Species-Specificity in Thermopreference and CO2-Gated Heat-Seeking in Culex Mosquitoes

Simple Summary Mosquitoes are cold-blooded insects whose body temperature and metabolism are largely affected by environmental temperature. These blood-sucking insects use heat emanating from their potential hosts to locate them for feeding, which is how they spread deadly diseases. They also use other cues, including exhaled carbon dioxide and other body odors emitted by the hosts. Interestingly, every species displays specific preferences for a range of ambient temperatures and blood hosts, which includes both warm- and cold-blooded animals. To better understand the role of heat in these contexts, we studied female mosquitoes of three species that differ in their location of origin and in their host preference: Culex territans, Cx. tarsalis, and Cx. quinquefasciatus. We analyzed their preference towards specific ambient temperatures and quantified their heat-seeking behaviors in the presence of CO2 at different concentrations. We found contrasting differences between these species, which reflects their native habitat and their host preference. Abstract Combining thermopreference (Tp) and CO2-gated heat-seeking assays, we studied the thermal preferendum and response to thermal cues in three Culex mosquito species exhibiting differences in native habitat and host preference (e.g., biting cold and/or warm-blooded animals). Results show that these species differ in both Tp and heat-seeking behavior. In particular, we found that Culex territans, which feed primarily on cold-blood hosts, did not respond to heat during heat-seeking assays, regardless of the CO2 concentration, but exhibited an intermediate Tp during resting. In contrast, Cx. quinquefasciatus, which feeds on warm blooded hosts, sought the coolest locations on a thermal gradient and responded only moderately to thermal stimuli when paired with CO2 at higher concentrations. The third species, Cx. tarsalis, which has been shown to feed on a wide range of hosts, responded to heat when paired with high CO2 levels and exhibited a high Tp. This study provides the first insights into the role of heat and CO2 in the host seeking behavior of three disease vectors in the Culex genus and highlights differences in preferred resting temperatures.


Introduction
Temperature is an important abiotic factor for all living organisms, especially affecting poikilotherms such as insects, whose body temperature is heavily dependent on envi-performed thermal gradient assays to quantify the resting temperature preference (T p ) and conducted free-flight heat-seeking assays in the presence and absence of CO 2 at various host concentrations for the three mosquito species.

Insect Rearing
Cx. quinquefasciatus (JHB strain, BEI Resources (NR-43025)) and Cx. tarsalis (YOLO strain, BEI Resources (NR-43026)) were reared from eggs hatched in larval trays (Bioquip Industries, Rancho Dominguez, CA, USA) and fed fish food (Hikari First Bites powdered fish food, Kyorin Food Industries, Kansai City, Japan) until they were collected at pupation. Cx. territans were collected as larvae at Mountain Lake Biological Station, Pembroke, VA, USA. The larvae were placed into larval trays and fed fish food as previously described until pupation. For each species, pupae of the same age were grouped in emergence funnel containers (Bioquip Industries, Rancho Dominguez, CA, USA), which were moved to a light box (opaque gray box fitted with an internal light source controlled by a timer) within the first two days after collection in order to entrain the adults on a 12 h/12 h light/dark cycle. All experiments were performed in the first 2 h of mosquitoes' peak activity, i.e., the first 2 h after the offset of the lights. All three species were maintained in climatic chambers (Percival) at 24-26 • C, 70% RH and provided with a 10% sucrose solution ad libitum and starved 24 h before the experiments occurred. Six to ten-day-old mated females were used for both the thermopreference and CO 2 -gated heat-seeking experiments.

Thermopreference Assays
The thermal gradient was adapted from the devices used in Ritchie et al. [41] and Verhulst et al. [20], consisting of an aluminum plate (86 × 25 × 3 cm, 6061 general purpose aluminum) with custom made acrylic covers (67 × 5 × 2.5 cm; Figure 1). Two temperature gradients (low: 14-34 • C or high: 27-47 • C) across the aluminum plate were created by two waterbaths (F500 Compact Recirculating Cooler, Julabo C-B17 Corio Open Heating Bath Circulator, Julabo, Seelbach, Germany) circulating water through both ends of the plate via copper piping embedded in the aluminum plate. The linearity of the temperature gradient for the surface of the aluminum plate was quantified and verified by placing iButtons across the whole gradient (N = 33; DS1923-F5# Hygrochron Temperature and Humidity, Embedded Data Systems, Lawrenceburg, KY, USA) to determine exact temperature ranges for both experimental conditions. Both the calibration and experiments were conducted 30-45 min after the water baths were turned on to allow for the gradient to be established and stable. In addition to measuring temperature on the plate surface, we inserted thermocouples (Proster Digital Two K-Type Thermocouple Temperature Thermometer) at regular intervals in the acrylic covers to measure the temperature above the plate (i.e., air temperature) to control for a vertical thermal gradient and to obtain more precise T p data depending on the landing position of the mosquito (i.e., either directly on the plate surface or on the side of the cover). No mosquitoes landed on the top portion of the cover. Moist rolled wipes (Kimwipe, CAT# 34120) were placed on each side of the acrylic cover to minimize the establishment of humidity gradients across the length of the thermal gradient. Humidity gradients created by the moist rollers were also quantified using i-buttons as mentioned above (low temperature gradient: 88-61% RH, average = 77%; high temperature gradient: 72-45% RH, average = 61%). Ten mosquitoes were released through an opening located in the middle of each acrylic cover and were allowed to adjust for 5 min before the experiment started. Up to four assays were conducted in parallel. After 30 min, a photograph of the thermal gradient was taken, and the position of the mosquitoes was reported to the calibration curve corresponding to each experiment to determine T p . The assays were conducted in a darkened room within one hour from the onset of the scotophase, i.e., mosquitoes' subjective nighttime. Gloves were worn during the handling of the mosquitoes and equipment to minimize the risk of contamination with human odors. conducted in parallel. After 30 min, a photograph of the thermal gradient was taken, and the position of the mosquitoes was reported to the calibration curve corresponding to each experiment to determine Tp. The assays were conducted in a darkened room within one hour from the onset of the scotophase, i.e., mosquitoes' subjective nighttime. Gloves were worn during the handling of the mosquitoes and equipment to minimize the risk of contamination with human odors. Figure 1. Schematic of the thermal gradient experimental setup. A temperature gradient was created on an aluminum plate (1) using a cold and a warm water bath on either side, connected with copper tubing (2). The mosquitoes were released into acrylic covers (3) through a hole in the top with a fitted acrylic plug (4). Each cover had moist wipe towels (5) and holes drilled on each side to minimize humidity build up (6). The acrylic enclosures are open on their bottom side to allow direct contact of mosquitoes with the aluminum plate.

CO2-Gated Heat Seeking Assays
The heat and CO2 seeking assay was performed based on methods adapted from previous studies [11,12]. The setup consisted of a mosquito rearing cage (Rearing and Observation Cage, 12" cube, Bioquip) with two Peltier elements (6 × 4 cm surface area; 12 V 5 A, Peltier Thermo-Electric Cooler Module and Heatsink Assembly, Part # 1335, Adafruit, New Your, NY, USA) equidistantly placed against the mesh lining one of the vertical walls of the cage. A 2 × 2 cm black square printed on standard printer paper (bright white, letter size; Gemini/Liberty paper, Los Angeles, CA, USA) placed at the center of both the Peltier elements served as a visual cue, which has been shown to enhance attraction for warm surfaces in Aedes mosquitoes [12,14] (Figure 2A). At the beginning of each 75 min long trial, 15 female mosquitoes were released into the cage and allowed to acclimate for 5 min. During acclimation, both Peltier elements were maintained at ambient temperature (23 °C). Post acclimation, one of the Peltier elements (Peltierwarm), chosen at random, was warmed up to deliver an increasing sequence of thermal stimuli between 30 and 50 °C with a 5 °C difference between successive stimuli, i.e., 30,35,40,45, and 50 °C. The sequence of thermal stimuli in the assay was not randomized as exposure to higher temperatures in the initial phase of the assay might affect mosquitoes' subsequent responses towards thermal stimuli at lower temperatures. Each of these thermal stimuli lasted for 10 min, following which the Peltier was cooled to ambient temperature for 5 min ( Figure 2B). A humidified air stream (6.80 m/s) was delivered into the cage from the center of its top side throughout the duration of the assay. A CO2 pulse (2100 ppm or 30000 ppm; 0.8 m/s, Gasco, Oldsmar, FL, USA) lasting 2 min accompanied the onset of each thermal stimulus and was injected into the humidified air circuit (Figure 2A). The other Peltier element (Peltierambient) was maintained at ambient temperature throughout the trial. The surface temperatures were monitored using a thermal imaging camera (C3, FLIR Systems, USA) and precisely controlled via a custom-built Arduino PID controller (Arduino Uno R3; Monster Moto Shield VNH2SP30). The circuit diagram and code for the PID controller used in this assay are available online (https://github.com/mosquitohub/Culex-Thermal-Biology.git, accessed on 10 January 2022). Gloves were used to release mosquitoes into the experimental setup to avoid contamination with host odors, and the Figure 1. Schematic of the thermal gradient experimental setup. A temperature gradient was created on an aluminum plate (1) using a cold and a warm water bath on either side, connected with copper tubing (2). The mosquitoes were released into acrylic covers (3) through a hole in the top with a fitted acrylic plug (4). Each cover had moist wipe towels (5) and holes drilled on each side to minimize humidity build up (6). The acrylic enclosures are open on their bottom side to allow direct contact of mosquitoes with the aluminum plate.

CO 2 -Gated Heat Seeking Assays
The heat and CO 2 seeking assay was performed based on methods adapted from previous studies [11,12]. The setup consisted of a mosquito rearing cage (Rearing and Observation Cage, 12" cube, Bioquip) with two Peltier elements (6 × 4 cm surface area; 12 V 5 A, Peltier Thermo-Electric Cooler Module and Heatsink Assembly, Part # 1335, Adafruit, New Your, NY, USA) equidistantly placed against the mesh lining one of the vertical walls of the cage. A 2 × 2 cm black square printed on standard printer paper (bright white, letter size; Gemini/Liberty paper, Los Angeles, CA, USA) placed at the center of both the Peltier elements served as a visual cue, which has been shown to enhance attraction for warm surfaces in Aedes mosquitoes [12,14] (Figure 2A). At the beginning of each 75 min long trial, 15 female mosquitoes were released into the cage and allowed to acclimate for 5 min. During acclimation, both Peltier elements were maintained at ambient temperature (23 • C). Post acclimation, one of the Peltier elements (Peltier warm ), chosen at random, was warmed up to deliver an increasing sequence of thermal stimuli between 30 and 50 • C with a 5 • C difference between successive stimuli, i.e., 30,35,40,45, and 50 • C. The sequence of thermal stimuli in the assay was not randomized as exposure to higher temperatures in the initial phase of the assay might affect mosquitoes' subsequent responses towards thermal stimuli at lower temperatures. Each of these thermal stimuli lasted for 10 min, following which the Peltier was cooled to ambient temperature for 5 min ( Figure 2B). A humidified air stream (6.80 m/s) was delivered into the cage from the center of its top side throughout the duration of the assay. A CO 2 pulse (2100 ppm or 30,000 ppm; 0.8 m/s, Gasco, Oldsmar, FL, USA) lasting 2 min accompanied the onset of each thermal stimulus and was injected into the humidified air circuit (Figure 2A). The other Peltier element (Peltier ambient ) was maintained at ambient temperature throughout the trial. The surface temperatures were monitored using a thermal imaging camera (C3, FLIR Systems, Wilsonville, OR, USA) and precisely controlled via a custom-built Arduino PID controller (Arduino Uno R3; Monster Moto Shield VNH2SP30). The circuit diagram and code for the PID controller used in this assay are available online (https://github.com/mosquito-hub/Culex-Thermal-Biology.git, accessed on 10 January 2022). Gloves were used to release mosquitoes into the experimental setup to avoid contamination with host odors, and the experiment was triggered and controlled remotely to prevent interference from the experimenter. Mosquitoes were attracted to the Peltier surface as well as the adjacent surfaces, perhaps owing to heat dissipation and convective currents. To account for attraction elicited by the dissipated heat, a target region (9 × 9 cm) around each of the Peltier elements was defined within which the number of mosquitoes that landed every 30 s throughout the trial was quantified by manually transcribing the videos. experiment was triggered and controlled remotely to prevent interference from the experimenter. Mosquitoes were attracted to the Peltier surface as well as the adjacent surfaces, perhaps owing to heat dissipation and convective currents. To account for attraction elicited by the dissipated heat, a target region (9 × 9 cm) around each of the Peltier elements was defined within which the number of mosquitoes that landed every 30 s throughout the trial was quantified by manually transcribing the videos. Schematic of ramping temperature steps of the "warm" Peltier element. The warm Peltier element was brought to ambient temperature (2 min) between each step (8 min), which increased the temperature by 5 degrees. Each step began with a 2-min pulse of CO2 (in gray).

Thermopreference Assays
Data from the thermopreference assays were imported into R [42] for analysis and visualization. In a first step, the distribution of the temperature preferred by mosquitoes for each experiment and species was compared to a uniform, continuous distribution by means of Kolmogorov-Smirnov tests. Then, the effect of the species and experiments on the preferred temperatures and relative humidities were analyzed by means of Linear Models (LM) with the species (3 levels: Cx. quinquefasciatus, Cx. tarsalis, Cx. territans) and gradient types (3 levels: constant, low, high) as categorical fixed predictors and a gaussian . Schematic of the heat-seeking experimental setup. Mosquitoes were released into a meshed cage with two Peltier elements directly applied to one side: one maintained at ambient temperature (23 • C, outlined in blue), and one set up to warm up as described in (B). A camera recorded the landing activity on the Peltier elements. Through the top of the cage, a tube delivered a constant flow of humidified air to which pulses of CO 2 were added as described in (B). Schematic of ramping temperature steps of the "warm" Peltier element. The warm Peltier element was brought to ambient temperature (2 min) between each step (8 min), which increased the temperature by 5 degrees. Each step began with a 2-min pulse of CO 2 (in gray).

Thermopreference Assays
Data from the thermopreference assays were imported into R [42] for analysis and visualization. In a first step, the distribution of the temperature preferred by mosquitoes for each experiment and species was compared to a uniform, continuous distribution by means of Kolmogorov-Smirnov tests. Then, the effect of the species and experiments on the preferred temperatures and relative humidities were analyzed by means of Linear Models (LM) with the species (3 levels: Cx. quinquefasciatus, Cx. tarsalis, Cx. territans) and gradient types (3 levels: constant, low, high) as categorical fixed predictors and a gaussian error distribution. Tukey post hoc tests with p value adjustment were used as a followup analysis for multiple comparisons, using the R packages lme4 (version 1.1-27.1 [43]), multcomp (version 1.4-17 [44]), and emmeans (version 1.7.1-1 [45]).
In our thermal preference assays, mosquitoes either landed on the substrate (i.e., the aluminum plate of the gradient) or the sidewalls of the acrylic covers. To compare the proportions of mosquitoes landing on either the substrate or the sidewalls of the apparatus we used a Generalized Linear Model with a binomial error distribution and a logit link. The species (3 levels: Cx. quinquefasciatus, Cx. tarsalis, Cx. territans) and gradient types (3 levels: constant, low, high) were used as categorical fixed predictors in the model. Post hoc pairwise comparisons between species and gradient types were achieved with the Tukey method for p value adjustment using the R packages lme4 [43] and emmeans [45]. Visualization of the location of each mosquito landed either on the substrate or the sidewalls of the gradient was achieved by scaling the coordinates between 0 (the minimum recorded value) and 1 (the maximum recorded value) according to the formula: where i represents each individual mosquito.

CO 2 -Gated Heat Seeking Assays
Data from the heat-seeking assay were analyzed using Generalized Linear Mixed Models with a Penalized Quasi-Likelihood approach (glmmPQL in R package MASS, version 7.3-54 [46]). The model assumed binomially distributed errors with proportion of mosquito landings on Peltier warm as the response variable. The species of Culex mosquitoes (Cx. quinquefasciatus, Cx. tarsalis, Cx. territans), CO 2 concentrations (2100 ppm and 30,000 ppm), and ambient (23 • C) vs. warm (30 • C to 50 • C) thermal stimuli are the categorical fixed predictor variables included as fixed effects in the model. The proportion of mosquito landings on Peltier ambient was included as a random effect. As this dataset involves repeated measurements on fifteen mosquitoes per assay across time points, to account for temporal correlations in the response variable, the residual correlation structure was incorporated in the model using Autoregressive order 1 (AR-1) function [47]. The random effect in the model represents variations in mosquito responses resulting from mere presence of the Peltier elements in the experiment cage in the absence of any thermal stimulus. A three-way interaction between the predictor variables was modeled to formally test for species-specific responses towards thermal stimuli set at ambient and host-like temperatures before and after CO 2 exposure. Post-hoc analysis for significant effects was performed using Tukey's HSD test and the reported P values are adjusted for multiple comparisons (Tukey's method) using the function emmeans (in R package emmeans, version 1.7.1-1 [45]). In a subsequent analysis, the mosquito responses to the thermal stimuli (10 min per stimulus; 30-50 • C) were compared across five 2-min intervals. A CO 2 pulse accompanied the first of the five 2-min intervals for every thermal stimulus. Finally, the proportion of mosquitoes landing on Peltier warm at 23 • C (5 min ahead of every thermal stimulus) was compared across thermal stimuli to test for the effects of multiple exposures to thermal stimuli on mosquito heat-seeking behavior. All results are presented as effect sizes with the corresponding 95% confidence intervals. Statistical significance was determined at an experiment-wise α = 0.05. We used R version 3.6.2 [42] to perform all the analyses and visualize the data (using R package ggplot2, version 3.3.5 [46]).

Thermopreference Assays
All three mosquito species displayed even and continuous distribution when provided with a constant temperature (i.e., 25 • C), which indicated no spatial preference, thigmotaxis, or bias relating to the setup and environment (Kolmogorov-Smirnov tests: p = 0.42, p = 0.08, and p = 0.28 for Cx. quinquefasciatus, Cx. tarsalis, and Cx. territans, respectively) ( Figure 3A). In the low gradient experiments, all species displayed a significant difference from a continuous distribution (Kolmogorov-Smirnov tests: p < 0.04). Culex tarsalis was more distributed throughout the aluminum plate than Cx. territans, which concentrated on the center of the gradient, yet they exhibited similar average T p (low: T p = 25.8 ± 5.6 • C and T p = 25.6 ± 4.9 • C, respectively; Tukey Contrasts for multiple comparisons of means: p = 0.9). Similarly, in the high gradient experiments, both species aggregated principally in the center of the gradient with Cx. tarsalis exhibiting a slightly higher average T p compared to Cx. territans (T p = 38 ± 5.6 • C and T p = 35.9 ± 5.2 • C, for Cx. tarsalis and Cx. territans, respectively) (Tukey Contrasts for multiple comparisons of means: p = 0.11) ( Figure 3B).
Although both Cx. territans and Cx. tarsalis' distributions were significantly different from continuous distributions, only Cx. tarsalis' was not different from a normal, gaussian, distribution (all Kolmogorov-Smirnov tests: p < 0.008; Shapiro-Wilk normality test: p = 0.037 and p = 0.113 for Cx. territans and Cx. tarsalis, respectively). Cx. quinquefasciatus behaved differently compared to Cx. tarsalis and Cx. territans in both low and high gradient experiments. Indeed, these mosquitoes showed a preference for the coolest spot available on the gradient (low: T p = 19.5 ± 4.6 • C and high: T p = 30.6 ± 5.9 • C), a behavior that was not observed during the control (i.e., constant temperature) experiment (Tukey Contrasts for multiple comparisons of means: p < 0.001 for all pairwise comparisons) ( Figure 3A-C). 0.04). Culex tarsalis was more distributed throughout the aluminum plate than Cx. territans, which concentrated on the center of the gradient, yet they exhibited similar average Tp (low: Tp = 25.8 ± 5.6 °C and Tp = 25.6 ± 4.9 °C, respectively; Tukey Contrasts for multiple comparisons of means: p = 0.9). Similarly, in the high gradient experiments, both species aggregated principally in the center of the gradient with Cx. tarsalis exhibiting a slightly higher average Tp compared to Cx. territans (Tp = 38 ± 5.6 °C and Tp = 35.9 ± 5.2 °C, for Cx. tarsalis and Cx. territans, respectively) (Tukey Contrasts for multiple comparisons of means: p = 0.11) ( Figure 3B). Although both Cx. territans and Cx. tarsalis' distributions were significantly different from continuous distributions, only Cx. tarsalis' was not different from a normal, gaussian, distribution (all Kolmogorov-Smirnov tests: p < 0.008; Shapiro-Wilk normality test: p = 0.037 and p = 0.113 for Cx. territans and Cx. tarsalis, respectively). Cx. quinquefasciatus behaved differently compared to Cx. tarsalis and Cx. territans in both low and high gradient experiments. Indeed, these mosquitoes showed a preference for the coolest spot available on the gradient (low: Tp = 19.5 ± 4.6 °C and high: Tp = 30.6 ± 5.9 °C), a behavior that was not observed during the control (i.e., constant temperature) experiment (Tukey Contrasts for multiple comparisons of means: p < 0.001 for all pairwise comparisons) ( Figure 3A-C).  The proportion of mosquitoes resting on the sides of the covers versus on the plate was higher in Cx. territans (89.1 ± 3.3%) compared to both Cx. tarsalis (66.1 ± 5.6%) and Cx. quinquefasciatus (69.0 ± 5.6%) for the low gradient experiments (Pairwise comparisons on the log odds ratio scale with Tukey method for p value adjustment: p = 0.0057 and p = 0.0256, respectively) ( Figure 4). However, the proportion was similar in the three species for the high gradient experiments (Cx. quinquefasciatus: 72.0 ± 5.3%; Cx. tarsalis: 75.6 ± 5.4%; Cx. territans: 70.8 ± 5.5%, respectively). (Pairwise comparisons on the log odds ratio scale with the Tukey method for p value adjustment: p > 0.99) (Figure 4). Interestingly, 20% of Cx. tarsalis mosquitoes were found knocked down on the warmer side of the gradient during the high gradient experiments, which did not occur in either of the other species.
mosquitoes' distribution along the gradient. Each dot represents the final resting position (i.e., Tp) of a single female mosquito. The boxes represent the upper and lower quartiles, and the black bars indicate the mean of each group. n.s. denotes mosquito distributions in the constant gradient experiments that were not significantly different from a uniform, continuous distribution (Kolmogorov-Smirnov tests, α = 0.05). Letters denote statistical differences between groups (Tukey post hoc tests for multiple comparisons, adjusted α = 0.05). Four replicates (n = 10; N = 40) have been used for the constant gradient. Ten replicates (n = 10, N = 100) per species have been used for the low and high gradients.
The proportion of mosquitoes resting on the sides of the covers versus on the plate was higher in Cx. territans (89.1 ± 3.3%) compared to both Cx. tarsalis (66.1 ± 5.6%) and Cx. quinquefasciatus (69.0 ± 5.6%) for the low gradient experiments (Pairwise comparisons on the log odds ratio scale with Tukey method for p value adjustment: p = 0.0057 and p = 0.0256, respectively) ( Figure 4). However, the proportion was similar in the three species for the high gradient experiments (Cx. quinquefasciatus: 72.0 ± 5.3%; Cx. tarsalis: 75.6 ± 5.4%; Cx. territans: 70.8 ± 5.5%, respectively). (Pairwise comparisons on the log odds ratio scale with the Tukey method for p value adjustment: p > 0.99) (Figure 4). Interestingly, 20% of Cx. tarsalis mosquitoes were found knocked down on the warmer side of the gradient during the high gradient experiments, which did not occur in either of the other species. Figure 4. Proportion of mosquitoes resting on the aluminum plate (i.e., "plate", green triangle) and on the side of the covers (i.e., "sidewalls", grey circle) for each of the temperature gradients and mosquito species tested. Four replicates (n = 10; N = 40) have been used for the constant gradient. Ten replicates (n = 10; N = 100) per species have been used for the low and high gradients.
In these experiments, while the humidity gradient was minimized by the introduction of moist wipes, a humidity gradient, negatively correlated with the temperature gradient (Pearson's product-moment correlation: R 2 = −0.95; p < 0.001). Given this strong correlation, the hygric preference across gradients and species mirrors the patterns observed with the thermal preferences, where Cx. quinquefasciatus significantly preferred more humid locations on the gradients (82.2 ± 0.648% RH and 67.5 ± 0.648% RH Figure 4. Proportion of mosquitoes resting on the aluminum plate (i.e., "plate", green triangle) and on the side of the covers (i.e., "sidewalls", grey circle) for each of the temperature gradients and mosquito species tested. Four replicates (n = 10; N = 40) have been used for the constant gradient. Ten replicates (n = 10; N = 100) per species have been used for the low and high gradients.
In these experiments, while the humidity gradient was minimized by the introduction of moist wipes, a humidity gradient, negatively correlated with the temperature gradient (Pearson's product-moment correlation: R 2 = −0.95; p < 0.001). Given this strong correlation, the hygric preference across gradients and species mirrors the patterns observed with the thermal preferences, where Cx. quinquefasciatus significantly preferred more humid locations on the gradients (82.2 ± 0.648% RH and 67.5 ± 0.648% RH for the low and high thermal gradients, respectively) than Cx. tarsalis (76.5 ± 0.621 and 59.2 ± 0.716% RH for the low and high thermal gradients, respectively) and Cx. territans (77.6 ± 0.645 and 61.8 ± 0.662% RH or the low and high thermal gradients, respectively). (Tukey post hoc tests: p < 0.001 for all comparisons). No significant differences were found between Cx. tarsalis and Cx. territans in the low (Tukey post hoc test: p = 0.788) and high gradients (Tukey post hoc test: p = 0.067), but within each species hygric preferences were significantly higher in the low than the high gradients (Tukey post hoc tests: p < 0.001 for all comparisons), reflecting the higher humidity levels correlated with lower temperatures ( Figure 5). post hoc tests: p < 0.001 for all comparisons). No significant differences were found between Cx. tarsalis and Cx. territans in the low (Tukey post hoc test: p = 0.788) and high gradients (Tukey post hoc test: p = 0.067), but within each species hygric preferences were significantly higher in the low than the high gradients (Tukey post hoc tests: p < 0.001 for all comparisons), reflecting the higher humidity levels correlated with lower temperatures ( Figure 5).

CO2-Gated Heat Seeking Assays
Culex territans did not respond to any thermal stimuli both in the presence and absence of CO2 at both 2100 ppm and 30,000 ppm ( Figure 6) (i.e., no mosquito landed on Peltierambient and Peltierwarm). Therefore, the responses of Cx. territans were excluded from subsequent analysis. The heat-seeking responses of both Cx. tarsalis and Cx. quinquefasciatus were contingent on exposure to CO2 (Table 1; Supplementary Tables S1  and S2). The response of both species towards Peltierwarm (30-50 °C) was significantly higher after exposure to CO2 at 30,000 ppm when compared to 2100 ppm (Table 1;  Supplementary Tables S1 and S2). Post exposure to CO2 at 30,000 ppm, the proportion of heat-seeking Cx. tarsalis and Cx. quinquefasciatus increased significantly with the temperature of Peltierwarm and was the highest at 40 and 45 °C ( Figure 6A; Table 1). Between the two species, post exposure to CO2 at 30,000 ppm, the magnitude of heatseeking response of Cx. tarsalis towards Peltierwarm (30-50 °C) was significantly higher when compared to Cx. quinquefasciatus (Chisq: 15.22, p < 0.01; Supplementary Tables S1 and S2). Post exposure to a lower concentration of CO2, i.e., 2100 ppm, significantly fewer Cx. tarsalis and Cx. quinquefasciatus responded to the thermal stimuli from Peltierwarm set between 30 and 50 °C ( Figure 6A; Table 1). Post exposure to 2100 ppm CO2, the heat-

CO 2 -Gated Heat Seeking Assays
Culex territans did not respond to any thermal stimuli both in the presence and absence of CO 2 at both 2100 ppm and 30,000 ppm ( Figure 6) (i.e., no mosquito landed on Peltier ambient and Peltier warm ). Therefore, the responses of Cx. territans were excluded from subsequent analysis. The heat-seeking responses of both Cx. tarsalis and Cx. quinquefasciatus were contingent on exposure to CO 2 (Table 1; Supplementary Tables S1 and S2). The response of both species towards Peltier warm (30-50 • C) was significantly higher after exposure to CO 2 at 30,000 ppm when compared to 2100 ppm (Table 1; Supplementary Tables S1 and S2). Post exposure to CO 2 at 30,000 ppm, the proportion of heat-seeking Cx. tarsalis and Cx. quinquefasciatus increased significantly with the temperature of Peltier warm and was the highest at 40 and 45 • C ( Figure 6A; Table 1). Between the two species, post exposure to CO 2 at 30,000 ppm, the magnitude of heat-seeking response of Cx. tarsalis towards Peltier warm (30-50 • C) was significantly higher when compared to Cx. quinquefasciatus (Chisq: 15.22, p < 0.01; Supplementary Tables S1 and S2). Post exposure to a lower concentration of CO 2 , i.e., 2100 ppm, significantly fewer Cx. tarsalis and Cx. quinquefasciatus responded to the thermal stimuli from Peltier warm set between 30 and 50 • C ( Figure 6A; Table 1). Post exposure to 2100 ppm CO 2 , the heat-seeking responses of Cx. tarsalis and Cx. quinquefasciatus were not significantly different between 30 and 40 • C ( Figure 6A; Table 1). However, the proportion of heat-seeking Cx. quinquefasciatus towards Peltier warm set at 45 and 50 • C upon exposure to 2100 ppm of CO 2 was significantly more than Cx. tarsalis (Table 1;  Supplementary Tables S1 and S2).
The heat-seeking response of both Cx. tarsalis and Cx. quinquefasciatus towards Peltier warm between 30 and 50 • C peaked at 2 min post exposure to CO 2 (at 2100 and 30,000 ppm) and declined consistently thereafter over the remaining 6 min ( Table 2;  Supplementary Tables S3 and S4). While the activity of the mosquitoes was not quantified in this assay, Cx. tarsalis and Cx. quinquefasciatus were actively flying during their exposure to CO 2 at 30,000 ppm. The magnitude of flight activity and the number of active mosquitoes in the two species were lower during exposure to 2100 ppm. Culex territans did not exhibit any flight activity during exposure to CO 2 at 2100 and 30,000 ppm.
Finally, to account for the effects of the increasing sequence of thermal stimuli in the assay, we compared the responses of Cx. tarsalis and Cx. quinquefasciatus towards Peltier warm at 23 • C before every thermal stimulus ( Figure 6). With every exposure to increasing thermal stimuli in Peltier warm , irrespective of the CO 2 concentration, significantly fewer mosquitoes moved away when Peltier warm was cooled to ambient temperature, i.e., 23 • C (Chisq: 17.98, p < 0.01; Table 3; Supplementary Tables S5 and S6), which could be due to the time interval (5 min) between thermal stimuli presentations. The heat-seeking response of both Cx. tarsalis and Cx. quinquefasciatus towards Peltierwarm between 30 and 50 °C peaked at 2 min post exposure to CO2 (at 2100 and 30,000 ppm) and declined consistently thereafter over the remaining 6 min ( Table 2;  Supplementary Tables S3 and S4). While the activity of the mosquitoes was not quantified in this assay, Cx. tarsalis and Cx. quinquefasciatus were actively flying during their exposure to CO2 at 30000 ppm. The magnitude of flight activity and the number of active mosquitoes in the two species were lower during exposure to 2100 ppm. Culex territans did not exhibit any flight activity during exposure to CO2 at 2100 and 30,000 ppm.

Discussion
In this study, we found differences in thermopreference across three species of Culex mosquitoes. Culex quinquefasciatus selected the cooler and more humid locations of the thermal gradient on both low and high temperature gradients, whereas Cx. territans and Cx. tarsalis displayed higher T p . However, the distribution of these two species was different between the two temperature ranges tested here. Culex tarsalis was more evenly distributed in the low gradient compared to the high gradient experiments, revealing a thermal preferendum for warmer ambient temperatures than those tested in the low thermal gradient. Culex territans, however, showed similar distribution patterns in the low and high gradients except that the coolest temperatures of the low gradient and the warmest temperatures of the high gradient tended to be avoided. Altogether, this suggests a preference for temperatures between 20 and 40 • C for this species.
A species' preference for the cooler or warmer resting temperatures may be related to the abiotic conditions associated with their natural environment. Culex quinquefasciatus is found mostly at low-to-moderate elevations throughout the tropical, subtropical, and warm temperate regions of the world [48,49]. Culex tarsalis is distributed across most of the USA in the subtropical, temperate, and desert regions, with the exception of the East coast and Southern Canada [21]. In contrast, Cx. territans is widely distributed throughout the Northern hemisphere and found in subtropical, temperate, and subarctic regions of the US, Canada, and Europe [21,22]. In addition to temperatures, each species experiences different humidity conditions in its native habitat, which may affect their risk of desiccation. Maintaining water balance is indeed particularly critical for mosquitoes and several Culex species can suppress water loss under unfavorable conditions and during diapause by adjusting their metabolic rates, changing their cuticle composition and by synthesizing HSPs [49][50][51][52]. Anderson and Hardwood [53] found that wild populations of Cx. tarsalis that diapause for longer periods of time tend to be more resistant to both cold and desiccation. Rinehart et al. [52] showed the same under laboratory conditions with Cx. pipiens reared under diapausing and non-diapausing conditions.
Mosquito strains used in the present study originate from various regions of the world that greatly differ in terms of annual average temperatures, rainfall and elevation. These abi-otic factors can lead to variations in both bionomics and genomics, which could explain the results obtained in the present study. The Cx. quinquefasciatus strain tested here derives from eggs collected in Johannesburg, South Africa, where annual temperatures are mild (max: 26 • C, min: 15 • C) with a long dry season (Source: NOAA). Here, we found that this species selected cooler resting temperatures in both thermal gradients, which also correlates with higher humidity levels. Culex tarsalis was collected in California (Yolo county, CA, USA) where temperatures can average 35 • C during the summer and where rainfall is relatively low (Source: NOAA). The warm and dry conditions this mosquito is accustomed to could have shaped its preference for warmer and less humid conditions (i.e., this strain is more adapted to hot and dry conditions). Culex territans was collected at Mountain Lake Biological Station (VA, USA), which is at high elevation with humid cold winters and mild summers (max: 29 • C, min: 17 • C in July) (Source: MLBS weather station). Our data show that this species had an intermediate T p compared to the two other species. As inter-populations differences in terms of thermotolerance, resistance to desiccation [7,50], and bionomics [49] have been highlighted in Culex spp., it would be interesting to conduct experiments with populations of the same species originating from other regions of the world to further examine the extent to which environmental conditions in the native habitat influence Culex mosquito T p .
In addition to climatic conditions in the native habitat influencing T p , we could also hypothesize that Cx. quinquefasciatus' selection for a cooler and more humid environment might also be due to host density and availability. Indeed, temperature may also affect the hosts' behavior (e.g., overall activity, sheltering). However, due to their ability to regulate their body temperature, endotherms, which Cx. quinquefasciatus feeds on, might be available at cooler temperatures. In contrast, in a cooler environment, ectotherms targeted by Cx. territans and to a lesser extent Cx. tarsalis, might not be active at all. As heat is an important host-seeking cue for mosquito species feeding on endotherms, a contrast between the temperatures of the environment and of the host is needed to trigger hostseeking as well as biting. Thus, by selecting a lower T p , Cx. quinquefasciatus might increase their chances of detecting the heat signature of a potential host available in the surrounding environment. Determining T p of other tropical species would be beneficial to determine if this is a typical preference for species in these regions.
Beyond influencing the sites that adult mosquitoes will choose for resting, the environmental temperature also affects their overall activity and host-seeking as well as blood-feeding behaviors [7][8][9][10]. In these contexts, thermosensation plays an important role at close range from the host as mosquitoes use the convective plumes generated by animals to guide their landing orientation before initiating blood-feeding [11][12][13][14][15][16]. In addition to thermal cues, mosquitoes use multiple host-specific olfactory, visual, and gustatory cues along with carbon dioxide plumes to identify and locate potential hosts for blood feeding [16,54,55], which allows for disease transmission. In Aedes aegypti, while responses to established convective plumes have been observed in choice assays [16], the addition of CO 2 was necessary to elicit landings on warm surfaces (i.e., Peltier elements) whose temperatures were transiently increased to natural host temperatures [13]. In the present study, the heat-seeking experiments revealed that Cx. tarsalis and Cx. quinquefasciatus also required the presence of CO 2 at elevated levels (30,000 ppm) to respond to an object warmer than the ambient temperature. These two species displayed comparable response profiles as a function of the CO 2 concentration, whereas Cx. territans did not show any marked responses, even at ecologically relevant (i.e., corresponding to their preferred host's) levels of CO 2 [56]. As Cx. quinquefasciatus feeds mainly on endotherms [33], and Cx. tarsalis is opportunistic but prefers endotherms [25][26][27][28], the similarity in their responses was expected. However, we found that the proportion of landing of Cx. quinquefasciatus was reduced at 50 • C, which reflects both their host preference as well as their lower T p compared to Cx. tarsalis. However, the continued interest in the Peltier warm around 50 • C in Cx. tarsalis may exemplify their opportunistic feeding nature as well as their higher T p . Culex territans' lack of response to the heat stimuli, regardless of the CO 2 concentration, also reflects their host preference, as amphibians have no thermal signature and expel low levels of CO 2 [56].
Several factors could contribute to these differences, including inter-specific variations in anatomical and morphological structures associated with host seeking as well as differences in the expression of receptors implicated in host detection [57,58]. Interestingly while Cx. quinquefasciatus and Cx. tarsalis are closely related species, Cx. territans is more phylogenetically distant from them [59][60][61][62][63]. Mosquitoes sense many aspects of their environment, including odorants, CO 2 , and heat, through specific receptors on their antennae, maxillary palps, tarsi and wings [64][65][66][67][68][69]. At the antennae level, a pair of thermoreceptors, one cold-sensitive and the other warm-sensitive, is housed together in a sensilla at the tip of the antennae. McIver [70] described structures in the first segment of the antennae in several Culex species and Ae. aegypti, which were later identified to mediate thermoception in Ae. aegypti mosquitoes [71]. Similar thermoreceptors, later characterized as TRPA1, have been identified in other mosquito species such as Anopheles gambiae [72] and Culex pipiens [73]. TRPA1, however, is known to be involved in mediating heat avoidance while Ir21a, a member of the ionotropic receptor (IR) family, has been found to be the primary receptor responsible for heat-seeking in An. gambiae and possibly other mosquito species [55]. Morphological studies showed that Cx. territans have fewer CO 2 -sensitive sensilla on their maxillary palps than Cx. tarsalis and Cx. pipiens [74], which could explain why, in the present study, Cx. territans did not respond to the heat stimuli combined with CO 2 . This could also be due to the absence of other sensory cues. Indeed, this species is known to use phonotaxis (i.e., frog calls) to locate their hosts [75]. In addition, our recent work has evinced the use of odors for host detection and blood feeding (Reinhold et al., in preparation). Gustatory receptors genes (Gr) are expressed in sensory neurons in sensilla on the maxillary palps and three GRs have been identified as critical for CO 2 sensing in several mosquito genera, including Anopheles, Aedes as well as Culex [76][77][78]. However, knowledge on receptors associated with host seeking, in particular for heat and CO 2 detection, in the three Culex species we focused on in this study remains limited, and future studies will be necessary to unravel the physiological and molecular mechanisms underlying host seeking in these disease vector insects.

Conclusions
In this study, we determined the T p and CO 2 -gated heat-seeking behaviors of three Culex spp., Cx. tarsalis, Cx. territans, and Cx. quinquefasciatus. We found that both the environmental conditions in the habitat of origin and host preferences of these three species impacted their T p and host seeking behavior. Future experiments will further dive into better understanding the thermal biology of these species at the behavioral and genetic levels.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/insects13010092/s1, Table S1: Analysis of deviance table for the Generalized Linear Mixed Model analyzing the proportion of mosquito landings on Peltier warm in the heat seeking assay (data in Table 1), Table S2: Pairwise contrasts for the proportion of mosquito landings on Peltier warm in the heat seeking assay (data in Table 1), Table S3: Analysis of deviance table for the Generalized Linear Mixed Model analyzing the proportion of mosquito landings on Peltier warm as a function of the duration of exposure to the thermal stimuli in the heat seeking assay (data in Table 2), Table S4: Pairwise contrasts for the proportion of mosquito landings on Peltier warm as a function of the duration of exposure to the thermal stimuli in the heat seeking assay (data in Table 2), Table S5: Analysis of deviance table for the Generalized Linear Mixed Model analyzing the proportion of mosquito landings on Peltier warm at 23 • C before every thermal stimulus in the heat seeking assay (data in Table 3), Table S6: Pairwise contrasts for the proportion of mosquito landings on Peltier warm at 23 • C before every thermal stimulus in the heat seeking assay (data in Table 3). Data Availability Statement: Data can be accessed online https://github.com/mosquito-hub/Culex-Thermal-Biology.git (accessed on 10 January 2022) and upon email request.