microRNAs Associated with Drought Response in the Bioenergy Crop Sugarcane (Saccharum spp.)

Sugarcane (Saccharum spp.) is one of the most important crops in the world. Drought stress is a major abiotic stress factor that significantly reduces sugarcane yields. However the gene network that mediates plant responses to water stress remains largely unknown in several crop species. Although several microRNAs that mediate post-transcriptional regulation during water stress have been described in other species, the role of the sugarcane microRNAs during drought stress has not been studied. The objective of this work was to identify sugarcane miRNAs that are differentially expressed under drought stress and to correlate this expression with the behavior of two sugarcane cultivars with different drought tolerances. The sugarcane cultivars RB867515 (higher drought tolerance) and RB855536 (lower drought tolerance) were cultivated in a greenhouse for three months and then subjected to drought for 2, 4, 6 or 8 days. By deep sequencing of small RNAs, we were able to identify 18 miRNA families. Among all of the miRNAs thus identified, seven were differentially expressed during drought. Six of these miRNAs were differentially expressed at two days of stress, and five miRNAs were differentially expressed at four days. The expression levels of five miRNAs (ssp-miR164, ssp-miR394, ssp-miR397, ssp-miR399-seq 1 and miR528) were validated by RT-qPCR (quantitative reverse transcriptase PCR). Six precursors and the targets of the differentially expressed miRNA were predicted using an in silico approach and validated by RT-qPCR; many of these targets may play important roles in drought tolerance. These findings constitute a significant increase in the number of identified miRNAs in sugarcane and contribute to the elucidation of the complex regulatory network that is activated by drought stress.

Introduction miRNAs are a class of small, non-coding RNAs, approximately 21 nucleotides in length, that are endogenous to both plants and animals [1,2] and function to regulate gene expression by sequence-specific interaction with target mRNAs [3,4]. In plants, each primary miRNA is transcribed by the RNA polymerase II enzyme [5] and forms an imperfect foldback structure that is processed into a stem-loop precursor (pre-miRNA) by Dicer-like 1 (DCL1), a nuclear RNase III-like enzyme. The pre-miRNA is then cleaved into a double-stranded RNA duplex, called miRNA/ miRNA*, by the same enzyme [3,6,7]. One of the strands, called the mature miRNA, is incorporated into the RNA-induced silencing complex (RISC), whereas the other strand, called miRNA*, is usually degraded. The incorporated mature miRNA guides the RISC to a target mRNA by base pairing, leading to mRNA cleavage or translational repression [3,4,7]. Plants miRNAs were initially described in Arabidopsis thaliana [7] and since then, an increasing number of miRNAs have been identified in plants. miRNAs are known to be involved in developmental and physiological processes such as flowering, leaf and flower differentiation and the auxin response [2,8]. Recently, the expression of a number of miRNAs has been found to be sensitive to abiotic and biotic stresses [9,10,11,12,13,14,15], with evidence that several play roles in drought stress [9,16]. To date, 34 sugarcane miRNAs are reported (http://www.mirbase.org). Considering that drought stress significantly reduces sugarcane yields [17], the identification of sugarcane miRNAs that respond to this stress will help to elucidate the molecular basis of drought stress tolerance in this important bioenergy crop.
Deep sequencing strategies have revolutionized the discovery of small RNAs, constituting the most effective method for plant miRNA detection [18]. This strategy has allowed the discovery of species-specific and miRNAs that are expressed at very low levels [19,20]. Because miRNAs are conserved across species, bioinformatics strategies [21,22] based on expressed sequence tags (ESTs) from sugarcane and genomic survey sequences (GSSs) from sorghum were used in this study and homology and structural similarity. In this work, our goal was to identify and characterize miRNAs that may be regulated by water deficit in sugarcane.

Molecular markers for drought in sugarcane plants
Two sugarcane cultivars differing in their tolerance to drought stress, RB867515 (higher tolerance, HT) and RB855536 (lower tolerance, LT), were grown in the greenhouse for three months. Water was withheld for 2, 4, 6 or 8 days, and stress symptoms, such as leaf rolling and senescence, appeared early on the second day. By the sixth day, almost all of the stressed plants were severely affected (Supporting Information, Figure S1 and Figure S2). Based on these results, we decided to evaluate the plants that were stressed at two and four days.
To characterize the degree of stress at the molecular level, we performed RT-qPCR (quantitative reverse transcriptase PCR) amplification of a sugarcane gene encoding a dehydrin (Fig. 1). The sugarcane assembled sequence (SAS) SCQGLR1085F11.g is induced by drought in sugarcane [23], and several homologues have been reported to be modulated by this stress in other species [24,25,26,27]. This gene was induced after two and four days of stress ( Fig. 1) in both sugarcane varieties. However, the cultivar with lower tolerance displayed greater induction at both experimental time points.

Transcriptome analysis of small RNAs in sugarcane
To identify the miRNAs involved in drought stress in sugarcane, eight small RNA libraries from mature leaves were sequenced using the Solexa technology. These libraries were representative of plants stressed for two and four days and control plants from both sugarcane cultivars. A total of 90 million reads were obtained, ranging from 8 to 15 million reads per library (Table 1), similar to the available sequencing data for citrus, sorghum and maize [19,28,29]. All Solexa reads were aligned against the GenBank (http://www.ncbi.nlm.nih.gov/) and RFam (http://rfam.sanger. ac.uk/) databases to sort the reads into categories ( Table 1). All of the sequences were then regarded as miRNAs for further analysis.
The size distribution of all of the sequences is presented in Fig. 2. The majority of the reads were 21 to 24 nt in length for all libraries, with 21 nt being the most redundant species, followed by 24 nt (Fig. 2). However, sequences of 24 nt were the most represented class of non-redundant species (Fig. 3). In the 21 nt reads, approximately 80% was composed of U or A at the first base position at the 59end in all libraries (Fig. 4), with equal number of each nucleotide.
Precursors were also identified among the sugarcane ESTs (Table 3 and Fig. 5) derived from six genomic loci because six different SASs were found to correspond to six different miRNA genes (Table 3) and their RNA sequences have the intramolecular capacity to fold into hairpin structures (Fig. 5). Five precursor sequences were from the SUCEST (http://sucest-fun.org/) database, whereas only one was found in the SoGI (http:// compbio.dfci.harvard.edu/) database (Table 3). Two distinct SASs were found for the ssp-miR167 precursor. The MFEs (minimum free energy) ranged from 2145 to 2338 kcal/mol, and the G/C content ranged from 42 to 49% (Table 3).

Analysis of sugarcane miRNAs modulated by drought stress
Among all of the miRNAs identified, seven were differentially expressed during drought (Fig. 6), of which six were differentially expressed after two days of stress (Table 4) and five were differentially expressed after four days of stress ( Table 5). The ssp-miR164 miRNA was differentially expressed only at two days of stress. Additionally, three miRNAs (ssp-miR164, ssp-miR397 and ssp-miR528) were up-regulated in the RB867515 (higher tolerance, HT) cultivar, and none were down-regulated by drought. In the RB855536 (lower tolerance, LT) cultivar, four miRNAs (ssp-miR164, ssp-miR394, ssp-miR399-seq 1 and ssp-miR1432) were down-regulated, and only one (ssp-miR397) was up-regulated by drought. Only ssp-miR397 displayed the same pattern in both cultivars, as it was induced after two days of water stress (Table 4). Real-time PCR of a sugarcane gene encoding a dehydrin (SCQGLR1085F11.g). RB867515 (higher drought tolerance) and RB855536 (lower drought tolerance) plants were irrigated (black bars) or subjected to drought stress by withholding irrigation (gray bars) for two (A) and four (B) days. Error bars represent the standard error (n = 4). * p,0.07 and ** p,0.05. Statistics was calculated between irrigated and drought treatments in each cultivar using the permutation mean test. The expression in irrigated RB867515 plants was considered as 1. doi:10.1371/journal.pone.0046703.g001 ssp-miR393 was only differentially expressed at four days of stress. Four miRNAs (ssp-miR394, ssp-miR397, ssp-miR399-seq 1 and ssp-miR528) were down-regulated by drought in the HT cultivar. Two other miRNAs (ssp-miR399-seq 1 and ssp-miR528) were down-regulated, and one (ssp-miR393) was up-regulated in the LT cultivar. Only miRNAs ssp-miR399-seq 1 and ssp-miR528 were down-regulated in both cultivars (Table 5).
Stem-loop RT-qPCR [30,31] was used to validate the expression of all seven differentially expressed miRNAs during drought according to the sequencing data. The expression patterns of five miRNAs (ssp-miR164, ssp-miR394, ssp-miR397, ssp-miR399-seq 1 and ssp-miR528) were validated by this approach. Among the 15 differentially expressed profiles observed with the sequencing approach (Fig. 6), five (33.33%) displayed the same expression profile using RT-qPCR (Fig. 7). The RT-qPCR analysis confirmed two out of eight (25%) profiles after two days of stress and four out of seven (57%) after four days. In the HT cultivar, five out of seven (71.4%) profiles were confirmed, whereas in the LT cultivar, no profile was confirmed. In all of the RT-qPCR profiles, the miRNA levels of the rehydrated plants demonstrated a tendency to return to the control levels ( Fig. 7).  Prediction of miRNA targets The mature sequences of the seven miRNAs modulated by drought were used to search for their targets in sugarcane (Table 6 and Supporting information, Table S1). All of these miRNAs had putative targets in the SUCEST database ( Table 6), several of which encode transcription factors (ssp-miR164, ssp-miR394, ssp-miR528 and ssp-miR1432), growth or development regulators (ssp-miR393), proteins associated with floral development (ssp-miR394) and several phosphatases, kinases, and oxidases (ssp-miR394, ssp-miR397, ssp-miR399-seq 1, ssp-miR528 and ssp-miR1432), among others (Supporting Information, Table S1). The targets had homologous proteins in several plant species, especially maize, rice and Brachypodium, suggesting that the pathways where these targets act may be conserved.

Target validation by RT-qPCR
RT-qPCR was used to validate the expression of seven putative target genes, one for each of the miRNAs modulated by drought (Supporting Information, Figure S3). To correlate the expression profiles of miRNAs and target genes we calculated the expression ratios between drought-stressed and control plants (Table 7). Considering the 15 treatments where miRNA expression were statistically significant (P,0.05, Fig. 6), we found that in seven treatments target genes had the expected profile, i.e., induced expression of miRNA and repressed expression of the target gene and vice-versa (Table 7, underlined ratios). In two out the seven target profiles, RT-qPCR expression was also statistically significant at P,0.05. This data suggest that in several cases sugarcane miRNA might regulate target genes by translational repression and not by mRNA degradation.

The sugarcane microRNAs
Although the miRNAs of several plant species have been recently studied, neither miRNA sequence identification nor an analysis of differential miRNA expression in response to drought stress has been performed in sugarcane. We identified 21 sugarcane miRNAs comprising 18 families using deep sequencing ( Table 2). Because the sugarcane genome has not yet been sequenced, Sorghum bicolor was chosen as a reference organism because of the high level of identity between the genes in both species [32,33,34]. One of the strategies for the characterization of miRNAs is to examine the phylogenetic conservation of their sequences [35], which permitted the identification and classification of the miRNAs from eight sugarcane libraries. The identification of these miRNAs was accomplished by precursor sequence folding into genuine hairpin structures of the S. bicolor sequences where the reads matched, and their classification was based on homology to S. bicolor miRNAs in the miRBase database.
In this study, only six putative precursors were found in the sugarcane EST databases, representing five miRNA families ( Fig. 5 and Table 3). Because such precursors have short lifetimes in   plants due to rapid processing by Dicer-like 1 in the nucleus [3,36], EST databases do not typically have sufficient coverage to facilitate the discovery of a great number of precursors. For example, in Brachypodium distachyon, a model organism for grass species, only 0.05% of the ESTs and 0.012% of the GSSs contain potential miRNAs [37]. The five families found in sugarcane have been previously described, and the secondary structures of the pre-miRNAs have been evaluated using an in silico approach [38]. The majority of the small RNAs in sugarcane are 21 or 24 nt in length, as expected in plants ( Fig. 2) [4]. The 24 nt species are the most abundant in the non-redundant pool of short reads in sugarcane (Fig. 3), and similar distributions have been reported in studies with several other organisms [18,39,40]. This distribution is likely because 24 nt RNAs are siRNAs and unlike miRNAs, a longer double strand of RNA can result in several different small sequences that act in post-transcriptional gene silencing (PTGS). It has been reported that uracil appears to be dominant as the first nucleotide at the 59 ends of mature miRNAs [41]. In sugarcane, the first positions of the mature sequences were equally composed of U and A (Fig. 4). The 59end nucleotide of a small RNA strand determines the identity of at least several of the recruited AGO proteins [42]. Thus, the 59end nucleotide has important implications for miRNA function because AGO1 preferentially associates with 21 nt RNAs with a U at the 59 end, resulting in PTGS activity. In contrast, AGO2 seems to preferentially associate with 21 nt RNAs that have an A at the 59 end, abolishing their silencing activity and rendering the miRNA functionally inert [43].

miRNAs associated with drought response in sugarcane
To identify miRNAs that are differentially expressed during drought, eight libraries from sugarcane leaf were analyzed: four from RB867515, a cultivar with high tolerance to drought, and four from RB855536, which has a lower drought tolerance. A total of seven miRNAs were differentially expressed during drought (Fig. 6). Six of these were differentially expressed at two days (Table 4), and five miRNAs were differentially expressed at four days of stress (Table 5). For five of these seven miRNAs, the differential expression profiles obtained by sequencing displayed the same expression profiles by RT-qPCR (Fig. 7): ssp-miR164, ssp-miR394, ssp-miR397, ssp-miR399-seq 1 and ssp-miR528. The lack of correlation observed for the other two miRNAs is likely due to cross-amplification of miRNA variants or to the presence of very similar miRNAs in the RT-qPCR samples.
The putative targets of the miRNA modulated by drought provide interesting clues to the drought response in sugarcane. For instance, a NAC transcription factor was found among the targets of ssp-miR164 (Table 6), in agreement with previous studies [44,45]. It is known that miR164 regulates the expression of five NAM/ATAF/CUC (NAC) proteins in Arabidopsis. This class of plant-specific transcription factors has important roles in development, growth and stress responses, such as cold, drought and pathogen attach [46,47,48,49,50]. In one study, overexpressing a member of the NAC gene family resulted in enhanced salt tolerance in tobacco [51]. In addition, NAC is responsible for transmitting auxin signals [52]. Another ssp-miR164 target was an MDR-like ABC transporter that also participates in auxin transport [53]. In this study, ssp-miR164 was differentially expressed in response to drought after only two days, suggesting  that this miRNA acts in the early stage of the stress response.
Considering that the targets of this miRNA presumably help sugarcane plants to withstand drought stress, we would expect that ssp-miR164 targets are repressed in both cultivars. This was the case for the LT cultivar ( Fig. 6 and 7), suggesting that both NAC and the MDR-like ABC transporter, and possibly auxin, play roles in the drought response. However, in the HT cultivar, ssp-miR164 was induced by drought. These data indicate that the two analyzed cultivars differ in their activation of molecular mechanisms in response to drought and that the target genes may play a more relevant role in the LT cultivar. It is known that miR393 is commonly up-regulated during drought stress in Arabidopsis, Oryza sativa, Medicago truncatula and Pinguicula vulgaris [54]. In this work, the ssp-miR393 miRNA was also up-regulated under drought stress in the LT cultivar after four days (Fig. 6). The predicted target of ssp-miR393 encodes a putative protein similar to TIR1, an auxin receptor in Arabidopsis thaliana. TIR1 recognized the 3-indol-acetic acid (AIA) and promote the degradation of the Aux/AIA repressor by a protein ubiquitin ligase that binds to a conserved domain of the repressor. This degradation releases the transcription of auxin-regulated genes [55]. Recently, the expression of TIR1 was related with the response to inorganic phosphate (Pi) in roots of Arabidopsis thaliana, [56]. Our data points to an involvement of auxin in sugarcane responses to drought. ssp-miR394 was down-regulated under drought stress in the two sugarcane cultivars (Fig. 6), indicating that this miRNA is important in the drought stress response independently of the plant genotype. The predicted target of ssp-miR394 is the gene encoding a glyceraldehyde-3-phosphate dehydrogenase (GAPDH); see Table 6. GAPDH catalyzes the oxidation of D-glyceraldehyde-3-P (D-G3P) to 3-phosphoglycerate (3-PGA) with the generation of NADPH in the sixth step of glycolysis [57]. Due to the increased need for available ATP and NADH 2 under drought conditions, glycolysis is usually enhanced. This hypothesis is supported by the fact that rice plants exposed to drought display increased levels of OsGAPDH transcripts [58]. In this context, a decrease in the ssp-miR394 level would increase the GAPDH level and consequently the ATP content.
We also found that the target of ssp-miR397 is a gene encoding a laccase (Table 6), in agreement with previous studies [59]. Laccases are glycoproteins with roles in lignin synthesis [60] and iron acquisition [61]. It has been proposed that laccases are involved in cell wall modifications, such as lignification, acting to reduce cell wall extensibility and elongation [62,63]. In maize [64] and tomato [65], increases in laccase transcripts were reported at high concentrations of NaCl. Under drought stress, ssp-miR397 was up-regulated in both cultivars on the second day (Fig. 6), suggesting a reduction in laccase expression levels. However, on the fourth day of drought stress, this miRNA was down-regulated in the HT cultivar (Fig. 6), likely leading to increased laccase levels, which would be reflected as a decrease in cell elongation due to lignin accumulation in the cell wall in response to stress. The complex expression pattern observed between the two cultivars indicates that lignification may act in different ways during drought stress. miR399 has been described as a negative regulator of inorganic phosphate concentration because it targets a pyrophosphatase [66]. In this study, we also found an inorganic pyrophosphatase 2like protein to be a target of ssp-miR399-seq 1 ( Table 6). Overexpression of AVP1 (Arabidopsis vacuolar pyrophosphatase gene) results in increased drought and salt tolerance in transgenic Arabidopsis, tomato, rice and cotton plants [67,68,69,70]. An increased proton pump activity by the vacuolar pyrophosphatase is the molecular explanation for drought resistance in these plants. This activity leads to a lower water potential in the plant vacuole and increases the activity of secondary transporters, preventing ion accumulation in the cytoplasm [70]. The down-regulation of ssp-miR399-seq 1 under drought stress in both sugarcane cultivars likely leads to increased pyrophosphatase levels. These data suggest that pyrophosphatases also play a role in sugarcane responses to drought stress and that this role is at least partially controlled by the microRNAs.
ssp-miR528 targets a gene encoding a UBX domain-containing protein ( Table 6). UBX domains have been identified in several proteins with diverse functions in eukaryotes. Among the 15 UBXcontaining proteins encoded by the Arabidopsis genome [71], PUX1 was shown to play important roles in plant growth and development. Loss-of-function mutant pux1 plants display accelerated growth in various plant organs, including roots and inflorescence shoots [72]. In this work, ssp-miR528 was downregulated under drought stress in both cultivars on the fourth day  ( Fig. 6), suggesting that this miRNA may be involved in the reduction of growth under this stress. The expression levels of ssp-miR397, ssp-miR399-seq 1 and ssp-miR528 were quite variable in the control plants (Fig. 6), and this variability is likely due to other environmental conditions and/or intrinsic changes, such as developmental influences.
A gene encoding a bZIP transcription factor was determined to be an ssp-miR1432 target (Table 6). Interestingly, several bZIPs are known to play important roles in conferring stress tolerance to plants [73] by regulating the expression of genes that are involved in mechanisms that are essential for stress adaptation, such as cytoplasmic ion homeostasis and osmotic adjustment [74]. A triple knockout in the genes encoding the bZIPs AREB1, AREB2, and ABF3 in Arabidopsis displayed increased tolerance to ABA and reduced drought tolerance [75]. ssp-miR1432 was down-regulated by drought at similar levels at all time points in both cultivars (Fig. 6), suggesting that the sugarcane bZIP activates genes that help sugarcane plants to overcome drought stress. However, because the expression patterns observed in both cultivars are similar, this transcription factor alone may not be responsible for the different levels of drought tolerance in RB867515 and RB855536.
It is notable that the expression patterns of the majority of the miRNAs did not display clear correlations with the differences in drought tolerance observed in the two sugarcane cultivars. Nevertheless, it is clear that the microRNAs differ considerably between the two cultivars, suggesting that the miRNAs with differential expression do not fully explain the drought higher tolerance observed in RB867515.
In summary, our results provide insight into the sugarcane microRNAs, highlighting the regulatory network triggered by drought stress in an important bioenergy crop. We found that transcription factors, kinases, phosphatases and chaperones may be targets of the miRNAs modulated by drought in sugarcane. Further work with transgenic sugarcane overexpressing or silencing the miRNAs or their targets will increase our knowledge of the molecular mechanism of drought stress response and tolerance in sugarcane.

Plant samples
Sugarcane cultivars RB867515 (high tolerance to drought) and RB855536 (low tolerance to drought) were obtained from RIDESA (Rede Universitária para o Desenvolvimento do Setor Sucroalcooleiro). Plants were grown in a greenhouse at the Federal University of Alagoas (Alagoas, Brazil, 9u459320S, 36u139090W) under normal irrigation for three months. To develop drought stress, water was withheld from the test plants. The rehydrated plants received normal watering for two days after the period without irrigation. The control plants received normal watering throughout the experiment. Tissue was collected from Leaf +1 (the highest unfolded leaf with a visible dewlap) in quadruplicate after 2, 4, 6 and 8 days from irrigated, non-irrigated (drought-stressed) and rehydrated plants. The samples were quick-frozen and maintained at 280uC until RNA extraction using the miRVana miRNA isolation kit (Life Technologies, USA) according to the manufacturer's protocol. A pool of two replicates was used for Solexa sequencing, whereas four individual plants were used for RT-qPCR validation.  Target Acc: accession number in the SUCEST or SoGI databases; Expectation: value assigned to the alignment of the mature miRNA and the target, ranging from 0 (perfect alignment) to 5; UPE: the energy required to open the secondary structure of the target at the recognition site (less energy means better accessibility to the target); Mature miRNA: miRNA mature size (in nucleotides); Target  Small RNA Sequencing RNAs from 16 to 27 nt long were selected by polyacrylamide gel electrophoresis, ligated with adaptors at both ends and the products used for cDNA synthesis at BGI (Beijing Genomic Institute, Tai Po, Hong Kong). The sequencing was performed using the Solexa platform at BGI.

Bioinformatics analysis
To identify the sequences corresponding to true mature miRNAs, we predicted their precursors (pre-miRNAs). Reads were preprocessed by removing adapters and discarding reads ,18 nt. The miRDeep-P program [21] was used to map the reads to the reference sequences (Sorghum bicolor genomic sequence and sugarcane ESTs). For a given mapped read, a 250 bp window was used from which reference sequences were extracted to predict RNA secondary structure [76]. The miRDeep core algorithm with a plant-specific scoring system based on known characteristics of plant miRNA genes was used to determine the secondary structure of the extracted reference sequences [35]. RNA sequences were classified as miRNA precursor candidates based on the following parameters: the folding of the RNA sequence into a characteristic stem-loop hairpin secondary structure, the position of the mature miRNA in one arm of the hairpin structure (allowing a maximum of six mismatches with the miRNA* sequence in the opposite arm), negative MFE values for the predicted secondary structures, and a G/C content in the 30-70% range [38], using the RNAfold [77]. After the normalization of the number of reads in each library, the expression of each miRNA was calculated using the Audic-Claverie method [78].
The targets of the miRNA were predicted using psRNATarget (http://plantgrn.noble.org/psRNATarget/) by searching for target genes based on both complementarity scoring and secondary structure analyses [79]. The BlastX algorithm [80] was used to find the hits in the Sugarcane Assembled Sequences (SAS) against the NCBI database to identify the coding strands because several SASs corresponded to the minus strand. Only hits complementary to the coding strand of the sugarcane mRNAs were selected.

Stem-loop reverse transcription and RT-qPCR validation
RT-loop primers (loop-RT), sequence-specific forward PCR primers (loop-FW) and universal reverse primers were designed following Chen et al., 2005 [30] for reverse transcription and PCR amplification of sugarcane miRNAs (Table S2). Reverse transcriptase reactions and cDNA amplification were performed as described by Varkonyi-Gasic et al., 2007 [31]. The miRNA and reference gene reverse transcriptase reactions were conducted under the same reaction condition.

RT-qPCR miRNA and target validation
To validate and quantitate the expression levels of the miRNAs in sugarcane leaf tissues, RT-qPCR was performed using SYBR Green PCR Master Mix (Life Technologies, USA) on a 7500 Real-Time PCR System (Life technologies, USA). Each PCR reaction (18 mL) included 2 mL cDNA, 10 mL SYBR Green Master Mix (16), 1 mL sequence-specific forward primer (10 mM), 1 mL universal reverse primer (10 mM) and 4 mL water. The sugarcane polyubiquitin gene [81] was used as a reference ( Table 7). The reactions were performed at 95uC for 10 minutes followed by 40 cycles of 95uC for 15 seconds and 60uC for 1 minute, with a final dissociation curve analysis. All reactions were run with four biological replicates, each in triplicate. The real-time PCR data analysis was performed by considering the reaction efficiencies to calculate the fold changes in miRNA levels using a web-based QPCR system [82].
For target validation, each reverse transcriptase reaction contained 2.5 mg of total DNA-free RNA, 1 mL oligo d(T) 17 VN (50 mM) and 1 mL of dNTPs mix (10 mM). The reaction was incubated for 10 minutes at 65uC and then laid on ice for 2 minutes. Subsequently, 56 First Strand Buffer, DTT, RNAse-Out and Superscript II enzyme (Life technologies, USA) were added. This reaction was incubated in a VerityTM Thermal Cycler (Applied Biosystems, USA) for 30 minutes at 42uC, followed by 15 minutes at 70uC. Forward and reverse specific PCR primers (Supporting Information, Table S2) were designed for amplification of the target genes. The reactions were carried out using the SYBRGreen PCR Master Mix (Applied Biosystems, USA) on 7500 Real Time PCR System (Applied Biosystems, USA) as described before.

Supporting Information
Text S1 Supporting information text. Physiological data. (DOCX)  Figure S1 Quantum potential efficiency (Fv/Fm) in sugarcane plants. The sugarcane cultivars RB855536 (LTlower drought tolerance) and RB867515 (HT-higher drought tolerance) were maintained under irrigation, without irrigation and without irrigation and then rehydrated, as indicated. * indicates differences between irrigated and drought-stressed plants; +indicates differences between irrigated and rehydrated plants.** and ++indicate p,0.01, and * and +indicate p,0.05 using the ttest. ns -not significant.
(PNG) Figure S2 Fluorescence quantum yield (WPSII) in sugarcane plants. The sugarcane cultivars RB855536 (LT -lower drought tolerance) and RB867515 (HT-higher drought tolerance) were maintained under irrigation, without irrigation and without irrigation and then rehydrated, as indicated. * indicates differences between irrigated and drought-stressed plants; +indicates differences between irrigated and rehydrated plants. ** and ++indicate p,0.01, and * and +indicate p,0.05 using the t-test. ns -not significant.
(PNG) Figure S3 Expression profile of the predicted target genes for the seven sugarcane miRNAs modulated by drought. The values are expressed as fold changes relative to the irrigated control for each gene. The bars represent the average of the irrigated plants (control, black bars) and drought-stressed plants (gray bars) for RB867515 (HT) and RB855536 (LT) after two (2) and four (4) days of stress: A) HT2; B) HT4; C) LT2; D) LT4. Error bars represent the standard deviation (n = 3). Statistics was calculated between irrigated and drought treatments using the t-test. * indicates differences between irrigated and droughtstressed plants, with p#0.05. (PNG) Table S1 Target prediction of the miRNAs differentially expressed in drought-stressed sugarcane plants. (All bioinformatics data without selection.) Target Acc: accession number in the SUCEST or SoGI databases; Expectation: value assigned to the alignment of the mature miRNA and the target, ranging from 0 (perfect alignment) to 5; UPE: the energy required to open the secondary structure of the target at the site recognition (less energy means better accessibility to the target); Mature miRNA: miRNA mature size (in nucleotides); Target start: the base position where the annealing with the miRNA starts; Target end: the base position where the annealing with the miRNA ends; Inhibition: the type of regulation by the miRNA; and Target description: description of the target according to a BLAST search in the GenBank database, including the name of the organism presenting the best hit. (PNG) Author Contributions