The piRNAs of C. elegans Are 21 Nucleotide RNAs

In Drosophila, zebrafish, mice and rats, Piwi proteins are associated with 24-30 nt piRNAs (Aravin et al., 2006; Brennecke et al., 2007; Girard et al., 2006; Grivna et al., 2006; Gunawardane et al., 2007; Houwing et al., 2007; Lau et al., 2006; Saito et al., 2006; Vagin et al., 2006; Watanabe et al., 2006). Searching for piRNAs in C. elegans we were unable to identify an abundant class of RNAs in this size range (data not shown). We therefore searched for piRNAs among the small RNAs previously identified in C. elegans: miRNAs, tncRNAs, endogenous siRNAs and 21U-RNAs (Ambros et al., 2003; Lau et al., 2001; Lee and Ambros, 2001; Lim et al., 2003; Ruby et al., 2006). To identify candidate piRNAs we tested if any of these short RNAs were dependent on Piwi. Surprisingly, we found that a 21U-RNA, 21UR-1, which was detected in RNA from wild-type animals was absent in RNA from two independent piwi mutants by northern blotting (Figure 1A). In contrast, expression of a ubiquitous miRNA, miR-52, was unaffected. To test if 21U-RNAs were generally absent in piwi mutants we generated libraries of 5′ monophosphate small RNAs from wild-type and piwi(n4357; n4358) mutants. High-throughput sequencing identified 1398 out of 5454 previously known 21U-RNAs (Ruby et al., 2006). We also identified a large number of candidate 21U-RNAs (data not shown). 21U-RNAs were either absent or dramatically under-represented in the piwi mutant library as compared to the wild-type library (Figure 1B). The most abundant 21U-RNA in the piwi sample had 8 reads as compared to 2127 reads in wild-type. We also assessed the expression levels of other small RNAs in wild-type versus piwi mutants and found no differences in miRNA expression, tncRNA expression, or a number of siRNA species including a 26 nucleotide siRNA (Figure 1C and data not shown). These data suggest that 21U-RNAs might be the piRNAs of C. elegans. To test this hypothesis directly, we generated a rabbit polyclonal antibody against PRG-1. After immunoprecipitation of PRG-1 using αPRG-1 serum from wild-type C. elegans adult whole cell extracts we detected 21U-RNAs by RT-PCR, but were not able to detect them from piwi mutant extracts or when using pre-immune serum (Figure 1D). Overall, 21U-RNAs are 100-fold enriched in PRG-1 immunoprecipitates ((Batista et al., 2008) and Discussion). As the high-throughput sequencing data suggested that 21U-RNAs were dramatically reduced in piwi mutants, we independently quantified the expression of seven 21U-RNAs by quantitative RT-PCR. As shown in Figure 1E, while the expression of a number of 21U-RNAs is dramatically reduced, some 21U-RNAs, including 21UR-1 were still detected in piwi mutants. For 21UR-1 we verified that the signal was specific by cloning and sequencing of RT-PCR products (data not shown). Taken together these data suggest that the 21U-RNAs are the piRNAs of C. elegans and we will refer to them as piRNAs below.

Open in a separate window

Figure 1

The 21U-RNAs are C. elegans piRNAs

(A) Northern blot showing that 21UR-1 is not detected in RNA (40 μg) isolated from two independent piwi double-mutants: piwi(n4357; n4358) and piwi(n4503; nDf57), whereas miR-52 is expressed in both mutants (same blot re-probed). Antisense DNA probes were used for 21UR-1 and miR-52. A U6 northern blot is shown as loading control. piwi(n4357; n4358) is an abbreviation for prg-1(n4357); prg-2( n4358). piwi(n4503; nDf57) is an abbreviation for prg-1(n4503); prg-2( nDf57).

(B) High-throughput sequencing reveals that the expression of many 21U-RNAs is dramatically reduced in piwi mutants. 21U-RNAs cloned from 5′ dependent wild-type and piwi(n4357; n4358) mutant libraries. Frequencies are shown for wild-type (blue) and piwi mutant (red) for the 400 most abundant 21U-RNAs in wild-type, plotted in the order of their wild-type frequency. Read frequencies were obtained by dividing the number of reads for a given 21U-RNA by the total number of reads from the same library (left-hand y-axis). The corresponding absolute number of reads are indicated in the right-hand y-axes. 21U-RNAs for which frequencies are shown include the 21U-RNA with most reads in the piwi mutant library (21UR-3224), which was sequenced 8 times in the piwi mutant and 2,127 times in wild-type.

(C) Expression of a 23 nucleotide antisense RNA (siR23-69) and a 26 nucleotide antisense RNA (siR26-263) is not affected in piwi mutants (northern blotting, 40 μg total RNA, antisense DNA probes).

(D) Immunoprecipitation followed by RT-PCR for 21UR-5101 reveals that 21U-RNAs are associated with PRG-1 in C. elegans extracts.

(E) Quantitative RT-PCR of seven 21U-RNAs demonstrates that Piwi is not essential for 21U-RNA biogenesis. Total RNA was extracted from 12 hour adult C. elegans. Expression levels shown are relative to levels in wild-type RNA. miR-52 expression was used as an internal control. Data are from three independent biological replicates. Error bars represent standard error of the mean.

piRNA Biogenesis is Independent of Many Genes in Other Small RNA Pathways

To identify additional genes involved in piRNA pathways in C. elegans we tested a panel of genes using mutants and RNAi for their effect on piRNA expression by northern blotting or quantitative RT-PCR (Figure 2 and Table S1). First, we checked piRNA expression in prg-1 and prg-2 single mutants. Using three independent alleles and RNAi experiments we found that only PRG-1 but not PRG-2 is required for piRNA expression (Figure 2 and Table S1). Next, we tested other Argonaute proteins; these included RDE-1, which is required for exogenous RNAi (Tabara et al., 1999), ALG-1 and ALG-2, which are redundantly required for miRNA function (Grishok et al., 2001), ERGO-1, SAGO-1, SAGO-2 and a group of five “MAGO” Argonaute proteins associated with endogenous siRNA pathways (Yigit et al., 2006). None of these Argonaute proteins were required for piRNA expression emphasizing the specificity of the requirement for PRG-1. In addition, we tested a number of other proteins involved in small RNA biology in C. elegans including all four known RNA-dependent RNA polymerases (RdRPs), RRF-1, RRF-2, RRF-3 and EGO-1 (Simmer et al., 2002; Smardon et al., 2000), none of which were required for piRNA expression. We also tested if the RNase III enzyme DCR-1, which is essential for the generation of siRNAs and germline development (Grishok et al., 2001; Ketting et al., 2001; Knight and Bass, 2001), is required for piRNA expression. As dcr-1 mutants are sterile, these experiments were carried out using homozygous mutant animals derived from heterozygous mothers or using RNAi. We found that DCR-1 is not required for piRNA expression (Figure 2B, Table S1). These data were also confirmed by high-throughput sequencing in dcr-1 mutants (Figure 2C). These data suggest that piRNAs are independent of many proteins involved in other small RNA pathways, including DCR-1.

Open in a separate window

Figure 2

piRNA biogenesis does not require many known small RNA pathway proteins

(A) 21UR-1 northern blotting of total RNA of wild-type and mutant young adult C. elegans. In the case of alg-1; alg-2(RNAi), alg-1 mutant L1 larvae were transferred to alg-2 RNAi feeding plates and young adult animals were harvested. A U6 northern blot is shown as loading control. See Table S1 in Supplemental Data for quantification of these results.

(B) 21UR-1 northern blotting of total RNA of wild-type and mutant young adult C. elegans. dcr-1 mutant animals used were homozygous animals derived from heterozygous mothers. To test for loss of DCR-1 activity in dcr-1 mutants, let-7 miRNA and pre-miRNA is shown. A U6 northern blot is shown as loading control.

(C) Distribution of 21U-RNAs on chromosome IV as detected by high-throughput sequencing of 5′ dependent wild-type, piwi(n4357; n4358), dcr-1 and mut-7 mutant libraries. Frequencies for a given 21U-RNA and locus were obtained by correcting the number of reads for multiple alignments and dividing by the total number of reads from the same library. Cumulative frequencies were plotted for non-overlapping 100 kb windows along chromosome IV.

Piwi and piRNAs are Restricted to the Male and Female Germline

We found that 21UR-1 expression is developmentally regulated with highest expression detected in young adults and adults by northern blot (Figure 3A). Additional piRNAs showed a similar expression pattern, as determined by quantitative RT-PCR (e.g. 21UR-5101, Figure 3B). Interestingly, prg-1 and prg-2 mRNAs show a similar pattern (Figure 3C, Figure S2). The observed expression pattern for piRNAs and piwi mRNA is consistent with expression in the germline. As Piwi proteins and piRNAs are thought to be exclusively expressed in germ cells in vertebrates (Carmell et al., 2007; Deng and Lin, 2002; Kuramochi-Miyagawa et al., 2004; Kuramochi-Miyagawa et al., 2001), but not in Drosophila (Brennecke et al., 2007; Cox et al., 1998; Gunawardane et al., 2007; Saito et al., 2006), we decided to test their germline restriction in C. elegans using a set of temperature-sensitive mutants (Figure 3D). 21UR-1 was absent from glp-4(bn2ts) and glp-1(e2144lf) mutant animals at the restrictive temperature, which are devoid of germ cells (Beanan and Strome, 1992). However, 21UR-1 levels were unchanged in glp-1(ar202gf,ts) mutants at the restrictive temperature, which are highly enriched in germ cells that have not yet entered meiosis (Pepper et al., 2003). 21UR-1 RNA was also detected in RNA from fem-1(hc17ts) (Kimble et al., 1984; Nelson et al., 1978) and fem-3(q22sd,ts) (Barton et al., 1987) mutants at the restrictive temperature that are devoid of sperm and oocytes respectively. The same restricted pattern was also observed for piwi mRNA (Figure 3E and Figure S2). Together these data suggested that C. elegans Piwi and piRNAs are restricted to the male and female germline.

Open in a separate window

Figure 3

Piwi and piRNAs are restricted to the male and female germline

(A) Profile of 21U-R1 expression during development. E, embryo. L1-L4, larval stages 1-4. YA, 12-hour adult. A, 24-48-hour adult. A U6 northern blot is shown as loading control.

(B) Quantitative RT-PCR of 21UR-5101. miR-52 expression was used as an internal control. Data are from three independent biological replicates. Error bars represent standard error of the mean.

(C) Quantitative RT-PCR of prg-1 mRNA. Actin mRNA was used as an internal control. Data are from three independent biological replicates. Error bars represent standard error of the mean.

(D) Northern blot showing that piRNA expression is restricted to the male and female germline. glp-1(gf), glp-1(lf), glp-4, fem-1 and fem-3 L1 larvae were grown to 12-hour adult stage at 25°C. 20 μg of total RNA was loaded in each lane. U6 and miR-52 northern blots are shown as loading controls.

(E) Quantitative RT-PCR of prg-1 mRNA. Actin mRNA was used as an internal control.

Piwi is Specifically Required for Silencing of Tc3 DNA Transposons

As Piwi proteins are required for transposon silencing in Drosophila (Aravin et al., 2001; Savitsky et al., 2006; Vagin et al., 2006) we examined transposon silencing in piwi mutants in C. elegans. We focused on the two most abundant DNA transposons in C. elegans, Tc1 and Tc3 (Figure 4A, B) (Consortium, 1998). First, we examined Tc1 transposase expression by quantitative RT-PCR using primers specific for 15 Tc1 loci (Figure 4C). We observed an approximately 50 fold increase in Tc1 transposase mRNA levels in mut-7 mutants, which show an elevated rate of transposition of Tc/mariner elements in the germline (Ketting et al., 1999). We did not, however, observe an increase in Tc1 transposase mRNA in piwi mutants. Surprisingly then, we found increased Tc3 transposase mRNA in two independent piwi mutants and three independent prg-1 mutants using primer pairs specific for 20 Tc3 loci (Figure 4D). Next, we assayed the germline excision rate of Tc1 and Tc3 transposons directly by the phenotypic reversion of unc-22 transposon insertion alleles (Ketting et al., 1999). As shown in Table 1, unc-22 reversion rates of Tc1, Tc3 or Tc4 insertion alleles were less than 10⁻⁶ in an otherwise wild-type background. In mut-7 mutant animals the reversion rate was increased 100 fold, as reported previously (Ketting et al., 1999). Strikingly, in prg-1 and piwi mutants Tc1 and Tc4 excision rates were not affected, but reversion of the Tc3 allele was increased 100 and 1000 fold, respectively (Table 1). These data suggest that PRG-1 and Piwi are powerful and specific suppressors of Tc3 transposition in the germline of C. elegans. As we found that Piwi and MUT-7 were both involved in germline transposon silencing we wondered if these proteins shared additional functions. We therefore tested if Piwi, like MUT-7, is required for germline transgene silencing (Kim et al., 2005). However, we found that germline transgene silencing is intact in piwi mutants (Figure S3). These data suggest that Piwi and MUT-7 have overlapping and distinct roles in the C. elegans germline.

Open in a separate window

Figure 4

Piwi is required to inhibit Tc3 transposase expression

(A) Diagram of the genomic structure of the two most common DNA transposons in C. elegans, Tc1 and Tc3. Tc1 and Tc3 are flanked by inverted repeats and encode a single spliced transcript for transposase. bp, base pairs. n, number of copies in the wild-type strain N2.

(B) Distribution of Tc1 and Tc3 transposons in the C. elegans genome.

(C), (D) Quantitative RT-PCR of Tc1 or Tc3 transposase mRNA. As the genomic copies of Tc1 and Tc3 have minor sequence variations, the number of transposon loci amplified by each qRT-PCR primer pair are shown (n). Actin mRNA was used as an internal control. Expression levels shown are relative to levels from wild-type RNA. Data are from three independent biological replicates. Error bars represent standard error of the mean.

Biogenesis of piRNAs

The striking difference in genomic organization of C. elegans piRNA clusters compared to vertebrates and Drosophila might suggest a divergent mechanism of biogenesis. We found that C. elegans piRNAs are Dicer independent (Figure 2B), as are Drosophila and zebrafish piRNAs (Houwing et al., 2007; Vagin et al., 2006). Taken together with the strong local piRNA strand bias in all three species (Brennecke et al., 2007; Houwing et al., 2007; Ruby et al., 2006; Vagin et al., 2006) it is unlikely that piRNAs are generated through a double-stranded RNA intermediate. Instead, piRNAs might be generated directly as primary transcripts. In C. elegans, this could be achieved by an RdRP. However, two observations argue against this hypothesis. First, we found no evidence for the dependence of piRNA expression on RdRPs (Table 1), although we cannot exclude redundancy. Second, piRNAs lack 5′ triphosphates, a hallmark of other RdRP derived small RNAs. We therefore favor a model in which piRNAs are derived from longer primary transcripts as is the case for miRNAs (Bartel, 2004). Interestingly, we show that Piwi is not essential for piRNA biogenesis, but that piRNA levels are dramatically reduced in piwi mutants (Figure 1E). This is analogous to C. elegans miRNAs, whose levels are reduced but are not absent in alg-1; alg-2 Argonaute mutants (Grishok et al., 2001). We therefore postulate that Piwi is not an essential component of the piRNA biogenesis pathway, but is required for piRNA accumulation or stability.

The piRNA genomic organization in Drosophila suggests that whole piRNA clusters could be transcribed as single primary transcripts and then processed into many mature piRNAs (Brennecke et al., 2007). Our finding that 21U-RNAs are piRNAs now suggests an alternative model. Many 21U-RNA loci are associated with a conserved upstream motif containing an eight nucleotide core consensus sequence CTGTTTCA ((Ruby et al., 2006) and Supplemental Data). This motif is located approximately 38 bases upstream of the base corresponding to the 5′ uridine of the 21U-RNA. Ruby et al. have postulated that this motif is part of a 21U-RNA promoter. We conclude that piRNAs might be transcribed as individual transcripts in C. elegans and perhaps other species. Alternatively, these motifs might also be involved post-transcriptionally in piRNA biogenesis. Having linked a conserved motif to piRNAs in C. elegans it will be important to search for such motifs in other species. Perhaps such a motif might be found only for a subset of “primary” piRNAs in Drosophila or vertebrates.

No Evidence for Ping-Pong in C. elegans

Recently, a compelling model for piRNA production in Drosophila named the Ping-Pong amplification loop has been put forward (Brennecke et al., 2007; Gunawardane et al., 2007). In this model Piwi/Aub is bound by a primary piRNA with a 5′ uridine. Target slicing by this complex followed by 3′ end processing results in the generation of a secondary piRNA that is antisense to the primary piRNA, overlaps by ten complementary nucleotides, and has an adenine at position 10. The loop is completed by AGO3, the third Piwi protein of Drosophila, which binds the secondary piRNA and given a suitable template, in turn generates piRNAs with 5′ uridines. This model is supported by the finding that Piwi/Aub bound piRNAs have a strong bias for 5′ uridine and AGO3 bound piRNAs a strong bias for adenine at position 10 (Brennecke et al., 2007; Gunawardane et al., 2007). To test the Ping-Pong hypothesis in C. elegans, we analyzed the overlap among 21U-RNAs and between 21U-RNAs and other C. elegans small RNAs (Figure 6A). We found few cases of overlap and no adenine peak at position 10 (Figure S6). We conclude that there is currently no evidence for a role of Ping-Pong to generate piRNAs or to couple piRNAs to endogenous siRNA pathways in C. elegans.

Piwi and piRNAs Act Upstream of Secondary siRNA Pathways in Tc3 Silencing

We suggest that C. elegans and Drosophila share primary piRNAs, but while amplification of small RNAs in Drosophila might be achieved through a Ping-Pong amplification loop, in C. elegans amplification might be achieved through secondary siRNAs and RdRPs (Figure 6B). We demonstrate that Tc3 transposon silencing, and associated siRNAs depend on Piwi and MUT-7. However, we cannot directly implicate a specific piRNA in Tc3 transposon silencing. One attractive candidate might be 21UR-139, which maps to the Tc3 transposase ORF and might act in cis. A second 21U-RNA also maps to the Tc3 locus (Batista et al., 2008). Both are sense relative to the transposase gene. We propose a speculative model to explain how sense piRNAs might be required for antisense siRNA production at the Tc3 locus. Piwi loaded with sense piRNAs might stimulate RdRP activity using Tc3 antisense transcripts as templates, in a manner analogous to siRNAs in RNAi amplification (Baulcombe, 2007). The sense transcripts would be targeted by antisense siRNAs, and act as a template for a second round of small RNA-stimulated RdRP activity (Figure 6B). Such a loop would maintain high levels of Tc3 siRNAs. In the case of the Tc3 transposase this might lead to PTGS against the mRNA. This is in agreement with the low number of siRNAs mapping to the intron as compared to the two exons of the Tc3 transposase (Figure 5B). In the case of the Tc3 TIR, this might involve chromatin-mediated TGS. If the TIR-associated siRNAs are functionally distinct from the siRNAs mapping to the Tc3 transposase ORF, these two pathways might be separable. Indeed, Batista et al. (2008) observed that prg-1 mutants lack only TIR-associated siRNAs, but not siRNAs associated with the transposase ORF (Batista et al., 2008). Together, these data suggest that PRG-1 and PRG-2 might have functionally distinct roles in Tc3 silencing. This is particularly intriguing as PRG-1, but not PRG-2 is required for piRNA accumulation (Figure 2A).

The amplification loop we propose is sequence independent, given an initiating 21U-RNA. Therefore the link between piRNA and siRNA pathways might not be restricted to Tc3. Indeed, we observe a positional bias for antisense siRNAs close to 21U-RNAs to map downstream of the 21U-RNA locus (Figure 6A). piRNA and siRNA pathways might be linked in other species too. Drosophila and vertebrates have lost secondary siRNAs and RdRPs, which are found in plants, yeast and C. elegans. However, recently endogenous siRNAs in mouse oocytes were shown to map to the same loci as piRNAs (Tam et al., 2008).

High-Throughput Sequencing and Data Analysis

We generated 5′-dependent libraries for N2 (2,719,949 reads), piwi(n4357; n4358) (764,960 reads), mut-7 (3,475,722 reads), dcr-1 (149,997 reads), and 5′-independent libraries for N2 (2,963,895 reads) and piwi(n4357; n4358) (3,017,027 reads) that were sequenced using the Illumina/Solexa platform. See Supplemental Data for additional information.

RNAi Experiments

RNAi experiments were carried out as reported previously (Fire et al., 1998; Timmons et al., 2001). Bacterial strains carrying plasmids expressing double-stranded RNA for the gene of interest were obtained from the Ahringer Laboratory genome-wide RNA library (Fraser et al., 2000; Kamath et al., 2003). All RNAi library inserts were confirmed by sequencing.

We thank Rob Shaw and Na An for strain management. We thank Andrew Hellman and Bob Horvitz for help generating prg-1 and prg-2 deletion strains. We thank Ericka Havecker, David Baulcombe, Leonie Kamminga, James Hadfield and Thomas Down for help with high-throughput sequencing. We thank Fuchou Tang for help with qRT-PCR assays. We thank Julie Ahringer, Jane Hubbard, Alla Grishok, Shohei Mitani and the Caenorhabditis Genetics Center (funded by the NIH National Center for Research Resources), University of Minnesota, Twin Cities, MN, USA for providing C. elegans strains. We thank Pedro Batista and Craig Mello for discussions on the manuscript and sharing of unpublished data. M.P.B. was supported by the Medical Research Council (MRC, UK). J.R.W. was supported by a Cancer Research UK studentship. L.D.G. was supported by an EPSRC fellowship (UK). S.T. is supported by Cancer Research UK and is a Royal Society-Wolfson Research Merit Award holder. This work was supported by a Cancer Research UK Programme Grant to E.A.M. (C13474).