Report

Developmentally Regulated piRNA Clusters Implicate MILI in Transposon Control

See allHide authors and affiliations

Science  04 May 2007:
Vol. 316, Issue 5825, pp. 744-747
DOI: 10.1126/science.1142612

Abstract

Nearly half of the mammalian genome is composed of repeated sequences. In Drosophila, Piwi proteins exert control over transposons. However, mammalian Piwi proteins, MIWI and MILI, partner with Piwi-interacting RNAs (piRNAs) that are depleted of repeat sequences, which raises questions about a role for mammalian Piwi's in transposon control. A search for murine small RNAs that might program Piwi proteins for transposon suppression revealed developmentally regulated piRNA loci, some of which resemble transposon master control loci of Drosophila. We also find evidence of an adaptive amplification loop in which MILI catalyzes the formation of piRNA 5′ ends. Mili mutants derepress LINE-1 (L1) and intracisternal A particle and lose DNA methylation of L1 elements, demonstrating an evolutionarily conserved role for PIWI proteins in transposon suppression.

Known piRNAs are not expressed until spermatocytes first enter mid-prophase (pachytene stage) at ∼14 days after birth (P14) (14). However, Mili expression begins in primordial germ cells at embryonic day 12.5 (5, 6), and transposons, such as L1, can be expressed in both premeiotic and meiotic germ cells (7, 8). We therefore probed a connection between Mili and transposon control by examining MILI-bound small RNAs in early-stage spermatocytes. Notably, MILI-associated RNAs could be detected at all developmental time points tested (Fig. 1 and fig. S1). Northern blotting revealed that pre-pachytene piRNAs join MILI before pachytene piRNAs become expressed at P14 (Fig. 1B). The appearance of pre-pachytene piRNAs was MILI-dependent, suggesting a requirement for this protein in either their biogenesis or stability (Fig. 1C). These results raised the possibility that MILI might be programmed by distinct piRNA populations at different stages of germ cell development.

Fig. 1.

MILI associates with distinct small RNA populations during spermatogenesis. (A) MILI-associated RNAs were analyzed from testes of 8-, 10-, and 12-day-old and adult (ad) mice. C, control immunoprecipitation. (B) Testes RNA (three top panels) or RNA from MILI immunoprecipitates (IP) (two bottom panels) from mice of indicated ages was analyzed by Northern blotting for a pre-pachytene piRNA, a pachytene piRNA, or let-7. The asterisk indicates the residual let-7 signal. (C) Northern hybridization of RNA isolated from P10 testes of WT mice and Mili-heterozygous and Mili-homozygous mutants, as indicated. The probe is as in (B). M, RNA size markers.

To characterize pre-pachytene piRNAs, we isolated MILI complexes from P10 testes and deeply sequenced their constituent small RNAs. Like pachytene populations, pre-pachytene piRNAs were quite diverse, with 84% being cloned only once. The majority of both pre-pachytene (66.8%) and pachytene (82.9%) piRNAs map to single genomic locations. However, a substantial fraction (20.1%) of pre-pachytene piRNAs had more than 10 genomic matches, as compared to 1.6% for pachytene piRNAs.

Annotation of pre-pachytene piRNAs revealed three major classes (Fig. 2A). The largest (35%) corresponded to repeats, with most matching short interspersed elements (SINEs) (49%), long interspersed elements (LINEs) (15.8%), and long terminal repeat (LTR) retrotransposons (33.8%). Although pachytene piRNAs also match repeats (17%), the majority (>80%) map uniquely in the genome, with only 1.8% mapping more than 1000 times (fig. S2). In contrast, 22% of repeat-derived pre-pachytene piRNAs map more than 1000 times and correspond closely to consensus sequences for SINE B1, LINE L1, and IAP retrotransposons (fig. S2). A second abundant class of pre-pachytene piRNAs (29%) matched genic sequences, including both exons (22%) and introns (7%). A third class matched sequences without any annotation (28%). All three major classes shared signature piRNA characteristics, including a preference for a uridine (U) at their 5′ end (>80%) (Fig. 2A).

Fig. 2.

Characteristics of pachytene and pre-pachytene piRNAs. (A) Genomic annotation of piRNA populations and the fraction of sequences with a 5′ U are indicated. rRNA, ribosomal RNA; ncRNA, noncoding RNA. (B) The top 20 pachytene and pre-pachytene piRNA clusters are displayed according to the fraction of uniquely mapped piRNAs produced (y axis) and cluster sizes (x axis). The cluster on chromosome 17, expressed at both stages, is marked with an arrow. Pre-pachytene clusters, enriched in repeat-derived piRNAs, are marked with arrowheads. (C) The density of uniquely mapped piRNAs on chromosome 7 is plotted for the plus and minus genomic strands (above and below the axis, respectively). (D) Transposable element fragments and piRNA density are plotted for a repeat-rich pre-pachytene piRNA cluster [“3” in (B)]. Elements, separated by type, are indicated as red boxes on the plus strand and as blue boxes on the minus strand. Gray shadowing indicates LINE and adjacent SINE fragments that closely match their consensus.

Pachytene piRNAs derive from relatively few extended genomic regions, with hundreds to thousands of different species encoded from a single genomic strand (1, 3, 4). Cluster analysis of pre-pachytene piRNAs yielded 909 loci, covering ∼0.2% of the mouse genome (5.3 megabases; table S1). Pachytene and pre-pachytene clusters show little overlap (Fig. 2, B and C, and table S1). Overall, pachytene clusters were larger, and each produced a greater fraction of the piRNA population than early clusters, which average 5.8 kb in size (Fig. 2B). Only 56.5% of uniquely mapped pre-pachytene piRNAs can be attributed to clusters, as compared to 95.5% in pachytene piRNA populations. Considered together, these results demonstrate that pre-pachytene and pachytene piRNAs are derived from different genomic locations, with pre-pachytene piRNAs being produced from a broader set of loci.

We were unable to intuit the functional importance of those pre-pachytene clusters that correspond solely to unannotated regions of the genome. The 28% of pre-pachytene piRNAs that correspond to protein coding genes were concentrated in 3′ untranslated regions (3′UTRs) (fig. S3) and showed a strong bias for certain loci, with 8% of the total coming from only 10 genes. However, these were invariably derived from the sense strand, implying that these piRNAs would be unable to direct posttranscriptional gene silencing. Using semiquantitative reverse transcription polymerase chain reaction (RT-PCR), we were unable to detect differences in the expression of genes corresponding to piRNA clusters by comparing wild-type (WT) to Mili-mutant animals. We do not yet understand what distinguishes genes that contribute piRNAs but do note that at least some have repetitive elements resident in their 3′UTRs (fig. S3).

Clusters that are rich in transposon sequences were among the most prominent, as judged by either their size or the number of piRNAs that they generate. Two of these were the largest prepachytene clusters (97 and 79 kb, respectively). Although uniquely mapping piRNAs were derived largely from one genomic strand, the mixed orientations of transposable elements within clusters led to the production of both sense and antisense piRNAs. As is observed in Drosophila, repeat-rich mouse piRNA clusters typically contained multiple element types, many of which comprise damaged or fragmented copies (9). In many repeat-rich clusters, the orientation of most elements was similar. For example, similarly oriented elements in the two longest clusters (Fig. 2D and table S1) resulted in the production of mainly antisense piRNAs, similar to the flamenco piRNA locus in Drosophila (9, 10).

We examined the possibility that pre-pachytene piRNAs might program MILI to repress transposon activity. Mili mutation showed substantial effects on L1 and IAP expression, with each increasing its levels by a factor of at least 5 to 10 (Fig. 3A). These studies were carried out at P10 and P14, before an overt Mili phenotype becomes apparent.

Fig. 3.

Mili regulates L1 and IAP elements. (A) Quantitative RT-PCR for IAP and L1 expression in testes from WT or Mili-null mice, as indicated. Expression was assessed at P10 and P14. Error bars indicate SD. (B) DNA was isolated from the tails or testes of Mili+/+, Mili+/–, or Mili–/– animals; digested with either a methylation-insensitive [Msp I (M)] or a methylation-sensitive [Hpa II (H)] restriction enzyme; and used in a Southern blot with a probe from the LINE-1 5′UTR. Arrowheads indicate bands arising from loss of methylation in the Mili-null animals. (C) Bisulfite sequencing of the first 150 bases of a specific L1 element was done in Mili+/– or Mili–/– animals.

Although posttranscriptional mechanisms likely contribute to silencing, CpG methylation is critical for transposon repression in mammals (1116). Both analysis with methylation-sensitive restriction enzymes and bisulfite DNA sequencing revealed substantial demethylation of L1 elements in Mili-mutant testes (Fig. 3, B and C). In the latter case, the ∼50% of L1 sequences that remain methylated in the mutant are likely derived from the somatic compartment. Considered together, our data suggest that pre-pachytene piRNAs might help to guide methylation of L1 elements, although we cannot distinguish between roles in de novo versus maintenance methylation.

In Drosophila, Piwi-mediated cleavage promotes the formation of secondary piRNAs (9, 17). This allows active transposons and piRNA clusters to participate in a feed-forward loop that both degrades transposon mRNAs and amplifies silencing (9). The presence of both sense and antisense piRNAs from mammalian transposable elements creates the potential for engagement of a similar amplification cycle. This cycle creates two tell-tale features. First, because Piwi proteins cleave targets opposite nucleotides 10 and 11 of the guide, piRNAs generated within the loop overlap their partners by precisely 10 nucleotides (9, 17). As predicted, we observed enrichment for piRNAs corresponding to L1 and IAP retrotransposons, in which the 5′ ends of sense and antisense partners are separated by precisely 10 nucleotides (Fig. 4, A and B). Second, because most piRNAs begin with a U, piRNAs produced by Piwi-mediated cleavage are enriched for adenine (A) at position 10 (9, 17). This bias was prevalent in L1- and IAP-derived piRNAs [the fraction of A at position 10 (10A) in Fig. 4, C and D]. For piRNAs to be cleavage-competent and active in the amplification cycle, they must retain a high degree of complementarity to their targets (fig. S4). Consistent with this hypothesis, piRNAs that map uniquely in the genome have a lower bias for 10A (e.g., 38.7% for non-5′U piRNAs matching LTR-containing retrotransposons) than do piRNAs with many (e.g., >1000) genomic matches (61.5%).

Fig. 4.

A Piwi-mediated piRNA amplification loop in mammals. L1 (A) and IAP (B) piRNAs were aligned to their consensus sequences allowing up to three mismatches, and distances separating 5′ ends of complementary piRNA were plotted. nt, nucleotide. Nucleotide biases were calculated for L1 (C) and IAP (D) piRNAs analyzed in (A) and (B). The fraction of A at position 10 was plotted both for piRNA classes that contain and lack a 5′ U. For each bar, the percentage of U or A residues that would be expected by random sampling is indicated by a solid line across the bar.

Our results suggest a conserved pathway through which a developmentally regulated cascade of piRNA clusters programs Piwi proteins to repress transposons in mammals. One key difference between transposon control in Drosophila and mammals is the role of cytosine methylation in maintaining stable repression. In plants, it is well established that small RNAs can guide methylation of complementary sequences (18, 19). The observations that Miwi2 (20) and Mili mutations strongly affect methylation of L1 elements and that MILI binds L1-targeted small RNAs suggest that mammals may also harbor an RNA-dependent DNA methylation pathway.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1142612/DC1

Materials and Methods

Figs. S1 to S4

Table S1

References

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article