Targeted Delivery of Gene Silencing in Fungi Using Genetically Engineered Bacteria

Abstract

Exploiting RNA interference (RNAi) in disease control through non-transformative methods that overcome the hurdle of producing transgenic plants has attracted much attention over the last years. Here, we explored such a method and used non-pathogenic bacteria as a versatile system for delivering RNAi to fungi. Specifically, the RNaseIII-null mutant strain of Escherichia coli HT115(DE3) was transformed with two plasmid vectors that enabled the constitutive or IPTG-inducible production of double-stranded RNAs (dsRNAs) against genes involved in aflatoxins production in Aspergillus flavus (AflC) or virulence of Botrytis cinerea (BcSAS1). To facilitate the release of the dsRNAs, the bacterial cells were further genetically engineered to undergo a bacteriophage endolysin R-mediated autolysis, following a freeze-thaw cycle. Exposure under in vitro conditions of A. flavus or B. cinerea to living bacteria or their whole-cell autolysates induced silencing of AflC and BcSAS1 in a bacteria concentration-dependent manner, and instigated a reduction in aflatoxins production and mycelial growth, respectively. In planta applications of the living bacteria or their crude whole-cell autolysates produced similar results, thus creating a basis for translational research. These results demonstrate that bacteria can produce biologically active dsRNA against target genes in fungi and that bacteria-mediated RNAi can be used to control fungal pathogens.

1. Introduction

RNA interference (RNAi) is a conserved mechanism of gene-silencing in nature with immense therapeutic potential. It is triggered by inverse transcripts that arecomplementary to the mRNA of a target gene, including double-stranded RNAs (dsRNAs) and small-interfering RNAs (siRNAs) that are produced by the DICER-mediated cleavage of dsRNAs into small pieces of 19–25 nucleotides [1]. Since its discovery, RNAi has been transformative to the study of gene function in numerous eukaryotic organisms and has created immense opportunities for its exploitation in medicine and disease control [2,3]. RNAi therapeutics is currently a rapidly evolving biotechnological field and despite the many challenges encountered along the way, 2018 marked a new era with the approval by the US Food and Drug Administration (FDA) of the first RNAi-based drug [2,4]. At present, several RNAi-based drugs that target diverse and rare human diseases are currently under development or undergoing phase I or phase II clinical trials. In addition, much effort is placed on the concomitant development of delivery strategies to target cells and tissues, which remains a key issue in the exploitation of RNAi therapeutics [2,5,6].

In crop protection, RNAi has mostly been utilized in the form of host-induced gene silencing (HIGS), an approach in which plants are genetically engineered to produce dsRNAs or siRNAs against genes in target pests and pathogens [7]. Several studies have highlighted the utility of HIGS in a variety of crops for the control of a wide-range of insect pests, nematodes, viruses, and fungal plant pathogens [7,8,9,10]. HIGS has also proven to be an effective method for reducing mycotoxin contamination in food and feed, including of the notorious aflatoxins [8,11,12,13,14,15]. Aflatoxins are secondary metabolites produced mainly by species of Aspergillus spp. and particularly by A. flavus and A. parasiticus [16,17]. At least 20 different aflatoxins have been described but aflatoxin B1 (AFB1), B2 (AFB2), G1 (AFG1) and G2 (AFG2) are the four most common types, with AFB1 and AFB2 mainly produced by A. flavus, and AFG1 and AFG2 by A. parasiticus [16,17]. They are highly toxic mycotoxins, well-known for their carcinogenicity and other serious illnesses related to liver damage and immunosuppression that they can cause in humans and livestock [18]. They enter the food chain mainly through contaminated agricultural commodities, with grain and nut crops such as maize, rice, and peanuts being major sources of contamination [17]. Estimates by the Food and Agriculture Organization (FAO) are that over 25% of the global food supply is contaminated by mycotoxins, including aflatoxins, whereas according to the World Health Organization (WHO), approximately 4.5 billion people are chronically exposed through their diet to these poisonous compounds, with often detrimental effects [16,17]. Thus, finding effective ways to mitigate contamination of food and feed by aflatoxins is crucial.

Despite the potential of HIGS for disease control, off-target effects, difficulties in plant transformation, and low public acceptance of transgene technology have restricted its application in the field [7,15,19,20]. Virus-induced gene silencing (VIGS), which exploits recombinant viruses as vectors to deliver RNAi molecules to plants, is an alternative to HIGS but to date it has mostly been utilized as a functional tool to study genes of interest in different crops, rather than for disease control [21,22,23]. The relatively recent discovery that externally applied dsRNAs and siRNAs can be taken-up by fungal cells from their environment and induce silencing in the transfected cells, has attracted much attention for its potential use as an alternative to HIGS for delivering RNAi to target pathogens [24]. Thus, spray-induced gene silencing (SIGS) has recently emerged as a novel, environmentally-friendly disease control method. At present, several attempts to generate RNAi-based biopesticides that utilize synthetic dsRNAs or siRNAs against genes in target pests and pathogens are underway [25,26]. Such RNA-based biopesticides can be customized to be species-specific or to target a broad range of pathogens, even across different taxonomic groups. Moreover, multiple genes can be simultaneously silenced, thus reducing the risk for resistance development [25,26,27]. However, a major limitation to the use of synthetic RNAi molecules as biopesticides is the cost of their large-scale production and formulation as well as the limited stability of the naked RNAi molecules under field conditions [26,27]. To overcome these limitations, a variety of synthetic materials have been developed to encapsulate and protect dsRNA from nuclease degradation. These include a double-layered hydroxide clay nanosheet that has been proved effective to protect dsRNA targeted against some viral diseases, or carbon nanotubes that successfully deliver siRNA and trigger posttranscriptional gene silencing in plant cells. Nevertheless, the use of these synthetic materials is still limited [3,28].

Here, we explored the possibility of using non-pathogenic bacteria as a versatile production and delivery system of RNAi to fungi. Bacteria-mediated delivery of RNAi has been successfully used to silence genes in Caenorhabditis elegans [29], insects [27], shrimps [30], and mammalian cells [31] but the utility of this approach for the control of fungi, to our knowledge, has not been exploited yet [26]. In our approach, a mutant of Escherichia coli lacking RNaseIII (an enzyme that degrades dsRNAs in bacterial cells) [29] was genetically engineered to express dsRNAs against A. flavus [32,33], and Botrytis cinerea, a notorious plant pathogen with a remarkably broad host range [34], and to further undergo a programmable autolysis [35], thus allowing the destruction of the bacterial cells and the release of the fungicidal dsRNAs. Our data show that the engineered bacteria accumulated high levels of dsRNAs, whereas subsequent treatment under in vitro or in planta conditions of A. flavus and B. cinerea with the living bacteria or the induced bacterial autolysates resulted in silencing of the complementary fungal genes and a reduction in production of aflatoxins or fungal growth, respectively.

2. Materials and Methods

2.1. Plasmid Construction

Plasmids pAGM4723-AflC₂₅₀ and pAGM4723-eGFP₃₅₀ (Figure S1) were constructed using modular cloning with the Golden Gate gene assembly [36,37]. Briefly, sense and antisense strands of AflC were PCR-amplified from the cDNA of this gene with primers pairs AflC-F/AflC-R and AflCr-F/AflCr-R, respectively (Table S1). The pKan^R promoter and rrnB-T1 terminator were obtained from plasmid pPROBE-gfp[tagless] [38] by PCR amplification with primer pairs Ppntii-F/Ppntii-R and rrnB_T1-term-F/rrnB_T1-term-R, respectively (Table S1). Sense and antisense strands of AflC were each cloned into a pICH41331 (Addgene MoClo Toolkit, Watertown, MA, USA) [36,37], in addition to pKan^R and rrnB-T1, by the mixture of BpiI (Thermo Fisher Scientific, Waltham, MA, USA) and T4 DNA ligase (New England Biolabs, Ipswich, MA, USA). The assemblies were done in a way that pKan^R and rrnB-T1 flanked the AflC₂₅₀ fragments. Subsequently, the promoter-sense AflC₂₅₀-terminator and promoter-antisense AflC₂₅₀-terminator assemblies were cloned into pICH47732 and pICH47742 (Addgene MoClo Toolkit, Watertown, MA, USA) [36,37], respectively, using BsaI instead of BpiI in the mixture. The endolysin R gene containing the pbla promoter and rrnB-T1 terminator was PCR-amplified from plasmid pAD-LyseR (Addgene Plasmid #99244) [35] with the primer pair EndoR_F/EndoR_R (Table S1) and cloned into pICH47751 (Addgene MoClo Toolkit, Watertown, MA, USA) [36,37] by BsaI and T4 DNA ligase. Finally, the AflC₂₅₀ dsRNA synthesis cassette and endolysin R gene were assembled and cloned into pAGM4723 (Addgene MoClo Toolkit, Watertown, MA, USA) [36,37], using BpiI. For cloning pAGM4723-eGFP₃₅₀, sense and antisense strands of eGFP were PCR-amplified from pICH41531 with primer pairs EGFP342-F/EGFP342-R and EGFP342-F-reversed/EGFP342-R-reversed, respectively (Table S1). pAGM4723-eGFP₃₅₀ was constructed by using the procedure described above.

Plasmids pAGM4723-BcSAS1₂₅₀ and pAGM4723-eGFP₂₅₀ (Figure S1) were constructed by adding a T7 promoter (5′–TAATACGACTCACTATAG–3′) in the forward and reverse primers in opposite directions. BcSAS1₂₅₀ was PCR-amplified from the cDNA of B. cinerea with the primer pair BcSAS1_T7_F/BcSAS1_T7_R (Table S1) and cloned into pICH47732 using BsaI and T4 DNA ligase in the reaction of the Golden Gate assembly. The BcSAS1₂₅₀ dsRNA synthesis cassette and endolysin R gene were assembled into pAGM4723 with BpiI and T4 DNA ligase. EGFP₂₅₀ was amplified from pAGM4723-eGFP₃₅₀ with primer pair G250_T7_F/G250_T7_R (Table S1) and subsequently following the same procedure described above to clone it into pAGM4723.

2.2. Bacterial and Fungal Strains Used in This Study, Their Growth Conditions, and In Vitro Bioassays

The bacterial delivery system was based on the RNaseIII-deficient E. coli strain HT115(DE3). The genotype of HT115 is E. coli [F-, mcrA, mcrB, IN(rrnD-rrnE)1, rnc14::Tn10(DE3 lysogen: lacUV5 promoter-T7 polymerase] [29]. This strain originates from the W3110 strain, a commonly used wild-type of the non-pathogenic E. coli K-12 strain. The HT115(DE3) strain was selected for its ability to produce stable dsRNA and to express genes placed under the control of promoters recognized by the IPTG-inducible T7 RNA polymerase. The fungal strains analyzed were A. flavus NRRL3357 and B. cinerea B05.10. Fungal cultures were established from frozen mycelia stored in 25% glycerol (v/v) at −80 °C and maintained on potato dextrose agar (Difco Laboratories, Detroit, NI, USA) (PDA) plates.

E. coli HT115(DE3) was incubated in Luria-Bertani (LB) liquid medium [10 g Bacto^TM Tryptone (Difco Laboratories, Detroit, NI, USA), 10 g NaCl, 5 g Bacto^TM Yeast Extract (Difco Laboratories, Detroit, NI, USA), 1 L H₂O, adjusted to pH 7.0] supplied with tetracycline (10 μg/mL) and kanamycin (50 μg/mL) at 37 °C at 300 rpm. The bacterial transformants HT115/AflC_Con, HT115/eGFP_Con, HT115/BcSAS1_Ind and HT115/eGFP_Ind were generated by the heat shock method in the presence of the plasmids pAGM4723-AflC₂₅₀, pAGM4723-eGFP₃₅₀, pAGM4723-BcSAS1₂₅₀, and pAGM4723-eGFP₂₅₀, respectively. Transformants were always grown in the presence of kanamycin (50 μg/mL) and when cultures reached an optical density at 600 nm (OD₆₀₀) of approximately 0.8, then isopropyl-β-D-1-thiogalactopyranoside (IPTG) (Sigma-Aldrich, St. Louis, MO, USA) was added at 1.0 mM final concentration to the cultures of HT115/BcSAS1_Ind and HT115/eGFP_Ind to induce dsRNA production when needed. Bacterial concentration was measured with the Spectronic Genesys 6 spectrophotometer (Thermo Electron Scientific Instruments LLC, Madison, WI, USA) as optical density at 600 nm (OD₆₀₀). Bacterial cultures were incubated until they reached the stationary phase, which was experimentally determined to be at an OD₆₀₀ of 1.2–1.3 (Figure S2). The volumes corresponding to a final OD₆₀₀ of 1, 0.5, 0.1, and 0.01 in 10 mL cultures were concentrated by centrifugation (3000× g, 5 min) and re-suspended in molecular grade water (G-Biosciences, St. Louis, MO, USA) to a final volume of 0.5 mL. To induce autolysis, the bacterial cells were subjected to a freeze-thaw cycle consisting of 15–20 min at −80 °C, followed by thawing for 60–90 min at 37 °C and subsequently treating the cell suspensions with the antibiotics ampicillin (100 μg/mL final concentration) and streptomycin (50 μg/mL) for 1 h, to assure that no living cells were present. Growth curves for the HT115/AflC_Con, HT115/eGFP_Con, HT115/BcSAS1_Ind, and HT115/eGFP_Ind strains as well as the HT115(DE3) strain (Figure S2), were calculated by measuring the OD₆₀₀ with the Genesys 6 spectrophotometer (Thermospectronic) every hour up to 9 h of growth, and then at 13 h and 24 h. All strains were grown in LB medium supplemented with tetracycline (10 μg/mL) and kanamycin (50 μg/mL). Kanamycin was not added to the HT115(DE3) culture, whereas when reaching OD₆₀₀ ≈ 0.8, strains HT115/BcSAS1_Ind and HT115/eGFP_Ind were induced with 1.0 mM IPTG. In each of the three experiments performed, the cultures were grown at 37 °C in 20 mL volumes with 250 rpm of shaking, with three replicates per strain.

To obtain fungal spores, A. flavus NRRL3357 was grown on PDA medium with continuous light at 37 °C for 3–5 days (non-aflatoxin production conditions), while B. cinerea was grown on PDA medium at 25 °C for 10–14 days. Spores were collected by scrapping mycelia with 2 mL of milli-Q water and filtrating them through two layers of cheesecloth. Spore concentration was measured with a hemocytometer. The spore suspension was subsequently added to 5 mL of YES liquid medium [2% Bacto^TM Yeast Extract (Difco Laboratories, Detroit, NI, USA), 6% sucrose, pH 5.8] to a final concentration of 10⁴ spores/mL. The spore suspension was then incubated at 37 °C (A. flavus) or 25 °C (B. cinerea), 250 rpm with continuous light. Once fungal spores had germinated (checked by microscopy at 6–8 h), the bacterial treatments were added to the fungal culture flasks. These bacterial treatments were either the living bacterial cells or their whole-cell autolysates produced in 10 mL and concentrated in 0.5 mL, as described above. The co-incubation conditions for both fungi were 25 °C, 250 rpm, in darkness. Fungal samples were harvested 12 h, 24 h, and 36 h after the bacterial treatment applications.

B. cinerea dry weight was measured after filtering the fungal culture with a vacuum filtration system using a 0.45 μm pore size nitrocellulose membrane (Thermo Fisher Scientific, Waltham, MA, USA). Once samples were completely dry, they were fully removed from the membrane with the help of a spatula and measured using a precision mg scale (Mettler Toledo, ML203E/03). Culture filtering by vacuum filtration, performed as described above, was also used to harvest the supernatant of A. flavus cultures, of which 1 mL was used to measure total content of aflatoxins by Enzyme-Linked Immunosorbent Assay (ELISA). Dried fungal mycelium was frozen in liquid nitrogen and stored at −80 °C before total RNA extraction with the TRizol^TM reagent (Invitrogen, Carlsbad, CA, USA).

2.3. Total RNA Extractions and RT-qPCR

Total RNA was extracted from A. flavus, B. cinerea, the inoculated plant tissues, and the bacterial strains using the TRIzol^TM reagent (Invitrogen, Carlsbad, CA, USA), according to the manufacturer’s instructions. Briefly, 1 mL of TRIzol^TM reagent (Invitrogen, Carlsbad, CA, USA) was added to 50–100 mg of ground plant tissue, fungal, or bacterial cells, and incubated for 5 min at room temperature. Then, 0.2 mL of chloroform was added to the samples and incubated for 3 min at room temperature, followed by centrifugation for 15 min at 12,000× g at 4 °C. The nucleic acids in the supernatant were precipitated by adding 500 μL of isopropyl alcohol (1:1) and centrifuging at 12,000× g at 4 °C for 10 min. Pellets were washed with 75% ethanol/water (v/v), re-suspended with DEPC-treated water (Invitrogen, Carlsbad, CA, USA), and treated with DNaseI according to the manufacturer’s instructions (Thermo Fisher Scientific, Waltham, MA, USA). Following total RNA extraction, samples were tested for integrity in 1.2% agarose gels and quantified using a Nanodrop Spectrophotometer (Nanodrop 2000, Thermo Fisher Scientific, Waltham, MA, USA). The absence of DNA contamination was examined in the samples by the lack of PCR amplification of the glyceraldehyde 3-phosphate dehydrogenase gene (GADPH) gene in A. flavus and of the BcActin gene in B. cinerea. cDNA was synthesized from purified total RNA using iScript^TM cDNA Synthesis Kit (Bio-Rad Laboratories, Hercules, CA, USA), following the manufacturer’s instructions. The SYBR Green kit SsoFast^TM EvaGreen^® (Bio-Rad Laboratories, Hercules, CA, USA) was used to carry out RT-qPCR reactions in a CFX96 Touch thermal cycler (Bio-Rad laboratories, Hercules, CA, USA) equipped with the CFX Maestro^TM software. The thermal profile was as recommended by the manufacturer (40 cycles of 98 °C for 3 s and 60 °C for 5 s), testing single amplification by a specific peak in the dissociation curve (0.5 °C increments every 10 s within a range of 65–95 °C). The GADPH gene, amplified with the primer pair qGADPH-F/qGADPH-R (Table S1), and the BcActin gene, amplified with the primer pair qBcActin-F/qBcActin-R (Table S1), were used as endogenous reference genes for A. flavus and B. cinerea, respectively. The primer pair qAflC-F/qAflC-R (Table S1) was used for AflC amplification, while primer pair qBcSAS1-F/qBcSAS1-R (Table S1) was used for BcSAS1. The AflC and BcSAS1 sequences of the amplicons generated by RT-qPCR do not match with the dsRNA sequence used in the treatments, to avoid artifacts. The relative expression level of each gene was calculated by the 2^−ΔΔCt method [39].

2.4. dsRNA Isolation from E. coli HT115(DE3)

Total RNA was extracted from E. coli HT115(DE3) cultures using the TRIzol^TM reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions and as described above. Then, dsRNAs were isolated from total RNA extraction, using one of the following two methods. In the first method, 20 μg of total RNA was treated at 37 °C for 45 min with 3 μL of DNaseI (Thermo Fisher Scientific, Waltham, MA, USA) and 1 μL of RNase A (Thermo Fisher Scientific, Waltham, MA, USA) in a 50 μL reaction volume containing 0.35 mM NaCl and 5 μL of 10 × DNase I buffer. dsRNAs were further purified with the RNA Clean & Concentrator Kit (ZYMO Research, Irvine, CA, USA) or the mirVana^TM miRNA isolation kit (Thermo Fisher Scientific, Waltham, MA, USA), following the specifications for <200 nucleotides RNA removal. In the second method, the MEGAscript^TM T7 Transcription Kit (Thermo Fisher Scientific, Waltham, MA, USA) was used from the step of nuclease digestion to remove DNA and ssRNAs from the filter cartridge provided with the kit, followed by purification of dsRNAs from the filter cartridge, which is able to remove free nucleotides and <200 nucleotides nucleic acid degradation products. Finally, dsRNAs were visualized by electrophoresis and their concentration was determined with a Nanodrop Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA).

To examine dsRNA stability over time in the bacterial autolysates and in living bacterial cultures, strain HT115/BcSAS1_Ind was grown to the stationary phase (OD₆₀₀ ≈ 1.2–1.3) in LB medium supplemented with tetracycline (10 μg/mL) and kanamycin (50 μg/mL). When the culture had reached an OD₆₀₀ ≈ 0.8, IPTG was added to a final concentration of 1.0 mM. The culture was subsequently diluted to an OD₆₀₀ of 1.0 in 10 mL volumes. After that, the 10 mL cultures were pelleted by centrifugation at 3000× g for 5 min and re-suspended in 0.5 mL of molecular grade water (G-Biosciences, St. Louis, MO, USA). The bacterial autolysates were produced as described above. Living bacterial cells were produced without the freeze-thaw cycle and the antibiotic treatment. Subsequently, the autolysates and the living bacterial cells were incubated at room temperature (≈ 25 °C) for 0 h, 1 h, 6 h, 24 h, 36 h, and 72 h, after which the debris was removed by centrifugation (21,000× g, 30 min, at 4 °C) and total nucleic acids were extracted from the supernatants with the TRIzol^TM method as described above. The nucleic acids were then treated with RNaseA and DNase I, and cleaned with the RNA Clean & Concentrator™ Kit, as described above. Finally, aliquots of 0.5 μg of RNA were loaded onto 2% agarose gels for electrophoresis.

2.5. Bioassays

Fresh maize seeds were surface sterilized as described previously [40]. Briefly, the entire seeds were treated with a 75% ethanol/water (v/v) solution for 30 s at room temperature, then rinsed with a milli-Q autoclaved water. Next, they were treated with a 2% hypochlorite solution for 5 min at room temperature and rinsed at least five times with milli-Q autoclaved water. Aliquots of 2 μL of a 50 × 10⁶ spore/mL suspension (milli-Q autoclaved water) of A. flavus were placed on a maize embryo and were incubated at 37 °C with light for 2 days (non-aflatoxin production conditions). After this time period, 10 μL of living or lysed bacteria cells obtained from the OD₆₀₀ 1.0 culture were placed at the same point of fungal inoculation and incubated for one more day at aflatoxin-production conditions (darkness at 25 °C). Humidity was maintained to over 80% by placing the maize seeds into boxes layered with sterile Whatman filter paper 1/5, which was hydrated with 1 mL of sterile water every 24 h. Subsequently, the maize embryos with the fungal mycelium were harvested for further analysis with RT-qPCR (40 cycles of 98 °C for 3 s and 60 °C for 5 s) and ELISA.

Nicotiana benthamiana leaves were inoculated with B. cinerea, as described previously [41]. Briefly, B. cinerea spores were collected from PDA plates incubated for 10–14 days, as described above. Spores were purified and re-suspended at a concentration of 10⁴ spores/mL in grape juice (1/2 strength). A drop of 4 μL of this spore suspension was inoculated onto both sides of detached 4–6 week-old N. benthamiana leaves, which were placed onto 1.5% (m/v) phytoagar (Thomas Scientific, Swedesboror, NJ, USA) medium trays. After 6 h, 10 μL of living or lysed bacteria cells obtained from the OD₆₀₀ 1.0 culture were placed at the same point of fungal inoculation, on the left side of the leaf blade for HT115/eGFP_Ind and the right side of the leaf blade for HT115/BcSAS1_Ind. Suspensions of living bacteria cells were applied with IPTG at a final concentration of 0.1 mM. Inoculated N. benthamiana leaves were incubated for 3 days. The relative lesion size between treatments in each leaf was determined by measuring the necrotic area using the ImageJ software. Then, they were collected for RT-qPCR analysis (40 cycles of 98 °C for 3 s and 60 °C for 5 s).

2.6. Analysis of Total Aflatoxins by ELISA

The total concentration of aflatoxins on E. coli-A. flavus co-cultures and maize seeds infected with A. flavus was analyzed by ELISA. The AgraQuant^® Total Aflatoxin Test Kit (4–40 ppb) (Romer labs^®, Tulln, Austria) was used for the ELISA analysis according to the manufacturer’s instructions. Briefly, aflatoxins were extracted with 70% (v/v) methanol from 1 mL of fungal culture supernatants or grinded inoculated maize seeds. Because the concentration of aflatoxins in the samples was unknown and it could fall outside of the range of the solutions of known aflatoxin concentration provided by the manufacturer (i.e., the standards), serial dilutions of the experimental samples were prepared (1:0, 1:10, 1:100, and 1:1000) in 70% (v/v) methanol, so that one of the serial dilutions would fall within the aflatoxin concentration range covered by the standards provided with the manufacture’s kit (0, 4, 10, 20, 30 and 40 ppb). The OD of each sample dilution at 450 nm (OD₄₅₀) was measured at the end of the ELISA reactions and the sample dilution whose OD₄₅₀ measurement was in the range of the manufacturer’s standards was selected. Subsequently, aliquots of 100 μL of the selected diluted samples and the standards in duplicate were added to 200 μL of the conjugate. After mixing, 100 μL were transferred to antibody-coated wells and incubated for 15 min at room temperature. The unbound conjugate was removed with deionized water during the washing step (5×). Once wells were dry, 100 μL of substrate were added into each well, incubated for 5 min, and then 100 μL of the stop solution were added. Microtiter plates were measured in an ELISA reader (Tristar LB941, Berthold Technologies, Bad Wildbad, Germany) at 450 nm. The concentration of aflatoxins in the samples was calculated using the formula of the calibration curve from the standards and multiplied by the specific dilution factor of the sample.

2.7. Statistical Analyses

Results represent the mean and the standard deviations of at least three biological replicates, which are indicated by error bars. Significant differences were determined using an unpaired t-test to compare the means between two groups, and a one-way ANOVA followed by Tukey’s Honestly Significant Difference (HSD) test to compare the means between three or more groups. The one-sample t-test was used to test differences between the relative aflatoxins production or fungal lesion size of each biological treatment with respect to the control treatment (designated as 1). ^o p < 0.10; * p < 0.05; ** p < 0.01, *** p < 0.001. Statistical analyses were performed using Statistix 10.0 software package (Analytical Software, Tallahassee, FL, USA).

3. Results

3.1. Bacteria Can Be Genetically Engineered to Produce dsRNAs against Target Genes in Fungi

For our studies, we have utilized the well-characterized E. coli HT115(DE3) strain, which lacks the rnc allele that encodes for the dsRNA-specific endonuclease RNaseIII, and is thus capable of synthesizing and preserving dsRNAs. The strain is also lysogenic for λDE3, thus enabling the transcription of genes placed under the control of the bacteriophage T7 promoter [29]. HT115(DE3) was genetically engineered to express dsRNAs against AflC in A. flavus [42] and BcSAS1 in B. cinerea [43]. These two pathogens were selected due to their importance as a producer of aflatoxins and a plant pathogen with a broad host range, respectively [32,34]. In addition, silencing following uptake of dsRNAs by their mycelia has been demonstrated for both fungi [12,13,15,24,40]. Likewise, AflC and BcSAS1 were selected as gene targets for RNAi because of their critical role in production of aflatoxins in A. flavus and virulence in B. cinerea, respectively. Specifically, AflC encodes a Type I polyketide synthase (pksA) needed for the synthesis of the backbone structure of aflatoxins in A. flavus, whereas BcSAS1 encodes for a member of the Rab family of small GTPases that is required for mycelial development and virulence in B. cinerea [42,43].

To further explore the versatility of the system, dsRNAs against AflC and BcSAS1 were produced in HT115(DE3) using two plasmid vectors that allowed the production of dsRNAs via different transcriptional configurations, i.e., (i) through the constitutive in trans (i.e., from separate plasmid loci) transcription of the sense and antisense strands of the selected target DNA template (Figure 1A) and (ii) through the IPTG-inducible in cis (i.e., from the same plasmid locus) bi-directional transcription of the DNA template (Figure 1B). Specifically, in the first configuration that was used to produce dsRNAs against AflC, a 250 bp cDNA fragment was cloned in opposite directions between the native promoter of the kanamycin resistance gene (pKan^R) and the transcriptional terminator T1 of the E. coli rrnB gene (rrnB-T1). A plasmid construct similar to this one but in which the AflC fragment was replaced by a 350 bp cDNA fragment of the enhanced green fluorescent protein (eGFP) was also prepared and used in control applications (Figure 1A and Figure S1A). In the second configuration that was used to produce dsRNAs against BcSAS1, two T7 promoters (pT7) were placed in an inverted orientation at the 5′ and 3′-end of a 250 bp cDNA fragment of the gene (Figure 1B). As for AflC, a plasmid construct in which the BcSAS1 fragment was replaced by a 250 bp eGFP cDNA fragment was also prepared and used as control (Figure 1B and Figure S1B). Following transformation into HT115(DE3) of the plasmid constructs and transcription, the first configuration produces in trans complementary sense and antisense ssRNA molecules from the DNA templates, which subsequently hybridize to form dsRNAs (Figure 1C). By contrast, in the second configuration sense/antisense transcript pairs are spontaneously formed in cis, thus creating dsRNAs (Figure 1C). Moreover, while in the first configuration the transcription of the target sequences is controlled by the constitutive pKan^R promoter, in the second configuration it is dependent on the induction of the T7 gene 1 in HT115(DE3) by IPTG and subsequent activation of the T7 promoter by the T7 RNA polymerase.

As both trans-kingdom transfer and environmental uptake of RNAi molecules by mycelial cells have been demonstrated to induce RNAi in fungi, in order to combine the benefits of the two approaches, we further genetically engineered the HT115(DE3) strain to allow the release of the dsRNA cargo through a programmable cellular autolysis step. For this purpose, a phage lambda endolysin gene (λR), placed under the control of the weakly constitutive pbla promoter of the β-lactamase gene and the rrnB-T1 terminator was cloned in the two plasmid vectors used for the production of the dsRNAs (Figure 1A,B and Figure S1A,B). When the inner bacterial cell membrane is intact, the endolysin protein accumulates in the cell without significantly impacting cell growth. However, a simple freeze-thaw cycle opens the inner membrane, thus enabling the endolysin to disrupt the bacterial cell wall and trigger autolysis [35]. Notably, although cell lysis results in the release of several lytic enzymes, the autolysis of the bacterial cell did not trigger a parallel degradation of the dsRNAs, which remained stable in the cell lysates up to 24 h, and partially stable up to 72 h of incubation at 25 °C. If the cells are not lysed, then dsRNAs remain stable in the living cells for at least for 72 h (Figure 1D). This indicates that both living bacteria and their lysates can be used to deliver dsRNAs against fungi.

3.2. Bacterially-Produced dsRNAs Induce RNAi in Fungi in a Bacteria Concentration-Dependent Manner

We initially sought to determine whether the HT115(DE3) bacterial strains that were genetically engineered to produce dsRNAs against AflC [HT115(DE3)(pAGM4723-AflC₂₅₀)], hereafter abbreviated as HT115/AflC_Con, or BcSAS1 [HT115(DE3)(pAGM4723-BcSAS1₂₅₀)], hereafter abbreviated as HT115/BcSAS1_Ind, were able to induce silencing of the targeted transcripts in A. flavus and B. cinerea, respectively. Two approaches for delivering the bacterially-produced dsRNAs against A. flavus and B. cinerea were utilized, i.e., trans-kingdom mediated transfer, facilitated by co-incubating living bacteria with fungal germlings (hereafter referred to as Treatment A), and environmental uptake from bacterial whole-cell autolysates (hereafter referred to as Treatment B).

To examine whether the physical interaction between fungi and bacteria can facilitate the trans-kingdom transfer of the bacterially-produced dsRNAs to fungi (Treatment A), 10 mL bacterial cultures with a final OD₆₀₀ of 1, 0.5, 0.1, and 0.01 that were prepared from serially-diluted cultures at stationary phase (OD₆₀₀ ≈ 1.2–1.3) (Figure S2), were concentrated to a volume of 0.5 mL and mixed with 5 mL of fungal germlings (10⁴ germlings/mL) in liquid cultures. Assessment of RNAi-induced gene silencing at 12 h of co-incubation between A. flavus and the HT115/AflC_Con strain, or between B. cinerea and the HT115/BcSAS1_Ind strain (hereafter referred to as test applications), showed a significant reduction in AflC and BcSAS1 transcript levels compared with the expression levels of these two genes when the two fungi were co-incubated with the HT115/eGFP_Con and HT115/eGFP_Ind strains, respectively (hereafter referred to as control applications) (Figure 2A,B, and Table S2). Overall, when germlings of A. flavus or B. cinerea were co-incubated with the bacterial concentrates obtained from the OD₆₀₀ 1.0, 0.5, and 0.1 cultures, a 21.7-fold, a 5.5-fold, and a 3.2-fold reduction in AflC expression, and a 23.3-fold, a 15.9-fold, and a 2.7-fold reduction in BcSAS1 expression were observed in the test applications relative to the control ones, respectively (Figure 2A,B and Table S2). These reductions in AflC and BcSAS1 transcript levels between the test and control applications were significant at α = 0.05 (Table S2) and positively correlated with the OD₆₀₀ of the bacterial cultures (AflC: r = 0.74, p = 0.0061; BcSAS1: r = 0.73, p = 0.007). Moreover, the reduction in target gene expression was presumably caused by the bacterially-produced dsRNAs alone, as the expression levels of AflC and BcSAS1 did not statistically differ among the control applications, or between the control applications and the axenic culture of A. flavus and B. cinerea (Figure S3A,B, and Tables S3 and S4). Taken together, the above results indicate that the two HT115(DE3) strains that were genetically engineered to produce dsRNAs against AflC or BcSAS1 are capable of inducing RNAi to A. flavus and B. cinerea in in vitro co-cultures of the two organisms, in a bacteria concentration-dependent manner.

As for Treatment A, a bacteria concentration-dependent response in the silencing levels of AflC and BcSAS1 (AflC: r = 0.80, p = 0.0017; BcSAS1: r = 0.77, p = 0.0037) was also observed when germlings of A. flavus and B. cinerea were incubated for 12 h under the same experimental conditions with the whole-cell autolysates of HT115/AflC_Con and HT115/BcSAS1_Ind, respectively (Treatment B) (Figure 2A,B and Table S2). Specifically, a significant (p < 0.05) reduction in the expression levels of AflC in the test relative to the control applications was obtained when A. flavus was incubated with lysates of HT115/AflC_Con collected from the OD₆₀₀ 1.0 and 0.5 cultures but not with the lysates collected from the OD₆₀₀ 0.1 and 0.01 cultures (Figure 2A and Table S2). On a quantitative basis, subtle changes in AflC expression between test and control applications were observed in Treatment B compared with Treatment A for all bacterial densities tested (Table S2), suggesting that environmental uptake of dsRNAs from bacterial whole-cell autolysates was less efficient in eliciting RNAi in A. flavus compared with applications of living bacteria. When considering BcSAS1, then a significant (p < 0.05) reduction in its expression level was observed when germlings of B. cinerea were co-incubated with lysates of HT115/BcSAS1_Ind collected from all four serially diluted cultures (i.e., cultures with an OD₆₀₀ of 1, 0.5, 0.1, 0.01) (Figure 2B and Table S2). Moreover, fold-changes in BcSAS1 transcript levels between test and control applications were lower in Treatment B compared with Treatment A for the OD₆₀₀ 1.0 and 0.5 cultures, but higher for the OD₆₀₀ 0.1 and 0.01 cultures (Figure 2B and Table S2). This suggests that higher densities of cell lysates could be interfering with the absorption of dsRNAs by the hyphae, thus making it appear that living bacteria are more potent in eliciting silencing. Finally, the inclusion of just the bacterial lysates in the liquid cultures had no effect on the expression of AflC or BcSAS1, as their transcript levels were essentially the same among control applications (Figure S3A,B and Tables S3 and S4).

In order to appreciate the silencing levels achieved against AflC and BcSAS1 in Treatments A and B, an additional application of purified dsRNAs was evaluated, consisting of 15 μg (12.5 nM) of dsRNAs against AflC or BcSAS1 applied to 5 mL of in vitro cultures (10⁴ germlings/mL) of A. flavus or B. cinerea, respectively. Compared with the expression levels of these two genes in control axenic cultures of A. flavus or B. cinerea, the application of the purified dsRNAs led to a 2.9-fold reduction in AflC transcript levels (Figure 2A) and an 8.9-fold reduction in BcSAS1 transcript levels (Figure 2B). These changes in gene expression are intermediate between those obtained in Treatments A and B at a bacterial OD₆₀₀ of 0.1 and 0.5. Thus, collectively our results indicate that the bacterial-based platform to produce dsRNAs can be effectively utilized to silence fungal genes, achieving results that are on par with those obtained with applications of naked dsRNAs.

3.3. Increased Contact Times between Fungi and dsRNA-Producing Bacteria Have a Positive Effect on Reducing Production of Aflatoxins in A. flavus and Mycelial Growth in B. cinerea

We next assessed whether an increase in contact time between HT115/AflC_Con and A. flavus or between HT115/BcSAS1_Ind and B. cinerea, would have an added positive effect on the dsRNA-mediated induced silencing of AflC and BcSAS1, respectively. For this purpose, concentrates of living bacteria or of their whole-cell autolysates from the OD₆₀₀ 1.0 cultures were added to 5 mL cultures of A. flavus or B. cinerea, and silencing levels of AflC and BcSAS1 were assessed after 12 h, 24 h, and 36 h of co-incubation (Table S5). All other conditions in these experiments were kept the same as in the in vitro experiments described above, including control applications.

Examination of the AflC transcript levels in just the control applications showed that the expression of this gene decreased over time in both Treatment A and Treatment B. The changes in AflC transcript levels from 12 h to 24 h and 36 h were significant at α = 0.05 (12 h to 24 h, Treatment A and B) or α = 0.10 (12 h to 36 h, Treatment A). However, there were not significant differences in AflC expression levels between 24 h and 36 h for neither Treatment A nor Treatment B (Figure S4A and Table S6). When assessing AflC silencing levels in the test relative to the control applications, then in Treatment A, a large 21.7-fold decrease in the AflC expression levels was observed at 12 h, but only a 5.2-fold and a 4.6-fold decrease were observed at 24 h and 36 h, respectively. The decreases in AflC expression at 12 h and 24 h were significant at α = 0.05, whereas the decrease at 36 h was significant only at α = 0.1. In a similar way, in Treatment B, a 4.1-fold, a 1.7-fold, and a 2.6-fold decrease in AflC expression were seen at 12 h, 24 h, and 36 h, respectively in the test relative to the control applications, of which only the decrease at 12 h was significant at α = 0.05 (Figure 3A and Table S5). Collectively, the data suggest that although the HT115/AflC_Con living bacteria or of their lysates strongly repressed AflC transcript levels during the first 12 h, the silencing effect was nearly alleviated by 24 h, despite the continuous presence of the bacteria or their lysates in the co-cultures. However, it is also possible that the absence of detectable silencing of AflC at 24 h and 36 h post bacterial applications is caused by the low expression levels of this gene at these two time points that diminishes the effects of the dsRNA treatment.

Since AflC encodes for a key enzyme in the biosynthetic pathway of aflatoxins [42], we next examined whether the observed HT115/AflC_Con-mediated silencing of AflC would influence their production. Quantification of aflatoxins content in co-cultures of Treatment A and B after 12 h, 24 h, and 36 h showed that, unlike AflC transcript levels, the total amount of these mycotoxins increased over time, most likely because of the parallel increase and accumulation in fungal biomass (Figure 3B and Table S7). However, when comparing aflatoxins content between test and control applications of either Treatment A or B, then a decrease was observed in the test applications that was more pronounced at 36 h, followed by 24 h and 12 h. Specifically, in co-cultures of Treatment A, a 64.6% and a 30.5% reduction in aflatoxins content was observed in the test relative to the control applications at 36 h and 24 h of co-incubation time, respectively, whereas at 12 h, a 13.3% increase was observed. The decrease at 36 h was significant at α = 0.05, whereas at 24 h it was significant at α = 0.1 and at 12 h it was not significant (Figure 3B and Table S7). In a similar way, in co-cultures of Treatment B, a 42.3%, 35.4%, and 0% reduction in aflatoxins content was observed in the test relative to the control application at 36 h, 24 h, and 12 h of co-incubation time, respectively. The differences were significant at α = 0.05 at 36 h and 24 h but not significant at 12 h (Figure 3B and Table S7). Overall, our results show that bacterially-produced dsRNAs against AflC can significantly impair production of aflatoxins over time, whereas the inverse correlation seen between AflC mRNA abundance and concentration of aflatoxins, suggests that there is likely a delay between AflC expression, and biosynthesis and secretion of aflatoxins.

When examining how increased contact times between B. cinerea and living HT115/BcSAS1_Ind or the lysates thereof affected BcSAS1 transcript levels and subsequent growth of the fungus, then a positive correlation was observed as well. Overall, BcSAS1 expression levels increased from 12 h to 24 h and 36 h in the control applications of both Treatment A and Treatment B. However, the differences in BcSAS1 expression between 12 h, 24 h, and 36 h were not significantly different from each other (Figure S4B and Table S8), indicating that the expression of this gene was essentially stable in the controls throughout the course of the experiment. When compared with the control applications, exposure of B. cinerea to living HT115/BcSAS1_Ind bacterial cells or their whole-cell autolysates resulted in a significant at α = 0.01 reduction in BcSAS1 transcript levels at 12 h and 36 h of co-incubation time. However, at 24 h, the reduction was significant only at α = 0.10 (Figure 3C and Table S5). Specifically, in co-cultures of Treatment A, a 23.3-fold and 14.5-fold decrease in BcSAS1 transcripts levels were observed in the test applications relative to the control ones at 12 h and 36 h, respectively whereas only a 4.3-fold reduction was seen at 24 h. In a similar way, in co-cultures of Treatment B, a 20.0-fold and a 25.6-fold reduction were observed at 12 h and 36 h, respectively, whereas a 2.9-fold reduction was seen at 24 h (Figure 3C and Table S5). These oscillations in BcSAS1 silencing levels from 12 h to 24 h and 36 h could represent an attempt by the fungal cells to partially recover from the effects of RNAi, perhaps through the activation of a regulatory feedback mechanism that increased BcSAS1 expression at around 24 h in an effort to compensate for the reduction in its mRNA levels. Irrespective of whether such a negative-feedback loop is operational for BcSAS1, overall, the results indicate that increased contact times between the fungus and the dsRNA-producing bacteria can have a positive impact on silencing of target genes, although this activity seems to be dependent on the gene target and possibly other parameters as well.

Since BcSAS1 is involved in mycelial development in B. cinerea [43], we also examined whether the observed HT115/BcSAS1_Ind-mediated silencing of BcSAS1 at 12 h, 24 h, and 36 h of co-incubation time between the fungus and the living bacterial cells or their autolysates would impact fungal growth. Indeed, a 16.0%, 27.4%, and 24.1% reduction in fungal dry weight was observed in the test relative to the control co-cultures of Treatment A at 12 h, 24 h, and 36 h of co-incubation time between the fungus and the bacteria, respectively (Figure 3D and Table S9). Although small, these changes were significant at α = 0.05 for 12 h and 24 h but not at 36 h (Table S9). In a similar way, a 39.2%, 61.3%, and 24.1% reduction in fungal dry weight was seen at 12 h, 24 h, and 36 h, respectively in the test co-cultures of Treatment B relative to the control ones (Figure 3D and Table S9). It is worth noticing that the oscillations observed in the reduction of fungal biomass at 12 h, 24 h, and 36 h inversely mirrored the fluctuations observed in the BcSAS1 silencing levels at these time points, suggesting that there could be a delay between inhibition of mRNA and phenotypic expression of growth impedance. This was more pronounced in Treatment B, as fungal growth in Treatment A was partially inhibited by growth of the bacteria, thus alleviating some of the observed differences. In this respect, fungal growth as determined by mycelial dry weight in the control co-cultures of Treatment A was reduced by 36.1% at 12 h (p = 0.068), 41.7% at 24 h (p = 0.003), and 16.9% at 36 h (p = 0.213) compared with Treatment B (Table S9). Despite these differences between Treatment A and Treatment B, overall the results demonstrate that applications of bacteria producing dsRNAs against BcSAS1 can effectively reduce mycelial growth of B. cinerea under in vitro conditions.

3.4. In Vivo Applications of dsRNA-Producing Bacteria Result in a Reduction of Aflatoxins Production by A. flavus and Disease Symptoms Caused by B. cinerea

The in vitro experiments demonstrated that HT115/AflC_Con and HT115/BcSAS1_Ind could be used as vectors for delivering RNAi against the target genes in A. flavus and B. cinerea, and thus impede production of aflatoxins and mycelial growth, respectively. Therefore, we next examined whether such an effect could also be achieved under in vivo conditions during fungal infections of host plants. For this purpose, A. flavus and B. cinerea were spot inoculated on maize seeds and leaves of N. benthamiana, respectively and after a period of incubation of 48 h for A. flavus and 6 h for B. cinerea, 10 μL of living bacterial concentrates or their whole-cell autolysates from the OD₆₀₀ 1.0 cultures were applied to the same fungal inoculation spots.

Examination of AflC transcript levels in A. flavus inoculated on maize seeds (Figure 4A) and exposed for 24 h at living HT115/AflC_Con bacteria or their autolysates, showed a significant (α = 0.05) 4.8-fold and a 6.5-fold reduction, respectively in its expression levels in the test relative to the control applications (Figure 4B and Table S10). Notably, AflC expression levels did not statistically differ between control applications of Treatment A and Treatment B (p = 0.441), indicating that silencing of AflC was caused by the dsRNAs produced by the HT115/AflC_Con bacteria. Moreover, the decrease in AflC transcript levels was further accompanied by a reduction in production of aflatoxins, as determined by parallel measurements of total aflatoxins content in the maize seeds (Figure 4C and Table S10). Specifically, for Treatment A, a 47.4% reduction was observed in the total amount of aflatoxins extracted from seeds treated with the HT115/AflC_Con strain compared with seeds treated with the HT115/eGFP_Con strain. In a similar way, for Treatment B, a 51.8% reduction in aflatoxins content was observed in seeds included in the test application relative to seeds included in the control one (Figure 4C and Table S10). The reduction in the total aflatoxins content between test and control applications clearly demonstrates that the HT115/AflC_Con strain can effectively obstruct the production of these mycotoxins by A. flavus, also under in vivo conditions.

As for AflC, application of the living HT115/BcSAS1_Ind cells or their whole-cell autolysates on N. benthamiana leaves inoculated with B. cinerea (Figure 5A) resulted in silencing of BcSAS1. Indeed, a 2.6-fold and 1.7-fold reduction in BcSAS1 expression was seen in the test applications relative to the control ones for Treatment A and Treatment B, respectively at 72 h post addition of the living bacteria or their whole-cell autolysates on N. benthamiana leaves inoculated with B. cinerea (Figure 5B and Table S11). These differences in BcSAS1 transcript levels between test and control applications were significant at α = 0.05, whereas BcSAS1 expression levels were the same between control applications of Treatment A and Treatment B (p = 0.820), indicating that the reduction in its transcript levels in the test applications was caused by the dsRNAs produced by the HT115/BcSAS1_Ind strain. Moreover, silencing of BcSAS1 was phenotypically expressed as a parallel reduction in the size of the infection lesions caused by B. cinerea on leaves of N. benthamiana (Figure 5A,C). Indeed, for Treatment A, a significant 35% reduction in lesion size between test and control application was observed, whereas in Treatment B, this significant reduction was by 25% (Figure 5C and Table S11). Collectively, these results demonstrate that applications of bacteria that produce dsRNAs against genes in target pathogens can be potentially used for disease control.

4. Discussion

Current practices to limit fungal diseases on plants rely heavily on synthetic chemical compounds but their constant use poses a threat to the environment and to human health, and further increases the risk of pathogen resistance development [44]. Host resistance can provide effective control of pathogens but may be limited in some plant species, and pathogens can often overcome resistance [45]. Therefore, alternative methods for disease control are urgently needed to meet the increasing demands in crop yields and quality of produce [44,46,47].

In this study, we exploited the systemic properties of dsRNAs and successfully utilized bacteria as an alternative to HIGS and SIGS methods to deliver RNAi to filamentous fungi. Bacteria have the advantage that they can be cultured in large volumes relatively inexpensively and protect the produced dsRNAs within their cells until their application [48,49]. Our results showed that even after bacterial lysis, the dsRNAs remained resistant to degradation for at least up to 24 h, and partially until 72 h, providing enough time for the dsRNAs to be absorbed by the fungal cells and induce silencing in target genes. The use of bacteria as vectors for delivering RNAi also circumvents the hurdle and limitations of producing and deploying transgenic plants in the field, whereas bacteria can be further engineered to be non-pathogenic microorganisms or to autolyze under selected induced conditions, thus increasing the safety of application [27]. Indeed, the utility of bacteria as delivery vehicles of RNAi has already been demonstrated against insects, nematodes, and even mammalian cell cultures [27,29,31] but, surprisingly, it had not been tested against fungi. We have now bridged this gap and show that bacteria that have been genetically engineered to produce dsRNAs against fungal genes can be used both in living form and as whole-cell autolysates to induce RNAi in fungi. Indeed, our data showed that even a single application of the recombinant bacteria was sufficient to achieve a significant reduction in production of aflatoxins by A. flavus or growth of B. cinerea, both under in vitro and in vivo conditions. This finding generates new opportunities for developing RNAi-based control methods against fungi of agronomic or medical importance by using non-pathogenic or attenuated bacteria as a dsRNAs delivery system.

In our studies, we have exploited the widely used HT115(DE3) E. coli strain as the host bacterium for the production of dsRNAs [29], but other bacteria species could be potentially engineered to deliver RNAi to fungi, including for example bacteria used in biological control, plant endophytic or symbiotic bacteria, and several others [48]. Indeed, symbiotic gut bacteria of insects were recently successfully engineered to deliver RNAi to their insect hosts [50], indicating that the approach is amendable to different bacterial species that can be used to meet the needs of specific applications. To further increase the versatility of the system and make it suitable for diverse applications, we utilized two recombinant expression vectors that enabled the production of dsRNAs in HT115(DE3) under constitutive or IPTG-induced conditions. Specifically, the first vector employed the constitutive Kan promoter and allowed the uni-directional transcription of the two complementary RNA strands from the target DNA template, which then anneal post-transcriptionally in the cytoplasm to form dsRNAs. The second vector was based on the bi-directional transcription of the target DNA sequence in a single reaction using the T7 promoter, which is activated upon binding of the T7 RNA polymerase to it. T7 RNA polymerase translation is in-terms dependent on initial IPTG-induction of transcription of the T7 gene in HT115(DE3) that encodes it. Inducible systems would generally provide higher efficiency, flexibility, and safety in use, as they enable the production and release of dsRNAs from bacteria under user-selected conditions. Although here we used the IPTG-activated T7 RNA polymerase-based expression system, a range of inducible promoters that can be triggered under specific conditions could be employed [51,52]. Constitutive promoters, on the other hand, increase the ease of dsRNA production and could be used when the persistent presence of recombinant bacteria producing dsRNAs is required to achieve a desirable and sustained silencing effect. Although the kinetics of dsRNAs production by the two expression systems was not part of our study, both expression systems were shown to yield sufficient amounts of biologically active dsRNAs that induced silencing in the target fungal genes. However, direct comparisons between the two systems could not be made, as different target sequences were cloned in the recombinant expression vectors, whereas dsRNA production in the IPTG-activated system would vary depending on the time of IPTG addition to the medium and its concentration [53]. Finally, although frequently utilized in RNAi studies, the production of hairpin-structured RNA (hpRNA) was not included in these studies because it was demonstrated that HT115(DE3) is able to process hpRNA into short fragments, in spite of lacking RNaseIII activity [54].

In order to combine the benefits or trans-kingdom transfer and environmental uptake of dsRNAs, we further engineered the HT115(DE3) E. coli strain to undergo programmed cellular autolysis induced by endolysin R, following a freeze-thaw cycle [35]. Both living bacteria and their autolysates induced significant silencing in the target fungal genes in a bacteria concentration-dependent manner, suggesting that further increasing bacterial densities could lead to an even higher effect. However, caution should be taken as increasing siRNAs concentrations past an optimum point does not necessarily lead to increased silencing levels and may even have an opposite effect due to saturation of the RNAi machinery [55,56]. Moreover, the silencing levels achieved by the two delivery methods were on par with those obtained with external applications of purified dsRNAs, indicating that even simple applications with living bacteria or their autolysates that bypass the purification of dsRNAs and their biological stabilization by binding to various materials [3,28], can be effective. However, whether one application was more effective than the other (i.e., living bacteria or autolysates) could not be established, since a combination of parameters, including the fungal species and bacterial concentration, had an effect on the observed silencing levels. For instance, while applications under in vitro conditions with living bacteria were seemingly more effective in inducing AflC silencing in A. flavus, applications with the bacterial autolysates were slightly more effective under in vivo conditions. When examining silencing of BcSAS1 in B. cinerea, then almost the opposite trend was observed, as applications under in vitro conditions of whole-cell autolysates were more effective at some bacterial concentrations, whereas living bacteria seem to have had a bigger effect under in vivo conditions. Such variation in the effectiveness of the two application methods depending on the fungal species could perhaps relate to the way and rate by which the bacterially-produced dsRNAs are taken-up by the fungal cells. Although it is still unclear how exogenous dsRNAs are internalized within fungal cells, several possible mechanisms may be in effect, including passive diffusion through the fungal cell wall, internalization through plasma membrane-associated receptors or protein channels and pores [57], or through the endocytosis or fusion of bacterial extracellular vesicles carrying the dsRNA [58]. It is plausible that the uptake of dsRNAs from the bacterial lysates is mediated by passive or active diffusion through the cell wall and membranes, whereas living bacteria translocate dsRNAs to the fungal cells through the secretion of extracellular vesicles in a way similar to plants [59]. Indeed, several studies have demonstrated that Gram-negative bacteria, including E. coli, release extracellular vesicles that contain diverse RNA molecules, including small RNAs (sRNAs) that can modulate gene expression in recipient eukaryotic host cells [24,58,60,61,62]. Irrespectively of the mechanism(s) that may be in effect, differences in their efficacy and utilization among different fungal species could perhaps explain the lack of correlation seen between applications of living or lysed bacteria, and silencing levels achieved in A. flavus and B. cinerea under in vitro and in vivo conditions.

Finally, an increase in contact time between the fungi and the living bacteria or their autolysates did not produce a clear benefit in terms of enhancing or maintaining silencing levels in AflC or BcSAS1. However, one might expect such an effect, at least in applications of living bacteria that continuously produce dsRNAs compared with the one-time dose of dsRNAs received by the fungal cells from the bacterial autolysates. A combination of various factors inherent to both the targeted genes and the fungal species could account for this lack of a stable silencing effect over time. Among others, these may include dsRNAs absorption rates, transcript abundance and mRNA turnover rate of the target genes, and the presence of regulatory feedback mechanisms that could restore gene expression upon silencing [63,64]. Indeed, upregulation of gene expression following RNAi is often reported as a cellular response, mainly as a result of saturation of the RNAi machinery [65], which in our case could be caused by the constant supply of dsRNAs. Nonetheless, although the increase in contact time between fungi and the bacteria did not lead to a significant further increase in AflC or BcSAS1 silencing levels, in contrast, it had a positive effect on reducing production of aflatoxins in A. flavus and mycelial growth of B. cinerea over time. This could possibly be due to the time needed for the inhibition of gene expression to materialize into an observable phenotypic effect, or because silencing reduces protein production at higher levels than mRNA levels, thus having a higher effect on phenotype. Indeed, discordances between mRNA downregulation and reduction in protein production have been documented before, thus leading to analogous disparities between changes in gene expression and phenotype severity [66,67].

Overall, in this study, we demonstrate that genetically engineered bacteria that produce dsRNAs against selected genes in fungi can be effectively used as vectors for delivering RNAi to fungi with high efficiency, thus opening new avenues for disease control.

5. Conclusions

• The E. coli HT115(DE3) strain can produce biologically active dsRNAs against target genes in fungi.
• The biologically active bacterial dsRNAs were successfully delivered to fungi through two different approaches: (i) treatment with living bacteria and (ii) treatment with bacterial lysates following a programmable cell autolysis step.
• In vitro applications of dsRNA-producing bacteria show that the degree of the RNAi effect on fungi is positively correlated with the concentration of the bacteria, while the phenotypic changes (i.e., reduction in production of aflatoxins in A. flavus and reduction of mycelial growth in B. cinerea) are positively correlated with the contact time.
• In vivo applications of dsRNA-producing bacteria result in a reduction of aflatoxins production by A. flavus on corn seeds and a reduction of disease symptoms caused by B. cinerea on N. benthamiana leaves.