Phenotypic adaptation to temperature in the mosquito vector, Aedes aegypti

Most models exploring the effects of climate change on mosquito‐borne disease ignore thermal adaptation. However, if local adaptation leads to changes in mosquito thermal responses, “one size fits all” models could fail to capture current variation between populations and future adaptive responses to changes in temperature. Here, we assess phenotypic adaptation to temperature in Aedes aegypti, the primary vector of dengue, Zika, and chikungunya viruses. First, to explore whether there is any difference in existing thermal response of mosquitoes between populations, we used a thermal knockdown assay to examine five populations of Ae. aegypti collected from climatically diverse locations in Mexico, together with a long‐standing laboratory strain. We identified significant phenotypic variation in thermal tolerance between populations. Next, to explore whether such variation can be generated by differences in temperature, we conducted an experimental passage study by establishing six replicate lines from a single field‐derived population of Ae. aegypti from Mexico, maintaining half at 27°C and the other half at 31°C. After 10 generations, we found a significant difference in mosquito performance, with the lines maintained under elevated temperatures showing greater thermal tolerance. Moreover, these differences in thermal tolerance translated to shifts in the thermal performance curves for multiple life‐history traits, leading to differences in overall fitness. Together, these novel findings provide compelling evidence that Ae. aegypti populations can and do differ in thermal response, suggesting that simplified thermal performance models might be insufficient for predicting the effects of climate on vector‐borne disease transmission.

because mosquitoes are ectothermic and the various mosquito and pathogen life-history traits that influence ultimate transmission, including larval survival, development rate, adult survival, fecundity, vector competence, biting rate, and pathogen development rate, are strongly temperature dependent (Delatte et al., 2009;Lambrechts et al., 2011;Mordecai et al., 2013Mordecai et al., , 2017Mordecai et al., , 2019;;Paaijmans et al., 2009Paaijmans et al., , 2010Paaijmans et al., , 2013;;Shapiro et al., 2017;Shocket et al., 2020;Suh et al., 2020;Tesla et al., 2018;Waite et al., 2019).One standard approach for exploring the consequences of climate warming on mosquito-borne disease is to develop and parameterize mechanistic models that integrate thermal performance curves for these traits into composite transmission metrics such as vectorial capacity or basic reproductive number (R 0 ) (Caminade et al., 2014;Cator et al., 2020;Johnson et al., 2015;Mordecai et al., 2013Mordecai et al., , 2017;;Murdock et al., 2016;Paaijmans et al., 2014;Shocket et al., 2020).An implicit assumption in many such models is that standard, fixed, thermal performance curves exist for a given mosquito species and associated pathogen/ parasite, and that these can be used to extrapolate over time and space.However, this assumption ignores the potential for adaptation to modify thermal performance curves of local populations (Couper et al., 2021;Sternberg & Thomas, 2014).Local adaption occurs when there is spatial variation in selection due to interactions with species or the environment leading to a relative increase in fitness in the local context (Sternberg & Thomas, 2014).If local adaptation leads to changes in thermal response curves, such as shifts in the optimum temperature or the critical thermal thresholds, "one size fits all" models will fail to capture current variation between populations and future adaptive responses to changes in temperature (Sternberg & Thomas, 2014), resulting in poor predictions of risk.
The evidence for local thermal adaptation in insects is mixed.
Some studies indicate the potential for populations to increase their thermal tolerance through evolutionary or plastic responses (Deutsch et al., 2008;Kingsolver et al., 2007;Larson et al., 2019;Overgaard & Sørensen, 2008), while others suggest that adaptation is limited (Bennett, McMillan, et al., 2021;Kellermann et al., 2009;Weaving et al., 2023).A common approach to estimate thermal tolerance is to examine knockdown rate (i.e., time to loss of motor function or death) following exposure to a stressful temperature.Thermal tolerance/knockdown measures have been shown to be a relevant proxy for fitness under field conditions (Jørgensen et al., 2019) and to be highly predictive of current distribution and likely future range (Overgaard et al., 2014;Sgrò et al., 2010).Populations can be compared using "common garden" experiments, whereby organisms collected across some types of environmental gradient are transferred to a shared common environment and traits compared (Lovell et al., 2022).A complementary approach is to use experimental passage studies (also referred to as experimental selection or evolution studies) to capture both the direct and indirect responses of organisms to temperature (Huey & Kingsolver, 1993).Previous experimental evolution studies in Drosophila indicate the potential to select for increases in heat tolerance (Bubliy & Loeschcke, 2005;Gilchrist & Huey, 1999;Mccoll et al., 1996).We are unaware of any such studies conducted on key mosquito vectors.
Here, we explore thermal adaptation in Aedes aegypti, the primary mosquito vector of arboviruses such as dengue, Zika, yellow fever, and chikungunya, through both comparative thermal tolerance and experimental passage studies.Our novel results provide strong evidence for existing variation in thermal performance between mosquito populations and the potential for phenotypic shifts in response to temperature change.Thus, our study challenges the prevailing application of "one size fits all" thermal performance models to explore the current and future impact of climate on mosquito-borne diseases.

| Mosquito collection
Aedes aegypti mosquitoes were collected from the field using ovitraps in five different locations in Mexico (Cabo San Lucas, Acapulco, Monterrey, Ciudad Juárez, and Jojutla) and compared to a standard laboratory population (Rockefeller strain) maintained at Penn State University.The field locations were chosen to capture a gradient of the landscape and climate driven primarily by variation in altitude (Figures S1 and S2; Table S1).Populations were founded with sufficient viable eggs collected from multiple ovitraps across the cities to yield at least 100 adult mosquitoes in the F1 laboratory generation.
Mosquitoes were reared for a generation (F2) in standard laboratory conditions (27°C, 80% humidity, 12:12 h photoperiod, 0.60 mg of bovine liver powder per 600 larvae, and ad libitum access to 10% sugar solution for the adults) prior to experimentation to remove the influence of any maternal effects.Two populations, Jojutla and Juárez, were reared for an additional generation (F3) to ensure a large enough population for subsequent experiments.

| Knockdown assays to estimate thermal tolerance
We followed methods developed recently by Ware-Gilmore et al. (2021) to examine the thermal tolerance of adult Ae. aegypti mosquitoes.These methods were adapted from numerous studies in Drosophila (Hoffmann et al., 2002;Overgaard et al., 2014;Sgrò et al., 2010) and were shown to be sufficiently sensitive to demonstrate the effects of infection with either dengue virus or the bacterial endosymbiont, Wolbachia, on thermal sensitivity (Ware-Gilmore et al., 2021).In brief, 3-to 4-day-old female mosquitoes were placed into individual sealed 40 mL glass vials.The vials were submerged into a tank filled with water at a regulated temperature of 41°C.Individuals were allowed 2 min to acclimate, after which they were monitored and the time to knockdown (immobility or death) was recorded.We monitored mosquitoes until all were knocked down.In the common garden experiment, we conducted six replicate runs of 10 mosquitoes giving a total of 60 mosquitoes per population.The passage experiment had 18 total replicates (6 per independent passaged line) of 10 mosquitoes for each temperature treatment (Dennington et al., 2023).

| Experimental passage
We used the F1 population of Ae. aegypti mosquitoes collected from Monterrey, Mexico, to establish six replicate lines, half of which were maintained at a standard insectary temperature of 27°C (80% humidity, 12:12 h photoperiod), which also approximates the overall mean temperature in Monterrey summer months, and the other half were maintained at an elevated temperature of 31°C (80% humidity, 12:12 h photoperiod) (Figures S1 and S2; Table S1).The elevated temperature of 31°C represents an increase of 4°C as might be expected under future climate warming.However, neither temperature simulates realistic environmental variation in the natural home environment and so both treatments were under some level of artificial selection to lab conditions.
Each replicate line was initiated with 600 first-instar larvae.
Larvae were added into 5.7 L containers containing 3 L of deionized water and 0.60 mg of bovine liver powder (MP Biomedicals) and placed in controlled temperature incubators (three replicate containers at 27°C and three containers at 31°C).Every other day we added 0.60 mg of bovine liver powder to each container until larvae began to pupate when we scaled the food to the number of remaining larvae.We removed pupae and placed them in a small cup (30 mL) with water from their original environment to allow for eclosion.Cups containing pupae were added to a large cage with ad libitum access to 10% sugar solution made with dextrose anhydrous and deionized water to sustain the adults as they emerged.We counted total pupae per container, along with the number of pupae that eclosed successfully.When the adult mosquitoes were of reproductive age (3-5 days after eclosion), any dead adult mosquitoes were counted.These measures were used to get an accurate count of surviving adult mosquitoes in each cage.To ensure balanced selection between lines and account for potential effects of genetic bottlenecks or drift that could result from different population sizes, adult mosquitoes were culled (3-5 days after eclosion) before blood feeding so that each line had the same number of mosquitoes.The lines were culled in pairs, for example, replicate one at 27°C was paired with replicate one in 31°C, for all 10 generations to ensure independence between replicates.Mosquitoes were culled with a 50:50 sex ratio to maintain possible differences in selection on sex.Once the adult cages were established, mosquitoes were fed a human blood meal every 4 days for 16 days using a standard membrane feeder.Eggs were collected every day and maintained on dry filter paper to prevent hatching.At the end of the 16-day egg laying period, the eggs were transferred to larval containers to initiate hatching, thus maintaining a uniform age structure.Mosquitoes were reared through to adult as described.
We followed this protocol for 10 generations (Figure S5).
At the end of the experimental period, we measured thermal tolerance for the six lines using the knockdown methods described above.In addition, to extend beyond this proxy variable and fully explore the effects of the passage treatments, egg-to-adult survival, mosquito development rate, mean adult survival, and fecundity were measured in mosquitoes reared in environmentally controlled incubators at 13,17,21,25,27,29,31,33,35, 37°C, each ±0.2°C and 80% ± 10% relative humidity.Eggs from the six passaged lines were hatched at 27°C.After 24 h, 200 first-instar larvae were put into 1.89 L containers with 1 L of deionized water and 0.20 mg of larvae bovine liver powder (MP Biomedicals) and placed in the respective incubator.We fed larvae 0.20 mg of liver powder every other day until pupation.Once larvae began to pupate, we scaled their food to the number of remaining larvae.We removed and counted living and dead pupae the day of pupation and placed them in a small cup (30 mL) with water from their original environment to allow for eclosion.Cups containing pupae were added to a small cage (17.5 cm 3 ) with ad libitum access to 10% sugar solution made with dextrose anhydrous and deionized water.We then counted the number of adults that eclosed every day.After 95% of females emerged, we blood fed females who were then 3-5 days old.We used blood from de-identified human donors (BioIVT, Corp.), so IRB approval and human subjects' approval were not needed.Immediately after blood feeding, we counted the total number of blood-fed females and placed up to 10 individual females into separate containers (50-mL polypropylene centrifuge tubes) lined with filter paper that contained 7-mL deionized water to measure individual fecundity.We recorded the day that females in individual containers first laid eggs and let them lay eggs for three total days, after which we removed them from their containers.We extracted the water from the containers to let the filter paper dry in their respective incubators, and then, we counted the eggs.We counted and determined the sex of the number of adults that died every day.We censored this experiment 4 weeks after the first egg lay at each temperature (Figure S5) (Dennington et al., 2023).

| Analysis
Kaplan-Meir survival curves, stratified by population, were used to visualize differences in knockdown when exposed to lethal temperatures, and a log-rank test was used to indicate whether there are differences between populations.In addition, Cox proportional hazard models were used to assess the differences in knockdown rates between experimentally passaged lines of Ae. aegypti.To assess the variation between assay replicates, we used log-rank analysis.
To analyze the thermal performance curves for fecundity and juvenile (egg to adult) survival, we fit symmetric thermal response functions to our data using a Bayesian approach following methods described in Johnson et al. (2015).Specifically, we assume that the mean fecundity and mean egg-to-adult survival can be described by a quadratic equation: While the mosquito development rate could be described by a Brière function: where T 0 is the thermal minimum (below which the function is zero), T m is the thermal maximum, and c is a positive constant that controls the curvature of the function (and thus the height of the curve for a given value of T 0 and T m ).We assume that observed values for each life-history trait must be non-negative, and are thus modeled with a truncated normal likelihood such that where τ is the precision (τ = 1/σ 2 ).We chose priors for T 0 and T m to restrict each trait to its biologically realistic range, specifically assuming that temperatures below 0°C and above 45°C were fatal (Johnson et al., 2015;Mordecai et al., 2013Mordecai et al., , 2017;;Shocket et al., 2020).For other parameters, we chose relatively uninformative priors (Data S1).
Egg to adult survival is a probability and must be constrained to lie between 0 and 1.Thus, we modeled these data as being binomially distributed where n is the number of total observations of which Y were successes (or survival) and the probability of a success at a particular temp, p, depends on temperature [f(T)], specifically being a piecewise quadratic, as above, but being constrained to be less than or equal to 1.
For the fecundity model, we first fit the function to all the fecundity data from both the arms (27 and 31°C) of the evolution experiment together, then with the two arms separately.The latter allowed all parameters, including the variance parameter of the truncated normal to be different between the two levels.Given that this was a relatively short experiment, it was also possible that only the limits (T 0 and T m ) would be different between the two experimental settings.To allow this, we allowed the mean function to vary between the H and L settings to have different limits while all other parameters are shared between them (i.e., we fit these as a hierarchical model, see Data S1).
We fit the models using Markov Chain Monte Carlo (MCMC) sampling as implemented in JAGS, using the R package R2jags (R Development Core Team, 2014; Plummer, 2014;Su & Yajima, 2012).
For each thermal response, we ran five MCMC chains with a 5000step burn-in and saved the subsequent 10,000 steps.We thinned the posterior samples by saving every eighth sample, for a total of 3125 posterior samples of parameters.We used these to produce samples from the posterior distribution of the fecundity function versus temperature.We summarized the relationship between temperature and fecundity by calculating the mean and 95% highest posterior density (HPD) interval for the function across temperatures.The HPD interval is a type of credible interval that includes the smallest continuous range containing 95% of the probability, as implemented in the coda package (Plummer et al., 2006).
Survival data for adult mosquitoes in each treatment were observed through a censoring time T cut ; that is, for each temperature, survival observations were censored 28 days after the first egg lay.
Thus, in each experiment, the proportion of mosquitoes were observed until they died, but a portion of them were still alive at the end of the experiment, and so the survival times were right-censored.We chose to model these data with a variation of a Bayesian Weibull survival model.We assume that the observed lifetimes, y i , at some particular temperature are drawn from a Weibull distribution with rate parameter and shape parameter k, with pdf, and corresponding CDF notated as f(y; , k).The median, m, of the Weibull is defined as which can be rearranged to solve for .For interpretability, we assume that the median lifetime decays exponentially with temperature across the experimental temperature range studied, that is, so that the median thermal performance curve, f(T), is where T is the temperature, and a and b are regression parameters to be estimated.
The likelihood is comprised of two components, one describing the individuals that were observed to die within the study period ( i = 1), and those who are censored (i.e., that survive the study period, i = 0).Thus, the likelihood is given by where the rate parameter is defined in terms of the median thermal performance curve, = (ln2) 1∕k ∕ g(T).We chose prior distributions to enforce positivity of the shape parameter and of the thermal performance parameters, but that are relatively uninformative, specifically, We used a temperature-dependent model for population growth r T i as previously described in Amarasekare and Savage (2012).
where, MDR is mosquito development rate, E is eggs per female per first gonotrophic cycle, is adult mortality rate, and j is juvenile mortality rate (Figure S11; Table S9).W(x) is the upper branch of the Lambert function.We combined the thermal performance curves for each trait to calculate temperature-dependent fitness, r T i , creating a unimodal curve.We followed Amarasekare and Savage (2012) and , truncated the results at r T i = 0 rather than allowing them to go negative.
We estimated the posterior distribution of r T i and used it to calculate the key temperature values for relative temperature-dependent population fitness.We used the mean and 95% credible intervals for the critical thermal minimum, maximum, and optimum temperatures for population fitness.

| RE SULTS
First, we examined thermal tolerance of five populations of Ae. aegypti collected from climatically diverse locations in Mexico that varied in temperature conditions due to differences in altitude and latitude (Figures S1 and S2; Table S1), together with a long-standing lab strain.Because we conducted multiple replicate assays, we examined inter-assay variation for each population using log-rank analysis (Table S2a).There was no significant difference between populations aside from Acapulco and Monterrey, which had significant inter-assay variation.However, the inter-assay variation was less than differences between populations, and the individual replicate assays for these populations were significantly different to the other populations (log-rank test of Acapulco replicates compared to other populations, χ 2 = 15.6, p = .02,df = 6; and Monterrey replicates compared to other populations, χ 2 = 15, p = .02,df = 6).Given these results, we combined replicates to form a single knockdown curve for each population to assess overall differences in thermal tolerance (Figure 1; models for individual lines and replicates can be found in Figures S3 and S4).There were significant differences in knockdown rates of the six Ae.aegypti populations (log-rank, χ 2 = 71.9,p < .0001,df = 5) (Figure 1).All five field-derived populations exhibited greater median knockdown times than the long-standing lab colony (Table S3a).This might indicate increased thermal tolerance of mosquitoes exposed to naturally fluctuating environmental conditions in which daily or seasonal maximum temperatures all exceed 30°C (Table S1), compared to the lab population which has been maintained at constant 27°C for many generations.Log-rank analysis also revealed multiple significant pairwise differences between the field populations (Table S3b), potentially consistent with local adaptation.However, with only five populations, it is not possible to conduct correlation-based analysis to infer causal effects of environmental variables.
To complement the comparison between spatially discrete populations and determine whether differences in temperature can drive differences in thermal tolerance, we next conducted an experimental passage study by establishing six replicate lines from one of the field populations of Ae. aegypti (Monterrey, Mexico) and maintaining half at a standard insectary rearing temperature close to the average summer temperature of the home environment (27°C), and half at an increased temperature to represent a warmer environment as might be expected through climate change (31°C).After 10 generations, we examined knockdown rate for the replicate lines (Figure 2).
Log-rank analysis revealed no significant inter-assay variation for the 27 or 31°C lines (Table S2b), so we pooled replicates to allow overall comparison between passage treatments.Lines maintained under selection in the lab exhibited an increase in thermal tolerance compared to the original host population that was tested after just one generation in the lab (models for individual lines and separated replicates can be found in Figures S6 and S7).Lab conditions differ to the original field conditions in multiple ways and so the extent to F I G U R E 1 Thermal tolerance of field-derived Aedes aegypti populations sampled from five locations in Mexico (Jojutla, Ciudad Juárez, Acapulco, Monterrey, and Cabo San Lucas, in order of decreasing thermal tolerance), together with a long-standing laboratory-adapted population.Tolerance was measured using a thermal knockdown assay in which adult mosquitoes were exposed to 41°C and the time until knockdown recorded.The p-value is based on a log-rank test to determine differences between knockdown curves.The colored lines represent the proportion of the mosquitoes resisting knockdown for 60 individual mosquitoes per population, with the shaded areas representing 95% confidence intervals.The dotted line indicates when 50% of individuals were knocked down.Individuals were monitored until all were knocked down.Data for each population individually are presented in Figures S3 and S4.
which this is a response to temperature versus diverse other selective pressures associated with lab adaptation is unclear.Importantly, however, our passaged lines also differed to one another, with those maintained at 27°C for 10 generations showing higher knockdown rate and hence less thermal tolerance (Cox PH, z = 2.33, p < .0001), than those maintained at 31°C (Cox PH, z = 0.43, p < .0001).These data confirm that differences in thermal tolerance between populations can be generated by differences in environmental temperature.
Moreover, differences between selected lines were not limited to the proxy fitness measure of thermal tolerance but were also apparent in multiple life-history traits.Specifically, there was a difference in the estimated thermal performance curves for egg to adult survival probability, fecundity, and median adult survival between the 27 and 31°C lines, indicated by both the empirical data and the Bayesian model fits (Figure 3a-d; replicate model fits can be found in Figures S8-S11).For egg to adult survival, we see a shift toward warmer temperatures of 0.2°C for the optimum and 1.7°C for the critical minimum temperature (CTmin) for lines passaged at 31°C, but a reduction of 1.2°C in the critical maximum temperature (CTmax).
Together, these results indicate an overall reduction in thermal performance breadth (the operative range) relative to lines passaged at 27°C.Adult survival also showed a small shift, with a cross-over in thermal performance curves indicating that the mosquito lines passaged at 27°C survived better at cooler temperatures than the 31°C lines, whereas those passaged at 31°C survived better at higher temperatures than the 27°C lines.For each of these traits (i.e., egg to adult survival probability, mosquito development rate, and adult survival), the deviance information criterion indicates that the best fit models are those that separate by selection treatment, indicating the differences between lines are significant (Tables S4-S8).For fecundity, there was a shift toward warmer temperatures of 0.8°C for the optimum, 0.9°C for CTmin, and 0.5°C for the CTmax lines passaged at 31°C compared with 27°C.As indicated by the deviance information criterion (Table S4), the fecundity model presented here and the model containing all data points without separating by treatment are equally as good at representing the data, but the posterior distributions show a decreased mean in lines selected at 27°C compared to those passaged at 31°C.Given the only difference between these lines was the passage temperature, these data from multiple traits confirm the potential for rapid phenotypic adaptation in response to local thermal regime.The exception was mosquito development rate, which showed no substantive differences between lines passaged at 27°C and those passaged at 31°C.
When we compile these life-history traits into an overall fitness model, we show that adaptation to a warmer temperature has an impact on population fitness (Figure 4; Table S9).Specifically, we show a horizontal shift in the curves so that those lines passaged at 31°C have an increase of 1.02°C for CTmin, 0.66°C for CTmax, and 0.24°C for the optimum (Topt), compared with lines passaged at 27°C.For each life-history trait included in the composite fitness metric, the deviance information criterion indicates that the best fit models are those that separate by passage temperature, indicating significance for the population growth fitness models.

| DISCUSS ION
In this study, we used lab experiments to shed light on how mos- F I G U R E 3 Life-history traits measured for Aedes aegypti mosquitoes reared at 13, 17, 21, 25, 27, 29, 31, 33, and 35°C.The data in blue are from the mosquitoes passaged at 27°C, while those in red are the mosquitoes passaged at 31°C.The data points for figures on the left show the means for each of the individual replicate mosquito lines, while the bars show the overall standard error of the grand means (data points for the passaged lines in the left-hand panels are offset by 0.5°C from one another to allow visual comparison).Thermal performance curves for life-history traits for the passaged mosquito lines fit using Bayesian inferences with weakly informative priors are shown on the right (Tables S5-S8).The blue line indicates mean model fits for the mosquitoes passaged at 27°C while the red line indicates mean model fits for the mosquitoes passaged at 31°C.The shaded areas indicate 95% credible intervals.Each mosquito line was tested between 13 and 37°C, but individuals did not survive and reproduce at 13 and 37°C.aegypti populations from Mexico.Then, we subjected one of these field-derived populations to experimental passage under different temperature conditions and found evidence for shifts in thermal performance after only 10 generations.Together, these novel results provide evidence that populations of Ae. aegypti mosquitoes can and do exhibit adaptation to environmental conditions.Thus, our study challenges the prevailing application of "one size fits all" thermal performance models to explore the current and future impact of climate on mosquito-borne diseases.
Local thermal adaptation has been extensively researched in numerous taxa but remains poorly characterized in mosquito vectors (Sternberg & Thomas, 2014).Egizi et al. (2015) argue that local adaptation to climate (and climate change) should be expected in mosquito vectors given that rapid adaptive evolution has been demonstrated in response to other selective agents such as day length (Urbanski et al., 2012) and insecticide applications (Egizi et al., 2015;Gatton et al., 2013;Ranson & Lissenden, 2016).Genetic signatures of thermal adaptation have been found in populations of Aedes japonicus japonicus in Hawaii and Virginia (Egizi et al., 2015) and in Ae. aegypti in Panama (Bennett, Sunday, et al., 2021).Also, population genetic studies on the key African malaria mosquito, Anopheles gambiae, have identified chromosomal inversion polymorphisms associated with environmental stress tolerance, including larval thermal tolerance (Ayala et al., 2014;Rocca et al., 2009).Other research has demonstrated significant variation in critical maximum respiratory temperatures between three spatially distinct populations of Culex tarsalis in California, with a positive correlation to mean daily maximum temperatures at each site (Reisen, 1995).Yet other studies on C. tarsalis (Reisen, 1995) and C. pipiens (Ruybal et al., 2016) in California have also identified significant between-population variation in the response of life-history traits to temperature but found no correlation with local temperatures (Bennett, McMillan, et al., 2021;Reisen, 1995;Ruybal et al., 2016).In the context of this limited body of work, our current study provides significant new empirical evidence for one of the most globally important mosquito vector species.
Our knockdown assays revealed differences in thermal tolerance between Ae. aegypti populations collected from the field.The patterns are consistent with numerous studies in Drosophila that show variation in quantitative traits such as heat and cold tolerance, body size, and development time along environmental gradients such as latitudinal clines (Hoffmann & Weeks, 2006;Sgrò et al., 2010).The effects we observe are unlikely to result from within-generation acclimation responses, since populations were reared for one to two generations under standardized conditions after collection from the field before evaluation (Weaving et al., 2023).However, because we examined a relatively small number of mosquito populations, we have limited power to correlate observed differences in temperature tolerance with specific features of their home environments and in principle, populations from the field could differ for multiple reasons.
To determine more clearly whether ambient temperature is likely to play a role in shaping thermal tolerance, we conducted an experimental passage study on one of the field-derived populations.This approach revealed that thermal tolerance is not a static trait but can shift over a few generations in response to ambient temperature.
This study is the first of its type, as far as we are aware, to be conducted with a major mosquito vector.The results are again consistent with studies on Drosophila along with corals and Daphnia, which indicate an increase in thermal tolerance in a relatively few generations in response to warming (Chakravarti et al., 2017;van Doorslaer et al., 2010).The mechanisms involved are unclear but are potentially linked to changes in factors associated with heat shock protein expression in response to thermal stress (Telonis-Scott et al., 2021).
The experimental passage study further showed that differences in ambient temperature can lead to shifts not only in thermal tolerance but also in thermal performance of a range fundamental life-history traits such as fecundity, juvenile survivorship, and adult survival.These life-history effects add support to the utility of thermal tolerance as a proxy fitness measure (Jørgensen et al., 2019;Overgaard et al., 2014;Sgrò et al., 2010).To standardize the possible effects of genetic drift or bottle necks, we maintained equivalent population sizes between paired replicate lines through selective culling of the 27°C treatment at each generation (see Section 2).
The average number of mosquitoes culled declined across generations, indicating ongoing adaptation to the laboratory environment and their respective thermal regimes (Figure S12).Nonetheless, we did not quantify genetic changes so we cannot confirm whether the effects we observe are a consequence of adaptive phenotypic plasticity, genetic change, or a combination of both.Phenotypic plasticity and genetic adaptation are both potentially important in enabling isolated populations to deal with environmental change (Arnold et al., 2019;Grant et al., 2017).Whatever the mechanism, the key result is that after 10 generations of passage, lines derived from a common background differed in their thermal performance, confirming potential for a dynamic response to the environment.Local adaptation to environmental temperature has been shown to affect multiple life-history traits in Drosophila species, including fecundity, immunity, egg viability, development rate, cuticle color, and other stress-related traits, even in populations with high gene flow (Lazzaro et al., 2008;Sarup et al., 2009;Sillero et al., 2014;Wittkopp et al., 2011).Our passage experiment revealed differences in rate and/or magnitude of adaptive response between traits.The fecundity thermal performance curves remained similar in terms of height and breadth but showed an overall separation between treatments consistent with passage temperature (i.e., a horizontal shift in the thermal performance curve in the 31°C regime resulting in higher CTmin, Topt, and CTmax temperatures than the 27°C regime).Similarly, the median adult survival curves showed a comparative increase in survival at lower temperatures in the 27°C passaged lines, and a comparative increase in survival at higher temperatures in the 31°C lines (note that within the temperature range studied, we found no evidence for a turnover in our measure of adult survival, but based on other studies, it is likely survival would be unimodal across a broader range (Mordecai et al., 2019)).The curves for larval survival had a narrower operative range and lower peak values in the high temperature treatment, suggesting a possible reduction in overall trait performance relative to the lower temperature treatment.
In contrast, thermal performance curves for development rate remained largely unchanged between passage treatments.Differences in response of individual traits to changes in environmental factors have been shown in many studies (Huang et al., 2007;Kellermann et al., 2015;Muñoz-Valencia et al., 2016;Schou et al., 2014;Simões et al., 2020).This finding cautions against the use of individual traits to infer the effects of climate or climate change, since variable correlation in thermal response between traits could potentially mediate population-level outcomes, as overall fitness depends on the composite impact on all traits.
When we integrate individual life-history traits into a fitness model, we show that there is an overall population response across temperature, with mosquitoes maintained in the higher temperature environment showing a horizontal shift in thermal performance resulting in an increase in relative fitness at warmer temperatures.The shift between fitness curves is smaller than the 4°C difference in ambient temperature, though it should be noted that at the thermal limits, even modest changes in performance curves still yield large relative differences in fitness (e.g., at the CTmin for the lines passaged at 31°C where the fitness is zero, the lines passaged at 27°C still achieve approximately 50% of their maximum fitness).The modest extent of the shift could be because adaption is still ongoing after 10 generations, because differences in stable mean temperatures around the optimum impose limited selection pressure on overall thermal performance, or because there are limits to adaptive potential within our founding population.The slightly reduced operative range for the lines passaged at 31°C compared with those passaged at 27°C, together with a small reduction in maximum fitness, follows previous observations that warm-adapted lineages tend to exhibit narrower thermal performance curves than their colder adapted counterparts (Alruiz et al., 2023;Deutsch et al., 2008;Gaitán-Espitia et al., 2013) and are indicative of thermodynamic constraints on the extent of adaptive potential (Verhulst et al., 2020).Ultimately, there will be some limits to adaptation and there are examples of both alpine and tropical species of Drosophila that are predicted to be unable to adapt to warming temperatures, leading to expectations of range loss (Hoffmann et al., 2003;Kellermann et al., 2009;Kinzner et al., 2019).
Research on thermal adaptation of mosquitoes has a basic scientific value, but one of the additional motivations for studying Ae. aegypti is to understand the implications for transmission of key arboviral diseases.The current study did not include infected mosquitoes so an important area for future research is to extend studies to examine the effects of local adaptation across all traits, including vector competence and parasite development rate to determine consequences for disease transmission.Moreover, even small changes in traits that have non-linear effects on transmission, such as biting rate and the interaction of adult longevity and pathogen development rate (Cator et al., 2020), could result in substantial effects on local force of infection.If there are significant trade-offs between traits, or certain traits are essentially fixed, the net effects of local thermal adaptation on transmission could be small relative to the overarching ecological influence of ambient temperature and other environmental variables.The current study provides important motivation to address this knowledge gap.
A further challenge in exploring associations with environmental temperature and transmission is that the microclimates experienced by mosquitoes can differ substantially from broader macroclimate measurements, depending on where the adult mosquitoes rest and local habitat features such as housing structure, land use, and vegetation cover (Murdock et al., 2017;Paaijmans & Thomas, 2011;Wimberly et al., 2020).Additionally, while there is minimal research on mosquito thermoregulatory behavior, particularly in field settings, there is some evidence to suggest that adult mosquitoes can limit exposure to thermal extremes through avoidance behaviors (Blanford et al., 2009;Kirby & Lindsay, 2004;Suh et al., 2020;Verhulst et al., 2020).The extent to which such behaviors might mediate the effects of changes in environmental temperature is unclear (Sillero et al., 2014).Other considerations, such as the influence of larval rearing temperatures on mosquito population thermal tolerance (Oliveira et al., 2021;Overgaard & Sørensen, 2008), potential interactions between larval resource availability and temperature on population fitness (Wimberly et al., 2020), the influence of microbial infection (Ware-Gilmore et al., 2021), or the cumulative, as opposed to instantaneous, effects of heat stress (Rezende et al., 2020), add further complexity.Together, these factors highlight the need for further research to improve overall understanding of mosquito thermal ecology in the field, including studies comparing the adaptive potential of mosquitoes across global populations, along with the genetic components that are driving differences in thermal tolerance between local populations (Rolandi et al., 2018;Sørensen et al., 2001).
In summary, despite substantial public health relevance, current understanding of the effects of environmental temperature on mosquito ecology and evolution remains extremely limited.The majority of temperature-based vector-borne transmission models ignore local thermal adaption, assuming instead that essentially every population of a specific vector-pathogen pairing will respond predictably based on a standardized set of thermal performance relationships.Our common garden experiments demonstrate standing phenotypic variation in thermal sensitivity between populations of Ae. aegypti mosquitoes.Our experimental passage data confirm the potential for rapid change in thermal sensitivity together with a suite of other ecologically relevant traits, in response to environmental temperature.As such, the current study provides the most compelling evidence to date for the potential for thermal adaptation in a key disease vector.If commonplace, such effects will tend to increase variation in the expected effect of climate and climate warming on mosquitoes and associated vector-borne diseases.
quito vectors might adapt to future climate change.First, we documented existing variation in thermal tolerance across five Ae.F I G U R E 2 Thermal tolerance of Aedes aegypti populations in the experimental evolution study.Tolerance was measured using a thermal knockdown assay in which adult mosquitoes were exposed to 41°C and the time until knockdown recorded.The colored lines represent the proportion of individuals resisting knockdown for 60 individual mosquitoes per replicate (with three replicates in each, 180 total individuals per treatment), with the shaded areas representing 95% confidence intervals.The dotted line indicates when 50% of individuals were knocked down.The p-value is based on a log-rank test to determine differences between knockdown curves.The gray line represents the baseline F1 population from the field prior to thermal passage in the laboratory (60 individual mosquitoes).The red line represents the replicate mosquito lines maintained for 10 generations at 31°C, while the blue represents the replicate mosquito lines maintained for 10 generations at 27°C.Results for individual replicate lines are presented in Figures S6 and S7.
(a) Egg to adult survival measured as the probability of individuals surviving to the adult stage.(b) Mosquito development rate is the inverse of the amount of time that it takes to reach the adult stage.(c) Fecundity is measured as individual egg production for the first gonotrophic cycle.(d) Model fits for estimated mean adult survival (in days) derived from measures of survival of adults, shown in the left panel as individual deaths at each temperature.Results for each individual replicate line are present in Figures S8-S10.

F
Temperature-dependent fitness (intrinsic rate of increase, r) derived from individual life-history traits measured for Aedes aegypti mosquitoes reared at 13, 17, 21, 25, 27, 29, 31, 33, and 35°C.The fitness model in blue is from the mosquito lines passaged at 27°C, while the model in red represents the mosquitoes passaged at 31°C.The lines indicate mean model fits, and the shaded areas indicate 95% credible intervals.