Classification of small RNAs
A total of 4,223,792 raw reads was generated by Solexa sequencing from the small RNA library prepared from pooled juvenile and mature plant materials subjected to various types of abiotic stress (Table and Table ). After removing adapters, cleaned reads amounted to 2,378,135 sequences corresponding to 670,645 unique sequences. The sequence lengths ranged from 17 to 32 nucleotides. Although some small RNAs were present about a thousand times in our dataset, most were sequenced only a few times. Small RNAs consisting of only one or two reads accounted for 32% and 53% of clean sequences, respectively.
Statistics of small RNA sequences from the Hevea brasiliensis small RNA library
List of the treatments applied to Hevea plants before small RNA isolation
Annotation of the Hevea small RNAs was attempted by BLASTN on the Arabidopsis genome. Most of the sequences did not map onto the Arabidopsis genome sequence. Of the mapped sequences, 26,721 were not annotated. Annotated sequences predicted 6,618 products of mRNA degradation, 6,800 transposable elements, 399 rRNAs, 138 products of pseudogene degradation, 73 ncRNAs, 39 tRNAs, 5 sn/snoRNAs and 57 miRNAs corresponding to 14 families.
Further mapping was carried out with all the miRNA sequences available in the PMRD database. In all, 18,430 sequences matched (0 or 1 mismatch) to conserved miRNAs in other species. The accessions were classified into 48 miRNA families (Table ). Since most sequences did not map against PMRD, the Hevea EST database was scanned using the LeARN pipeline. This strategy led to the identification of 10 additional miRNA families, based on the stem-loop structure, which were not described in other species (Table ). The distribution of the 670,645 unique accessions showed that the most abundant accessions had in descending order 27, 19, 24, 26 and 17 nucleotides (Figure ). Accessions with 17 and 19 nucleotides came from products of mRNA degradation and transposable elements according to the mapping results against the Arabidopsis genome sequence (data not shown).
Length distribution of unique accessions in the small RNA dataset in Hevea brasiliensis.
Characteristics of conserved and putative novel miRNA families in Hevea brasiliensis
Reads of conserved and putative novel miRNAs were distributed according to their nucleotide lengths (Figure ). The 48 conserved miRNA families mostly had 24 and 23 nucleotides (Figure ). For the 10 putative novel miRNA families, the most represented lengths were 24, 27 and 26 nucleotides (Figure ).
Length distribution of reads matching miRNA families in Hevea brasiliensis. (A) conserved miRNA (B) putatively novel families.
Of the 48 conserved miRNA families, 12 were predominant with more than 1,000 reads, including 3 families that displayed more than 20,000 reads. The latter corresponded to HbmiR159/319 and HbmiR408 (Figure ). By contrast, 36 families were less represented in the dataset with 18 families having fewer than 100 reads: HbmiR1310, HbmiR162, HbmiR168, HbmiR1863, HbmiR2118, HbmiR2910, HbmiR2914, HbmiR2915, HbmiR2916, HbmiR393, HbmiR394, HbmiR395, HbmiR399, HbmiR444, HbmiR476, HbmiR482, HbmiR828 and HbmiR845 (Figure ).
Figure 3 Abundance of the 48 conserved miRNA families in Hevea brasiliensis with 0 and 1 mismatch on a minimum length of 19 nucleotides. A) The sequencing frequency of a conserved miRNA family in the small RNA library above 1000 reads. B) The sequencing frequency (more ...)
A detailed analysis of length distribution within the Hevea miRNA families revealed some discrepancies (Figure and Figure ). Firstly, the 20-22 nucleotide and 23-27 nucleotide size classes were detected for both conserved and novel miRNA families. Secondly, the most abundant size for miRNAs was 21-24 nucleotides, except for HbmiR169 (27 nucleotides), HbmiR2911 (26 nucleotides), HbmiR482 (25 nucleotides), HbmiR472 (25 nucleotides), HbmiRn8 (25 nucleotides) and HbmiRn9 (27 nucleotides). Thirdly, the putative novel miRNAs occurred at low levels with fewer than 500 reads, and at extremely low levels for HbmiRn1 and HbmiRn2, with only three sequences (Figure ).
Number of reads according to the length distribution of accessions of conserved miRNA families in Hevea brasiliensis.
Number of reads according to the length distribution of accessions of putative novel miRNA families in Hevea brasiliensis.
Detection of single-nucleotide mutations in miRNA sequences
For all the conserved miRNA families, single-nucleotide modifications in the sequences were analysed in detail based on the data presented in [30
]. In Arabidopsis
, seventeen families are concerned by RNA editing events and nine families by 5' deletion and 3' uridylation modifications (called below -1 + UU modification). In Hevea
, of the seventeen families, four families displayed single-nucleotide mutations at the sites of post-transcriptional RNA modifications described in Arabidopsis
, namely AtmiR156, AtmiR159a, AtmiR164a, and AtmiR172a (Table ). By contrast, no -1 + UU modification was observed in Hevea
Frequency of single-nucleotide mutation and of a combination of 5' deletion and 3' uridylation (-1 + UU) for four miRNA families from Hevea
Identification of precursor transcripts for conserved and putative novel miRNA families from the hevea brasiliensis transcriptome
Conserved and putative novel miRNA precursors were sought in the Hevea
clone PB260 transcript sequence databases obtained from various organs using the LeARN pipeline [51
]. Eight conserved miRNA families that mapped against RNA sequences displayed a stem-loop structure (Table and Additional Table 1). A single miRNA precursor transcript was found for the HbmiR156, HbmiR159, HbmiR396, HbmiR476 and HbmiR2910 families. Two miRNA precursor transcripts were identified for the HbmiR166, HbmiR319 and HbmiR408 families. In addition, the LeARN pipeline predicted 10 putative novel miRNAs with their unique precursors (Table and Additional Table 2). Mapping the sequences of the small RNA dataset on the precursor sequences led to the identification of miRNA*for all eight conserved families (HbmiR156, HbmiR159, HbmiR166, HbmiR319, HbmiR396, HbmiR408, HbmiR476 and HbmiR2910) and for the five putatively new families HbmiRn3, HbmiRn4, HbmiRn8, HbmiRn9 and HbmiRn10. Stem-loop reverse transcription polymerase chain reactions (RT-PCR) were successfully performed on seven precursors for the conserved families and on nine for the putatively new families (Figure ).
List of precursors for conserved microRNAs identified in the Hevea clone PB260 transcriptome sequences
List of putatively new miRNA precursors identified in the Hevea clone PB260 transcriptome sequences
Gel electrophoresis of stem-loop RT-PCR. A) Stem-loop RT-PCR from seven precursors of conserved miRNA family. B) Stem-loop RT-PCR from nine precursors of putatively new miRNAs.
Comparison of the gene ontology classification of predicted targets for conserved and putative novel miRNA families
Computational prediction of miRNA targets led to the identification of 1,083 sequences for the 48 conserved families and 705 for the 10 putative novel miRNA families. Their involvement in biological processes and their molecular functions were attributed using Gene Ontology (Figure and ). Most of the GO terms were represented in the same proportions for predicted targets from conserved and putative novel miRNA families. Target genes of the putative novel miRNA families all had GO terms of conserved miRNA families for biological processes and molecular functions. In biological process terms, the putative novel miRNA families had two additional terms (growth, cell wall organization or biogenesis). The number of miRNA target genes was smaller for putative novel than for conserved miRNAs for some GO terms: cellular process (25% as opposed to 27%), metabolic process (22% as opposed to 27%) and biological regulation (10% as opposed to 12%) (Figures and ). This number was larger for putative novel miRNA targets compared to conserved targets for cellular component organization (5% as opposed to 3%), localization (7% as opposed to 4%) and response to stimulus (10% as opposed to 8%) (Figures and ). As regards the GO terms for molecular function, although the distribution for antioxidant activity was the same for the 2 classes of miRNAs (1%), a decrease in the proportion of transcription regulator activity (4% as opposed to 7%) and an increase in structural molecule activity (4% as opposed to 2%) and transporter activity (7% as opposed to 4%) were observed for the putative novel miRNA families (Figures and ).
Pie chart representation of Gene Ontology terms at level 2 for putative targets of conserved (A and B) and novel miRNAs (C and D). A and C) Gene Ontology terms for biological processes, B and D) Gene Ontology terms for molecular functions.
Comparison of predicted target genes for conserved and putative novel miRNA families involved in "response to stimulus", and in "antioxidant" and "transcription regulation" activities
As the management of oxidative stress is crucial for rubber productivity, we set out to more effectively understand the mechanisms of redox homeostasis. We therefore gave priority to predicted miRNA-targeted genes involved in the "response to stimulus" (biological process) and in "antioxidant" and "transcription regulator" activities (molecular process) according to the GO annotations (Tables and ).
List of predicted targets for conserved miRNAs involved in response to stimulus, and in antioxidant and transcription regulator activities according to GO terms
List of predicted targets for putative novel miRNAs involved in response to stimulus, and in antioxidant and transcription regulator activities according to GO terms
Twelve members of the AP2/ERF domain-containing transcription factor family were putatively targeted by HbmiR156, HbmiR159, HbmiR172, HbmiR393, HbmiR395, HbmiR396, and HbmiR408. Two HD-ZIP III domain-containing protein members (HbmiR165/166, HbmiR172), a transcription factor with a bZIP domain (HbmiRn4), three NAC domain-containing proteins (HbmiR164, HbmiR165 and HbmiRn7), three Squamosa promoter binding proteins (HbmiR156/157), four WRKY transcription factors (HbmiR156/157, HbmiR390, HbmiR444), four Auxin Response Factor (ARF); HbmiR160 and HbmiRn5) and five GRAS family transcription factors (HbmiR170, HbmiR171, HbmiR396 and HbmiRn5) were also potentially targeted.
Sequences encoding protein involved in protein degradation, such as cysteine protease (HbmiR396) and ubiquitin protein ligase (HbmiRn6 and HbmiRn7), were identified for genes classified in the "response to stimulus" category.
Three sequences related to ABA signalling were also identified: an ABA-insensitive protein (HbmiR166, and HbmiR2910), and an HVA22-like protein (HbmiR398). In addition, one sequence encoding CTR1, a negative regulator of the ethylene signalling pathway, was targeted by HbmiRn5. MiRNA biogenesis was also highlighted by the presence of a cap binding protein predicted to be targeted by HbmiR167. Interestingly, two enzymes involved in natural rubber biosynthesis, HMG-CoA reductase and 4-hydroxyl-3-methylbut-2-enyl diphosphate reductase, were predicted to be targeted by HbmiR159 and HbmiR444, respectively. For the "antioxidant activity" category, several ROS-scavenging enzymes were identified. The chloroplastic and peroxisomal isoforms of CuZnSOD were putatively targeted by HbmiR398 and HbmiR159, respectively. Another enzyme involved in glutathione biosynthesis, gamma glutamyl cysteine ligase (GCL), was predicted to be targeted by HbmiRn5. Rboh (or NADPH oxidase) is an enzyme involved in ROS production, which is likely to be targeted by two miRNAs (HbmiR2914 and HbmiR476). In addition, targets that are subjected to or are involved in the redox status of proteins, such as genes encoding proteins involved in lignin synthesis, cynnamoyl CoA reductase, lignin forming anionic peroxidase precursor and 4-coumarate:CoA ligase, were also found to be targeted by HbmiR159, HbmiR408 and HbmiR398, respectively.
The cleavage site for three targets has been experimentally validated for a chloroplastic CuZnSOD, a Squamosa promoter binding protein and an ARF (Table ). For these targets, the cleavage site is located at the canonical 10th nucleotide.
List of target genes experimentally cleaved by miRNA