Overview of Pinellia miRNA microarray

To identify miRNAs in the Pinellia species, a miRNA microarray was used; this microarray contained probes complementary to known mature miRNAs from 43 plant species in the miRBase database [27]. The miRNA chip was hybridized with Cy5 and Cy3 probe pairs of P. pedatisecta and P. ternata, respectively. On the macro level, 101 unique plant miRNAs from 22 miRNA families were commonly identified from P. pedatisecta and P. ternata. Most of the identified miRNA families (miR159, miR156, miR166, miR167, and miR319) were highly conserved with numerous members, whereas several known but non-conserved miRNA families (miR528, miR894, miR1432, miR1450, miR2919, miR2936, and miR3946) had just one member (Figure 1) [28]. In addition, we investigated the homology between the identified miRNAs in this study and known miRNAs in other species. To a large extent, the Pinellia miRNAs were identified based on their high homology with O. sativa miRNAswhich considered a reference species for studying monocotyledons(Figure 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Number of miRNA members in each family which commonly identified from P. pedatisecta and P. ternate by microarray.

https://doi.org/10.1371/journal.pone.0075978.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. The homology of identified miRNAs in other plant species.

Values on Y axis indicate the number of homological miRNA between Pinellia and other plant species.

https://doi.org/10.1371/journal.pone.0075978.g002

Differential Expression of P. pedatisecta miRNAs

Although several dynamic factors influence miRNA function, the miRNA expression patterns were considered the major components that determine miRNA activity [29]. Thus, we compared the accumulation of known miRNA families using microarray signals to identify further the differential miRNA expression between P. pedatisecta and P. ternata. Based on a cut-off value and p < 0.01, 21 miRNAs from 12 families (Table 1) were differentially expressed. Among these, 14 miRNAs (from the miR156, miR166, miR168, miR397, miR528, miR535, and miR1450 families) were more abundant in P. pedatisecta (log₂ fold change greater than 1), whereas the remaining 7 miRNAs (from the miR894, miR165, miR159, miR396, and miR2919 families) showed lower expression (log₂ fold change less than −1). Previous studies reported that miR156, miR159, miR165/166, miR168, miR396, and miR397 participate in developmental regulation, transcription factor activity and stress adaptation by targeting SBP [30], MYB [31], the homeodomain-leucine zipper [32], AGO [33], growth response factor (GRF) [34], and laccase [35].

miRNA validation by qRT-PCR

qPCR was widely considered to be more accurate and quantitative methods for miRNA expression profiling than microarrays [36]. Based on the miRNA microarray results, 7 miRNAs differentially expressed in P. ternata and P. pedatisecta were selected for qRT-PCR analysis to confirm the presence and expression of the miRNAs. As shown in Figure 3, all miRNAs were detected by qRT-PCR and most results of qRT-PCR were consistent with the microarray data. In particular, miR159f, miR396e, and miR894 were more abundant in P. ternata, whereas miR528 and miR397b were more abundant in P. pedatisecta. However, the expression patterns of miR156g and miR535 differed between the microarray and qRT-PCR analyses. This finding may be attributed to the different sensitivity and specificity of the two technologies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Validation and expression analysis of miRNAs by qRT-PCR (P.pedatisecta/P. ternate).

Blue indicates the miRNA relative expression generated from microarray; Red indicates the miRNA relative expression tested by qRT-PCR. The 5.8s rRNA was chosen as the endogenous control. The error bars indicate the standard deviations obtained from three independent experiments of qRT-PCR.

https://doi.org/10.1371/journal.pone.0075978.g003

Prediction of differentially expressed miRNA targets

Plant miRNAs usually regulate gene expression by guiding targeted mRNA cleavage [37]. Thus, the biological functions of miRNAs may be completely understood by discovering their target genes. To date, the Pinellia genome has yet to be sequenced, and no expressed sequence tags are available. According that conserved miRNA may regulated conserved targets [28], We used the Rice MSU RNA database (version 7.0) to computationally predict potential differentially expressed miRNA targets. Three computational target prediction algorithms (WMD3 http://wmd3.weigelworld.org/cgi-bin/webapp.cgi?page=TargetSearch;project=stdwmd, Targetfinder and psRNAtarget http://plantgrn.noble.org/psRNATarget/) were used. The Results revealed that 159 potential genes were targeted by 18 differential expression miRNAs. As shown in Table S1, 11 out of 18 miRNAs targeted transcription factors, thereby suggesting that these miRNAs mediate negative post-transcriptional regulation.

Based on the obtained miRNA–target pairs, the miRNA-mediated regulatory networks were constructed using Cytoscape [38]. According to their miRNA expression patterns, these networks could be classified into two categories (Figure 4). The results support the well-known miRNA-mediated regulation. For example, the miR156 family was more abundant in P. pedatisecta, and different members of this family were found to regulate the SPL genes (LOC_Os06g45310, LOC_Os08g39890, LOC_Os11g30370, and so on) from the SBP-box gene family (Figure 4A and Table S1) which consistent with the previous reports [37,39,40]. The genes LOC_Os05g41166 and LOC_Os01g12700 belonging to MYB family transcription factor were identified as targets of the miR159b, miR159e, and miR159f that influence flowering and male fertility [7]. Other miRNA targets are involved in diverse physiological and metabolic processes including growth regulating factor, laccase, START domain containing protein, growth regulating factor protein, ATPase-like domain containing protein, PP2Ac, and PWWP domain containing protein (Table S1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Differential-expression microRNA-mediated regulatory networks.

(A) Network mediated by high expression microRNAs in P. pedatisecta. (B) Network mediated by low expression microRNAs in P. pedatisecta. The predicted regulatory relationships between miRNAs and targets were denoted by blank lines. The targets of blue nodes were annotated at the bottom and all the networks were constructed using Cytoscape [35].

https://doi.org/10.1371/journal.pone.0075978.g004

Functional analysis of the miRNA targets

To better understand the functional roles of predicted miRNA target genes, we subjected these genes to functional enrichment analysis by using AgriGO, a promising method for uncovering the miRNA-gene regulatory network on the basis of biological process and molecular function [41] (via ‘‘Singular Enrichment Analysis’’; species, Oryza sativa MSU 7.0; reference, Rice MSU 7.0 nonTE transcript). The genes targeted by differentially expressed miRNAs were strongly associated with development (57.97%), transcription factor activity (37.68%), reproduction (33.33%) and cell differentiation (21.74%) (Table S2). Many miRNA families were involved in the same biological process. For example, miR156, miR168 and miR396 participated in “multicellular organismal development” within the “Biological Process” category (Figure 5A). while the target genes of miR156, miR159, and miR165 were enriched with the GO term “transcription factor activity” within the “Molecular Function” category (Figure 5B). In addition, We observed that the enriched GO terms ‘‘reproductive structure development,’’ ‘‘anatomical structure development,’’ ‘‘post-embryonic development,’’ and ‘‘multicellular organismal process” were potentially related to ‘‘organ development’’. It indicated that these differentially expressed miRNAs were associated with organ development in Pinellia.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. GO (Gene Ontology) term enrichment analysis of the target genes of differential expression miRNAs between Pinellia ternata and Pinellia pedatisecta.

(A) Analysis of the targets of the differential expression microRNAs within the ‘“Biological Process”’ category. (B) Analysis of the targets of the differential expression microRNAs within the “Molecular Function” category. This analysis was performed by using the online tool agriGO, selecting the “Oryza sativa MSU 7.0” as a background.

https://doi.org/10.1371/journal.pone.0075978.g005

Identification of miRNA targets by degradome sequencing

Accurate validation of miRNA targets is important to thoroughly reveal the biological roles of miRNAs. The degradome sequencing approach was applied to identify the gene targets of P. pedatisecta miRNAs and elucidate the potential biological functions of its miRNAs [17]. A total of 19 429 349 raw reads, with 93% (18 137 269) of the sequences having the 5′ ends of uncapped, polyadenylated RNAs were generated.

We used the rice genome for the degradome sequencing study. After removing the respective adaptor sequences and pairing these with O. sativa cDNA, 29% (5 328 833) of the distinct degradome reads from 35 430 cDNAs were perfectly matched to the rice genome. Subsequently, the 15 nt upstream and downstream of these mapped cDNAs sequences were extracted to generate the 30 nt target signatures, which were designated as “t-signatures.” These t-signatures were collected to identify and classify the cleaved miRNA targets by the CleaveLand pipeline [42]. The targets of 18 miRNAs from 4 miRNA familieswhich identified by the Pinellia miRNA double-color microarray were detected (Table 2). 7 members of miR156 family and miR157d were identified tocleave the same gene: LOC_Os08g39890, which encodes for SPL14 (SBP-box gene family). Furthermore, The result showed that miR156g could target LOC_Os09g31438 which encodes for SPL17. miR529 has four types of potential gene targets, including glucose-1-phosphate adenylyltransferase, SAUR30 (an auxin-responsive member of the SAUR gene family), RNA recognition motif-containing proteins, and retrotransposon proteins (Figure 6). miR529 apparently performs its additional functions in Pinellia by targeting multiple genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. Identified miRNA targets by degradome sequencing are presented in the form of t-plots.

The T-plots which were referred to as ‘target plots’ by German et al indicated the distribution of the degradome tags along the sequences of targets (top). The alignment of mRNA: miRNA show the matching degree between miRNA and its target sequence (bottom). The two dots indicate matched RNA base pairs; one dot indicates a GU mismatch. The red spot and arrow represents the sliced target site. The y-axis measures the intensity of the cleavage signal, and the x-axis indicates the position of the cleavage signal on a specific transcript. (A) Regulation of SBP-box gene family by miR156 and miR157 family identified by degradome sequencing. (B) Different target genes cleaved by the miR529 family.

https://doi.org/10.1371/journal.pone.0075978.g006

To determine whether SPL14 could be regulated by miR156, we compared the present results with those reported in the literature. The SPL14 target of miR156 was previously validated by experiments of several independent laboratories [43,44,45,46]. Jiao [45] mapped the miR156-directed cleavage sites of SPL14 using 5′ RACE. The sequencing of 14 randomly chosen clones showed that 13 clones had 5′ ends of the cleaved fragments in the central region of the miR156 target site, thereby confirming that SPL14 can be precisely cleaved by miR156. However, miR529 shares a 14 nt to 16 nt sequence that is significantly homologous with miR156. Given the high sequence similarity, miR529 was predicted to cleave the same genes LOC_Os08g39890 and LOC_Os09g31438 at an overlapping site. By considering that miRNA-guided cleavage occurs precisely between the 10th and 11th nucleotide from the 5′ end of miRNA, we analyzed the results of 5′ RACE by Jiao et al. [45]. We speculated that two clones were generated during 5′ RACE by miR529 but not by miR156 (Figure 7). SPL14 in plants may be targeted by the combined activity of both miR156 and miR529. In addition, miR156 predominantly regulates SBP transcription. However, whether miR156a and miR529b target the same gene SPL14 remains to be verified by more methods.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 7. Cleavage analyses with respect to SPL14 targeted by miR529 and miR156.

(1) SPL14 mRNA(LOC_Os08g39890) aligned with miR529 and miR156. (2) miR529 and miR156 sequence alignments. *means homology sequence(3). Cleavage sites of SPL14 transcript by miR156 were determined by Jiao et al using 5’ RACE assay. The number of clones which represent the cleavage frequencies is indicated above the sequence.

https://doi.org/10.1371/journal.pone.0075978.g007

Plant materials and total RNA extraction

P. ternata and P. pedatisecta plants were maintained in a greenhouse at the Zhejiang Sci-Tech University. Leaves with same developmental stages, plant size, and vigor from both P. ternata and P. pedatisecta plants were selected and quickly frozen in liquid nitrogen. Total RNA was extracted with the mirVana™ miRNA Kit (Ambion, Austin, TX, USA), according to the manufacturer’s protocol. After quantification, the integrity of the isolated RNA was validated by separating the major rRNA bands on agarose gels.

miRNA Microarray Design

Based on the known miRNAs in the miRBase database, we designed a miRNA microarray containing 1957 unique plant miRNAs, representing 2521 well-characterized miRNA from 43 plant species (http://microrna.sanger.ac.uk/). Among the 2521 miRNAs, 190 were from A. thaliana, 324 from O. sativa, 198 from P. patens, 196 from soybean, 174 from M. truncatula, 202 from Z. mays, and the remaining 1237 from 37 other plant species. The unique mature miRNAs were spotted onto a µParaflo™ microfluidic microchip using in situ parallel synthesis and RNA hybridization–optimized probes. Each chip contained two replicates of each probe. For microarray quality control, 5S rRNA was designed as the internal positive control, whereas blank and non-homologous nucleic acids served as negative controls. In addition, eight quality control probes were synthesized to perfectly match or to have a single-nucleotide mismatch with eight external spiked-in synthetic RNAs (20-mer).

Microarray hybridization and data analysis

The miRNA microarray experiment was performed according to the protocol provided by LC Sciences. Total RNA (1 µg) was used as the starting material for microarray assay. The small RNAs (<300 nt) were size-fractionated using a YM-100 microcon centrifugal filter (Millipore) from the total RNA extracted from leaf tissues. After the small RNAs were 3′-extended with a poly(A) tail using poly(A) polymerase, whereas an oligonucleotide tag was ligated to the poly(A) tail for later fluorescent dye staining. To balance hybridization melting temperatures, we chemically modified the detection probes. Hybridization was performed in a µParaflo® microfluidic chip station (LC Sciences) in 100 µL 6× SSPE buffer (0.90 mol/L NaCl, 60 mmol/L Na₂HPO₄, and 6 mmol/L ethylenediaminetetraacetic acid; pH 6.8) containing 25% (v/v) formamide; samples were incubated at 34 °C overnight. After hybridization, we washed each microarray in 0.1× SSPE buffer and used the tag-specific Cy3 and Cy5 dyes to label P. ternata and P. pedatisecta.

The hybridization images were digitized using an Axon Gene Pix 4000B Microarray Scanner (Axon/Molecular Device, Sunnyvale, CA, USA) and analyzed using the Array-Pro image analysis software (Media Cybernetics). Raw data were used by first subtracting the background and then normalizing the signals with a locally weighted regression (LOWESS) filter (miRNA microarray data were deposited into the NCBI-GEO with accession no. GSE48322). Two criteria were used to accept a miRNA signal as detectable: (1) a signal intensity higher than three times the background standard deviation, and (2) spot CV < 0.5 (where CV = signal standard deviation/signal intensity). Signals from P. ternata and P. pedatisecta plants were compared using the paired, two-tailed Student’s t-test; only signals with p < 0.05 and differential expression greater than a threefold increase or decrease were considered significant (Cebeci and Budak, 2009).

Computational prediction of miRNA targets

To predict the genes targeted by differentially expressed miRNAs, three computational target prediction algorithms (psRNAtarget http://plantgrn.noble.org/psRNATarget/, Targetfinder and WMD3 http://wmd3.weigelworld.org/cgi-bin/webapp.cgi?page=TargetSearch;project=stdwmd) were used. The miRNA sequences were reverse-complemented and matched against the Rice MSU (version 7.0) mRNA database. Each mismatch is converted to a weighted score according to the criterion in Jones-Rhoades and Bartel [58]. Mismatched pairs or single nucleotide bulges scores 1, and G:U pairs scores 0.5. The mismatched and G:U pair scores were doubled within the core segment to reflect the importance of complementarity to the miRNA 5’ end for target site function [59]. Perfect matches were given a score of “0”, and a cutoff score of “4” was selected to minimize the number of non-authentic targets. The data predicted by these algorithms were combined and the overlaps were calculated(Table S1).

Expression validation of miRNA using qRT-PCR

Quantitative RT-PCR was used to validate the differences in the expression levels of miRNAs in P. pedatisecta and P. ternata. Total RNA was isolated from the leaves of P. pedatisecta and P. ternata and treated with RNase-free DNase I to remove DNA contamination from the RNA. These samples were simultaneously collected with those for degradome sequencing. Based on the microarray results, 7 miRNAs with expression levels showing at least a 1.5-fold change were selected, and primers were designed according to Varkonyi-Gasic et al. [60] (Table S3). Total RNA (250 ng) was reverse-transcribed into the miRNA cDNA by RTase M-MLV (Takara Bio, D2639A) using stem-loop primers by following the manufacturer’s protocol. qRT-PCR with SYBR Green (Invitrogen, 11733-038) was performed using the Applied Biosystems 7900 apparatus with the following cycling parameters: 50 °C for 2 min, 95 °C for 2 min, followed by 40 cycles of denaturing at 95 °C for 15 s and annealing at 60 °C for 30 s, with a plate read between each cycle. A disassociation stage (melting curve analysis) was then performed. The threshold cycle (C_t) was defined as the cycle number at which the fluorescence signal exceeded the fixed threshold. For normalized sample C_t values, the 5.8s rRNA was used (GenBank ID: AF469037.1; forward primer, 5′-GATGAAGAACGTAGCGAAATG-3′; reverse primer, 5′-TCGATGGTTCACGGGATT-3′) was used. The 2^–ΔΔCt method was used to analyze the relative changes in gene expression based on the qRT-PCR experiments.

Degradome library construction and data analysis

To identify the potential target mRNAs, a degradome library using cleaved ends of polyadenylated transcripts was constructed from P. pedatisecta leaves based on the method described previously by German [17]. Briefly, approximately 1 mg of poly(A)-enriched RNA was directly ligated to a chimeric 5′ RNA adapter (5′-GUUCAGAGUUCUACAGUCCGACGAUCCAGCAG-3′) containing a 3′ Ecop 15I (NEB, R0646S) recognition site using T4 RNA ligase (Promega, M1051). Reverse transcription was carried out to generate first-strand cDNA using an oligo(dT) with a 3′-adapter sequence (5′-CGAGCACAGAATTAATACGACT₍₁₈)V-3′). Subsequently, a short PCR reaction was used to amplify the cDNA using the primers 5′-CGAGCACAGAATTAATACGACT-3′ and 5′-GTTCAGAGTTCTACAGTCCGACGATCCAG CAG-3′. After purification and digestion with Ecop 15I, a double-stranded 3′-DNA adapter (top, 5′-p-NNTCGTATGCCGTCTTCTGCTTG-3′; bottom, 5′-CAAGCAGAAGACGGCATACGA-3′) with degenerate nucleates in the overhang region was ligated to the Ecop I digestion products using T4 DNA ligase (NEB, M0202S). These products were recovered by PAGE and PCR amplification (18 cycles of 98 °C for 10 s, 60 °C for 30 s, and 72 °C for 15 s.). The final cDNA library was the purified and sequenced using a Solexa/Illumina genome analyzer (LC Sciences, Hangzhou, China). Raw sequencing reads were obtained using the Illumina Pipeline (version 1.5) software to remove adaptor sequences and low-quality reads. In addition, to easily analyze the miRNA targets and RNA degradation patterns, t-plots were built according to the distribution of signatures (and abundances) along these reads [61]. The identification and classification of categories of the sliced miRNA targets were processed according to the CleaveLand pipeline (Axtell Laboratory, Pennsylvania State University, USA).