The Tudor protein Veneno assembles the ping-pong amplification complex that produces viral piRNAs in Aedes mosquitoes

Abstract

PIWI-interacting RNAs (piRNAs) comprise a class of small RNAs best known for suppressing transposable elements in germline tissues. The vector mosquito Aedes aegypti encodes seven PIWI genes, four of which are somatically expressed. This somatic piRNA pathway generates piRNAs from viral RNA during infection with cytoplasmic RNA viruses through ping-pong amplification by the PIWI proteins Ago3 and Piwi5. Yet, additional insights into the molecular mechanisms mediating non-canonical piRNA production are lacking. TUDOR-domain containing (Tudor) proteins facilitate piRNA biogenesis in Drosophila melanogaster and other model organisms. We thus hypothesized that Tudor proteins are required for viral piRNA production and performed a knockdown screen targeting all A. aegypti Tudor genes. Knockdown of the Tudor genes AAEL012437, Vreteno, Yb, SMN and AAEL008101-RB resulted in significantly reduced viral piRNA levels, with AAEL012437-depletion having the strongest effect. This protein, which we named Veneno, associates directly with Ago3 in an sDMA-dependent manner and localizes in cytoplasmic foci reminiscent of piRNA processing granules of Drosophila. Veneno-interactome analyses reveal a network of co-factors including the orthologs of the Drosophila piRNA pathway components Vasa and Yb, which in turn interacts with Piwi5. We propose that Veneno assembles a multi-protein complex for ping-pong dependent piRNA production from viral RNA.

Figure 1.

Orthologous Tudor genes in Drosophila melanogaster and Aedes aegypti. On the left, a neighbor joining tree based on TUDOR domains from A.aegypti (red) and D. melanogaster (blue) is shown. Numbers indicate bootstrap values for 1000 iterations; only values >500 are shown. In the middle, predicted domain structures of Tudor proteins are drawn schematically, with TUDOR domains shown in black, zinc fingers in blue, putative RNA binding domains in green, domains associated with helicase activity in orange and all other domains in red. Numbers at the top indicate protein length in amino acids. On the right, protein domains other than TUDOR domains are presented, ordered from amino to carboxyl terminus, as indicated in the middle panel. As the AAEL008101 gene produces two splice variants encoding TUDOR domains of slightly different composition (PB and PC), both were included as separate entities in the multiple sequence alignment. B-box, B-box type zinc finger (Zf) domain; C2CH, C2CH-type Zf domain; C2H2, C2H2-type Zf domain; DEAD, DEAD box domain; HA2, Helicase-associated domain; Hel-C, helicase C domain; KH, K homology RNA-binding domain; LOTUS, OST-HTH/LOTUS domain; MBD, Methyl-CpG-binding domain; MYND, MYND (myeloid, Nervy, DEAF-1)-type Zf domain; PHD, PHD-type Zf domain; RING, RING-type Zf domain; RMI1_N, RecQ mediated genome instability domain; RRM, RNA recognition motif; SMN, survival motor neuron domain; SNase¸ Staphylococcal nuclease homologue domain; UBA, ubiquitin associated domain.

Figure 2.

Loss of vpiRNA production upon knockdown of several Tudor genes. Tudor genes were knocked down in Aag2 cells by dsRNA transfection after which small RNA production of (+) strand SINV and histone H4 mRNA (H4)-derived piRNAs was assessed using northern blot analyses. dsRNA targeting luciferase (dsLuc) and Piwi5 were used as negative and positive controls, respectively. U6 snRNA was used as a loading control. Aedes proteins that have a clear one-to-one ortholog with similar domain composition are named after their Drosophila orthologs. The heat map depicts relative changes in NSP4 and Capsid viral RNA abundance and Tudor/Piwi5 knockdown efficiencies as determined by RT-qPCR. All expression values were normalized to SINV-infected dsLuc control samples. Gray boxes indicate samples for which no RT-qPCR was performed.

Figure 3.

Veneno is required for efficient biogenesis of vpiRNAs. (A and B) Normalized read counts of 25–30 nt piRNAs (A) and 21 nt siRNAs (B) mapping to the SINV genome (top row), Ty3-gypsy transposons (second row), BEL-Pao transposons (third row) and histone H4 mRNA (bottom row) upon knockdown of Veneno (dsVen) and control knockdown (Firefly Luciferase, dsLuc). Virtually no siRNA-sized reads mapping to histone H4 mRNA were found (∼200 reads per library), and these are therefore not shown. (C) Nucleotide bias at the first 20 positions of the 25–30 nt small RNA reads mapping to sense strand (upper panel) and antisense strand (lower panel) of the indicated RNA substrates in dsLuc libraries (n = number of reads). (D) The probability of 5′ overlap between piRNAs from opposite strands in dsLuc and dsVen libraries for the indicated classes of piRNAs. For bar charts in A and B, read counts of three independent libraries were normalized to the amount of miRNAs present in those libraries and analyzed separately for the sense (black) and antisense (gray) strands. Bars indicate mean ± standard deviation. Two-tailed student’s t-test was used to determine statistical significance (*P < 0.05; **P < 0.01, ***P < 0.001). To generate sequence logos and 5′ overlap probability plots shown in C and D, reads of three independent libraries were combined.

Figure 4.

Veneno accumulates in cytoplasmic Ven-bodies. (A) Schematic representation of Ven transgenes used in IFA experiments. (Full: amino acid [aa] 1–785; C91: aa 91–785; C199: aa 199–785; C234: aa 234–785; C400: aa 400–785; N670: aa 1–670; N399: aa 1–399; red lines indicate sequences of low amino acid complexity, rich in asparagines [N] or glutamines [Q]). (B–H) Representative confocal images of Aag2 cells expressing transgenes drawn schematically in (A). Scale bar represents 10 μm. (I) The cytoplasm of 46–56 individual cells expressing GFP-tagged transgenes was traced as depicted and the mean and standard deviation of signal intensity was determined to calculate the coefficient of variation as a measure of signal granularity. (J) Scatter dot plot shows the GFP-signal granularity for individual cells; the red lines indicate the mean.

Figure 5.

Characterization of a multi-protein complex containing the ping-pong partners Ago3 and Piwi5. (A and B) Enrichment of 25–30 nt small RNAs mapping to the SINV sense (A) and antisense (B) strand, relative to input, in small RNA sequencing libraries of IP of the indicated PIWI proteins. (C and D) Nucleotide bias at the first 20 positions of 25–30 nt small RNA reads mapping to sense strand (upper panel) and antisense strand (lower panel) in Ago3-IP (C) and Piwi5-IP (D) libraries. (E) Protein lysates from SINV-infected (+) and uninfected (−) Aag2 cells transfected with expression plasmids encoding GFP or GFP-Ven before (Input) and after GFP-IP were analyzed for (co)purification of endogenous Ago3 and Piwi5, as well as the GFP-tagged transgene by western blot. The asterisk indicates a non-specific band. (F and G) Identification of Ven-interacting proteins in lysates from both mock- (F) and SINV-infected Aag2 cells (G) by label-free quantitative (LFQ) MS. Permutation-based FDR-corrected t-tests were used to determine proteins that are statistically enriched in the Ven-IP. The LFQ-intensity of GFP-Ven IP over a control IP using the same lysate and non-specific beads (log₂-transformed) is plotted against the −log₁₀P-value. Interactors with an enrichment of log₂ fold change > 4.3; −log₁₀ P value > 1.5 are indicated. Proteins in the top right corner represent the bait protein in green (Ven) and its interactors. Orthologs of known piRNA biogenesis factors in Drosophila melanogaster are indicated in red and interacting proteins present in both mock- and SINV-infected pulldowns are shown in bold font. Where available, interacting proteins were named according to their ortholog in D. melanogaster. In case of uncharacterized orthologous Drosophila proteins, we assigned the Vectorbase GeneID to the protein. (H) Reciprocal IPs of GFP-Ven, RFP-Vasa, V5-3xFlag-Yb, Ago3 and Piwi5 using antibodies targeting GFP, RFP, V5, Ago3 and Piwi5, respectively. Samples were probed with antibodies against GFP, RFP, Flag, Ago3, Piwi5 and α-tubulin, as indicated. (I) Lysate from Aag2 cells stably expressing GFP-Ven was fractionated on a 10–45% sucrose gradient. Protein fractions were size separated and stained using antibodies against GFP, Ago3 and Piwi5. RNA samples from those fractions were analyzed by northern blot analysis, using probes targeting abundant (+) strand ping-pong dependent piRNAs produced from the S-segment of the bunyavirus Phasi Charoen-like virus (PCLV) and U6 snRNA. All fractions contain proteinaceous material as is evidenced by silver staining (Supplementary Figure S5C); spliceosomal ribonucleoprotein complexes are enriched in fractions 11–16, as evidenced by the presence of U6 snRNA.

Figure 6.

Ven-Ago3 interaction is mediated by sDMA-recognition. (A) Schematic representation of Veneno transgenes used in Ago3 co-IP experiments. (C206: amino acid [aa] 206–785; C234: aa 234–785; C400: aa 400–785; 206–669: aa 206–669; 206–399: aa 206–399; 2Δ: C206-G463A/Y465A; Full-2Δ: G463A/Y465A; 4Δ: C206-G463A/Y465A/D483A/N486A; red lines indicate sequences of low amino acid complexity, rich in asparagines [N] or glutamines [Q]). (B) Sequence corresponding to a part of the first TUDOR domain with residues indicated that were mutated in the 2Δ and 4Δ transgenes. Residues indicated in blue bold font are predicted to be involved in sDMA recognition. (C) Lysates from Aag2 cells expressing the indicated GFP-tagged Ven transgenes were subjected to GFP-IP and subsequently analyzed for co-purification of Ago3 by western blot. α-Tubulin serves as loading control. Asterisks indicate non-specific bands. (D) Lysate from Aag2 cells stably expressing GFP-Ago3 and transiently transfected with a plasmid encoding RFP-Ven was immunoprecipitated using GFP-, RFP- and empty beads. Western blots were stained using antibodies for GFP, RFP and symmetrical dimethylated arginines (sDMA). The asterisk indicates a non-specific band. (E) Representative confocal image of Aag2 cells expressing GFP-tagged Ven-2Δ-mutant; scale bar represents 10 μm. (F) Schematic model of the identified multi-protein complex responsible for ping-pong amplification of exogenous (viral) and endogenous (transposable element, TE)-derived piRNAs. The thickness of the arrows reflects the relative contribution of the complex to processing of different RNA substrates. N, nucleus; C, cytoplasm.