A fungal milRNA mediates epigenetic repression of a virulence gene in Verticillium dahliae

Abstract

MiRNAs in animals and plants play crucial roles in diverse developmental processes under both normal and stress conditions. miRNA-like small RNAs (milRNAs) identified in some fungi remain functionally uncharacterized. Here, we identified a number of milRNAs in Verticillium dahliae, a soil-borne fungal pathogen responsible for devastating wilt diseases in many crops. Accumulation of a V. dahliae milRNA1, named VdmilR1, was detected by RNA gel blotting. We show that the precursor gene VdMILR1 is transcribed by RNA polymerase II and is able to produce the mature VdmilR1, in a process independent of V. dahliae DCL (Dicer-like) and AGO (Argonaute) proteins. We found that an RNaseIII domain-containing protein, VdR3, is essential for V. dahliae and participates in VdmilR1 biogenesis. VdmilR1 targets a hypothetical protein-coding gene, VdHy1, at the 3′UTR for transcriptional repression through increased histone H3K9 methylation of VdHy1. Pathogenicity analysis reveals that VdHy1 is essential for fungal virulence. Together with the time difference in the expression patterns of VdmilR1 and VdHy1 during fungal infection in cotton plants, our findings identify a novel milRNA, VdmilR1, in V. dahliae synthesized by a noncanonical pathway that plays a regulatory role in pathogenicity and uncover an epigenetic mechanism for VdmilR1 in regulating a virulence target gene.

This article is part of the theme issue ‘Biotic signalling sheds light on smart pest management’.

1. Introduction

Verticillium dahliae is one of the major causal pathogens of Verticillium wilt, which leads to devastating losses of many economically important crops worldwide [1–3]. Verticillium wilt is a soil-borne fungal disease attacking the crop vascular system and its control is particularly challenging because the fungus is able to survive in adverse environments as a resting structure known as microsclerotia for over a decade in the soil. In addition, V. dahliae poses a broad host range of more than 400 plant species [4,5]. Cotton, an important economic crop, is one of its major hosts. Germination of V. dahliae microsclerotia in soil is stimulated by root exudates. We have recently reported that V. dahliae germinates large numbers of hyphae on the root surface, a few of which differentiate into hyphopodia and penetration pegs to pierce cotton roots during initial colonization. Once a certain number of hyphae have extended into xylem vessels, the hyphae rapidly proliferate in the vessels, which secrete elicitor and effector proteins and trigger the Verticillium resistance response that ultimately develops into wilt disease symptoms with a typical wilting leaf phenotype [6–9].

Small RNAs (sRNAs)-induced RNA interference (RNAi) is a widespread biological process leading to sequence-specific degradation or translational repression of target mRNAs (posttranscriptional gene silencing (PTGS)), chromatin modification or transcriptional gene silencing (TGS) [10,11]. The common feature of RNAi pathways is that Argonaute (AGO)-bound sRNAs guide protein effector complexes to complementary targets to mediate gene silencing. RNAi tends to be highly conserved in eukaryotes, including throughout most of the fungal kingdom, which consists of a large number of diverse groups [12–15]. As in the fission yeast Schizosaccharomyces pombe, RNA-induced transcriptional silencing (RITS) represses transcription by recruiting a histone methyltransferase to the target genes in heterochromatin [16–18]. The posttranscriptional gene silencing phenomenon was first discovered and designated quelling in a model fungus, Neurospora crassa [19]. It is caused by transgenic DNA, which can induce sequence-specific degradation of homologous mRNAs. After that, quelling-defective mutants (qde) were generated and qde genes were isolated, and the corresponding proteins QDE1 (RdRP, RNA-dependent RNA polymerase), QDE2 (AGO) and QDE3 (RecQ DNA helicase) were functionally characterized [20–23]. Two partially redundant Dicer-like (DCL) proteins involved in the production of sRNA were also identified in N. crassa [24]. Another silencing mechanism, termed meiotic silencing by unpaired DNA (MSUD), protects the genome during the sexual cycle in N. crassa [25]. RNAi and sRNA have also been reported in some important pathogenic filamentous fungi, such as Magnaporthe oryzae [26–28] and Fusarium graminearum [29]. Notably, RNAi pathways in fungi are complex and mainly function in genomic defence, heterochromatin formation and gene regulation [12,30]. DCL and AGO proteins have been found in many fungi, and RNAi-based target gene repression is widely applied in functional genome research in various fungi. Recently, a non-canonical RNAi pathway that involved an RNaseIII-like protein, R3B2, which promotes mRNA degradation, has also been reported in Mucor circinelloides [31]. Thus, RNAi pathways exist in most fungi, similar to other eukaryotes.

sRNAs, defined by their length of 18–24 nucleotides, play important roles in growth and developmental processes, stress response and pathogenicity [32–36]. Among them, miRNAs are a class of interior sRNAs found in eukaryotes and that function in many processes, including development, biotic and abiotic responses and defence [37]. MiRNA was thought to be absent in fungi until the recent identification of miRNA-like RNAs (milRNAs) in N. crassa [38]. Four major milRNAs in N. crassa are produced by at least four different biosynthetic pathways using a distinct combination of factors, including DCLs, QDE2, QDE2-interacting protein (QIP) and MRPL3 (an RNaseIII domain and dsRNA recognition motif-containing protein) [38]. QDE1 and QDE3 are not required for the production of these milRNAs. Moreover, most plant and animal miRNAs are transcribed by RNA polymerase II (Pol II), while transcription of the major milRNAs in N. crassa relies on Pol III [39]. milRNAs exist in Fusarium oxysporum, Sclerotinia sclerotiorum, Puccinia striiformis f. sp. tritici and Zymoseptoria tritici [40–43]. However, studies have mainly focused on bioinformatic predictions. The biogenesis and function of the predicted milRNAs are rarely investigated.

Recently, a natural RNAi transmission between fungal pathogens and host plants, a phenomenon of trans-kingdom RNAi in plant–fungi interactions, has been reported [44–46]. We have found that cotton plants export endogenous miRNAs, including miR166 and miR159, into V. dahliae hyphae to target fungal virulence genes for silencing [7]. We also successfully applied trans-kingdom RNAi by host-induced gene silencing (HIGS) technology in cotton to defend against Verticillium wilt caused by V. dahliae [44]. These data imply that the RNAi system is active in V. dahliae and have aroused great interest in investigating this fungal RNAi system.

In the present study, we characterized the RNAi components and small RNAs from the V. dahliae genome (strain V592). A novel milRNA, named VdmilR1, was identified to be possibly produced by a noncanonical RNAi component, an RNaseIII-containing protein. The gene encoding a hypothetical protein, named VdHy1, was identified to be a target of VdmilR1 for epigenetic repression; VdHy1-knockout mutants had reduced virulence in cotton plants. These data, together with the expression patterns of VdmilR1 and VdHy1 during the infection process, uncover a VdmilR1/VdHy1 module plays a role in V. dahliae pathogenicity in cotton plants.

2. Results

(a). Identification of potential RNA silencing components in Verticillium dahliae

Based on the genome sequences of VdLs.17, a V. dahliae isolate from lettuce, we identified three homologs of RdRP proteins (VDAG_02528 for RdRP1, VDAG_03247 for RdRP2 and VDAG_07629 for RdRP3), which contained an RdRP domain; two homologs of AGO proteins (VDAG_10484 for AGO1 and VDAG_01530 for AGO2), which possessed conserved PAZ and PIWI domains, and a canonical DCL protein (VDAG_00471 for DCL1), which contains an RNA helicase domain, dsRBD and two RNaseIII catalytic domains (figure 1a). An atypical DCL protein (VDAG_06945), lacking one RNaseIII domain, was also identified and named DCL2 (figure 1a). The evolutionary relationships of V. dahliae RdRPs, AGOs and DCLs with those from other fungal species [47] and Arabidopsis were established by phylogenetic analysis of protein sequences. Members of V. dahliae RNAi components were grouped clearly with known members that function in quelling or MSUD pathways (electronic supplementary material, figure S1), although sexual reproduction has not been characterized in V. dahliae, which is classified as a species of Fungi Imperfecti. RdRP1, AGO1 and DCL1 were grouped with quelling factors, while RdRP2, AGO2 and DCL2 were grouped with proteins from the MSUD pathway. RdRP3 was in a separate group, in which proteins contained related domains but with unknown function, independent of the quelling and MSUD classes.

Figure 1.

RNA-silencing components and the profiles of small RNAs in V. dahliae. (a) Conserved domains of RNA-silencing components RdRP, DCL and AGO in V. dahliae. (b) The length distribution of total Vd-sRNAs and Vd-sRNAs after removing 3′ or 5′ truncated homologues in V. dahliae. (c) Ratio of the first nucleotide preference of Vd-sRNAs in V592. (d) Predicted stem-loop structures of seven highly expressed (RPM > 10) VdmilRNA precursors in V592, including VdMILR1. The mature VdmilRNA sequences are labelled in red. (e) Small RNA gel blotting of VdmilR1 in V592 detected by using the LNA™ (Locked Nucleotide) enhanced probe.

(b). Small RNA analysis and prediction of potential miRNA-like genes in Verticillium dahliae

We then isolated total RNA from V. dahliae for sRNA extraction and high-throughput sequencing. After removing low-quality reads, we obtained 11 213 899 reads, 52.43% of which perfectly matched to the VdLs.17 genome. The perfectly matched sRNAs, Vd-sRNAs, with length between 18 and 30 nt, were included in our analysis.

The length distribution and 5′-terminus nucleotide preferences of Vd-sRNAs were assayed. Vd-sRNAs were mainly 18–25 nt in length and did not show a preference for a specific size (figure 1b), but they had a high proportion for guanine at the 5′ end (1G) (figure 1c). These data contrast with those from N. crassa sRNAs, which are 19–25 nt long, with a preference for 21–22 nt, and had a strong preference for 1 U [38]. We, therefore, aligned Vd-sRNAs sequences and found that many shorter Vd-sRNAs (18–20 nt) were homologous to longer Vd-sRNAs with 3′ or 5′ truncation, suggesting possible exonuclease-mediated digestion after Vd-sRNA production. Correspondingly, the most Vd-sRNAs were 21–24 nt after removing the shorter homologous sequences (figure 1b).

Mireap and miRDeep [48] were used to appraise potential miRNAs in the V. dahliae genome. Seven highly expressed Vd-sRNAs (RPM > 10) mapping loci could form stem-loop structures (figure 1d). The lengths of the putative precursors were within 70–400 bp. These candidate milRNAs were 20–23 nt in length.

Small RNA gel blotting was carried out to examine the accumulation of the high-RPM miRNA candidates. Unexpectedly, we failed to detect any small RNA signal using regular oligo probes. Therefore, an LNA™ (Locked Nucleotide) enhanced probe that has been suggested to be more sensitive and specific than ordinary probe was applied for the blotting analysis. With this approach, we were able to detect one of milRNA candidates (figure 1e), which was predicted to originate from a locus in chromosome 8 at the gene link region between 2083943 and 2083870 (supercontig1.14: 753992:754065). We named it VdmilR1.

(c). Validation and transcription analysis of VdmilR1 precursor

As shown in figure 1d, the predicted precursor of VdmilR1 (VdMILR1) could form typical stem-loop structures. To verify whether it could produce mature VdmilR1, we overexpressed VdMILR1 under constitutive expression from the Tef promoter in wild-type V. dahliae V592 to obtain the Tef-VdMILR1 strain, which had no obvious developmental defect but did show dense growth of hyphae (figure 2a). Higher accumulation of VdmilR1 was detected in Tef-VdMILR1 compared with wild-type V592 (figure 2b), verifying that VdMILR1 is a functional precursor encoding VdmilR1.

Figure 2.

VdmilR1, produced by an RNaseIII protein, VdR3, epigenetically represses its target virulence gene VdHy1. (a) Colony morphology of wild-type V592, overexpression strains Tef-VdMILR1 and Tef-VdR3, and knockout mutant Δvdhy1 and its complement strain Δvdhy1/VdHy1 on PDA plates after two weeks of incubation. We found no obvious developmental defect but the slightly dense growth of hyphae was observed in Tef-VdMILR1 and Tef-VdR3. Δvdhy1 reduced hyphal growth and melanin production, and the complemented strain Δvdhy1/VdHy1 restored the colony morphology. (b) Detection of VdmilR1 in Tef-VdMILR1 by small RNA gel blotting. Higher accumulation of VdmilR1 was detected in Tef-VdMILR1 compared with wild-type V592. (c) Pol II is required for the production of VdmilR1. Schematic of the overlapping locus of VdMILR1 and Asp-tRNA in the genome (left). The arrows indicate the transcription direction. Small RNA gel blotting shows the level of VdmilR1 in the indicated strains (right) treated with α-amanitin (Pol II inhibitor) or H₂O (as solvent control), or with ML-60218 (Pol III-specific inhibitor) or DMSO (solvent control). The rRNA in the bottom panel shows equal loading of RNA samples. (d) Small RNA hybridization shows that mature VdmilR1 was markedly increased in Tef-VdR3 strains. (e) Quantitative RT-PCR (qRT-PCR) of VdHy1 in wild-type, Tef-VdMILR1 and Δvdmilr1. The V. dahliae tubulin beta chain gene was used as the internal control. Error bars show ±s.d. from three replicates. The level of VdHy1 mRNA in V592 was set to 1. (f) Chromatin immunoprecipitation (ChIP) assay at VdHy1 chromatin using H3K9me3-specific antibody. The top schematic shows the genomic structure of VdHy1 and the relative positions of the primers (P1–P3) used for ChIP assays and the VdmilR1 target site. Black box: coding sequence. White boxes: 5′UTR and 3′UTR sequences. ChIP-qPCR data were normalized to a sample of input DNA. The V. dahliae tubulin beta chain gene was used as the internal control. Error bars show ±s.d. from three replicates. The values in V592 were arbitrarily designated 1. (g) Disease symptoms of cotton plants infected with V592, Tef-VdMILR1, Δvdhy1 or complemented strain Δvdhy1/VdHy1 at 25 days postinoculation (dpi). (h) The expression patterns of VdmilR1/VdHy1 during the infection process. The V. dahliae tubulin gene and V. dahliae U6 were used as internal controls for VdHy1 and VdmilR1, respectively. Error bars show ±s.d. from three replicates. Data obtained from the time point of 0.5 (first lanes) were set to 1.

In most eukaryotes, miRNAs are transcribed by RNA polymerase II (Pol II) [49–52], while Pol III is responsible for tRNA and 5S rRNA transcriptions [53,54]. Interestingly, in N. crassa, the four major milRNAs are transcribed by Pol III [39]. Because the VdMILR1 locus overlapped with an Asp-tRNA-coding locus (figure 2c), we investigated which RNA polymerase was responsible for the transcription of VdMILR1 in V. dahliae by using chemical inhibitors of Pol II and Pol III and examining the accumulation of VdmilR1. As shown in figure 2c, compared with the solvent (H₂O) treatment (control), Pol II inhibitor treatment reduced the accumulation of VdmilR1, whereas Pol III inhibitor treatment increased, rather than decreased, VdmilR1 compared with the solvent (DMSO) treatment control, suggesting that inhibition of Pol III promoted VdMILR1 transcription. Notably, DMSO as the solvent of Pol III inhibitor seemed to repress the expression of VdmilR1, consistent with the previous report in N. crassa in which DMSO also reduced the expression of the four major milRNAs [39]. These results demonstrate that Pol II-, but not Pol III-, mediates VdMILR1 transcription in V. dahliae, and inhibition of Pol III promotes Pol II-mediated VdMILR1 transcription at the VdmilR1/tRNA overlapping loci. Our data also imply divergent milRNA transcription mechanisms in fungi.

(d). An RNaseIII domain-containing protein, but not the canonical DCL and AGO proteins, are necessary for VdmilR1 biogenesis

In plants and animals as well as in N. crassa, DCL proteins have conserved roles in directing small RNA synthesis. It is worth mentioning that Neurospora AGO protein forms the RNA induced silencing complex (RISC) complex to guide siRNA to degrade homologous transcripts and also exerts slicer activity to produce the functional strand of siRNA [55]. In N. crassa, QDE2 (AGO) is also involved in the biogenesis of the major milRNAs, milR-1 and milR-2 [38].

We thus generated the deletion mutants Δdcl1, Δdcl2, Δago1 and Δago2 by using homologous recombination methods (electronic supplementary material, figure S2a,b). The growth phenotypes of these mutants on potato dextrose agar (PDA) were different from those of wild-type V592 (electronic supplementary material, figure S2c), suggesting that these RNAi components play roles in hyphal and spore developmental processes in V. dahliae. Accumulation of VdmilR1 was examined. Surprisingly, no decreases in VdmilR1 accumulation were detected in either deletion mutant (electronic supplementary material, figure S2d). By contrast, the VdmilR1 level was increased in Δdcl1, Δdcl2 and Δago2 compared to that of wild-type V592 (electronic supplementary material, figure S2d). These results indicate that the synthesis of VdmilR1 was independent of these canonical RNAi proteins in V. dahliae. The notably increased VdmilR1 in Δdcl1, Δdcl2 and Δago2 mutants ruled out redundant functions and suggested that DCLs and AGO2 might have negative effects on the production of milRNAs, at least VdmilR1, in V. dahliae.

To investigate whether another RNaseIII domain-containing protein was responsible for VdmilR1 synthesis, we searched the V. dahliae genome and found VdR3(VDAG_04981), which contained an RNaseIII domain without other known conserved domain. We were unsuccessful at obtaining a VdR3 deletion mutant, suggesting that the VdR3 coding gene is an essential gene for V. dahliae. We therefore overexpressed VdR3 under the Tef promoter in wild-type V592 to obtain the Tef-VdR3 strain, in which higher expression of VdR3 was detected (electronic supplementary material, figure S2e). Tef-VdR3 colonies exhibited no obvious developmental defects but showed dense growth of hyphae on PDA medium (figure 2a). Small RNA hybridization showed that the mature VdmiR1 level was markedly increased in Tef-VdR3 strains (figure 2d), indicating that VdR3 plays a role in VdmilR1 biogenesis. Together with the previous finding of a Dicer-independent non-canonical pathway for milRNA maturation in Mucor circinelloides, which requires the RNaseIII-like protein R3B2 [31], our data reveal that an RNaseIII domain-containing protein VdR3 is also involved in the biogenesis of milRNAs, at least VdmilR1, biogenesis in V. dahliae.

(e). VdmilR1 targets a hypothetic gene, VdHy1, by epigenetic regulation

To investigate the biological functions of VdmilR1 and to identify the genes of V. dahliae targeted by VdmilR1, we combined computational prediction with RNA-ligase-mediated-5′ rapid amplification of cDNA ends (RLM-5′ RACE). Several genes were predicted using the transcript sequences of the VdLs.17 strain. However, none of the predicted targets were verified by RLM-5′RACE, suggesting that VdmilR1 might not mediate target cleavage. We therefore compared the mRNA levels of these genes between the wild-type V592 and Tef-VdMILR1 strains. Reduced mRNA expression was detected in Tef-VdMILR1 for one predicted target, VdHy1 (VDAG_08333) (figure 2e), which encodes a hypothetic protein with neither functional annotations nor known protein domains. Consistent with this finding, we detected increased expression of VdHy1 in the Δvdmilr1 strain (figure 2e), a VdmilR1-knockout mutant created by homologous recombination method (electronic supplementary material, figure S2a,b).

The predicted VdmilR1 target site was located in the 3′UTR region of VdHy1. As no VdmilR1-mediated cleavage was detected, we hypothesized that reduced transcription of VdHy1 was regulated by VdmilR1 through epigenetic control. Methylated histone H3 Lys9 (H3K9) is a conserved hallmark of heterochromatin [56]. Thus, we examined the status of histone methylation at the VdHy1 locus in the wild-type V592, Tef-VdMILR1 and Δvdmilr1 strains by chromatin immunoprecipitation using H3K9me3 antibody followed by quantitative PCR (ChIP–qPCR). A series of primers at the VdHy1 locus were designed (figure 2f). As shown in figure 2f, compared to V592, the repressive histone mark H3K9me3 in VdHy1 was more enriched in the Tef-VdMILR1 strain and less enriched in Δvdmilr1, demonstrating that reduced VdHy1 transcription in Tef-VdMILR1 resulted from the high level of H3K9me3 enrichment at the VdHy1 locus, while increased VdHy1 transcription in Δvdmilr1 correlated to the lesser H3K9me3 enrichment at the VdHy1 locus.

To validate whether the target site was required for VdmilR1-mediated regulation of VdHy1 expression, we conducted a green fluorescence protein (GFP) reporter sensor assay with the wild-type 3′UTR of VdHy1 and a VdmilR1-resistant mutant 3′UTR of VdHy1 (electronic supplementary material, figure S3a), designated GFP-3′UTR and GFP-3′UTRm, respectively. The expression of GFP-3′UTR and GFP-3′UTRm was examined in the wild-type V592 and Tef-VdMILR1 strains in at least three transformants for each construct. In wild-type V592 transformants, the average expression level of GFP-3′UTR was slightly higher than that of GFP-3′UTRm (electronic supplementary material, figure S3b). This slight difference was presumably owing to the low level of VdmilR1 in V592. Remarkably, in Tef-VdMILR1 transformants, significantly lower average expression level of GFP-3′UTR was detected compared with that of GFP-3′UTRm (electronic supplementary material, figure S3b), indicating that increased VdmilR1 resulted in decreased GFP-3′UTR transcript and that the VdmilR1 target site in the 3′UTR of VdHy1 was required for the GFP-3′UTR reduction. Taken together, our data demonstrate that VdmilR1 targets the 3′UTR of VdHy1 for transcriptional repression through increasing histone H3K9 methylation.

(f). A VdmilR1/VdHy1 module plays a role in Verticillium dahliae pathogenicity in cotton plants

Finally, we tested the functional role of VdmilR1-mediated VdHy1 repression in V. dahliae. We generated a VdHy1 deletion mutant, Δvdhy1 (electronic supplementary material, figure S2a,b), whose colonies showed reduced hyphal growth and melanin production on PDA medium (figure 2a). The Δvdhy1 mutant exhibited markedly reduced virulence in cotton plants without induction of wilt symptoms, in contrast to the wild-type V592 (figure 2g). Both colony morphology and pathogenicity were restored in the Δvdhy1/VdHy1 complemented strain (figure 2a,g). These data demonstrate targeted disruption of VdHy1 and indicate that VdHy1 plays a role in fungal development and is a virulence gene of V. dahliae (figure 2a,g). Consistent with those findings, the VdmilR1-overexpressing Tef-VdMILR1 strain, in which VdHy1 transcription was reduced (figure 2e), also exhibited a certain degree of reduced virulence in cotton plants (figure 2g).

To further investigate the VdmilR1/VdHy1 regulatory module in pathogenicity during cotton infection, we examined the expression patterns of VdmilR1 and VdHy1 at different time points after V. dahliae infection. VdHy1 expression was rapidly induced at early time points and reached a high level at 3 days postinfection (dpi), then was greatly reduced at 10 dpi (figure 2h). The expression level of VdmilR1 was likely maintained at a basal level at the early time points and began to be induced at 3 dpi, then reached a high level at 10 dpi, consistent with the great reduction of VdHy1 at this time point. At 20 dpi, both VdmilR1 and VdHy1 returned to basal levels (figure 2h). These results demonstrate that VdHy1 is required for the early infection and repressed along with increased VdmilR1 at 10 dpi during infection. Together with the virulence effect of VdHy1 on cotton infection (figure 2g), our data reveal a regulatory role of VdmilR1/VdHy1 in V. dahliae pathogenicity in cotton plants.

3. Discussion

In this study, the use of high-throughput sequencing led to the discovery of a large number of sRNAs and several potential milRNA genes in V. dahliae. VdmilR1 detected by RNA gel blotting was functionally studied. We found that transcription of the VdmilR1 precursor VdMILR1 gene depends on Pol II and that an RNaseIII domain-containing protein, VdR3, is likely involved in the synthesis of VdmilR1. VdmilR1 targets a virulence gene, VdHy1, at its 3′UTR for transcriptional silencing through histone H3K9 methylation. A VdmilR1/VdHy1 module plays a regulatory role in pathogenicity during V. dahliae infection.

Even though the potential precursor genes in V. dahliae share similarities with animals' and plants' miRNA features, such as the precursor genes being able to form hairpin structures and the most abundant produced sRNAs residing in the stem region of the hairpin structure, the sRNAs are not named miRNAs but are instead called milRNAs, following the nomenclature in Neurospora [38], because VdmilR1 production in V. dahliae is independent of the canonical RNAi proteins, such as the DCL and AGO proteins in plants and animals, that govern conventional miRNA production [33,57]. We found that VdR3, a putative RNaseIII enzyme, participates in the production of VdmilR1. The exact function of VdR3 in VdmilR1 biogenesis remains to be investigated. VdR3 might have multiple functions. Because overexpression of VdR3 increased VdmilR1, while increased VdmilR1 reduced the virulence of V. dahliae in cotton plants, a VdR3-overexpressing strain increased virulence in cotton plants (electronic supplementary material, figure S2f), showing that VdR3 has other functions in addition to VdmilR1 biogenesis. The fact that VdR3 contains only one RNaseIII domain suggests that VdR3 might form complexes with another factor(s), e.g. DCL2, which lacks one RNaseIII domain. The increased, rather than decreased, VdmilR1 level in some V. dahliae RNAi mutants (electronic supplementary material, figure S2d) implies that mutation of, for instance, DCL2 or AGO2, might release VdR3 and promote a VdR3-containing complex to induce VdmilR1 biogenesis. VdR3 is likely an essential gene for V. dahliae with multiple functions, making V. dahliae RNAi research complicated and challenging. Nevertheless, together with the fact that MRPL3, an RNaseIII domain protein, is an important component of milRNA biogenesis in Neurospora [38] and the fact that an RNaseIII-like protein, R3B2, promotes mRNA degradation in M. circinelloides [31], our findings suggest that putative RNaseIII enzyme-mediated but Dicer-independent sRNA biogenesis pathways commonly exist in the fungal kingdom.

Intriguingly, we have recently found that host miRNAs are exported into V. dahliae, targeting its virulence genes for cleavage [7]. However, in this study, we found that VdmilR1-mediated reduced VdHy1 expression did not result from target cleavage but rather increased histone H3K9 methylation on the target VdHy1 sequence. In the fission yeast S. pombe, sRNA-mediated RITS represses transcription by recruiting a histone methyltransferase to the target genes in heterochromatin [17,18]. VdmilR1 targets the 3′UTR of VdHy1 likely through limited base-pairing interaction, similar to animal miRNAs that usually share less extensive sequence complementarity with their targets at the 3′ UTR [37]. Although animal miRNA-target interactions often result in the inhibition of translation, rather than epigenetic repression at transcriptional level, some mammal miRNAs regulate gene expression at the transcriptional level in the nucleus [58]. miRNAs can directly modulate gene transcription in the nucleus through the recognition of specific target sites in promoter regions [59]. Nevertheless, we cannot rule out that cleavage-independent translational repression exists in V. dahliae, as neither reduced transcript levels nor cleavage were detected for other predicted VdmilR1 targets.

The detection of the time difference in the expression patterns of VdmilR1 and VdHy1coincidently reflects the fungal infection cycle. Verticillium dahliae is a hemibiotrophic fungus that grows preferably in vascular tissues of host plants from the parasitic phase to the saprophytic phase, when conidia and microsclerotium are produced [60,61]. We previously found that, upon successful penetration of V. dahliae into the roots, internal hyphal growth within the xylem vessel cells began at 3 dpi and extensively proliferated by 5–10 dpi [60,61]. The greatly increased VdHy1 at 3 dpi is coincident with the requirement for the initial rapid proliferation of hyphae at this time point. Once fungal biomass increases over time, Verticillium avails itself of VdmilR1 regulation to prevent VdHy1-related virulence in the host and keep its host alive during the biotrophic phase of the infection. Later, regaining basal expression levels of VdmilR1 and VdHy1 is presumably required for conidia and/or microsclerotium formation during the saprophytic phase when wilt symptoms appear, as observed in cotton plants infected with wild-type V592, but not with VdHy1 deletion mutant strains.

Although the exact function of VdHy1 in virulence remains to be investigated, the VdmilR1/VdHy1 regulatory module provides essential information as well as an example case for further study of milRNA functions in fungal growth and pathogenicity. Our data demonstrate that RNAi machinery in V. dahliae also operates at an epigenetic repression level, in addition to mRNA degradation mediated by plant host-exporting miRNAs, such as miR166 and miR159 [7]. Our finding also provides a new strategy to investigate key trans-kingdom miRNAs/sRNAs involved in plant–fungi interaction and to exploit new efficient targets to protect cotton against Verticillium wilt disease, a major threat to the cotton industry.

4. Methods

(a). Fungal culture conditions and infection assays

The virulent defoliating V. dahliae isolate V592, isolated from cotton in Xinjiang, China, was used in this study. The culture conditions of V592 and the conidia production for infection assays were described previously [61]. For plant infection, cotton plants (Xinluzao No. 16) were used in infection assays to evaluate the effect of V592 and transformants on virulence using our laboratory's unimpaired root-dip inoculation method, as described in our previous research [62]. The infection assay was repeated at least three times. Disease grades on cotton leaves were classified into five levels of severity of disease symptoms during fungal invasion: 0 = no visible wilting or yellowing symptoms, 1 = one or two cotyledons wilted or dropped off, 2 and 3 = one or two true leaves wilted or dropped off, 4 = all leaves dropped off or the whole plant died.

(b). Identification of the pivotal components of the silencing pathway in Verticillium dahliae

Putative homologs of the RDR, DCL and AGO proteins in V. dahliae were searched for with the Search Conserved Domains program (Batch CD-search) [63], using V. dahliae protein sequences (ASM15067v2) as input. The amino acid sequence of each potential RNA-silencing protein was aligned with ClustalX2 [64]. A maximum-likelihood phylogenetic tree was constructed with aligned sequences by Molecular Evolutionary Genetics Analysis (MEGA) [65]. Dendro scope [66] was used to draw the tree.

(c). Small RNA deep sequence and data mining

The raw reads, which removed the adaptor and low-quality reads, were mapped to the Verticillium spp. genome (PRJNA28529), using Short Oligonucleogide Analysis Package (SOAP) alignment software [67]. The perfectly matched sRNAs, Vd-sRNAs, with lengths between 18 and 30 nt, were included in our analysis. Mireap (https://sourceforge.net/projects/mireap/) and miRDeep [48] were used to appraise potential miRNAs. The expression of potential miRNAs was normalized by the reads per million (RPM). The targets of miRNAs were predicted by psRNATarget [68] with default parameters.

(d). Construction and transformation

To generate the knockout plasmids pGKO-dcl1, pGKO-dcl2, pGKO-ago1, pGKO-ago2, pGKO-VdHy1 and pGKO-VdMILR1, upstream and downstream genomic sequences were amplified with the specific primer pairs: dcl1-up-F/R, dcl1-dn-F/R; dcl2-up-F/R, dcl2-dn-F/R; ago1-up-F/R, ago1-dn-F/R; ago2-up-F/R, ago2-dn-F/R; VdHy1-up-F/R, VdHy1-dn-F/R; VdMILR1-up-F/R, VdMILR1-dn-F/R (electronic supplementary material, table S1). Both sequences were inserted into a position flanking of the hygromycin resistance cassette of the vector pGKO with the USER enzyme to generate knock-out plasmids, and transformation was performed as described previously [69,70].

To generate the overexpression plasmids Tef-VdMILR1 and Tef-VdR3, the wild-type genes VdMILR1 and VdR3 were amplified from the V592 genome with the primer pairs of VdMILR1-if-F/R and VdR3-if-F/R, respectively (electronic supplementary material, table S1). The PCR products were inserted into an EcoRI/BamHI-linearized pNEO binary vector under the constitutive promoter Tef by infusion enzyme [9,62]. Agrobacterium tumefaciens-mediated transformation was conducted as previously described [62].

To generate the complementary plasmid pNEO-VdHy1, the genomic sequence of VdHy1, including 2 kb upstream of the start codon, was amplified with the primer pair VdHy1-HE-F1/R1 (electronic supplementary material, table S1). The fragment was inserted into a HindIII/EcoRI-linearized pNEO binary vector. pNEO-VdHy1 plasmid was transformed into knockout mutant Δvdhy1 to produce the complemented strain Δvdhy1/VdHy1.

To generate GFP reporter sensor constructs, the wild-type 3′UTR of VdHy1 was amplified with primer pair: VdHy1-3′UTR-if-F1/R1 (electronic supplementary material, table S1). The fragment was inserted into EcoRI/BamHI-linearized pNEO to produce the GFP-3′UTR plasmid. To generate the VdmilR1-resistant mutant, the primers VdHy1 m-F/R were used, and plasmid GFP-3′UTRm was produced by the QuickChange Lightening Site-Directed Mutagenesis Kit (210519, Agilent, Santa Clara, CA, USA) according to the manufacturer's instructions.

(e). RNA polymerase II and RNA polymerase III inhibitor treatments

α-Amanitin (A2263, Sigma-Aldrich, St. Louis, MO, USA) was used as an inhibitor of RNA polymerase II transcription. ML-60218 (557403, Millipore, Billerica, MA, USA) was used as an inhibitor of RNA polymerase III transcription. α-Amanitin was dissolved in distilled water at a concentration of 1 mg ml⁻¹. The final concentration used in tissue treatment was 2 µg ml⁻¹. ML-60218 was dissolved in DMSO at a concentration of 20 mg ml⁻¹. The final concentration used in tissue treatment was 20 µg ml⁻¹. Culture medium containing an equal volume of H₂O or DMSO was used as a control. Conidia were germinated and grown for 3 days in liquid Czapek-Dox medium before the addition of α-Amanitin or ML-60218. Tissue was treated at room temperature for 2 h in the dark (to prevent light degradation of the pol III inhibitor) before harvest.

(f). RNA extraction, small RNA gel blotting and quantitative real-time PCR analysis

For small RNA gel blotting, total RNA was extracted using the TRIzol reagent according to the manufacturer's instructions and dissolved in 50% formamide deionized. A total of 40–60 µg of total RNA was separated by electrophoresis on 17% denaturing polyacrylamide gel and electrically transferred to Amersham Hybond™-N⁺ membrane (RPN303B, GE Healthcare, USA). The probes were labelled with [γ-³²P]ATP using polynucleotide kinase (M0201 V, NEB, USA). Signal intensity was quantified using ImageQuant TL software (GE Healthcare, USA).

For qRT-PCR of high-molecular-weight RNA, total RNA was extracted using an RNA extraction kit (TR02-150, GeneMark, Taiwan, China) and dissolved in RNase-free H₂O. Two micrograms of total RNA was reverse-transcribed into cDNA using the GoScript™ Reverse Transcription System (Promega, Madison, WI, USA). qPCR was performed using EvaGreen qPCR mastermix (abm, Vancouver, Canada). The Verticillium tubulin beta chain gene (VDAG_10074) was used as the internal control. For miRNA qRT-PCR, total RNA was extracted using miRNA extraction kit (DP501, TIANGEN, Beijing, China) and 1 µg of total RNA was reverse-transcribed into cDNA using the miRcute Plus miRNA First-Strand cDNA Synthesis Kit (KR211, TIANGEN, Beijing, China). qPCR was performed using the miRcute Plus miRNA qPCR Detection kit (FP411, TIANGEN, Beijing, China).The Verticillium U6 gene was used as an endogenous control. qRT-PCR analysis was performed in a Bio-Rad CFX96 Real-Time system. The primers were VdR3-qRT-F/R; VdHy1-qRT-F1/R1; VdTub-qRT-F/R; eGFP-qRT-F/R; VdmilR1-qRT-F and VdU6-qRT-F (electronic supplementary material, table S1). At least three biological replicates and three technical replicates within an experiment for each sample were performed.

(g). Chromatin immunoprecipitation assays

Fungal spores (0.5 g fresh weight) were harvested and used for the chromatin immunoprecipitation (ChIP) assay. ChIP was performed as described [71]. The wild-type V592 and the transformed lines Tef-VdMILR1 and Δvdmilr1 were cultured in liquid Czapek-Doxmedium for 3 days following filtering with Miracloth (Millipore, Billerica, MA, USA). Acquired spores were fixed with 1% formaldehyde for 15 min. For IP, 1 µg antibody of H3K9me3 (ab8898, Abcam, Cambridge, UK) was used to pull down chromatin containing the specific methylation mark. Then, qPCR was performed to determine the enrichment of VdHy1 DNA immunoprecipitated in the ChIP. Gene-specific primers were VdHy1-pro-F/R, VdHy1-qRT-F1/R1 and VdHy1-qRT-F2/R2 (electronic supplementary material, table S1). After amplifying VdHy1 regions and the reference gene VdTub region from the purified enriched DNA sample and from the input, the relative enrichment was calculated by using the $2^{-\mathrm{\Delta }\mathrm{\Delta }{C}_t}$ method as follows: ΔC_t (input sample) = C_t (target region) – C_t (reference region); ΔC_t (enriched DNA sample) = C_t (target region) – C_t (reference region); ΔΔC_t = ΔC_t (enriched DNA sample) – ΔC_t (input sample); fold-ratio enrichment enriched DNA sample/inputsample = $2^{-\mathrm{\Delta }\mathrm{\Delta }{C}_t}$ [72].

Data accessibility

This article has no additional data.

Authors' contributions

H.-S.G. and Y.J. designed experiments. Y.J., P.Z., T.Z. and S.W. performed experiments. J.-H.Z. performed sRNA computational informatics analysis. H.-S.G., Y.J. and J.-H.Z. analyzed data. H.-S.G. and Y.J. discussed the results and wrote the paper.

Competing interests

The authors have declared that no competing interests exist.

Funding

This work was financially supported by the Strategic Priority Research Program of the Chinese Academy of Sciences (XDB11040500 to H.-S.G.), National Natural Science Foundation of China (31730078 to H.-S.G. and 31700131 to Y.J.)