Impact of extrinsic incubation temperature on natural selection during Zika virus infection of Aedes aegypti

Arthropod-borne viruses (arboviruses) require replication across a wide range of temperatures to perpetuate. While vertebrate hosts tend to maintain temperatures of approximately 37°C - 40°C, arthropods are subject to ambient temperatures which can have a daily fluctuation of > 10°C. Temperatures impact vector competence, extrinsic incubation period, and mosquito survival unimodally, with optimum occurring at some intermediate temperature. In addition, the mean and range of daily temperature fluctuations influence arbovirus perpetuation and vector competence. The impact of temperature on arbovirus genetic diversity during systemic mosquito infection, however, is poorly understood. Therefore, we determined how constant extrinsic incubation temperatures of 25°C, 28°C, 32°C, and 35°C control Zika virus (ZIKV) vector competence and population dynamics within Aedes aegypti and Aedes albopictus mosquitoes. We also examined diurnally fluctuating temperatures which more faithfully mimic field conditions in the tropics. We found that vector competence varied in a unimodal manner for constant temperatures peaking between 28°C and 32°C for both Aedes species. Transmission peaked at 10 days post-infection for Aedes aegypti and 14 days for Aedes albopictus. The effect of diurnal temperature was distinct and could not have been predicted from constant temperature-derived data. Using RNA-seq to characterize ZIKV population structure, we identified that temperature alters the selective environment in unexpected ways. During mosquito infection, constant temperatures more often elicited positive selection whereas diurnal temperatures led to strong purifying selection in both Aedes species. These findings demonstrate that temperature has multiple impacts on ZIKV biology within mosquitoes, including major effects on the selective environment within mosquitoes. Author Summary Arthropod-borne viruses (arboviruses) have emerged in recent decades due to complex factors that include increases in international travel and trade, the breakdown of public health infrastructure, land use changes, and many other factors. Climate change also has the potential to shift the geographical ranges of arthropod vectors, consequently increasing the global risk of arbovirus infection. Changing temperatures may also alter the virus-host interaction, ultimately resulting in the emergence of new viruses and virus genotypes in new areas. Therefore, we sought to characterize how temperature (both constant and fluctuating) alters the ability of Aedes aegypti and Aedes albopictus to transmit Zika virus, and how it influences virus populations within mosquitoes. We found that intermediate temperatures maximize virus transmission compared to more extreme and fluctuating temperatures. Constant temperatures increased positive selection on virus genomes, while fluctuating temperatures strengthened purifying selection. Our studies provide evidence that in addition to altering VC, temperature significantly influences the selective environment within mosquitoes.

159 To assess coding region-specific signatures of selection, we analyzed d N /d S for each viral 160 protein coding sequence independently. In both mosquito species exposed to diurnal 161 temperatures, d N /d S was much less than 0.1 only within the NS5 coding sequence (

Increased extrinsic incubation temperatures drive viral variant fixation in mosquitoes 320
We identified 8 consensus mutations (5 non-synonymous and 3 synonymous) in multiple 321 mosquitoes during the course of this study. These all were present in the input population at low 322 frequencies (0.01-0.35, Table 1) and rose in frequency during mosquito infection. Four consensus changes were found in both Aedes species, and in mosquitoes held under most 324 temperature regimes. L330V E (Fig S1A) is within domain III of the envelope protein, which 325 plays a role in host cell receptor binding and entry [41]. W98G NS1 (Fig S1B) is a surface 326 exposed aromatic to aliphatic amino acid change on the wing section of NS1, which contributes 327 to cellular membrane association [42]. M220T NS1 (Fig S1C) replaces a sulfur containing side 328 group with a hydroxylic side group and is located on the loop surface the NS1 172-352 329 homodimer [43]. G83 NS5 (Fig S1D) is a synonymous mutation found in the middle of the 330 coding sequence for the NS5 methyltransferase domain active site. Moreover most of these 331 substitutions were not particularly conservative and may be of functional significance.

332
While none of these mutations increased significantly in frequency at 25°C, mosquito exposure 333 to 35°C or diurnal temperatures cause some of these mutations to rise in frequency,

334
sometimes in a species-dependent manner, again highlighting the temperature-dependence 335 of variant frequencies in our studies.

336
To assess the fitness implications of these and other mutations that were repeatedly detected in 337 mosquitoes, we engineered individual mutations into a ZIKV infectious clone and conducted in 338 vivo competition studies at 25° and 35°C. The most notable finding from these studies is that 339 higher temperatures tended to favor frequently detected mutations, which is consistent with our 340 data on variant frequencies, consensus level changes and the strength of purifying selection.

341
Accordingly, we conclude that the variants we examined are more likely to reach high frequency 342 at higher temperature.

343
The work presented here was designed to address the hypothesis that temperature determines 344 not only the efficiency with which mosquitoes transmit arboviruses (which has been well 345 established for decades) but that it also influences virus evolutionary dynamics.   ZIKV and an average of < 8x coverage across the genome indicating little to no contamination, 478 sequencing bleed through, or index hopping (S1 Table). Only variants in the coding sequence

479
(nt position 108-10379), with100x coverage or greater and a cut off of 0.01 frequency were used 480 for analysis to account for low coverage (reads per genome position) in the 3' and 5' 481 untranslated regions (S1 Table, Fig S2).

482
Data analysis was performed using custom Python and R code integrated into the RPG using the iSNV frequency (p) at each nucleotide position (s):

499
The mean S from all sites s is used to estimate the mutant spectra complexity. Divergence was 500 calculated using F ST to estimate genetic divergence between two viral populations as described