Antiviral adaptive immunity and tolerance in the mosquito Aedes aegyti

Mosquitoes spread pathogenic arboviruses while themselves tolerate infection. We here characterize an immunity pathway providing long-term antiviral protection and define how this pathway discriminates between self and non-self. Mosquitoes use viral RNAs to create viral derived cDNAs (vDNAs) central to the antiviral response. vDNA molecules are acquired through a process of reverse-transcription and recombination directed by endogenous retrotransposons. These vDNAs are thought to integrate in the host genome as endogenous viral elements (EVEs). Sequencing of pre-integrated vDNA revealed that the acquisition process exquisitely distinguishes viral from host RNA, providing one layer of self-nonself discrimination. Importantly, we show EVE-derived piRNAs have antiviral activity and are loaded onto Piwi4 to inhibit virus replication. In a second layer of self-non-self discrimination, Piwi4 preferentially loads EVE-derived piRNAs, discriminating against transposon-targeting piRNAs. Our findings define a fundamental virus-specific immunity pathway in mosquitoes that uses EVEs as a potent and specific antiviral transgenerational mechanism.


89
Here we demonstrate that Piwi4 is upregulated in somatic tissues in adult female Ae. Aegypti 90 following blood meal and restricts dengue virus replication in vivo. In infected cells, Piwi4 is 91 required for the accumulation of mature v-piRNAs and binds preferentially to antisense v-92 piRNAs (corresponding to the anti-genome viral RNA) and not to anti-transposon piRNAs.    159 is involved in v-piRNA and v-siRNA maturation but appears to make a minor contribution to 160 their initial production (see Discussion).

162
Piwi4 binds to antisense v-piRNAs 163 Because maturation of piRNAs is linked to their association with PIWI proteins [22], we then 164 assessed whether Piwi4 could bind to v-piRNAs. To address this, we performed Piwi4-165 immunoprecipitation (IP), using a Piwi4-FLAG construct transiently expressed in SINV infected 166 Aag2 cells, followed by small RNA sequencing. To control for background resulting from the 167 high abundance of small RNAs in the cells and potential confounding effects due to Piwi 168 protein overexpression, we normalized small RNAs associated with Piwi4 to small RNAs 169 present in the input sample. Contrary to previous reports that did not normalize to input small 170 RNA content, we found that Piwi4 bound to v-piRNAs and that Piwi4-associated small RNAs 171 were significantly enriched for antisense SINV-derived piRNAs ( Figure 3A, antisense vs sense 172 v-piRNAs, 2 , p < 2.2e-16). In addition, Piwi4-associated v-piRNAs displayed a uracil bias at 173 . CC-BY 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted October 9, 2018. ; https://doi.org/10.1101/438911 doi: bioRxiv preprint their first position ( Figure 3B), a signature of Piwi-bound piRNAs. In contrast, Piwi4 did not 174 significantly associate with v-siRNAs (Z score for 21 nt sense and antisense v-siRNAs 175 enrichment -0.2 and -0.6 respectively) nor TE-derived piRNAs ( Figure 3C).

176
To further validate that Piwi4 preferentially bound to antisense v-piRNAs, we performed three 177 additional IP followed by small RNA sequencing experiments in SINV infected Aag2 cells: two 178 using the Piwi4-FLAG construct and one negative control IP experiment in Aag2 cells 179 transiently expressing eGFP. Deep-sequencing analysis of small RNAs in Piwi4-IP and input 180 fractions confirmed that Piwi4 preferentially binds to antisense v-piRNAs (antisense vs sense 181 v-piRNAs, 2 , p < 2.2 e-16 and p < 4.2e-9, for Experiment 1 and 2 respectively) with a U1 bias,

247
Next, we examined whether vDNA synthesis and recombination with transposon elements can 248 discriminate between self (host) and nonself (viral) RNAs. Only a small fraction of known 249 expressed Aag2 mRNAs were found to have derived sequences in the episomal DNAs reads 250 (~7.5% of known expressed mRNAs; 1093 out of 14612). Furthermore, there was no 251 correlation between mRNA and episomal DNA abundance (R = 0.0045, Figure 5G and Figure   252 5supplement E). We detected no reads derived from the top 54 most abundant mRNAs, 253 which account for 32% of all transcripts, and found no correlation between mRNA length and 254 their enrichment in episomal DNA (R = 0.0786, Figure 5H). Indeed, the number of SINV vDNA 255 reads was higher than for the most abundant host mRNA-derived episomal DNA (586 and 256 256 reads, for SINV and Aag2 transcript AAEL006357 respectively). Our results indicate that 257 reverse transcription and recombination with retrotransposons is a highly specific mechanism, 258 which preferentially converts viral RNA into vDNA. Together, these data suggest a model in 259 which vDNA is acquired by the specific incorporation of viral RNA into retrotransposon 260 . CC-BY 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted October 9, 2018. ; https://doi.org/10.1101/438911 doi: bioRxiv preprint replication complexes, followed by reverse transcription and recombination between the 261 transposon and virus genome.

264
If indeed reverse transcription and recombination are followed by integration into the host 265 genome, this process could provide mosquitoes with a mechanism of immunological memory.

266
Integration of vDNA derived from non-retroviral RNA viruses, known as endogenous viral

298
To test if EVEs are used to produce functional antiviral piRNAs, we analyzed EVE-derived 299 small RNAs produced in Aag2 cells. The vast majority of small RNAs mapping to EVEs are 24-300 30 nt in length (>95%) and are almost exclusively antisense relative to the pseudo-ORF 301 (>99%) (Fig. 6C, i). Furthermore, EVE-derived piRNAs showed a strong bias for uracil at the 302 1 st position ( Figure 6 -supplement A) and were protected from beta-elimination, suggesting 303 that they are mature piRNAs, methylated on their 3'end ( Figure 6C, i). Importantly, knockdown 304 . CC-BY 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted October 9, 2018. ; https://doi.org/10.1101/438911 doi: bioRxiv preprint of Piwi4 and Ago3, but not Ago2, resulted in a loss of EVE-derived piRNAs ( Figure 6C, ii).

305
Thus, the length, methylation, and sequence bias indicate that small RNAs derived from EVEs 306 are bona fide piRNAs and their production require Piwi4 and Ago3. Because Piwi4 specifically 307 binds to antisense v-piRNAs produced during acute infection ( Figure 3A and C), we examined 308 whether Piwi4 also preferentially associates with antisense piRNAs during persistent infection.

309
Analysis of the new, highly contiguous, Aag2 genome assembly revealed that the existence of

347
suggesting that persistently infecting CFAV genome is cleaved by these anti-sense piRNAs 348 . CC-BY 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted October 9, 2018. ; https://doi.org/10.1101/438911 doi: bioRxiv preprint (Figure 7supplement B). We thus sought to determine whether piRNAs derived from EVEs 349 are capable of mediating silencing. To this end, we inserted into the 3' UTR of a Renilla 350 luciferase (Rluc) reporter gene ~500 bp sequences corresponding to Ae. aegypti genome-351 encoded EVEs in either the sense or antisense orientation ( Figure 7A, ii inset). We selected 352 four independent sequences with distinct piRNA expression levels ( Figure 7A, i inset).

353
Transfection of Rluc-EVE reporters into Aag2 cells demonstrated that Rluc expression was 354 significantly reduced when the inserted EVE sequence was in the sense orientation as 355 compared to the control antisense orientation ( Figure 7A, ii). Furthermore, efficiency of 356 silencing correlated with piRNA expression levels ( Figure 7A, i). Thus, antisense EVE-derived 357 piRNAs can silence mRNAs containing complementary sequences.

358
We next tested whether EVE-derived piRNAs can protect from virus infection. We first inserted

436
. CC-BY 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted October 9, 2018. ; https://doi.org/10.1101/438911 doi: bioRxiv preprint In Aag2 cells, EVE-containing piRNA clusters produce more piRNAs than loci without 437 EVEs[18]. However, little is known about the regulatory mechanisms behind EVE-derived 438 piRNAs production. Preferential binding of Piwi4 to EVE-derived v-piRNAs does not extend to 439 neighboring TE-targeting piRNAs ( Figure 6E). Therefore, sorting of EVE-piRNAs to Piwi4 440 might be regulated in a target-specific manner, that is able to discriminate between TE and

480
. CC-BY 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted October 9, 2018. ; https://doi.org/10.1101/438911 doi: bioRxiv preprint 481 . CC-BY 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted October 9, 2018.

612
Chi square (X 2 ) tests were performed on 2x2 contingency tables for the different size or 613 strandness of v-piRNAs using the chisq.test in R.

614
Affinity purification (AP) 5 x 10 6 Aag2 cells were seeded in 10 cm dishes and allowed to 615 attach overnight. Cells were transfected with expression plasmids using Transit2020 (Mirius Bio) 616 using the manufacturer's instructions. 24 hours post transfection cells were washed with dPBS 617 three times, scraped off the dish in IP-buffer (pH 7.5 @ 4°C, 10 mM Tris, 2.5 mM EDTA, 250 618 . CC-BY 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted October 9, 2018. ; https://doi.org/10.1101/438911 doi: bioRxiv preprint mM NaCl, and cOmplete protease inhibitor, Roche), and centrifuged at 2000 rcf for 5 min at 619 4°C. Cell pellets were resuspended in 300 μl lysis buffer (IP-buffer + 0.5% NP-40) and 620 incubated at 4°C for 30 min and then centrifuged at @ 12000 rcf for 10 min at 4°C. The 621 supernatant was added to 50 μl of Protein A conjugated beads (Sigma) and rotated for 1 hr at 622 4°C. Lysate was adjusted to 1% NP40 (by adding 4 volumes IP-buffer) and then transferred to 623 50 μl of anti-FLAG conjugated beads (Sigma, Cat #F2426) and rotated for 6-16 hr at 4°C.

635
The additional Piwi4 pull down experiments were performed similarly except for the following.

636
After the three washes, small RNAs were first released from the beads by adding 20 mg/mL 637 Proteinase Kand 80 U/mL murine RNase Inhibitors for 1h at 55˚C, followed by TRIzol extraction.

638
Cloning of the small RNAs were performed as described below except that there was no size       CC-BY 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted October 9, 2018. ; https://doi.org/10.1101/438911 doi: bioRxiv preprint DNA Library sample prep kit, Illumina) and sequenced on an HiSeq 4000. Bioinformatic 664 analyses were carried out as described [47]. More than 50% comes from the genome (

695
Reverse transcriptase inhibitor treatments was as carried out as described in [23].

696
. CC-BY 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under

714
The EVE with the lower E-value was chosen for further analysis to predict EVEs that 715 overlapped. Several Blast hits to viral protein genes were identified as artifacts because of their 716 homology to eukaryotic genes (e.g. closteroviruses encode an Hsp70 homologue). These    . CC-BY 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted October 9, 2018. ; https://doi.org/10.1101/438911 doi: bioRxiv preprint       CC-BY 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted October 9, 2018. ; https://doi.org/10.1101/438911 doi: bioRxiv preprint plasmid prep kit (used for circDNA isolation) in extracting genetic material from mitochondria. B.