Mechanism of siRNA production from repetitive DNA

Repetitive DNA loci are a major source for the production of eukaryotic small RNAs, but how these small RNAs are produced is not clear. Yang et al. show that DNA tandem repeats and double-strand breaks are necessary and, when both are present, sufficient to trigger gene silencing and siRNA production. In addition to siRNA production, the quelling pathway also maintains tandem repeats by regulating homologous recombination.

Abstract

RNAi is a conserved genome defense mechanism in eukaryotes that protects against deleterious effects of transposons and viral invasion. Repetitive DNA loci are a major source for the production of eukaryotic small RNAs, but how these small RNAs are produced is not clear. Quelling in Neurospora is one of the first known RNAi-related phenomena and is triggered by the presence of multiple copies of transgenes. Here we showed that DNA tandem repeats and double-strand breaks are necessary and, when both are present, sufficient to trigger gene silencing and siRNA production. Introduction of a site-specific double-strand break or DNA fragile site resulted in homologous recombination of repetitive sequences, which is required for gene silencing. In addition to siRNA production, the quelling pathway also maintains tandem repeats by regulating homologous recombination. Our study identified the mechanistic trigger for siRNA production from repetitive DNA and established a role for siRNA in maintaining genome stability.

RNAi is a gene silencing mechanism conserved from fungi to mammals (Baulcombe 2004; Catalanotto et al. 2006; Buhler and Moazed 2007; Ghildiyal and Zamore 2009). The RNAi-related pathways are dependent on siRNAs generated from dsRNA by Dicer cleavage. An Argonaute (Ago) family protein associates with siRNAs to mediate post-transcriptional or transcriptional gene silencing. In eukaryotes, a major function of RNAi is to act as a host defense mechanism against transposons and viral invasion (Napoli et al. 1990; Sijen and Plasterk 2003; Siomi et al. 2008; Wang et al. 2010; Chang et al. 2012). Consistently, a significant portion of eukaryotic small RNAs are produced from repetitive DNA loci in fungi, plants, and animals and silence transposon sequences (Siomi et al. 2008; Ghildiyal and Zamore 2009). How these small RNAs are specifically produced from repetitive DNA loci is not clear.

Quelling, a post-transcriptional gene silencing mechanism in the filamentous fungus Neurospora crassa, was one of the first RNAi-related phenomena characterized (Romano and Macino 1992; Cogoni et al. 1996; Catalanotto et al. 2006; Chang et al. 2012). Similar to the process of cosuppression in plants, quelling is achieved by random transformation of a transgene, which results in the silencing of expression of homologous genes post-transcriptionally (Napoli et al. 1990; Romano and Macino 1992). Because the silencing triggered by quelling is associated with the presence of multiple copies of a transgene and is transient due to the loss of transgene or reduction of transgene copy numbers, it was proposed that quelling is triggered by repetitive transgenes. Most Neurospora strains lack highly repetitive DNA sequences except for the ribosomal DNA (rDNA), suggesting that quelling is a genome defense mechanism against transposon replication. Supporting this conclusion, the RNAi pathway was shown to suppress proliferation of a LINE-like transposon in Neurospora (Nolan et al. 2005).

In the quelling pathway, quelling-deficient-1 (QDE-1) is a dual-functional enzyme with both DNA-dependent RNA polymerase (DdRP) and RNA-dependent RNA polymerase (RdRP) activities (Cogoni and Macino 1999a; Lee et al. 2010). QDE-1 first acts as a DdRP to produce single-stranded aberrant RNA (aRNA) from the repetitive DNA loci and then uses its RdRP activity to convert aRNA into dsRNA. The RecQ DNA helicase QDE-3 and the ssDNA-binding complex RPA are required for this process (Cogoni and Macino 1999b; Nolan et al. 2008; Lee et al. 2010). After dsRNA is processed by Dicer, the resulting siRNAs are loaded onto the Ago protein QDE-2 to mediate post-transcriptional gene silencing of homologous RNAs (Catalanotto et al. 2004; Maiti et al. 2007).

We previously discovered that treatment of Neurospora with a DNA damage-inducing agent results in expression of a class of small RNAs, named qiRNAs, from the rDNA locus, which contains ∼200 copies of rDNA (Lee et al. 2009, 2010). Similar to quelling-induced siRNA synthesis, the production of qiRNAs also requires QDE-1, QDE-3, Dicer, and RPA. More recently, we showed that the homologous recombination (HR) process is required for qiRNA production and quelling (Zhang et al. 2013, 2014). Because repetitive sequences can provide abundant donor sequences for HR, our results suggested that HR is a mechanism that allows the genome to distinguish repetitive DNA sequences from nonrepetitive sequences. Since our discovery of qiRNA, DNA damage-induced small RNAs have been reported in Arabidopsis, Drosophila, and mammals (Francia et al. 2012; Michalik et al. 2012; Wei et al. 2012), suggesting that DNA damage is a common trigger for small RNA production in eukaryotes.

What is the genomic feature of repetitive DNA required for gene silencing, and what is the mechanism that allows the specific production of siRNAs from the repetitive DNA loci? Do siRNAs have additional functions other than gene silencing? In this study, we sought to determine the mechanism of siRNA production from repetitive DNA. We demonstrate here that DNA tandem repeats and a double-strand break (DSB) are both required for siRNA production. In addition, we show that the RNAi pathway regulates HR of repetitive sequences. This study thus identified the trigger for siRNA production from repetitive DNA and established a role for siRNAs in maintaining genome stability.

Results

Quelling is associated with tandem transgene repeats

Quelling in Neurospora is triggered by the transformation of multiple copies of a transgene and results in the silencing of the homologous endogenous gene through the RNAi pathway (Romano and Macino 1992; Cogoni et al. 1996; Fulci and Macino 2007). Quelling efficiency correlates with transgene copy numbers, and silencing is frequently unstable due to loss of the transgene. To investigate the genomic nature of the quelling phenomenon, we carried out quelling assays by randomly transforming a truncated albino-1 (al-1) fragment-containing plasmid, pBSKal-1, into a qde-1^ko strain in which QDE-1 expression can be induced by the addition of quinic acid (QA). The inducible expression of QDE-1 allowed us to identify true quelled strains because random transformation may also yield inverted al-1 repeats, from which dsRNA and siRNA can be produced independently of QDE-1 and QDE-3 (Goldoni et al. 2004).

Pronounced silencing of the endogenous al-1 gene was observed in the presence of QA in ∼30% of the resulting transformants, as indicated by their white conidia and hyphae (Fig. 1A, top panel). Modest silencing was observed in the absence of QA because of low QDE-1 expression due to the leakiness of the qa-2 promoter. Similar to earlier reports (Romano and Macino 1992; Cogoni et al. 1996), Southern blot analysis of al-1 revealed that in addition to the endogenous al-1 gene, multiple copies of al-1 transgenes were integrated into the genome. In addition, there were ladder-like multiple al-1-specific signals in each quelled strain (Fig. 1A, bottom panel). In contrast, in the unquelled strains from the same transformation, the al-1 transgene copy numbers were low, and only one or two al-1 transgene-specific bands were observed. Quantitative PCR (qPCR) analyses of al-1 copy numbers from 27 independently generated quelled strains showed that there were an average of 14 copies of al-1, ranging from five to 38 copies, in each strain. Because Neurospora is multinucleated and not all nuclei have the transgene, the actual transgene copy number in silenced nuclei should be higher than this estimate.

Figure 1.

Tandem repeats of the al-1 transgene are associated with quelling. (A) Multiple copies of the al-1 transgene are associated with quelling. (Top panel) Slants of the qde-1^ko; MycQDE-1 strains after transformation with the al-1 transgene. The quelled transformants display white/yellow conidia and aerial hyphae in the presence of QA due to silencing of the al-1 gene. The unquelled transformants are orange as is the control strain (con), which is the host strain without al-1 transformation. The addition of QA induces expression of MycQDE-1. (Bottom panel) Southern blot analysis using an al-1 fragment as the probe shows the profiles of transgenes in the indicated strains. The arrow indicates the endogenous al-1 signal. (B) Progenies of a quelled strain have different quelling phenotypes and transgene profiles. (C) Schematic diagram showing transgene integration sites identified by whole-genome sequencing.

Microconidia were purified to obtain homokaryotic progenies of quelled transformants. As shown in Figure 1B, al-1 silencing was observed in most of the progenies, but silencing was lost in some due to the loss of al-1 transgene. Although the ladder-like al-1 transgene signals remained in the strains with the silencing phenotype, the signal patterns varied from progeny to progeny. Together, these results are consistent with the conclusion that quelling requires multiple copies of a transgene and indicate that the transgene and its copy numbers are highly unstable.

How are multiple copies of al-1 transgenes arranged in the genome? Although it was long assumed that the transgene copies might form tandem repeats (Cogoni et al. 1996), the presence of multiple al-1 transgene signals also suggested the possibility that the copies are integrated in many different loci of the genome. To determine the locations of the transgene copies, we pooled individually isolated genomic DNA from 43 independently generated al-1 quelled strains and performed high-throughput whole-genome deep sequencing (Fig. 1C). The total sequencing reads obtained were predicted to cover >99% of the genome for every strain. Despite such a high coverage, there were only a total of 61 al-1 genomic integration sites (1.4 per strain) identified. These sites appear to be randomly distributed in the genome and do not appear to have any preference for genome context. This result strongly suggests that most of the transgenes in each strain form tandem repeats at one integration site and are not integrated at multiple genomic loci. Because HR is required for quelling (Zhang et al. 2013, 2014), the ladder-like al-1 signals in the quelled strains are therefore likely due to HR events among transgene repeats, which would result in different transgene copy numbers in different nuclei.

Tandem repeats are not sufficient for silencing but can trigger silencing after DNA damage agent treatment

Quelling assays are normally performed by using a random non-HR-based transformation method, which cannot control the integration site, copy number, or nature of the repeat of the transgene. To determine whether the tandem repeat itself is sufficient to trigger quelling, we generated a series of constructs that contained one (1xal-1) to seven (7xal-1) tandem repeats of a promoter-less 1.3-kb al-1 ORF fragment (Fig. 2A). In addition, constructs that have the qa-2-inducible promoter that can drive sense, antisense, or convergent transcription of the al-1 fragment were also generated. These constructs were individually targeted to the his-3 locus of a wild-type strain, which, in most but not all cases, resulted in single-copy integration of the construct by HR. Surprisingly, no al-1 silencing was observed in a vast majority of the transformants (Fig. 2B), demonstrating that tandem repeats (up to seven) by themselves were not sufficient to trigger silencing.

Figure 2.

Tandem repeats are not sufficient for silencing but can trigger silencing after DNA damage agent treatment. (A) A diagram showing the al-1 fragment used to generate al-1 tandem repeat constructs. (B) A table showing the al-1-containing his-3 targeting constructs and the summary of al-1 silencing results of the transformants for each construct. (C,D, top panels) Slants of the transformants from the 1xal-1 (C) and 7xal-1 (D) transformants. (Bottom panels) Southern blot analyses using the al-1 fragment as the probe, showing the profiles of transgenes in the indicated strains. (WT) A wild-type control strain. The low levels of the endogenous al-1 bands in some lanes are due to unequal loading of DNA samples. (E, top panel) Slants of the indicated strains in the presence or absence of histidine. (Middle panel) Small RNA Northern blot analysis results showing the production of al-1-specific siRNA production in the indicated strains. (Bottom panel) Ethidium bromide-stained gel showing the RNA loading in the indicated samples. (F) Small RNA Northern blot analysis results showing the production of siRNA in a histidine dose-dependent manner in the 7xal-1 strain but not in the other strains. Ethidium bromide-stained gels are shown below each Northern blot. The dsal-1 strain produces al-1 siRNA by expression from an al-1 inverted repeat.

Silencing was observed in a very low percentage of the transformants (Fig. 2B–D). To determine the genomic nature of these silenced transformants, Southern blot analyses were performed for a few randomly selected strains that were transformed with constructs containing one or seven copies of the al-1 repeat. In contrast to the unsilenced strains, which, as expected, only have a single site and single-copy integration of the indicated al-1 construct at the his-3 locus, ladder-like al-1 signals were seen in the silenced transformants (Fig. 2C,D, bottom panels), indicating that the silencing in these strains was due to nonhomologous integration of the constructs that form multiple repeats of the construct. These results indicate that the low frequency of gene silencing is due to low frequency of non-HR of the transgene in the his-3 targeting transformation method.

We previously showed that treatment of Neurospora with a DNA-damaging agent triggers production of qiRNAs from a repetitive rDNA locus (Lee et al. 2009, 2010). To determine whether DNA damage can trigger the production of siRNA from tandem repeats at a non-rDNA locus, we treated a strain that contains a single copy of 7xal-1 with histidine, a known DNA-damaging agent in Neurospora that inhibits ribonucleotide reductase (Pendyala and Wellman 1975; Lee et al. 2009). As shown in Figure 2E, siRNA was not observed in the wild-type strain or the 7xal-1 strain without histidine treatment, but siRNA production was induced by histidine in the 7xal-1 strain. Consistent with the siRNA production, silencing of al-1 was also observed in the 7xal-1 strain, as indicated by the color change of the conidia and aerial hyphae after histidine treatment. Histidine induced siRNA production in a dose-dependent manner in the 7xal-1 strain but failed to induce siRNA in strains with a single copy of al-1 repeat (Fig. 2F). These results indicate that DNA damage treatment can specifically trigger siRNA production from a locus with tandem repeats.

A DNA DSB in an al-1 tandem repeat triggers silencing

Histidine is a mild DNA-damaging agent that results in replication stress by inhibiting nucleotide biosynthesis in Neurospora (Pendyala and Wellman 1975; Lee et al. 2009). To determine the nature of DNA damage that triggers siRNA production, we introduced a cleavage site for the homing endonuclease I-SceI (TAGGGATAACAGGGTAAT) into one of the al-1 tandem repeats at the his-3 locus (Fig. 3A). The yeast enzyme I-SceI expressed very poorly in Neurospora, likely due to the strong codon biases of the Neurospora genome (Zhou et al. 2013). To express I-SceI in Neurospora, we codon-optimized the I-SceI gene based on Neurospora codon preferences. Because there is no endogenous I-SceI site in the Neurospora genome, the expression of I-SceI results in a DSB only at the introduced I-SceI site.

Figure 3.

I-SceI-induced DSB in the al-1 tandem repeat triggers gene silencing and siRNA production. (A) Schematic diagrams showing the al-1 his-3 targeting construct and the location of the I-SceI site. (B) PCR amplification of the 1xal-1 sequence spanning the 3XI-SceI cut site from strains containing a 1xal-1(3XI-SceI) construct with/without I-SceI expression. The expression of I-SceI was achieved by introducing an I-SceI expression construct into the indicated strains. (C) Southern blot analysis results using an al-1 probe, showing the 1xal-1(I) strain with/without I-SceI expression. (D) Small RNA Northern blot analysis showing the lack of siRNA production from the 1xal-1(I) strain. (E) Southern blot analysis results using an al-1 probe of the 2xal-1(I) strain with/without I-SceI expression. (F, top panel) Photo of the slants showing that I-SceI induced al-1 silencing in the 7xal-1(I) strains. (Bottom panel) Southern blot analysis results using an al-1 probe of the 7xal-1(I) strain with/without I-SceI expression. (G) Small RNA Northern blot analysis showing the induction of siRNA production by I-SceI expression from the 7xal-1(I) strain. The ethidium bromide-stained gels show equal loading of RNA samples. The dsal-1 and wild-type (WT) strains served as positive and negative siRNA controls, respectively.

To determine whether the expressed I-SceI can create a site-specific DSB in Neurospora, we first examined a strain [1xal-1(3XI-SceI)] that has one al-1 repeat and three concatenated I-SceI recognition sites. PCR of the al-1 region spanning the I-SceI sites (Fig. 3B) and DNA sequencing showed that the expression of I-SceI specifically led to a reduction in the length of the I-SceI-containing fragment in the 1xal-1(3XI-SceI) strain, but not in the strain lacking I-SceI expression, due to loss of one or two I-SceI sites as a result of DNA repair. This result demonstrated that the codon-optimized I-SceI expressed in Neurospora results in a DSB at the introduced site.

We then examined the effect of I-SceI expression in strains that carry one, two, or seven tandem al-1 repeats with an I-SceI site. I-SceI expression did not cause detectable al-1 silencing or siRNA production in strains with only one or two repeats, although in the 2xal-1(I) (I stands for I-SceI site) strains, an al-1 transgene band that corresponds to the size of one al-1 repeat was observed (Fig. 3C–E). In the 7xal-1(I) strains, the I-SceI expression led to al-1 silencing (as indicated by the yellow or white aerial hyphae and conidia), al-1 siRNA production, and the appearance of ladder-like al-1 transgene signals in Southern blot (Fig. 3F,G). The ladder-like al-1 transgene signals are similar to those seen in the quelled strains. The ladder-like transgene signals are due to HR events of tandem repetitive sequences, which result in a population of nuclei with different repeat numbers.

To further confirm our conclusion, we introduced an artificial DNA fragile site (FR) that contains multiple (CGG)n trinucleotides (n = 133) next to the al-1 tandem repeats (Fig. 4A; Fu et al. 1991). As expected, the presence of the fragile site resulted in al-1 silencing in ∼35% of the transformants, which all exhibited ladder-like al-1 transgene signals (Fig. 4B [top panel, lanes 1–6], D). A probe specific for the regions covering the fragile site (probe FR) (Fig. 4B, bottom panel) showed that all of these strains retained fragile sites of various lengths. The transformants that lacked al-1 silencing had lost either the fragile site or the al-1 repeat (Fig. 4B, lanes 7,8).

Figure 4.

Artificial fragile site in an al-1 tandem repeat triggers silencing. (A) A schematic diagram of the 7xal-1 construct with the fragile site (FR). (B, top panel) Photo of the slants of the indicated strains showing al-1 silencing in the 7xal-I-FR transformants. (Middle panel) Southern blot analysis results using the al-1 probe for the indicated strains. (Bottom panel) Southern blot analysis results using a probe to the fragile site. (C) Silencing phenotypes and al-1 genomic profiles in strains with a 7xal-1-FR-S or 7xal-1-FR-M construct at the his-3 locus. Wild-type and 7xal-1 strains without the fragile site served as negative controls. (D) Northern blot analysis showing the production of siRNAs in the 7xal-1-FR transformants. The ethidium bromide-stained gel shows equal loading of RNA samples. The dsal-1 and wild-type (WT) strains served as positive and negative siRNA controls, respectively. (E,F) ChIP assays using the phospho-γH2A antibody showing the association of γH2A at the repetitive al-1 region in 7xal-1-FR transformants (left) or the al-1 quelled strains (right). (*) P < 0.05.

Because the length of the trinucleotide repeats in the fragile site correlate with replication stalling and DSBs (Eichler et al. 1994; Jarem et al. 2010), we examined the effects of fragile sites with fewer CGG repeats: FR-S (n = 33) and FR-M (n = 67). No detectable silencing and no ladder-like al-1 transgene signals were observed in the FR-S strain (Fig. 4C, left). On the other hand, weak but detectable al-1 silencing and weak ladder-like al-1 transgene signals were seen in the FR-M strains (Fig. 4C,D).

To confirm the presence of a DSB due to the presence of the fragile site, we examined the accumulation of γH2A, a marker of DSBs (Lobrich et al. 2010), near the fragile site by chromatin immunoprecipitation (ChIP). As expected, there was significant enrichment of γH2A near the fragile site in the 7xal-1-FR strain (Fig. 4E), confirming the presence of DSBs. Similarly, significant enrichment of γH2A was also found at the al-1 transgene site in previously obtained quelled strains (Fig. 4F), indicating that quelling is also a result of DSBs. Together, these results further confirmed the requirement of DSBs in triggering siRNA production from the tandem repetitive sequences. In addition, the tight correlation between silencing and the presence of various transgene HR products further supports the role of HR in the small RNA production.

A comparison of the colors of silenced strains revealed that the presence of the longest fragile site (n = 133) resulted in more robust al-1 silencing than did the I-SceI-induced DSB (Figs. 3F, 4B). This difference is likely due to the fact that the repair of the I-SceI cleavage can result in deletion or mutation of the I-SceI site. As a result, the effect of I-SceI would be transient.

DSB-induced siRNA production profile at the tandem repeat region

To evaluate the siRNA production profile after DSB induction, we carried out small RNA sequencing in strains with and without tandem repeats and mapped the small RNA reads to the genome. No siRNA production was observed in the wild-type strain or strains with only one copy of the al-1 fragment with or without the I-SceI expression at the his-3 locus (Fig. 5A), indicating that DSB alone is not sufficient to trigger small RNA production. In contrast, in the strain containing 7xal-1 at the his-3 locus, siRNA production was observed at the transgene site only when I-SceI was expressed (Fig. 5B). Similarly, the presence of the fragile site next to the 7xal-1 repeats also resulted in robust siRNA production at the transgene site (Fig. 5C).

Figure 5.

Small RNA sequencing reveals siRNA production profiles from the tandem repeats after DSB. Histograms showing siRNAs of the indicated strains mapped to the al-1 transgene at the his-3 locus and the endogenous al-1 locus. Each repeat of the al-1 fragment is indicated by one yellow triangle. The Y-axis represents the number of normalized small RNA reads per million sequences. Plus (+) and minus (−) values represent the reads of small RNAs derived from sense and antisense strands, respectively. The X-axis indicates the position (in kilobases) of the regions analyzed. + I-SceI and − I-SceI indicate with or without I-SceI expression, respectively. + FR and − FR indicate with or without the fragile site in the construct. (A) siRNA distributions in the wild type (WT) and the strain with one copy of the al-1(I) fragment at the his-3 locus. (B) The comparison of siRNA distribution between the wild type and the strain with 7xal-1(I) at the his-3 locus. (C) siRNA distributions in the wild type and the strains with 7xal-1 or 7xal-1-FR at the his-3 locus.

At the his-3 locus, the DSB-induced small RNA reads (each read was mapped only once to the genome) covered the entire al-1 repeat and were also found in ≤3-kb flanking regions, indicating that siRNA production spread outside the repetitive regions. In contrast, the reads mapped to the endogenous al-1 locus were solely limited to the repeated al-1 region, indicating that siRNAs were only produced from the repetitive transgene locus, and the mapped reads at the endogenous al-1 locus were false reads. Together, these results further demonstrated the requirement of both DSBs and tandem DNA repeats in siRNA generation.

The siRNA production pathway regulates HR of the repetitive DNA

The production of DSB-induced siRNA from the repetitive DNA prompted us to examine whether the RNAi pathway has an additional role in genome stability. To test this, we introduced the 7xal-1-FR construct into dicer^dko, qde-1^ko, and qde-2^ko strains. Dicer and QDE-1 are required for the production of siRNA, whereas the Ago QDE-2 is required for siRNA-mediated gene silencing. As shown in Figure 6A, silencing from the 7xal-1-FR construct was completely abolished in the dicer^dko, qde-1^ko, and qde-2^ko strains, indicating their essential roles in this pathway. As expected, al-1 siRNA production was abolished in the dicer^dko and qde-1^ko strains.

Figure 6.

The quelling pathway contributes to the maintenance of tandem repeats by regulating HR. (A) Phenotypes of the 7xal-1-FR-containing strains in the wild-type (WT) and the quelling mutant backgrounds. (Top) Photo of slants. (Middle) al-1 siRNA Northern blot analysis results. (Bottom) Ethidium bromide-stained gel. The dsal-1 and wild-type (WT) strains served as positive and negative siRNA controls, respectively. (B–E) Southern blot analysis results showing the genomic profiles of the 7xal-1-FR transgene in wild-type (B), dcl^dko (C), qde-1^ko (D), and qde-2^ko (E) backgrounds. (F) A table summarizing the al-1 Southern blot analysis results on the 7xal-1-FR transgene profiles in the indicated strains. (G,H) al-1 Southern blot analysis results showing the genomic profiles of the 7xal-1(I) transgene after I-SceI induced DSBs in wild-type (G) and qde-1^ko (H) strains. (I) al-1 Southern blot analysis results showing the genomic profiles of the 7xal-1-FR transgene in the ku80^ko,qde-1^ko double mutant.

Southern blot analyses of the obtained transformants carrying the 7xal-1-FR in the wild-type and mutant backgrounds were compared. In the wild-type background, ∼35% of the transformants exhibited al-1 silencing and showed the expected pattern of al-1 transgene: the presence of the predicted full-length 7xal-1 band with a ladder of bands corresponding to the reduced al-1 copy number (Fig. 6B, lanes 1–4). For strains with no detectable al-1 silencing (for example, Fig. 6B, lanes 5–8), the patterns of the al-1 transgene ladders were abnormal. The patterns had one to three major transgene bands of various sizes and lacked the ladder-like signal seen in the silenced strains. This result suggests that the 7xal-1 tandem repeats subjected to DSB are unstable and can be lost due to recombination among the repeats. In contrast to wild-type transformants, almost none of the 7xal-1-FR transformants of the dicer^dko, qde-1^ko, and qde-2^ko strains exhibited the expected transgene pattern, and almost all showed the abnormal transgene patterns, suggesting rapid loss/abnormal repair of al-1 repeats in these quelling mutant strains (Fig. 6C–F).

To further confirm these results, we also compared the genomic profiles of the 7xal-1 with an I-SceI site after I-SceI expression in the wild-type strain and qde-1^ko mutant. As shown in Figure 6G, the expression of I-SceI resulted in the ladder-like 7xal-1 transgene signals in most of the wild-type transformants. In the qde-1^ko transformants, however, the transgene profiles were abnormal, and the ladder-like signals disappeared (Fig. 6H). Together, these results strongly suggest that the quelling pathway maintains the tandem repeats and regulates the recombination of the repetitive sequences.

Removal of the nonhomologous end-joining (NHEJ) repair pathway rescues the abnormal genomic profile of al-1 tandem repeats in a quelling mutant

HR and NHEJ are two major pathways for DSB repair (Sancar et al. 2004). In Neurospora, Ku80 is one of the essential components required for NHEJ (Ishibashi et al. 2006). To determine whether the abnormal genomic profiles of the al-1 tandem repeats in the quelling mutants are due to elevated NHEJ events after DNA damage, we created a ku80^ko,qde-1^ko double mutant and introduced the 7xal-1-FR construct at the his-3 locus by HR. In contrast to the qde-1^ko single mutant, most of the ku80^ko,qde-1^ko transformants displayed the full-length 7xal-1 bands with ladder-like signals (Fig. 6I) despite the lack of al-1 silencing (data not shown). This result suggests that the abnormal genomic profiles of the tandem repeats are mostly due to NHEJ, and the DSB-induced siRNA promotes HR events of the tandem repeats.

Discussion

Repetitive DNA loci are a major genomic source for small RNA production in eukaryotes. In this study, using Neurospora as a model system, we demonstrated that tandem repeats and DSBs are both necessary and, when both are present, sufficient to trigger gene silencing and siRNA production. Furthermore, we showed that the RNAi pathway helps to maintain the repetitive sequences in the genome by promoting HR. These results uncovered the mechanistic trigger of siRNA production from repetitive DNA and established a role for RNAi in maintaining genome stability.

Consistent with previous studies (Romano and Macino 1992; Cogoni et al. 1996), we showed that quelling in Neurospora is tightly associated with multiple copies of a transgene. The transient nature of quelling is due to loss or reduction of transgene copy number. Whole-genome sequencing of the quelled strains indicated that the multiple copies of a transgene form tandem repeats at a single site in the genome, indicating that the formation of tandem repeats rather than transgene integration at multiple sites is required for quelling. However, by generation of a series of strains with defined repeat numbers at a defined genetic locus, we showed that the tandem repeat alone is not sufficient for gene silencing and siRNA production (Fig. 2).

By introducing a defined DSB or a DNA fragile site at the tandem repeat region, we demonstrated that a DSB is necessary and sufficient for gene silencing and siRNA production (Figs. 3–5). These results demonstrate that a DSB in the tandem repeat locus is the initial trigger for the activation of the quelling pathway (Fig. 7). In quelled strains and in tandem repeat-containing strains with gene silencing, Southern blots exhibited ladder-like transgene signals. The bands correspond to different transgene copy numbers due to HR events between individual repeat sequences. Supporting this conclusion, we previously showed that the HR process is required for quelling (Zhang et al. 2013, 2014). Thus, HR of the tandem repeat region may result in recombination intermediates that can be specifically recognized by the quelling pathway, likely QDE-3, a putative RecQ DNA helicase homologous to the Werner/Bloom syndrome proteins (Cogoni and Macino 1999b). QDE-3 may resolve the recombination intermediates into ssDNA using its DNA helicase activity and may recruit ssDNA-binding protein RPA and the dual-functional enzyme QDE-1 to the site (Lee et al. 2010). QDE-1 uses its DdRP and RdRP activities to first produce ssRNA and then dsRNA to activate the downstream RNAi pathway. As a result, this pathway ensures that siRNAs are specifically produced from the repetitive DNA loci, which can be formed by transposon replication, but not from nonrepetitive parts of the genome.

Figure 7.

A model showing the proposed mechanism of repeat-induced siRNA production. Replication stress or DNA damage triggered DSBs in the tandem-repetitive genome locus. DSBs result in HR of repetitive sequences. The recombination intermediates of the HR process can be specifically recognized by the quelling pathway, likely QDE-3, which may resolve the recombination intermediates into ssDNA and recruit RPA and QDE-1. QDE-1 uses its DdRP and RdRP activities to first produce ssRNA and then dsRNA to activate the downstream RNAi pathway. The nuclear QDE-2 associates with siRNAs and promotes the HR process between the repetitive sequences. The cytoplasmic siRNA-associated QDE-2 results in post-transcriptional gene silencing.

In addition to the mechanism of small RNA production, our study also uncovered a role for the RNAi pathway in maintaining genome stability by regulating the HR process. In mutants that are deficient in RNAi or siRNA production, the ladder-like transgene signals were absent, and the tandem repeats in the genome were not maintained after DSB as a result of rapid loss of the repeats (Fig. 6). This observation is similar to the hyperrecombination phenotype of yeast rad51 (Aguilera 1995; Fasullo et al. 2001). Importantly, such a defect of the RNAi mutant was rescued when the NHEJ pathway was blocked (Fig. 6I). These results suggest that the siRNAs induced by DSBs are involved in maintaining the repeats and promoting the HR process between the repetitive sequences or suppressing NHEJ events. Consistent with this observation, it was previously observed that the quelling mutants have reduced rDNA copy numbers (Cecere and Cogoni 2009). Consistent with such a nuclear role of RNAi, we found that a significant amount of QDE-2 protein resides in the nucleus (Supplemental Fig. S1). Together, these results suggest that in addition to post-transcriptional gene silencing, siRNAs produced from repeat regions also help maintain genome stability.

How does the RNAi pathway regulate the DNA repair process? In fission yeast, plants, and animals, small RNAs mediate epigenetic modifications of chromatin to cause transcriptional gene silencing (Buhler and Moazed 2007; Zhang and Zhu 2011; Castel and Martienssen 2013). However, the quelling pathway and epigenetic regulation of chromatin appear to be independent of each other in Neurospora (Chicas et al. 2004, 2005; Freitag et al. 2004). Furthermore, we failed to detect significant levels of DNA/histone H3K9 methylation and H3K27 methylation of the tandem repeats at the his-3 locus (Supplemental Fig. S2; data not known). These results suggest that siRNAs may not act by regulating the chromatin modifications. Because of the requirement of the Ago QDE-2 in maintaining the tandem repeats, it is likely that siRNA brings QDE-2 to the nascent transcript after DSB formation. In mammalian cells, it was very recently shown that AGO-2 interacts with RAD51 and is important for the localization of RAD51 to DSB sites (Gao et al. 2014). Thus, siRNA may target the HR machinery to the DSB to promote HR. On the other hand, it was recently shown in yeast that RNA transcripts can mediate HR of chromatin (Keskin et al. 2014). Thus, QDE-2 may also target RNA transcripts made from the repetitive region to regulate their role in HR.

Repeat-associated small RNAs are found in almost all eukaryotes. In addition, DNA damage-induced small RNAs have been reported in Arabidopsis, Drosophila, and mammals (Francia et al. 2012; Michalik et al. 2012; Wei et al. 2012), suggesting that DNA damage may be a common trigger for small RNA production in eukaryotes. Interestingly, in plants and mammalian cells, the reporter constructs used for the detection of small RNA production after DSB formation have two tandemly arranged repeats flanking the I-SceI sites (Wei et al. 2012). Thus, as in Neurospora, the presence of repeats and the HR process may also be required in the production of siRNAs in other organisms.

Repetitive DNA is known to be a major source of genome instability in various organisms due to hyperrecombination events as a result of replication stress (Bzymek and Lovett 2001; Vader et al. 2011). In Neurospora, the repetitive rDNA locus is a site of frequent chromosome breakage (Butler 1992), and we previously showed that the replication process is required for qiRNA production (Zhang et al. 2013). Here we demonstrated that a tandem repeat alone resulted in siRNA production when subjected to replication stress treatment. Thus, repetitive DNA is sensitive to replication stress, which can lead to DSBs and siRNA production. Dicers have been shown to be involved in maintaining the genome stability and rDNA integrity in Drosophila and fission yeast (Peng and Karpen 2007; Castel et al. 2014). Therefore, small RNA and RNAi components appear to have a broad role in maintaining genome stability in eukaryotes.

Materials and methods

Strains, growth conditions, and constructs

The wild-type strain used in this study was FGSC 4200(a). Quelled transformants were obtained by cotransformation of a mixture of 2 μg of pBSKal-1 (an al-1 fragment-containing plasmid) and 0.5 μg of pBT6 (a benomyl-resistant gene-containing plasmid) to the qde-1^ko; qa.MycQDE-1 strain. The qde-1^ko; qa.MycQDE-1 strain was generated previously (Lee et al. 2010). The benomyl-resistant transformants were picked onto minimal slants supplemented with QA to identify the quelled (white or yellow conidia and aerial hyphae) transformants. A liquid medium of 1× Vogel’s, 0.1% glucose, and 0.17% arginine with or without 0.01 M QA (pH 5.8) was used for liquid cultures. For liquid cultures containing histidine, a final concentration of 0.5 mg/mL was used.

The his-3 locus targeting vector pDE3dBH was used to construct the tandem repetitive al-1 plasmids. dsal-1, wild-type (301-6, his-3), qde-1^ko (his-3), qde-2^ko (his-3), and dicer^dko (his-3) strains were generated previously (Choudhary et al. 2007; Maiti et al. 2007). The qde-1^ko Ku80^ko his-3 strain was generated in this study by crossing qde-1^ko and Ku80^ko his-3 strains. The his-3 locus targeting plasmids were introduced into the indicated strains. The (CGG)n fragile sites were created by self-annealing PCR reaction as described previously (Samadashwily et al. 1997) and were ligated to a blunt-end T vector. The (CGG)n fragments of different lengths were selected and inserted into the pDE3dBH.nxal-1constructs at the EcoRI site.

ChIP assay

Anti-γH2A (phospho-S129) and anti-H3K9me3 antibodies (ab15083 and ab8898, Abcam) were used to perform immunoprecipitation assays. The wild-type extract was used as control. qPCR was carried out to quantify the levels of DNA in the precipitates. The relative DNA-binding levels of ChIP assays were determined by comparing the relative enrichment of the chromatin-immunoprecipitated DNA between the quelled strain and a wild-type strain. Relative enrichment of the chromatin-immunoprecipitated DNA was normalized to the actual al-1 copy numbers in each quelled strain or transformant. actin levels in the chromatin-immunoprecipitated DNA were used to normalize the loading in different samples.

Southern blot analyses

Genomic DNA was extracted and purified as previously described (Mohn et al. 2009). Approximately 10 μg of genomic DNA was used in the Southern blot analyses. PstI (absent in the al-1 fragment) was used to digest the genomic DNA for al-1 pattern detection. BamHI and EcoRV were used to digest the genomic DNA for (CGG)n length detection. The digested DNA was separated on a 0.8% agarose gel by electrophoresis and then transferred onto a Hybond N⁺ (GE Healthcare) membrane. Prehybridization, hybridization, and washing of the blot were performed as recommended by the manufacturers. Probes were labeled with ³²P by PCR using dCTP [α-³²P] from PerkinElmer. The primers used in labeling the probes were probe_al_F, GCGTTACTTCAAGACGGATCGCA; probe_al_R, CATCAACGCCACCAACAAAGGG; probe_a_F, TTTGGCCCACGACTTCTTC; probe_a_R, CATCAACGCCACCAACAAAG; probe_b_F, GGGAAACGCCTGGTATCTTTA; and probe_b_R, TCAGATCTGCCGGTCTCCCTATAG.

Northern blot and qPCR assays

Small RNA was prepared and enriched from total RNA using 5% polyethylene glycol (MW8000) and 500 mM NaCl as previously described (Maiti et al. 2007). RNA concentration was measured using a NanoDrop (Thermo Scientific). Small RNA was separated by electrophoresis through 16% polyacrylamide, 7 M urea, and 0.5× tris-borate EDTA (TBE) gels. Equal amounts of small RNA (20 μg) were loaded in each lane. UV-cross-linking was used to fix small RNAs to a Hybond N⁺ membrane (GE Healthcare). siRNA hybridization was carried out as described previously (Yang et al. 2013). qPCR was performed using iTaq SYBR Green Supermix with ROX (Bio-Rad) as previously described (Choudhary et al. 2007).

Genome sequencing

Whole-genome sequencing was performed by The McDermott Center Next-Generation Sequencing Core at University of Texas Southwestern. Genomic DNA was randomly sheared into small fragments using Covaris E210, and paired-end 2 × 100-base-pair (bp) sequencing was performed on Illumina HiSeq. After DNA sequencing, the cleaned reads were mapped to the reference Neurospora genome sequence (http://www.broadinstitute.org/annotation/genome/neurospora/GenomeDescriptions.html#Neurospora_crassa_OR74A) using SAMtools (Li et al. 2009). Reads that could not fully map to either the Neurospora genome or plasmid used for transformation were aligned against the Neurospora genome and plasmid using NCBI-BLAST with parameters for short sequences to determine the junction sites of transgene insertion.

Small RNA sequencing

Small RNA sequencing was performed by The McDermott Center Next-Generation Sequencing Core at University of Texas Southwestern. The cleaned reads were aligned to the Neurospora genome as well as the introduced plasmid sequences. Small RNA densities in the regions of interest were calculated and plotted within 100-bp sliding windows with a step size of 1 bp using R ( http://www.R-project.org). Sequence data from this study are available through the NCBI Short Read Archives (accession no. SRP051839)

Genomic DNA extraction and methylated DNA immunoprecipitation (MeDIP) assay

Genomic DNA was extracted as previously described (Mohn et al. 2009). For MeDIP assays, 5–10 μg of genomic DNA was sonicated into small fragments, and 1 μg of the 5-methylcytosine monoclonal antibody (Active Motif, catalog no. 39649) was used to perform the MeDIP assay as previously described (Pomraning et al. 2009). MeDIP samples were analyzed with qPCR. Relative enrichment of DNA was calculated as the ratio of MeDIP sample to input, and ratios were normalized to the actual al-1 copy numbers in each quelled strain or transformant. actin levels were used to normalize the loading in different samples.