Pre-piRNA trimming safeguards piRNAs against erroneous targeting by RNA-dependent RNA polymerase

SUMMARY

The Piwi/Piwi-interacting RNA (piRNA) pathway protects genome integrity in animal germ lines. Maturation of piRNAs involves nucleolytic processing at both 5′ and 3′ ends. The ribonuclease PARN-1 and its orthologs mediate piRNA 3′ trimming in worms, insects, and mammals. However, the significance of this evolutionarily conserved processing step is not fully understood. Employing C. elegans as a model, we recently discovered that 3′ trimming protects piRNAs against non-templated nucleotide additions and degradation. Here, we find that worms lacking PARN-1 accumulate an uncharacterized RNA species termed anti-piRNAs, which are antisense to piRNAs. Anti-piRNAs associate with Piwi proteins, are 17–19 nucleotides long, and begin with 5′ guanine or adenine. Untrimmed pre-piRNAs are misdirected by the terminal nucleotidyltransferase RDE-3 and RNA-dependent RNA polymerase EGO-1, leading to the formation of anti-piRNAs. This work identifies a class of small RNAs in parn-1 mutants and provides insight into the activities of RDE-3, EGO-1, and Piwi proteins.

Graphical Abstract

In brief

The maturation of piRNAs involves 3′ nucleolytic processing across animal species. Pastore et al. demonstrate that untrimmed piRNAs in C. elegans parn-1 mutants are targeted by the terminal nucleotidyltransferase RDE-3 and the RNA-dependent RNA polymerase EGO-1 and serve as templates for the generation of anti-piRNAs.

INTRODUCTION

The Piwi protein, a germline-enriched Argonaute, and Piwi-interacting RNAs (piRNAs) are essential for gametogenesis and fertility in all animals studied to date.^1-4 The primary role of the piRNA pathway is to safeguard the integrity of the germline genome by suppressing foreign sequences such as transposable elements and viral genomes.^5-9 Piwi/piRNA complexes, via base-pairing interaction, target foreign transcripts and trigger epigenetic silencing.^10,11 While piRNAs vary in genomic origin, length, and sequence across species, the mechanisms underlying their maturation are somewhat conserved.^1-4 In organisms ranging from worms to insects and mammals, single-stranded precursor transcripts are produced from piRNA genes. These precursors are then transported to perinuclear nuage, germ granules, or the surface of mitochondria, where they undergo extensive nucleolytic processing to yield mature piRNAs with specific lengths.^1-4 The final step in piRNA maturation is the 2′-O-methylation at their 3′ terminus, catalyzed by a conserved methyltransferase.^12-14

With a powerful genetic and molecular toolkit, C. elegans has emerged as one of the crucial organisms for studying the piRNA pathway. C. elegans expresses a single functional Piwi protein known as PRG-1. Mature piRNAs are commonly referred to as 21U-RNAs due to their strong propensity for a 5′ uridine (U) and a primary length of 21 nucleotides (nt).^15-19 Types I and II piRNAs are found in C. elegans and its sibling species C. briggsae, with type I piRNAs being notably much more abundant.^16,20 Type I 21U-RNAs in C. elegans are encoded by over 15,000 individual minigenes organized within two large clusters at chromosome IV.^15-18 Promoters of many type I piRNA genes contain an 8-nt consensus Ruby motif, which is associated with chromatin factors including PRDE-1, SNPC-4, TOFU-4, and TOFU-5.^21-24 Type II piRNA genes share transcription start sites with actively transcribed genes across the genome but lack the Ruby motif.¹⁶ Unlike other organisms, piRNA precursors in C. elegans, known as capped small RNAs (csRNAs), are relatively short, typically spanning 25–29 nt in length.¹⁶ Akin to other species, C. elegans piRNA biogenesis involves nucleolytic processing at both the 5′ and the 3′ termini.^1-3 The 5′ cap and the first 2 nt of csRNAs are removed by a trimeric Schlafen-domain nuclease.²⁵ A few nucleotides are resected from their 3′ ends, followed by 2′-O-methylation catalyzed by the methyltransferase HENN-1.^26-29

Despite the variation in length and sequence across species, piRNAs in worms, insects, and mammals all undergo 3′ trimming. For instance, our previous investigations led to the identification of PARN-1 as the specific exonuclease responsible for generating the mature 3′ ends of piRNAs in C. elegans.²⁸ In parn-1 mutant worms, piRNAs with 3′ extensions accumulate, while their 5′ cap and first 2 nt are correctly removed.²⁸ A parallel study showed that poly(A)-specific ribonuclease-like domain-containing 1 (PNLDC1), a member of the PARN family, trims piRNA in Bombyxmori.³⁰ Subsequent studies demonstrated that PNLDC1 acts on piRNAs in mice.^31-33 Notably, exome sequencing of infertile male patients identified mutations in PNLDC1, linking these mutations to defects in piRNA processing and azoospermia.^34,35 Collectively, these studies emphasize the importance of piRNA 3′ nucleolytic processing across various species.

PRG-1/piRNA complexes target a broad spectrum of germline transcripts as well as certain transposable elements.^10,36-38 The targeting by PRG-1/piRNAs sets off a cascade of events. RNA-dependent RNA polymerases (RdRPs) are recruited to target RNAs that serve as templates for the production of secondary small interfering RNAs (siRNAs). These siRNAs are typically 22 nt long and bear a 5′ guanine residue, hence referred to as 22G-RNAs.^39-41 22G-RNAs are then loaded onto a large group of worm-specific Argonautes (WAGOs).^{10,11,36-38,42} Similarly, exogenous double-stranded RNAs (dsRNAs) can also induce the production of WAGO 22G-RNAs. In this process, endonuclease RDE-8 plays a pivotal role in cleaving the target RNAs.⁴³ RDE-3, a terminal nucleotidyltransferase (TENT),⁴⁴ adds alternating uridine (U) and guanosine (G) nucleotides to the 3′ termini of cleaved transcripts, forming what is known as poly(UG) tails.^45,46 RdRPs are subsequently recruited to the RNAs bearing poly(UG) tails, utilizing them as templates to synthesize 22G-RNAs.⁴⁵ This amplification cycle serves as a potent mechanism for propagating and maintaining epigenetic silencing across generations.^45,47 Thus far, RDE-3 and RdRPs are believed to act on long RNAs that are targeted by piRNAs or dsRNAs.

Here we report that, in C. elegans, deficiency for PARN-1 leads to the accumulation of a unique class of RNAs that are antisense to piRNAs, referred to as anti-piRNAs. The formation of anti-piRNAs involves a series of enzymatic activities. RDE-3 catalyzes the addition of non-templated UG dinucleotide and poly(UG) sequences at the 3′ terminus of untrimmed piRNAs, although the UG additions are not a prerequisite for the production of anti-piRNAs. EGO-1, one of the functional RdRPs in C. elegans, utilizes untrimmed piRNAs as templates for synthesizing anti-piRNAs. Furthermore, a seed-gate structure within PRG-1 appears to limit the elongation of anti-piRNAs, resulting in 17- to 19-nt anti-piRNAs. In conclusion, our findings provide a mechanistic insight into enzymatic activities of RDE-3 and EGO-1 and highlight the indispensability of piRNA 3′ processing.

RESULTS

Anti-piRNAs accumulate in parn-1 mutants

Acting as a 3′-to-5′ exonuclease, PARN-1 generates the mature 3′ end of piRNAs in C. elegans.²⁸ Worms deficient for PARN-1 accumulate untrimmed piRNAs with 3′ extensions and non-templated nucleotides at their 3′ end.^28,48 It is well established that single-stranded precursors are produced from piRNA genes.^1-4 In our previous analyses of small RNA sequencing (smRNA-seq) data, we focused exclusively on sense reads that mapped to piRNA-producing loci.^28,48 By examining both sense and antisense reads that are uniquely mapped to the reference genome (Wormbase release WS279), we serendipitously discovered a previously unrecognized class of RNA species antisense to piRNAs in parn-1(tm869) mutants, which we termed anti-piRNAs (Figure 1A). In the wild-type strain, small RNA reads predominantly matched the sense strand of piRNA genes (Figures 1A and 1B; Table S1). In contrast, antisense reads were readily detected in parn-1 mutants, although their abundance (reads per million total mapped reads or RPM = 541) was much lower than that of sense reads (RPM = 28,567) (Figure 1B; Table S1). Both piRNA and anti-piRNA levels were dramatically reduced in prg-1(tm872); parn-1 double mutants (Figures 1A and 1B; Table S1), suggesting that the accumulation of piRNAs and anti-piRNAs depends on PRG-1 protein.

Figure 1. Anti-piRNAs are produced in parn-1 mutants.

(A) Browser view showing reads mapping sense and antisense to piRNA loci in wild type and parn-1 and parn-1; prg-1 mutants. Sense sequences are shown in blue, antisense sequences are shown in magenta. C. elegans piRNA clusters are highlighted with gray boxes. Data are presented as the mean of n = 2 biological replicates for wild type and parn-1. And parn-1; prg-1 data come from a single sample. Note that scales for sense and antisense reads are adjusted differently.

(B) Bar plots showing the total levels of piRNA (blue) and anti-piRNA matching reads (magenta) in wild type and parn-1 and parn-1; prg-1 mutants. Data are displayed as the mean ± standard deviation of two biological replicates for wild type and parn-1. parn-1; prg-1 comprises a single biological replicate.

(C) Browser view showing sense and anti-piRNA reads mapping to 21UR-3975 in wild type and parn-1 mutants. Data are displayed as independent biological replicates.

(D) Bar plot showing piRNA levels under untreated and oxidizing conditions in parn-1 mutants. Data are displayed as the mean ± standard deviation of two biological replicates. **p < 0.01, two-tailed t test.

(E) Same as (D) but showing anti-piRNA level.

(F) Bar plot showing the RPM of length and first nucleotide of antisense sequences in wild-type input and PRG-1 IP samples. Data were obtained from one biological replicate.

(G) Same as (F) but in parn-1 mutants.

See also Figure S1 and Tables S1 and S2.

smRNA-seq from parn-1 mutants revealed that anti-piRNAs originate from 1,934 piRNA loci (Table S2). We next inspected specific examples. Consistent with previous findings,^28,48 piRNAs such as 21UR-3975 were primarily 21 nt in wild type, while untrimmed 21UR-3975 with 3′ extensions are found in parn-1 mutants (Figure 1C). Anti-piRNAs were exclusive to parn-1 mutants and were absent in wild-type strains (Figure 1C). Curiously, 5′ ends of anti-piRNAs initiated from various positions, while their 3′ ends mapped almost exclusively within the annotated piRNA gene (Figure 1C). In both wild-type and parn-1 animals, piRNAs are known to bear 2′-O-methylation at their 3′ terminus.^26-29,48 To determine the methylation status of anti-piRNAs, we performed sodium periodate-mediated oxidation assays and sequenced oxidation-resistant RNA species. RNAs with terminal 2′-O-methylation are resistant to oxidation, while unmethylated RNAs possess vicinal diols at their 3′ terminus, rendering them susceptible to oxidation and poor substrates for small RNA cloning. In parn-1 mutants piRNAs were enriched by oxidation upon normalization to total number of reads (Figure 1D), suggesting that untrimmed piRNAs possess terminal 2′-O-methylation.^28,48 However, anti-piRNAs were strongly depleted by oxidation (Figure 1E), suggesting the absence of terminal 2′-O-methylation at anti-piRNAs.

The observation that the accumulation of anti-piRNAs requires PRG-1 prompted us to investigate whether anti-piRNAs are loaded onto PRG-1 (Figures 1A and 1B). To this end, we recovered PRG-1 complexes by immunoprecipitation (IP) from wild-type and parn-1 extracts and subsequently sequenced PRG-1-bound small RNAs. In line with previous findings,²⁸ wild-type piRNAs displayed characteristic 5′ U and a length of 21 nt, while untrimmed piRNAs with 5′ U were enriched in parn-1 mutants from PRG-1 IP (Figures S1A and S1B).

Next, we assessed antisense reads mapping to piRNA-producing loci. Some basal levels of 22G-RNA reads were detected in the input of the wild-type sample, but these were absent in the PRG-1 IP (Figures 1F and S1C). In addition, we observed 21U-RNA reads that overlapped with piRNA-producing loci in the wild-type PRG-1 IP (Figures 1F and S1D). For instance, 21U-RNA reads mapped to the reverse strand of 21ur-11958 (Figure S1D). However, the characteristic Ruby motif was found in the approximately 40 nt upstream of both the 21U-RNA sequence and 21ur-11985 (Figure S1D). We conclude that these 21U-RNA reads likely originate from unannotated piRNA genes in the complementary strand. These unannotated piRNA genes were also reported by a recent study.⁴⁹ In stark contrast to the wild-type strain, PRG-1 IP from parn-1 mutants revealed a substantial enrichment of anti-piRNAs (Figure 1G). Notably, these anti-piRNAs predominantly initiated with A or G residues (Figure 1G). Moreover, the size distribution of anti-piRNAs exhibited a broad range, peaking at 17–19 nt (Figure 1G). We conclude that the size distribution and the 5′ nucleotide composition of anti-piRNAs are distinct from the well-characterized 21U-RNAs or 22G-RNAs.

piRNA::anti-piRNA duplexes are loaded onto PRG-1

We proceeded to develop an approach to probe the formation of piRNA::anti-piRNA duplexes within PRG-1 by combining PRG-1 IP and RNA ligation (Figure 2A). Briefly, anti-PRG-1 antibody was used to enrich the PRG-1 complex.¹⁵ The purified complex was treated with tobacco acid pyrophosphatase (TAP), an enzyme converting 5′ capped or polyphosphorylated RNA into 5′ monophosphorylated RNA. Proximal RNAs within the PRG-1 complex were then ligated to create hybrids. To enrich these hybrids, we isolated RNAs longer than 26 nt, which were subsequently subjected to small RNA cloning and deep sequencing (Figure 2A).

Figure 2. piRNA::anti-piRNA duplexes are loaded onto PRG-1.

(A) Schematic illustrating PRG-1 IP followed by ligation. Following PRG-1 IP, RNAs bound to PRG-1 were treated with TAP. Adjacent RNAs were ligated together using RNA ligase. Chimeric RNAs were then subjected to RNA isolation, adapter ligation, and deep sequencing.

(B) Flowchart showing library statistics and the number of piRNA::anti-piRNA chimeras from PRG-1 IP/ligation in parn-1 mutants.

(C) Specific example showing sequencing reads of ligation between 21UR-10633 and an associated anti-piRNA (red). The raw sequence count of this chimera measured by smRNA-seq is shown.

(D) Same as in (C) but showing locus 21UR-3276.

(E) Same as in (D) but showing locus 21UR-1615.

See also Figure S2.

Our experimental effort in wild-type and parn-1 backgrounds yielded ~2.77 million and ~5.82 million sequence reads, respectively (Figures 2B and S2). Of these reads, a substantial portion mapped to the genome reference (Figures 2B and S2). Among reads that failed to map to the genome, we searched for hybrid reads composed of a piRNA sequence and an anti-piRNA sequence. No piRNA::anti-piRNA hybrids were identified in wild type (Figure S2). In parn-1 mutants, we recovered 1,858 hybrid reads, which were distributed across 1,533 piRNA-producing loci (Figure 2B). In silico folding on hybrid reads revealed the evidence for the presence of piRNA::anti-piRNA duplexes (Figures 2C-2E). In summary, our smRNA-seq data and bioinformatic analyses uncovered anti-piRNA species that are generated upon the loss of PARN-1. These anti-piRNAs form duplexes with untrimmed piRNAs, which are loaded onto PRG-1 (Figure 2). Notably, the abundance of anti-piRNAs was diminished in prg-1; parn-1 double mutants compared with parn-1 single mutants (Figures 1A and 1B), suggesting that anti-piRNAs are predominantly associated with and protected by PRG-1. Moreover, we showed that anti-piRNAs are 17–19 nt in length, possess 5′ A or G, but lack 2′-O-methylation 3′ residues (Figures 1E and 1G), distinguishing them from their piRNA counterparts or WAGO 22G-RNAs.

RDE-3 modifies the 3′ terminus of untrimmed piRNAs

Next, we aimed to identify the enzymatic factors responsible for anti-piRNA production. We hypothesized that RNA polymerases may play a role in generating anti-piRNAs, given their preference for initiating transcription with a purine (A or G) as the first base in nascent transcripts.^50-52 Initially, we considered the possibility that anti-piRNAs, akin to piRNA precursors, are produced by RNA polymerase II. However, by mining transcription start site databases,^16,53 we found that anti-piRNA reads did not coincide with transcription start sites. This led us to explore an alternative hypothesis: anti-piRNAs are synthesized by RdRPs using piRNAs as templates. It is known that exogenous dsRNAs or endogenous triggers can induce the cleavage of target RNAs.^43,45 Following cleavage, RDE-3 adds poly(UG) tails to the 3′ terminus of cleaved RNAs (Figure 3A).^45,46 RdRPs, including RRF-1 and/or EGO-1, initiate the synthesis of 22G-RNAs, which are templated directly from poly(UG) RNAs (Figure 3A).⁴⁵ A speculative but intriguing possibility is that the 3′ termini of untrimmed piRNAs could not be properly accommodated by PRG-1 and therefore are susceptible to RDE-3 and RdRP(s).

Figure 3. anti-piRNAs are targeted by RDE-3 and EGO-1.

(A) Schematic illustrating 22G-RNA synthesis in wild-type C. elegans. Following target cleavage, RDE-3 is recruited to the 3′ end of target RNAs and adds stretches of poly(UG) repeats These repeats recruit the RdRP complex composed of EGO-1 or RRF-1 (RNA-dependent RNA polymerase), EKL-1 (Tudor domain protein), and DRH-3 (RNA helicase) for 22G-RNA production.

(B) Bar plot showing the frequency of non-templated nucleotide additions (3′ tailing) to piRNAs in wild type and parn-1 and parn-1; rde-3 mutants. piRNA tailing frequency is derived using the equation $\frac{RPMoftailed{group}_i}{TotalpiRNARPM}$, where RPM tailed group_i is the abundance of reads with tail i (mono-A, mono-G, mono-C, mono-U, GU or GU repeats, UG or UG repeats, and others), and Total piRNA RPM is the total abundance of all reads mapping to piRNA loci (perfectly matching and tailed reads). Data are displayed as the mean of two biological replicates.

(C) Specific example showing the sequence of sense piRNA and anti-piRNA sequences mapping to 21UR-4864. The graphic shows both perfectly matched and tailed piRNAs, anti-piRNAs, and anti-piRNAs mapping to tailed sense piRNAs. Non-templated nucleotides are shown in blue (U), red (A), orange (C), and green (G). Nucleotides matching the endogenous sequence are shown in gray. Anti-piRNAs with nucleotides mapping to tailed sense piRNAs are highlighted with a gray bar for clarity. RPMs are displayed as the mean of two biological replicates.

(D) Bar plot showing the levels of anti-piRNAs (normalized to miRNA) in parn-1, parn-1; rde-3, parn-1; ego-1^RNAi, parn-1; rrf-1, parn-1; ekl-1^RNAi, parn1; drh-3^TS, parn-1; drh-3^RNAi, and parn-1; oma-1^RNAi. Data are displayed as the mean ± standard deviation of two biological replicates. *p < 0.05, ***p < 0.001, two-tailed t test.

See also Figure S3.

Our previous study revealed abundant 3′ non-templated nucleotide additions, also known as 3′ tailing, in parn-1 mutants.⁴⁸ At that time, our customized bioinformatic pipeline allowed us to detect up to three occurrences of non-templated nucleotides.⁴⁸ For this investigation, we employed the short-read aligner Tailor,⁵⁴ which allowed us to unbiasedly determine non-templated nucleotide additions to RNAs regardless of their length. In line with prior findings,⁴⁸ a small fraction of piRNAs exhibited 3′ non-templated additions in wild-type worms. In parn-1 mutants, the frequency of non-templated nucleotide additions increased to 10.14% (Figure 3B). Monouridylation (mono-U) was the most abundant modification in both wild-type and parn-1 strains, while monoguanylation (mono-G) of piRNAs became abundant in parn-1 mutants (Figure 3B).⁴⁸ Both UG dinucleotide/repeat (0.43%) and GU dinucleotide/repeat (0.58%) were detected in piRNAs upon loss of PARN-1 (Figure 3B). Strikingly, mono-G, UG, and GU additions were significantly reduced in rde-3 (ne3370); parn-1 double mutants compared with their levels in parn-1 single mutants, while mono-A and mono-U levels remained comparable between these two strains (Figure 3B).

Loss of mono-G in rde-3; parn-1 mutants was unexpected, given that a previous cellulo-tethering assay suggested that RDE-3 catalyzes poly(UG) additions.⁴⁶ To further investigate this matter, we assessed the nucleotide composition at positions 1 or 2 nt upstream (−1 or −2 position) of the 3′ non-templated nucleotides (Figures S3A and S3B). The Tailor pipeline defined the −1 or −2 nucleotides as templated nucleotides, as they matched the genome reference. However, they could also be non-templated nucleotides if the added 3′ tails matched the genome reference purely by chance. In the group of piRNAs with mono-A tails, there was an overrepresentation of A at both the −1 and the −2 positions, implying that poly(A) may be added to the 3′ terminus of untrimmed piRNAs (Figures S3A and S3B). U was overrepresented at the −1 position in groups with mono-G and GU tails, but not in other groups (Figure S3A). No such strong bias was observed at the −2 position (Figure S3B). These findings suggest that U addition may precede G and GU addition. It is also possible that piRNAs terminating with U are preferred substrates for RDE-3.

Finally, we inspected specific piRNA-producing loci. For example, in parn-1 mutants, sense reads mapped to the 21ur-4864 locus exhibited 3′ extensions and non-templated nucleotide additions, including mono-U and GU. Antisense reads initiated from various positions. On some occasions they were complementary to sense reads with the non-templated nucleotides (Figure 3C). In agreement with the genome-wide analysis (Figure 3B), the non-templated GU addition at 21UR-4864 was absent in rde-3; parn-1 mutants (Figure 3C). Taken together, these findings suggest that RDE-3 catalyzes 3′ non-templated UG dinucleotide and repeat additions to untrimmed piRNAs.

EGO-1 uses untrimmed piRNAs as templates to initiate anti-piRNA production

To determine whether RdRPs are responsible for generating anti-piRNAs, we depleted individual RdRP components to assess their impact on anti-piRNA production. Two RdRPs, RRF-1 and EGO-1, function partially redundantly in the germ line to produce 22G-RNAs using endogenous transcripts or transposons as templates.^39-41 RRF-1 and EGO-1 individually form a complex with the Tudor-domain protein EKL-1 and RNA helicase DRH-3.^40,55 While rrf-1 is a non-essential gene, EGO-1, EKL-1, and DRH-3 are required for viability.^39-41,55

To deplete RdRP components in the parn-1 mutant background, we utilized a loss-of-function allele of rrf-1(ok589)⁵⁶ and a temperature-sensitive allele of drh-3(ne4253)⁴⁰ and performed knockdown of ego-1, drh-3, and ekl-1 using RNA interference (RNAi).⁵⁵ To assess the impact of these depletions, we measured overall 22G-RNA levels as well as two major categories of germline 22G-RNAs: CSR-1 class 22G-RNAs and WAGO class 22G-RNAs.^55,57 As expected, depletion of rde-3, rrf-1, ego-1, drh-3, and ekl-1 resulted in a reduction in 22G-RNA levels (Figure S3C). Since the degree of reduction varied across samples (Figure S3C), we decided to normalize anti-piRNA reads against microRNA reads instead of total mapped reads to accurately quantify anti-piRNA abundance (Figure 3D). While rde-3; parn-1 mutants displayed the strongest depletion of WAGO-1 22G-RNAs (Figure S3C), their anti-piRNA levels were only modestly decreased (Figure 3D), suggesting RDE-3 contributes to but is not required for anti-piRNA production. Deletion of rrf-1 resulted in a significant reduction in WAGO-1 22G-RNAs, but not in CSR-1 22G-RNAs (Figure S3C). Interestingly, we observed a slight increase, albeit not statistically significant, in anti-piRNA levels in rrf-1 mutants (Figure 3D). On the other hand, ego-1-depleted animals exhibited an ~12.4-fold decrease in anti-piRNA levels relative to the control (Figure 3D), suggesting EGO-1 is the primary RdRP for anti-piRNA production. Consistent with the idea that EGO-1, EKL-1, and DRH-3 form an RdRP complex,⁴⁰ mutation in drh-3 and depletion of either ekl-1 or drh-3 by RNAi resulted in a reduction in WAGO-1 22G-RNAs, CSR-1 22G-RNAs, and anti-piRNAs (Figures 3D and S3C). Taken together, our studies provide evidence that the EGO-1-containing RdRP complex initiates the de novo synthesis of anti-piRNAs that are templated directly from untrimmed piRNAs.

The seed-gate structure within PRG-1 hinders anti-piRNA elongation

Since the EGO-1/EKL-1/DRH-3 complex normally produces 22G-RNAs,⁴⁰ we wondered why it can generate anti-piRNAs that are 17–19 nt in length (Figure 1G). We first measured the distance between the 5′ ends of piRNAs (5′ U) and the 5′ ends of anti-piRNAs (Figure 4A). While the piRNA size in parn-1 mutants showed a broad distribution (22–29 nt),²⁸ the 5′ piRNA-to-5′ anti-piRNA distance exhibited a narrower range (25–29 nt) (Figure S4A). This finding indicates that longer isoforms of untrimmed piRNA are preferred RdRP substrates. We then measured the distance between the 5′ ends of piRNAs and the 3′ ends of anti-piRNAs (Figure 4A). The distance exhibited a narrow range and peaked at 9 nt (Figure 4B). In other words, the last nucleotide of anti-piRNAs primarily base-paired with the 10^th nucleotide of the corresponding piRNAs (Figure 4B). This observation motivated us to test two models for the generation of anti-piRNA 3′ ends: (1) the RdRP complex uses entire piRNAs as templates to generate RNA duplexes, and PRG-1 cleaves antisense RNAs at the phosphodiester bond linking nucleotides 9 and 10 (Figure 4C), and (2) the RdRP polymerization reaction is terminated due to the steric hindrance imposed by a certain structure(s) within PRG-1 (Figure 4C).

Figure 4. PRG-1 seed-gate may block RdRP from synthesizing anti-piRNAs.

(A) Schematic illustrating the information derived from measuring the distance from the 5′ or 3′ ends of anti-piRNAs (red) to the 5′ ends of sense piRNAs (black). The 5′-to-3′ (piRNA-to-anti-piRNA) distance describes where anti-piRNA synthesis stops; conversely, the 5′-to-5′ distance describes where anti-piRNA synthesis begins.

(B) Line plot showing the 5′-to-3′ distance of piRNAs to anti-piRNAs in parn-1, parn-1; prg-1^D654A, and parn-1; prg-1^SG_mut. Data are displayed as distances relative to the 5′ end of the piRNAs (position 0). The y axis is the percentage of reads, with the 5′-to-3′ distance on the x axis. The gray box highlights a 5′-to-3′ distance of 8 to 9 nt. Data are displayed as the mean of two biological replicates. g, guide RNA.

(C) Schematic illustrating two models by which 3′ ends of anti-piRNAs are generated. In the first model, synthesis of anti-piRNAs (red) goes beyond g10 of piRNAs (black) and PRG-1 cleaves anti-piRNAs between g9 and g10. In the second model, RdRP elongation is halted due to steric hinderance by PRG-1 protein itself.

(D) Cartoon representation of C. elegans PRG-1 (Ce-PIWI, AlphaFold prediction) superimposed on Ephydatia fluviatilis (Ef-PIWI, cryo-EM structure; PDB: 7KX9). Shown in orange and gray are the piRNA and target RNA bound to Ef-PIWI, respectively (I). (II) Zoomed-in cartoon showing the Ce-PIWI catalytic triad DDH, highlighted in magenta. PRG-1 D654 is mutated to alanine. (III) Zoomed-in cartoon showing the Ce-PIWI seed-gate (dark blue) and Ef-PIWI seed-gate (dark green). Highlighted in magenta are g9 and g10 of the piRNA. The PRG-1 seed-gate (L332-T350) was replaced with a Gly₆ linker.

See also Figure S4.

To test the first model, we assessed anti-piRNA production in catalytically inactive PRG-1 mutants. PRG-1 contains evolutionary conserved catalytic residues (DDH) essential for its slicer activity,⁴² although the slicer activity is dispensable for piRNA-mediated silencing.^36,37,47 Previous in vitro assays have shown that PRG-1 slicer activity is abolished by mutating its DDH triad to DAH.³⁷ CRISPR-Cas9 genome editing was used to introduce the DAH mutation (D654A) in the prg-1 genomic locus in a previous study (Figure 4D).⁴⁷ We observed that levels of piRNAs and anti-piRNAs in prg-1^D654A(gg630); parn-1 were comparable to those in parn-1 mutants (Figures S4B and S4C). Importantly, the distance between the 5′ ends of piRNAs and the 3′ ends of anti-piRNAs in prg-1 (D654A); parn-1 peaked at 9, similar to what was observed in parn-1 mutants (Figure 4B). Collectively, these findings suggest that PRG-1 slicer activity does not play a role in the production of anti-piRNAs.

To explore the second model, we examined the PRG-1 structure that is predicted by AlphaFold (Figure 4D).⁵⁸ The AlphaFold program reported high confidence scores for the PAZ, MID, and PIWI domains of PRG-1. We superimposed the predicted PRG-1 structure on the cryoelectron microscopy (cryo-EM) structure of the Piwi-piRNA-target complex from the sponge Ephydatia fluviatilis (Ef-Piwi) (Figure 4D).⁵⁹ Within the Ef-Piwi, Drosophila Piwi, and silkworm Siwi structures, a distinct a helix, referred to as the seed-gate, is present near the piRNA seed region (positions 2 to 8).⁵⁹ This seed-gate has been observed to move upon binding to targets and is essential for targeting fidelity.⁵⁹ The seed-gate structure was observed in PRG-1 (Figure 4D). Notably, it was in close proximity to positions 9 and 10 of piRNAs, where most of the anti-piRNAs terminate (Figure 4D). We therefore tested the idea that the seed-gate structure might obstruct RdRP-mediated anti-piRNA elongation. The disruption of the seed-gate (PRG-1^SG_mut, in which L332-T350 are replaced with a Gly₆ linker) in prg-1(how32); parn-1 led to a decrease in piRNA and anti-piRNA levels compared with parn-1 mutants (Figures S4B and S4C). Although it is important to note that this mutation could potentially alter the conformation of PRG-1, the disruption of the PRG-1 seed-gate resulted in the production of longer anti-piRNAs and a 1-nt shift in the distance between the 5′ ends of piRNAs and the 3′ ends of anti-piRNAs (Figure 4B). We conclude that the seed-gate within PRG-1 may impose steric hindrance that blocks RdRP activity, thereby hindering anti-piRNA elongation.

piRNA-target interactions promote the generation of anti-piRNAs

It is noteworthy that only a subset of piRNAs (n = 1,934) in parn-1 mutant strains were templated to generate anti-piRNAs (Table S2). Furthermore, there was a poor correlation between the abundance of anti-piRNAs and their corresponding piRNAs (Pearson’s ρ = 0.05) (Figure 5A), suggesting that the abundance of piRNA is not the determinant factor for anti-piRNA production. This raises the question of why certain piRNAs were selectively targeted by the RdRP complex. Previous studies demonstrated that base-pairing interactions between piRNAs and their targets can recruit RdRPs to target RNAs, which serve as templates to produce 22G-RNAs.^10,11,36,37 This led us to explore the possibility that piRNA::target interactions might also facilitate the synthesis of anti-piRNAs by recruiting RdRPs to untrimmed piRNAs.

Figure 5. piRNA::target interactions facilitate anti-piRNA synthesis.

(A) Scatterplot comparing the levels of sense and antisense sequences for a given piRNA (n = 1,934) in parn-1 mutants. The solid red line represents no difference. The upper and lower dashed red lines represent a fold difference of 2 and 0.5, respectively. Pearson’s correlation test was used to derive ρ.

(B) Boxplot showing the CLASH count per piRNA for piRNAs with anti-piRNA mapping sequences (n = 1,650) and the control set. The control set consists of 1,650 random piRNAs with abundance comparable to that of the experimental set but lacking anti-piRNA sequences. This random sampling process was repeated 1,000 times to obtain the distribution of the median CLASH count. ****p < 0.0001 two-tailed t test.

(C) Schematic illustrating the CLASH chimera between 21UR-4864 (black) and w03h9.2 (red). Shown in the blue box are the CLASH count detected for this interaction and the predicted free energy (ΔG) of the interaction.

(D) Schematic illustrating the w03h9.2 genomic locus, 21UR-4864 target site in the w03h9.2 transcript (red), the amino acids this sequence encodes, and deletion of the target site (w03h9.2_del).

(E) Genome browser view showing sense (blue) and antisense (magenta) reads mapping to 21UR-4864 in parn-1; w03h9.2_WT and parn-1; w03h9.2_del. Data are displayed as individual biological replicates.

See also Figure S5

The assessment of the target::piRNA interactome was performed using CLASH, an approach combining cross-linking, ligation of piRNAs and their targets, and subsequent sequencing of hybrid molecules.^60,61 In silico folding was conducted to determine the base-pairing interaction within hybrid reads, and the Gibbs free energy (ΔG) of the most energetically favorable interaction can be subsequently deduced (Figure S5A).⁶⁰ It is important to acknowledge that the CLASH experiment was performed in the wild-type background, but not in parn-1 mutants.⁶⁰ However, considering that the seed sequence of piRNAs largely governs piRNA::target interactions,⁶⁰ we posited that hybrid reads derived from wild type could serve as a proxy for piRNA:: target interactions in parn-1 mutants. Therefore, we cross-referenced piRNAs with anti-piRNA sequences defined in this study with piRNA::target interactions previously revealed by the CLASH experiment.⁶⁰ When focusing on robust piRNA::target interactions (ΔG ≤ −15 kcal/mol), we observed that of the 1,934 piRNAs containing anti-piRNA sequences, 1,650 exhibited detectable CLASH counts. To create a control group for comparison, we randomly selected 1,650 piRNAs with abundance comparable to that of our experimental set but lacking anti-piRNA sequences. This random sampling process was repeated 1,000 times to obtain the distribution of the median CLASH count (Figure 5B). Importantly, the CLASH counts for piRNAs with anti-piRNAs were found to be significantly higher than those in the control group (Figure 5B). This finding suggests a positive correlation between piRNA::target interactions and the synthesis of anti-piRNAs.

Next, we aimed at determining the causal relationship between piRNA::target interactions and anti-piRNA production. Our previous study established the base-pairing interaction between 21UR-4864 (piRNA) and the endogenous W03H9.2 transcript (target RNA).⁴⁸ This interaction was characterized by abundant CLASH reads and extensive base-pairing interactions (ΔG = −29.08 kcal/mol) (Figures 5C and 5D).⁴⁸ In this study, we discovered anti-piRNAs originating from the 21ur-4864 locus in parn-1 mutants (Figure 5E). We reasoned that if anti-piRNAs are indeed induced by base-pairing interactions between piRNAs and their targets, disrupting the interaction between 21UR-4864 and W03H9.2 would have a negative impact on anti-piRNA production. Using CRISPR-Cas9 genome editing, we generated a deletion allele of w03h9.2 (w03h9.2_del) that specifically deleted the 21UR-4864 binding site from W03H9.2 mRNA and removed seven amino acids from the W03H9.2 protein (Figure 5D).⁴⁸ Consistent with the idea that piRNA induces 22G-RNA production,^10,11,36,37 the deletion of the 21UR-4864 target site (w03h9.2_del) resulted in a reduction in 22G-RNAs that are antisense to the W03H9.2 transcript in parn-1 mutants (Figure S5B). When examining the expression of anti-piRNAs that are antisense to 21UR-4864, we found that w03h9.2_del mutants displayed 2.95-fold reduction in anti-piRNAs levels compared with the control group (Figure 5E). Taken together, our findings provide evidence that piRNA::target interactions promote the production of anti-piRNAs in parn-1 mutants.

PRG-1 loaded with piRNAs::anti-RNAs is partially defective in 22G-RNA production

We set out to assess the physiological relevance of anti-piRNAs. Because the base-pairing interactions between anti-piRNAs and piRNAs could potentially hinder piRNAs from recognizing their intended targets, we hypothesized that PRG-1, when loaded with piRNAs::anti-RNA, cannot effectively induce 22G-RNA synthesis. To test this hypothesis, we defined target sites for 1,934 piRNAs with detectable anti-piRNAs using the published CLASH data (Figure 6A).⁶⁰ This analysis revealed a total of 16,030 potential piRNA target sites (Figure 6A). There was an increase in 22G-RNA levels within a ±30 nt window of predicted targeting sites in parn-1 mutants compared with prg-1; parn-1 mutants (Figure 6A), This observation implies that the group of 1,934 piRNAs can engage PRG-1 to trigger 22G-RNA production. This result aligns with our expectation, considering that piRNAs are more abundant than their corresponding anti-piRNAs, and only a fraction of piRNAs engage in base-pairing with anti-piRNAs.

Figure 6. PRG-1 loaded with piRNAs::anti-RNAs is partially defective in 22G-RNA production.

(A) The density of 22G-RNAs within a 60-nt window around piRNA target sites for piRNAs with anti- piRNAs (red) and without anti-piRNAs (blue). The solid line shows the 22G-RNA density in parn-1 mutants (n = 2) and the dashed line shows density in prg-1; parn-1 mutants (n = 1). Data are presented as the mean of two biological replicates for piRNAs with anti-piRNAs and the mean of 10 random samplings for piRNAs lacking anti-piRNAs. Shaded areas indicate the standard deviation. The 5′ U of piRNAs is centered at position 0 and indicated by a green arrow

(B) Schematic illustrating the synthesis of 22G-RNAs and anti-piRNAs in wild-type and parn-1 mutant strains

We next aimed to investigate whether the presence of anti-RNAs hinders 22G-RNA production. To establish a control set for comparison, we randomly selected 16,030 piRNA target sites associated with piRNAs lacking anti-piRNA reads. This random sampling process was performed 10 times to enhance the rigor and robustness. We observed that 22G-RNA production was less pronounced in the group of 1,934 piRNAs compared with the control set (Figure 6A). In conclusion, our findings suggest that PRG-1 loaded with piRNAs::anti-RNAs may indeed exhibit a partial deficiency in 22G-RNA production.

DISCUSSION

From nematodes to mammals, the maturation of piRNAs invariably involves 3′ nucleolytic processing.^28,30-33 Our previous study demonstrated that the evolutionarily conserved ribonuclease PARN-1 mediates piRNA 3′ trimming in C. elegans.²⁸ In this study, we report that untrimmed piRNAs in parn-1 mutants were modified by RDE-3 and templated by EGO-1/EKL-1/DRH-3 complex to produce anti-piRNAs (Figure 6B). In the wild type, piRNA targeting may trigger the cleavage of target RNAs whose 3′ terminus is subject to poly(UG) addition.^45,46 RdRP complexes are recruited to poly(UG) targets, which serve as templates to produce 22G-RNAs that are subsequently loaded to WAGOs (Figure 6B). In parn-1 mutants, however, the 3′ terminus of untrimmed piRNAs may fail to be accommodated properly by the PAZ domain of PRG-1. This results in the erroneous recognition and modification of these long untrimmed piRNAs by RDE-3. Upon recruitment, RdRP complexes not only use targets as templates to generate 22G-RNAs, but also use untrimmed piRNAs (with or without UG/GU additions) as templates to synthesize anti-piRNAs (Figure 6B). Interestingly, while 22G-RNA production depends on the RdRPs EGO-1 and RRF-1, the synthesis of anti-piRNAs primarily relies on EGO-1 (Figure 3D). The elongation of anti-piRNAs is halted by the seed-gate structure of PRG-1, resulting in the synthesis of 17- to 19-nt anti-piRNAs that associate with the PRG-1 protein (Figure 6B). With regard to the physiological significance of anti-piRNAs, we speculate that the base-pairing interaction between anti-piRNAs and piRNAs blocks piRNAs from binding to targets. Consistent with this idea, piRNAs with anti-piRNA reads are partially defective in producing 22G-RNA at the piRNA target sites (Figure 6A). Collectively, our findings underscore the critical role of nucleolytic processing at the piRNA 3′ terminus and unveils a novel class of small RNAs.

Although anti-piRNAs were exclusively identified in the parn-1 mutant background, their discovery sheds light on the enzymatic activities of RDE-3, RRF-1, and EGO-1. Previous studies described RDE-3 as the terminal nucleotidyltransferase adding alternating U and G to long RNAs during RNAi.^45,46 Our data provide additional insights into RDE-3 substrates and activities. First, RDE-3 can target small RNAs that are not properly processed, indicating that substrate lengths do not determine RDE-3 targeting. Second, RDE-3 has the capacity to catalyze UG dinucleotide additions (Figure 3B), which raises the possibility that RDE-3 may not exhibit high processivity in vivo. It is important to note that our small RNA cloning method may not capture piRNAs with long poly(UG) tails, as these long tails form atypical RNA quadruplex complexes, which may hinder adaptor ligation.⁶² Alternatively, some cellular ribonucleases may shorten piRNA poly(UG) tails. To comprehensively characterize RDE-3′s enzymatic activities and define its substrate specificity, further investigations using biochemical assays are warranted.

Two RdRPs, RRF-1 and EGO-1, are thought to function partially redundantly in the production of 22G-RNAs.⁴⁰ However, our findings suggest that they have distinct roles in anti-piRNA synthesis. Specially, RRF-1 is dispensable for anti-piRNA production, while EGO-1 emerges as the primary RdRP responsible for anti-RNA production (Figure 3D). Previous studies have showed that RRF-1 exhibited stronger binding to poly(UG) tails compared with EGO-1.^45,46 This suggests that poly(UG) tails may be crucial for recruiting RRF-1 to target RNAs but less so for EGO-1 recruitment (Figure 3D). This idea aligns with our observations of modest anti-piRNA depletion in rde-3 mutants and robust anti-piRNA depletion in ego-1 mutants. Furthermore, our data reveal that EGO-1 initiates anti-piRNA synthesis with either an A or a G as the first base, occurring at similar frequencies (Figure 1G). When recruited to RNA targets, EGO-1 may generate secondary siRNAs starting with either A or G. The preference for 5′ G found in 22G-RNAs is possibly dictated by WAGO proteins. In vitro biochemical experiments and structural analysis will be needed to establish the distinct activities of RRF-1 and EGO-1.

Limitations of the study

The current study is subject to several limitations. First, anti-piRNAs are found exclusively in the parn-1 mutants when the 3′ end of piRNAs is not properly trimmed (Figures 1 and 2). We have not yet identified the conditions, such as environmental stresses or transposon activation, that are conducive to the production of anti-piRNAs in wild-type worms. Second, while the complete absence of a 21UR-4864 target site resulted in reduced anti-piRNA production (Figure 5), we did not systematically investigate the impact of piRNA::target base-pairing interactions on anti-piRNA production. Finally, although our genetic experiments support the model that untrimmed piRNAs were modified by RDE-3 and templated by EGO-1 to generate anti-piRNAs, and that the PRG-1 seed-gate structure impedes anti-piRNA elongation (Figure 6), conclusive biochemical data validating this model are lacking. Future investigations, incorporating biochemical assays and structural analysis, are therefore warranted.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests regarding resources and regents used in this manuscript can be directed to and will be fulfilled by the Lead Contact, Wen Tang (tang.542@osu.edu).

Materials availability

All materials generated in this study are available from the lead contact without restriction.

Data and code availability

• This paper analyzes existing, publicly available data. These accession numbers for the datasets are listed in the key resources table. Small RNA sequencing data from N2 and parn-1 young adult animals have been deposited at NCBI: PRJNA683039.⁴⁸ PRG-1 CLASH sequencing data have been deposited at NCBI: SRP131397.⁶⁰ Small RNA sequencing from parn-1; prg-1 double mutants have been deposited at NCBI: PRJNA291851.²⁸ All sequencing data generated in this study have been deposited at NCBI: GSE244073.

• The small RNA sequencing pipeline used to preprocess, align, and preform initial analysis of the data is available under https://doi.org/10.5281/zenodo.8302723 or at github.com/benpastore/nextflow_smRNA. This pipeline was constructed using Next-flow. ⁷⁴ Custom code generated for analysis in this study are available at https://doi.org/10.5281/zenodo.10304098 or github. com/benpastore/2023_antipiRNA.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-PRG-1	Batista et al.15	PMC2570341
Bacterial and virus strains
Bacteria: OP50	Caenorhabditis Genetics Center	https://cgc.umn.edu/strain/OP50
Bacteria: ego-1 RNAi food	Ahringer RNAi Collection (Source Biosciences)	Kamath et al.63
Bacteria: drh-3 RNAi food	Ahringer RNAi Collection (Source Biosciences)	Kamath et al.63
Bacteria: ekl-1 RNAi food	Ahringer RNAi Collection (Source Biosciences)	Kamath et al.63
Chemicals, peptides, and recombinant proteins
TRI Reagent	Thermo Fisher Scientific	Cat#AM9738
1-Bromo-3-chloropropane	Sigma Aldrich	Cat#B62404
PIR-1	Li et al.64	PMC6961543
T4 RNA Ligase 1	New England Biolabs	Cat# M0437
T4 RNA Ligase 2, truncated	New England Biolabs	Cat# M0242
SUPERaseIn	Thermo Fisher Scientific	Cat#AM2694
Q5 High-Fidelity DNA Polymerase	New England Biolabs	Cat#M0491
Ex TaqDNA Polymerase	TaKaRa	Cat#RR001C
Go TaqDNA Polymerase	Promega	Cat#M300A
dNTPs	Roche	Cat#3622614001
Sodium Hypochlorite Solution	Fisher Scientific	Cat#SS2901
Sodium meta-Periodate	Fisher Scientific	Cat#S398-100
Sodium Tetraborate Decahydrate	Fisher Scientific	Cat#S246-500
Boric acid	Sigma Aldrich	Cat#B0394-100G
glycerol	Fisher Scientific	Cat#AC327255000
Protein A/G dynabeads	Thermo Fisher Scientific	Cat# 80105G
pRF4 injection marker, rol-6(su1006)	Ghanta et al.65	PMC8417391
Critical commercial assays
mirVana miRNA Isolation Kit	Thermo Fisher Scientific	Cat#AM1560
SuperScript IV Reverse Transcriptase	Thermo Fisher Scientific	Cat#18090200
Deposited data
Raw and analyzed original sequencing data generated for use in this study	This study	GEO: GSE244073
Small RNA sequencing from N2 and parn-1	Pastore et al.48	PRJNA683039
CLASH data	Shen et al.60	SRP131397
Small RNA sequencing from parn-1; prg-1 double mutants.	Tang et al.28	PRJNA291851
Experimental models: Organisms/strains
C. elegans: N2 Bristol (wild-type)	CGC	CGC
C. elegans: parn-1 (tm869) V	Outcrossed with N2. Pastore et al.48	PMC8459939
C. elegans: prg-1 (tm872) I; parn-1 (tm869) V	Tang et al.28	PMC4785802
C. elegans: rrf-1 (ok589) I; parn-1 (tm869) V	This study	NA
C. elegans: rde-3 (ne3370) I; parn-1 (tm869) V	This study	NA
C. elegans: drh-3 (ne4253) I; parn-1 (tm869) V	This study	NA
C. elegans: prg-1 (how32) I; parn-1 (tm869) V	This study	NA
C. elegans: prg-1(gg630) I; parn-1(tm869) V	This study	NA
C. elegans: w03h9.2 (how33) II; parn-1(tm869) V	This study	NA
Oligonucleotides
Oligos used in this study	This study	Table S4
Recombinant Protein
Alt-R S.p. Cas9 Nuclease V3	Integrated DNA Technologies	Cat# 1081058
Software and algorithms
Bowtie version 1.2.3	Langmead et al.65	http://bowtie-bio.sourceforge.net/manual.shtml
BEDtools version 2.26.2	Quinlan et al.66	https://bedtools.readthedocs.io/en/latest/
Trim Galore version 0.6.4 d	Martin67	http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/
Tailor	Chou et al.54	https://github.com/jhhung/Tailor
BEDOPS version 2.4.39	Neph et al.68	https://bedops.readthedocs.io/en/latest/
kentUtils	Kent et al.69	https://hgdownload.soe.ucsc.edu/downloads.html#source_downloads
Integrative Genomics Viewer (IGV)	Robinson et al.70	https://igv.org/
R version 4.0.3	The R Project for Statistical Computing	https://www.r-project.org/
dplyr version 1.0.2	Wickham et al.71	https://dplyr.tidyverse.org/
ggplot2 version 3.3.2	Wickham et al.71	https://ggplot2.tidyverse.org/
tidyr version 1.1.2	Wickham et al.71	https://tidyr.tidyverse.org/
Vienna RNA package version 2.3.5	Gruber et al.72	https://www.tbi.univie.ac.at/RNA/
Nextflow	Di Tomasso et al.73	https://www.nextflow.io/
Nextflow smRNA sequencing pipeline	This study	https://doi.org/10.5281/zenodo.8302723
Custom codes specific to this study	This study	https://doi.org/10.5281/zenodo.10304097

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

The Bristol strain N2 was designated as the wild-type C. elegans strain.⁷⁵ All other strains and allele information are listed in the key resources table and Table S3. All strains were fed a diet of E. coli OP50 and maintained on Nematode Growth Media (NGM). All animals were maintained at 20 °C unless otherwise indicated.

METHOD DETAILS

CRISPR/Cas9 genome editing

PRG-1^SG_mut strains were generated using CRISPR/Cas9 genome editing and single stranded DNA donors in the parn-1 mutant background.⁷⁶ Cas9 loaded with gRNA targeting the PRG-1 seed gate genomic sequence was microinjected to parn-1 adult animals along with ssDNA donors with 35 homology sequences. The PRF4 vector containing the dominant allele of rol-6 served as an injection marker.⁷⁶ Guide RNA, DNA donor and genotyping primer sequences can be found in Table S4.

RNA interference by feeding of double stranded RNA

The HT115 RNAi feeding strains were picked from the C. elegans Ahringer RNAi library.⁶³ For RNA purification experiment 60,000 synchronized L1 larvae were platted on 150 mm NGM plates containing 50 μg/ml ampicillin and 5 mM IPTG seeded with HT115 bacteria against target genes. After reaching adulthood animals were collected with M9 and subjected to RNA extraction and small RNA enrichment.

PRG-1 immunoprecipitation and RNA ligation

60,000 synchronous N2 and parn-1 adult were collected, suspended and homogenized in one volume of IP buffer (110 mM Potassium acetate, 2 mM magnesium acetate, 0.1% Tween 20, 0.5% Triton, 1mM DTT, 20 mM HEPES-KOH, pH 7.5) with complete protease inhibitors (Roche) using FastPrep-24 homogenizer (MP Biomedicals). Lysates were cleared by two rounds of centrifugation at 14,000 x g for 10 min. Cell extracts were incubated with 12 μg anti-PRG-1 antibody¹⁵ and 60 μl Protein A/G dynabeads (Thermo Fisher Scientific) at 4°C for 3 h. Beads were washed three times with IP buffer with protease inhibitors and once with wash buffer (150 mM NaCl, 2 mM magnesium acetate, 50 mM Tris-HCl pH 7.5). Beads were incubated with Tobacco Acid Pyrophosphatase (Epicentre) and washed once with wash buffer. In a 150 μl reaction, beads were incubated with 0.5 U/ μl T4 RNA ligase 1 (NEB) in 1x ligase buffer (containing 1 mM ATP, 1 U/uL RNase inhibitor and 10% PEG-8000) on the rocker at 16°C overnight. RNA was recovered by treatment with proteinase K (2.0 mg ml⁻¹ in 0.5% (w/v) SDS, 40 mM EDTA, 20 mM Tris-HCl, pH 7.5) at 50°C for 10 min, followed by extraction with TRI Reagent (Sigma) and ethanol precipitation. To enrich the ligated product, RNAs ranging from 26 to 50 nts were purified from the 15% polyacrylamide/7M urea gel and subject to small RNA cloning.

RNA extraction and small RNA enrichment

60,000 L1 larvae were plated to 150 mm NGM plates supplemented with 2 mL of concentrated OP50 food. Approximately 64 hours after plating L1 larvae, day 1 adults with eggs were collected from plates using M9 solution. Harvested animals were washed twice with M9 and once with water. Animals were then suspended in TRI Reagent solution (Sigma) and frozen at −80°C until RNA extraction was performed. To extract RNAs animals suspended in TRI Reagent solution were thawed and lysed using a Bead Mill (Thermo Fisher Scientific). Bromo-chloropropane (Sigma) was added to the lysed worms and centrifuged at 12,000 x g for 15 minutes. Following centrifugation, the aqueous phase was pipetted to an equal volume of cold isopropanol and placed at −20°C for 1 hour to precipitate RNAs. RNAs were pelleted by centrifugation at 20,000 x g for 15 minutes at 4°C. Total RNAs were washed with 75% EtOH twice and suspended in RNAse free water. Small RNAs were enriched from total RNAs using the MirVana miRNA isolation kit according (Thermo Fisher Scientific).

Oxidation of small RNAs with NaIO₄

5 μg small RNAs were oxidized with 25 mM sodium periodate (Fisher Scientific) in borax/boric acid buffer (0.06 M borax, 0.06 M boric acid, pH 8.6) in the dark at 20°C for 30 minutes. 20 μL of 50% glycerol was added and incubated for an additional 15 minutes to quench the sodium periodate. Oxidized RNAs were eluted from the reaction using EtOH precipitation at −20°C. Eluted RNAs were suspended in RNAse free water and subsequently used in small RNA cloning.

Small RNA-sequencing library preparation

Small RNAs were treated with RNA phosphatase PIR-1 to remove γ and β phosphates from the 5′ triphosphorylated RNAs.⁶⁴ Monophosphorylated RNAs were ligated to 3’ adapters (rAppAGATCGGAAGAGCACACGTCTGAACTCCAGTCA/3ddC/, IDT) using T4 RNA ligase 2 in 25% PEG 8000 (NEB) at 15°C overnight. A 5′ adapter (rArCrArCrUrCrUrUrUrCrCrCrUrArCrArCrGrArCrGrCrUrCrUrUrCrCrGrArUrCrU, IDT) was then ligated to RNAs to the product using T4 RNA ligase 1 (NEB) for 4 hours at 15°C. Ligated products were reverse transcribed using SuperScript IV Reverse Transcriptase (Thermo Fisher Scientific) to convert RNA to cDNA libraries. cDNA libraries were amplified by PCR and subsequently sequenced on an Illumina Novaseq platform (SP2 x 50 bp) at the OSU Comprehensive Cancer Center genomics core.

Small RNA-sequencing data analysis

Adapters and reads with poor quality were removed using TrimGalore.⁶⁷ Identical sequences were counted and collapsed into fasta format using a customized shell script. Processed reads were aligned to a reference containing rRNAs, tRNAs, snRNA and snoRNAs. Reads aligning to this reference were excluded from downstream analysis as they likely represent contaminating RNA degradation products and can skew sample normalization. Reads were then aligned to the genome reference (WormBase, WS279) as well as a reference containing the sequences of exon-exon junctions using Bowtie.⁶⁵ Reads aligning to the genome or junction reference were converted to bed format using BEDOPS.⁶⁸ Alignments were intersected to a bed file containing gene annotations using BEDtools intersect.⁶⁶ Reads overlapping with genomic features were filtered according to the following rules using a customized python script: 1) 22G/A siRNAs: reads mapping antisense to protein coding gene, lincRNA, pseudogenes or transposable element exons with a length of 21-23 nt and containing a 5’ G or A. The read must align perfectly but may map to up to 1000 genomic locations. 2) piRNAs: reads mapping sense to piRNA genes, containing a 5’ U of any length. The 5’ end of the read must map within 0,1, or 2 nucleotides relative to the annotated 5’ end of the piRNA gene. Reads mapping to piRNAs must map perfectly and uniquely. 3) Anti-piRNAs: reads mapping antisense to piRNAs. No filters were placed on 5’ nt identity or length. The read must align perfectly and uniquely. Anti-piRNAs were further filtered after preliminary analysis to be 17-19 nucleotides in length with a 5’ G/A. 4) miRNAs: Reads mapping to annotated miRNAs in the sense orientation with any 5′ nt and length. Following this filtration raw read counts were normalized to the total number of reads mapping to the genome or miRNAs, and then scaled to 1,000,000 total reads. The per-gene read count was then aggregated for each gene in each sample and used in downstream analysis. To generate BigWig files, bedfiles containing normalized reads passing filters for each genomic feature described above were converted to BedGraph files using BEDOPS.⁶⁸ BedGraph files were converted to BigWig files using tools from UCSC.⁶⁹ BigWig files were visualized in IGV.⁷⁰ 4,012 CSR-1 targets and 1,350 WAGO-1 targets were determined as described.^55,57,77

Analysis of PRG-1 IP and RNA ligation data

Raw reads from PRG-1 IP/RNA ligation experiments were processed as described above with unaligned reads retained for downstream analysis. Retained reads were aligned to a piRNA reference in which the annotated 3’ end of piRNAs was extended 50-nt. Alignments were filtered to select only those that: 1) mapped in the sense orientation, 2) mapped within position 0, 1, or 2 relative to the annotated 5’ terminus of the reference and 3) contained a tail or unaligned sequences that were at least 10-nt in length. Wild-type piRNA sequences had to precisely span 21 nucleotides, while parn-1 samples allowed sense sequences of 21 nucleotides and longer. Unaligned sequences were subjected to an additional re-alignment step, demanding an antisense orientation to the piRNA reference. The identification of genuine piRNA::anti-piRNA ligation products was achieved through a customized Python script, merging reads containing both sense and antisense piRNA mapping sequences.

Analysis of non-templated nucleotide additions

Processed and collapsed fasta reads were aligned to the genome reference (WormBase, WS279) allowing zero mismatches. Reads failing to align underwent a secondary alignment with Tailor,⁵⁴ followed by conversion to bed files and assignment to specific genomic coordinates using BEDtools intersect.⁶⁶ The same rules used for filtering reads overlapping genomic features described above were also used in the analysis of tailed reads. After annotating genes associated with tailed reads, further filtration involved: 1) retaining reads with 1-nucleotide tails. 2) for reads having tails of 2 to 5 nucleotides, requiring the edit distance (e.g. number of mismatches between the reference and read) to be equal to the length of tails. Reads failing this but being at least 3 nucleotides long were retained if their tail consisted of two alternating nucleotides (e.g., GUG/UGU). 3) preserving reads of at least 5 nucleotides entirely composed of a single nucleotide or two perfectly alternating nucleotides. Tailed reads meeting these criteria were normalized using the same constants as perfectly matched reads, as previously described. Tailed reads underwent further noise reduction by considering their presence in both sequencing replicates and the abundance of perfectly matched counterparts mapping to the corresponding genes. To filter according to abundance the Z score of perfectly matched reads mapping to a given gene was calculated in each genetic background, genes with a Z score ≥ 1 were retained for downstream analysis.

CLASH data analysis

Published PRG-1 CLASH data were analyzed as previously described.⁶⁰ Briefly, Adapters and reads with poor quality were removed using TrimGalore.⁶⁷ Processed reads were collapsed, and aligned to the C. elegans reference. After identifying CLASH chimeras, the ΔG (kcal/mol) of piRNA::target interactions were calculated using tools from the Vienna RNA package (version 2.3.5).⁷² To ensure the inclusion of confident piRNA::target interactions, those with a ΔG exceeding −15 kCal/mol were excluded from subsequent analysis. 1,650 piRNAs with anti-piRNA matching reads had detectable CLASH counts.⁶⁰ Equitable comparisons were conducted by randomly sampling an equal number of piRNAs from the overall piRNA pool, ensuring that their abundance distributions closely matched those of the 1,650 piRNAs in our experimental group. This random sampling process was iterated 1,000 times to obtain the distribution of median CLASH count per piRNA for each sampling iteration, which was compared to the experimental set.

Metagene analysis of 22G RNA density around piRNA target sites

A total of 16,030 piRNA target sites were identified for piRNAs with anti-piRNA mapping based on previously published PRG-1 CLASH data⁶⁰. These CLASH defined piRNA target sites were extended 30-nt upstream and 30-nt downstream of the piRNA 5’ U. For each piRNA target site, 22G-RNA density in parn-1 mutants was calculated. The control set comprised of randomly selected 16,030 piRNA target sites for piRNAs lacking anti-piRNA reads. This random sampling process was performed 10 times to enhance the rigor and robustness of the analysis.

QUANTIFICATION AND STATISTICAL ANALYSIS

The details regarding published and custom tools used for quantification and statistical analysis are specified in the STAR★Methods. Specific information regarding the statistical tests used can be found in the figure legends. Data were tested for normality before preforming parametric statistical tests. Statistical tests used are indicated in the Figure legends. Statistical significance was defined using a p-value alpha of 0.05. Plotting and analysis in R relied heavily on the R packages ggplot2 and dplyr.⁷¹

Highlights.

• Loss of C. elegans PARN-1 leads to accumulation of anti-piRNAs

• EGO-1 uses untrimmed piRNAs as templates to produce anti-piRNAs

• The seed-gate structure within the Piwi protein hinders anti-piRNA elongation

• piRNA-target interactions facilitate the generation of anti-piRNAs