One-year-old mulberry cutting seedlings of Husang 32 (M. multicaulis) were grown in a growth chamber at 26°C with 12 h of light and 90% humidity. Arabidopsis thaliana (Col-0) plants were incubated under standard greenhouse conditions (22°C, humidity 90% and 12 h of light).

The sixth leaves of five plants were sampled and mixed together, and the total RNA from the leaves sampled for transcriptome was isolated using TRIzol^® reagent (Invitrogen, Carlsbad, CA, United States). The cDNA library was constructed and sequenced according the method described by Xie et al. (2013). The sequences obtained by high-throughput sequencing were chosen to predict the mul-miR3954 targets using psRNATarget¹. The predicted miRNA targets were validated using the degradome sequencing data of the leaves of Husang 32 with the software package, CleaveLand 3.0, as previously described (Addo-Quaye et al., 2008). The Coding Potential Calculator (CPC²), CNCl and Pfam were used to predict transcripts with coding potential, and the transcripts with CPC score<-1 and CNCl score <0 were discarded. The remaining transcripts with length >200 nt and open reading frames (ORFs) <80 bp were considered as lncRNAs. The secondary siRNAs were identified based on the sequences from small RNA library sequencing data of the leaves of Husang 32 according to the method described before (Gai et al., 2014).

The TRIzol^® reagent (Invitrogen, Carlsbad, CA, United States) was used to extract total RNAs of the leaf samples, and the RNAs extracted were digested with DNase I before being used. qRT-PCR was performed on the Rotor-Gene 3000A instrument (Corbett Research, Sydney, Australia) with the SYBR Premix Ex TaqTM kit (Takara, Shanghai, China) for mRNA and the PrimeScriptTM miRNA qPCR Starter Kit Ver.2.0 (Takara, Shanghai, China) for miRNAs according to the manufacturer’s protocol. The actin and U6 genes were chosen as reference genes for normalization of mRNA and miRNA, respectively. The comparative cycle threshold (Ct) method (Livak and Schmittgen, 2001) was used to revaluate the relative gene expression levels. All samples were assayed in triplicate. The primes used for qRT-PCR are given in the Supplementary Table S1.

Total RNAs extracted were firstly separated on (1.2% w/v) formaldehyde denatured agarose gel and then were transferred onto nylon Hybond N membrane. The blots were hybridized with the probes labeled by digoxigenin (DIG) using the PCR DIG Probe Synthesis Kit (Roche, Germany). Prehybridization, hybridization with probe, washing, and detection steps were performed according to the protocol described by Liu et al. (2008). For the mul-miR3954, si165369, si47311, si161579 and si24218 northern blot analyses, sRNAs were extracted with miRCURY^TM RNA Isolation Kit (Takara, Shanghai, China). Specific digoxigenin-labeled RNA probes complementary to the mul-miR3954 and each siRNA were synthesized by Roche Ltd (Roche, Shanghai, China). Prehybridization, hybridization, washing, and detection were performed as previously described method (Yang et al., 2008). The probes used are given in the Supplementary Table S2.

The prepared proteins were mixed with 5 × SDS–PAGE sample buffer and heated for 3 min at 95°C and then were subjected to SDS/PAGE on 12% (w/v) polyacrylamide gels. After electrophoresis, proteins were transferred electrophoretically onto nitrocellulose membranes for protein gel blot analysis. Polyclonal anti-MuCML27 antibodies were generated by immunizing rabbits with purified MuCML27 protein, and the protein was detected by western blots using horseradish peroxidase (HRP)-conjugated polyclonal anti-MuCML27. Western blot analysis was conducted according to the procedure described by Poppenberger et al. (2011).

To study the function of si161579, the vector OE-si161579 over-expressing si161579 was constructed as described by Wang et al. (2010) with modifications. Firstly, a genomic fragment of the ath-miR165 backbone of 176 base pairs was amplified using primers miR165F and miR165R and cloned into pBI121. The mature ath-miR165-producing DNA region was substituted by si161579 using overlapping PCR with the combination of primers si161579 I, si161579 II, si161579 III, si161579 IV, miR165F and miR165R as described before (Schwab et al., 2006). The sequences of the primers used for the vector construction are listed in Supplementary Table S3.

The mature mul-miR3954-producing DNA region was cloned from mulberry DNA, and the RNA isolated from leaves of mulberry was used to synthesize cDNA which was used as template to clone the MuCML27 and MuLnc1 genes, respectively. Primers (Supplementary Table S3) used to amplify the genes were designed based on the gene sequences (Supplementary Table S4). The genes cloned were then ligated individually into the binary vector pBI121 under the control of the CaMV 35S promoter. Then the vector was introduced into Agrobacterium tumefaciens strain GV3101 and transformed into WT Arabidopsis plants by floral dipping. Transformants were selected and confirmed by qRT-PCR and northern blot. The T3 generations from independent ectopic expression transgenic lines were randomly chosen for further functional studies.

The 4-week-old Arabidopsis plants were used for resistance analysis. Inoculation with Pst DC3000 was conducted by spraying the bacterial suspensions (10⁵ CFU mL^-1) of Pseudomonas syringae pv tomato DC3000 (Pst DC3000) onto the rosette leaves. Bacterial growth within the leaf was monitored by a serial-dilution assay (Katagiri et al., 2002). Detached rosette leaves were inoculated with 2-mm-diameter mycelium plugs taken from the actively growing colonies of Botrytis cinerea grown on PDA medium and placed in a covered Petri dish to retain moisture and incubated at 22°C. Each treatment was conducted independently at least three times.

For drought treatments, Arabidopsis seedlings were grown in soil under regular watering regime for 4 weeks. Then the seedling plants were irrigated with 60 gL^-1 PEG-6000 to simulate drought stress. For salt tolerance assay, 4-week-old Arabidopsis seedlings were irrigated with 1/8 concentrated MS salt solution supplemented with 250 mM NaCl. All the experiments were performed three times and three replicates for each treatment were used.

Fresh leaf material was boiled in 3% sulfosalicylic acid for 10 min with constant shaking. After cooling and filtering, the filtrate was used for proline content assay following the method described by Shan et al. (2007). For malondialdehyde (MDA) concentration measurements, fresh leaf material was homogenized using 5% (w/v) trichloroacetic acid and the homogenate was centrifuged at 12,000 × g for 15 min. The supernatant was mixed with an equal volume of thiobarbituric acid (0.5% in 20% [w/v] trichloroacetic acid) and then the mixture was boiled for 25 min at 100°C, followed by centrifugation for 5 min at 7,500 × g. The supernatant was used for MDA concentration measurement according to the procedure as described previously (Peever and Higgins, 1989). These experiments were repeated at least three times.

Based on the transcriptome information attained, one lncRNA, MuLnc1, was identified. Interestingly, MuLnc1 was predicted to be a target of mul-miR3954 (Figure 1A), and this was further experimentally verified by mRNA degradome sequencing (Figure 1B). To explore whether MuLnc1 functioned as a target transcript of mul-miR3954, the correlation between the expression levels of MuLnc1 and mul-miR3954 in mulberry tissues were examined by qRT-PCR. The results showed that MuLnc1 was expressed at relatively high level in the ripe fruits, but its transcript level was very low in the roots, phloem and leaves (Figure 2A), whereas the expression level of mul-miR3954 was very low in the ripe fruits, but its transcript level was relatively high in the roots, phloem and leaves. This suggested that MuLnc1 and mul-miR3954 expressions were highly tissue-specific, and there was a very good negative correlation between their expression levels (Figure 2B), and this is in accord with the function of miRNA in guiding target mRNA cleavage.

Interestingly, some siRNAs in our small RNA library sequencing data were found to match to the MuLnc1 transcript downstream of mul-miR3954 binding site (Figure 3A), and northern blot results confirmed the existence of these siRNAs (Figure 3B). Thus, MuLnc1 transcript might be cleaved by mul-miR3954 and generate the secondary siRNAs. Since the efficient genetic transformation system has not been established in mulberry trees, and the MuLnc1 and mul-miR3954 do not exist in Arabidopsis, to further explore whether the MuLnc1 is targeted by mul-miR3954, the MuLnc1 and MUL-MIR3954 genes were cloned and introduced into Arabidopsis plants, respectively. Genome PCR analysis results showed that both the MuLnc1 and MUL-MIR3954 genes were integrated into the Arabidopsis genome (Figures 4A,B), and qRT-PCR and northern blot analysis results indicated that the MuLnc1 transcript and mature mul-miR3954 were successfully expressed in the transgenic Arabidopsis (Figures 4C–F). The hybrids between MuLnc1 and MUL-MIR3954 transgenic Arabidopsis were produced and used to analyze the cleavage of MuLnc1 transcript by mul-miR3954. qRT-PCR and northern blot assays showed that there was no difference in the expression levels of mul-miR3954 between the hybrid and MUL-MIR3954 transgenic plants (Figure 5A). However, the expression levels of MuLnc1 were significantly lower in the hybrid plants than those in the MuLnc1 transgenic plants (Figure 5B). This suggested that mul-miR3954 can target and cleave MuLnc1 transcript in the hybrid plants. Moreover, four siRNAs, si47311, si161579 and si24218, generated from MuLnc1 transcript were detected in the hybrid plants, but not detected in the MuLnc1 and MUL-MIR3954 transgenic plants (Figures 5C–E). Therefore, mul-miR3954 can cleave MuLnc1 transcript and trigger the production of secondary siRNAs from MuLnc1 transcript in vivo. Moreover, we found that these siRNAs generated from the downstream of the mul-miR3954 binding sites were in a 21-nt “phased” pattern which is a characteristic feature of ta-siRNA. Thus, these siRNAs were potential ta-siRNAs.

Analysis of the mRNA degradome data indicated that the siRNA, si161579, generated from MuLnc1 transcript can target the mulberry calmodulin-like protein 27 gene (MuCML27) (Supplementary Figure S1). To examine whether the si161579 is functional in vivo, the expression level of MuCML27 in different tissues of mulberry was determined. The results showed that the expression level of MuCML27 was relatively high in the ripe fruits, but it was low in the roots, phloem and leaves (Figure 6). This suggested that there was a negative correlation between MuCML27 and siRNA-triggering mul-miR3954, and the si161579 generated from MuLnc1 transcript is functional and able to down regulate its target. To further verify the cleavage of MuCML27 by si161579, the artificial vector OE-si161579 over-expressing si161579 was constructed first and then was transformed into Arabidopsis thaliana. Analyses of qRT-PCR and northern blot showed that the si161579 can be expressed in the transgenic Arabidopsis thaliana plants (OEs) successfully (Figures 7A,B). Moreover, the MuCML27 gene was also cloned and transformed into Arabidopsis thaliana. qRT-PCR, northern blot and western blot analyses showed that the transgenic plants (OEc) expressed the MuCML27 gene successfully (Figures 7C–E). Then the OEsiR-ARF vector constructed was transformed into the OEc plants, and the transgenic lines (OEsc) obtained were used to explore the si161579–mediated MuCML27 gene silencing. qRT-PCR and northern blot analyses showed that the si161579 was successfully expressed in the OEsc transgenic Arabidopsis plants (Figures 7F–G). Interestingly, the mRNA and protein abundances of MuCML27 in the OEsc plants were significant lower than those in the OEc plants (Figures 7H–J). These results indicated that the si161579 can silence and down-regulated the expression of MuCML27 gene in the OEsc plants, and the silencing may occur at the level of transcription.

Recent evidences suggest that several members of calmodulin-like protein (CMLs) play important roles in both abiotic and biotic responses including plant-microbe interactions (Perochon et al., 2011; Zeng et al., 2015). To find out whether the MuCML27 gene is involved in plant responses to pathogens, the OEc and OEsc plants were inoculated with B. cinerea and Pst DC3000, respectively. Rosette leaves were detached from 4-week-old Arabidopsis plants and inoculated with B. Cinerea, and development of disease symptoms was analyzed 4 days after inoculation (DAI). The results showed that expanding disease yellow lesions were yielded in the leaves of wild type and OEsc plants, and the symptoms in the leaves of wild type plants were the same as those in the OEsc plants. However, the disease spot size in the inoculated leaves of OEc plants was smaller than that in the inoculated leaves of wild type and OEsc plant (Figure 8A). When the plants were inoculated with Pst DC3000, there were severe disease symptoms showing gray-brown lesion with chlorosis in the wild type and OEsc Arabidopsis plants. However, the disease symptoms were not evident in the leaves of OEc Arabidopsis plants (Figures 8B,C), although mild chlorosis or necrosis was occasionally observed. To determine whether the depressive symptom development reflected the restricted growth of Pst DC3000 inside the leaves, the bacterial growth in the inoculated leaves was monitored. Arabidopsis leaf samples were taken at 36 h after inoculation, and the strain Pst DC3000 reaches the concentrations of 10⁵ colony-forming units (CFU) cm^-2 of leaf in the inoculated leaves of wild type Arabidopsis. Growth of the strain in the inoculated leaves of OEsc plants was essentially the same as that in the inoculated leaves of wild type Arabidopsis. Whereas growth of Pst DC3000 strains was extremely inhibited in the inoculated leaves of OEc plants (Figure 8D). Therefore, these results indicated that overexpressing MuCML27 can also enhance plant resistance both to Pst DC3000 and B. cinerea, and si161579 indeed silences the MuCML27 gene and inhibits the functions of MuCML27 in plant response to biotic stress.

To explore whether silencing of MuCML27 mediated by si161579 is associated with plant tolerance to abiotic stress, we compared the drought and salt tolerance of the OEc, OEsc and wild-type Arabidopsis. When the plants were exposed to drought stress, the leaves of wild-type and OEsc plants became severe rolling and wilting at 12 days after withholding water. However, the leaves of OEc plants showed less rolling and wilting symptoms as compared with the wild-type and OEsc ones during the drought stress process (Figure 9). When the plants were exposed to salt stress, the leaves of wild-type and OEsc plants became wilted and curled after treatment with 250 mM NaCl for 2 weeks and the plants had a severe retardation in their growth. Whereas the growth inhibition level in the OEc plants was relatively less (Figure 9). In addition, the proline and malondialdehyde (MDA) contents in the leaves of wild type and OEc, OEsc plants under salt and drought stresses were also measured. The results showed that under salt and drought treatments, the proline level was higher but the MDA content was significantly lower in the OEc plants compared to that in the OEsc and wild type plants (Figure 10). These demonstrated that MuCML27 transgenic plants exhibit increased both drought and salt tolerance, and silencing of MuCML27 guided by si161579 inhibited the functions of MuCML27 in plant response to drought and salt stresses.