Transposon defense by endo-siRNAs, piRNAs and somatic pilRNAs in Drosophila: contributions of Loqs-PD and R2D2

Abstract

Transposable elements are a serious threat for genome integrity and their control via small RNA mediated silencing pathways is an ancient strategy. The fruit fly Drosophila melanogaster has two silencing activities that target transposons: endogenous siRNAs (esiRNAs or endo-siRNAs) and Piwi-interacting small RNAs (piRNAs). The biogenesis of endo-siRNAs involves the Dicer-2 co-factors Loqs-PD, which acts predominantly during processing of dsRNA by Dcr-2, and R2D2, which primarily helps to direct siRNAs into the RNA interference effector Ago2. Nonetheless, loss of either protein is not sufficient to produce a phenotype comparable with a dcr-2 mutation. We provide further deep sequencing evidence supporting the notion that R2D2 and Loqs-PD have partially overlapping function. Certain transposons display a preference for either dsRBD-protein during production or loading; this appeared to correlate neither with overall abundance, classification of the transposon or a specific site of genomic origin. The endo-siRNA biogenesis pathway in germline operates according to the same principles as the existing model for the soma, and its impairment does not significantly affect piRNAs. Expanding the analysis, we confirmed the occurrence of somatic piRNA-like RNAs (pilRNAs) that show a ping-pong signature. We detected expression of the Piwi-family protein mRNAs only barely above background, indicating that the somatic pilRNAs may arise from a small sub-population of somatic cells that express a functional piRNA pathway.

Figure 1. Length distribution of transposon-matching small RNAs identified in this study.

A) Reads of each library originating from soma and germline were mapped to the reference containing a transposon sequence collection. Transposon matching small RNAs were analyzed for their size distribution and normalized to total genome matching reads. The normalized counts were expressed as reads per thousand (RPT). B) The steady state transcript levels of 297, TNFB, roo and blood transposable elements were examined by qRT-PCR. RNA was isolated from three biological replicates of heterozygous loqs-D and r2d2 mutants separated in somatic and germline tissue, respectively. The doc, 412 and copia transposons were included for comparison. Ct-values for each transposon were normalized to the rp49 control (delta Ct). Values are mean ± SD (n = 3). C) The length distribution of transposon matching small RNAs in r2d2 and loqs-D mutants after exclusion of roo, 297, TNFB and blood transposons.

Figure 2. Analysis of strand asymmetry and 5’-nucleotide preference in deep sequencing data.

A) The thermodynamic asymmetry was calculated for transposon mapping endo-siRNAs of the indicated genotypes. We calculated the difference in free energy of duplex formation at either end of the presumed siRNA precursor for each sequence read using the nearest neighbor method , then calculated the average difference (ΔΔG0'). A positive value indicates that on average the 5’ ends of the reads are less stably base paired than the opposite ends. B) The relative frequency for each nucleotide at the 5’-end is depicted as a function of genotype and RNA treatment.

Figure 3. Changes in processing and loading of small RNAs matching to individual transposons in r2d2 and loqs-D mutants.

Transposon mapping endo-siRNA were normalized to total genome matching reads and expressed as reads per million (RPM). Each dot in the plot represents an individual transposon consensus sequence. The upper two panels compare of heterozygous r2d2 and loqs-D mutants during processing (left) and loading (right) for soma and germline. The lower panels compare homozygous r2d2 with homozygous loqs-D mutants. For example, a higher amount of endo-siRNAs in r2d2 homozygous mutant than in loqs-D homozygous mutants means that these endo-siRNAs are r2d2 independent but loqs-D dependent. They are thus situated below the diagonal, whereas transposons that require loqs-D but not r2d2 will fall above the diagonal.

Figure 4. Analysis of endo-siRNA and piRNA origin with respect to a collection of transposon master control loci in the Drosophila genome.

Fifteen transposon containing genomic regions were reported as master regulators of transposon activity . The reads from each library were separated in endo-siRNAs (21 nt) and piRNAs (24-27 nt) by their length, then mapped allowing only those reads that matched uniquely among these clusters to be retained. The counts were normalized to cluster length as well as to total genome matching reads (reads per kilobase per million mapped reads, RPKM). Cluster 2 (chromosome X; 20A), 8 (chromosome X; 20 A-B, also known as flamenco), 13 (chromosome 3LHet) and 15 (chromosome 3LHet) are shown. In the soma cluster 2 showed an excessive amount of a unique sequence at a particular location in the homozygous mutant r2d2 sample after β-elimination. This likely represents a technical artifact and was therefore omitted. We specifically label the results after exclusion of the special sequence with **.

Figure 5. Analysis of steady state level of transposons by qRT-PCR.

RNA was isolated from heterozygous and homozygous r2d2 and loqs-D mutants. DNA was digested with DNase I, the RNA was reverse transcribed and used for transposon profiling by qRT-PCR. Each transposon was normalized to the average of rp49 and gapdh controls and depicted as the fold change of homozygous to heterozygous mutant in soma and germline, respectively (p<0.05(*), p<0.009(**) student’s T-test, n = 3).

Figure 6. Orientation bias for pilRNAs in soma and piRNAs in germline.

Small RNA libraries generated with β-eliminated RNA samples were mapped to the transposon sequence collection. The RPM for sense (+) and antisense (–) transposon matching small RNAs for 23 nt to 29 nt were depicted for soma (A) and germline (B) to demonstrate the orientation bias. Note that the apparent increase of somatic pilRNAs is due to the removal of certain miRNAs and endo-siRNAs in homozygous mutants, which are either less efficiently produced and/or mis-directed into Ago1. Upon β-elimination, these RNAs no longer contribute to the sequenced pool, hence other RNA classes appear to be more abundant. We did not observe this increase if untreated libraries were analyzed.