The uniform seeds of maize inbred line LH196 were selected for use. After 2 h of sunlight insolation, 1 h of water flushing, 8 min of immersion disinfection with 70% of alcohol, and 6 min of soaking disinfection with 10% H₂O₂, the seeds were washed with diluted water three times at 1 min each time. The seeds were transferred to a layer of filter paper in a petri dish and then covered with another wet filter paper sprayed with diluted water. Germination was accelerated under dark with a temperature of 28°C for 2 days in a plant growth chamber. The germinated seeds were then planted in plastic pots containing vermiculite and cultured under conditions of 14 h light/10 h dark with an air temperature of 30°C light/25°C dark in the plant growth chamber. At the stage of third true leaves, healthy and uniformly growing plants were selected and divided into two groups. These groups were transferred to a hydroponic device and then they were individually treated with different nutrient solutions including 0 mM NaCl (CK) and 250 mM NaCl (salt). The other components of the solution were the same as previously described (Xu Z. et al., 2011). The leaves and roots were harvested after 12 h of treatment, immediately frozen in liquid nitrogen and stored at -80°C for high-throughput sequencing and degradome analysis. Other seedlings were treated 10 days in nutrient solutions including 0 mM NaCl as CK or 250 mM NaCl as salt treatment before harvested for further phenotype measurement, with the nutrient solutions being replaced by new ones every day.

After 10 days of treatment, we harvested the shoots and roots separately. The control group and the treatment group each contained nine randomly taken seedlings and the fresh weight and dry weight were recorded. Anthocyanin was extracted as follows: four seedlings of CK and salt treatment were harvested, respectively; the stems were weighed after the leaves and roots were removed; the stems were then cut into short pieces at length of 2–3 mm and they were put into an EP tube; 10 mL of extraction solution (HCl: methyl alcohol = 1.99, v/v) were added, and the lid was closed and kept at 4°C for 2 days; the tube was shaken twice each day. The relative content of the anthocyanin of the stems was calculated [relative contents of anthocyanin = (OD₅₃₀-0.25 × OD₆₅₇) × the volume of extract/the weight of stem] (Zhao, 2013) on an Excel spreadsheet.

The total RNA was extracted from the four types of samples, namely, the leaf of CK (LC), the leaf under salt treatment (LS), the root of CK (RC), the root under salt treatment (RS) via Trizol reagent (Invitrogen, Carlsbad, CA, United States) according to the manufacturer’s protocol. The RNA concentration and purity were measured using a BioPhotometer (Eppendorf, HH, GER), and the RNA quality and integrity were assessed by agarose gel electrophoresis. The RNAs were treated with RNase-free DNase I (TaKaRa, Dalian, China) to remove the genomic DNA.

Four strand-specific RNA libraries with a size of 18–30 nt were prepared as previously described (Sunkar and Zhu, 2004; Hafner et al., 2008), and submitted to the BGI (Beijing Genomics Institute, Shenzhen, China) for sequencing. The sRNA digitalization analysis based on HiSeq high-throughput sequencing with the Illumina HiSeq 2000 sequencing system utilizes the sequencing by synthesis (SBS), which can decrease the loss of nucleotides caused by the secondary structure. It is also noted for its small sample requirement, high throughput, and high accuracy with a simply operated automatic platform. Through such analysis, one can obtain millions of sRNA sequence tags in one process, can comprehensively identify sRNAs of certain species in certain conditions, and predict novel miRNAs and construct a sRNA differential expression profile between samples, which is a powerful tool for sRNA function research. The experimental process of sRNA sequencing is shown in Supplementary Figure S1. The 49 nt sequence tags from HiSeq sequencing will undergo data cleaning analysis first, which includes getting rid of the low-quality tags and 5′ adaptor contaminants from the 50 nt tags to obtain credible clean tags. The length distribution of the clean tags and both the common and specific sequences between samples will be further summarized. A standard analysis will be performed to categorize the clean tags as potentially known miRNAs vs. novel miRNAs, which would be those which cannot be placed into any previously known category. After obtaining this miRNA result, target prediction for miRNAs, GO enrichment, and a KEGG pathway for target genes will be analyzed. The procedure is shown in Supplementary Figure S2. The undesirable reads were excluded from the raw data to generate clean reads. The sRNA tags were mapped to the genome by the Short Oligonucleotide Analysis Package (SOAP) (Li et al., 2008)¹ to analyze their expression and distribution on the maize genome². The sequences that were perfectly matched were identified for the next analysis. Non-coding RNAs (ncRNAs) that were annotated as rRNAs, tRNAs, snRNAs, and snoRNAs were identified by alignment and a BLAST search against the Rfam (10.1) and GenBank databases (Benson et al., 2013).

Approximately 150 mg of total RNA were used for the purification of polyadenylated RNA molecules with oligo(dT) cellulose (Ambion). In addition, degradome libraries were built as previously described (German et al., 2008). By using the Illumina HiSeq 2000 sequencing system, degradome sequencing utilizes a SE50 sequencing strategy and produces 49 nt raw reads. The 3′ adaptor was trimmed before the bioinformatics analysis to obtain real degradome fragments with a length of 20–21 nt (Folkes et al., 2012). All of the data herein have been deposited in the National Center for Biotechnology Information, United States (NCBI). After pre-processing, the clean tags are generated and stored. Clean tags are classified by the alignment to the database and to remove the ncRNAs. Finally, the miRNA–mRNA pairs are identified and mapped to the reference genes. Clean tags are mapped to the genome by SOAP2.20 and then the genome distribution of the tags is analyzed (see Supplementary Figure S3) (German et al., 2008, 2009). Degradome sequencing provided the resource for us to predict the target genes, and the GO analysis of the entire target genes are shown in Supplementary Figure S4.

The expression of miRNAs and their target genes are quantified by an iCycler iQTM real-time PCR detection system (Bio-Rad), following the manufacturer’s instructions. To accurately estimate the regulation by miRNA of its target gene, forward primer and reverse primers located in both sides of the predicted cleavage site were designed with Primer 3 software³. The cDNAs were reverse-transcribed from the prepared RNAs according to PrimeScript^TM RT Master Mix (TaKaRa) and they were then diluted into same concentration. qRT-PCR with three statistical and three biological repeats were run with the Power SYBR Green PCR Master Mix (Applied Biosystems). Maize a-tubulin 5 (Tub5) was selected to be an internal standard. A 2 μL volume of cDNA was used in the 25 μL PCR mixture. The mixture was incubated at 95°C for 3 min, followed by 50 cycles of 95°C for 15 s, and 60°C for 30 s.

Maize seedlings grown under a 250 mM NaCl solution treatment provided a phenotype of salt stress: the leaves turned yellow and wilted, the aerial roots grew to a large quantity, and the branches of the roots increased and became thinner (Figure 1A). All of the parts of the maize inbred line LH196 seedling biomass were significantly reduced, but both the root/shoot ratio of the fresh weight and dry weight increased significantly (Figure 1C). Under the salt treatment, the relative content of anthocyanin in the stem had a significant increase of approximately 1.5 times (Figures 1B,D).

Approximately 100.7 million raw reads of sRNA and degradome sequencing data were generated from the four libraries. The data involving the raw reads, clean reads, and unique clean reads and the matching sequences of the maize genome across the entire sequencing of the samples are listed in Supplementary Table S1. The similarity of the four maize sRNA libraries, leaf control, leaf salt, root control, and root salt (LC, LS, RC, and RS), is appreciable (see Supplementary Table S2). The unique matching sequences were used in further steps. All four sRNA libraries show a similar trend, which indicates that the 24 nt group has the highest abundance of size distribution according to the unique reads (Figure 2). The distribution of sRNAs on every chromosome is quite uniform (Figure 3), with the sRNA reading counts being close to each other. The abundance of extremely conserved miRNAs has been determined, such as miR156, miR159, miR164, miR166, miR167, miR168, and miR398, which was very similar to previous reports (Zhao et al., 2013). For instance, the abundance of zma-miR156 and zma-miR167 was 40560.15 reads per million (RPM = specific miRNA reads^∗10⁶/total reads) and 3714.984 RPM, respectively.

Our sRNA sequencing data covered the known miRNAs very well. Combining two leaf libraries (LC, LS) with the other two root classes (RC, RS), a total of 81 unique mature miRNA sequences belonging to 25 known miRNA families in miRBase (Release 21: June 2014) (Kozomara and Griffiths-Jones, 2014) were sequenced, whereas in all of the libraries, there were 37 of 81 unique mature miRNAs detected.

The identification of new miRNAs supplemented the database of maize miRNA. Many potentially new miRNAs were identified from the four libraries, which have not been registered on the miRBase previously (Release 21). More than 200 novel miRNAs were found in the leaves, and approximately 150 potentially new miRNAs were detected in the roots. Seventy-six of them were identified in both the leaves and the roots (see Supplementary Table S3).

It is well known that miRNAs are highly conserved in plants. According to the alignment of all of the unique clean reads with less than four mismatches from four libraries (LC, LS, RC, and RS) against all of the known plant miRNAs in miRBase (Release 21), a total of 1040 known plant miRNAs were identified from these four maize libraries; 762 and 726 were from the leaves and roots, respectively, and 448 miRNAs were common between the leaves and roots. Among these 1040 miRNAs, 314 and 278 miRNAs were specific to the leaves and roots, respectively. We used both the fold change (absolute value ≥ 1) and P-value (significant ^∗, P-value ≤ 0.05; extremely significant ^∗∗,P-value ≤ 0.01) to define the significance of the expression. A total of 400 (52.49%) and 455 (62.67%) miRNAs showed a significantly different expression in the leaves and roots, respectively, under different treatments (see Supplementary Table S4).

A total of 37 potential new miRNAs were selected based on the criteria of an absolute fold change ≥ 2 between the salt treatment and the control (Zhao et al., 2013) and standardized sequencing reads ≥ 1 RPM (see Supplementary Table S5). Fifteen of the 37 miRNAs showed significant differences between the salt treatment and the control in both leaves and roots; 13 showed significant difference only in the leaves, whereas 9 showed significant difference only in the roots. Among these potential miRNAs, mir-29 and mir-36 were classified as new members of the miR167 family and miR164 family, respectively (Figure 4). Many novel miRNAs identified from two or more sRNA libraries also have a relatively high abundance, and their expression levels changed significantly under the salt treatment. For instance, mir-250 was significantly down-regulated in both the leaves and roots under the salt treatment, whereas mir-330 was up-regulated significantly in the leaves but down-regulated in the roots (see Supplementary Table S5).

Among all of the novel miRNAs, mir-29 and mir-36 were classified as new members of the miR167 family and miR164 family, respectively. Four unique mature miRNA sequences were detected, namely, zma-miR167a, a-5p, zma-miR167e-3p and zma-miR164a-3p. In the miR167 family, mir-29 was the only significantly changed species, and in miR164s, mir-36 was up-regulated significantly under salt treatment in the leaves, whereas others were down-regulated (Table 1). The results suggest that mir-29 and mir-36 may play a primary role among zma-miR167 species and zma-miR164 species in responses to high salinity, respectively. The other six novel species, namely, mir-250, mir-316, mir-17, mir-330, mir-205, and mir-189, caught our attention. They showed a highly significant difference between the control and the salt treatment in the leaves and roots (Table 1). These eight novel miRNAs together with their targets and functions are listed in Table 2.

Eight novel miRNAs were randomly selected for qRT-PCR to test the reliability of sRNA sequencing. Compared with the expression of the control miRNAs, the expression pattern of most miRNAs under the salt treatment by deep sequencing was similar to qRT-PCR in both the leaf and root, respectively (Figure 5). The correlation test showed that the expression of the eight miRNAs exhibited significant positive correlations between qRT-PCR and deep sequencing (R² = 0.6413; and R² = 0.7465) (Figure 5A).

We examined the response of the targets of the eight novel miRNAs (mir-29, mir-36, mir-205, mir-250, mir-316, mir-330, mir-17, and mir-189) to salt treatment in the leaves, roots or both by qRT-PCR with specific primers (see Supplementary Table S6) with Zea mays alpha-tubulin 5 used as the internal standard. The relative expression levels of GRMZM2G012479, GRMZM2G046092, GRMZM2G017290, and GRMZM2G012160 before salt treatment were very low, but the four genes were 5.5-fold, 7.5-fold, 5.3-fold, and 10.5-fold up-regulated after salt treatment in the roots, respectively. Interestingly, GRMZM2G046092 and GRMZM2G012160 were down-regulated in leaves under the salt treatment. However, GRMZM2G130062, the target of novel mir-189, was up-regulated in the leaves but down-regulated more than fivefold in the roots. In addition, some targets were up-regulated or down-regulated both in the leaves and roots (Figure 6). These results suggest that the interaction between miRNAs and their targets may play various roles in different tissues of maize in response to salinity. Thus, in this study, a possible internal regulation pathway was proposed for four species of maize novel miRNAs in response to salt stress (Figure 7).