As is the case with many pathogen-host systems, the capacity for transmission of arboviruses generally increases with increases in pathogen load in mosquitoes. For this reason, one would predict that, in the absence of opposing selective forces, evolution would favor maximal replicative fitness of arboviruses in mosquito vectors. Conventional wisdom predicts interaction of vectors and the pathogens they carry should generally be benign , implying that there should be little constraint on intrahost fitness in invertebrate hosts. Experimental evolution studies with St. Louis encephalitis virus (Flaviviridae: Flavivirus) demonstrate an inability for further adaptation to Cx. pipiens, suggesting this virus may indeed have achieved its evolutionary potential in this system , yet similar studies with WNV have demonstrated the capacity for further adaptation of this virus to Cx. pipiens. Although there have been modest adaptive and consensus-level genetic changes since its introduction to the U.S., WNV, like many arboviruses, has remained remarkably static over time [42, 45–48]. As has been shown with some systems, this could be partially attributed to differential selective pressures resulting from host cycling [16, 21, 22, 49], yet studies with mosquito-adapted WNV MP20 demonstrate that host-specific adaptations without significant fitness trade-offs in vertebrate hosts are attainable, suggesting other adaptive constraints may exist . In this study, using WNV MP20 and its parental strain (WNV WT), we show that arbovirus adaptation could be further constrained by the coupling of intrahost fitness and virulence in mosquitoes. Although a correlation between virulence and viral load, as well as strain-specific differences in vertebrate virulence for WNV and other arboviruses is well established [50–52], virulence of arboviruses in invertebrate systems has not previously been considered a dynamic trait contributing significantly to pathogen evolution. The lack of studies in this field is likely a result of the historic assumption that the invertebrate immune response is relatively generic, yet recent advances in mosquito genetics and invertebrate immunity have revealed complex interactions between vector-borne pathogens and their invertebrate hosts [10, 53]. In addition, a number of studies have documented highly variable levels of arbovirus vector competence among populations of individual mosquito species, demonstrating the specificity of arbovirus-mosquito interactions and host-virus genotype by genotype outcomes [54, 55]. This work establishes that strain-specific interactions with invertebrate hosts have the potential to be substantial forces shaping both vector and arbovirus evolution and adaptation.
Our results clearly demonstrate virulence resulting in decreased survival for MP20-exposed Cx. pipiens relative to both unexposed and WT-exposed Cx. pipiens. The fact that this decreased survival was not measured with WT-exposed mosquitoes demonstrates strain-specificity and establishes that virulence was a by-product of experimental evolution studies selecting for this high fitness strain . For exposed individuals with detectable infections, a clear cost of infection was measured. In addition, a direct correlation between viral load and virulence beyond 14 days exposure was measured, demonstrating that arbovirus intrahost replicative fitness may be coupled with virulence in an invertebrate host. As predicted with the higher fitness Cx. pipiens-adapted MP20 strain, overall viral loads, as well as infection and dissemination rates were higher. Yet, interestingly, decreased survival for MP20-exposed mosquitoes was also measured for individuals with both relatively low levels of infection as well as with the MP20-resistant group, for which there were no detectable WNV infections. Previous studies with both WNV  and Dengue virus have also demonstrated that fitness costs can be associated with resistance to infection , yet the cost for resistance measured in our study was measured only with MP20-exposed mosquitoes, demonstrating that the decreased fitness of resistant mosquitoes is not likely due simply to a coupling of WNV resistance and low fitness of mosquitoes, but instead a direct result of strain-specific exposure and, subsequently, defence against establishment of infection. These results suggest that the magnitude of invertebrate defence against establishment of arbovirus infection may be specific and that the cost for such defence may be directly correlated to strain virulence. Further studies investigating strain-specific immune response will be required to understand variation in the mechanisms and/or extent of immune gene activation that correlates with the costs of immune deployment. With RNA interference (RNAi) as the primary immune response to arbovirus infection in mosquitoes , it is possible that WNV MP20 elicits a more robust RNAi response which may be costly to the mosquito, yet this warrants future investigation. Regardless of mechanism, this result implies that chronic exposure of mosquito populations to arboviruses could have measurable effects on mosquito fitness and, subsequently, selective pressures, which are independent of vector competence. Recent studies in Ae. aegypti mosquitoes demonstrate that genes involved in RNA virus defence are subject to high levels of positive selection, suggesting a host-virus evolutionary arms race similar to what has been previously described for highly pathogenic microbes in vertebrate systems 
Previous studies with WNV have demonstrated a fecundity cost of infection for Cx. tarsalis, but not Cx. pipiens mosquitoes . Although in the current study we again did not measure differences in overall fecundity in Cx. pipiens, our results do demonstrate an association between WNV infection and substantial alterations to reproductive patterns, particularly with MP20-exposed individuals. Specifically, MP20-exposed mosquitoes maximized early egg output at a later cost. This alteration in egg production could be viewed as an adaptation to maximize reproductive output in the face of decreased fitness, particularly since both bloodmeal digestion and egg production would be extremely costly for mosquitoes whose fitness is already compromised . Indeed, increased early egg production resulted in increased hatch rates and similar overall reproductive output for WNV MP20-exposed individuals, despite both decreased survival and egg production beyond week 2 of the study. What is not considered in a controlled laboratory rearing setting is the uncertainty of successful feeding, oviposition and egg hatching, which in nature is largely dependent on fluctuating environmental conditions and likely maximized by producing multiple egg rafts over time . For this reason, the reproductive strategy of MP20-exposed mosquitoes could be less productive in a natural setting. Another consequence of these altered reproductive patterns, at least from the perspective of the pathogen, is decreased probability of vertical transmission. Since Culex mosquitoes do not generally take a bloodmeal prior to overwintering  and vertebrate hosts are not known to be capable of developing significant persistent or recrudescent infective viremia [61, 62], the capacity for vertical transmission of WNV in mosquitoes is likely critical for maintenance in temperate regions that experience significant seasonal breaks in transmission [63–65]. Since the majority of egg rafts derived from MP20-infected mosquitoes were produced following the first week of infection, the number of females vertically transmitting was lower than that of WT-infected mosquitoes. Although our overall rates of vertical transmission were low for both groups, and generally comparable to what has been previously reported [64, 66–68], what was remarkable is that all 4 WNV-infected mosquitoes that produced a 3rd or 4th egg raft vertically transmitted. Since previous studies evaluating vertical transmission have generally been done en masse, without knowledge of the reproductive history of individuals producing positive larvae, it is possible that the potential for vertical transmission of WNV and other flaviviruses in mosquitoes have been significantly underestimated.
As with fecundity, overall blood feeding rates among groups were similar, yet differences in timing of bloodmeal acquisition, which could be significant in WNV transmission, were associated with both exposure and infection status. WNV infection significantly increased the likelihood of bloodmeal acquisition in the first feeding following exposure, particularly with the WNV MP20-exposed mosquitoes. As these mosquitoes are anautogenous, bloodmeal acquisition is a requirement for egg maturation; thus these differences generally correlate to fecundity differences and, like fecundity, early feeding enhancement results in a subsequent decrease in rates in the following weeks. Enhanced blood feeding with infection, as has been shown with WNV in Cx. tarsalis as well as with Plasmodium infection of An. gambiae, could be viewed as a manipulation by the pathogen to increase transmission potential, yet since the likelihood of transmission increases with time, the decreased feeding beyond 7 days measured here, particularly with the MP20-exposed mosquitoes, is likely to instead decrease WNV transmission potential. This decreased probability of transmission could be further enhanced by the fact that, on average, MP20-infected mosquitoes consumed smaller bloodmeals, likely as a result of decreased feeding times.
The method of selection utilized in creation of the MP20 strain required only that a low proportion of surviving individuals transmit WNV to be used for subsequent passages . These selection criteria, although sufficient to select strains with superior replicative fitness and, therefore, transmissibility, could also tolerate modest levels of virulence. Here, by calculating vectorial capacity, we have shown that this level of virulence would inhibit transmission potential on the population level, such that wildtype WNV would have an advantage in terms of invasion and population spread, despite the increased intrahost fitness of WNV MP20 in Cx. pipiens. This cost of virulence could contribute to the overall dampened rates of evolution and partially explain why similar ‘adaptive’ arbovirus strains do not emerge and persist as readily as would be predicted for pathogens with the capacity to so rapidly explore sequence space. The notion that individual strain fitness and vector virulence may be coupled, as predicted by the trade-off hypothesis and demonstrated here, could fundamentally change our understanding of how vector-virus interactions work to shape the evolutionary trajectories of arboviruses and other vector-borne pathogens.
What remains unknown is how variation in natural populations of mosquitoes may affect susceptibility to virulence. It is well documented that colonization can be detrimental to population fitness and it is possible that field populations could tolerate more fit strains without significant costs [70, 71]. There are also likely to be significant temporal and generational variations in fitness which may have profound effects on vector-virus interactions and subsequent outcomes of infections. Differences in terms of the cost of resistance and blood feeding behaviour are evident when contrasting results presented here to previous evaluation of the effect of wildtype WNV exposure on Cx. pipiens. Although these experiments were separated by ~4 years and it not surprising that significant changes to the colony population could occur over that time, this demonstrates that spatial and temporal variation precludes our capacity to make broad assumptions about the outcomes of vector-virus interactions. In whole, these results demonstrate that a greater comprehension of the complexity and specificity of interactions between vectors and pathogens will be required if we are to better characterize the evolution of these systems.