The use of deep-sequencing technology was efficient to identify 256 miRNAs of Glycine max. These miRNAs were identified from eight different libraries from precursors with stem-loop secondary structures that also map to the soybean genome (Additional file 1). They were detected from water deficit and rust libraries and were characterized as following: detected for the first time, already detected in some plant species, conserved in soybean, or a variant of a known miRNA (isoform). From these analyses, we found 24 novel families that had not been detected before, six families that had already been detected in Coniferophytes, Embryophytes and Magnoliophytes (dicotyledons and monocotyledons), and 22 conserved soybean families. In terms of conserved soybean miRNAs, we only detected 20 known miRNAs in our sequencing. This small number of known miRNA genes detected in our libraries could be due to the two filters used in our processing. These filters may have missed some known, conserved soybean miRNAs because they discarded reads with low frequency and those with more than five matches in the genome.
We detected 121 miRNAs with additional nucleotides in the 3' or 5' terminus compared to the recorded mature miRNA. These miRNA variants (isomiRNAs) were very common in our population of small, detected RNAs. Out of the isomiRNAs, we observed 21 pairs of sense and antisense miRNAs. The duplex presents the antisense strand paired to the corresponding miRNA with two nucleotides 3' overhangs (Additional file 1). This shows that the sense and antisense miRNAs originated from DCL1 processing and supports their validation as true miRNAs [26, 55, 56].
In addition, we validated the conserved miRNAs in our libraries based on homology to known miRNAs in miRBase. The phylogenetic conservation of miRNA sequences is one rule proposed by Ambros et al.  to characterize miRNAs. In this study, we established new miRNAs in soybean that were already detected in other plants species. However, as opposed to some studies that only blast the candidate to the known miRNA mature sequence, our identifications were determined by precursor sequence folding and verification of the genuine hairpin structures.
The complexity of the plant response to biotic and abiotic stresses involves many genes and biochemical and molecular mechanisms, and adaptation to these stresses is achieved through regulating gene expression at the transcriptional and post-transcriptional levels. With regard to post-transcriptional regulation, miRNAs are associated with water deficit response in others plants, but this was the first time that differential expression of these small RNAs were observed in soybean during water deficit. In order to validate 11 of the novel miRNAs detected in sequencing by the RT-qPCR method, we constructed primers stem-loop and analyzed their expression during abiotic and biotic stresses (Figure 1). We observed that several miRNAs were up-regulated during the water deficit in the sensitive genotype (Figure 1A). However, during the same stress, these miRNAs had a different expression in the tolerant genotype. This distinct miRNAs behavior between the two contrasting genotypes under the same conditions could be involved with the drought-tolerance that is observed in the tolerant genotype. One of these miRNAs with this expression pattern is the new MIR-Seq11. Interestingly, MIR-Seq11 was predicted to target peroxidase protein. As known, stress conditions can produce excess concentrations of reactive oxygen species (ROS), resulting in oxidative damage at the cellular level . The increase of this miRNA in the sensitive genotype, when subjected to water deficit, could be one of the factors associated with vulnerability of these sensitive plants. Whereas in tolerant genotype during the two conditions, the expression levels of MIR-Seq11 are lower than in the sensitive cultivar during stress. This situation could indicate that the unchangeable MIR-Seq11 levels in the tolerant genotype may be related to its drought-tolerance capacity.
Another interesting point is the expression of a novel miRNA MIR-Seq07 that showed increased expression levels during the water deficit stress for both genotypes. This result allows us to associate this miRNA with water deficit stress mechanism independently of the genotype background. Our computational approach showed that one of the loci targeted by MIR-Seq07 corresponds to a fructose-bipfosphato-aldolase enzyme which is a constituent of both the glycolytic/gluconeogenic pathway and the pentose phosphate cycle in plants . Therefore increase and/or activation of aldolase appear to be implicated in the plant growth mainly through promotion of the glycolytic pathway function to synthesize ATP . Since, MIR-Seq07 expression was increased during the stress condition in both genotypes and assuming that it can inhibit or degrade aldolases, it could be associated to metabolism decreasing during water deficit in roots.
Plants possess several adaptive traits to support pathogen attacks. In Glycine max, ASR is responsible for significant losses in soybean growth areas. Nevertheless, no study investigating miRNAs and ASR disease had been preformed to date. To determine if miRNAs act as key factors during rust infection or for resistance maintenance, we performed expression analyses with the same 11 miRNAs during mock and infected conditions in two different genotypes (Figure 1B). In general the miRNAs under the fungus infection were down-regulated in the susceptible genotype (except MIR482bd-3p). For example, MIR-Seq11, MIR-Seq13 and MIR-Seq15 which had predicted peroxidases, oxidoreductases and translational initiation factor respectively as targets proteins, were down regulated when infected with ASR. The peroxidases enzymes help to metabolize H2O2 in higher plants, and these proteins, as also others proteins with oxidoreductase activity, have already been reported to be up-regulated after pathogen infection and especially after ASR , indicating a possible involvement of MIR-Seq11 and MIR-Seq13 with the responses to ASR infection. Considering, that a translational initiator factor was predicted to be targeted by MIR-Seq15, we could speculate about the participation of this miRNA in the protein synthesis machinery.
In the resistant plants, most of the miRNAs analyzed by RT-qPCR (except MIR482bd-3p, MIR-Seq07, MIR-Seq15ab) did not vary across the mock and rust infection. Surprisingly, MIR-Seq07 was the unique miRNA that was down-regulated during the fungi infection for both genotypes analyzed in our study. We already mentioned that the MIR-Seq07 had predicted protein target related to metabolism and thus its possible association with water stress. However MIR-Seq07 also had predicted LRRs (leucine-rich repeats)-domain target which are known to be present in disease resistance proteins [59, 60]. This suggested a good candidate for the investigation of the miRNAs' regulatory mechanisms during ASR stress. Although we investigated the expression patterns of some miRNAs detected in our sequencing and predicted the target genes that it regulates, additional experimental approaches must be addressed to confirm these hypotheses.