Spectral preferences of mosquitos are altered by odors Open Access

Vision as a sensory modality plays an important role in many if not most ecological interactions among insects (Warrant, 2019). Mosquitos, like most insects, use vision in a variety of ecological interactions, including predator avoidance, mate finding, oviposition, locating vertebrates hosts and vectoring the pathogens of disease (Clements, 1999; Hawkes et al., 2022). However, vision is used in concert with many other sensory cues including odor, heat and humidity depending on the behavioral context. For example, in the context of vertebrate host finding in mosquitos, it has been shown that CO₂ can induce visual search behaviors (van Breugel et al., 2015; Carnaghi et al., 2021). In the absence of CO₂, mosquitos are not attracted towards dark high-contrast objects; however, after exposure to CO₂, mosquitos will approach and investigate these objects. Other cues such as heat, odor and humidity govern landing and biting (McMeniman et al., 2014; Cardé, 2015; Sumner and Cardé, 2022; Sumner et al., 2023; Giraldo et al., 2023). This CO₂-gated approach of mosquitos to visual stimuli is wavelength dependent, with mosquitos attracted to the cyan, orange and red spectral bands and showing some ability to discriminate between green and red stimuli of matched perceptual brightness (Alonso San Alberto et al., 2022).

Mosquitos have a visual system with many similarities to those of other dipterans (Hawkes et al., 2022), but we lack detailed information about their photoreceptors, unlike other flies such as Calliphora, Musca or Drosophila (Hardie, 1985; Sharkey et al., 2020). Like these flies, mosquitos have compound eyes made up of hundreds of individual units know as ommatidia consisting of six outer (R1–6) and two inner (R7,8) photoreceptor cells (Brammer, 1970). However, in higher flies, opsin expression varies among ommatidial types in a random mosaic across most of the compound eye, whereas in mosquitos, ommatidial types are highly regional with a single type in each eye region (Hu et al., 2009). Of the 10 opsins that have been identified in Aedes aegypti, five are expressed in the compound eye (Giraldo-Calderón et al., 2017). These include a pair of longwave green-sensitive opsins, a blue- and a UV-sensitive opsin as well as an opsin homologous to the Drosophila Rh7 that in mosquitos is also sensitive to green wavelengths (Hu et al., 2009, 2011, 2012, 2014). The outer photoreceptors always express the same green-sensitive opsin while the expression of the central photoreceptors varies across the eye regions, with the majority of the eye expressing green- and UV-sensitive opsins, with blue sensitivity limited to a small ventral stripe and the very dorsal portion of the eye. This opsin expression mirrors electroretinogram studies in Ae. aegypti showing two sensitivity peaks in the green (∼525 nm) and in the UV (∼350 nm) (Muir et al., 1992b). We lack sensitivity data for individual mosquito photoreceptors, and it is currently unclear whether the multiple longwave-sensitive opsins could provide the underpinnings to allow for the behavioral discrimination between green and red spectral bands.

Different odors have been shown to shift spectral preferences in other insects (Reisenman et al., 2000; Yoshida et al., 2015; Brodie et al., 2015; Balamurali et al., 2019; Bolton et al., 2021), which could be one reason for the conflicting results in behavioral studies examining spectral preference in mosquitos. Many studies have identified red and black as attractive colors (Fay and Prince, 1968; Muir et al., 1992a), including a recent study examining CO₂ gating of visual responses (Alonso San Alberto et al., 2022), but other colors have also been found to be attractive (Brett, 1938; Smart and Brown, 1956; Snow, 1971). If odors beyond CO₂ could gate or shift the spectral preference of mosquitos, this would seem adaptive as reflectance spectra of resources important to mosquitos can differ considerably. For example, human skin reflects a considerable amount of light in the 600–700 nm range (Stamatas et al., 2004) whereas the reflectance of many mosquito-pollinated flowers is highest between 500 and 600 nm (Alonso San Alberto et al., 2022). To investigate these potential shifts, we refined earlier techniques to make use of novel LED synthesizers (Belušič et al., 2016; Egri et al., 2020), allowing for much tighter control of stimulus wavelength and intensity, enabling refined estimates of spectral preference independent of the confounding effects of stimulus intensity. Using proven 3D tracking methods, we were then able to demonstrate that the addition of odors characteristic of human hosts, floral nectar sources or potential oviposition sites to a CO₂ plume could alter the responses of female Ae. aegypti to visual stimuli.

Photographic measurements (https://doi.org/10.17632/fdr7znz5dh.2) were taken with a Nikon D5100 DSLR (spectral sensitivity characterized in Jiang et al., 2013) fitted with a AF-S DX Nikkor 35 mm f/1.8 G lens. The captured raw images were decoded in FIJI (Schindelin et al., 2012), using the DCRAW plugin (https://www.dechifro.org/dcraw/) in a manner that preserved sensor linearity.

Spectra from the LED synthesizer and within the wind tunnel (Fig. 1A,B; Fig. S1; https://doi.org/10.17632/fdr7znz5dh.2) were measured using a calibrated spectrophotometer (USB-2000, Ocean Optics Inc., Dunedin, FL, USA) with reflectance spectra being calibrated against a 99% Spectralon reflectance standard (SRS-99-010, Labsphere, Inc., North Sutton, NH, USA).

Mosquitos (Aedes aegypti: Rockefeller strain) were provided by BEI Resources (Manassas, VA, USA) and were raised at the University of Washington campus in mixed sex groups of approximately 200 adults. Mosquitos were provided with sucrose ad libitum and maintained at 26–28°C, 80% relative humidity and a photoperiod of 12 h light:12 h dark. Females were used in behavioral experiments 7–8 days post-emergence and were assumed to be mated. Previous research had established a mating rate of over 90% with cohabiting females of this age (Alonso San Alberto et al., 2022). Before use in experiments, mosquitos were cold anesthetized, males removed, separated into groups of 35–50 females and deprived of sucrose for a period of 16–24 h. Mosquitos were used in experiments during an approximately 6 h period centered on the mosquito's subjective sunset. This time of day was chosen as mosquito flight activity increases in the approach of sunset and in the absence of a lights-off cue, this activity remains high for several hours following sunset (Clements, 1999).

These experiments made use of the same wind tunnel and real-time tracking system as described in previous papers (van Breugel et al., 2015; Vinauger et al., 2019; Zhan et al., 2021; Alonso San Alberto et al., 2022) and its construction and use are detailed in Alonso San Alberto et al. (2023). Any deviations from this setup and procedure are fully detailed below. Behavioral experiments took place in a low-speed wind tunnel (ELD Inc., Lake City, MN, USA), with a working section of 224×61×61 cm high and a constant laminar flow of 40 cm s⁻¹ (Fig. 1D). We used two short-throw projectors (LG PH450 U, Englewood Cliffs, NJ, USA) and rear projection screens (SpyeDark, Spye, LLC, Minneapolis, MN, USA; Fig. S1A,C) to provide low-contrast gray horizons on each side of the tunnel. The projectors provided ambient light at a level of ∼3 lx across the 420–670 nm range. Fabric liners were positioned on the floor of the working section to provide both a low-contrast background and a projection surface for the LED synthesizer arrays. The custom-sewn liner consisted of a piece of white cotton broadcloth (Fig. S1D,E; Jo-Ann Stores, LLC, Hudson, OH, USA) with 10 strips of black tulle (Fig. S1D,E; Jo-Ann Stores LLC) appliquéd at regular intervals. Sixteen cameras (Basler AC640gm, Exton, PA, USA) fitted with infrared (IR) pass filters (Kodak 89B, Kodak, Rochester, NY, USA) were positioned so that multiple cameras covered all areas of the working section, allowing mosquito trajectories to be recorded at 60 frames s⁻¹. IR backlights (B07ZZ2LJKY, 360DigitalSignage, Shenzhen, GD, China) were installed below and at the sides of the wind tunnel and diffused by the side screens and fabric floor to provide a uniform bright background in the IR to optimize mosquito tracking while falling well outside the visual sensitivity range of the mosquitos (Fig. S1B). The temperature within the working section was a constant 22.5°C and ambient CO₂ outside the wind tunnel was ∼400 ppm (Alonso San Alberto et al., 2022).

The CO₂, filtered clean air and odor-laden air were delivered using three mass flow controllers (MC-200SCCM-D, Alicat Scientific, Tucson, AZ, USA) whose output was combined and delivered to the working section via a single outlet located upwind of the visual stimuli at a height of 20 cm (Fig. 1D). The mass flow controllers were supplied with clean filtered air or CO₂ via gas canisters (Linde Gas & Equipment Inc., Burr Ridge, IL, USA). The timing and flow rates of CO₂, filtered clean air and odor-laden air were independently controlled by the same Python script that controlled visual stimuli. All stimulus series included a pre-CO₂ and post-CO₂ period to serve as a baseline of mosquito behavior in the absence of attractive stimuli. Preliminary testing indicated that the CO₂ concentration (0%, 1%, 5% or 10%) in the plume significantly increased both activation and recruitment to visual stimuli (Fig. 2A,B) but had no effect on spectral preference (Fig. 2C). For these reasons, we elected to perform all bioassays using a plume with 10% CO₂, despite this concentration being greater than what would be expected in the vicinity of vertebrate hosts (Geier et al., 1999), sources of floral nectar (Peach et al., 2019a), or oviposition sites.

Odor-laden air was directed through a 1 liter container enclosing the odor source, with the odor, when present, composing 10% of air being released from the outlet. Odor sources were changed in between runs, and the odor containers were removed and cleaned with 95% ethanol. Runs testing a floral odor used a freshly cut common tansy (Tanacetum vulgare) inflorescence, with 10–15 composite flowers and its stem inserted into a water-filled vial (20 ml). These flowers were collected from the vicinity of the University of Washington campus in Seattle, WA, USA. Common tansy was chosen as its scent had been previously demonstrated to be attractive to Ae. aegypti (Peach et al., 2019a,b).

Human foot odor has been shown to be behaviorally attractive to Ae. aegypti (Lacey et al., 2014; Bello and Cardé, 2022) and is easily collected using nylon socks (Njiru et al., 2006; Okumu et al., 2010). In runs testing a human odor, a single nylon sock was used as an odor source. The socks were worn for a period of 8–12 h by one of the investigators (A.J.B.). Odor collection was limited to a single individual to eliminate the potential of interindividual variability in odor composition (Ellwanger et al., 2021).

Plant infusions have been demonstrated to be effective oviposition attractants for Ae. aegypti (Reiter, 1991; Mwingira et al., 2020). Following the methods of Ritchie et al. (2014), we created an extract using 2 g of dried alfalfa in 1.2 l of water that was aerobically aged for at least 7 days and was then used up until the extract was 14 days old; 100 ml of this extract was added to the odor container.

Humidity is an important near-field cue for mosquitos (Cardé and Gibson, 2010; Laursen et al., 2023) and is a component of human sweat, stimuli from flowers and oviposition sites such as those mimicked by the tested plant infusions. To determine whether the relative humidity of the plume changed when sources of humidity were present in the odor container, relative humidity measurements (SHT4x sensor, Sensirion, Stäfa, Switzerland) were conducted, with results indicating that this aqueous odor source did not add a detectable amount of humidity to the plume (Fig. S2A). Behavioral investigations further demonstrated that humidity, in the form of 100 ml of water in the odor container, in the absence of CO₂ had no effect on the recruitment rate and its presence in combination with CO₂ did not alter the spectral preference of Ae. aegypti (Fig. S2B,C).

To gain more fine control of the spectral composition and intensity of visual stimuli, we moved away from the paper targets used previously (van Breugel et al., 2015; Zhan et al., 2021; Alonso San Alberto et al., 2022; Sumner and Cardé, 2022). Inspired by other systems used to generate visual stimuli for insects using LEDs (Belušič et al., 2016; Egri et al., 2020), we created a pair of LED synthesizers to generate visual stimuli. The synthesizers consisted of an array of 17 different LEDs with peak wavelengths ranging from 390 to 743 nm (Fig. 1B). While it would have been preferable to extend the range of stimuli into the UV, we were constrained by the UV transmission of the acrylic floor of the wind tunnel (Fig. S1C). These LEDs were mounted in a hemispheric dome aimed at a central diffuse reflector made of cotton broadcloth (Fig. 1A,C). The combination of the diffuse reflection and the aperture in the top of the dome allowed us to project the light from each LED to the same circular portion of the wind tunnel floor.

The intensity of the LEDs from each of the 17 color channels in both LED synthesizers was independently controlled using pulse width modification (PWM) using custom Arduino sketches uploaded to an Arduino Uno (Rev3, Adafruit, New York, NY, USA) coupled with four breakout boards (PCA9685, Adafruit) via a Python script (https://doi.org/10.17632/fdr7znz5dh.2). A combination of photographic and spectrographic measurements was used to tune each LED channel to produce isoquantal illumination of ∼3.5×10¹¹ photons cm⁻² s⁻¹ as measured at the surface of the wind tunnel floor. With the exception of the color intensity ramp bioassays, this intensity (1.0) was used for all monochromatic stimuli.

In order to create dark visual stimuli, it was necessary to project the light of these LED synthesizers onto a pair of black tulle targets formed by seven concentric circles of tulle (Fig. 1C,D). These circles approximated the inverse of the Gaussian intensity distribution of the light from the LED synthesizers. This arrangement prevented any intensity artifacts along the edge of the visual stimuli which would be present if all the tulle layers were the same diameter. These tulle targets also served to increase the saturation of colors produced by the LED synthesizers, as the light from the synthesizers only passed through the tulle layers once, while reflections from the fabric floor needed to pass through these layers twice (Fig. S1F).

The LED synthesizers also allowed for the creation of composite spectra using multiple LED channels. We created achromatic gray stimuli by approximating the spectral composition of DLP projectors providing ambient illumination (Fig. S1A). This composite spectrum, in combination with the tulle target, allowed us to create a neutral gray achromatic stimuli with an intensity (1.0) that closely matched the white fabric background (Fig. 1D, inset). We included several controls with achromatic gray stimuli of different intensities to serve as positive and negative controls, as well as to demonstrate the effect of achromatic intensity (Fig. 3). Mid-gray stimuli with an intensity midway (0.5) between the background intensity (1.0) and the unilluminated tulle target (0.0) were used as a common control stimulus across all experiments (Figs 3–,6; Figs S3, S4, S6).

We performed two main experimental series, spectral sweeps where each of the 17 LED color channels was tested at the same intensity (Figs 4 and 5) and color intensity ramps where we selected a limited set of LED color channels and displayed them at a range of intensities (0.00–3.00; Fig. 6). These experimental series were performed with CO₂ alone and with the combination of CO₂ and floral, host and oviposition site odors. In both experimental series, the order of stimuli was alternated so that stimuli appearing near the beginning of one series would appear near the end of the next series, thereby controlling for any increases or decreases in mosquito responsiveness to the visual stimuli over an experimental trial. Similar to previous wind tunnel experiments (Alonso San Alberto et al., 2022), these experimental series included a clean air only period before and after the main portion of the series (Fig. 3; Figs S3 and S4). Runs, and all of their component trajectories, were excluded from analysis based on three independent criteria: (1) strong responses occurred to visual stimuli during these clean air-only periods (suggesting odor contamination), (2) visual responses were generally low throughout the experiment even during odor exposure (i.e. problems with CO₂ release or experimental animals) or (3) if visual responses were much lower in one half of the series than the other (suggesting the run was falling outside the sunset activity peak).

A 3D real-time tracking system (van Breugel et al., 2015; Stowers et al., 2017) was used to track the mosquitos' trajectories. As described above, for each experimental trial we released a group of 50 mosquitos as this number was found to be the best compromise between minimizing interaction between individuals while still providing ample opportunities to capture mosquito responses to visual stimuli (Alonso San Alberto et al., 2022). This tracking system was unable to maintain mosquito identities for extended periods of time; however, this non-independence was accounted for using mixed statistical models (see below). Previous studies have showed similar behavior between groups and singular mosquitos when responding to stimuli within this wind tunnel (Alonso San Alberto et al., 2022). Analyses were restricted to trajectories that were at least 90 frames (1.5 s) long. Occupancy maps (Fig. 1F) were calculated by taking the number of mosquito occurrences within each 0.3 cm² square of the wind tunnel and dividing the number by the total number of occurrences in all squares, yielding a percentage of residency. To evaluate the behavioral responses to the visual stimuli, three different metrics were analyzed: (1) the number of mosquito trajectories, (2) the proportion of trajectories responding to stimuli and (3) preference indices. To assess the propensity of mosquitos to fly during a particular set of stimuli, we compared the number of recorded trajectories among stimuli, henceforth referred to as activation. We characterized mosquito preference and recruitment by trajectory passage through a pair of fictive cylinders centered on each visual stimuli (diameter: 14 cm, height: 4 cm). A sensitivity analysis demonstrated that this volume best captured the mosquitos responding to visual stimuli (Alonso San Alberto et al., 2022). To examine the relative number of mosquitos approaching the visual stimuli, henceforth referred to as recruitment, we compared the proportion of trajectories entering either the test or control volumes. Both of these analyses were done at the stimulus level; however, we evaluated preference for each trajectory (Fig. 1E) with a preference index. This index was used to evaluate and compare the preference for the test stimulus over the control stimulus irrespective of the strength of the mosquito recruitment. The index was defined as the amount of time a mosquito trajectory spent in the test stimulus volume minus the amount of time spent in the control stimulus volume, divided by the total time it spent in both volumes. During analysis (see below), overall preference indices were then estimated for each stimulus, weighting the contributions of trajectories by the total time spent in the stimuli volumes.

As CO₂ is known to be a critical cue sensitizing response to visual cues (van Breugel et al., 2015; Carnaghi et al., 2021), we first wanted to confirm its effects in our bioassays. Across all experiments, in the absence of CO₂ we saw relatively low levels of activation (Fig. 2A; Figs S2B and S5A), with mosquitos showing a reduced propensity for flight as measured by the number of trajectories recorded within a given stimulus period. Less than 1% of mosquito trajectories were recruited to either the test or control stimulus when CO₂ was absent, with most runs seeing no trajectories investigating the visual stimuli (Figs 2B and 3; Figs S2B, S3, S4 and S5B). Recruitment was low even when highly attractive low-intensity (i.e. black) visual cues were used. Recruitment rates without CO₂ were so low relative to CO₂ rates that it made estimating preference impractical. In the presence of CO₂, recruitment rates increased 7- to 17-fold depending on the stimulus (Fig. 3). The increase in recruitment was lowest when both stimuli were gray (1.0) with an intensity closely matching the white fabric background. The increase was much greater for stimuli with a greater contrast with background. We did observe statistically significant declines in both activation (post hoc contrast with Šidák adjustment, z=−13.61, P<0.0001) and recruitment (post hoc contrast with Šidák adjustment, z=−10.14, P=0.015) over the course of the spectral sweep experiment as evidenced by the difference between the initial and final response to an unilluminated black tulle target, despite a potentially small confounding effect of differing control stimuli. However, the effect of these declines was accounted for in experiments by alternating the order of stimulus presentation. We also observed that while recruitment did decline sharply after the CO₂ period ended, it did not fully return to the rate seen before the CO₂ release, with the post-CO₂ recruitment being approximately double that occurring pre-release.

As our past work did not examine how differences in CO₂ concentrations in an odor plume could affect mosquito recruitment (Alonso San Alberto et al., 2022), and the amount of CO₂ varies considerably among different resources (Geier et al., 1999; Peach et al., 2019a), we evaluated the mosquito recruitment when they were provided with a plume of 0%, 1%, 5% and 10% CO₂ (Fig. 2). We found that CO₂ concentration resulted in a statistically significant increase in both mosquito activation (Fig. 2A; likelihood-ratio test, χ²=125.17, d.f.=15, P<0.0001) and recruitment (Fig. 2B; likelihood-ratio test, χ²=305.67, d.f.=15, P<0.0001). Activation increased across all CO₂ concentrations but had a relatively high baseline as compared with recruitment (Fig. 2A). Recruitment also differed across all CO₂ concentrations, with a smaller increase between 5% and 10%. The effect of CO₂ was also smaller for gray stimuli with minimal contrast with the fabric background (Fig. 2B). Despite the strong effect of CO₂ on recruitment, we did not observe any statistically differences in preference for visual stimuli among the different CO₂ concentrations (Fig. 2C; likelihood-ratio test, χ²=12.55, d.f.=10, P=0.25). Preference in this and other experiments was defined as the difference in time spent in the test stimulus volume relative to the control volume normalized by the time spent in either volume. These results were largely consistent with the spectral and intensity preferences observed in the spectral sweep and intensity ramp experiments.

We also examined the effects of wavelength and intensity on mosquito activation and the rates of mosquito recruitment (Fig. 2; Figs S3 and S4). While stimulus wavelength (likelihood-ratio test, χ²=58.69, d.f.=64, P=0.66) and intensity (likelihood-ratio test, χ²=115.78, d.f.=105, P=0.22) did not statistically significantly affect mosquito activation, both wavelength (likelihood-ratio test, χ²=164.36, d.f.=64, P<0.0001) and intensity (likelihood-ratio test, χ²=224.79, d.f.=105, P<0.0001) had a significant effect on recruitment in the spectral sweep (Fig. S3) and intensity ramp experiments (Fig. S4), respectively. Stimuli with high intensity (Fig. S4) or stimuli at or near the green range (Fig. S3; 475–575 nm) where mosquitos have higher spectral sensitivity resulted in decreased mosquito recruitment. However, these effects on recruitment were small compared with the effects on visual preference (compare Fig. 4 with Fig. S6).

In the spectral sweep, we found that the odor of human feet and alfalfa plant extract significantly increased both mosquito activation (likelihood-ratio test, χ²=75.06, d.f.=51, P=0.0158) and recruitment to visual stimuli (Fig. S3; likelihood-ratio test, χ²=87.76, d.f.=51, P=0.0010). However, we did not observe similar increases with floral odor. In contrast with these results, in the intensity ramp experiments, we observed no increase with odor in either activation (likelihood-ratio test, χ²=87.38, d.f.=72, P=0.10) or recruitment (Fig. S4; likelihood-ratio test, χ²=81.83, d.f.=72, P=0.20). As it was operationally complicated to alternate between odors, different odors were tested at different times. Therefore, it is probable that these differences in recruitment observed during the spectral sweep experiments could be attributed to differences in responsiveness among the different cohorts of experimental animals.

To better evaluate the effect of odor on mosquito recruitment, we performed an additional experiment where we tested foot odor concurrently with and without CO₂. In this experiment, we found the addition of foot odor to the CO₂ plume as compared with the CO₂ plume alone resulted in a small though statistically significant increase in recruitment (Fig. S5B; a priori contrast, z=−2.48, P=0.0129). This increase was small relative to the increase seen in the spectral sweep and intensity ramp experiments. Additionally there was a small but statistically significant increase in recruitment with foot odor over filtered air in the absence of CO₂ (Fig. S5B; a priori contrast, z=−2.42, P=0.0153). Foot odor also decreased mosquito activation when CO₂ was present (Fig. S5A; a priori contrast, z=3.34, P=0.0008).

We also performed a similar experiment to examine the potential confounding effects of humidity produced by some of our odor sources. Despite humidity being an import host cue (Cardé and Gibson, 2010), we did not observe an increase in activation or recruitment when humidity was presented in the absence of CO₂ (Fig. S2B,C). This lack of response might be due to the low amount of humidity (Fig. S2A), or that the humidity was not co-localized with the visual cues and humidity is a short-range cue (Laursen et al., 2023). However in the presence of CO₂, humidity actually decreased both activation (a priori contrast, z=2.16, P=0.0310) and recruitment (a priori contrast, z=3.70, P=0.0002), perhaps due to more mosquitos being attracted to the odor outlet (and source of humidity) rather than to the visual stimuli. We also did not detect any statistically significant differences in spectral preference with humidity as compared with CO₂ alone (Fig. S2D).

Our prior investigation of mosquito spectral preference used paper targets with broad spectral composition and was unable to fully control for differences in intensity (Alonso San Alberto et al., 2022). Refining those earlier methods, we used narrowband isoquantal stimuli generated by a pair of LED synthesizers. Unlike the other experiments, these spectral sweep experiments had stimuli spanning a wide spectral range, allowing us to better characterize spectral preference. We also found that preference measurements were more consistent across cohorts as compared with recruitment (compare Fig. 4 and Fig. S6 with Figs S3 and S4)

In the presence of CO₂ alone, mosquitos exhibited a preference for shorter wavelengths (390–420 nm), with the preference declining gradually as the stimulus wavelength approached the green range (475–575 nm), before increasing and subsequently plateauing above 600 nm (Fig. 4A). Despite a decreased preference for stimuli in the green range (preference index of ∼0.15), all the examined colored stimuli in the spectral sweep were more attractive than the moderately attractive medium gray control. Outside the green range, we saw preference indices approaching 0.50, similar to the attraction we observed for the highly attractive unlit black tulle target (Fig. 6A).

In contrast to our previous work that only tested CO₂, here we conducted spectral sweep experiments with different olfactory stimuli. We found that odors other than CO₂ resulted in statistically significant differences in the spectral preference of Ae. aegypti (Figs 4 and 5; likelihood-ratio test, χ²=93.62, d.f.=51, P=0.0003). Floral odors (Fig. 5A) resulted in a non-significant negative preference shift in the lower wavelength range (390–450 nm), and a positive shift in the green range (475–575 nm). The odor of a plant infusion (oviposition lure) also increased the attraction in the green range, but this shift appears to be more towards the longer green wavelengths than with floral odor (Fig. 5C). There was also no shift in preference in the short wavelength range with the plant infusion odor. In contrast to the other odors, foot odor generally increased the preference to all stimuli, with a statistically significant shift in all wavelength ranges (Fig. 5B).

As has been noted previously (van Breugel et al., 2015; Alonso San Alberto et al., 2022), the preference of mosquitos for a visual stimulus is highly dependent on its intensity. The results of the intensity ramp experiments clearly demonstrated that darker stimuli were highly preferred (Fig. 6A). We were able to model and fit these behavioral responses of mosquitos to both achromatic and color stimuli with a sigmoid model (Fig. S6A). This modeling allowed us to quantify and compare this decline among different wavelengths. We found that the effect of intensity on mosquito preference was wavelength dependent, with the sharpest declines with increased intensity seen with the 527 nm stimuli (Fig. 6B). Stimuli with wavelengths further from the green showed more shallow declines, with the 621 nm stimuli showing no appreciable decline with increased intensity. We also examined the effect of odor on the relationships, fitting a similar sigmoid model (Fig. S6B–D). Here, we saw significant flattening of these intensity relationships with the addition of foot odor (Fig. 6C,E,F). In contrast, floral odor and alfalfa extract had comparably small effects on these intensity relationships (Fig. 6C,D,E,G) and in some cases these odors even had a marginal negative effect on preference (Fig. 6D,E).

Previous behavioral investigations of vertebrate host finding in mosquitos have demonstrated that this process involves the integration of olfactory, visual, thermal and tactile cues. Color, or more specifically the spectral composition of visual cues, has also been shown to be important during host finding. This study builds on this previous work by refining previous characterization of spectral preference in the presence of CO₂, and by investigating the effects of odors beyond CO₂ on the spectral preference of Ae. aegypti. These odors suggestive of vertebrate hosts, floral resources and oviposition sites all resulted in shifts in the spectral preference of mosquitos. However, these shifts differed between odors, with floral and oviposition site odor increasing the preference for green wavelengths (475–575 nm), and vertebrate host odor increasing the preference for all wavelengths. These differential shifts suggest that spectral preference is context dependent.

Consistent with long-standing observations that mosquitos are attracted to dark (i.e. black) high-contrast visual stimuli during vertebrate host search (Kennedy, 1940; Sippell and Brown, 1953; Muir et al., 1992a; van Breugel et al., 2015), floral foraging (Peach et al., 2019b) and oviposition (Snow, 1971), we found that Ae. aegypti females showed a strong preference for similar dark stimuli (Fig. 6). We also found that, similar to previous studies (Brown, 1954; Muir et al., 1992a; Alonso San Alberto et al., 2022), mosquitos preferred longer wavelength colors (>580 nm, orange, red) over shorter wavelengths (<570 nm). However, unlike previous studies, the use of our LED-generated stimuli as compared with fabric or paper targets allowed us to disentangle the spectral content or chromaticity of the stimulus from its intensity and contrast with the background. This decoupling better reveals an inverse gaussian shaped curve of preference centered in the green with increased attraction to both blue and violet as well as red and oranges colors (Fig. 4A). Our past work (Alonso San Alberto et al., 2022) observed a preference for cyan wavelengths (470–510 nm) that was not observed in this study. This difference with our previous study may be accounted for by the narrow spectral range of the LED stimuli with an intensity more closely matching the other colors. As shown by the more pronounced declines with green wavelengths (Fig. 6B), mosquitos are maximally sensitive (at least among the tested wavelengths) to green wavelengths. This result mirrors the action spectrum shown in ovipositing Ae. aegypti (Snow, 1971), where green light had the greatest effect on oviposition preference. Our results suggest that the decreased preference to green reflects the innate preference for dark low-intensity stimuli, and does not demonstrate color discrimination.

While the responses of Ae. aegypti to red wavelengths (Fig. 6B,G) showed no evidence of sensitivity to light beyond 600 nm, this result does not preclude such sensitivity as we tested only a relatively narrow range of intensities. As the sensitivity of opsins decreases in an exponential manner as you move away from the peak sensitivity (Stavenga et al., 1993), we would expect residual sensitivity to extend well into the 600–700 nm range. At higher intensities, optomotor responses have been demonstrated in Anopheles gambiae with wavelengths >600 nm (Gibson, 1995). Female Ae. aegypti have also been shown to discriminate between green and red stimuli of matched intensities, further suggesting sensitivity to wavelengths beyond 600 nm (Alonso San Alberto et al., 2022).

This study is the first to examine how odors beyond CO₂ can gate and change the visual preferences of mosquitos. Mosquitos in different behavioral contexts are searching for different resources (vertebrate hosts, nectar sources, oviposition sites), and the visual appearance of these resources can differ greatly (Alonso San Alberto et al., 2022). It has been suggested that odor could gate responses to other resources similar to the way CO₂ gates the response to vertebrate host cues (Alonso San Alberto et al., 2022; Shannon et al., 2024). In this study, we found that odors associated with both attractive floral resources (Peach et al., 2019a) and oviposition sites (Ritchie et al., 2014) significantly increased mosquito preference for green (475–575 nm) stimuli (Fig. 5). These responses were also gated by CO₂, but we did find that concentrations similar to those possibly emitted by flowers (Peach et al., 2019a) were sufficient to activate visual responses (Fig. 2). Many flowers including the common tansy used in our experiments have a relatively high reflectance in this range (Arnold et al., 2010; Peach et al., 2019b), and an increased preference to this range of wavelengths would be expected to increase floral foraging success. The increase in sensitivity to green wavelengths for the oviposition lure is less clear but might indicate water enriched with plant matter. In contrast with the other odors, human foot odor increased visual response across the entire range of tested wavelengths (Fig. 5). This result is unsurprising as host odor synergizes with CO₂ to increase mosquito response to visual stimuli (Lacey et al., 2014; Cardé, 2015).

Coupling between the olfactory and visual systems in mosquitos has already been demonstrated, with CO₂ stimulation greatly increasing both visual tracking (Barredo et al., 2022) and physiological responses in the lobula (Vinauger et al., 2019). In Diptera, this visual neuropil responds to moving objects, seems to have a role in target detection and relays chromatic information to the central brain (Nordström and O'Carroll, 2006; Trischler et al., 2007; Aptekar et al., 2015; Lin et al., 2016). In contrast, visual stimuli do not seem to modulate responses within the antennal lobe, where olfactory information is processed in mosquito brains (Vinauger et al., 2019). Given this established pathway between these two processing centers, it seems likely that other odors could similarly modulate mosquito visual responses. It has been hypothesized that the low acuity of mosquito vision might preclude object identification, leaving odor as the primary source of information (Vinauger et al., 2019).

From the observed spectral preferences (Fig. 4A), it seems likely that in the presence of CO₂ alone, Ae. aegypti visual response is dominated by achromatic contrast cues, with the level of contrast being variable across the mosquito visual range as a result of differences in spectral sensitivity. The observed spectral preferences are inversely proportional to the electroretinogram-determined spectral sensitivity of the Ae. aegypti compound eye, with a primary sensitivity peak in the green at ∼525 nm and a secondary peak in the UV at ∼350 nm (Muir et al., 1992b). Like in other Dipterans, the R1–6 outer photoreceptors are the most numerous type in mosquitos, and these photoreceptors dominate electroretinogram responses (Minke et al., 1975). While these photoreceptors express a green-sensitive opsin (Hu et al., 2012), it seems likely that they also express a UV-sensitizing pigment similar to the outer photoreceptors in Brachyceran flies, giving them UV sensitivity (Stavenga et al., 2017). Given this broadened sensitivity, the visual response of mosquitos to targets could be explained through the responses of R1–6 photoreceptors alone. However, it seems that a subset of the R7 central photoreceptors may be sufficient if the function of the outer photoreceptors is disrupted (Zhan et al., 2021).

While the preference of mosquitos with CO₂ alone seems dominated by the responses of R1–6 photoreceptors, the shifts in spectral preference with other odors suggest that this general preference for dark contrasting stimuli is modified by input from the central photoreceptors. The opsin expression in Ae. aegypti central photoreceptors shows that R8 cells throughout most of the eye express a green-sensitive opsin (Hu et al., 2012), with R7 cells co-expressing either a green- and blue-sensitive opsin in the dorsal and ventral stripe regions or a green- and UV-sensitive opsin in the remainder of the eye (Hu et al., 2009, 2011, 2014). In Culex pipiens, electroretinogram with chromatic adaptation suggests the presence of photoreceptors primarily sensitive to green and UV wavelengths (Peach et al., 2019b). Taken together, this suggests that mosquitos could discriminate by comparing the inputs of R7 and R8 cells in an opponent manner similar to that seen in Drosophila (Schnaitmann et al., 2020). This opponency could allow green wavelengths to have a positive input on spectral preference that modifies but does not negate the preference for dark objects. Such a positive input could explain the increased preference for green wavelengths seen in the presence of odor (Fig. 5). Our results suggest that this input is gated by odor and not a general aspect of the mosquito visual response.

Despite the observed shifts in preference in the green range (475–575 nm) during the spectral sweep experiments (Fig. 5), we did not observe a flattening of the intensity preference relationship with these wavelengths in the intensity ramp experiments (Fig. 6E). It may be that the strong general preference for dark objects is masking the effect in these experiments, and this spectral discrimination could be more apparent when mosquitos are presented with a choice between two differently colored quasi-monochromatic stimuli rather than a choice between one such stimulus and an achromatic control. It is also possible that the size of this spectral discrimination could be modulated by physiological state. For example, younger mosquitos are more likely to seek out floral resources, and older mosquitos are more predisposed to blood feeding (Foster and Takken, 2004; Peach et al., 2019a). We elected to eliminate physiological state as a source of variation by testing only 7 day old non-blood-fed females. We would have expected younger and gravid mosquitos to respond more strongly to the floral (Peach et al., 2019a) and plant extract oviposition lures (Ritchie et al., 2014), respectively.

Traps integrating visual and olfactory cues have been demonstrated to improve the capture of hematophagous insects such as tsetse flies and kissing bugs (Green, 1986; Reisenman et al., 2000; Torr and Vale, 2015). Visual cues are also an important aspect of mosquito trap design (Bidlingmayer, 1994; Kline, 2006); however, the interplay of vision and olfaction is not generally considered. In addition, mosquito trap design has been largely static with little change over the last 50 years and unattractive colors (white, green, blue) are often employed (Bidlingmayer, 1994; Kline, 2006). The shifts in spectral preference we observed with odor suggest that odor and visual cues in traps could be better tailored to target different species or different groups of mosquitos (i.e. gravid mosquitos, newly emerged mosquitos). There remains considerable work to be accomplished to refine mosquito trap design to make better use of both color and odor.

Building on work on olfactory gating of visual cues in mosquitos (Vinauger et al., 2019; Alonso San Alberto et al., 2022), this study demonstrated that odors other than CO₂ can shift the visual responses of Ae. aegypti. The shifts in preference to these odors suggest that different odor cues could gate visual response in a context-dependent manner, priming mosquitos to respond to differ resources in their environment. Our results further suggest that only the outer photoreceptors relating to achromatic responses are involved in responses to visual cues in the presence of CO₂; however, the addition of other odors can recruit input from the central photoreceptors. We only examined a limited set of odors; future research should examine a wider set of odors and dissect these to discover which odorants are driving the observed shifts. As the gating of visual responses by CO₂ differed among species (Alonso San Alberto et al., 2022), and mosquitos vary greatly in their ecology and host preferences, it seems likely that olfactory gating differs among mosquito species and could offer many potential avenues for study. The study of visual responses in mosquitos and their modulation by olfaction would also be aided by a deeper and more mechanistic understanding of their photoreceptors and the potential intensity, polarization (Bernáth and Meyer-Rochow, 2016) and chromatic channels available as inputs for mosquito behavior.

We would like to thank N. Salimi and K. Hanssen for assistance with behavioral bioassays, Binh Nguyen for mosquito rearing and care work, and two anonymous reviewers for their constructive comments. We are also grateful for the discussions and advice provided by G. Van Susteren, C. Ruiz, G. Tauxe, F. Chen and O. Akbari.