Polymorphism analyses and protein modelling inform on functional specialization of Piwi clade genes in the arboviral vector Aedes albopictus

Current knowledge of the piRNA pathway is based mainly on studies on Drosophila melanogaster where three proteins of the Piwi subclade of the Argonaute family interact with PIWI-interacting RNAs to silence transposable elements in gonadal tissues. In mosquito species that transmit epidemic arboviruses such as dengue and chikungunya viruses, Piwi clade genes underwent expansion, are also expressed in the soma and cross-talk with proteins of recognized antiviral function cannot be excluded for some Piwi proteins. These observations underscore the importance of expanding our knowledge of the piRNA pathway beyond the model organism D. melanogaster. Here we focus on the emerging arboviral vector Aedes albopictus and we couple traditional approaches of expression and adaptive evolution analyses with most current computational predictions of protein structure to study evolutionary divergence among Piwi clade proteins. Superposition of protein homology models indicate possible high structure similarity among all Piwi proteins, with high levels of amino acid conservation in the inner regions devoted to RNA binding. On the contrary, solvent-exposed surfaces showed low conservation, with several sites under positive selection. Analysis of the expression profiles of Piwi transcripts during mosquito development and following infection with dengue serotype 1 or chikungunya viruses showed a concerted elicitation of all Piwi transcripts during viral dissemination of dengue viruses while maintenance of infection relied on expression of primarily Piwi5. Opposite, establishment of persistent infection by chikungunya virus is accompanied by increased expression of all Piwi genes, particularly Piwi4 and, again, Piwi5. Overall these results are consistent with functional specialization and a general antiviral role for Piwi5. Experimental evidences of sites under positive selection in Piwi1/3, Piwi4 and Piwi6, that have complex expression profiles, provide useful knowledge to design tailored functional experiments.

with PIWI-interacting RNAs to silence transposable elements in gonadal tissues. In mosquito species 23 that transmit epidemic arboviruses such as the Dengue and Chikungunya viruses, Piwi clade genes 24 underwent expansion, are also expressed in the soma, and code for proteins that may elicit antiviral

28
Here we focus on the emerging arboviral vector Aedes albopictus and we couple traditional 29 approaches of expression and adaptive evolution analyses with most current computational 30 predictions of protein structure to study evolutionary divergence among Piwi clade proteins.

31
Superposition of protein homology models indicate high structure similarity among all Piwi proteins, 32 with high levels of amino acid conservation in the inner regions devoted to RNA binding. On the 33 contrary, solvent-exposed surfaces showed low conservation, with several sites under positive

143
Piwi genes the presence of the PAZ, MID and PIWI domains, the hallmarks of the Piwi subfamily of 144 Argonaute proteins [35]. For Ago3, Piwi1/3, Piwi2, Piwi4 and Piwi6, single transcript sequences that 145 correspond to predictions based on the identified DNA sequences were retrieved (S1 Dataset).

146
Sequencing results of the transcript from Piwi5 showed a sequence 27 bp shorter than predicted on 147 the reference genome, due to a 45bp gap followed by a 18b insertion, 110 and 333 bases after the 148 ATG starting codon, respectively. This transcript still includes the PAZ, MID and PIWI domains. The 149 presence of this transcript was further validated by northern-blot (Fig 1). For Piwi7, the transcript 150 sequence also appears shorter than predicted (Fig 1). Alignment and phylogenetic analyses, in the      [39,40]. To evaluate the selective pressures acting along these genes, we analysed the 173 polymorphism pattern in Ae. albopictus samples from wild-collected populations and from the 174 Foshan reference strain. Synonymous and non-synonymous mutations were found for each gene in 175 all populations (Fig 2), with Piwi1/3 displaying the lowest polymorphism (Table 1)    Finally, to gain insight on how variable Piwi genes are in comparison to slow-and fast-220 evolving genes of Ae. albopictus, we collected variability data of sets of genes previously identified 221 to have slow and high evolutionary rates [42]. For each population, we compared the overall level of 222 polymorphism (LoP) of the Piwi genes and of a dataset of fast-evolving genes (FGs) to that measured 223 for a dataset of slow-evolving genes (SGs) (Pischedda et al., 2019). Our results indicate that Piwi4,

224
Piwi6 and Piwi7 have LoP values comparable to those of FGs, while Ago3 and Piwi5 do not 225 significantly deviate from the LoP values of SGs. Piwi1/3 appears to be conserved (Fig 3).

Computational predictions of molecular structures 227
The functional significance of the mutations under selection, as well as that of all the shared 228 missense mutations in the PAZ and PIWI domains, was tested by computing predictions of three-229 dimensional molecular structures of the Piwi proteins using the most-recent X-ray crystallography 230 structure of Argonaute proteins as templates [43,44]. Homology modelling revealed high structural 231 conservation among the seven Piwi proteins despite sequence heterogeneity (S2 Fig.; Fig 4.A).

232
Similarly, to D. melanogaster, the highest levels of amino acid sequence conservation were 233 found in the regions that, based on homology modelling, define the inner pocket of Argonaute 234 molecular assembly where the RNA binds. Significantly lower sequence conservation was found on 235 the proteins surface (Fig 4.B). Based on our computational predictions, we could not detect 236 aminoacidic polymorphisms that would affect RNA binding or processing, suggesting that all Ae.  (Table 2.B) on the homology models showed that all variant amino acids were in 239 regions distant from the predicted RNA-binding and/or processing sites, suggesting that these 240 mutations are unlikely to alter protein folding, but could influence its stability.

261
Overall, at the adult stages, Ago3 and all Piwi genes were more expressed in females than 262 males. Expression in ovaries was higher than in the corresponding carcasses, in both sugar-and 263 blood-fed females. Differences in carcasses vs. ovaries expression were more pronounced after 264 blood-meal for Ago3, Piwi1/3 and Piwi6, while expression of Piwi2 was doubled in sugar-fed vs. 265 blood-fed ovaries.

Piwi genes expression following viral infection 267
Finally, we assessed whether the expression pattern of Piwi genes was altered upon DENV 268 and CHIKV infection (Fig 5.B). Clear differences in the expression pattern of Piwi genes was seen 269 both when comparing data from CHIKV-versus DENV-infected samples, and carcasses versus 270 ovaries. In ovaries, during CHIKV infection all Piwi genes were significantly up-regulated compared 271 to both sugar-and blood-fed mosquitoes. Four days post infection (dpi), the expression of Ago3, 272 Piwi1/3, Piwi6 and Piwi7 was between 4 to 10 folds higher than that of Piwi2, Piwi4 and Piwi5, which 273 nevertheless were upregulated with respect to ovaries of sugar-and blood-fed mosquitoes. An 274 opposite profile was seen in the carcasses, where all Piwi genes, particularly Piwi1/3 and Piwi4, 275 were down-regulated. At 4 dpi, CHIKV has already disseminated throughout the mosquito body, has 276 reached the salivary glands and is able to be transmitted. CHIKV viral titer was reduced ten folds by

342
Overall, these observations and our expression analyses support the hypothesis of an early 343 activation of the piRNA pathway following DENV infection, but a late activation after CHIKV infection.

344
Additionally, our expression analysis is consistent with a generalist antiviral role for Piwi5, which is 345 elicited both during DENV and CHIKV infection [20], but suggest a more prominent role for Piwi6 346 and Piwi1/3 or Piwi4 and Ago3 during infection with DENV and CHIKV, respectively.

Mosquitoes 349
Aedes albopictus mosquitoes of the Foshan strain were used in this study [10]. Mosquitoes

350
were reared under constant conditions, at 28 °C and 70-80% relative humidity with a 12/12 h light/dark cycle. Larvae were reared in plastic containers at a controlled density to avoid competition 352 for food. Food was provided daily in the form of fish food (Tetra Goldfish Gold Colour). Adults were 353 kept in 30 cm 3 cages and fed with cotton soaked in 0.2 g/ml sucrose as a carbohydrate source. Adult 354 females were fed with defibrinated mutton blood (Biolife Italiana) using a Hemotek blood feeding 355 apparatus. Mosquitoes from Mexico and La Reunion island were collected in 2017 as adults and 356 maintained in ethanol 70% before shipment to Italy. All samples were processed at the University of 357 Pavia.

Mosquito infections 359
Foshan mosquitoes were infected with either DENV serotype 1, genotype 1806 or with

367
Four boxes containing 60 one-week-old females were exposed to an infectious blood-meal 368 composed by 2 mL of washed rabbit red blood cells, 1 mL of viral suspension and 5 mM of ATP. The 369 titer of the blood-meal was 10 7 PFU/mL for CHIKV and 10 6.8 PFU/mL for DENV. Fully engorged 370 females were placed in cardboard boxes and fed with a 10% sucrose solution. Mosquitoes were 371 incubated at 28 °C until analysis.

372
In parallel, mosquitoes were fed with uninfected blood-meal or kept on a sugar-diet and grown 373 in the same conditions. Thirty mosquitoes were killed to be analyzed at days 4 and 14 post-infection 374 (pi) for CHIKV, and at days 4 and 21 dpi for DENV.

375
To estimate transmission, saliva was collected from individual mosquitoes as described in

381
The number of infectious particles in saliva was estimated by focus fluorescent assay on C6/36 Ae.

387
At the same time points mosquitoes that had been fed a not-infectious blood or kept on a 388 sugar diet were sampled and dissected as above.

Bioinformatic identification of Piwi genes in the Ae. albopictus genome 390
The sequences of the Ae. aegypti Piwi genes [52] were used as query to find orthologs in the 391 reference genome of the Ae. albopictus Foshan strain (AaloF1 assembly) and in the genome of the

Structure of Piwi genes 404
DNA extracted from whole mosquitoes and dissected ovaries [55] was used as template in 405 PCR amplifications to confirm the presence and the genome structure of each bioinformatically-406 identified Piwi gene. Primers were designed to amplify each exon, with particular attention to detect 407 differences between paralogous Piwi genes (S1 Table). The DreamTaq Green PCR Master Mix 408 (Thermo Scientific) was used for PCR reactions with the following parameter: 94 °C for 3 minutes,   Table). PCR reactions were performed using a High Fidelity

Northern Blot analysis 432
10µg of total RNA from a pool of 10 sugar-fed females was run in a 1% x 2% 433 agarose/formaldehyde gel (1 g agarose, 10 ml 10x MOPS buffer, 5.4 ml 37% formaldehyde, 84.6 ml 434 DEPC water). Gels were washed twice in 20x SSC for 15 minutes prior to blotting. RNA was 435 transferred to an Amersham Hybond-N+ nylon membrane (GE healthcare) using 20x SSC and

532
The funders had no role in study design, data collection and interpretation, or the decision to 534 submit the work for publication.