B. cinerea strains used in this study (Table 1) were grown and spores were collected as described in Zhang and van Kan (2013). Tomato (S. lycopersicum cv. Moneymaker) were grown in a greenhouse at 20°C for five to six weeks, and detached compound leaves were used for inoculation assays.

B. cinerea conidia were diluted in Potato Dextrose Broth (PDB, 12 g/l) medium to 1000 conidia/μl and 2 μl droplets containing the conidia suspension were inoculated on detached tomato leaves essentially as described by Zhang and van Kan (2013). The inoculated compound leaves were inserted in wet floral foam and incubated in closed containers at 20°C with relative humidity of ~100% under natural light, before being examined for the virulence assay or sampled for further purposes. For the virulence assay, each tomato leaflet was inoculated with three droplets on every leaf half for one B. cinerea strain. For the sampling of B. cinerea (or mock) - infected tomato leaves that was used for mRNA and sRNA sequencing, 10 droplets of 2 μl conidia suspension (or only PDB) were inoculated on both leaf halves at ~1 cm from the central vein. Four leaflets of one compound leaf were inoculated at the same time, and one leaflet was sampled by excising the central vein and collecting the remaining leaf tissue at each defined time point (t=0; 12; 16 and 24 hpi). For the sampling B. cinerea (or mock) - infected tomato leaves that was used for the quantitative reverse transcription PCR (RT-qPCR), the leaflets were inoculated in six to eight circular areas and area included five 2 μl droplets of conidia suspension (or only PDB). The inoculated areas were excised by a cork borer with a diameter of 15.6 mm at each time point (t=0; 12; 16, 24 and 36 hpi). More details on the inoculation and sampling design were described in Qin et al. (2022).

Fungal mycelium or tomato leaf samples were frozen in liquid nitrogen and used for extraction of small RNA using the mirPremier^® microRNA Isolation Kit (Sigma-Aldrich) while mRNA was isolated using the Maxwell^® 16 LEV Plant RNA Kit (Promega).

Single-end Illumina sequencing was applied to all sRNA and mRNA samples by Vertis Biotechnologie AG (Martinsried, Germany) on a strand-specific library with read length of 75 nt. Sequence processing and bioinformatic analyses of data are described in Qin et al. (2022).

Differentially expressed genes (DEGs) in the fungus were identified by comparing fungal mRNA levels in the infected tomato leaf tissues with a B. cinerea liquid culture, as described in our recent study (Qin et al., 2022). Differentially produced sRNAs in tomato were analyzed by comparing tomato sRNA levels in the B. cinerea-infected tomato leaf tissues with the mock-inoculated leaf tissues at the same time point, as described in Qin et al. (2022). The thresholds for up- or down- regulation were calculated similarly for both mRNA and sRNA, using the DEseq2 algorithm through a negative binomial distribution to calculate the p-value. The significance was defined by thresholds that consisted of a p-value of lower than 0.05 and a log2 fold-change of higher than 1 or lower than -1 for up- or down-regulation, respectively. To determine the origin of the sRNA reads, the reads were mapped to both the B. cinerea and S. lycopersicum genomes. Bowtie was used as mapping tool with the constraint to only map reads without mismatches (-v 0). Reads that had a perfect match to the B. cinerea genome were also mapped to the S. lycopersicum genome and vice versa. Reads mapping perfectly on both genomes were labelled as ‘shared reads’ and were discarded, while reads that mapped to only one genome were labelled as ‘unique reads’. Target prediction of sRNAs was performed using sRNAs extracted from the ‘unique reads’. The tool psRNATarget was used to predict the plant sRNA targets on the fungal mRNA population (Dai et al., 2018), with settings adjusted to the default Schema V2 2017. The ‘expectation’ was set to match a free energy threshold of -20 kCal/mol (expectation 3). Both UTRs and CDS were tested for target sites. The sequences of the UTRs and CDS were obtained using the coordinates of the ASM83294v1 release-51 B. cinerea annotation. Predicted sRNA-target pairs were filtered by expression keeping only the sRNA-target pairs that showed sRNA up-regulation and target mRNA down-regulation.

Synthesis of cDNA from mRNA was performed using M-MLV reverse transcriptase (Promega). For reverse transcription of sRNA, the qScript microRNA cDNA Synthesis kit (Quanta Bioscience) was used. RT-qPCR was performed using SensiMix SYBR Hi-ROX Kit (Bioline). Primer combinations described in Supplementary Table 1 were used in RT-qPCR to quantify levels of mRNAs and sRNAs. The transcript level of a ribosomal protein encoding gene Bcrpl5 from B. cinerea was used to normalize fungal mRNA levels. A sRNA of the tomato U6 spliceosomal RNA component was used to normalize plant sRNAs. The threshold cycle (CT) values were determined by Bio-Rad CFX Manager 3.1 and fold-changes calculated using the 2^−ΔΔCt method (Pfaffl, 2006).

B. cinerea mutant strains used in this study were generated by PEG-mediated protoplast transformation as described by (Kars et al., 2005) with minor modifications. The constructs of donor templates were made by the yeast recombination method as described by Schumacher (2012). For the construction of plasmids as well as the amplification of donor templates, PCR was performed with the primer sets shown in Supplementary Table 1 using the Expand™ High Fidelity PCR System (Sigma). After obtaining transformed colonies on hygromycin-selective plates, the screening of transformants was performed by PCR with primer sets indicated in Supplementary Table 1 using the GoTaq^® G2 DNA Polymerase (Promega). The coding region of the Bcspl1 gene from the mutants used in this study was sequenced to verify whether or not the sRNA target site contained the 5-nucleotide substitution.

Differential expression analyses of fungal genes were performed by comparing gene expression during infection of tomato leaves with a B. cinerea liquid culture, as described in Qin et al. (2022). More than 3000 B. cinerea genes (one quarter of the annotated genes) were differentially expressed at 24 hpi, as compared to growth in liquid culture, of which 1724 were up-regulated and 1282 were down-regulated (Supplementary Data 1). Among the up-regulated genes of B. cinerea at this time point of infection were genes encoding five polygalacturonases and 103 other Carbohydrate-Active enZYmes (CAZymes), 51 proteases, 63 membrane transporters, 5 proteins involved in signaling and 12 putative transcription factors. Moreover, the cluster of Bcboa genes encoding biosynthetic enzymes for the production of the polyketide phytotoxin botcinic acid were upregulated at 24 hpi, as compared to the in vitro liquid culture. Meanwhile, the in planta down-regulated genes (at 24 hpi) included genes encoding 36 CAZymes, 27 proteases, 39 membrane transporters, 33 proteins involved in signaling and 93 putative transcription factors. Moreover, there was significant down-regulation in planta at 24 hpi of melanin biosynthetic genes (Schumacher, 2016), of the gene cluster encoding biosynthetic enzymes for production of the sesquiterpene phytotoxin botrydial (Porquier et al., 2016), as well as the non-ribosomal peptide synthase gene Bcnrps8 and six polyketide synthase genes along with their flanking genes involved in synthesis of (yet unknown) secondary metabolites (Supplementary Data 1).

Within the entire dataset, approx. 15% of the B. cinerea genes (1713 genes) that were predicted targets of the tomato sRNAs indeed displayed significant down-regulation in at least one of the time points (12, 16 and 24 hpi) as compared to the liquid culture (Qin et al., 2022). These fungal transcripts were, on average, predicted to each be targeted by 19.3 unique tomato sRNAs (Supplementary Data 2). The list of in planta downregulated genes that are predicted to be silenced by tomato sRNAs includes genes encoding 63 CAZymes, 43 proteases, 46 membrane transporters, 46 proteins involved in signaling and 101 putative transcription factors (Supplementary Data 2). Only 30 fungal genes were predicted to be targeted by a single tomato sRNA, including transcripts encoding fumarase BcFUM1, glutathione S-reductase BcGST8, catalase BcCAT7, phosphatidylserine decarboxylase BcPSD, melanin biosynthetic enzyme BcBRN2 and the cell death-inducing effector BcSPL1.

After examining the in silico prediction of B. cinerea genes targeted by tomato sRNAs based on the sequencing dataset, we aimed to validate and establish in more detail the host sRNA – fungal mRNA profiles during the early infection. We performed new experiments to inoculate tomato leaves with B. cinerea and sampled at seven time points within the first 24 hpi. sRNA and mRNA were extracted from these samples, followed by quantification of the expression levels of selected tomato sRNAs (Sl-sRNAs) and their matching target mRNAs in B. cinerea (Bc-mRNAs) by reverse transcription-quantitative PCR (RT-qPCR). We selected three Sl-sRNAs and their predicted target Bc-mRNAs (Table 2) for molecular validation. The selection of sRNA-mRNA pairs was based on the following criteria: i) the predicted target mRNA showed sufficient reads in the sequencing dataset at all infection time points; ii) the predicted target mRNA was significantly down-regulated at one or more time point(s) as compared to the liquid culture; iii) the target gene might contribute to fungal infection; iv) the tomato sRNA was up-regulated in the early stages of infection and should be derived from a transposon locus. The following three B. cinerea genes were chosen for further analysis: the 5-oxoprolinase gene Bcoxp1 is a homolog of the Fusarium graminearum oxp1 gene which was reported to be involved in development and virulence (Yang et al., 2018); the gene Bccnd1, encoding a secreted effector protein that is expressed in a calcineurin-dependent manner (Viaud et al., 2003) and is homologous to GAS1 and GAS2 effectors of Magnaporthe grisea, expressed in appressoria and required for full virulence (Xue et al., 2002); and the cerato-platanin gene Bcspl1, encoding an effector that induces plant cell death and is important for virulence (Frías et al., 2011).

Molecular quantification results from RT-qPCR indicated correlations between Sl-sRNAs and their matching Bc-mRNA targets (Figure 1). Expression levels of the three tested B. cinerea genes displayed upregulation at 12 hpi, and they reached their lowest level at 20 hpi but increased again at 24 hpi. The down-regulation of the three Bc-mRNAs coincided with, or followed shortly after, an increase of the levels of their corresponding Sl-sRNAs. Specifically, the lowest expression of the three Bc-mRNAs occurred at 20 hpi, while Sl-sRNAs levels were high at 14 hpi or 16 hpi (Figure 1). Interestingly, levels of these three selected Sl-sRNAs all displayed an approximate doubling in the early stage of fungal infection between 12 and 14 hpi, or between 14 and 16 hpi.

From B. cinerea genes that were predicted to be targeted for silencing by tomato sRNAs, we selected one gene for experimental validation of a causal relation between the presence of the tomato sRNA and the down-regulation of its target B. cinerea mRNA. Selection of the gene was based on three criteria: its transcript should decrease at some time during infection, as compared to previous time point(s) in the dataset; the transcript should be (predicted to be) targeted by a single tomato sRNA, in order to minimize the impact of multiple sRNA-mRNA interactions; the gene should participate in virulence of B. cinerea. The Bcspl1 gene was selected as it is the predicted target of a single tomato sRNA (sRNA1187) and encodes a cell-death inducing effector protein with a role in virulence (Frías et al., 2011). Only 30 genes were predicted to be targeted by a single tomato sRNA (Supplementary Data 2) and Bcspl1 was the only gene in this list that had been reported to participate in virulence. During B. cinerea infection on tomato leaves, Bcspl1 displayed a peak in transcript level at 12 hpi and was ~10-fold down-regulated at 20 hpi (Figure 1). The lower transcript level of Bcspl1 coincided with a transient peak in the level at 16 hpi of the tomato sRNA1187 (Figure 1), which is produced from a transposable element on tomato chromosome 6 (Figure 2A).

In order to disrupt the ability of Bcspl1 mRNA to interact with sRNA1187, we aimed to generate a B. cinerea mutant carrying a Bcspl1 allele with mutations in the sRNA target site. A substitution of five nucleotides would result in a change of free energy of the hybrid between sRNA1187 and the Bcspl1 mRNA from -28 kCal/Mol (wild type) to -11 kCal/Mol (mutant), without changing the encoded protein (Figure 2B). The cut-off for free energy of RNA hybrids is -20 kCal/Mol (Marín and Vaníč, 2011), implying that this substitution would fully abolish sRNA-mRNA hybrid formation. We checked that the synonymous substitutions in the Bcspl1 target site would not result in a sequence that could inadvertently be targeted by different tomato sRNAs in the dataset. A gene replacement construct was generated that encompassed the entire Bcspl1 gene (containing the desired five-nucleotide substitution) and a part of the downstream gene Bcin03g00510 with a hygromycin-resistance cassette (hph) inserted in the intergenic region (Figure 3A). Transformation of this construct to wild type B. cinerea strain B05.10 yielded seven transformants of which two contained the hph cassette in the target locus and five were ectopic transformants (Figure 3B). The Bcspl1 gene from both transformants was amplified and sequenced. Transformant #5 contained the desired substitution and was named spl1-5mnt. Transformant #1 contained a wild type Bcspl1 sequence, presumably as a result of recombination with the target locus downstream of the sRNA target site (orange dashed lines in Figure 3A). This transformant was designated spl1-wt and served as a control transformant to exclude an impact on transcript levels caused by introducing a hph cassette close to the Bcspl1 locus.

We hypothesized that if the tomato sRNA would indeed participate in silencing of Bcspl1 in B. cinerea during infection, the synonymous substitutions in the sRNA target site would result in a distinct Bcspl1 transcript profile during infection, i.e., the downregulation between 16 and 24 hpi (as observed in Figure 1) would be abolished. This might result in a higher virulence of the fungal mutant if the increased production of BcSPL1 protein would enable the fungus to trigger host cell death more effectively.

We inoculated tomato leaves with both the spl1-5mnt B. cinerea mutant and the control spl1-wt transformant and sampled the tomato leaf samples at four timepoints between 12 and 36 hpi. RT-qPCR was performed to quantify the Bcspl1 expression in spl1-5mnt and spl1-wt mutants. Contrary to the hypothesis, the Bcspl1 transcript profile in spl1-5mnt was similar to that in spl1-wt (Figure 4A), indicating that the downregulation of Bcspl1 at 24 hpi was not abolished despite the substitution in the predicted target site for sRNA1187. Additionally, the level of tomato sRNA1187 was not significantly different between the leaf tissues infected by spl1-wt or spl1-5mnt isolates (Figure 4B). Infection assays were performed to compare the virulence of both transformants to each other and to wild type strain B05.10. As shown in Figure 5, there was neither a difference in virulence between spl1-5mnt and spl1-wt, nor between spl1-wt and wild type B05.10. These experiments did not provide any evidence for participation of tomato sRNA1187 in the downregulation of Bcspl1 mRNA during infection.