An RNAi-Based Control of Fusarium graminearum Infections Through Spraying of Long dsRNAs Involves a Plant Passage and Is Controlled by the Fungal Silencing Machinery

Meeting the increasing food and energy demands of a growing population will require the development of ground-breaking strategies that promote sustainable plant production. Host-induced gene silencing has shown great potential for controlling pest and diseases in crop plants. However, while delivery of inhibitory noncoding double-stranded (ds)RNA by transgenic expression is a promising concept, it requires the generation of transgenic crop plants which may cause substantial delay for application strategies depending on the transformability and genetic stability of the crop plant species. Using the agronomically important barley—Fusarium graminearum pathosystem, we alternatively demonstrate that a spray application of a long noncoding dsRNA (791 nt CYP3-dsRNA), which targets the three fungal cytochrome P450 lanosterol C-14α-demethylases, required for biosynthesis of fungal ergosterol, inhibits fungal growth in the directly sprayed (local) as well as the non-sprayed (distal) parts of detached leaves. Unexpectedly, efficient spray-induced control of fungal infections in the distal tissue involved passage of CYP3-dsRNA via the plant vascular system and processing into small interfering (si)RNAs by fungal DICER-LIKE 1 (FgDCL-1) after uptake by the pathogen. We discuss important consequences of this new finding on future RNA-based disease control strategies. Given the ease of design, high specificity, and applicability to diverse pathogens, the use of target-specific dsRNA as an anti-fungal agent offers unprecedented potential as a new plant protection strategy.


Introduction
Fusarium graminearum [19]. While these results provided proof-of-concept that RNAi-based plant protection is an effective strategy for controlling diseases caused by devastating necrotrophic pathogens, the broad applicability of this transgenic method remains questionable due to the persisting weak acceptance of GMO strategies for food and feed production in many countries. More important, a broad application of this transgenic approach is hampered by the lack of transformability of various crop plants and the missing genetic stability of the silencing trait. Here we investigate the potential and the mechanism of an RNAi-based crop protection strategy using direct spray applications of CYP3-dsRNA to target F. graminearum. We show that the 791 nt long dsRNA is taken up by the plant and transferred in an unmodified form via the vascular system to fungal infection sites where it is processed by the fungal RNAi machinery as a prerequisite for its antifungal activity. We show a strong correlation between accumulation of CYP3-dsRNA at infection sites, silencing of CYP51 expression, and fungal inhibition.

Spray-induced gene silencing (SIGS) of Fusarium genes
To provide a proof of concept, we conducted an experiment targeting the expression of the jellyfish green fluorescent protein (GFP) in the GFP-expressing F. graminearum strain Fg-IFA65 GFP [20] by using a GFP-specific 720 nt long dsRNA (GFP-dsRNA, S1 Fig). Detached barley leaves were locally sprayed with 20 ng μL -1 GFP-dsRNA or Tris-EDTA buffer (TE, control) and drop-inoculated 48 h later with Fg-IFA65 GFP in the distal (non-sprayed) leaf segment. Confocal microcopy showed strong GFP fluorescence associated with fungal mycelia on TEtreated control leaves at six days post inoculation (dpi) (Fig 1A). In contrast, fluorescence ( Fig  1B) and GFP transcripts ( Fig 1C) were largely absent in mycelia grown on leaves that were locally sprayed with GFP-dsRNA, although mycelial growth was unrestricted as evidenced by light microscopy. This observation clearly demonstrates the possibility of targeting a gene of an attacking microbe via SIGS.
To further explore the potential of SIGS, we assessed the silencing efficiency of CYP3-dsRNA, which targets the three Fusarium genes CYP51A, CYP51B, and CYP51C. The 791 nt long CYP3-dsRNA contains complementary fragments of these genes starting with N-terminal CYP51B, followed by CYP51A and CYP51C [19]. Leaves were sprayed with CYP3-dsRNA and 48 h later drop-inoculated directly onto the sprayed area with Fg-IFA65. At six dpi, CYP3-dsRNA-treated leaves developed brownish lesions that were substantially smaller than those on TE-or GFP-dsRNA-sprayed leaves that served as control in this experiment (Fig 2A). Quantitative real-time PCR (qPCR) analysis of fungal DNA levels, based on the ratio between fungal tubulin and plant ubiquitin, confirmed reduced fungal growth on CYP3-dsRNA-treated leaves ( Fig 2B). To confirm that inhibition of Fusarium growth by CYP3-dsRNA was provoked by sequence-specific gene silencing, expression of all the three fungal CYP51 genes was assessed. At six dpi, total RNA was isolated from infected leaves and the levels of CYP51A, CYP51B and CYP51C transcripts were measured by qPCR and normalized to the expression of the fungal ßtubulin gene. Consistent with the concept of spray-induced gene silencing, we found that the relative amounts of CYP51 transcripts were reduced on average by 58% (CYP51A), 50% (CYP51B), and 48% (CYP51C) in leaves sprayed with CYP3-dsRNA vs. the GFP-dsRNA control ( Fig 2C).

SIGS confers strong resistance against Fusarium in distal leaf parts
Mobile cell non-autonomous inhibitory RNAs that spread gene silencing into adjacent cells and tissues have been observed in various plants [21][22][23]. Encouraged by the observed reduction in GFP fluorescence in Fg-IFA65 GFP upon infection of leaf segments that did not receive a direct GFP-dsRNA spray (see Fig 1), we tested whether locally sprayed CYP3-dsRNA confers gene silencing in Fusarium infecting distal, non-sprayed segments of barley leaves. To this end, the upper part of detached leaves (local tissue) was sprayed with 20 ng μL -1 CYP3-dsRNA, GFP-dsRNA, or TE, while the lower part (distal tissue) was covered by a plastic tray to prevent direct dsRNA contamination. After 48 h, the distal, non-sprayed part of the leaves was dropinoculated with Fg-IFA65 GFP ; six days later, resistance to fungal infection was assessed. Distal leaf areas of CYP3-dsRNA-treated leaves developed substantially smaller lesions as compared to leaves sprayed with GFP-dsRNA or TE (S2 Fig) indicating that the silencing signal was basipetally transported. Consistent with this finding, the amount of fungal DNA as determined by qPCR was greatly reduced in the distal leaf area as compared to the control treatments ( Fig  3A). The relative amounts of fungal CYP51A, CYP51B and CYP51C transcripts were strongly reduced on average by 72% (CYP51A), 90% (CYP51B), and 71% (CYP51C) as compared with control (GFP-dsRNA) treatment ( Fig 3B). Confocal microscopy of fungal inoculation sites in distal leaf areas confirmed that, on TE-treated leaves, Fg-IFA65 GFP conidia had germinated and colonized tissue next to the inoculation site ( Fig 3C). In contrast, fungal mycelia on CYP3-dsRNA-treated leaves were only visible at the inoculation sites, and the surrounding leaf tissue was free of infection hyphae (Fig 3D). Consistent with this, the large number of fungal conidia with very short germ tubes at the inoculation sites of CYP3-dsRNA-treated leaves indicated that fungal germination was strongly impaired. We also tested whether the silencing signal was transported in the acropetal direction. Segments were sprayed with 20 ng μL -1 CYP3-dsRNA and subsequently drop-inoculated in the distal leaf area. Fusarium infections were also reduced in the acropetal experimental set up (S3A  Forty-eight hours after spraying, distal, non-sprayed leaf segments were drop-inoculated with Fg-IFA65 GFP (20 μL of a solution containing 2 x 10 4 conidia mL -1 ). GFP silencing efficiency was visualized 6 dpi using confocal microscopy. (C) GFP transcripts were quantified by qPCR at 6 dpi. The reduction in fungal GFP expression on leaves sprayed with GFP-dsRNA and infected with Fg-IFA65 GFP compared with TE-sprayed controls was statistically significant (***P < 0.001; Student´s t test). Bars represent mean values ± SDs of three independent experiments. Scale bars represent 100 μm.

Uptake and processing of sprayed CYP3-dsRNA
To further explore the SIGS mechanism, we investigated whether the spray-applied long CYP3-dsRNA is translocated in the plant tissue and/or processed by the plant's silencing machinery independent of fungal infections. Following CYP3-dsRNA treatment, local (sprayed) and distal (non-sprayed) leaf segments were harvested separately at 24, 48, 72, or 168 h after spraying. Northern blot analysis detected unprocessed 791 nt CYP3-dsRNA in both local and distal tissue (Fig 4A), showing that the long dsRNA is systemically translocated within the plant. In the local (sprayed) segment, CYP3-dsRNA was detected over the full time range, while it accumulated only transiently at early time points (24 h) after spraying in the Detached second leaves of three-week-old barley were sprayed evenly with CYP3-dsRNA, TE (mock control), and GFP-dsRNA (negative control), respectively. After 48 hours, leaves were drop-inoculated with 2 × 10 4 conidia mL −1 of Fg-IFA65 onto the sprayed area and evaluated for necrotic lesions at 6 dpi. (B) The relative amount of fungal DNA at 6 dpi as measured by qPCR was reduced in CYP3-dsRNA-treated leaves compared to control leaves. Bars represent mean values ± SDs of three independent experiments. The reduction of fungal growth on CYP3-dsRNA vs. TE-or GFP-dsRNA-sprayed leaves was statistically significant (*P < 0.05; Student´s t test). (C) Gene-specific qPCR analysis of fungal CYP51A, CYP51B, and CYP51C transcripts at 6 dpi (corresponding to 8 d after spraying). The reduction in fungal CYP51 gene expression on CYP3-dsRNA-sprayed leaves as compared with GFP-dsRNA-sprayed controls was statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001; Student´s t test). distal (non-sprayed segments). This accumulation profile is consistent with the idea that the vast bulk of the CYP3-dsRNA fraction was absorbed via the cut surface of the detached leaf. Moreover, CYP3-dsRNA-derived 21 nt and 22 nt small interfering (si)RNAs also accumulated over the whole time range after spraying in the local leaf segments, demonstrating that CYP3- week-old barley were sprayed evenly with CYP3-dsRNA, TE, and GFP-dsRNA, respectively. After 48 h, the non-inoculated, semi-systemic (distal) tissue was drop-inoculated with 2 × 10 4 conidia mL −1 of Fg-IFA65 GFP ; the relative amount of fungal DNA in distal tissue, as measured by qPCR at 6 dpi, was reduced in CYP3-dsRNA-treated leaves. Bars represent mean values ± SDs of three independent experiments. The reduction of fungal growth on CYP3-dsRNA-sprayed leaves was statistically significant (**P < 0.01; Student´s t test). (B) Gene-specific qPCR analysis of CYP51A, CYP51B, and CYP51C transcripts at 6 dpi in distal leaf areas. Bars represent mean values ±SDs of three independent sample collections. The reduction in CYP51 expression in leaves sprayed with CYP3-dsRNA compared with GFP-dsRNA-sprayed controls was statistically significant (**P < 0.01, ***P < 0.001; Student´s t test). (C,D) Microscopy of fungal growth at semi-systemic sites of drop-inoculation with Fg-IFA65 GFP . (C) Successful fungal colonization (green) on TE-sprayed leaves. Profuse hyphal growth is seen inside the cells (plasma membrane stained with RH414 is highlighted in red) (D) Hyphal formation is strongly reduced and confined to the inoculated leaf area on CYP3-dsRNA-sprayed leaves. Impaired spore germination was observed in the area around the inoculation site while the surrounding cells remained free of colonization. (IF, infection hyphae; GS, germinating spore). Photographs for C and D were taken at 6 dpi. dsRNA was partly processed by the plant (Fig 4B). In this experiment, Northern analysis could not detect siRNAs in distal leaf parts, probably because the technique was not sensitive enough.
To further investigate uptake and transport of sprayed CYP3-dsRNA, it was labeled with the green fluorescent dye ATTO 488 (CYP3-dsRNA A488 ) and sprayed onto barley leaves. The biological activity of CYP3-dsRNA A488 was indistinguishable from non-labeled CYP3-dsRNA as evidenced by reduced fungal infection and strong silencing of fungal CYP51 genes upon spray application (S4A- S4C Fig). Moreover, using confocal laser scanning microscopy, a green fluorescent signal was detected in the vascular tissue at 24 hours after spraying leaves with 20 ng μl -1 CYP3-dsRNA A488 . In leaf cross-sections, fluorescence was seen in the xylem (Fig 5A-5C). Inspection of longitudinal leaf sections revealed that the fluorescence was not confined to the apoplast but also was present in the symplast of phloem parenchyma cells, companion cells, and mesophyll cells, as well as in trichomes and stomata (Fig 5D-5I). When CYP3-dsRNA A488 -sprayed leaves were inoculated with Fg-IFA65, the fluorescent signal also was detectable inside fungal conidia and germ tubes ( Fig 5H) and fungal mycelium (Fig 5J and S5  Fig). Together these data show that CYP3-dsRNA is taken up by the plant and is transferred via the plant vascular system. Systemic translocation within the plant and accumulation by the fungus also raised the possibility that CYP3-dsRNA is processed by the fungus into inhibitory siRNAs to eventually target fungal CYP51 genes.
To test this possibility, we first profiled CYP3-dsRNA-derived siRNAs in infected and noninfected leaves. Small RNA sequencing (RNAseq) analysis revealed distinctly different CYP3-dsRNA-derived siRNA profiles in mock-vs. Fg-IFA65-infected local and distal (non-sprayed) leaf segments (Fig 6A and 6B) with higher numbers of reads of CYP3-dsRNA-derived siRNAs in infected leaves, and highest numbers of reads in locally-inoculated vs. distally-inoculated leaves. These data suggest that CYP3-dsRNA also is processed by the fungus and that the fungal silencing machinery is involved in SIGS and reduced fungal infections. Detection by RNAseq of CYP3-dsRNA-derived siRNA in the distal (non-sprayed) part of leaves also supported our interpretation that northern analysis failed to detect low amounts of these siRNAs due to sensitivity problems.

Fungal DICER-LIKE-1 is required for efficient SIGS in systemic leaf areas
To further substantiate involvement of the fungal silencing machinery, we generated a fungal dcl-1 mutant (Fg-IFA65 Δdcl-1 ) that is deficient for DICER-LIKE 1 (S6 Fig), a critical component of the fungal silencing machinery that produces siRNA from long dsRNA stretches. Fg-IFA65 Δdcl-1 and the wild type Fg-IFA65 were indistinguishably virulent on TE-sprayed barley leaves (Fig 7A), showing that fungal DCL-1 is not required for successful leaf infections. However, in contrast to Fg-IFA65, the mutant Fg-IFA65 Δdcl-1 also heavily infected distal areas of CYP3-dsRNA-treated barley leaves (Fig 7B), suggesting that the mutant strain is not amenable to SIGS. We concluded that the fungal silencing machinery appears to be indispensable for CYP3-dsRNA-mediated SIGS at systemic areas in the barley-Fusarium graminearum pathosystem. To further confirm that FgDCL-1 is required for CYP51 target gene silencing, levels of CYP51A, CYP51B and CYP51C transcripts were compared by qPCR in the wild type vs. the dcl-1 mutant on infection of CYP3-dsRNA sprayed leaves. The relative amounts of transcripts were reduced in Fg-IFA65 on average by 50% (CYP51A), 70% (CYP51B), and 40% (CYP51C) as compared with TE (control) treatment. In contrast, expression of CYP51 targets was not reduced in the Fg-IFA65 Δdcl-1 mutant (Fig 7C).
We additionally conducted an in vitro experiment to further demonstrate the requirement of FgDCL-1 for CYP3-dsRNA-mediated silencing of fungal CYP51 genes. Mycelia of axenic cultures of Fg-IFA65 and Fg-IFA65 Δdcl-1 were treated with CYP3-dsRNA. Subsequently, expression of CYP51 genes was recorded. Consistent with the leaf assay, the relative amounts of fungal CYP51A, CYP51B and CYP51C transcripts were reduced in the wild type Fg-IFA65 but not in  the Fg-IFA65 Δdcl-1 mutant (Fig 7D). Confirmatory total sRNAs profiling by RNAseq in axenically-grown Fg-IFA65 revealed a range of sRNAs originating from CYP3-dsRNA (Fig 7E-7G and S2 Table), further proving that the fungus can process CYP3-dsRNA. Suspiciously, the majority of siRNA species mapped to sites in the CYP51A gene fragment of the CYP3-dsRNA. Further work must show if this profile is a result of the physical structure of the dsRNA.

CYP3-dsRNA-derived siRNAs also confer SIGS
The failure to detect CYP3-dsRNA-derived siRNA in the distal area of CYP3-dsRNA-sprayed leaves by northern analysis along with the compromised SIGS phenotype of the mutant Fg-IFA65Δdcl-1 suggested that the concentration of siRNA in the distal leaf parts was too low to mediate silencing of CYP51 genes in the fungus. Alternatively, Fusarium is generally unable to absorb siRNA from barley leaves. To address these possibilities, we sprayed barley leaves with high concentration of CYP3-dsRNA-derived siRNAs (20 ng μl -1 ) and subsequently inoculated local (sprayed) and distal (non-sprayed) leaf segments with Fg-IFA65. We found that the fungus was strongly inhibited by these siRNAs both in the local (S7A and S7C Fig

SIGS-mediated fungal inhibition is independent of innate immune responses
In mammalian cells, perception of certain dsRNAs via toll-like receptors triggers an inflammation response [24,25]. Therefore, we assessed whether CYP3-dsRNA elicits an innate immune response called pattern-triggered immunity (PTI) [26], when sprayed onto barley leaf segments. To this end, expression of barley genes that are indicative of the canonical defenserelated salicylate-and jasmonate-dependent pathways [27] was evaluated. Expression of salicylate-responsive pathogenesis-related 1 (HvPR1) and jasmonate-responsive S-adenosyl-l-methionine:jasmonic acid carboxyl methyltransferase (HvJMT) in TE-treated leaves was strongly induced by Fg-IFA65, but not by CYP3-dsRNA treatment (Fig 8). Furthermore, Fg-IFA65-induced expression of either gene was much lower in CYP3-dsRNA-treated leaves as compared with TE-treated leaves. This result strongly suggests that CYP3-dsRNA does not induce PTI in barley, and that the SIGS mechanism does not rely on activation of canonical defense pathways. The finding also is relevant when considering the fitness cost, and thus yield, of the SIGS strategy.

Discussion
In this study, we show that delivery of long noncoding double-stranded RNA targeting the three CYP51 genes of the necrotrophic ascomycete fungus Fusarium graminearum via spray application effectively reduces the development of the pathogen on barley leaves. Thus, our work further supports the idea that RNA could be used as a chemical treatment to control plant diseases.
Next to the economic and ecologic consideration about the deployment of antimicrobial RNAs as a new plant protection measure, elucidating the molecular mechanisms of RNAbased disease control is a key for successful future implementation. While plant-derived transgene-mediated silencing of target genes in plant pathogens and pests (a mechanism known as  (HvPR1) and (B) S-adenosyl-l-methionine:jasmonic acid carboxyl methyltransferase (HvJMT). Both genes are highly responsive to Fg-IFA65 but not to CYP3-dsRNA or TE treatment. Please note that a combined treatment of CYP3-dsRNA followed by Fg-IFA65 48 h later also did not induce these marker genes, which shows independently that fungal development on CYP3-dsRNA-treated leaves is strongly inhibited. host-induced gene silencing [HIGS]) has been frequently reported [12,14,19], few examples have demonstrated the efficiency of exogenous RNA delivery to kill plant attackers. HIGS virtually is based on the plant's silencing machinery, whereas the mechanism of gene silencing by exogenously delivered dsRNA constitutes a more complex situation. For instance, diverging questions are possible involvement of the silencing machineries of host and/or the pathogen/ pest, the requirement of local and remote transport of channeled dsRNA molecules, and the problem of dsRNA transport at the apoplast-symplast interface.
Using the F. graminearum-barley pathosystem as a model to study the mechanism of exogenously applied inhibitory dsRNA was motivated by the fact that Fusarium Head Blight and Crown Rot cause serious problems worldwide including food and feed safety issues due to mycotoxin contamination of cereals and maize. Focusing on fungal CYP51 genes as targets for silencing was reasonable because our previous work provided proof-of-concept that transgenemediated silencing of these genes effectively reduced fungal development in Arabidopsis and barley. More than that, direct treatment of F. graminearum with inhibitory dsRNA matching CYP51 gene sequences had been demonstrated to inhibit fungal development in axenic cultures [19].
The previous finding of impaired fungal growth in axenic cultures, when treated with a 791 nt dsRNA (CYP3-dsRNA) targeting the three fungal genes CYP51A, CYP51B, and CYP51C, let us speculate that the fungus can process CYP3-dsRNA into siRNA that interfere with the expression of CYP51 genes. Gene annotation of F. graminearum's genome (http://www. broadinstitute.org) predicted genes coding for two ARGONAUTE-like proteins, two DICERlike proteins and five RNA-dependent RNA Polymerases (RDR), suggesting that the RNAi pathway is functional [28]. Consistent with these findings, RNAseq analysis of axenically grown F. graminearum, treated with CYP3-dsRNA, showed high numbers of reads of CYP3-dsRNA-derived siRNA (Fig 7). Together these data showed that F. graminearum possesses a functional gene silencing system, which is a prerequisite for disease control by SIGS.
To test the antifungal activity of CYP3-dsRNA and their siRNA derivatives, we used a detached leaf assay that enabled us to assess fungal growth in local (directly sprayed) and distal (semi-systemic, non-sprayed) leaf segments. Using this approach, we could demonstrate that inhibitory dsRNA translocated via the plant vascular system and eventually was absorbed by the pathogen from leaf tissue (Fig 5). The profile of inhibitory dsRNA accumulation, as demonstrated by northern blot analysis (Fig 4) and RNAseq (Fig 6), showed that both long CYP3-dsRNA and plant-processed CYP3-dsRNA-derived siRNA accumulate in the plant vascular system, though translocation of siRNA seems to be less efficient and thus siRNA concentration at the remote infection sites probably was not high enough to induced SIGS. Consistent with this notion, in planta produced CYP3-dsRNA-derived siRNAs was detected in the distal leaf parts only by the more sensitive RNAseq technique but not by northern analysis. Nevertheless, spraying high concentrations of CYP3-dsRNA-derived siRNA (20 ng μL -1 ) induced the SIGS process (S7 Fig), demonstrating that the fungus is able to absorb siRNAs from barley leaves.
Because of the less efficient movement of siRNAs, transport and translocation of the unprocessed 791 nt dsRNA has a critical role in the SIGS process, demonstrated by a compromised SIGS phenotype of the Fusarium mutant Fg-IFA65 Δdcl-1 at distal leaf segments (Fig 7) but not at local, directly sprayed leaf segments (S8 Fig). Compromised DICER activity resulted in the fungus inability to cleave CYP3-dsRNA into siRNA, thus interrupting the RNA interference mechanism in case the concentration of CYP3-dsRNA-derived siRNA is not sufficiently high. Our finding that the unprocessed long dsRNA could be absorbed from leaf tissue has further implications for future disease control strategies using dsRNA. There are good arguments that application of longer dsRNAs might be more efficient than application of siRNAs given there more efficient translocation. Long dsRNA would be processed into many different inhibitory siRNA by the fungus, which not least also could be an issue when considering the risk of compound resistance emerging in a pathogen under field test conditions. Thus, based on our findings, further research is required to establish rules for optimal dsRNA structures, considering e.g. molecule lengths, combinatorial order of gene fragments, target sites in a given gene target, and the number of genes targeted by one dsRNA. Supporting the requirement for more information as to the design of dsRNA probes, RNAseq analysis revealed that most of the CYP3-dsRNA-derived siRNAs, accumulating in the axenic fungal mycelium, were not equally distributed at the CYP3-dsRNA scaffold, but accumulated at the CYP51A gene fragment (Fig 7 and  S2 Table). Further analysis is required to explain this bias in the production of siRNAs from CYP3-dsRNA.
Our results are also consistent with the view that inhibitory dsRNA is more effectively absorbed by the fungus through infection hyphae that have intimate contact to plant tissue (compare CYP51 gene expression in Figs 2 and 3). How these hyphae differ from the fungal germ tubes and extracellular hyphae is however yet unclear although distinct biochemical modifications of fungal hyphae that penetrate host plants and are involved in nutrient uptake have been demonstrated [29,30]. Thus it is likely that these specialized, leaf tissue colonizing hyphae show dsRNA uptake that is superior over germ tubes.
Our data are consistent with reports showing that silencing signals in plants are mobile [31,32]. The design of our experiments based on the previous finding that sRNAs, just as viroids [33], preferably move via the vascular system in the source-to-sink direction although some reports discussed transport in the opposite route (for review see [21]). Source-to-sink movement is one reason why the phloem rather than the xylem is generally considered as the conduit for movement of the silencing signal. This hypothesis is supported by the finding that the xylem sap, which transports water and ions, commonly is free of RNA [34]. However, spray application of dsRNA onto detached leaves cannot be compared with the situation in an intact leaf. Exogenously applied dsRNA first reached the apoplast, including the xylem (Fig 5), and subsequently translocated into the symplast by a yet unknown mechanism. Consistent with this, we could also demonstrate acropetal movement of the silencing signal that resulted in the inhibition of the fungus in distal not sprayed leaf areas (S3 Fig). Apoplastic movement of RNA has been proposed, e.g. to explain how maternally expressed siRNAs could be transferred from the endosperm of developing seeds into the symplastically isolated embryo [35].
Regardless of how target-specific inhibitory RNAs are applied-by transgene delivery (HIGS) or spray (SIGS)-the use of target-specific inhibitory RNAs to confer plant protection potentially is an alternative to conventional chemicals because they are i) highly specific and solely depending on their nucleotide sequence and ii) can be developed against an unlimited range of pathogens provided that the RNAi machinery is in place. Given the accumulation of dsRNA in the plant phloem (Fig 5), sucking insects also are realistic SIGS targets as their efficient control by HIGS has been largely demonstrated [36,37]. Certainly, many questions have to be addressed in the future to eventually judge the agronomical potential of SIGS, including the costs of RNA applications and their stability under field conditions. Hence, more research is required to develop application strategies, including improved uptake by compound design and use of chemical formulations. Another yet unassessed issue is the risk that microbial strains become insensitive to a commercial dsRNA product. Such scenario could probably be resolved by application of dsRNA mixtures that target different regions in one gene or even different genes. Moreover, a commercial dsRNA product should be designed not to have off-target effects in other organisms that might be relevant in the respective agroecosystem, including beneficial fungi and bacteria. In this respect, it is important to understand that both plants [12] and fungi [38] support the production of secondary siRNAs, meaning there is a potential for transitivity and amplification. It is therefore possible that low abundance inhibitory dsRNA sprayed onto the plant triggered a large silencing effect in Fusarium via these secondary RNAs. Importantly, when considering the regulatory issue of RNA-based plant protection it is crucial to emphasize that the principles of SIGS and HIGS rely on the mechanisms found for transkingdom communication in mutualistic and parasitic interactions, and thus is a natural phenomenon [12,13,14].
Apart from the dsRNAs prospects in future plant protection strategies, there is an additional technological potential in developing new pesticides. The simple phenotyping adopted by the SIGS screens renders them a powerful tool for genetic studies to assess compound targets with high efficiency and low costs. Thus, the present data provide essential scientific information on a fundamentally new plant protection strategy, thereby opening novel avenues for improving crop yields in an environmentally friendly and sustainable manner.

Plant and fungal materials
The spring barley (Hordeum vulgare) cultivar (cv.) Golden Promise was grown in a climate chamber under 16 h light photoperiod (240 μmol m -2 s -1 photon flux density) at 18°C/14°C (day/night) and 65% relative humidity. The wild type Fusarium graminearum strain Fg-IFA65 was described earlier [20]. Plates were incubated at room temperature under constant illumination from one near-UV tube (Phillips TLD 36 W/08) and one white-light tube (Phillips TLD 36 W/830HF). Fungal strains were cultured on synthetic nutrient-poor agar (SNA)-medium [39].
For generation of the DICER LIKE-1 Fg-IFA65 Δdcl-1 knock-out mutant, a homologous recombination strategy was used. The two homologous recombination segments USS and DSS (~1 kb each), representing promoter and termination regions of the FgDCL-1 gene, were selected based on the sequence information available at the Fusarium graminearum genome database (http://www.broadinstitute.org), and were PCR amplified. Primers used for USS (DCL_1_USS_KpnI_F and DCL_1_USS_KpnI_R), and DSS (DCL_1_DSS_HindIII_F and DCL_1_DSS_HindIII_R) are listed in S1 Table. The USS and DSS were cloned into flanking sites of the hph cassette of the pPK2 binary vector [38] using USER enzyme mix (New England Biolab, Inc., Ipswich, MA, USA) in Escherichia coli. The resultant plasmid was confirmed for proper orientation of cloned inserts in the vector by PCR conducted using USS/DSS and vector-specific primers and then by sequencing the PCR products.
The pPK2::ΔFgdcl-1 binary plasmid containing two Fgdcl-1 USS and DSS was transformed into Agrobacterium tumefaciens strain LBA4404 by electroporation, and transformants were analyzed by conducting restriction analysis. The ATMT of F. graminearum was based on a modified protocol [42]. Briefly, A. tumefaciens LBA4404 containing the pPK2::ΔFgdcl-1 plasmid was grown overnight in LB medium at 28°C (25 μg mL -1 kanamycin, 25 μg mL -1 rifampicin, and 5 μg mL -1 tetracycline). The next day, 10 ml of LB medium supplemented with above mentioned antibiotics and 200 μM acetosyringone was inoculated with 100 μl of the A. tumefaciens culture. This A. tumefaciens cell suspension with an OD 600 of 0.5 to 0.7 was mixed with F. graminearum conidial suspensions (10 5 −10 6 mL -1 ) in liquid SNA medium in equal proportions [1:1(v/v)]. Aliquots of 200 μl of the mixture were spread on black filter paper circles (Grade 551; Whatman Inc., Piscataway, NJ, USA), which were overlaid on SNA plates containing 200 μM acetosyringone and incubated for 2 days in the dark at RT until mycelial growth was observed on the filter paper. Transformants were selected on SNA medium supplemented with 50 μg mL -1 of hygromycin B (Sigma, St. Louis, MO, USA) and 300 μg mL -1 ticarcillin (Fisher Scientific, Pittsburgh, PA, USA).

dsRNA synthesis
The stacked clone (CYP51 B-A-C) [19] covering sequences of the three cytochrome P450 lanosterol C-14α-demethylase genes CYP51A (FGSG_04092), CYP51B (FGSG_01000), and CYP51C (FGSG_11024) from F. graminearum was used as template for the synthesis of a 791 nt long CYP3-dsRNA [19]. The pLH6000-Ubi::sGFP plasmid [43] was used as template for the synthesis of a 720 nt long GFP-dsRNA (S1 Fig). dsRNA was generated using MEGAscript RNAi Kit (Invitrogen) following MEGAscript protocols. Primer pairs T7_F and T7_R with T7 promoter sequence at the 5`end of both forward and reverse primers were designed for amplification of dsRNA (S1 Table).

Spray application
Detached leaf assay: Detached leaves of three-week-old barley plants were transferred into square Petri dishes (120 x 120 x 17mm) containing 1% agar. For spray application, dsRNA was diluted in 500 μL water to a final concentration of 20 ng μL -1 , corresponding to 10 μg dsRNA per plate. For sRNA application, the reaction mixture of DICER-cleaved dsRNA was used at a final concentration of 10 μg siRNA diluted in 500 μL -1 water per plate. Leaves were sprayed using a spray flask (10 mL capacity). Each dish containing six detached leaves was evenly sprayed. For the semi-systemic setup, detached leaves were covered before spraying with a plastic tray leaving only the upper part (approximately 1 cm) uncovered. After spraying, dishes were kept open until the surface of each leaf was dried (approximately 2 h). After an indicated lag time, leaves were drop-inoculated with 20 μL of 2 × 10 4 fungal conidia mL −1 . Closed dishes were incubated for 6 d at approximately 20°C on the lab bench. Alternatively, one-week-old barley seedlings were sprayed in the first leaf stage with 20 ng μL -1 dsRNA, and spray-inoculated three weeks later with 2 × 10 4 conidia mL −1 of Fg-IFA65. Inoculated plants were grown for three weeks in a growth chamber before evaluating the infection symptoms.

In vitro axenic cultures
Fg-IFA65 and Fg-IFA65 Δdcl-1 were cultured on synthetic nutrient SNA-medium. Plates were incubated at room temperature under constant illumination from one near-UV tube (Phillips TLD 36 W/08) and one white-light tube (Phillips TLD 36 W/830HF). Conidia were harvested from one-weak-old cultures with a sterile glass rod and sterile water [19]. CYP3-dsRNA was added to the fungal samples. Plates were incubated at room temperature. Gene expression studies were performed 24 h post CYP3-dsRNA treatment.

Quantification of fungal infection and transcript analysis
The relative amount of fungal DNA was measured using qPCR to quantify fungal infection. DNA was extracted using the CTAB method [44]. Expression analysis of the three fungal CYP51 genes as well as plant defense marker genes PR1 and JMT, respectively, was performed using qPCR. RNA extraction from infected leaves was performed with TRIzol (Invitrogen) following the manufacturer's instructions. Freshly extracted mRNA was used for cDNA synthesis using QuantiTect Reverse-Transcription kit (Qiagen). For qPCR 10 ng of cDNA was used as template in the Applied Biosystems 7500 FAST realtime PCR system. Amplifications were performed in 7.5 μL of SYBER green JumpStart Taq ReadyMix (Sigma-Aldrich) with 0.5 pmol oligonucleotides. Each sample had three repetitions.
To quantify the amount of fungal DNA, primers were used for assessing expression of the fungal β-tubulin gene (FGSG_09530) with reference to barley ubiquitin gene (S1 Table). Primers were used for assessing expression of target CYP51 genes with reference to β-tubulin gene (S1 Table). After an initial activation step at 95°C for 5 min, 40 cycles (95°C for 30 sec, 57°C for 30 sec, 72°C for 30 sec) were performed. Ct values were determined with the 7500 Fast software supplied with the instrument. Transcript levels of β -tubulin gene were determined via the 2-Δ Δ Ct method [45] by normalizing the amount of target transcript to the amount of reference transcript.
Small RNA library production and sequence analysis

Northern blot analysis
For Northern blot analysis, 8 ng of total RNA from local region and 80 ng of systemic region or negative control (TE-mock) and 10 pg of in vitro transcribed CYP3-dsRNA was loaded onto a 6% denaturing polyacrylamide gel with DNA Molecular Weight Marker VIII (Roche), transferred to a nylon membrane. CYP3-dsRNA and U2 snRNA were detected using the DIG Labeling and Detection System (Roche) following the manufacturer's instructions. Chemiluminescence was detected using X-ray films.
The CYP3-dsRNA probe was created by PCR with CYP3-dsRNA forward and reverse primer (S1 Table) on the stacked clone (CYP51 B-A-C) [19] using PCR DIG Labeling Mix (Roche). U2 snRNA loading control was amplified from cDNA created from total RNA using qScript Flex cDNA Kit (Quanta BioSciences) and primers U2 forward (TACCTTTCTCGGC CTTTTGG and U2 reverse (CAGCAGCAAGCTAC TGTGGT). Gel purified probes were hybridized in NorthernMax Prehybridization/Hybridization Buffer (Ambion) at 45°C over night.

Confocal microscopy of fluorophore distribution
Twenty-four h after spraying fluorescing dsRNA were imaged using a Leica TCS SP2 (Leica Microsystems, Wetzlar, Germany) equipped with a 75-mW argon/krypton laser (Omnichrome, Chino, CA) and a water immersion objective (HCX APO L40x0.80 W U-V-l objective). Fluorescing dsRNA were imaged using a LSM 880 (Zeiss Microscopy GmbH, Jena, Germany) with the 488 nm laser line of an argon multiline laser (11.5 mW) and a HeNe 594 nm (1.3 mW) laser. Images were taken with a 20x objective (Plan-Apochromat 20x/0.8). Lambda stacks were created using the 32 channel GaAsP detector. Reference spectra with each pure fluorescence dye were recorded. The sample was inspected in Online Fingerprinting mode. Specific areas of the sample were imaged in lambda mode followed by Linear Unmixing with ZEN software (Zeiss, Jena, Germany). Fluorescent labeling of the dsRNA was performed using the Atto 488 RNA Labeling Kit (Jena Bioscience, Jena, Germany) following the manufacturer's instructions. Leaves were sprayed with the labeled dsRNA and 24 h later drop-inoculated with 2 × 10 4 Fg-IFA65 conidia mL −1 . To assess whether dsRNA has an effect on fungal morphology, leaves were inoculated with Fg-IFA65 GFP and infected leaves were analyzed at 6 dpi.
For observation of phloem tissue, cortical cell layers were removed down to the phloem from the lower side of the main vein of a mature leaf. The leaf surface and longitudinal-as well as cross sections were stained with 4.3 μM of the membrane dye RH-414 (-N-(3-triethylammoniumpropyl)-4-(4-(4-(diethylamino)phenyl)butadienyl)pyridiniumdibromid) and/or with 5 μg mL -1 of the fungal hyphae dye wheat germ agglutinin (WGA) Alexa Fluor 1 594 conjugate (Invitrogen) for at least 10 min.
RH-414, WGA Alexa Fluor 594, and the autofluorescence of cell walls and chloroplasts were excited by the 564-nm line of the argon/krypton laser, while GFP and ATTO 488 were excited with the 488-nm line. For observation at the 590 nm and 510 nm wave lengths, respectively, a long pass filter was used. Digital images were processed with Adobe Photoshop to optimize brightness, contrast, and color and to enable an overlay of the photomicrographs.

Statistical analysis
Analyses were performed in SigmaPlot 12 (Systat Software) using Student´s t-tests after data were tested for normality distribution (Shapiro-Wilk test). Both the fungal dicer-like-1 mutant Fg-IFA65 Δdcl-1 and the wt strain (Fg-IFA65) were inhibited by CYP3-dsRNA as compared to TE treatment. Photographs were taken at 6 dpi. (C) Quantification of fungal DNA by qPCR analysis confirmed the macroscopic analysis. Bars represent mean values ±SDs of two independent sample collections. The reduction of fungal growth in samples treated with CYP3-dsRNA compared with mock-treated TE controls was statistically significant ( Ã P < 0.05, ÃÃ P < 0.01; Student´s t test). (TIF) S1