Two tomato near-isogenic cultivars (cv.), the resistant Motelle (Mot, I2/I2) and susceptible Moneymaker (MM, i2/i2), described previously, were used in this study (Ouyang et al., 2014). Tomato seedlings growing at 25°C with a 16/8-h light/dark cycle for 2 weeks were used for all experiments. The wild-type F. oxysporum f. sp lycopersici (race 2) (FOL) strain was FGSC 9935. Two-week-old tomato seedlings were removed from soil and roots incubated in a solution of FOL conidia at a concentration of 1 × 10⁸/ml for 30 min. Water-treated tomato seedlings were used as the negative control. Forty seedlings were used for each treatment. Plants were then replanted in soil and maintained in a green house at 25°C for 24 h with constant light as described previously (Ouyang et al., 2014). Plants were removed from soil, and roots were rinsed gently and excised, then immediately frozen in liquid nitrogen and stored at -80°C. In order to control for experimental variation, all experiments were repeated three times.

Total RNA was isolated from roots using TRIzol^® Reagent (#15596026, Life Technologies, CA, United States) according to the manufacturer’s recommendations. For miRNA Northern blot analysis, 20 μg of total tomato root RNA was resolved using urea polyacrylamide gel electrophoresis (PAGE). MiRNA-specific oligonucleotide probes (Table 1) were end-labeled using γ-32P-ATP (#M0201, New England Biolabs, Ipswich, MA, United States). Blots were stripped and reprobed using at most one additional miRNA probe. The upper section of the blot was used for the loading control using a U6 oligonucleotide probe. All blots were imaged using a PhosphorImager (Molecular Dynamics/GE Life Sciences, Pittsburgh, PA, United States) and band intensities quantified using Imagequant software.

Transcript levels for FRG3 were determined using both Northern blot analysis and quantitative reverse transcriptase PCR (qRT-PCR). For Northern analysis, 10 μg of total RNA was resolved on 1% agarose gels and processed as described previously (Ouyang et al., 2014). Probe templates were prepared by amplification of cDNA using specific primers in PCRs (Table 1). Probes were labeled using the random priming method according to the manufacturer’s protocol (#U1100, Promega, San Luis Obispo, CA, United States). All blots were stripped and reprobed using 18S RNA probe as a loading control. Blots were imaged and band intensities quantitated as described above.

For qRT-PCR analysis, 1 μg of total RNA was used for first strand cDNA synthesis with Murine Leukemia Virus Reverse Transcriptase (M-MLV; #4368813, Life Technologies, Grand Island, NY, United States). Diluted cDNA was employed as the template for qRT-PCR (iQ5, Bio-Rad, Philadelphia, PA, United States) using Action as internal control. FRG3 gene-specific primers are listed in Table 1.

DNA fragments encoding sly-miR482d-3p and sly-miR482e-3p and the target gene FRG3 were inserted into vector GATEPEG100. All constructs were transformed into Agrobacterium tumefaciens strain GV3101. N. benthamiana plants were maintained in a green house at 25°C with a 16/8-h light/dark cycle. A. tumefaciens cultures were grown in liquid LB medium with selection (Qiao et al., 2013). All constructs were co-infiltrated into N. benthamiana leaves. After 40 h, infiltrated leaves were harvested, and cellular protein was extracted (Kocken et al., 1993). Western blot analysis with Anti-FLAG Antibody (#635691, Clontech, Mountain View, CA, United States) was conducted as described previously (Ouyang et al., 2014).

MiRNA-target validation was performed using a 5′ RACE assay (Ouyang et al., 2014). Briefly, total RNA was isolated from infiltrated N. benthamiana leaves. The Poly (A⁺) mRNA fraction was directly ligated using an RNA Oligo adaptor. Reverse transcription was performed, followed by semi-quantitative PCR using gene-specific primers (see Table 1). PCR products were cloned into the pGEM-T Easy Vector (#A1360, Promega) and sequenced.

Virus-induced gene silencing was utilized to suppress expression of FRG3 using TRV-based vectors (pTRV1 and pTRV2) (Ouyang et al., 2014). Briefly, the 3-UTR of FRG3 was amplified using gene-specific primers (Table 1) and cloned into the pTRV2 vector. Vectors for silencing of the Phytoene Desaturase (PDS) gene were used as a positive control (Mantelin et al., 2011). Four weeks after infiltration, leaves were harvested, RNA isolated, and the degree of silencing determined using qRT-PCR. The same plants were then infected with FOL or water for phenotypic analysis. Disease symptoms of VIGS plants were assessed after four more weeks. Genomic DNA was isolated from leaves and used for determining relative levels of FOL using qPCR of the rRNA intergenic spacer region (IGS) (Validov et al., 2011).

A phylogenetic analysis was carried out for the six members from the protein I2 family combining with four NBS-LRR proteins which were reported by our group previously (Ouyang et al., 2014). All amino acid sequences were obtained from the Sol Genomics database¹. All sequences were aligned using ClustalW. The evolutionary history was inferred by using the Maximum Likelihood method based on the JTT matrix-based model (Jones et al., 1992). The percentage of trees in which the associated taxa clustered together is shown next to the branches. Initial tree for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model, and then selecting the topology with superior log likelihood value. Point accepted mutation (PAM) was used for an amino acid transition matrix. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. The analysis involved 12 amino acid sequences. All positions containing gaps and missing data were eliminated. There were a total of 204 positions in the final dataset. OsGAPDH was used as a root by midpoint method. Evolutionary analyses were conducted in MEGA7 (Kumar et al., 2016).

All data were subjected to Student’s t-test analysis by SPSS 11.5 (SPSS Company, Chicago, IL, United States).

MiRBase21 has seven miR482 entries including miR482a, b, c, d-5p, d-3p, e-5p, and e-3p of which five correspond to miR482s targeting NBS-LRRs and two correspond to the complementary miR^∗ sequences (miRBase21, Kozomara and Griffiths-Jones, 2014). SlmiR482 is unusual among microRNA families, in that most members are 22 rather than 21 nucleotides, and have more sequence variability than other miRNA families (Figure 1A) (de Vries et al., 2015). To characterize the response of the slmiR482 family to FOL in tomato, we performed small RNA Northern blot analysis with specific probes for all seven members of the slmiR482 family. Interestingly, the RNA blot results revealed that all seven members were expressed and regulated differentially in two tomato cultivars plants after FOL infection, which differed from our previous sRNA-seq results (unpublished data) (Figure 1B and Supplementary Figures S1–S7). Levels of slmiR482e-5p were suppressed significantly in both Moneymaker and Motelle upon FOL infection. SlmiR482b, slmiR482d-3p, and slmiR482e-3p were down-regulated in Motelle, but, on the contrary, were up-regulated slightly in Moneymaker after FOL treatment. However, slmiR482d-5p presented the opposite pattern, with decreased levels in Moneymaker, and increased amounts in Motelle significantly, when treated with FOL (Figures 1B,C). These findings suggest that slmiR482 family members play different roles during the defense of tomato against the pathogen FOL.

Based on the results from both sRNA-seq and small RNA northern blot analysis, slmiR482d-3p and slmiR482e-3p were determined to be down-regulated in Motelle but upregulated in Moneymaker after infection, suggesting that both may negatively regulate levels of resistance gene mRNAs and/or their translation in tomato. Therefore, we focused our attention on the targets of slmiR482d-3p and slmiR482e-3p.

We utilized the psRNATarget algorithm (Dai and Zhao, 2011) to predict targets of the slmiR482 family. For each member, we found several potential targets in the tomato genome (Supplementary Table S1). Intriguingly, Solyc12g099060 was a putative target for both slmiR482d-3p and slmiR482e-3p, with a two-nucleotide shift in the binding site (Figure 2A). Solyc12g099060 is predicted to encode an NBS-type resistance protein containing coiled-coil (CC) and P-loop domains. The two miRNAs are predicted to bind in the P-loop region of the transcript (Figure 2B).

It is well known that the dominant I2 locus in tomato, introgressed from the wild tomato species S. pimpinellifolium, confers resistance against FOL race 2 (Simons et al., 1998). Six homologs, including I2, I2C-1, I2C-2, I2C-3, I2C-4, and I2C-5 were identified at the I2 locus in tomato (Simons et al., 1998). To clarify the genetic homology between the I2 family and Solyc12g099060, phylogenetic analysis was performed based on the amino acid sequences. Our results showed that the I2 family clusters in a separate group from Solyc12g099060, as well as the four NBS-LRR genes that we studied previously, Solyc08g075630, Solyc08g076000, Solyc05g008650, and Solyc09g018220 (Ouyang et al., 2014). Solyc08g075630 and Solyc08g076000 were targeted by slmiR482e-3p (slmiR482f), and Solyc05g008650 and Solyc09g018220 were targets of slmiR5300. Surprisingly, Solyc12g099060 shares a more recent common ancestor with Solyc09g018220 than it does with the other targets of 482e-3p (Figure 2C).

To test the possibility that the presence of FOL would affect the expression of Solyc12g099060, we checked the transcript level of Solyc12g099060 under water/FOL treatments in both tomato varieties using both total RNA Northern blot analysis and quantitative RT-PCR. Basal expression levels of Solyc12g099060 mRNA were 1.5-fold higher in Motelle relative to Moneymaker (Figure 3A). Importantly, Solyc12g099060 mRNA was induced almost eightfold in Motelle after treatment with FOL, but only slightly increased in Moneymaker under the same conditions (Figure 3A). The Northern blot results (Figure 3A and Supplementary Figure S8) were consistent with those obtained using quantitative RT-PCR (Figure 3B). These findings support the conclusion that expression of Solyc12g099060 is induced by FOL infection in both tomato cultivars.

Based on the analysis above, we named Solyc12g099060 FRG3 (FOL Resistance Gene 3).

To verify how the miRNA regulates the FRG3 target, we implemented Agrobacterium-mediated transient co-expression experiments in N. benthamiana. FRG3 and slmiRNAs were inserted into a binary construct containing a FLAG-tag (for FRG3). Vectors containing FRG3 alone or slmiR166, which does not recognize FRG3, served as negative controls.

Total RNA was extracted from N. benthamiana leaves infiltrated with the Agrobacterium strains. We performed qRT-PCR to check the transcript level of FRG3 during co-expression with a miRNA. In the presence of slmiR482e-3p, levels of the FRG3 transcript were greatly decreased (Figure 4A). We then checked for possible translational control of FRG3 by the miRNA using Western blot analysis with antibody against the FLAG-tag. Our data showed that levels of FRG3 protein were down-regulated significantly by the presence of slmiR482e-3p (Figure 4B). Surprisingly, co-expression of FRG3 and slmiR482d-3p in N. benthamiana did not lead to a detectable change in FRG3 transcript or FRG3 protein (data not shown). These results strongly suggest that slmiR482e-3p, but not slmiR482d-3p, is responsible for the down-regulation of the FRG3 target gene. The observation of suppressed FRG3 mRNA levels is consistent with slmiR482e-3p acting on FRG3 mainly at the transcript stability level.

To verify the cleavage site on the target mRNA, we performed RNA ligase mediated 5′ rapid amplification of cDNA ends (5′RACE) PCR analysis to detect the product of slmiR482e-3p mediated cleavage of FRG3 mRNA after transient coexpression in N. benthamiana. Our results indicate that the major cleavage site in FRG3 catalyzed by slmiR482e-3p occurred after the tenth nucleotide from the 5′ end of the miRNA (Figure 4C).

To explore a possible role for FRG3 in resistance to FOL, we employed a TRV-based VIGS system using the extreme 3′ end of the open reading frame (ORF) of FRG3 and a portion of the 3′ untranslated region to knock-down the expression of FRG3 in the resistant cultivar Motelle. As a positive control, Phytoene Desaturase (PDS) TRV-silenced plants (TRV-PDS) were generated in parallel (Mantelin et al., 2011). The photobleached phenotype was observed in TRV-PDS plants 4 weeks after TRV infection, signifying that silencing of the PDS gene had occurred. Therefore, all VIGS plants were treated with FOL 4 weeks after TRV infection. Motelle plants treated with water and Motelle plants transduced with empty TRV vector served as negative controls. Disease phenotypes were scored 4 weeks after FOL infection.

For all VIGS plants, transcript levels of FRG3 were quantified by qRT-PCR prior to FOL infection. The data indicated that the mRNA level of FRG3 was down-regulated by ∼60–80% in VIGS plants compared to control Motelle plants (Figure 5A). We also checked the mRNA level of Mi-1 and I2 to eliminate the possibility of off-target effects during VIGS (Figure 5A). All FOL-treated VIGS plants grew more slowly than control plants treated with water (Plant 1) and exhibited severe leaf wilting discoloration disease symptoms (Figure 5B; Plants 2–4).

To estimate the biomass of FOL in infected VIGS plants, we amplified the rRNA IGS from genomic DNA isolated from tomato leaves using qPCR. Our data demonstrates that FOL levels were elevated significantly in inoculated FRG3 VIGS plants, particularly in Plants 2 and 3 (Figure 5C). This result is consistent with observed disease severity symptoms.