Redundancy of the Two Dicer Genes in Transgene-Induced Posttranscriptional Gene Silencing in Neurospora crassa

Abstract

RNA interference (RNAi) in animals, cosuppression in plants, and quelling in fungi are homology-dependent gene silencing mechanisms in which the introduction of either double-stranded RNA (dsRNA) or transgenes induces sequence-specific mRNA degradation. These phenomena share a common genetic and mechanistic basis. The accumulation of short interfering RNA (siRNA) molecules that guide sequence-specific mRNA degradation is a common feature in both silencing mechanisms, as is the component of the RNase complex involved in mRNA cleavage. During RNAi in animal cells, dsRNA is processed into siRNA by an RNase III enzyme called Dicer. Here we show that elimination of the activity of two Dicer-like genes by mutation in the fungus Neurospora crassa eliminates transgene-induced gene silencing (quelling) and the processing of dsRNA to an siRNA form. The two Dicer-like genes appear redundant because single mutants are quelling proficient. This first demonstration of the involvement of Dicer in gene silencing induced by transgenes supports a model by which a dsRNA produced by the activity of cellular RNA-dependent RNA polymerases on transgenic transcripts is an essential intermediate of silencing.

RNA-mediated gene silencing refers to a series of mRNA sequence-specific degradation mechanisms occurring in eukaryotic cells. Often, mRNA degradation is activated as a consequence of the introduction of either DNA or RNA. The first evidence of these phenomena was obtained in plants in which the introduction of transgenes induced posttranscriptional gene silencing (PTGS) of homologous endogenous genes in a mechanism called cosuppression (36). Transgene-induced gene silencing phenomena are widespread among plants, and analogous mechanisms are present also in fungi (7). mRNA degradation can be induced not only by transgenes but also by the introduction of double-stranded RNA (dsRNA) molecules in the well-known phenomenon of RNA interference (RNAi) (23). These gene silencing phenomena induced either by transgenes or dsRNA share common features, suggesting that these mechanisms may originate from an ancestral process probably devoted to the host genome defense against invasive molecules of nucleic acids, such as transposons and viruses (29, 38).

A key feature of the PTGS phenomena is the accumulation of small RNA molecules, 21 to 25 nucleotides (nt) long, called short interfering RNAs (siRNAs). These small RNAs were first identified in association with PTGS in plants (20). Subsequently, it was shown that, when added to Drosophila extracts, large dsRNA molecules are processed into small RNA molecules with a similar size of 21 to 23 nt (15, 35). These siRNAs are subsequently incorporated into a large multiprotein complex, termed the RNA-induced silencing complex (RISC), which guides the degradation of homologous mRNA molecules by cleaving them at the center of the region that is complementary to the guide siRNAs (21).

Biochemical analysis of the RISC allowed the identification of the fundamental protein component Argonaute-2 (22) that has been proposed to have an “adapter” function in binding siRNAs, committing them to the RNAi pathway. Argonaute-2 belongs to a family of proteins that are characterized by a Piwi domain and a PAZ (Piwi/Argonaute/Zwille) domain (4) and are, therefore, referred to as the PPD proteins (PAZ and Piwi domain). Strikingly, genetic approaches have led to the identification of other PPD proteins that are necessary for both RNAi in Caenorhabditis elegans (RDE-1) (34) and transgene-induced silencing in Neurospora crassa (QDE-2) (5), indicating an extremely high level of conservation of the mRNA degradation machinery among different gene silencing pathways.

Studies carried out in Drosophila demonstrated that siRNAs are generated by the action of a dsRNA-specific endonuclease (1). This enzyme, named Dicer due to its ability to digest dsRNA into uniformly sized small RNAs, belongs to an RNase III subclass characterized by two RNase domains as well as an ATP-dependent RNA helicase domain. Typically, Dicer cleaves dsRNA into 21-nt fragments with 2-nt 3′ overhangs, and in Drosophila it has been shown to interact directly with the RISC complex (22). It has been suggested that the interaction of Dicer with the RISC complex is required in order to supply the RISC with siRNA. Moreover, it has been found that Dicer is involved in the biogenesis of endogenous small RNAs, or micro-RNAs (miRNAs) (24), which are involved in translational control of several mRNAs required in development (19, 26, 27).

Although the Dicer family of proteins is conserved among all eukaryotes and is required for RNAi, as yet no evidence has been provided that this class of enzyme is also required for gene silencing phenomena induced by transgenes. The fact that 21- to 25-nt RNAs in both sense and antisense orientation are accumulated in transgenic plants (20) and fungi (6) that show gene silencing has led to the following proposals: first, a dsRNA molecule must be produced as a consequence of the presence of the transgene and, second, analogously to what has been observed in RNAi, such dsRNA must be processed by Dicer. Current models (11, 37) propose that transgenes may activate gene silencing by producing aberrant single-stranded transcripts that are converted in dsRNA by cellular RNA-dependent RNA polymerases (RdRPs). Consistent with these models, putative RdRPs have been found to be essential for transgene-induced PTGS both in plants (SGS2/SDE1) (13, 30) and fungi (QDE1) (9). However, it is still uncertain how RdRPs recognize aberrant RNAs as templates and whether RdRPs are indeed able to produce dsRNAs.

Recent data have shed light on the activity of QDE1, an RdRP homologue involved in PTGS (quelling) in the fungus N. crassa. Purified recombinant QDE1 protein catalyzes RNA-dependent RNA polymerization on single-stranded RNA (ssRNA) templates (28). However, QDE1 appears to possess two different activities: it either synthesizes long RNA chains that produce template-length RNA duplexes, or it synthesizes short 9- to 21-mer cRNA oligonucleotides distributed along the entire ssRNA template. The polymerization of short RNAs appears significantly more efficient than the synthesis of full-length dsRNA, suggesting that this reaction mode may be relevant for the QDE1 function in vivo. The fact that at least a subset of the short RNA molecules produced by QDE1 have a size similar to that of siRNAs has suggested that QDE1 may directly provide the RISC nuclease complex with short RNA guides (28). According to this model, a Dicer-like activity could either not be required at all for quelling in Neurospora or, alternatively, may have only an ancillary function in increasing quelling efficiency, perhaps by producing additional siRNA on the limited amount of dsRNA synthesized by QDE1.

Previous extensive genetic screens carried out in N. crassa have identified only three loci (qde-1, qde-2, and qde-3) essential for PTGS (10), and the fact that a putative Dicer gene was not included among these may reinforce the notion that Dicer is dispensable in quelling. Moreover, recent data have shown that the inactivation of the Arabidopsis Dicer homologue, CARPEL FACTORY, does not compromise transgene-induced gene silencing (18). However, since four different Dicer homologues are present in the Arabidopsis genome, a certain degree of redundancy may be expected, making it difficult to prove participation of the Dicer homologues in PTGS in plants.

Here, we have investigated the involvement of a Dicer-like protein in the PTGS mechanism in N. crassa. We first demonstrated that dsRNA molecules incubated in vitro with N. crassa cell extracts were processed into short RNA with a size of ∼21 nt. Consistent with a Dicer-like activity, the processing of dsRNA was found to be energy dependent. Moreover, we found that the Dicer-like activity is constitutively present in Neurospora independent of either the activation of gene silencing or a functional gene silencing machinery. Searches of the Neurospora genome sequence revealed the presence of two Dicer-like homologous genes (dcl-1 and dcl-2). Gene replacement was used in order to produce both single (dcl-1 or dcl-2) and double (dcl-1/dcl-2) mutants. We found that while single mutants are quelling proficient, the double Dicer mutant was completely impaired in quelling, indicating that Dicer activity is absolutely required for gene silencing in Neurospora but that the two Dicer genes are redundant in the quelling pathway. Consistent with these results obtained in vivo, cell extracts from single mutants but not the double mutant had processing activity.

MATERIALS AND METHOD

N. crassa strains

The N. crassa wild-type (WT) strains 74-OR23A and 74-OR8a were obtained from the Fungal Genetics Stock Center (FGSC), University of Kansas, Kansas City (strains FGSC 987 and FGSC 988, respectively), and used for all transformation experiments. The qde mutants were previously described (10).

Media and growth conditions

Strains were grown in Vogel's minimal medium for Neurospora (NMM), as described elsewhere (14), plus benomyl (1 μg/ml) or hygromycin (at a concentration of 0.2 mg/ml in slant and liquid media or 0.3 mg/ml in solid media), as required. Ascospores from crosses were heat activated to induce germination at 60°C for 30 min, as previously described (14).

To obtain forced heterokaryon strains, Basta-resistant conidia from a WT strain were mixed with benomyl-resistant conidia from the Dicer transformant (DT) strain and incubated on minimal media plus 0.1 mM Basta (Aventis-Italia SpA) and 1 μg of benomyl per ml. This double selection allows growth of only forced heterokaryotic strains.

Isolation and characterization of Dicer mutants

To obtain a dcl-1⁻ strain, we performed site-specific insertional mutagenesis by transforming the N. crassa WT mating type a strain (74-OR8a) with a linear DNA fragment containing two sequences homologous to the upstream and downstream regions of the dcl-1 genomic locus on either side of a hygromycin resistance cassette (Fig. 1).

FIG.1.

dcl-1 and dcl-2 knockouts. (A) Schematic depiction of dcl-1 and dcl-2 loci before and after homologous recombination with a linear DNA molecule. White box, genomic dcl-1 and dcl-2 loci; black box, hygromycin B resistance cassette; grey box, region of homology between dcl gene flanking sequences and the branches of the linear DNA; up1 and up2, sequences upstream of dcl-1 and dcl-2 loci, respectively; down1 and down2, sequences downstream of dcl-1 and dcl-2 loci, respectively; arrows, restriction sites for XhoI and BamHI; solid lines, predicted bands; dashed lines, ³²P-labeled probes. (B) Southern blot analysis of WT and four independent dcl-1 transformants. Genomic DNA was digested with XhoI and hybridized with a DNA probe complementary to the upstream region of the dcl-1 gene. Transformant in lane 12 shows a band of ∼2.4 kb predicted for the dcl-1 knockout. (C) Southern blot analysis of WT and eight independent dcl-2 transformants. Genomic DNA was digested with BamHI and hybridized with DNA complementary to the upstream region of the dcl-2 gene. Transformants in lanes 6 and 19 show a band of ∼2.0 kb predicted for the dcl-2 knockout. (D) Identification of the dcl-1/dcl-2 double mutant. The contemporaneous presence of dcl-1 and dcl-2 mutations was monitored by a double Southern analysis of genomic DNAs from the WT strain and five ascospores originating from a cross between dcl-1⁻/helper and dcl-2⁻ strains. The Southern blotting conditions used to identify both the dcl-1⁻ locus and dcl-2⁻ locus are as reported in Materials and Methods. The dashed arrows point to the predicted sizes of the dcl-1 and dcl-2 knockouts. The stars indicate the double mutant strain.

A Southern analysis (Fig. 1) on a WT strain and four independent hygromycin-resistant transformants allowed us to isolate the knocked out strain. Genomic DNA preparations from WT and four transformant strains were digested with XhoI and hybridized with a DNA ³²P-labeled probe corresponding to the genomic region upstream of the 5′ region used to direct recombination at the dcl-1 locus. The presence of unique XhoI sites in both the endogenous, nondisrupted dcl-1 gene and in the hygromycin cassette of the replacement construct allows discrimination between recombinant and WT loci. The recombinant strain displaying the correct hybridization pattern was then purified. A similar procedure was used to disrupt the dcl-2 gene into the N. crassa WT mating type A strain (74-OR23A). The linear DNA fragment used to knock out this gene contains the flanking 5′ and 3′ noncoding regions of dcl-2 harboring the hygromycin cassette in place of the open reading frame. In order to identify a disrupted dcl-2 gene, DNA preparations of eight different transformants were analyzed by Southern blotting. The probe comprised a region upstream of the 5′ region used to direct recombination at the dcl-2 locus that recognizes both WT and dcl-2⁻ loci but can differentiate between the two. A BamHI restriction site that was present in the hygromycin gene and was used as part of the construct to disrupt the dcl-2⁻ gene was absent in the WT, which made such differentiation possible. Two independent dcl-2⁻ strains were recovered (Fig. 1).

Preparation and analysis of N. crassa cell extracts for Dicer activity

N. crassa extracts were prepared from mycelia after they were powdered in liquid nitrogen exactly as previously described (17). Extracts were not pretreated with micrococcal nuclease in Dicer experiments. dsRNA templates were prepared by annealing an equimolar concentration of sense and antisense RNA originating from the radiolabeled in vitro transcription of a portion of the al-1 gene cloned in both orientations downstream of a plasmid T3 RNA polymerase promoter. ssRNA was removed from dsRNA by treatment with RNase One (Promega). The susceptibility of ssRNA and the resistance of dsRNA to RNase One were verified by gel electrophoresis. Following nuclease treatment, dsRNA was extracted with phenol-chloroform, precipitated with ethanol, and resuspended in water. dsRNA or single-strand sense and antisense RNA controls were incubated with extracts (final volume, 50 μl) under standard reaction conditions (17). Not adding exogenous ATP, GTP, creatine phosphate, and creatine kinase to the reaction mixtures accomplished energy depletion. Following the incubation period, samples (15 μl) were removed and deproteinized by extraction with phenol-chloroform. This step was accomplished by adding the samples to tubes containing 40 μl of phenol-chloroform and 25 μl of water and vortexing immediately. After separating the phases by centrifugation, aliquots of the aqueous phase were combined with denaturing loading buffer, denatured by heating to 95°C for 5 min, and examined by denaturing gel electrophoresis in urea-acrylamide gels (15 and 6% acrylamide gels were used for the analysis; see Fig. 2 and 5, respectively). Radiolabeled species were visualized by phosphorimaging.

FIG. 2.

Dicer activity in N. crassa extracts. Denaturing acrylamide gel analysis of the dsRNA cleavage assay of different N. crassa strain lysates incubated with ³²P-radiolabeled double-stranded molecules for the times indicated. (A) The WT lysate was incubated in the presence (+) or absence (−) of an energy regeneration system. (B) ³²P-radiolabeled dsRNA was incubated with protein extracts from the WT, a quelled strain, and the qde-1⁻, qde-2⁻, and qde-3⁻ mutant strains, and the progression of the reactions was monitored at different times. The size of the RNA was estimated by using 25-nt DNA oligonucleotides as size markers.

FIG. 5.

Dicer activity in N. crassa with analysis of WT, dcl-1, dcl-2, and dcl-1/dcl-2 strains. Cell extracts were incubated with radiolabeled dsRNA for 0, 30, or 90 min (T₀, T₃₀, and T₉₀, respectively) in the presence (+) or absence (−) of an energy regenerating system as indicated, and the RNA was examined by denaturing gel electrophoresis as described in Materials and Methods. Decade RNA markers (Ambion) labeled with ³²P were used as size standards (M).

Plasmid constructions

PCR amplification of WT N. crassa DNA was carried out by using Amplitaq DNA polymerase (Perkin-Elmer) with pairs of forward and reverse primers for upstream and downstream regions of both dcl-1 and dcl-2 loci. Forward and reverse primers for the upstream dcl-1 region contained KpnI and XhoI restriction sites, respectively, whereas forward and reverse primers for its downstream region contained HindIII and NotI sites, respectively. The corresponding primers for dcl-2 have the following restriction sites: KpnI and XhoI, upstream forward and reverse primers, respectively; and SpeI, and NotI, downstream forward and reverse primers, respectively. The following primers were used (added restriction sites are underlined): D1 forward upstream, 5′-CGG GGT ACC CCG CAA AGC TCA TCT GTA AGT GG-3′; D1 reverse upstream, 5′-CGC CGC TCG AGC GTG GTG AAG TGA TGA TTC TC-3′; D1 forward downstream, 5′-AGA AGG CTA TCA AGC TTT TGG AGG GCA TGA GCG T-3′; D1 reverse downstream, 5′-ATA AGA ATG CGG CCG CAA ATG CCG AAG GAA GCA ACA-3′; D2 forward upstream, 5′-CGG GGT ACC CGA GGA ATC ACT TCG TGA TAA G-3′; D2 reverse upstream, 5′-CGC CGC TCGAGT TGG GCA TCT GGC TTC AAG C-3′; D2 forward downstream, 5′-GGA CTA GTC AGT CTC TAC TCC GGT CAT G-3′; and D2 reverse downstream, 5′-ATA AGA ATG CGG CCG CTT GCA ACG AGC GCA GAT TGC C-3′.

The PCR products were gel purified, digested with the appropriate enzymes, and both upstream and downstream fragments for a single locus were ligated into plasmid pCSN44 (33a) to place a fragment on each side of the hygromycin resistance expression cassette. These plasmids were double digested with KpnI and NotI. It was possible by gel separation to purify linear constructs without vector sequences; these constructs were used to transform two mating types of N. crassa (mata for dcl-1 and matA for dcl-2) to hygromycin resistance, thereby knocking out dcl-1 and dcl-2, respectively.

The pIR plasmid was constructed by inserting an inverted repeat corresponding to the al-1 gene into an N. crassa expression cassette under the control of the qa-2 gene promoter. The inverted repeat was obtained by ligating, in the opposite orientation, two PCR fragments corresponding to the al-1 gene sequence. The two PCR fragments were obtained by using a common primer at the 5′ end complementary to bases 1053 to 1072 of the al-1 gene and containing a XmaI restriction site (5′-TCC CCC CGG GGG GAT ACC GCT TCG ACC AAG GTC C-3′; restriction site is indicated by underlined nucleotides) and two different primers (5′-TGC CGG AAT TCC GGC ACG TTG AAG AAG TCG TGG-3′ and 5′-ATC CGG AAT TCC GGA TCA TCA GGC AGG GCC TGC-3′) corresponding to the 3′ end and each containing an EcoRI restriction site (at underlined nucleotides) complementary to bases 2380 to 2398 and 1969 to 1990 of the al-1 gene, respectively. The PCR products were digested with EcoRI and ligated to each other. The resulting inverted repeat was digested with XmaI and cloned in plasmid pMYX2 (3) under the control of the qa-2 promoter.

Southern blotting

To identify knockout strains, 5 μg of chromosomal DNA digested with different restriction enzymes (XhoI or BamHI) was fractionated by electrophoresis on a 0.8% agarose gel. The DNA was transferred onto GeneScreenPlus (New England Nuclear) filters by capillary blotting. Filters were prehybridized and hybridized at 65°C according to GeneScreenPlus procedures. The ³²P-labeled probes used were prepared by using a Random Primed DNA labeling kit (Roche) as described by the manufacturer. Diagrams of the fragments used as probes are shown in Fig. 1A. Both probes were 200 nt long and were synthesized by PCR amplification with 20-mer oligonucleotides from chromosomal DNA. The following primers were used: D1 forward probe, 5′-AAG GGG TAT GTT AGT GTG G-3′; D1 reverse probe, 5′-CAT TGT CTT GGT AGT GAA GG-3′; D2 forward probe, 5′-TGG GCT GTA TGG TTT TGG GG-3′; and D2 reverse probe, 5′-TCG AGG TTT CTT ACA CCG GG-3′.

Northern blotting

Total RNA was extracted from frozen mycelia. The RNA was electrophoresed on agarose gel, transferred onto GeneScreenPlus filters, and probed with a riboprobe specific for the antisense al-1 sequence. RNA probe was transcribed in vitro with ³²P-labeled UTP (50 μCi per 20-μl reaction volume; specific activity, 3,000 Ci/mmol) (New England Nuclear). Filter hybridization was performed at 60°C overnight in a hybridization solution containing 5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), 5× Denhardt's solution, 50% formamide, 1% sodium dodecyl sulfate (SDS), and 15 μg of salmon sperm per ml. Membrane was washed twice in a solution containing 2× SSC and 0.1% SDS for 15 min at room temperature, twice with the same solution at 42°C for 15 min, and in a solution containing 0.1× SSC and 0.1% SDS for 15 min at 68°C. Finally, membrane was incubated at 37°C for 30 min with 25 μg of RNase A per ml and 10 U of RNase T₁ per ml to remove unspecific background.

Small RNA purification and hybridization were performed as described elsewhere (6).

RESULT

Identification and characterization of Dicer-like activity in N. crassa protein extracts

Protein extracts obtained from N. crassa-germinated conidia were tested for the presence of a dsRNA processing activity in in vitro assays. Radiolabeled dsRNA was incubated with N. crassa lysates at 25°C, and the time course of the reaction was monitored after 0, 30, and 90 min. The result of the dsRNA cleavage assay was then analyzed by denaturing acrylamide gel electrophoresis.

It has been shown that the Dicer activity of both Drosophila embryo lysates (31) and of immunoprecipitated Dicer protein (1) require an ATP regenerating system containing GTP, ATP, creatine phosphate, and creatine kinase for efficient dsRNA processing. To verify the necessity of ATP in the N. crassa system, we repeated the dsRNA cleavage assay in the presence and absence of this ATP regenerating system.

As shown in Fig. 2A, we found that dsRNA molecules were processed when incubated with N. crassa protein extracts into short RNA fragments with a discrete size of approximately 25 nt. The dsRNA processing requires the dynamic production of ATP (Fig. 2A), just as for Dicer activity in Drosophila. A constant source of energy is probably required for this family of proteins to resolve the complex structures of the dsRNA molecules to permit dsRNA processing.

The Dicer activity was also assayed in a transgenic strain that exhibited PTGS. As shown in Fig. 2B, the level of dsRNA processing activity is similar to that found in a WT untransformed strain, suggesting that an ongoing silencing is not required in order to activate the Dicer activity that, by contrast, appears constitutively present. Moreover, we also analyzed extracts from three quelling-defective (qde) strains (Fig. 2B). These mutants are defective in various genes (qde-1, qde-2, and qde-3) required in different steps of the quelling pathway: qde-1 gene encodes an RdRP, qde-2 corresponds to a protein of the PPD family which is predicted to be part of the RISC complex, and qde-3 encodes a RecQ DNA helicase whose function has still to be defined but may be involved in transgene recognition. We observed that a dsRNA processing activity is still present in all three qde mutants, indicating that the Dicer functions independently from these three components of the silencing machinery and, more generally, that it does not require a functional quelling apparatus.

Dicer genes in N. crassa

Using a BLASTP algorithm, we searched the database of the nearly complete N. crassa genome and found two putative Dicer homologues, corresponding to two annotated putative proteins in the Whitehead Institute database (www-genome.wi.mit.edu/annotation/fungi/neurospora/). The alignment of these putative proteins by the ClustalW program shows a reciprocal percentage of identity of 23% and a percentage of identity with the Drosophila Dicer-1 protein of 23% for one predicted protein and 26% for the other. The two Dicer-like genes were then called dcl-1 (NCU08270.1) and dcl-2 (NCU06766.1). Both predicted proteins contain all four domains characteristic of the Dicer family of proteins. Two RNase III domains (RNase IIIa and RNase IIIb) are within the C-terminal portion, and the RNA helicase and DEAD-box ATP binding domains are within the N-terminal portion. An additional domain with an unknown function (duf283) is present in both proteins. DCL1 contains 1,584 amino acids and has a predicted mass of 179 kDa. DCL2 contains 1,520 amino acids with a predicted mass of 170 kDa. DCL2 also contains a dsRNA binding domain found in most, but not all, Dicer homologues; this domain is missing in DCL1 (Fig. 3).

FIG. 3.

Dicer domains in N. crassa Dicer proteins. Black bars correspond to the full protein sequence. The boxes correspond to the identified domains, each with its starting and stopping amino acid. Both DCL1 and DCL2 of Neurospora contain a DEAD box, a helicase C domain (hel C), a duf283 domain with an unknown function, and two RNase III domains (RNase IIIa and RNase IIIb). DCL2 contains also a dsRNA binding domain. aa, amino acids.

Quelling in Dicer knockout mutants

In order to verify the involvement of Dicer in the quelling mechanism, we knocked out each of the two Dicer genes by using a gene replacement strategy as described in Materials and Methods (Fig. 1A to C).

To assess the involvement of dcl-1 and dcl-2 in quelling, we cotransformed each of the two knockout mutants with plasmid pX16 (10) containing a portion of the al-1 gene involved in carotenoid biosynthesis. Transformation of N. crassa with this plasmid has previously been shown to be able to induce silencing of the endogenous al-1 gene. Silencing of the al-1 gene is easily scored by the appearance of transformants showing an albino (white or yellow) phenotype. Quelling frequency was calculated as the percentage of transformants showing an albino phenotype of the total number of independent transformants examined. As shown in Table 1 both dcl-1 and dcl-2 single mutants showed a quelling frequency of about 20%, which is comparable with the silencing frequency observed in a WT background, indicating that both mutants were proficient in quelling. This result suggested that either (i) Dicer is not required for quelling or (ii) the Dicer genes are redundant in the silencing pathway. In order to distinguish between these possibilities, we created a double dcl-1/dcl-2 mutant and tested it for its ability in quelling. The double mutant was obtained by crossing opposite mating types of the two original single mutants dc1-1⁻ and dcl-2⁻. To overcome the possible problem of female sterility of the mutants, we made forced heterokaryons between each of the two mutants by using a helper strain that can participate nutritionally and functionally in the cross but does not participate genetically. Strains deriving from five ascospores were analyzed by Southern blotting to verify the presence of the two mutated alleles (see Materials and Methods) (Fig. 1D). The identified double mutant strain together with the single mutants originated in the sister progeny were tested for their ability to support quelling. The single Dicer mutants again showed a frequency of gene silencing comparable to that of the WT. In contrast, the double mutant showed no quelling (Table 1), indicating that the two Dicer homologues are an essential part of the silencing pathway but that their function is redundant in the quelling machinery.

TABLE 1.

Quelling efficiency in WT, dcl-1⁻, dcl-2⁻, and dcl-1⁻/dcl-2⁻ mutants

Plasmid	Percentage of transformants showing quelling (total no. of transformants examined)
	WT	dcl-1−	dcl-2−	Sister progenya
				dcl-1−/dcl-2−	dcl-1−	dcl-2−
pX16 + pMYX2	20 (346)	19 (351)	20 (312)	10 (325)	20 (200)	20 (250)
pMYX2	0 (325)	0 (360)	0 (410)	0 (390)	0 (230)	0 (210)

^aStrains derived from the same cross.

dsRNA as intermediate in quelling

Previous results have shown that qde-2 mutant strains accumulate siRNA normally, while qde-1 and qde-3 mutants are both blocked in the accumulation of siRNA (6), indicating that these two genes are required upstream of the production of siRNA. The finding that Dicer is required for quelling implies that dsRNA is an intermediate in the gene silencing pathway. This suggests a model by which the transgenic transcripts are first converted into dsRNA by QDE1 and then the dsRNA is processed by the two DCL proteins to produce siRNA. The siRNA would then be incorporated into a complex (RISC) that also contains QDE2. This model predicts that the expression of dsRNA would enter the quelling pathway downstream of RdRP but that in order to induce silencing it should require both DCL and QDE2. In order to test this model, we expressed a dsRNA as a hairpin RNA molecule by introducing a plasmid containing an inverted repeat (pIR) corresponding to the al-1 gene under the control of the qa-2 gene promoter, which is inducible by quinic acid. pIR was introduced by transformation into several Neurospora strains containing different genetic backgrounds. Both qde-1 and qde-3 mutants transformed with pIR displayed quelling proficiency (Table 2). Thus, the RdRP activity of QDE1 is dispensable when a preformed dsRNA is expressed. By contrast, the qde-2 mutant remained completely impaired in gene silencing, confirming that the qde-2 gene is required downstream of siRNA accumulation and that this gene is also necessary for dsRNA-induced gene silencing. Moreover, consistent with the redundancy of the two Dicer homologues in the silencing pathway shown above, only the dcl-1/dcl-2 double mutant is defective in silencing mediated by dsRNA, while the two single mutants remained proficient.

TABLE 2.

dsRNA-induced gene silencing

Strain	No. of silenced transformants/total no. of transformants	Silencing abilitya
WT	54/70	+
qde-1−	87/112	+
qde-2−	0/85	−
qde-3−	57/83	+
dcl-1−	130/180	+
dcl-2−	63/81	+
dcl-1−/dcl-2−	0/73	−

^a+, quelling proficiency; −, impaired gene silencing.

The double Dicer mutant is impaired in dsRNA processing both in vivo and in vitro

The observation that the dcl-1/dcl-2 double mutant was defective both in transgene-induced and dsRNA-induced gene silencing led us to the hypothesis that the two Dicer proteins were involved in cleaving dsRNA molecules into siRNAs. To further investigate this point, we analyzed at the molecular level whether the dcl-1/dcl-2 mutant is indeed blocked in dsRNA processing. The dcl-1/dcl-2 mutant was transformed with the pIR plasmid in order to produce a dsRNA corresponding to the al-1 gene. Because this mutant is blocked in dsRNA-induced gene silencing, the resulting transformed strain (called DT1) retains a WT (orange) phenotype due the expression of the endogenous al-1 gene. The strain DT1 was then used to produce a heterokaryotic strain with a WT untransformed strain. The resulting heterokaryon contains nuclei from both the DT1 and WT strains in the same cytoplasm. Restoration of the quelling of the al-1 gene produced an albino (white) phenotype in this heterokaryon, indicating that the Dicer knockout alleles are recessive and are complemented by the WT alleles in the heterokaryon. A Northern analysis was performed on the DT1 and WT strains as well as the resulting heterokaryon (Fig. 4), with the aim of examining the level of the dsRNA and the accumulation of siRNAs. While dsRNA is accumulated to a high level in the DT1 strain without any detectable siRNAs, in the heterokaryotic strain, the level of full-length dsRNA is reduced and there is a corresponding accumulation of siRNAs (Fig. 4).

FIG. 4.

dsRNA and siRNAs present in a double Dicer mutant strain. (A) Northern blot of total RNAs extracted from the DT1 transformant, WT, and a heterokaryotic strain (DT1 plus WT) made by using a sense riboprobe corresponding to the al-1 sequence. (B) Northern blot of low-molecular-weight enriched RNA preparations from strains as indicated in panel A made by using a hydrolyzed riboprobe corresponding to the al-1 gene.

Consistent with the data obtained in living cells, radioactive in vitro-generated dsRNA incubated with cell extracts from the double dcl-1/dcl-2 mutant was not processed, and the accumulation of short RNA ∼25 nt long appeared completely blocked (Fig. 5). By contrast, WT and both single Dicer mutants showed a Dicer-like activity that produced short RNA ∼25 nt long, consistent with the observation that the two single mutants were both quelling and dsRNA-induced PTGS proficient. However, while the dcl-1 mutant showed a dsRNA processing activity efficiency similar to that of the WT strain, the accumulation of siRNA, although present, was reduced in a dcl-2 mutant compared to the level in the WT (Fig. 5). Production of the ∼25-nt species was energy dependent. An additional ∼16-nt species was produced in an energy-dependent manner in all extracts; however, this species, but not the ∼25-nt species, could also be obtained from ssRNA (data not shown).

DISCUSSION

It is commonly thought that PTGS phenomena induced either by dsRNA (RNAi) or by transgenes (responsible for quelling in fungi and cosuppression in plants) originates from an ancient mechanism probably devoted to genome defense. Supporting this notion, PTGS phenomena show a common mechanistic and genetic basis. Both RNAi and transgene-induced PTGS pathways require members of the PPD family (16). Moreover, in both these phenomena, siRNAs are considered diagnostic elements of the silencing process as well as being cofactors necessary to direct the sequence-specific RNA degrading activity of the nuclease complex called RISC. During RNAi in animal cells, it was found that dsRNAs are processed into siRNAs by the action of an RNase III enzyme called Dicer (1). In transgenic plants and fungi, it has been proposed that dsRNA could originate from transgenic transcripts (11), frequently called aberrant RNAs, that are converted into dsRNA by the action of RdRPs. Such enzymes are in fact required for PTGS both in fungi and in plants.

Putative Dicer homologues are also present in plant and fungi genomes; however, until now no direct evidence of an involvement of this class of enzyme in PTGS has been provided. This lack of evidence on the role of Dicer in transgene-induced gene silencing may have several explanations. Recent results on the in vitro activity of QDE1, an RdRP required for quelling in Neurospora, have shown that QDE1 is able to synthesize long dsRNA molecules by copying full-length ssRNA templates (28). Surprisingly, QDE1 also has another activity synthesizing short 9- to 21-nt copies of the ssRNA template. Moreover, this reaction has been found to be considerably more efficient than the synthesis of long full-length dsRNA. These observations have led to the proposal (28) that at least a subset of the QDE-1 reaction products, close to 19 to 21 nt, could provide the ideal RNA cofactors for a RISC nuclease complex. According to this model, siRNA could be directly produced by QDE1 on aberrant transgenic RNA templates, thereby rendering the Dicer-like activity dispensable in quelling in Neurospora. The fact that the previous exhaustive genetic screens in Neurospora failed to isolate silencing mutants defective in the two Dicer homologues could apparently support the above model. The two Dicer homologues could be involved in other cellular processes such as the production of miRNAs that are involved in development. It is suggestive that mutation of the Arabidopsis Dicer-1 homologue, CARPEL FACTORY, blocks miRNA production but does not compromise PTGS (18). However, both in Neurospora and Arabidopsis, the existence of several Dicer paralogs (four in Arabidopsis) would mask their role in silencing if their functions were redundant.

We found that dsRNA molecules incubated with N. crassa extracts were promptly processed into short RNA approximately 21 nt long. Similar to the situation in Drosophila and C. elegans, this processing activity requires ATP (1, 25). The meaning of this ATP requirement is still under debate; it has been proposed that ATP plays a role in increasing the efficiency of siRNA formation by promoting a structural rearrangement of Dicer, required for the next round of substrate dsRNA binding and catalysis. Alternatively, it is possible that ATP can promote the release of the produced siRNA from Dicer to downstream components of the PTGS pathway, such as the RISC nuclease. The Dicer-like activity present in Neurospora appears to be constitutively present since it depends neither on the triggering of gene silencing by transgenes nor on a functional silencing pathway. In fact, both the WT strain and qde mutants retain the dsRNA processing activity.

The in vitro dsRNA processing into short RNA was completely abolished in a strain in which both the Dicer homologues found in the Neurospora genome were mutated. Each predicted gene product possesses all of the domains previously found in other proteins belonging to the Dicer family, namely, two RNase III domains located at the C terminus and an RNA helicase domain coupled with a DEAD-box ATP binding site at the N-terminal end. While the homologue DCL1 is predicted by the PSORT program (http://psort.nibb.ac.jp/) to be a nuclear protein, DCL2 is predicted to be cytoplasmic and contains a dsRNA binding domain that is apparently missing in DCL1. Although the analysis of the two DCL proteins could suggest different activities and/or cellular localizations, both proteins appear able to process dsRNA in vitro. Moreover, they appear completely redundant in quelling in vivo. In effect, only extracts from a double mutant were unable to process the dsRNA, while the dsRNA processing activity can be seen in the extracts from the single mutants. However, the dcl-1 mutant strain displayed dsRNA processing activity comparable to that of the WT strain, whereas the dcl-2 mutant had reduced activity.

More importantly, the double Dicer mutant is totally blocked in transgene-induced gene silencing, since no transformants showing quelling have been isolated after the introduction of transgenic al-1. By contrast, the two single Dicer mutants displayed a frequency of silencing comparable to that of the WT strain. These results indicate that Dicer is required for quelling and, more generally, is necessary for PTGS induced by transgenes. The redundancy of the two DCL proteins in the quelling pathway may suggest that both in Neurospora and plants multiple gene silencing pathways exist that can be juxtaposed. This could explain why until now no indication of Dicer involvement in transgene-induced PTGS has been provided. The observation that the dcl-2 mutant displayed decreased Dicer activity in vitro could suggest that the DCL2 protein has the major role in the dsRNA processing. This idea contrasts with the finding that the dcl-2 mutant did not show any reduction in quelling frequency, suggesting either that the residual processing activity in dcl-2 is sufficient in vivo to fully support silencing or that in vitro results do not entirely reflect the situation in vivo. For instance, DCL1 could be unstable and/or not easily extracted in protein preparations, thus leading to an apparent reduction of Dicer-like activity in in vitro essays. However, the finding that both in vivo and in vitro the two dcl genes need to be eliminated in order to completely block the production of siRNAs suggests that both proteins are able to process dsRNA molecules. The fact that DCL1 apparently lacks a dsRNA binding domain may indicate either that DCL1 acts in vivo in concert with other dsRNA binding protein(s) or that a still unknown domain present in DCL1 allows the interaction with the dsRNA substrates.

It has been proposed that Dicer proteins function as dimers in the processing of dsRNA substrates (2). This raises the possibility that the Neurospora DCL1 and DCL2 may form active heterodimers. Although heterodimerization cannot be excluded, the finding that mutations on single dcl genes do not compromise the generation of siRNA suggests that each protein may function independently from one another, perhaps forming functional homodimers.

The involvement of Dicer in quelling also demonstrates that dsRNA molecules are essential intermediates in the PTGS pathway. Above, it was suggested that dsRNA could arise from the activity of the RdRP QDE1 that uses single-stranded transgenic RNA templates. In support of this idea, we found that the expression of dsRNA is able to bypass the qde-1 requirement. Thus, it is not clear if QDE1 produces short RNA molecules in vivo. However, if such short RNAs were made in vivo, they are not sufficient to form a functional RISC complex. It is tempting to speculate that the short RNAs produced by QDE1 may have a function different from that of siRNAs. For instance, in C. elegans (33) it has been shown that small RNAs are used as primers by RdRP to convert target mRNA into dsRNA, leading to an amplification mechanism that enhances the strength of RNAi. Moreover, we observed that the qde-3 gene, encoding a recQ DNA helicase (12), was no longer required for dsRNA-induced silencing, suggesting for this gene also a role upstream of the formation of the dsRNA intermediate. Molecular analysis confirmed that the double Dicer mutant is indeed impaired in dsRNA processing in vivo. We found that dsRNA was accumulated in the dcl-1/dcl-2 mutant without any detectable accumulation of siRNA. By contrast, complementation of the two mutations that occurs in heterokaryosis with a WT strain resulted in a reduction of the dsRNA level coupled with an accumulation of siRNA leading to the restoration of silencing of al-1 gene.

Further research will be necessary to answer questions relating to the specificity, if any, of the two dcl genes in different gene silencing pathways. For instance, it will be interesting to find out if either or both dcl genes are also involved in meiotic silencing by unpaired DNA (8, 32), a quelling-related gene silencing phenomenon occurring in Neurospora in the sexual phase of the life cycle.