Detection and perception of generic host volatiles by mosquitoes: responses to CO2 constrains host-seeking behaviour

Natural selection has favoured specialization in anthropophilic mosquito host choice, yet in the absence of human hosts, females feed on a selected range of vertebrates. For host recognition, we hypothesize that mosquitoes primarily rely on generic host volatiles. Detection and perception of such compounds would provide the mosquito with a flexible, yet constrained, odour coding system that could delineate host preference. In this study, we show that the quintessential generic volatile for host-seeking, carbon dioxide, activates and attracts the malaria mosquito, Anopheles coluzzii, and the arbovirus vectors, Aedes aegypti and Culex quinquefasciatus, within boundaries set by the dynamic range and coding capacity of the CO2-sensitive olfactory receptor neurons. These boundaries are sufficiently broad to elicit behavioural responses to various hosts within their preferred host range. This study highlights the significance of the sensitivity of the carbon dioxide detection system and its regulation of host seeking and recognition.


Background
Mosquitoes that transmit infectious diseases often express a marked, inherent host preference [1][2][3][4][5][6]. Host preference studies of the African malaria vector, Anopheles gambiae sensu lato, and the arbovirus vectors, Aedes aegypti and Culex quinquefasciatus, show that natural selection favours a restricted host breadth [3]. Despite this, there remains sufficient plasticity in host preference to provide a mechanism by which mosquitoes can adapt to different environmental conditions [1,3,5], which is an important variable regulating disease transmission by predominantly anthropophilic mosquitoes [6]. This indicates that there is both a cost and benefit to maintaining plasticity [1] and/or that these species are physiologically limited in the capacity to be plastic.
Olfaction is the principal sense by which mosquitoes locate their hosts [3,7]. Host discrimination and selection is a sequence of behaviours that includes activation, long-and short-range attraction, and landing on the host [8]. Initial recognition of an upwind host relies on the detection of minute fluctuations in carbon dioxide (CO 2 ) concentration, which elicits activation and subsequent attraction in host-seeking mosquitoes [8][9][10], even in the absence of other host odours [8]. Emitted by all vertebrates, CO 2 also gates the attraction to host odours over a range of distances in host-seeking mosquitoes [11,12]. Observed interspecific variation in the behavioural response to CO 2 may be attributed to differences in the dynamic range of the CO 2 -chemosensory system for each mosquito species [9,13,14], and the underlying mechanism regulating the CO 2 dynamic range is one of sensory constraint [15,16].
We hypothesize that such limitations could be generated by a series of constraints on the sensory system used to detect and discriminate between potential host species at different distances, from activation and attraction (this study) to short-range acceptance (see companion paper: [17]). Here, we test the hypothesis that host preference correlates with the receptive range of the CO 2 -chemosensory system, and suggest that the behavioural response of anthropophilic mosquitoes to CO 2 is constrained by limits in sensory acuity.

Single sensillum recordings
The maxillary palps of Ae. aegypti, An. coluzzii and Cx. quinquefasciatus are covered with capitate peg sensilla, variously described as peg sensilla or basiconic sensilla, each housing three olfactory receptor neurons (ORNs) [16,[18][19][20]. In all species, the ORN with the largest amplitude is, by convention, referred to as the A cell, and has previously been shown to be an absolute detector of CO 2 below 1200 ppm [15,16,19,20]. Electrophysiological recordings from this neuron were made and analysed as previously described [16].
A continuous humidified stream of synthetic air (Strandmöllen AB, Ljungby, Sweden), lacking CO 2 , was passed over the maxillary palp (2 l min −1 ) via a glass tube (7 mm i.d.). Carbon dioxide was introduced into the air stream through a hole (2 mm i.d.) in the glass tube, 11 cm upstream of the maxillary palps. Delivery of CO 2 was regulated by two-way Teflon solenoid valves (Teddington, Skogås, Sweden), controlled via the digital output of an IDAC-4 (Syntech, Germany). Each valve was connected to separate gas cylinders containing metered amounts of CO 2 (150, 300, 600, 1200, 2400, 4800 ppm) and oxygen (20%), balanced by nitrogen (Strandmöllen AB). A pulsed stimulus train of CO 2 was used, with stimulation for 1 s and an interstimulus interval of 1 s.

Flight tube bioassay
Behavioural responses to pulsed CO 2 stimuli were assessed in a glass flight tube bioassay (80 × 9.5 cm i.d.), as previously described [16], with minor modifications (figure 3b). Briefly, the tube assay was illuminated from above with white light at 280 lux for the diurnal Ae. aegypti, while red light (40 lux) was used for the nocturnal An. coluzzii and Cx. quinquefasciatus. Experiments for each species were conducted during their period of peak host-seeking activity [17]. Charcoal filtered humidified air (25 ± 2°C, RH 65 ± 2%) flowed through the flight tube at 30 cm s −1 . To ensure a laminar flow and a homogeneous plume structure, the air passed through a series of stainless steel mesh screens prior to entering the flight tube (figure 3b). A pulsed flow of pure CO 2 (Strandmöllen AB), regulated by a stimulus controller (SEC-2/b, Syntech, Germany), was introduced into a pulse generator. Homogenized CO 2 pulses were delivered, using the same pulsation protocol as for the physiological experiments above, at the desired concentration (600, 1200, 2400 or 4800 ppm) through mixing pure CO 2 with pressurized air at 4.5 l min −1 rsos.royalsocietypublishing.org R. Soc. open sci.
in the pulse generator, as previously described [16]. The concentration of CO 2 in the flight tube was measured using a CO 2 analyser (LI-820, LICOR Biosciences, Lincoln, NE, USA). Controls consisting of exposing individual mosquitoes in the flight tube to non-pulsed ambient CO 2 levels (385.4 ± 6.28 ppm) that varied minimally over the duration of the experiment (1.3 ± 0.60 ppm) were run daily.
Individual mosquitoes were kept in glass release chambers (7 × 2.6 cm i.d.), covered with stainless steel mesh on one side and a cotton plug on the other, in the bioassay room for 24 h prior to the experiments [16]. The following times were measured: the time after opening the release chamber to takeoff (flight activation), the time from take-off to upwind flight directed towards the odour source (halfway, 40 cm), and the time from halfway to source contact. The maximum time recorded was 120 s. Thirty individuals of each species were observed at each concentration of CO 2 . To minimize the effect of daily variation in baseline activity and responses to odours, an equal number of test and control individuals were observed each day.

Statistical analysis
Repeated measures 2-way ANOVA, followed by a Bonferroni post hoc test was performed to compare the physiological activity among the species. The behavioural data were treated in two ways. Twoway ANOVAs, followed by a Tukey post hoc test, were used to compare the time to response among treatments and controls, as well as across species, using GraphPad Prism v.

Physiological response to carbon dioxide
Stimulation with single pulses of CO 2 with increasing concentrations elicited a dose-dependent response in the A cell of the capitate peg sensilla of all mosquito species (figure 1). Pulsed stimuli of CO 2 induced a phasic-tonic response from the A cell that remained unaltered and dependent on the stimulation. At concentrations of 600 and 1200 ppm CO 2 , the A cell of all species was able to detect and track the pulsed stimuli (figure 2). While, at higher concentrations all species were able to detect the pulses, only Ae. aegypti and An. coluzzii were able to track the stimuli, i.e. fire in response to CO 2 throughout each pulse (figure 2). The A cell of Cx. quinquefasciatus, while detecting pulse onset rapidly, adapted to each CO 2 pulse and with subsequent pulses affecting its ability to disadapt, thus limiting its capacity to track the stimuli at the higher concentrations.

Behavioural response to carbon dioxide
The behavioural responses to pulsed CO 2 stimuli differed between species. The time to activation decreased in the presence of pulsed CO 2 compared with the controls (figure 3a; upper panel). Comparing among the species, the time to activation was increased in Cx. quinquefasciatus compared to Ae. aegypti and An. coluzzii, at concentrations above 1200 ppm (2400 ppm: t-ratio −4.88, p < 0.0001; t-ratio −8.48, p < 0.0001; 4800 ppm: t-ratio −3.72, p = 0.0035; t-ratio −7.53, p < 0.001, respectively; figure 3a; upper panel). All mosquitoes of each of the three species were activated in both the presence of pulsed CO 2 and the constant ambient CO 2 control experiments ( figure 4; upper panel).
In response to pulsed CO 2 , the observed trend was a decrease in time to reach halfway along the flight tube (halfway) compared to the controls, for both Ae. aegypti and An. coluzzii as the concentration of CO 2 increased, resulting in a significant difference at 4800 ppm for Ae. aegypti ( figure 3a; middle panel)    Cx. quinquefasciatus, at concentrations exceeding 600 ppm CO 2 (1200 ppm: t-ratio −4.72, p < 0.0001; t-ratio −4.55, p = 0.0002; 2400 ppm: t-ratio −4.03, p = 0.0014; t-ratio −4.32, p = 0.0005; 4800 ppm: t-ratio −3.47, p = 0.0098, respectively; figure 3a; middle panel), with one exception. At the highest CO 2 concentration tested, An. coluzzii took as long as Cx. quinquefasciatus to reach halfway, which was significantly slower than Ae. aegypti (4800 ppm: t-ratio −3.16, p = 0.0248; figure 3a; middle panel). In general, the time to make source contact did not differ between the controls and the pulsed CO 2 stimuli for all three species halfway in the flight tube was between 93% to 100% of the tested individuals of Ae. aegypti and An. coluzzii for all concentrations tested, whereas that of Cx. quinquefasciatus declined from 100% to 70% as concentration increased ( figure 4; middle panel). Similarly, 87% to 100% of Ae. aegypti and An. coluzzii made source contact after flying upwind to pulsed CO 2 (figure 4; lower panel), whilst 83% to 43% of the tested Cx. quinquefasciatus made source contact to increased concentrations of pulsed CO 2 (figure 4; lower panel).

Discussion
Host choice by mosquitoes is, in part, regulated through senses that have been adapted to preferred hosts, and sensory constraint is a mechanism by which host breadth is regulated [7,21]. Here, we support the previous finding that detection and perception of CO 2 by the olfactory system play a vital role in the activation of host-seeking behaviour [8][9][10][11][12][13][14][15][16]. The response characteristics of the CO 2 -detecting ORNs differ, however, among mosquito species, correlating with differential behavioural outputs. The data provided here emphasize that CO 2 affords host recognition cues to mosquitoes, and that the detection and perception of CO 2 provide mosquitoes with a dynamic, yet constrained, coding system for host finding.

Constraint in detection limits the responsiveness to CO 2
The physiological and behavioural responses of the studied mosquito species to CO 2 differed. In Cx. quinquefasciatus, the limited sensory ability to continuously respond throughout the pulses of CO 2 at concentrations exceeding 1200 ppm, approximating that emitted by a large mammal [13], constrained the behavioural response, particularly activation. The concentration at which the first response was detected when stimulated with single pulses, above ambient CO 2 levels, and the slope of the dose-response curve of the CO 2 -sensitive ORN in Cx. quinquefasciatus, indicate that CO 2 sensing is more acute at lower ecologically relevant concentrations in this species. The increased sensitivity and reduced dynamic range of the CO 2 -sensitive ORNs in Cx. quinquefasciatus, compared with Ae. aegypti and An. coluzzii, when challenged with multiple pulses of high CO 2 concentrations, correlates with differences in host preference breadth. Whereas all three species feed on human hosts, they also demonstrate plasticity in feeding behaviour. The hosts of Ae. aegypti and An. coluzzii include a range of mammals, whereas Cx. quinquefasciatus shifts between humans and birds, depending on host availability [3]. It appears that the preference of Cx. quinquefasciatus for birds has exerted a selective pressure on the CO 2 -chemosensory system to activate in response to and to follow intermittent contacts with CO 2 filaments, which, because of the size of the birds, are smaller, of lower average concentration and probably rarer [22]. Having such an acute CO 2 -chemosensory system may have put restrictions on the dynamic range of the CO 2 -sensitive ORNs, as the neurons rapidly adapt, and disadapt more slowly, when stimulated with intermittent pulses of CO 2 at high (greater than or equal to 2400 ppm) concentrations. It is likely that the CO 2 -sensory machinery has become saturated, reducing the sensitivity to repetitive stimulation. Similar restrictions have been described for pheromone-responsive ORNs in moths, showing that the tracking ability of single ORNs and the behavioural response to repeated stimuli is dependent on adaptation-disadaptation mechanics (for review see [23]). The constrained response of ORNs to high CO 2 concentrations, when provided in multiple pulses, does not impede the activation and attraction of Cx. quinquefasciatus to CO 2 emitted by humans, as they readily enter houses and tents with a sleeping person, where the measured CO 2 concentration at the entrance of the tents was at or below 1500 ppm [24], which is within the dynamic range of their CO 2 -chemosensory system (this study). However, the time to activation and halfway, described in this study, increased in the presence of CO 2 at or above 1200 ppm compared to controls, emphasizing that the CO 2 -chemosensory system is constrained at elevated CO 2 concentrations over the short range. The CO 2 -sensitive ORNs of Ae. aegypti and An. coluzzii are less acute at concentrations above ambient CO 2 levels, which is in agreement with previous studies [15,19], indicating that these highly anthropophilic mosquitoes are less dependent on CO 2 alone in favour of other host cues as a basis for host selection. The dynamic range of their CO 2 -sensitive ORNs is wide, with a threshold at or below ambient concentration (350-400 ppm) and the neurons do not reach their maximal response at the highest dose tested (4800 ppm). Behavioural analysis also shows that both species are activated and attracted to the full concentration range of CO 2 tested, which is in line with previous reports [14,25]. Thus, the CO 2 -chemosensory systems of Ae. aegypti and An. coluzzii are equipped to detect a wide range of CO 2 emissions, from rare and intermittent CO 2 signals at a distance of several metres from a potential host [26][27][28] to amounts equivalent to that emitted by a human or other large mammalian host [27]. In conclusion, host selection by Cx. quinquefasciatus is constrained by the dynamic range of their CO 2 -sensitive ORNs dictating the behavioural response, in particular activation, of this species. This may define the breadth of intrinsic preference and allow for behavioural plasticity within these limits.

Conclusion
While the role of CO 2 in activating and attracting mosquitoes to potential hosts is well characterized [11,12], this study highlights the importance of CO 2 , within natural release rates, in regulating host seeking and recognition. The cross-species comparison revealed the importance of analysing the response properties and tuning of CO 2 -sensitive ORNs together with how this may affect the behavioural output. From a vector control perspective, this is essential when developing lures for optimal attraction of specific mosquito species in efforts to control and monitor populations.
Data accessibility. All data that are relevant to the study are reported within the article.