world's population. The activity of certain miRNAs significantly affects traits of agronomic importance in rice, such as tiller growth, early flowering, panicle and grain production.^31,32 However, rice yields are severely compromised by the fungal pathogen Magnaporthe oryzae, the causal agent of the rice blast disease.³³ Indeed, rice blast is considered the most important fungus-caused disease in plants in terms of scientific and economic relevance.³⁴ Both PTI and ETI are involved in immunity against the rice blast fungus.³⁵ Although the rice/M. oryzae pathosystem has been extensively analyzed at the molecular and cytological levels, miRNA-mediated defense responses have not been fully described in this pathosystem. The function of distinct rice miRNAs in blast disease resistance has only recently been demonstrated, for miR160a, miR398b and miR7695.^17,36 Target genes for rice miRNAs have been mainly predicted by computational approaches, and only a small fraction of targets has been experimentally validated.

Owing to the scientific and economic importance of the rice/M. oryzae pathosystem, this study focused on rice miRNAs regulated by the rice blast fungus. We prepared sRNA and degradome libraries from rice tissues (leaves, roots) treated, or not treated with M. oryzae elicitors. Use of high-throughput Illumina sequencing revealed a comprehensive picture of the miRNA transcriptome in each tissue and the response of rice miRNAs to fungal elicitors. Furthermore, 9 novel miRNAs from rice were identified. A degradome approach was used to identify transcriptome-wide miRNA targets in rice. An interesting observation was the identification of miRNA-guided cleavage of Conserved Peptide upstream Open Reading Frames (CPuORF)-containing transcripts. We provide evidence for miRNA-mediated regulatory networks controlling sRNA biogenesis and functioning, and hormone signalling cross-talk in the rice response to fungal elicitors. The data will serve as foundation for studies addressing fundamental molecular mechanisms that govern rice immunity.

Results

Genome-wide expression profiling of rice miRNAs

To obtain a genome-wide survey of miRNAs in rice and their responsiveness to fungal elicitors, we constructed sRNA libraries from rice leaves and roots treated or not with elicitors obtained from the rice blast fungus M. oryzae. Elicitors are widely used to trigger typical defense responses in numerous plant species, including rice.³⁷ Elicitor-treated and mock-treated tissues were harvested at 2 times of elicitor treatment (30 min and 2 h). Illumina Solexa sequencing of sRNA libraries generated 118810219 reads (approx. 62.5 and 56.3 million reads from leaf and root libraries, respectively) (Fig. S1A). After trimming the adapter sequences and short sequences, a total of 47866683 and 41541308 mappable sRNA sequences were obtained from leaves and roots, respectively. SRNAs then underwent a BLAST search against the known non-coding RNA families (rRNAs, tRNAs, small nuclear RNAs and small nucleolar RNAs) deposited in the Rfam Genbank databases, and distinct sRNAs belonging to these categories were removed. For details about the bioinformatic analysis of sequencing data, see Supplementary Methods. A total of 7660460 and 3022421 unique sequences were generated from leaf and root libraries, respectively (Fig. S1A). Consistent with the typical sRNA distribution in plants, most of the sRNAs were 18–27 nt in length in the 2 tissues. The highest abundance was found for 24-nt sequences in terms of both total abundance and unique sequences (Fig. S1B, C).

SRNA sequences perfectly mapping the rice genome were searched against the miRBase database (version 21, http://www.mirbase.org)³⁸. Currently, miRBase has 713 registered rice miRNAs representing 334 families. Following the current annotation, the mature miRNAs identified in our sequencing dataset originating from opposite arms of the same pre-miRNA are denoted with a -5p or -3p suffix (miRNA or miRNA* species). A total of 705 miRNAs and miRNAs* representing 332 known miRNA families were identified in leaf libraries (Table S1). In roots, 696 miRNAs corresponding to 326 miRNA families were detected (Table S1). Thus, the sequencing depth obtained in this study was sufficient for a comprehensive differential expression analysis of rice miRNAs and presumably also for the identification of novel miRNAs from rice.

By analyzing the Illumina sequencing data set, we obtained the expression profile of rice miRNAs in each tissue (control libraries). Members of the miR166, miR168, and miR396 families were the most abundant in rice leaves (Fig. 1A, left panel). In roots, miR166 family members and miR2863b showed the highest expression (Fig. 1A, right panel). However, miR2863 accumulated at a relatively low level in leaves (Table S1). Differences in the expression of miRNAs and/or among members of a particular miRNA family in one or another tissue might indicate a functional divergence in these tissues.

Figure 1.

Expression profiling of known miRNAs from rice. (A) Expression of known miRNAs in leaves and roots of rice plants. Reads retrieved from the Solexa/Illumina sequencing dataset for each family member in control libraries were normalized to the total count of reads obtained in the corresponding library. Only the most abundantly expressed miRNAs are presented. Asterisks denote miRNAs examined in (B and C). Details on miRNA expression in each tissue are in Supplemental Table 1. (B, C) Expression of miRNAs identified in small RNA libraries from rice tissues by Northern blot analysis (B) or stem-loop RT-qPCR (C). Lower panel in (B) shows ethidium bromide staining of RNA samples. Oligonucleotides used as probes in (B) are indicated on the right side.

To validate the expression pattern of miRNAs obtained by deep sequencing, we randomly selected 8 miRNAs with differential accumulation in leaf or root rice tissues. Northern blot analysis confirmed the expression of miRNAs showing high and moderate abundance in our sequencing dataset (e.g., miR156a, miR172c, miR2863b, miR5076), whereas the expression of low-abundant miRNAs (e.g., miR1857, miR2096, miR5147, miR5819) was validated by stem-loop RT-qPCR followed by nucleotide sequencing (Fig. 1C; details on miRNA abundance are in Table S1). According to the miRBase registry (release 21), the low-abundant miRNAs (miR1857, miR2096, miR5147, miR5819) are identified only in rice.

A detailed analysis of Illumina sequencing data of sRNA libraries also revealed the presence of miRNA sequences representing new members of known miRNA families. They were named as miR1861p, miR2120b, miR5801c, and miR6245b (Fig. 2; the nucleotide sequence of these precursors is in Fig. S2).

Figure 2.

Precursor structures of novel members of known miRNA families. Small RNA sequences recovered from the Solexa/Illumina sequencing data mapping into these structures are represented by black bars. The nucleotide sequences of these precursor structures are in Supplementary Figure 2.

Identification of novel miRNAs from rice

The identification of previously uncharacterized miRNAs from rice was a major objective of this study. Toward this end, we used the miRDeep-P software with default parameters.³⁹ Briefly, we computationally predicted miRNA stem loop precursor structures and searched for sRNAs mapping opposite to each other at both strands of the hairpin (e.g., miRNA/miRNA* sequences) having the characteristic 2-nt 3′-overhangs, a signature of DICER cleavage.^38,40 Mismatches between sRNA reads and the rice genome were not allowed. In this way, we identified 9 loci that fulfilled the fold-back structure criterion for miRNA precursors (Fig. 3A; precursor nucleotide sequences are in Fig. S3; expression data are in Table S2). The names assigned in the miRBase registry for the novel miRNAs identified in this work were osa-miR11336, osa-miR11337, osa-miR11338, osa-miR11339, osa-miR11340, osa-miR11341, osa-miR11342, osa-miR11343 and osa-miR11344.

Figure 3.

Precursor structures and detection of novel miRNAs from rice. (A) Precursor structures of novel miRNAs. Small RNA sequences mapping into these structures are represented by black bars. Additional information on the nucleotide sequence and chromosomal location is in Supplemental Figure 3. (B) Northern blot analysis of novel miRNAs. Total RNA samples (200 µg) were analyzed. The same blots were successively stripped and re-probed with³²P-end-labeled oligonucleotides. Except for miR11341, which was detected in roots, all novel miRNAs accumulated in rice leaves. Lower panels show ethidium bromide staining of RNA samples.

SRNA Northern blot analysis was carried out to validate the newly identified miRNAs from rice (Fig. 3B). Both miRNA-5p and miRNA-3p species were detected for 8 of the 9 miRNAs identified, which further supports that they represent previously uncharacterized miRNAs from rice. These miRNAs accumulated at relatively low levels in rice tissues as judged by both the low number of reads found in the Illumina sequencing data (Table S2), and the large amount of RNA needed for their detection by Northern blot analysis. Five of the novel miRNAs located in intergenic regions, whereas 4 mapped to the intronic region of a gene (Fig. 3A). None of these novel miRNAs have obvious orthologs in any other plant species for which genomic sequences are available (NCBI database) suggesting that they might represent novel rice-specific miRNAs.

Elicitor-responsiveness of rice miRNAs

The sequencing frequencies for miRNAs in our sRNA libraries, as calculated from normalized libraries, were used to investigate the elicitor-responsiveness of rice miRNAs. Figure 4A shows representative examples of elicitor-induced alterations in accumulation of known miRNAs (detailed information on the abundance for all known miRNAs identified in each tissue and condition is in Table S1). Consistent with the role of miRNAs as modulators of gene expression, a dynamic response occurs on the accumulation of rice miRNAs in response to fungal elicitors. In some cases, the elicitor-responsiveness of a particular miRNA showed the same trend of expression in the 2 tissues (e.g., miR2863b, down-regulated), whereas in other cases, a miRNA showed a different response to elicitors depending on the tissue (e.g., miR6255, up-regulated in leaves but downregulated in roots) (Fig. 4A). Also, a different response could be observed for a particular miRNA at one or another time of elicitor treatment.

Figure 4.

Elicitor-responsiveness of known miRNAs from rice. (A) Expression analysis of known miRNAs in leaves and roots at 30 min or 2 h of elicitor treatment (light and dark bars, respectively) as determined by the logarithm of fold change (elicitor vs. control). Representative examples are shown (see Table S1 for detailed information on the expression of the complete list of rice miRNAs). Asterisks denote miRNAs examined in (B). (B) Northern blot analysis of miR156a, miR529b and miR5078 in control and elicitor-treated rice leaves. Total RNA samples (70 µg) were analyzed. Oligonucleotides used as probes are indicated on the right. c, control; e, elicitor.

The elicitor-responsiveness of known miRNAs accumulating at high or moderate level in rice tissues (as judged by the number of reads in the sRNA sequencing data set) was further validated by Northern blot analysis (Fig. 4B). These miRNAs included miR156a, miR529b and miR5078. A similar trend in the response to elicitors was observed when comparing results obtained by Northern blot analysis and sequencing data. Like known miRNAs, the newly identified miRNAs also showed a dynamic response to treatment with fungal elicitors (Fig. S4; Table S2).

Collectively, analysis of sRNA deep sequencing data revealed alterations in the accumulation of an important number of rice miRNAs in response to treatment with M. oryzae elicitors.

Identification of target genes of elicitor-regulated rice miRNAs

Most plant miRNAs have extensive complementarity to their target genes and regulate their expression predominantly through mRNA cleavage. The slicing activity on their target typically occurs between the nucleotides 10 and 11 from the 5′ of the miRNA, and the resulting 3′ fragment of the target mRNA possesses a free 5′ monophosphate. These cleavage products can be recovered by RNA ligase-mediated ligation, whereas the full-length cDNA with a 5′ cap structure or other RNAs lacking the 5′ monophosphate group are not compatible for ligation. This property was exploited to validate miRNA-mediated cleavage of target transcripts.⁷ Later on, a degradome sequencing technology was developed for high-throughput miRNA target identification in plant species.^41,42

To identify target genes of elicitor-regulated miRNAs, we generated degradome libraries from both control and elicitor-treated rice leaves (30 min and 2 h of treatment; same plant material used for construction of the sRNA libraries). Illumina sequencing of degradome libraries yielded 128.3 million reads, of which 51568 unique signatures could be mapped to the rice transcriptome (statistics of degradome sequencing are in Table S3). The miRNA cleavage sites were identified by using the PAREsnip platform developed for discovery of miRNA/target interactions evidenced through degradome sequencing.⁴³ The workflow used to identify miRNA targets is in Figure 5A. To visualize the cleavage events within the target mRNAs, we plotted the abundance of each signature as a function of its position in the target transcript (target-plots, or t-plots; representative t-plots are in Fig. S5A). The identified targets were classified into 5 categories based on the relative abundance of signatures at the target site and along the transcript (categories 0, 1, 2, 3 and 4).⁴³ This analysis identified 602 targets for 299 of the 332 miRBase-annotated rice miRNA families. Transcripts showing high abundance reads at potential cleavage sites, described as categories 0, 1 and 2, represent the strongest evidence for true cleavage products.⁴³ Accordingly, in this work we considered only targets in categories 0, 1 and 2. Representative miRNA targets confirmed by degradome analysis for known rice miRNAs are in Table 1 (the complete list of target genes in categories 0, 1 or 2 is in Table S4).

Figure 5.

Confirmed target genes for rice miRNA using degradome sequencing. (A) Pipeline for target identification in degradome sequencing. (B) MapMan annotation of miRNA targets validated by degradome analysis in control and elicitor-treated leaves. Only the validated target genes identified by degradome sequencing in categories 0, 1 and 2 were considered. CM, carbohydrate metabolism. (C) Schematic of the Conserved-Peptide upstream Open Reading Frame (CPuORF3)-containing OsbZIP38 gene targeted by miR5819. Validation by 5′RACE of miRNA-mediated cleavage of CPuORF3-containing bZIP transcripts is shown.

 

Among the miRNA targets identified by degradome sequencing, signatures associated with conserved miRNA targets were the most abundantly represented. Most were classified as category 0, thus confirming the accuracy of our degradome analysis (Fig. S6). Conserved miRNAs targeted several family members (e.g.,, miR156, miR164 and miR169), whereas non-conserved miRNAs tended to target few genes (e.g., miR2104, miR5075, miR5518 and miR5809). Cleavage events corresponding to non-conserved targets mediated by conserved miRNAs were also discovered. For instance, besides SPL genes, miR156b target genes encoding a DUF260 domain-containing protein and O-methyltransferase (Table S4). MiR159abcdef targeted MAP kinase 8 in addition to MYB transcription factor genes. Degradome analysis also revealed an NBS-LRR disease resistance gene as the target gene for miR167. Therefore, besides having conserved targets, deeply conserved miRNAs might also have specialized functions by regulating non-conserved targets. Similar to reports by other authors,^44,45 we found no clear association between the miRNA level and the cleavage frequency of target transcripts.

For a functional characterization of all miRNA targets identified, we performed a MapMan analysis⁴⁶ of validated target genes for miRNAs identified in control or elicitor-treated rice tissues. In elicitor-treated tissues, genes targeted by miRNAs showed a strong enrichment in the subcategory of biotic stress (Fig. 5B). Among the target genes with known functions in biotic stress responses were chitinase (targeted by miR426), lipid transfer protein (LTP, targeted by miR5819), disease resistance (targeted by miR164e, miR167, miR1439 and miR1850), and a D-mannose binding lectin protein (targeted by miR1439) (Table 1). Other target genes of note are those involved in detoxification systems, such as copper/zinc superoxide dismutase (SOD1 and SOD2, targeted by miR398), L-ascorbate oxidase (targeted by miR528), glutaredoxin (targeted by miR531a), and glyoxalase (targeted by miR1425 and miR1861c) (Table 1).

MapMan analysis revealed an increase in miRNA target genes classified as “Signaling” in elicitor-treated tissues compared to non-treated tissues (3% in control tissues; 9% in elicitor-treated tissues), with a decrease in those in the category of “Development” (14% in control tissues; 8% in elicitor-treated tissues). The functional category of carbohydrate metabolism was enriched in elicitor-treated but not control tissues. Likely, the target genes identified in degradome libraries are components of the elicitor-induced signal transduction pathways leading to activation of defense responses.

Of note, miR393 is well known to cleave the TIR/AFB2 clade of auxin receptors during PTI against P. syringae in Arabidopsis plants.²⁰ Consistent with this finding, TIR1 and AFB2 (OsFBL21 and OsFBL16, respectively) were found as targets of miR393 in our degradome analysis (Table 1). Interestingly, we found that, in addition to TIR1 and AFB2, miR393 also cleaves Suppressor of gene silencing 3 (SGS3) transcripts (Table 1). A soybean homolog of Arabidopsis SGS3 was previously reported as the target of a soybean miRNA (Soy_25).⁴⁷

Degradome analysis revealed targets for both the miRNA and miRNA* (i.e., miRNA-5p and miRNA-3p sequences) for 19 miRNAs supporting that their miRNA* sequences may be functional (Table 1). Some examples are miR160e, miR393b, miR530, miR1428e and miR2098. Moreover, target analysis revealed the regulation of distinct mRNA targets by 2 different miRNAs (Table S5). These targets included transcripts encoding RDR2 (RNA-dependent RNA polymerase 2) (targeted by miR818cd and miR1850.2, AP2 domain-containing protein (targeted by miR171abcdi and miR1426), tyrosine protein kinase (targeted by miR818cd and miR1436), glyoxalase (targeted by miR1425 and miR1861c), and a transcriptional regulator (targeted by miR531 and miR5075). Co-regulation of target transcripts by 2 or more miRNAs was previously reported.⁴⁴ However, whether the miRNAs targeting the same gene identified in this work act in concert to regulate target gene expression remains to be determined, as does whether cleavage of their corresponding target transcripts is biologically relevant or merely a neutral event.

Regulation of CPuORF-containing transcripts by rice miRNAs

Upstream open reading frame (uORF)-containing genes represent a specific class of selectively translated genes in which a short peptide (sPEPs) encoded by the uORF sequence modulates translation of the downstream major ORF. Typically, uORFs lie upstream of the main protein coding region (i.e., in the 5′ UT region of the mRNA). Although uORFs are a common feature in many eukaryotic mRNAs, including plants, those encoding conserved uORFs (CPuORFs) are relatively rare and occur in less than 1% of angiosperms.^48,49

Of interest, 3 CPuORF-containing genes were identified among the target genes for rice miRNAs. These genes include CPuORF3-OsbZIP38 (targeted by miR5819), CPuORF4-OsbZIP27 (targeted by miR5075), and CPuORF7-SAM decarboxylase (targeted by miR2101) (Table 1; Fig. S7). Of them, miR5819 cleaves OsbZIP38 transcripts within the nucleotide sequence corresponding to the short peptide encoded by CPuORF3, which is located at the 5′ UT region of OsbZIP38 (Fig. 5C). Consistent with results obtained by degradome analysis, 5′-RACE identified cleavage fragments at the expected site of CPuORF3-bZIP38 transcripts, which further supports that these transcripts are cleaved by miR5819 (Fig. 5C). The cleavage site for miR5075 and miR2101 located at the coding region of OsbZIP27 or the 3′-UT region of SAM decarboxylase, respectively (Fig. 7). The three miRNAs targeting CPuORF-containing genes (miR5819, miR5075, miR2101) are rice-specific according to the current miRBase registry (release 21).

Pathways regulated by elicitor-responsive miRNAs in rice

Combined sRNA and degradome analysis allowed us to obtain a comprehensive list of miRNA/target interaction pairs in rice and revealed interesting regulatory networks mediated by elicitor-responsive miRNAs. We provide evidence for an important number of miRNAs (and their corresponding target genes) involved in sRNA biogenesis and functioning machinery, including the miRNA pathway (miR162/DCL1; miR168/AGO1), hc-siRNA pathway (miR818 and miR1850/RDR2; miR1847/AGO4), and ta-siRNA pathway (miR168/AGO1 and miR393/SGS3) (Fig. 6A). As well, we identified miRNA/target gene pairs involved in several subnetworks associated with hormone signaling and crosstalk between defense-related hormones, namely ET, SA, JA and auxin signaling as well as polyamine biosynthesis (Fig. 6B, C). For instance, we detected a miR1846-guided regulation of ACC oxidase (ACO; final step of ethylene biosynthesis), and several miRNAs that regulate the expression of ethylene-responsive genes (ERF33, ERF43, ERF73, ERF90, EREBP56, and EREBP124) (Fig. 6B). Furthermore, we identified distinct miRNA/target gene pairs controlling the conversion of S-adenosyl-L-methionine (SAM) to the specific precursor molecules required for the production of Me-SA, Me-JA, or polyamines. These signaling pathways are connected with the ethylene signaling pathway through SAM. These findings will be discussed in more detail below.

Figure 6.

Overview of regulatory networks in which miRNA/target pairs function during the rice response to fungal elicitors. Small RNA and degradome sequencing data were used to establish regulations in the indicated pathways. All target genes were supported by degradome sequencing. miRNAs targeting these genes are boxed. Except for EDF1, all the indicated target genes are classified in categories 0, 1 or 2 in the degradome analysis.⁴³ (A) Small RNA biogenesis and functioning. AGO, ARGONAUTE; DCL, DICER-like; RDR, RNA-dependent RNA polymerase; SGS3, Suppressor of gene silencing. (B) Ethylene signaling pathway and crosstalk with salcylic-acid and jasmonic-acid signaling pathways, and polyamine biosynthesis. EDF1, ethylene response DNA binding factor 1; EREBP, ethylene-responsive element binding protein; ERF, ethylene responsive factor; ETR, ethylene receptor; JMT, jasmonic acid carboxyl methyltransferase; Me-JA, methyl-jasmonic acid; Me-SA, methyl-salicylic acid; SAM, S-adenosylmethionine; SAMT, salicylic acid carboxyl methyltransferase. (C) Auxin signaling pathway. ACC, 1-aminocyclopropane-1-carboxylate; ACO, ACC oxidase; ARF, auxin response factor; TIR1, TRANSPORT INHIBITOR RESPONSE 1.

Targets of novel miRNAs from rice

To further understand the biological function of the newly identified miRNAs from rice, we used the degradome sequencing dataset and identified target genes for 6 of the 9 novel miRNAs (Table 2). All these target transcripts have degradome tags by previously described parameters (representative t-plots for some target genes for novel miRNAs are in Fig. S5B). Unlike conserved miRNAs, the target genes for novel miRNAs were not enriched in transcription factors. Most of the miRNA targets for novel miRNAs from rice are involved in biotic stress, whereas other target genes are involved in vesicle transport, cell wall, lipid metabolism and development. The identification of target genes for the newly identified miRNAs provides further evidence that they are bona fide miRNAs from rice. Target genes not being detected for the 3 novel miRNAs could be due to the low abundance of the target transcript, inefficient miRNA-directed cleavage, or miRNA-guided translational repression of target genes.

 

Discussion

In this study, we present a comprehensive characterization of rice miRNAs that are regulated by elicitors from the rice blast fungus M. oryzae. We identified miRNAs representing 332 and 326 miRNA families in sRNA libraries from rice leaves and roots, respectively. Most contained one or more members with elicitor responsiveness. We previously reported a group of rice miRNAs belonging to 63 miRNA families that respond to treatment with M. oryzae elicitors.¹⁷ Here, in-depth analysis of the miRNA transcriptome validated previous data and significantly extended the list of elicitor-regulated rice miRNAs. The large number of miRNAs showing elicitor responsiveness and their dynamic response to elicitor treatment reflects the complexity of processes that are under miRNA regulation. Presumably, dynamic alterations of miRNA accumulation would allow for a fine-tuning of host gene expression, which would then contribute to maintain timely and appropriate levels of target transcripts in the cell. These findings also support that the rice response to M. oryzae elicitors is controlled, at least in part, by miRNA-mediated cleavage of an important number of rice genes. Some of the miRNAs identified in this study may represent rice-specific miRNAs, some having specific functions in PTI responses against the blast fungus. Recently, Li et al.¹⁹ reported miRNAs that differentially respond to blast infection in resistant and susceptible rice varieties. However, in that study an important number of miRNAs mapped to the M. oryzae genome (525%– depending on the sRNA library) indicating that the identified miRNAs were of both rice and/or M. oryzae origin. Use of fungal elicitors excludes the possibility of miRNAs of fungal origin being in the sRNA sequencing data set.

Deep sequencing of the sRNA transcriptome also allowed the confident identification of 9 novel miRNAs from rice (miR11336 to miR11344). Moreover, new members of known miRNA families are described. While potentially sacrificing several bona fide miRNAs, we used a highly stringent approach to avoid false positives. In this way, we identified both miRNA-5p and -3p sequences with a 2-nt 3′ overhand, a signature of DCL activity, in our sRNA sequencing dataset for all 9 novel miRNAs. Although the newly identified miRNAs were expressed at relatively low levels, Northern blot analysis confirmed the accumulation of 8 novel miRNAs. Furthermore, in this work, we took advantage of the public sRNA high-throughput (HTP) sequencing data of rice. We searched for the sRNA species mapping to the novel miRNA candidates in rice DCL1 interference transgenic lines (dcl1-IR plants).⁵⁰ The data sets included GSM520637 prepared from seedlings of rice DCL1 RNAi transgenic lines and GSM520640 generated from control wild-type plants, all retrieved from GEO (http://www.ncbi.nlm.nih.gov/geo/). Compared to the wild-type plants, the dcl1–1R plants accumulated lower levels of miR11336 (both −3p and −5p sequences), miR11337 (−5p sequence), miR11338 (−3p sequence), miR11339 (both −3p and −5p sequences), miR11340 (−3p sequence) and miR11344 (−5p sequence) (Figure S8). The dependency on DCL1 for accumulation of these sRNAs further supports that they are bona fide novel miRNAs.

Degradome analysis confirmed target genes that are subjected to miRNA-guided regulation in rice, which included conserved and non-conserved targets for known miRNAs. GO analysis of target genes for miRNAs confirmed by degradome analysis revealed enrichment of genes involved in stress responses in elicitor-treated rice tissues. For instance, ROS species are known to be produced under infection conditions, which can rapidly damage biomolecules, so a balance must be maintained between the production and scavenging of ROS. As expected, we found that genes involved in protection against oxidative stress, such as SOD1 and SOD2, glutaredoxin, ascorbate oxidase, or glyoxalase, were among the target genes of elicitor-regulated miRNAs (Table 1). One elicitor-regulated identified was miR398, which targets SOD1 and SOD2. Overexpression of miR398b in rice enhances resistance to M. oryzae.¹⁹Ascorbate oxidase (targeted by miR528) is an apoplastic enzyme that catalyzes the first step in degradation of ascorbic acid, thus providing the major redox buffering capacity of the apoplast. Ascorbate oxidase has also been proposed to catalyze the oxidative decarboxylation of auxin, which suggests a role in regulation of auxin levels.⁵¹ In tobacco plants, oxidation of apoplastic ascorbic acid has been associated with loss of the auxin response and susceptibility to P. syringae.⁵² The identified miRNA/target gene pairs involved in protection against oxidative stress might function in a concerted manner to maintain the dynamic balance of ROS levels, thereby maintaining the physiological redox status of the plant cell during elicitor treatment.

Among the confirmed targets for elicitor-regulated miRNAs, we identified components of the plant basal defense response, such as LTP, chitinase and PPR-containing protein genes.^53-55 We also validated a miRNA-guided cleavage of rice mannose-binding lectin and remorin genes (targets of miR1439 and miR2925, respectively) (Table 1). Plant mannose-binding lectins are crucial for plant defense signaling during pathogen attack by recognizing specific carbohydrates on pathogen surfaces.⁵⁶ Remorin proteins are found in the plasma membrane in specialized compartments known as membrane rafts, which are platforms for signal transduction during plant–microbe interactions.⁵⁷ Our degradome analysis also identified an important number of transcription factors. Presumably, each transcription factor might further regulate a set of other genes; thus, elicitor-regulated miRNAs might control numerous genes and processes through a complex gene regulation network. In addition, our degradome analysis identified targets for 6 of the 9 novel miRNAs reported in this work. The targets included disease resistance genes (RPP13, MLO), defensin, PPR proteins, ankyrin, and vesicle-associated membrane protein genes.

However, for many miRNAs, we did not identify any target. Taking into account that we limited the search in degradome libraries to categories 0, 1, and 2 (PAREsnip), it is likely that many other miRNA/target gene interactions will be missed. Several possibilities could explain why target genes for miRNAs were not detected. As previously mentioned, some target genes might be difficult to detect because of low abundance or because they guide translational repression of target genes. Alternatively, because we analyzed only 2 times of elicitor treatment, a time-delay between upregulation of a particular miRNA and target cleavage may occur, as was previously reported for some miRNA/target interactions.⁵⁸

Although star strands of the duplex have been traditionally considered nonfunctional, evidence here presented supports that the star strand might also have a role in silencing. Thus, we identified miRNA and miRNA* sequences for 19 rice miRNAs for which degradome data confirmed the existence of targets for the 2 members of the miRNA/miRNA* duplex. A well-known example of a miRNA in which the 2 sRNAs of the duplex are functional is the Arabidopsis miR393. miR393 targets TIR1, whereas miR393* regulates a Golgi-localized SNARE protein (MEMB12) that mediates secretion of PR1.⁵⁹ In other studies, miR171* was found to silence the expression of the Arabidopsis SUVH8 gene, also known as SET DOMAIN GROUP21.⁶⁰ Confirmation of target genes for both miRNA and miRNA* of distinct rice miRNAs support the operation of miRNA*-mediated mRNA cleavage mechanisms in rice. Whether the 2 members of the duplex contribute to PTI in rice remains to be determined.

miRNA-mediated gene regulatory networks in the rice response to fungal elicitors

To shed more light on the functional role of elicitor-regulated miRNAs, we searched for regulatory networks enriched in miRNAs that were supported by the identification of the corresponding target genes by degradome analysis. As previously mentioned, we identified an important number of miRNA/target gene pairs involved in sRNA biogenesis and functioning, namely miRNAs, hc-siRNAs and ta-siRNAs (see Fig. 6A). Each pathway contains 2 or more steps in which a miRNA/target pair participates for generating a specific class of sRNA or accommodation of the sRNA in the RISC silencing complex. The observed elicitor-regulated expression of miR162 and miR168 (targeting DCL1 and AGO1 transcripts, respectively) is consistent with a regulation of the miRNA machinery itself by fungal elicitors. Presumably, an adjustment of miR162 and miR168 levels by elicitors would contribute to maintenance of appropriate levels of DCL1 and AGO1 and, accordingly, miRNA functioning. Supporting this hypothesis, alterations in miR168 and/or AGO1 expression have been described in different plant–pathogen interactions or in response to treatment with elicitors.^61-63 Also, dcl1 and ago1 mutants are compromised in PTI responses or flg22-induced disease resistance in Arabidopsis.^36,64

Hc-siRNAs guide DNA methylation at target genomic loci via RNA-directed DNA methylation (RdDM), which reactivates transposons and transcription of silenced genes.⁶⁵ RDR2 and AGO4 are components of the hc-siRNA pathway. Here, we found that RDR2 and AGO4 transcripts were regulated by the activity of distinct miRNAs (both miR818 and miR1850 target RDR2, whereas miR1847 targets AGO4). This evidence suggests the existence of miRNA-mediated control of the production and functioning of hc-siRNAs in plants. In other studies, the contribution of RdDM in Arabidopsis antibacterial defense was documented.⁶⁶ Viral and bacterial infections can also modify hc-siRNAs production and alter DNA methylation.^67-69

Our findings also point to a possible regulation of ta-siRNA production by elicitor-regulated miRNAs. We found regulation of SGS3 and AGO1 (by miR393 and miR168, respectively), whose activity is known to be important in the generation of ta-siRNAs.⁷⁰ That SGS3 is involved in pathogen resistance is further supported by Arabidopsis sgs3 mutants exhibiting enhanced susceptibility to pathogen infection.⁷¹

Overall, profiling of rice miRNAs and their target genes reinforce the existence of self-regulatory mechanisms of the miRNA pathway while revealing a miRNA-mediated regulation on the hc-siRNA and ta-siRNA machinery as part of the rice response to fungal elicitors.

We can draw conclusions regarding miRNA-mediated regulatory networks involved in hormone signaling and crosstalk in hormone signaling. The phytohormones ET, SA, and JA as well as polyamines play an important role in disease resistance, including resistance to M. oryzae in rice. The methyl esters of JA and SA (Me-JA, Me-SA) can trigger defense responses in plants.^54,72 In this study, we identified an important number of miRNA/target gene pairs that control ET signaling or connect the ET pathway with the Me-JA, Me-SA and polyamine biosynthetic pathways. These pathways are connected at the level of SAM, which serves as a precursor for the production of ET, JA- or SA-methyl derivatives, and polyamines. In particular, distinct SAM-dependent carboxyl methyltransferases use SAM as the methyl donor in the biosynthesis of Me-SA (SAMT, SA carboxylmethyl transferase) or Me-JA (JMT, JA carboxylmethyltransferase). Of note, genes encoding SAMT and JMT were under miRNA regulation (miR5512ab and miR1428bcd, respectively) (see Fig. 6B). Moreover, SAM is decarboxylated by the activity of SAM decarboxylase (the first enzyme involved in the production of polyamines), and SAM decarboxylase transcripts were also under miRNA regulation (miR2101, see Fig. 6B). Thus, we can propose a regulatory mechanism in which distinct miRNA/targets operate for the control of SAM level and SAM distribution to sustain defense-related signaling pathways.

Another network in which several miRNA/target gene pairs participate involves auxin signaling. Degradome analysis revealed target genes that function upstream (auxin perception by F-box auxin receptors) or downstream (ARF6, ARF8, ARF9, ARF10, ARF18, ARF20 and ARF22) of the auxin signaling pathway. These genes are regulated by elicitor-responsive miRNAs. We also revealed a miR5809-guided cleavage of GH3, which encodes indole-3-acetic acid (IAA)-amido synthetase. The GH3 enzyme catalyzes the synthesis of IAA-amino acid conjugates, thus providing a mechanism for maintaining auxin homeostasis by conjugating excess IAA to amino acids. Thus, treatment with fungal elicitors may alter the accumulation of miRNAs known to modulate auxin signaling pathways.

In addition, miRNAs are key regulators of auxin response pathways associated with developmental programmes in plants.⁷³ Auxin interacts with other phytohormone signaling pathways during plant development. MiRNA-mediated regulation of auxin signaling contributes to antibacterial resistance in Arabidopsis, as illustrated by regulation of TIR/AFB2 auxin receptor genes by miR393.²⁰ In other studies, repression of the auxin response pathway increased the susceptibility to necrotrophic fungi.⁷⁴ In rice, auxin homeostasis regulates the expression of rice defense genes and resistance to blast fungus.⁷⁵ In addition to reprogramming host developmental processes, elicitor-regulated accumulation of these miRNAs might well contribute to regulation of defense responses directly or indirectly via cross-talk between auxin and other defense-related hormones.

Hormonal cross-talk has emerged as a major player in regulating tradeoffs between growth and PTI-mediated defense.⁷⁶ The observed connections between ET signaling and the SA and JA signaling pathways via miRNA activities raises some interesting questions regarding the possible role of miRNAs in regulating plant development and defense responses in rice. Because miRNAs provide quantitative regulation of target gene expression rather than on–off regulations, elicitor-mediated regulation of gene expression might help fine-tune host gene expression in reprogramming developmental programs and defense while avoiding the fitness costs associated with the expression of host defense responses. This process would then be part of the adaptive strategy of plants to pathogen infection. A fine-tuned regulation of immune responses, PTI and/or ETI, would avoid negative effects in plant traits of agronomical interest, such as biomass and seed production. The literature contains several examples of miRNAs that regulate growth and development and also mediate plant responses to biotic or abiotic stress.⁷⁷ As an example, Arabidopsis miR396 is a developmental regulator in the reprogramming of root cells during cyst nematode infection.⁷⁸ Understanding the specific function of rice miRNAs controlling plant responses to pathogen infection and developmental cues will provide powerful tools to optimize the growth–defense balance, which, in turn, will help to improve rice productivity. How plants coordinate the dual function of certain sRNAs in development and defense deserves further investigation.

Regulation of CPuORF-containing genes by elicitor-responsive rice miRNAs

uORFs are cis-acting RNA elements involved in translational regulation of the main coding region located downstream of the uORF sequence encoding the conserved short peptide (sPEP). By analogy to miRNAs, which are riboregulators of gene expression, uORF-encoded short peptides function as “peptoregulators” that mediate translational control of downstream genes.⁷⁹ Plant uORFs are classified along evolutionary lines, and for a fairly small fraction, the peptide sequence was conserved during evolution, the so-called CPuORFs. Most genes with CPuORFs have regulatory functions, and transcription factors are over-represented among these genes.^48,49 We found that 2 CPuORF-containing bZIP transcription factors from rice, OsbZIP38 and OsbZIP27, are under miRNA regulation (targeted by miR5819 and miR5075, respectively). Of note, the target site of miR5819 locates at the nucleotide sequence encoding the sPEP, as revealed by degradome analysis and further validated by 5′RACE. We present evidence for regulation of the CPuORF7-containing SAM decarboxylase gene by miR2101. From these findings, we propose miRNA-guided regulation of CPuORF-containing genes during elicitor treatment. Thus, this regulatory mechanism represents an additional layer of control in the refined regulatory system based on functioning of CPuORF-encoded short peptides.

Certain CPuORF-encoded sPEPs are activated by metabolic signals (i.e., sucrose, polyamines), which offers a path for metabolic control of gene expression.⁴⁹ In particular, sucrose can repress translation of AtbZIP11 and AtbZIP2 (members of the S1-group of bZIP transcription factors).⁸⁰ Indeed, members of the group S of bZIP transcription factors from different species contain a sucrose-regulated CPuORF in their 5′ UT region, also known as Sucrose Control-uORFs (SC-uORFs). The SC-uORFs can be found in long (e.g.,, AtbZIP11 and AtbZIP2) and short versions (e.g., AtbZIP3), the later representing the more conserved C-terminal of SC-uORFs that are known to mediate sucrose-induced repression of translation of the bZIP gene (Fig. 7A). Mutations in the long AtbZIP11 that creates a shorter conserved uORF (C-terminal conserved amino acids) still allow for sucrose-induced translational control, indicating that the C-terminal amino acids are essential for sucrose regulation.⁸⁰ Of interest, the 2 CPuORF-containing transcription factors we identified that are under miRNA regulation (OsbZIP38 and OsbZIP27) contain the short version of the SC-uORF (Fig. 7A). This observation suggests sucrose-regulated translational control in the expression of the 2 rice bZIP transcription factors. Sucrose is a signal molecule for activation of plant defense responses.^81-83 A priori, the interaction of the miRNA regulation of CPuORF-containing genes and translational control by sucrose would allow these regulatory genes to respond in a flexible way to rapidly changing stimuli that affect sucrose levels in plants, including pathogen infection. Thus, we propose a model for miR5819 function in regulating OsbZIP38 expression (Fig. 7B). This regulatory network integrates miRNA function and metabolic regulation of gene expression, thereby representing a novel regulatory network potentially involved in plant immune responses. Overall, we provide important clues to further understand the miRNA-mediated and metabolic regulation of CPuORF-containing genes in plants.

Figure 7.

miR5819-mediated regulation of OsbZIP38 in rice in response to treatment with fungal elicitors. (A) Alignment of the conserved Sucrose Control-uORF (SC-uORF) amino acid sequences present in the 5′ UT region of the group S bZIP-type transcription factors in Arabidopsis (AtbZIP11 and AtbZIP2), and rice miRNA-regulated bZIP transcription factors identified in this study (OsbZIP38, OsbZIP27). Dark and light gray indicate different amino acids. (B) A model depicting the regulation of OsbZIP38 expression by miR5819 and CPuORFs. Treatment with fungal elicitors regulates miR5819 accumulation, which in turn negatively regulates the accumulation of OsbZIP38 transcripts. The target site of miR5819 locates at the nucleotide sequence encoding the short peptide (sPEP, encoded by CPuORF3). Sucrose can modulate translation of AtbZIP11 and AtbZIP2 (members of the S1-group of bZIP transcription factors) via SC-uORF (Wiese et al. 2004). Whether OsbZIP38 is translationally controlled by sucrose remains to be determined.

To conclude, our results support that miRNAs and their corresponding target genes can be considered an integral part of the rice response to M. oryzae elicitors. Disease-resistant rice plants can be obtained by altering the expression of the miRNA or its target gene. Since pathogen attack is one of the primary causes of crop losses worldwide, unravelling the miRNA-mediated mechanisms underlying pathogen resistance of plants has profound significance.

Materials and Methods

Plant material and elicitor treatment

Rice (Oryza sativa japonica cv. Nipponbare) plants were grown at 28° ± 2°C under 16-h/8-h light/dark cycles. Elicitors from the M. oryzae strain 18.1 were prepared as previously described⁸⁴ and used at a final concentration of 300 µg/mL. In all experiments, mock treatments were performed. The plant material was harvested at 30 min and 2 h of elicitor treatment. Three biological replicates were analyzed. Each sample represented a pool of approximately 150 rice plants.

Construction of sRNA and degradome libraries and sequencing

Total RNA was extracted from rice tissues with use of TRI Reagent solution (Ambion, Austin, TX, USA). In all, 24 sRNA libraries were prepared by using the TruSeq™ Small RNA kit (Illumina Inc., CA, USA) from leaves and roots that had been treated or not with fungal elicitors (2 times of elicitor treatment each tissue, 30 min and 2 h; 3 biological replicates per sample).

Four degradome libraries were prepared from control and elicitor-treated rice leaves (2 times of elicitor treatment, 30 min and 2 h) as previously described.⁴² Small RNA and degradome libraries were individually sequenced on an Illumina Genome Analyzer (HiSeq2000). All the small RNA and degradome sequence data have been deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc = GSE66611). Bioinformatic analysis of sequencing data is detailed in Supplemental data. Gene ontology (GO) analysis and GO enrichment of target genes identified in the degradome libraries involved use of MapMan (http://mapman.gabipd.org/web/guest/mapman).

Identification of novel miRNAs

miRDeep-P with default parameters was used to identify novel miRNAs from rice as described.³⁹ For this, all the sRNA sequences were mapped to the rice genome (O. sativa, version 7.0; http://rice.plantbiology.msu.edu/). Next, genome sequences spanning the putative miRNA, 500 nt upstream and downstream sequences, were extracted and used for fold-back secondary structures by use of RNAfold with default parameters (Vienna package 2.1.0; http://rna.tbi.univie.ac.at/cgi-bin/RNAfold.cgi). Sequences that met the criteria for recognition of candidate miRNA precursors described by Meyers et al.⁴⁰ were considered.

Expression analyses

For Northern blot analysis of rice miRNAs, total RNAs were fractionated in a 17.5% denaturing polyacrylamide gel containing 8 M Urea, transferred to nylon membranes (Hybond-N, GE Healthcare, United Kingdom), and probed with [γ³²P]ATP-labeled oligonucleotides (Table S6). Hybridization signals were detected by use of Phosphorimager (BioRad, CA, USA). Synthetic RNA oligonucleotides were loaded as size markers.

Stem-loop RT-qPCR was used for miRNA expression analysis as described. Further experimental details can be found in Supplemental data.

Modified 5′-RNA ligase-mediated RACE for mapping mRNA cleavage sites

RNA ligase-mediated 5′ RACE was performed as described in Llave et al. 2011⁸⁵ with specific primers listed in Supplementary Table 6 and the First-Choice RLM kit (Ambion, Austin, TX, USA). Amplification products were cloned, and at least 15 independent clones were sequenced.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

We thank Dr. M-Y Liu from the VYM Genome Research Center, National Yang-Ming University, Taipei, Taiwan, for assistance in degradome data generation.

Funding

This work was supported by the Spanish Ministry of Economy and Competitiveness (MINECO) and the European Regional Development's funds (FEDER) [BIO2012-32838 to BSS] and the CSIC/NSC (Spanish National Research Council/National Science Council of Taiwan)-Cooperative Research Project-Formosa Program [2009TW0041 to BSS and Y-ICH]. We also thank the Generalitat de Catalunya (Xarxa de Refèrencia en Biotecnologia) for substantial support. P. Baldrich was a recipient of a Ph.D grant from the “Ministerio de Ciencia e Innovación, Formación de Personal Investigador, BES-2010-032879).

Authors' Contributions

PB carried out most of the experimental work and data analyses. SC participated in small RNA library preparation. BSS coordinated the design and execution of this study and wrote the manuscript. M-TW, T-TL and Y-ICH participated in degradome analysis. Y-ICH also participated in the design of this study and critically revised the manuscript.

Supplemental Material

Supplemental data for this article can be accessed on the publisher's website.