miR482 Regulation of NBS-LRR Defense Genes during Fungal Pathogen Infection in Cotton

In this study, we characterized the miR482 family in cotton using existing small RNA datasets and the recently released draft genome sequence of Gossypium raimondii, a diploid cotton species whose progenitor is the putative contributor of the Dt (representing the D genome of tetraploid) genome of the cultivated tetraploid cotton species G. hirsutum and G. barbadense. Of the three ghr-miR482 members reported in G. hirsutum, ghr-miR482a has no homolog in G. raimondii, ghr-miR482b and ghr-miR482c each has a single homolog in G. raimondii. Gra-miR482d has five homologous loci (gra-miR482d, f-i) in G. raimondii and also exists in G. hirsutum (ghr-miR482d). A variant, miR482.2 that is a homolog of miR2118 in other species, is produced from several GHR-MIR482 loci in G. hirsutum. Approximately 12% of the G. raimondii NBS-LRR genes were predicted targets of various members of the gra-miR482 family. Based on the rationale that the regulatory relationship between miR482 and NBS-LRR genes will be conserved in G. raimondii and G. hirsutum, we investigated this relationship using G. hirsutum miR482 and G. raimondii NBS-LRR genes, which are not currently available in G. hirsutum. Ghr-miR482/miR482.2-mediated cleavage was confirmed for three of the four NBS-LRR genes analysed. As in tomato, miR482-mediated cleavage of NBS-LRR genes triggered production of phased secondary small RNAs in cotton. In seedlings of the susceptible cultivar Sicot71 (G. hirsutum) infected with the fungal pathogen Verticillium dahliae, the expression levels of ghr-miR482b/miR482b.2, ghr-miR482c and ghr-miR482d.2 were down-regulated, and several NBS-LRR targets of ghr-miR482c and ghr-miR482d were up-regulated. These results imply that, like tomato plants infected with viruses or bacteria, cotton plants are able to induce expression of NBS-LRR defence genes by suppression of the miRNA-mediated gene silencing pathway upon fungal pathogen attack.

Introduction microRNAs (miRNAs) are 20-24 nucleotides (nt) long small noncoding RNAs and are processed from MIRNA genes that are transcribed by RNA polymerase II. In plants, the primary transcript (pri-miRNA) of a MIRNA gene is processed by the RNase III-like enzyme DICER-LIKE1 (DCL1) into a hairpin structure, miRNA precursor or pre-miRNA, which is further cleaved by DCL1 to produce a miRNA/miRNA* duplex from the stem region of the hairpin. The miRNA/miRNA* duplex is assembled into a RNA induced silencing complex (RISC). Once incorporated into the RISC, the mature miRNA binds to complementary sites in its target genes through base pairing and causes degradation and/or translational repression of the target mRNAs, which depends on the level of complementarity between miRNA and its targets [1]. miRNAs are posttranscriptional repressors of gene activity; however, depending on the role of their target genes, they can be positive or negative regulators of a certain biological process. Since the discovery of miRNAs in plants a decade ago, miRNAs have been demonstrated to play a critical role in many aspects of plant development, biological processes and stress responses [2,3,4,5,6].
Pathogen infection triggers transcriptional reprogramming in host plants. Plants have evolved multiple layers of mechanisms to protect themselves from pathogen attacks, including non-host specific resistance, PAMP (pathogen-associated molecular pattern)-trigged immunity (PTI) and effector-triggered immunity (ETI) [7]. A number of miRNAs in Arabidopsis have been shown to either positively or negatively regulate elicitor flg22-induced callose deposition, a signature of the PTI defense response [8]. miR393 has been implicated in bacterial PTI through repressing auxin signaling [9]. Arabidopsis plants elicited by flg22 showed induction of MIR393 transcription and down-regulation of miR393 targets, including three F-box auxin receptors TIR1 (Transport Inhibitor Response 1), AFB2 (Auxin signaling F-Box proteins 2) and AFB3, and consequently increased resistance to Pseudomonas syringae [9]. More recently, miRNAs have been shown to be directly involved in regulation of disease resistance (R) genes [10,11,12]. The N gene from Nicotiana benthamiana [13], which encodes a TIR (the Toll and Interleukin-1 Receptor) type of nucleotide binding site (NBS)-leucine-rich-repeat (LRR) receptor protein that confers resistance to tobacco mosaic virus (TMV), was found to be cleaved by nta-miR6019 and nta-miR6020 [11]. Transient expression of N-targeted miRNAs in N. benthamiana attenuates N-mediated resistance to TMV [11], indicating that nta-miR6019 and nta-miR6020 play an important role in regulating disease resistance in N. benthamiana. Further bioinformatic mining identified several miRNAs, including miR482 and miR2118, which target members of different R-gene families in tomato, potato, soybean and Medicago truncatula [10,11]. miR482 and miR2118 are partially overlapping and both were predicted to target the conserved sequences encoding the P-loop motif of the NBS-LRR receptors [10,11,12], so would be expected to suppress the expression of large numbers of NBS-LRR defense genes. Using a transient assay system, it has been shown that N. benthamiana mRNAs encoding NBS-LRR proteins could be silenced by tomato miR482 [12].
miR482, miR2118 and nta-miR6019 belong to a specific type of miRNA that is 22-nt long and generated from pre-miRNAs containing asymmetric bulges in the miRNA/miRNA* duplex. This type of miRNA has been demonstrated to be the trigger for production of phased 21-nt secondary small RNAs from their target transcripts through the RDR6/DCL4 pathway [14,15]. Therefore, miR482-, miR2118-and nta-miR6019-mediated cleavage of target disease resistance genes is expected to cause not only decay of their target mRNAs but also production of phased secondary small RNAs from the targeted R-genes. This has been verified in tomato and M. truncatula [10,11,12]. Furthermore, at least one of the secondary small RNAs generated from a miR482 targeted NBS-LRR gene has been shown to target mRNA encoding another defense-related protein [12], indicating that the secondary small RNAs generated from NBS-LRR genes can possess the characteristic of trans-acting siRNAs (ta-siRNAs) [16,17]. Plants infected with viruses or bacteria showed a reduced level of miR482 and an increased level of mRNAs of miR482 targets, suggesting that the miR482-mediated silencing cascade is suppressed by pathogen attack and may be a defense response of plants [12].
Cotton is the most important textile fiber crop in the world. The cotton genus (Gossypium spp.) consists of 50 species, including 45 diploids and five allotetraploids. Two A-genome species (G. arboreum and G. herbaceum) and two AD-genome species (G. hirsutum and G. barbadense) were independently domesticated and cultivated for their fibers [18]. All the allotetraploids originated from relatively recent interspecific hybridization events between an Agenome-like ancestral African species similar to modern G. arboreum or G. herbaceum and a D-genome-like Central American species similar to modern G. raimondii [19]. Because of the importance of fiber, miRNA identification in cotton has mainly focused on fiber tissues, particularly tissues related to fiber initiation [20,21,22], although other tissues such as embryogenic callus have been used [23]. In the miRBase v20 (updated in June 2013), only 86 cotton miRNAs have been annotated (80 in G. hirsutum, one each in G. arboreum and G. herbaceum, and four in G. raimondii). This paucity is largely due to the lack of cotton genome sequences and use of only a narrow range of tissues in miRNA identification. This has changed with the release of the draft genome assembly of G. raimondii, a D-genome diploid cotton species that is the closest living relative of the ancestral D t -genome in tetraploid cotton [24,25], from which 28 conserved and 181 non-conserved miRNA families were predicted [24]. Another study investigated cotton miRNAs responsive to Verticillium dahliae infection [26]. V. dahliae is a fungal pathogen causing the soil-borne vascular disease verticillium wilt (VW), one of the major cotton diseases worldwide. The small RNA populations in V. dahliaeinoculated cotton (G. hirsutum and G. barbadense) roots were sequenced and compared with those from mock-treated roots. This investigation identified a number of conserved miRNAs, including miR2118, and 14 new cotton miRNAs, which were upor down-regulated upon V. dahliae infection [26].
In this study, we characterized the cotton miR482 family using published small RNA datasets [10,26] and the genome sequence of G. raimondii [24], identified G. raimondii NBS-LRR genes potentially targeted by miR482/miR482.2, and analyzed V. dahliae-induced expression changes of miR482/miR482.2 and their predicted NBS-LRR targets in G. hirsutum. We found that about 12% of the G. raimondii NBS-LRR genes are potential targets of miR482/miR482.2, and that miR482-mediated cleavage of NBS-LRR genes triggers production of phased secondary small RNAs. Four members of the ghr-miR482 family and several NBS-LRR genes were down-regulated and up-regulated in cotton seedlings infected with V. dahliae, respectively, implying that miR482-mediated silencing of NBS-LRR genes is released in cotton upon fungal pathogen infection to activate disease defense.

The Cotton miR482 Family
The first member of the cotton miR482 family was identified in G. hirsutum [20]. Three more members of this family, ghr-miR482a and ghr-miR482b in G. hirsutum and gra-miR482 in G. raimondii, were later reported [21]. By searching the published cotton small RNA datasets, we found that gra-miR482 also exists in G. hirsutum; therefore, there are at least four miR482 members in the G. hirsutum genome. In order to characterize the cotton miR482 family and keep the original nomenclature as much as possible, we renamed the first miR482 identified by Kwak et al. [20] as ghr-miR482c and the G. hirsutum counterpart of gra-miR482 as ghr-miR482d ( Figure 1). Ghr-miR472 reported by Kwak et al. [20] is in fact ghr-miR482a. The miR482 reported by Wang et al. [22] is a variant of ghr-miR482b because its 1 st -20 th nucleotides are identical to the 3 rd -22 nd nucleotides of ghr-miR482b ( Figure 1). We designated it ghr-mi482b.2 (similar to miR2118 in other plant species). Using ghr-miR482b.2 as a query and searching against G. hirsutum small RNA datasets, we found two isoforms of ghr-miR482b.2. One is identical to ghr-miR482d except for the last two nucleotides; another has no corresponding ghr-miR482 sequence found in G. hirsutum. We named these two as ghr-miR482d.2 and ghr-miR482e.2, respectively ( Figure 1). We found no similar variants for ghr-miR482a and ghr-miR482c in G. hirsutum.
To determine whether these G. hirsutum miRNAs exist in the G. raimondii genome, and how many loci are able to generate these miRNAs in the G. raimondii genome, we searched the G. raimondii genome for sequences that exactly match with ghr-miR482 and predicted hairpin structure using the genomic sequence surrounding the miRNA. We found that the G. raimondii genome contains a single locus each for gra-miR482b and gra-miR482c. The sequence identical to ghr-miR482d is detected at five loci that are able to form a stem-loop structure in the G. raimondii genome. They were named as gra-miR482d, f, g, h, and i ( Figure 2A; Figure S1). No homolog of ghr-miR482a was found in the G. raimondii genome. Correspondingly, gra-miR482b.2, gra-miR482d.2, gra-miR482f.2, gra-miR482g.2, gra-miR482h.2, and gra-miR482i.2 as well as gra-miR482e.2 were found in the G. raimondii genome. When allowing one mismatch, we found five more loci each containing a gra-miR482d isoform and predicted to form a hairpin structure in the G. raimondii genome ( Figure S2); however, these new gra-miR482d isoforms were not found in any of the published cotton (G. hirsutum, G. barbadense and G. arboreum) small RNA datasets so are not further analysed.
Gra-miR482h/miR482h.2 and Gra-miR482e.2 are Tandem miRNAs Both gra-miR482h/miR482h.2 and gra-miR482e.2 are located on scaffold 9 of G. raimondii, and are just 164 bp apart from each other. The G. raimondii genomic sequence containing these two miRNAs is predicted to form two hairpins. Gra-miR482h/ miR482h.2 and gra-miR482e.2 are located within each of these two hairpins ( Figure 2A). Because these two miRNAs are closely located, they could be generated from the same primary transcript. A gene model (Gorai.009G338400, no potential function annotated) is predicted in the region containing gra-miR482h/ miR482h.2 and gra-miR482e.2 in the G. raimondii genome although it has no homologous gene in other plants so is presumably non-coding. By searching the cotton gene index (DFCI) database (http://compbio.dfci.harvard.edu/cgi-bin/tgi/ gimain.pl?gudb = cotton), we found a tentative consensus sequence TC257237 (1446 bp in length), which was assembled from a few G. raimondii expressed sequence tags (ESTs) and is fully reverse complementary to the G. raimondii genomic sequence containing gra-miR482h/miR482h.2 and gra-miR482e.2. This finding suggests that TC257237 is likely to be the primary transcript for gra-miR482h/miR482h.2 and gra-miR482e.2 and that its orientation is probably mis-annotated in the DFCI database ( Figure S3).
By mapping G. hirsutum small RNAs to the G. raimondii sequence containing gra-miR482h/miR482h.2 and gra-miR482e.2, we found that the hairpin harbouring gra-miR482e.2 contains two phased miRNA/miRNA* duplexes, one is gra-miR482e.2/ miR482e.2* and another is gra-miRCan1/miRCan1* (the small RNA with more reads was designated gra-miRCan1; Figure 2A). This second miRNA appears to be unique to the D and D t genomes as when we cloned the same region from diploid and tetraploid cotton species the miRCan1/miRCan1* duplex was only found in the D (G. raimondii) and D t (G. hirsutum) genomes ( Figure S4). In comparison to the D and D t genome sequences, the A genome (G. arboreum) sequence contains three single nucleotide polymorphisms (SNPs) in miRCan1, which would abolish the base pairing between miRCan1 and miRCan1* and change the hairpin structure of the A genome sequence ( Figures 2B, S4). Two potential targets, including a gene encoding a leucine-rich repeat kinase, were predicted for gra-miRCan1 (Table S1).

NBS-LRR Defense Genes are Targets of miR482 in Cotton
In tomato and M. truncatula, miR482 and miR2118 have been shown to regulate numerous NBS-LRR defense genes through targeting the conserved sequences encoding the P-loop motif of the NBS-LRR receptors [10,12]. In cotton, only one EST (NP673173) encoding an NBS domain protein has been predicted to be a target of ghr-miR482a/miR482b [21]. This is because no NBS-LRR receptor encoding genes have been reported and only a very limited number of disease resistance related ESTs are available in cotton, although partial genomic sequences for a number of NBS-LRR gene analogues have been amplified in both diploid and allotetraploid cottons based on the sequence information of the conserved motifs in other plant species [27,28,29].
Release of the G. raimondii genome sequence provided us the opportunity to estimate the number of NBS-LRR genes in cotton and how many are targets of miR482. According to the results reported by Paterson et al. [24], the G. raimondii genome contains 300 NBS-LRR genes. We found that 36 (12%) of these 300 NBS-LRR genes were potentially targeted by various numbers of the gra-miR482 (or ghr-miR482) family members. Of these 36 candidate NBS-LRR targets, 16 were uniquely targeted by only one member of the gra-miR482 family and 20 were targeted by at least two members (Table S1). Of the members of the gra-miR482 family, gra-miR482b/miR482b.2 had fewer NBS-LRR targets than other members (Table S1). Four predicted targets of ghr-miR482/ miR482.2 were selected for cleavage analysis using the approach of rapid amplification of cDNA ends or 59 RACE. Gorai.007G320500 and Gorai.009G033000 were confirmed to be targets of ghr-miR482 whereas Gorai.007G319800 was found to be targeted by ghr-miR482d.2. For Gorai.002G044500 the cleavage sites were not found at the expected positions but at their nearby positions ( Figure 3).
The P-loop sequence of the NBS-LRR genes targeted by members of the ghr-miR482 family is shown in Figure 4A. It is clear that the last three amino acids GKT that complement the first seven nucleotides of ghr-miR482 were completely conserved in all predicted NBS-LRR targets. According to the consensus nucleotide sequences of the top three P-loops that were found in the majority of NBS-LRR genes targeted by ghr-miR482, a variable nucleotide was always observed at the 3 rd position of a codon ( Figure 4B). These characteristics of the target sites suggest that gain and loss of regulation of NBS-LRR genes by ghr-miR482 is possible without changes to the protein sequence. In addition, several genes with a diverse function, including a gene encoding a MYB-domain containing protein, were also predicted to be targeted by member(s) of the miR482 family (Table S1).

Generation of phased Secondary Small RNAs in NBS-LRR Genes
To further confirm the authenticity of the predicted targets and to know whether phased secondary small RNAs were generated from the cleaved targets, we mapped four publically available cotton small RNA datasets (GSM699074-GSM699077 from V. dahliae-infected and mock-treated roots of G. hirsutum and G. barbadense) onto the G. raimondii genome and analysed the distribution of small RNAs across all the predicted targets of ghr-miR482/miR482.2 using the method previously described [17,30]. Of the 36 candidate NBS-LRR targets, the majority had small RNAs mapped to, and nine showed a phased distribution of those small RNAs (phase score .2) spreading downstream from the primary target site of ghr-miR482/miR482.2 ( Figures 5, S5). Distribution of small RNAs from the first four phases after the ghr-miR482d cleavage site in Gorai.011G075600 is shown in Figure 5. In this case, the 5' end of the first phased 21-nt small RNA was aligned with the expected cleavage site (the 10 th nucleotide counting from the 5' end) of ghr-miR482d rather than that of ghr-miR482d.2, suggesting that production of the phased small RNAs was triggered by ghr-miR482d-mediated cleavage. 21-nt small RNAs corresponding to the first phased register of ghr-miR482 were also observed in Gorai.007G320300 and Gorai.007G320500 although no phase signal was bioinformatically detected for these two genes probably because the number of phased small RNAs were below the threshold. For Gorai.007G320500, ghr-miR482mediated cleavage was also confirmed by cleavage analysis (Figure 3). For Gorai.007G320500 and Gorai.009G033000, apart from the ghr-miR482-mediated cleavage, additional cleavage sites located downstream and in phase with the cleavage sites of ghr-miR482d were found (Figure 3), further suggesting that these genes behave like TAS loci in Arabidopsis. To know whether, like in tomato plants infected with viruses or bacteria, the expression levels of miR482 were down-regulated in cotton plants infected with fungal pathogen V. dahliae by root dipping (Figure S6), we first performed small RNA northern blots using RNA isolated from the susceptible Sicot71 cultivar (G. hirsutum). Expression of ghr-miR482a/miR482e.2 was detected in leaves collected at 1 dpi from plants infected with V. dahliae by root dipping, but no V. dahliae-induced down-regulation of ghr-miR482a/miR482e.2 was observed ( Figure 6A). We were unable to detect expression of these two miRNAs in roots. We then performed the more sensitive miRNA stem-loop qRT-PCR to analyze the expression levels and changes of individual members of the ghr-miR482 family in response to V. dahliae infection. In leaves and roots, ghr-miR482b was the most abundantly expressed while ghr-miR482a was the least expressed under mock infection ( Figures 6B-6H). Upon V. dahliae infection, a decrease in expression was observed for ghr-miR482b/miR482b.2 and ghr-miR482c in both leaves and roots, as well as for ghr-miR482d.2 in roots ( Figures 6C, 6D, 6F, 6G), whereas the expression levels of ghr-miR482a, ghr-miR482d and ghr-miR482e.2 were unchanged ( Figures 6B, 6E, 6H). To determine whether this down-regulation of ghr-miR482/miR482.2 caused up-regulation of their possible NBS-LRR targets, expression changes of 11 NBS-LRR genes that were predicted targets of ghr-miR482/miR482.2 were analysed. Ten of these 11 NBS-LRR genes were found to be induced upon V. dahliae infection in either leaves or roots, or both (Figure 7). Significant induction in both leaves and roots was observed for Gorai.002G044900, Gorai.007G320500 and Gorai.011G075600 (Figures 7B, 7E, 7J), and significant induction in leaves or roots sequence difference between the A t and D t genome at the primer annealing regions. miRCandidate1 (miRCan1) and miR482e.2 are highlighted in blue and red, respectively. Three SNPs that abolished the production of miRCan1 in the A genome are boxed. doi:10.1371/journal.pone.0084390.g002 Figure 3. Ghr-miR482/miR482.2-mediated cleavage of NBS-LRR genes. Cleavage analysis was performed using 59 RACE. Two bands were amplified and sequenced for Gorai.007G320500 whereas a single band was sequenced for other three genes. Gorai.007G320500 is also a predicted target of ghr-miR482c. The sequences of ghr-miR482d and ghr-miR482d.2 are underlined and showed in italic, respectively. The expected cleavage sites of ghr-miR482 and ghr-miR482. These results suggest that the NBS-LRR target genes annotated in G. raimondii are conserved in G. hirsutum, and that an induced expression of NBS-LRR target genes in V. dahlia-infected G. hirsutum was a result of repression of ghr-miR482/miR482.2 biogenesis. Gorai.008G112600, a predicted target of ghr-miR482a-c (Table  S1), remained unchanged ( Figure 7G). One possibility is that the miR482-target sites of Gorai.008G112600 and its G. hirsutum homolog are different and the latter is no longer a target of ghr-miR482a-c in G. hirsutum. Another possibility is that Gorai.008G112600 is a target of ghr-miR482a, which was lowly expressed and not V. dahlia-infection responsive, rather than ghr-miR482b/c because it is a better target of ghr-miR482a based on prediction (Table S1). In addition, the possibility of translational repression also could not be ruled out.

Discussion
miR482 is a diversified miRNA family whose abundance varies significantly among different plant species. It is highly expressed in Solanaceae species, but rarely detected in monocotyledonous species [12]. Arabidopsis has no miR482 annotated, but its miR472 is closely related to miR482 found in other species. Interestingly, some MIR482 loci are able to produce a miR482.2 variant, whose 1 st -20 th nucleotides are overlapping with the 3 rd -22 nd nucleotides of miR482. This variant was designated miR2118 in some plant species [10]. In rice, miR2118 is highly expressed in developing inflorescence and triggers production of phased siRNAs [31,32] but no miR482 has been reported. According to the published cotton small RNA data [10,20,21,22,26] we analysed, the miR482 family of G. hirsutum has at least four members, and two members (ghr-miR482b and ghr-miR482d) have a corresponding ghr-miR482.2. In addition, ghr-miR482e.2 but not its corresponding ghr-miR482e is found in the G. hirsutum small RNA datasets we analysed. By blasting the G. raimondii genome sequence, we found that ghr-miR482a has no homolog and thus could be a newly evolved ghr-miR482 member in G. hirsutum. Although ghr-miR482b and ghr-miR482c each have a single match in the G. raimondii genome, ghr-miR482d has five matches (gra-miR482d, fi), suggesting that ghr-miR482d could be generated from multiple loci in the D t genome of G. hirsutum. All seven members of the ghr-miR482 family (Figure 1) are found in G. arboreum [10], suggesting that they may also be expressed from the A t genome of G. hirsutum. In addition, we identified at least five precursors of new isoforms of ghr-miR482d in the G. raimondii genome, but these isoforms were not found in the G. hirsutum small RNA datasets we analysed. It could be that these ghr-miR482d isoforms are expressed at very low levels or in tissues that have not been used in generating the small RNA data. Another possibility is that they have been lost in G. hirsutum after polyploidization.
miRNAs have been shown to play an important role in response to biotic and abiotic stresses in plants [3,5,6]. Several miRNAs have been demonstrated to be regulators of the pathways related to plant immunity. For example, miR393 was induced by flg22 to restrict Pseudomonas syringae growth by repressing auxin signalling [9]. A Brassica-specific miRNA, bra-miR1885, was specifically induced by infection with TuMV (turnip mosaic virus), but not with CMV (cucumber mosaic virus), TMV (tobacco mosaic virus) or Sclerotinia sclerotinorium, a necrotrophic fungal pathogen. Bra-miR1885 has been shown to cleave a TIR-NBS-LRR (TNL) type disease resistance gene [33], although the function for the pathogen induced expression of bra-miR1885 and cleavage of disease resistance gene is yet to be investigated.
Expression of miR482 was found to be suppressed in tomato plants infected with viruses or bacteria, and consequently some of its disease resistant NBS-LRR target genes were induced two to three-folds [12]. Based on this observation and on the experiment, in which N. benthamiana mRNAs encoding NBS-LRR proteins were found to be silenced by tomato miR482, the authors have proposed that plants are able to exploit virus-and bacteriumderived suppressors of RNA silencing to induce expression of defense-related genes to achieve non-race-specific resistance against viral and bacterial pathogens [12]. Disease resistance proteins may have a cost to plants [34] because if unregulated they can trigger autoimmunity in the absence of pathogen infection and inhibit plant growth [11]. Plants have thus evolved the miR482-NBS-LRR regulatory loop as a counter mechanism to minimize the cost of over-expression of NBS-LRR genes in the absence of a pathogen, and to ensure rapid induction of disease resistance proteins upon pathogen attack.
In this study, we asked the question whether the mechanism uncovered in tomato for viral and bacterial pathogens is conserved for fungal pathogen in cotton. We confirmed some of the predicted targets were cleaved by ghr-miR482/miR482.2 ( Figure 3) and found that miR482-mediated cleavage of NBS-LRR genes triggers production of phased siRNAs (Figures 5, S5). Furthermore we found that V. dahliae infection resulted in down-regulation of some members of the ghr-miR482 family and up-regulation of their NBS-LRR targets (Figures 6-7). Individual members of the miR482 family behaved differently in response to V. dahliae infection. Three members remained unchanged and the highest expressed and significantly down-regulated member, ghr-miR482b, seems to have no contribution towards the induction of the NBS-LRR genes examined. The exact reason for this still needs further investigation, but ghr-miR482b has the least number of predicted NBS-LRR targets and has no unique target. Of the 36 predicted NBS-LRR targets, only two are targets of ghr-miR482b and both are better targeted by other members of the ghr-miR482 family (Table S1).
The miR482-mediated regulation of NBS-LRR genes seems to be finely modulated in a few different ways. The target site (P-loop) of miR482 is one of the conserved motifs in the NBS-LRR proteins. The miR482 family is thus expected to regulate the expression level of a number of NBS-LRR genes. We found that ,12% of NBS-LRR genes in the G. raimondii genome are potential targets of the miR482 family, and that 10 of the 11 analysed could be induced upon V. dahliae infection. On the other hand, the number of NBS-LRR genes regulated by miR482 in each species could be evolutionally determined by the balance between minimizing the disadvantageous effect of over-expression of NBS-LRR genes in the absence of a pathogen and maximizing the induction of NBS-LRR genes in the presence of a pathogen. This is supported by the observation that both the miR482 sequences and the predominant amino acid sequences of the Ploops of NBS-LRR proteins in cotton are different from those in tomato, soybean and M. truncatula [10,12], a result of co-evolution of miR482 and their potential NBS-LRR targets in each species. RNAs generated from the region immediately after the P-loop, which was cleaved by ghr-miR482d. C) Genomic sequence of the P-loop and the first four 21-nt phases (P1-P4). All small RNAs generated from the region corresponding to phases P1 to P4 are shown with the small RNAs in precise phased registers shown in bold. The number after each small RNA represents the counts of the corresponding small RNA in the small RNA datasets used in this study. The sequences underneath the pink and blue lines are ghr-miR482d and ghr-miR2118d, respectively. The expected cleavage sites of ghr-miR482d and ghr-miR2118d are indicated by blue and pink arrow, respectively. doi:10.1371/journal.pone.0084390.g005 miR482 Regulation of NBS-LRR Genes in Cotton Gain and loss of regulation of NBS-LRR genes by miR482 could be an on-going evolutionary event because for each type of P-loop the nucleotide variation was always found at the synonymous 3 rd position of a codon ( Figure 4). Furthermore, diverse and finely controlled biogenesis of miR482/miR482.2 seems to play a role in the modulation of miR482-mediated regulation of NBS-LRR genes. Some MIR482 loci generate both miR482 and its variant miR482.2, and some generate only either miR482 or miR482.2. In addition, gra-miR482h/miR482h.2 and gra-miR482e.2 on scaffold 9 are contained in a single transcript (Figure 2A).
When we were performing northern blot analysis, we assumed that the overall expression level of all members of the ghr-miR482 family could be detected by using one representative member of ghr-miR482 and ghr-miR482.2 (ghr-miR482a and ghr-miR482e.2 used), respectively, as probe. In view of the expression level of ghr-miR482b and its significant reduction in the V. dahliaeinfected leaves ( Figure 6C), this assumption seems to be incorrect. There are three mismatches between ghr-miR482a and ghr-miR482b/miR482d, although only one mismatch between ghr-miR482a and ghr-miR482c. Four mismatches are present between ghr-miR482e.2 and ghr-miR482b.2, and six between ghr-miR482e.2 and ghr-miR482d.2 ( Figure 1). Therefore, the signal detected by northern blot might be only for ghr-miR482a and ghr-miR482e.2, a result consistent with that of miRNA stemloop qRT-PCR ( Figures 6B, 6H).
In conclusion, on the basis of characterization of the miR482 family and identification of NBS-LRR genes targeted by ghr-miR482/miR482.2, we demonstrated that V. dahliae infection represses certain members of the miR482 family and induces expression of specific disease resistance NBS-LRR genes in cotton. This suggests that the miR482-NBS-LRR regulatory loop is part of the immune responses induced not only by viral and bacterial pathogens but also by fungal pathogens.

Plant Materials and Verticillium Dahliae Inoculation
G. hirsutum (varieties Sicot71 and Texas Marker-1, or TM-1) and G. arboreum (variety Yunnanzhongmian, or YZ) plants were raised from seeds and grown in glasshouse at 2862uC with 16 hrs day and 8 hrs night regime. One-true-leaf whole seedlings (including roots) of TM-1 and YZ were used in DNA isolation. The same stage seedlings of Sicot71 were used in V. dahliae inoculation by root dipping. This was done by submerging the roots of cotton plants into a suspension of V. dahliae conidia for 5 minutes. The inoculated plants were then planted in soil in 8-cm pots. The V. dahliae inoculum was prepared by growing V. dahliae in potato dextrose broth (PDB, 1/2 strength) on a shaker (25uC) for 9 days. The conidial suspension was diluted to ,10 8 spores/ml with full strength PDB before inoculation. Leaf and root samples were separately collected from V. dahliae infected and mock-treated (dipping in water) plants at one day-post-inoculation (dpi).
Identification of MIR482 in the G. raimondii Genome G. raimondii short sequences with 0-1 mismatch in comparison to their closest member of the miR482 family of G. hirsutum were first identified using each ghr-miR482 or ghr-miR482.2 as a query. Two-hundred base pairs of G. raimondii sequence flanking each identified short sequence were then retrieved and subjected to hairpin structure prediction using RNA-fold. A MIRNA-like hairpin contains a ghr-miR482, ghr-miR482.2 or their isoform was considered as MIR482 or MIR482.2 of G. raimondii.
Identification of Gra-miR482 Targets in the G. raimondii Genome and Confirmation of ghr-miR482/miR482.

2mediated Cleavage of NBS-LRR Genes
Putative targets of gra-miR482/miR482.2 were first predicted based on the annotated G. raimondii transcripts [24] using psRNATarget (http://plantgrn.noble.org/psRNATarget/) [35], and were then manually checked and selected based on the following criteria: no mismatch (except G::U pair) at positions 10 and 11 (relative to the 5' end of miRNA); no two consecutive mismatches but one mismatch flanked by a G::U pair allowed; with a score #3.5 (based on 1 for mismatch and G::U pair at position 10 or 11; 0.5 for G::U pair at positions other than positions 10 and 11).
Confirmation of ghr-miR482/miR482.2-mediated cleavage of NBS-LRR genes was carried out using total RNA by 59 RACE (rapid amplification of cDNA ends) as previously described [36]. Briefly, ,2 mg of RQ1 DNase (Promega) treated total RNA isolated from Sicot71 was ligated with 100 pmol of the 59 adaptor (5' AACAGACGCGUGGUUACAGUCUUG 3') using T4 RNA ligase (NEB) in the presence of 1 mM of ATP and 1 ml of RNaseOUT (Invitrogen). Two rounds (primary and nested) of PCR were performed using gene specific reverse primers and a universal forward primer (GUS_RACEf; Table S2) that annealing to the 59 adaptor. Bands with an expected size were gel purified using the MinEluteH Gel Extraction Kit (Qiagen) and inserted into pCRH4-TOPOH vector (Invitrogen). At least four clones were sequenced for each ligation using the T7 primer.

Northern Blot and qRT-PCR
Approximately 25 mg of total RNA isolated using the hot borate approach [37] was used in small RNA northern blot analysis as previously described [36]. The antisense sequences of ghr-miR482a and ghr-miR482e.2 were used as probes. Because of similar sequences between miR482 and miR482.2, we reasoned that the expression of miR482 and miR482.2 could not be unambiguously detected by the miR482 and the miR482.2 antisense probe, respectively; therefore, both probes were used together. miR482 stem-loop qRT-PCR was performed according to the approach reported previously [38]. The expression levels of NBS-LRR genes in V. dahliae infected and mock-treated cotton leaves and roots were analysed as previously described [39]. Briefly, reverse transcription was performed using 1 mg of RQ1 DNase (Promega) treated total RNA, random primer or stem-loop RT primer, and SuperScript III reverse transcriptase (Invitrogen). The first-strand cDNA reaction was diluted 10 folds prior to qPCR and 4.6 ml of the diluted cDNA was then used as the PCR template. Reverse transcriptase negative controls were performed for each reverse transcription (RT) reaction to make sure there is no genomic DNA contamination. Each miRNA or gene was analysed using three biological replicates each with three technical replicates. Relative expression levels of the target genes were calculated using 2 2DCt . Significance was analysed by t-test. Cotton Histone 3 (Accession no. AF024716) was used as the reference for normalization [21]. All qRT-PCR reactions were run on the ABI PRISM TM 7900HT Fast Real-Time PCR System (ABI) using SYRBH GreenER TM qPCR SuperMix (Invitrogen). Oligos used in northern blot and qRT-PCR analyses are shown in Table S2.