Eight sRNA-seq data sets produced from B. cinerea-infected tomato were downloaded from SRA database according to the accession numbers (GSM1101912, GSM1101913, GSM1101914, GSM1101915, GSM1101916, GSM1101917, SRR1482408 and SRR1463412). Among them, GSM1101912, GSM1101914 and GSM1101915 were respectively produced from the B.cinerea-inoculated leaves at 0h, 24 h and 72 h by Weiberg et al. (2013), whereas SRR1482408 and SRR1463412 were produced from the mock- and B.cinerea-inoculated tomato at 7 days (Jin and Wu, 2015). Tomato genome sequences were downloaded from ftp://ftp.solgenomics.net/(version:build_2.40). The datasets of virulence factors in fungal pathogens were downloaded from http://sysbio.unl.edu/DFVF/index.php (Lu et al., 2012).

To identify B. cinerea-inducted sRNA from tomato, two sRNA-seq datasets which were sequenced from B. cinerea-inoculated (TD7d) and mock-inoculated (TC7d) tomato leaves at 7 days post-inoculation (dpi) were downloaded from NCBI SRA database with the accession number SRP043615 (Jin and Wu, 2015). The unique sequences in each library were extracted and combined into 1 sRNA library for sRNA identification; all reads in both libraries that were exactly mapped to tomato genomic sequences but unmapped to tRNA, rRNA and B.cinerea genome sequence were extracted as tomato sRNA. The raw read count of these unique sRNAs were retrieved from each sRNA libraries and normalized to reads per millions (RPM). The sRNA sequences with a minimum RPM of 10 in each library were extracted for different expression analysis. The upregulated sRNAs with the Log2(TD7d/TC7d) of >1 and and p-value <0.001 in B. cinerea-infected tomato were considered as the candidates of B. cinerea-induced sRNAs in tomato. Then these candidates were confirmed by sRNA-seq datasets produced from B.cinerea-inoculated tomato at 24 hpi and 72 hpi comparing with mock-inoculated tomato by Weiberg et al. (2013). These B.cinerea-induced sRNAs were considered as miRNAs through sequences alignment with known miRNAs deposited in miRBase database (Kozomara and Griffiths-Jones, 2011). The remaining sRNAs were considered as unknown siRNA. The target genes of the identified sRNA including miRNAs and siRNAs were predicted by using psRNATarget server against cDNA sequences of B. cinerea (Dai and Zhao, 2011). The virulence factors were analyzed for the putative targets by using blastp with the parameters of E-value <1e-5, identity >40% and coverage >70%.

The seeds of tomatoes cv. Micro Tom and Nicotiana benthamiana were seeded directly into the soil with a 12h:12h photoperiod at ~22°C in a greenhouse. This work used 6-week-old plants. Potato dextrose agar was used for the cultivation of B. cinerea. Conidiospore has collected from 2-week-old infected tomato washed with distilled water two times and adjusted the concentration of 5 × 10⁶ conidiospores/mL for bioassay. Based on the sequence shown in Table S2 , both single strand (ss) sRNAs and double strand (ds) sRNAs were synthesized by GenePharma (Shanghai, China). Each sRNA was added to the conidiospore suspension at a final concentration of 10 µM and immediately drop-inoculated onto five tobacco leaves or tomato leaves for several days. The leaves in control were inoculated with conidiospores in the same manner but with water or negative control RNA (NC RNA, 5′-UUCUCCGAACGUGUCACGUTT-3′. has no target gene in B. cinerea). The experiment was repeated three times. We analyzed the intensity of infection of B. cinerea control and treatment with sRNA by imaging and Trypan blue staining (Woods-Tör et al., 2018).

Conidiospore germination was detected using the cellophane strip method described by Bilir et al. (2019). Briefly, cellophane strips were cut into 1.5 cm2, sterilized by high pressure, and then placed onto MS media in Petri dishes and were added with 10 µL of B. cinerea conidiospore with sRNA or NC-RNA treatment. After incubating for 12 h:12 h photoperiod at 22°C, conidiospores were examined under a light microscope.

Total RNAs were extracted using TRNzol-A+ reagent (TIANGEN, Beijing, China), and a Nano Drop 2000 spectrophotometer was used for RNA quantification. The steps of reverse transcription for the mRNA and sRNA are different. For mRNAs, reverse transcription was performed using the PrimeScript RT Reagent Kit with gDNA Eraser (Takara, Dalian, China) according to the manufacturer’s recommendations. A similar reaction without reverse transcriptase was performed as a control to confirm the absence of genomic DNA in subsequent steps. For miRNAs, we added poly(A) using E. coli poly(A) polymerase (NEB, Beijing, China). 3’ RT-Primer (Invitrogen) was used as the reverse transcription primer for the following reverse transcription according to the manufacturer’s protocol.

SYBR Green PCR was performed in accordance with the manufacturer’s instructions (NEB, Beijing, China). In brief, 1 μL of cDNA template was added to 10 μL of 2X Master Mix (NEB, Beijing, China), and 10 μM of specific primers and ddH₂O were added to a final volume of 20 μL. The reaction was pre-denatured for 3 min at 94 °C, followed by 50 cycles of 94°C for 30 s and 58°C for 30 s. All reactions were performed in triplicate, and controls (no template) were included for each gene. Threshold cycle (Ct) values were automatically determined using the ABI 7500 Real-Time PCR System (USA). The confirmation of amplicon specificity was based on the melt curve at the end of each run. Fold changes were calculated using the 2^-ΔΔCt method, where 2^-ΔΔCt = (Ct, target − Ct, inner) Treatment − (Ct, target – Ct, inner) Control (Livak and Schmittgen, 2001). All oligos used in this study are listed in Supplementary Table S5 .

Pre-miR396a was synthesized and then cloned into the pBIN438 vector with BamH I and Sal I. The cDNA sequences of B. cinerea genes targeted by miR396a were obtained using the RT-PCR method and then inserted into the pEarleyGate 100 expression vector down-stream of the CaMV 35S promoter region. The target sites of miR396a-5p in the target genes were deleted using overlap PCR. Subsequently, the mutated sequences were also inserted into the pEarleyGate 100 vector with expression driven by CaMV 35S promoter.

Transient expression experiments were performed in accordance with the method described by Meng et al. (2020). The Agrobacterium tumefaciens strain GV3101 was transformed with the constructs p35S::MIR396a, p35S::target (normal target site), and p35S::target^mu (target site deletion). Transformed A. tumefaciens samples were harvested and resuspended in infiltration medium [10 mM MgCl2, 10 mM MES (pH 5.6), 200 μM acetosyringone] with the OD600 adjusted to approximately 0.2 for 3-4 h. The A. tumefaciens was infiltrated into 2-week-old N. benthamiana leaves, which were harvested 2 days later.

The statistical analysis was performed with SPSS statistical software 22.0 (United States). All the results were expressed as the Means with SDs from three independent experiments. The t-test was selected and the P-values < 0.05 were considered statistically significant.

Two sRNA-seq datasets were sequenced from B. cinerea-inoculated (TD7d), and mock-inoculated (TC7d) tomato leaves at 7 days post-inoculation (dpi) were downloaded from the NCBI SRA database with the accession number of SRP043615 to identify tomato-derived sRNAs targeting B. cinerea (Jin and Wu, 2015). By using 10 normalized RPM sRNA reads as a cutoff, a total of 1373 sRNAs, ranging from 20 nt to 24 nt, were identified in both libraries ( Table S1 ). Among them, 53 sRNAs were upregulated in B. cinerea-infected leaves compared with mock-infected leaves and then considered as B. cinerea-induced sRNA candidates ( Table S2 ). Moreover, 21 out of the 53 sRNAs were upregulated in B. cinerea-inoculated tomato libraries reported by Weiberg et al. (2013) ( Table S2 , Table 1 ), but the remaining 32 sRNAs could not be confirmed ( Table S2 ). Of these 21 expression-confirmed sRNAs, six had been reported as miRNAs in miRBase, and the remaining 15 were unknown, which were labeled as siR1–siR15 ( Table 1 ). In addition, six B. cinerea-induced sRNAs including 2 miRNAs (miR156d-5p and miR396a-5p) and four siRNAs (siR3, siR10, siR13, and siR14) were confirmed by two or more B. cinerea-treated datasets reported by Weiberg et al. (2013) ( Table 1 ).

We identified potential target genes of these 21 B. cinerea-induced sRNAs in silico and focused on those associated with fungi virulence factors to understand the roles of these sRNAs in cross-kingdom. The results showed that a total of 149 transcripts of B. cinerea were identified as putative targets by psRNAtarget with default parameters (http://plantgrn.noble.org/psRNATarget/) ( Table S3 ). By searching the database of virulence factors in fungal pathogens, we revealed that 82 transcripts targeted by the 21 sRNAs shared high sequence similarity with 63 virulence factors, indicating their potential roles in the pathogenicity of B. cinerea ( Table S4 ). Moreover, 13 transcripts were targeted by nine sRNAs, namely, miR396a-5p, miR482b, siR2, siR3, siR5, siR8, siR9, siR14, and siR15, which were involved in “Grey mould” ( Table 2 ). The transcripts were targeted by the remaining 12 sRNAs, namely, miR157a-5p, miR156d-5p, miR5300, miR396a_3p, siR1, siR4, siR6, siR7, siR10, siR11, siR12, and siR13, which were reported as virulence factors in other fungal pathogens.

To confirm the expression of these identified sRNAs induced by B. cinerea, 6 sRNAs (miR156d-5p, miR396a-5p, siR3, siR10, siR13 and siR14) supported by two or more other B. cinerea-treated datasets were selected to measure their expression patterns in B. cinerea-inoculated tomato leaves at 24 hours post-inoculation (hpi) using quantitative reverse transcription PCR (qRT-PCR). Results showed that all of these sRNAs were upregulated in B. cinerea-infected tomato leaves compare with mock-treated leaves except for miR156d-5p ( Figure 1A ), showing a consistent expression pattern with sRNA-seq datasets (Weiberg et al., 2013; Jia et al., 2015). Of these 5 B. cinerea-induced sRNAs, only 3 sRNAs might target four virulence factor coding genes of grey mould disease, including Bcin15p02590.2 and Bcin04p04920.1 targeted by miR396a-5p, Bcin04p03120.1 targeted by siR3 and Bcin01p06260.1 targeted by miR14 ( Table 2 ). Therefore, expression patterns of these 4 coding genes were also investigated in B. cinerea inoculated on the potato leaves comparted with that on the the PDA plates the by using qRT-PCR. Results showed that all of these four targets were down-regulated ( Figure 1B ), showing a negative regulation with their sRNAs.

Evidences showed that exogenous RNA-induced gene silence (ERIGS) is an efficient method to control grey mould disease by adding B. cinerea gene-targeted sRNAs (Wang et al., 2016; Wang et al., 2017; Meng et al., 2020; Qiao et al., 2021). Therefore, we used ERIGS to test the antifungal activities of these 3 cross-species sRNAs against B. cinerea. These three cross-species sRNAs (miR396a-5p, siR3 and siR14) were added separately into the B. cinerea conidiospore solution to a final concentration of 10 μM and then drop-inoculated to at least six tobacco leaves for 3 days. Result showed that the average diameter of necrosis reached ~11 mm on B. cinerea-inoculated leaves with water or NC RNA treatments ( Figures 2A, B ). On the leaves of sRNA-treated B. cinerea, three sRNAs had smaller necrosis than those mock treated ( Figure 2A ). Among them, siR14 hardly had a necrotic diameter (4.1 mm), followed by miR396a-5p (5.8 mm), and the remaining siR3 had a similar necrotic diameter (~7.5 mm, Figure 2B ). Correspondingly, the virulence factors coding genes (Bcin15p02590.2 and Bcin04p04920.1) targeted by miR396a-5p were significantly down-regulated in B. cinerea treated by miR396a-5p, but no expression changes in B. cinerea treated by the other two sRNAs ( Figures 2C–E ). Similarity, the expression of Bcin04p03120.1 targeted by siR3 and Bcin01p06260.1 targeted by siR14 were only down-regulated in B. cinerea treated by siR3 and siR14, respectively ( Figures 2C–E ). These results suggesting that these 3 sRNAs control grey mould disease by targeting the virulence genes of B. cinerea.

Evidences showed that adding exogenous sRNAs can inhibit the spore germination and hence infection of pathogenic fungi by inducing RNAi pathway (Bilir et al., 2019; Meng et al., 2020). Therefore, Germination assays were performed at 24 h, 48 h and 72 h to understand the effects of the 3 sRNAs on the conidiospore germination of B. cinerea. Results showed that most of the spores incubated with NC-RNA germinated successfully and the mycelia grew well within 24 hours, but the spores treated with each of these 3 sRNAs (miR396a-5p, siR3 and siR14) failed to germinate ( Figure 3 ). These results suggest that RNA has a significant inhibitory effect on spore germination of B. Cinerea.

miR396a-5p which has a medium antifungal activity in all of three sRNAs was selected in the subsequent assay for the effects of sRNA concentration. miR396a-5p was added to the conidiospore solution of B. cinerea with a final concentration of 5, 10 and 20 μM, and then inoculated to tobacco leaves for 4 days. For the single stand sRNA, miR396a-5p with 5μM has a largest necrosis diameter (13 mm); followed by 10μM; miR396a-5p treated with 20 μM has a minimal necrosis (8 mm) ( Figures 4A, C ). These results indicated that high concentration of sRNA has better anti-B.cinerea activity. A similar result was observed for the double strand miR396a-5p ( Figure 4B, C ). These results showed that the suppression by double strand miR396a-5p (ds-miR396a-5p) was more efficient than that by single strand miR396a-5p (ss-miR396a-5p) ( Figure 4C ). In addition, the antifungal activity of ds-miR396a-5p were also investigated on tomato leaves through inoculating conidiospore solution of B. cinerea with 10 uM ds-miR396a-5p. Results showed that the necrosis area of B. cinerea were significantly reduced in ds-miR396a-5p-treated leaves compared with that in NC RNA-treated leaves ( Figure 5 ), exhibiting strong antifungal activity.

To understand the anti-B. cinerea mechanism, transient co-expression was performed to validate two B. cinerea coding genes targeted by miR396a-5p which was predicted by psRNAtarget ( Table 2 , Table S3 ). We firstly constructed two types of the target gene vectors, which contained either a target site or lacked a target site, to further validate the identified targets of miR396a-5p ( Figure 6A ). target constructs together with the pre-miR396a were transiently co-expressed in N. benthamiana. As the control, each of two wild type target constructs was also transiently expressed in N. benthamiana. Compared with transient expression of the control, the relative expression levels of Bcin15g02590.2 and Bcin04g04920.1 without miR396a-5p target sites were not repressed by co-expression with pre-miR396a ( Figure 6B ). However, the transcript levels of Bcin15g02590.2 and Bcin04g04920.1 containing a target site were significantly decreased when co-expressed with pre-miR396a ( Figure 6B ). The results suggested that both virulence genes of B. cinerea could be negatively regulated by miR396a.