JEB desktop wallpaper calendar 2016

JEB desktop wallpaper calendar 2016

Moment-to-moment flight manoeuvres of the female yellow fever mosquito (Aedes aegypti L.) in response to plumes of carbon dioxide and human skin odour
Teun Dekker, Ring T. Cardé


Odours are crucial cues enabling female mosquitoes to orient to prospective hosts. However, their in-flight manoeuvres to host odours are virtually unknown. Here we analyzed in 3-D the video records of female Aedes aegypti mosquitoes flying in a wind tunnel in response to host odour plumes that differed in spatial structure and composition. Following a brief (∼0.03 s) encounter with CO2, mosquitoes surged upwind and, in the absence of further encounters, counterturned without displacing upwind. These patterns resemble moth responses to encounter and loss of a filament of pheromone. Moreover, CO2 encounters induced a highly regular pattern of counterturning across the windline in the horizontal (crosswind) and vertical planes, causing the mosquito to transect repeatedly the area where CO2 was previously detected. However, despite the rapid changes across all three axes following an encounter with CO2, the angular velocities remained remarkably constant. This suggests that during these CO2-induced surges mosquitoes stabilize flight through sensors, such as the halteres and Johnston organs, sensitive to Coriolis forces. In contrast to the instantaneous responses of the mosquito CO2, a brief encounter with a filament of human skin odour did not induce a consistent change in mosquito flight. These differential responses were reflected in further experiments with broad plumes. A broad homogeneous plume of skin odour induced rapid upwind flight and source finding, whereas a broad filamentous plume of skin odour lowered activation rates, kinetic responses and source finding compared with homogeneous plumes. Apparently, yellow fever mosquitoes need longer continuous exposure to complex skin-odour blends to induce activation and source finding.


Mosquitoes are renowned for their ability to locate their hosts using odours. In search of novel mosquito and vector intervention techniques, much research on mosquito olfaction has targeted the identification of the suite of odours and the sensory system for their detection and processing (Ghaninia et al., 2008; Syed and Leal, 2009; Carey et al., 2010). This task is formidable, considering the number and structural diversity of compounds emanating from humans alone (e.g. Krotoszynski et al., 1977; Cork and Park, 1996; Bernier et al., 2001). Although there is progress in deciphering how a mosquito processes host odours (Carey et al., 2010; Okumu et al., 2010a), and how this is affected by host preference (e.g. Dekker and Takken, 1998; Dekker et al., 2001a; Dekker et al., 2002), our understanding of the mechanisms mosquitoes use to orient upwind along the plume of host odours remains poorly understood (reviewed by Cardé and Gibson, 2010).

Our current view of how flying insects use odours in source location is primarily based on decades of experiments with male moths orienting to female pheromone and more recent work on Drosophila fruit flies. The first detailed study on insect flight responses to odour, however, was Kennedy's (Kennedy, 1940) landmark wind-tunnel study with the yellow fever mosquito Aedes aegypti (L.). Kennedy demonstrated that the spatial distribution of odours does not ‘guide’ the insect along the plume; instead, odours induce upwind flight, the progress of which is gauged by visual feedback. This process is termed odour-mediated optomotor anemotaxis. Following the identification of many moth pheromones beginning in the 1960s, male moths became a principal system for studying odour-mediated flight of insects, partly because of the reliable orientation response of male moths to female pheromone. In the moth-centric model, male moths respond rapidly to encountering a filament of pheromone by surging upwind. If the moth fails to intercept another filament within several hundred milliseconds, upwind progress gradually slows, and the track assumes a zigzag form. Eventually upwind displacement ceases, causing the moth to cast (zigzag without making upwind progress) (Baker, 1990; Mafra-Neto and Cardé, 1994; Vickers and Baker, 1994) (reviewed by Cardé and Willis, 2008). Reiterative encounters at a frequency above 5 Hz can induce nearly straight upwind flights (Mafra-Neto and Cardé, 1994; Vickers and Baker, 1994). The temporal acuity of the male moth olfactory circuitry is high, resolving successive encounters of pheromones and antagonists separated by only a few milliseconds (Baker et al., 1998; Vickers et al., 2001).

The moth–pheromone system, however, may not typify orientation of all insects flying along odour plumes, or orientation to all kinds of odour sources. Moth pheromone is released from a small point source, and males are under strong selective pressures to respond to and locate females quickly. The pheromone is comprised of relatively few components (typically one to four), and these blends are thought to be under stabilizing selection for fidelity of composition. The presence and ratios of odours from most other sources are not so constrained for constancy, and these blends may be released from different locations, so that the plume itself can be spatially variable. These differences in plume dynamics and composition may be reflected in differences in in-flight responses to odours. Although recent studies on the orientation of Drosophila melanogaster to vinegar odour (Budick and Dickinson, 2006) highlighted similarities in orientation mechanisms between moths and flies, there can be differences in flight responses to defined odour plumes (Baker and Vickers, 1997; Justus and Cardé, 2002). Reactions to differing plume structures also may also underlie differences in capture efficiency that are contingent on trap design (e.g. Macaulay and Lewis, 1977; Webster et al., 1986).

Differential orientation mechanisms may reflect limitations or species-specific characteristics of their sensory systems, particularities in their life history or habitat, and/or characteristics of the odour sources to which they orient. Identification of ‘attractants’ should, therefore, be accompanied by studies of the moment-to-moment response and trap entry requirements of insects to odours and visual cues. This is especially important in the wake of the discovery of the receptor proteins (olfactory and ionotropic receptors) in many insect species, which has accelerated the identification of insect behaviour-modifying substances (Vosshall et al., 1999; Hallem and Carlson, 2006; Benton et al., 2009; Carey et al., 2010).

Few studies have detailed odour-mediated flight in other insects besides moths and fruit flies, and accordingly our knowledge of differences among insects is comparatively sparse. Mosquitoes are in particular need of research to determine factors important in their orientation to host odours and capture in odour-baited traps. Such traps could be pivotal in future intervention efforts, either in monitoring of intervention strategies or directly in control (Okumu et al., 2010a). Although odours are a principal cue used by mosquitoes in host finding, there are many factors affecting host orientation that are not directly related to the olfactory stimulus itself or its ‘attractiveness’. These factors can be as important in capturing mosquitoes as the ‘correct’ odour-blend composition. In a laboratory study, odour-port entry of Aedes aegypti and Anopheles gambiae s.s. in response to CO2, skin odour and their combination was differentially influenced by odour plume structure (Dekker et al., 2001b). Factors at the core of this difference likely also affected the up to fourfold difference in trapping efficiency of several models of commonly used mosquito traps (Cooperband and Cardé, 2006a; Cooperband and Cardé, 2006b). In another study, approach of mosquitoes to odour-baited entry traps (OBETs) was monitored with electric nets and compared with actual capture. Using a human as an odour source, OBETS caught the more endophilic Anopheles arabiensis much more efficiently than the exophilic Anopheles quadriannulatus, demonstrating that other factors, either related to plume structure and/or visual responses, affect orientation and ultimately trap capture, and in a species-specific manner (Torr et al., 2008).

In the present study we revisited odour orientation in the same mosquito species that Kennedy used in his pioneering experiments 70 years ago. We analyzed in three dimensions (3-D) the manoeuvres of A. aegypti to host odour plumes that differed in composition and structure. We used thin ribbon-like odour plumes to analyse the response to brief encounters with odour (using a ribbon plume) (see Mafra-Neto and Cardé, 1994), and a broad host-odour plume to analyse the response to entering and exiting broad plumes. Alignment of flight track sections with respect to single plume encounters and entering or leaving a broad plume permitted a moment-to-moment analysis of the flight tracks of A. aegypti in response to odours. We demonstrate here that CO2 and skin odour have differential effects on A. aegypti's activation rate, flight manoeuvres and ability to locate the odour source.



We used the Rockefeller strain of A. aegypti. Mosquitoes were reared at 80% relative humidity under a 14 h:10 h light:dark cycle. The 14 h ‘day’ included a 1 h artificial dusk period. Adults were kept in 30×30×30 cm gauze cages (Bugdorm-I, BioQuip, Gardena, CA, USA) and provided with a 6% glucose solution. Larvae were reared on Tetramin® fish food (Melle, Germany). We tested 10–20 day-old non-blood-fed, mated female mosquitoes that had not had prior exposure to host odours in a bioassay. Mosquitoes were transferred to release cages 12 h before testing. They had no exposure to host odours from 12 h prior to the onset of the experiments. Each release cage contained four female mosquitoes. Experiments were conducted during the first 5 h of photophase.

Experimental setup and testing procedures

We tested the flight of mosquitoes in a laminar flow wind tunnel (Fig. 1) [see Dekker et al. (Dekker et al., 2005) for details]. Before the start of an experiment, a release cage containing experimental mosquitoes was placed on the release platform on the wind-tunnel floor 130 cm downwind from the upwind screen. The opening of the cage faced downwind and was placed against a screen, which prevented mosquitoes from departing before the start of the experiment. After 3 min the platform was lifted and turned slowly 180 deg upwind. The final position of the release cage was such that the ribbon odour plume intercepted the centre of the cage. We tested mosquitoes for 3 min and recorded their behaviour with two Sanyo® VCB 3512T cameras (Chatsworth, CA, USA) set at a 1/100 s shutter speed and equipped with 6 mm lenses, one from the side and one from the below the wind tunnel. The camera views were synchronized with an Event & Video Control Unit (Peak Performance Technologies Inc., Centennial, CO, USA), overlaid and recorded on one tape with a Sony® EVO-550H Hi-8 tape recorder. The 3-D flight coordinates of the flight tracks were obtained with Motus (Peak Performance Technologies Inc.) at 30 Hz and salient flight parameters were calculated (see Data analysis).


We tested CO2 and odour from human skin at various concentrations. Test concentrations of CO2 (0.05, 0.2, 0.8, 4% and pure) were obtained by using flowmeters (Cole-Parmer, Vernon Hills, IL, USA) to mix either 100% or 4% CO2 with clean air, all from pressurized cylinders. Measurements with a photoionization detector (mini-PID, Aurora Scientific Inc., Aurora, ON, Canada) mixing propylene (a tracer gas) and air verified a homogeneous mixing before the odour-laden air entered the plume-generating device [see Justus et al. (Justus et al., 2002) for details of the PID method].

Skin odour was obtained by inserting a human arm (T.D.) in a 10 cm diameter glass tube containing a flow of 3 l min–1 (ribbon plume, turbulent plume, dilutions of skin odour) or 30 l min–1 (100% skin odour, homogeneous plume) clean air from a pressurized cylinder (see Dekker et al., 2005). A stainless steel fan at one end of the tube created a counter flow of approximately 400 l min–1 over the arm, which ensured high uptake and homogenization of the skin odour. ‘Hand’ odour was obtained by inserting the experimenter's hand into the plume-generating device through a tube (Dekker et al., 2005). Three hours before the experiments the arm was rinsed with tap water for 1 min. To minimise possible odour adsorption, the tube leading to the plume generators was only 15 cm long and made of Teflon®. Reduced concentrations of skin odour were obtained by splitting the skin-odour-laden air (verified using bubble flow meters, which have a negligible resistance). To create a diluted turbulent plume, the effluent was supplemented with clean air to a total of 3 l min–1. As a control in the homogeneous skin-odour experiments (see Broad plumes), we also created a homogeneous skin-odour plume by inserting a hand (T.D.) in the plume generator upwind of the laminising screens (Dekker et al., 2005). Except for the hand used in the skin-odour tube, we used Fisherbrand® (Pittsburgh, PA, USA) latex examination gloves during the experiments to avoid contamination with any experimental device. The screens were replaced whenever the concentration of skin odour of a new experiment was higher than in the previous experiment. Screens were washed thoroughly with water and soap at the end of each experimental day.

Fig. 1.

(A) Wind tunnel setup and (B–D) plume generators used. The turbulent plume generator (B) consisted of a 14 cm diameter glass ring with sixteen 1 mm holes on the inner side equidistant from each other, through which the odour exited into the flight chamber. Odour was pumped into the plume generator at at rate of 3 l min–1. The homogeneous plume generator (C,D) was placed behind the stainless steel laminising screens. It had two inlets, one for the odour and the other for a clean air ‘jet’ to homogenize the mixture. Alternatively, we inserted a hand (treatment: hand) from the side of the wind tunnel through a tube (D), upwind from the laminising screens and the Honeycomb® laminiser, into the homogeneous plume generator. (E) Sample propylene density plots in the turbulent plume measured using a mini-photoionization detector after subtraction of the background values. Five random measurements inside and one at the edge of the 15 cm diameter plume at 50 cm downwind reflect the fluctuating signal generated by the plume-generating device. Sampling rate, 100 Hz.

Ribbon plumes

We used ribbon plumes (a thin, pencil-like odour stream) (see Mafra-Neto and Cardé, 1994; Mafra-Neto and Cardé, 1995) to determine the response to a single brief encounter with odour. A pipette was inserted in the flight chamber 5 cm from the upwind screen. A 20 ml min–1 flow rate through the pipette ensured that flow from the pipette entered the main air stream in the wind tunnel isotropically. A skin-odour plume was created by splitting the 3 l min–1 skin-odour-laden air (see above) into a fraction of 20 ml min–1 just before it entered the wind tunnel. We used Teflon® tubing and the distance to the pipette (15 cm) was kept as short as possible to minimize adsorption. The structure of the plume was checked daily with TiCl4 ‘smoke’.

The laminar flow through the wind tunnel allowed for accurate measurement of the diameter and the precise position of the plume. The plume was visualized using TiCl4 ‘smoke’ and video records were used to verify its dimensions and position. The plume had a diameter of approximately 0.5 cm, which implies that a mosquito flying crosswind through the plume at a flight speed of 30 cm s–1 would be in contact with odour for approximately 0.03 s. The plume was 25 cm from the side and 18 cm above the wind tunnel floor.

Broad plumes

We used two kinds of broad plumes to establish the effect of plume structure on flight behaviour.

Turbulent plume

We created a turbulent plume by pushing the odour-laden air through a turbulent plume generator consisting of a 14 cm diameter glass ring. The glass tubing was 5 mm in diameter and had sixteen 1 mm holes equidistantly spaced on the inside of the glass ring. The plume generator was placed 5 cm from the upwind end of the flight chamber (Fig. 1B). For concentrations other than 4% and 100%, we mixed CO2 with medical air to obtain the desired concentration. The plume structure of the turbulent plume was analyzed by simulating the plume structure with air containing propylene as chemical tracer gas. A mini-PID, placed in the flight chamber at different points from the source along the upwind, lateral and vertical axes, measured the propylene concentration of the mixture at 100 Hz. The resulting propylene concentration plots (Dekker et al., 2005) (see Fig. 1E for the broad turbulent plume) demonstrate that the plume was turbulent, i.e. high and low concentration peaks were interspersed with clean air.

Homogeneous plume

We created a homogeneous plume by pushing the odour-laden air into a plume generator placed behind the two stainless steel fine-mesh screens (Fig. 1C,D). To ensure a homogeneous mixture, clean air at 4 l min–1 was introduced via a Pasteur pipette into the plume generator, just downstream of the point where the desired odour was introduced. To obtain the desired concentration of CO2, we adjusted the flow rate of 100% or 4% CO2, assuming a background concentration of 0.035% and a flow through the plume generator of 360 l min–1. The homogeneity of the plume was verified using air containing propylene as the chemical tracer gas in conjunction with the mini-PID. We analyzed the plume structure at different points from the source along the upwind, lateral and vertical axes at a sampling rate of 100 Hz. The resulting propylene concentrations plots (Dekker et al., 2005) demonstrate that the plume was homogenous in the centre and slightly turbulent along its outer envelope.

The laminar flow through the wind tunnel allowed for accurate estimation of the diameter and the position of the turbulent and homogeneous plumes. The plumes were cylindrical and had a diameter of approximately 14 cm. The centre of the plume was 25 cm from the side and 17 cm above the wind tunnel floor. By turning the release cage at the start of the experiment, the cage was centred in the plume. Mosquitoes exiting the cage once it was turned were in contact with the plume.


Ribbon plume experiments

In the ribbon plume experiments we analysed the response of mosquitoes to a single brief encounter with host odour. Two series were conducted. N-values (see also under analysis) are given in brackets and reflect the number of mosquitoes analysed for source finding and track analysis, respectively.

In the skin-odour series, we compared the modulation of flight by brief encounters (i.e. crossing the ribbon plume) of skin odour (N=244, 30) and CO2 (N=240, 86). Clean air ribbon plumes served as control (N=204, 30).

In the CO2 series, we compared the modulation of flight by brief encounters of CO2 at different concentrations: 0.05% (N=64, 13), 0.2% (N=64, 17), 0.8% (N=64, 22), 4.0% (N=64, 33) and 100% (N=64, 23).

Because mosquitoes rarely landed on the tip of the pipette, source finding in the ribbon plume experiments was defined as arriving within a 10 cm perimeter around the tip of the pipette.

Broad plume experiments

In the broad plume experiments we analysed the effect of plume structure and odour concentration on activation, upwind flight and source finding by A. aegypti. The following series were performed. N-values (number of mosquitoes) refer to total tested, used for activation, source finding and track analysis, respectively.

The homogeneous skin-odour series consisted of five treatments: clean air (N=68, 59, 31, 17), 100% skin odour (N=72, 64, 55, 36), 20% (N=71, 67, 30, 28), hand (N=71, 64, 57, 49) and 0.4% CO2 (72, 59, 58, 37).

The homogeneous CO2 series consisted of five treatments: 0.05% (N=48, 45, 43, 25), 0.1% (N=56, 55, 55, 30), 0.4% (N=53, 40, 48, 26), 1% CO2 (N=48, 40, 43, 31) and hand (N=48, 46, 43, 23). Higher concentrations of CO2 could not be tested, because at higher concentrations the expansive cooling of CO2 made the plume's position unreliable.

The turbulent skin-odour series consisted of four treatments: 100% skin odour (N=70, 60, 55, 28), 20% skin odour (N=72, 60, 27, 17), 4% CO2 (N=72, 59, 59, 36) and clean air (N=70, 60, 48, 23).

The turbulent CO2 series consisted of six treatments: 0.05% (N=56, 49, 24, 20), 0.2% (N=55, 54, 39, 24), 0.8% (N=56, 54, 51, 31), 4.0% (N=54, 50, 49, 30), 100% CO2 (56, 53, 53, 30) and clean air (N=55, 50, 20, n/a).

Data analysis

The 3-D flight coordinates of the flight tracks were obtained with Motus (Peak Performance Technologies, Inc.) at 30 Hz. Some flight tracks were excluded from analysis for two reasons. First, flying mosquitoes disrupt the plume, which makes the relationship between plume contact and flight manoeuvres of those mosquitoes downwind tenuous. This happened frequently as we placed four mosquitoes in each release cage and usually several mosquitoes were flying simultaneously. Therefore, tracks from those mosquitoes that clearly flew in the wake of another mosquito were excluded. Second, a few tracks were excluded when frame dropout rates (caused by inadequate visual resolution for tracking) were high. Slight digitising errors (inaccuracies) were corrected for by smoothing the data with the cubic spline algorithm, a method that is particularly well suited for data that are parabolic in nature (Jackson, 1979). The behavioural parameters were calculated with several custom-made programs created in Visual Basic®. Table 1 lists the parameters used for flight track analysis. Fig. 2A,B illustrates the triangle of velocities used in flight track analysis, whereas Fig. 2C shows an example 3-D flight track, its XY, XZ and YZ projections, as well as the displacement over time in X, Y and Z of a mosquito flying in response to a 4% CO2 plume. Simultaneously, TiCl4 ‘smoke’ was used to visualize the precise position of the mosquito with respect to the plume. Note the regular track reversals in the plots of Y and Z over time.

Table 1.

Flight track parameters

Fig. 2.

(A) Illustration of the triangle of velocities modified from Marsh et al. (Marsh et al., 1978). α, track angle; β, course angle. (B) Terms used for Coriolis forces (rotational torques) in different planes. (C) Example track of a mosquito flying in response to a 4% CO2 plume. The left panel shows the projections in XY, XZ and YZ, whereas the right panel shows the displacement across all three axes over time. The red circle in the YZ plot indicates the position of the ribbon plume.

Tracks were analysed and aligned with respect to contacting the ribbon plume. The precise position of the plume was verified using TiCl4 ‘smoke’ and the mini-PID (details see above). Flight parameters over 0.1 s (three frames) intervals with respect to such odour encounters were averaged within flight tracks.

The data were log transformed, checked for normality and day effects, and analysed in Statistica (StatSoft, Inc., Tulsa, OK, USA) with a repeated-measures ANOVA, followed by a Fisher's least significant difference (LSD) post hoc test to determine differences between means of treatments. Contrasts were used to test for significant changes in a parameter within a treatment after plume contact (repeated measures).

Fig. 3.

Percentage source finding in the ribbon-plume series (±s.e.m.). (A) Skin-odour series; (B) CO2 series. Source finding is defined as reaching a 10 cm diameter sphere (area) around the Pasteur pipette. Significant differences between treatments are denoted by different letters above the bars. N-values are indicated on the bars. (C) Sample tracks to a ribbon plume of 4% CO2. The light ribbon in the centre signifies the plume, and red circles the plume contacts. Green circles show the position of the mosquito at 0.1 s intervals.

We created a simulated flight track using the mean X, Y and Z displacement of the mosquito. To calculate such tracks, the program alternated the sign of the Y and Z displacement at the mean track reversal (zigzag) frequency. The resulting simulated flight track is analogous to the filament-response model's ‘building blocks’ of moth response to pheromone (see Introduction).

A Weibull distribution was used to describe the activation rate of A. aegypti (Crawley, 1993). A shape parameter, α, allows the take-off rate (‘hazard’) to increase (α>1) or decrease (α<1) over experimental time, starting with a constant rate (exponential distribution, α=1) (Aitkin et al., 1989; Crawley, 1993). We used a censoring indicator for mosquitoes that did not take off within the experiment, allowing for ‘non-responders’ to contribute to the survivorship function. Differences between the survivorship curves were assessed with χ2-values.

The percentage of mosquitoes that left the release cage and reached the source was arcsine square-root transformed and statistically analysed with a two-way ANOVA, followed by an LSD post hoc test to determine significant differences between the means.


Response to encountering an odour filament

Contacting a ribbon plume of CO2 induced a rapid upwind surge (reduced track angle and increased flight speed; see sample tracks, Fig. 3C), which brief encounters of skin odour failed to induce. This difference also was reflected in the source finding rate (Fig. 3A,B). Although source finding with skin odour encounters did not differ from the clean air control, it was positively correlated with CO2 concentration and reached a plateau at 4% CO2 (Fig. 3B).

Tracks were analyzed and aligned with respect to contacting the ribbon plume using a computer algorithm. This procedure brought out a pronounced difference in the moment-to-moment responses induced by CO2 and skin odour (Figs 46). Note that these figures do not represent flight tracks, but mean flight parameters over time. Contacting a ribbon plume of air or skin odour did not have an effect on any track parameters, whereas a single brief contact with 4% CO2 induced rapid upwind turning (reduced track angle; Fig. 4A) for 0.5 s, after which mosquitoes slowly reverted to more crosswind headings. Casting ensued within 2 s after contact with a CO2 filament (i.e. the last 0.5 s of the plots shown in Figs 46).

During upwind surges mosquitoes reduced their vertical and, to a lesser extent, crosswind displacement (Fig. 5A), while increasing their muscle-generated torque (increase in upwind displacement and airspeed; Fig. 5A, Fig. 6A). The groundspeed and 3-D flight speed increased little, yet significantly during surging (positive orthokinetic response, Fig. 6A). The surge-cast response was dependent on CO2 concentration (Fig. 4B, Fig. 5B). Although the upwind speed and duration of the surge was similar across all concentrations, at the lower CO2 concentrations mosquitoes redirected their headings towards the crosswind more slowly (Fig. 4B) and less consistently (larger envelope in the figure). This is also apparent in Fig. 5B, which demonstrates that at lower concentrations the deceleration of upwind progress was lower and more variable, as well as in Fig. 6B, where both the increase and decrease in airspeed was less pronounced and more variable at lower CO2 concentrations. Casting was rarely observed at the two lowest concentrations, 0.05% and 0.2% (Fig. 4B).

We also calculated track reversal (zigzag) frequency in the Y (crosswind) and Z (vertical) planes following plume contact. These demonstrate that CO2 induced mosquitoes to zigzag across the windline in a highly regular fashion, the frequency of which slowly decreased with time elapsed after plume contact (Fig. 7A). More interestingly, CO2 induced a significant increase in the frequency of vertical track reversals at approximately 1.6- to 1.9-fold that of the crosswind zigzag frequency, which remained elevated for many seconds after plume contact. Mosquitoes displayed a tightly arranged frequency of track reversals in Y and Z planes at all CO2 concentrations, although the frequency increased with concentration (P<0.05 for 0.05% compared with 4% and 100%; Fig. 7B). From the flight track parameters we simulated a flight track following contact with a 4% CO2 filament (Fig. 7C). The simulation represents an average mosquito's response to contacting a CO2 filament. Each dot represents 1/30th of a second. The figure illustrates the mosquito's upwind surge (increased upwind projection), its kinetic response (increased spacing of the dots), as well as the reversal into casting over time. It also illustrates how the mosquito, through its manoeuvres, effectively transects in YZ the area where the plume was previously sensed. Skin odour and clean air did not induce such a regular Y and Z pattern of track reversals (Fig. 7A,B).

Fig. 4.

Flight angles of mosquitoes in the ribbon-plume series. (A) The change in track, 3-D and course angles over time after contacting a ribbon plume of clean air (control), skin odour or 4% CO2; (B) track angles of mosquitoes after contacting the ribbon CO2 plume. Grey shading denotes the standard error of the mean. Values (deg) in red differ significantly from those in black. N-values are indicated in brackets. Note that these graphics do not represent flight tracks, but mean track angles over time.

Next we analyzed the surge's angular velocity (degrees turned per second) and Coriolis forces (degrees turned per centimetre travelled). These parameters are informative about the centrifugal forces acting on the body of the mosquito. During the upwind surge and reversal into casting, the Coriolis forces (Fig. 8A) and angular velocity (Fig. 8B) remained remarkably constant, despite the major changes in the projection and kinetics of its flight in X, Y and Z. This was especially true for the yaw, the 3-D Coriolis forces (combined yaw, pitch and roll) and 3-D angular velocity (XYZ). In contrast, despite the little change in flight in response to skin odour and clean air, the Coriolis forces and angular velocities and their projections were more variable than with CO2.

Response to broad plumes

In subsequent experiments we analysed the response of mosquitoes to broader plumes of CO2 and skin odour. In four experimental series we tested the effect of plume structure and concentration on activation rate, source finding, flight kinetics and track angles [data on the homogeneous plume are from Dekker et al. (Dekker et al., 2005)].

The activation rate of mosquitoes is displayed in Fig. 9A,B. Activation by CO2 was slightly dependent on concentration. In the turbulent CO2 series, activation by 0.05 and 0.2% CO2 was lower than by the other concentrations. This could be due in part to the dilution of CO2 through the plume-generating device, as in the homogeneous series low CO2 concentration induced full activation rates. However, turbulent CO2 at 100% also induced a lower activation rate, as did 1% homogeneous CO2. At these levels of CO2, sensory adaptation may cause rapidly reduced activation over time. In contrast, compared with CO2, turbulent skin odour was a weak activator, but activation rates were much increased when skin-odour plumes were presented homogeneously. The percentage of non-responders was inversely related to the activation rate (Fig. 9A).

The patterns of activation and source finding were generally similar (Fig. 9A,C). Source finding was higher with CO2 than skin odour in all series. In addition, source finding was higher when skin odour was presented as a homogeneous compared with a turbulent plume. Note that because the amount of skin odour carried in the effluent stream leading to the plume generator was the same in the homogeneous and turbulent plumes, the instantaneous concentrations of filaments within the turbulent plume would be higher than the concentration of skin odour in the homogeneous plume.

Fig. 5.

X, Y and Z displacement after plume contact in ribbon-plume experiments. (A) Skin-odour series; (B) CO2 series. The standard error is displayed as an envelope around the means. Values (cm s–1) in red denote a significant difference from initial values. N-values are indicated in brackets. Note that these graphics do not represent flight tracks, but displacement over time.

Fig. 10 displays the ground, air and 3-D flight speeds of mosquitoes flying in response to plumes of different structure and content. Although plumes of different CO2 concentrations and structure induced similar flight speeds (Fig. 10A,B), both concentration and structure of skin-odour plume affected flight speeds. Whereas homogeneous plumes of 100% skin odour and ‘hand’ (Fig. 10A) induced flight speeds similar to CO2, turbulent plumes of skin odour reduced flight speeds compared with CO2 (Fig. 10B). Furthermore, flight speeds at dilutions of homogeneous skin-odour plumes were lower than undiluted. Further, it should be noted that flight speeds were relatively constant, differing by a maximum of 20%.

Next we analyzed the track angle from moment-to-moment as mosquitoes entered or exited the plume (Fig. 11). Within 0.2 s upon entering or exiting a turbulent skin odour plume, mosquitoes turned upwind or downwind, similar to CO2 (100% skin odour, Fig. 11A, top panel). Similar results were obtained with homogeneous plumes (100% skin odour and hand, Fig. 11A, lower panel). No such changes were observed with diluted skin-odour plumes. After entering a turbulent CO2 plume, mosquitoes aimed their thrust more due upwind at all concentrations of CO2. However, the latency decreased with increasing concentration, i.e. within 100 ms, except for 0.2% CO2 (within 200 ms) or 0.05% (Fig. 11B, top panel). In contrast, homogeneous plumes of CO2 induced turning upon entering or exiting the plume within 100 ms at all concentrations, although mosquitoes exiting a homogeneous 1% CO2 plume changed their track angle only after 400 ms (Fig. 11B, lower panel).


Odour-mediated orientation in insects has been described primarily using male moth responses to female pheromones, and more recently D. melanogaster to fruit odours (reviewed in Cardé and Willis, 2008) (see also Budick and Dickinson, 2006; Budick et al., 2008). Resulting models emphasize the importance of both odorant and clean air in upwind flight. Single encounters induce an upwind surge, which slowly morphs into casting unless a new filament is encountered (Baker, 1990; Mafra-Neto and Cardé, 1994; Vickers and Baker, 1994). Male moths faithfully resolve such intermittent signals at astonishingly high frequency in behavioural response (Baker et al., 1998), in peripheral detection (Bau et al., 2002), and in the central nervous system (Vickers et al., 2001). Because homogeneous clouds lack filament–clean air successions, they fail to induce the chain of repetitive surge-cast behaviours, thereby generally disrupting upwind flight and source finding (Kennedy et al., 1980; Willis and Baker, 1984) (but see also Justus and Cardé, 2002).

Fig. 6.

Ground speed, flight speed and airspeed changes following a single plume contact in ribbon-plume experiments. (A) Skin-odour series; (B) CO2 series. The standard error is displayed as an envelope around the means. Values (cm s–1) in red denote a significant difference from initial values. N-values are indicated in brackets. Note that the graphics in this figure do not represent flight tracks, but the mean speed over time.

Fig. 7.

Track reversal and simulated flight track in ribbon-plume experiments. (A) Track reversal frequency in Y and Z over time after contacting a single filament of CO2 or skin odour. The standard error is displayed as an envelope around the means. (B) Track reversal frequency after contacting a ribbon plume of CO2 at various concentrations. Asterisks indicate a significant difference between the Y and Z frequencies. (C) Simulated flight track of A. aegypti after CO2 filament contact in XY, XZ and YZ. The track is based on mean track values. The orange circle in the YZ plot indicates the position of the plume. N-values are indicated in brackets and on the bars.

In mosquitoes moment-to-moment responses in relation to plume structure have been studied only in A. aegypti (Dekker et al., 2005), and the question thus arises to what extent odour-mediated flight in other mosquitoes is similarly constrained by plume structure, although there are suggestions for such effects in previous work (Dekker et al., 2001b). Here we demonstrate that plume structure affects odour-mediated flight of A. aegypti from moment-to-moment in an odour-specific manner.

Response to filament encounters differs between odours

Following contact with a filament of CO2, mosquitoes showed pronounced orthokinetic (increased flight speed and airspeed) and anemotactic responses (turning upwind). The pattern of upwind surging and reversion to casting in response to CO2 was quite similar to the reaction of male moths encountering a single pheromone filament (Mafra-Neto and Cardé, 1994; Vickers and Baker, 1994), with an upwind surge of ca. 0.1 s, and after ca. 0.5 s gradually reverting to casting. The response of mosquitoes to CO2 fits the surge-cast model (Baker, 1990), and indicates a reliance on on–off cues (Vickers et al., 2001). A similarity between A. aegypti's response to CO2 and moth responses to pheromone was also inferred from the lack of upwind progress of mosquitoes exposed to homogeneous clouds of CO2 in a narrow flight tube (Geier et al., 1999). Such a strong coherence in the response dynamics across such diverse taxa may indicate optimal strategies for plume tracking. It may also indicate similar constraints in odour processing and thus similar behavioural optima in odour source finding.

In contrast, after contacting the skin-odour ribbon plume, there was no change in the averaged flight parameters. Only rarely did we observe any response to brief encounters of skin odour. Instead, mosquitoes required longer exposure in broad plumes for upwind flight to ensue. In broad plumes of skin odour, higher levels of activation, upwind flight and source finding were observed when skin odour was presented as a continuous as opposed to a turbulent cloud, which also contrasts with male moth orientation to pheromone. Apparently, there is not a single model that fits in-flight responses to all odours, even within a species.

Fig. 8.

Coriolis forces and angular velocities in ribbon plume experiments. (A) Coriolis forces (deg cm s–1) and their axonal projections, yaw (across the Z-axis), pitch (across the Y-axis) and roll (across the X-axis); (B) angular velocities (deg s–1) and their 2-D projections over time after plume contact. The standard error is displayed as an envelope around the means. N-values are indicated in brackets.

Three points emerge from our observations on skin-odour and CO2-induced responses. First, CO2 induces a stronger and faster response than skin odour; second, CO2 induces this response independent of other host odours; and third, the response to CO2 seems to be highly similar across a 10,000-fold dilution range.

First, we observed that the response to skin odour requires an exposure longer than that to CO2 to induce upwind flight in A. aegypti, which indicates that responses to complex blends require more than ∼0.05 s to be recognized as a potential ‘host’. Time constraints in response and/or recognition are supported by experiments on odour-conditioned honeybees (Apis mellifera) using the proboscis extension response (PER) (Dekker, 2002; Ditzen et al., 2003; Fernandez et al., 2009). These studies all calculated that honeybees require ∼0.4 s for adequate recognition. Dekker furthermore found that, whereas honeybees required ∼0.4 s for PER, trains of 0.1 s pulses at a frequency of 3 Hz and higher, such as may occur in plumes disrupted by turbulence, also readily induced PER (Dekker, 2002). However, these reports were based on responses to learned odours, which may involve alternative olfactory pathways with different time constraints than innate odour responses. Yet, in the present study we found an apparent dichotomy between the innately attractive odours of CO2 and human skin, underscoring the possibility that time constraints underlie the response of mosquitoes to skin odour, but not (to the same extent) CO2. It could be that complex cues impose time constraints on processing, which in turn would imply that the response to CO2 bypasses these processes, even though in A. aegypti CO2 sensitizes the skin-odour circuitry (Dekker et al., 2005).

Second, it is remarkable that CO2, although part of the suite of host odours and detected by only 2% of the olfactory sensory neuron array (Grant et al., 1995; Grant and O'Connell, 1996), singly induces the full strength of a behavioural response typical for ‘attraction’. It is generally thought that odours in blends act additively or, more frequently, synergistically, and individual odours are at best only marginally effective in attracting insects compared with the complete blend. This seems to be the rule for mosquito attractants as well, and may have hampered progress in finding effective lures. An exception is CO2, as it does not require other odours to induce the full strength and sequence of orientation behaviours that lead to host finding in A. aegypti. This raises the question of how and where in the host orientation process other host odours come into play. ‘Synergism’ in mosquito orientation has been reported for CO2 in combination with several odours, most notably lactic acid in an olfactometer (Smith et al., 1970) and 1-octen-3-ol in a field suction trap (Takken and Kline, 1989). However, because only the end result, capture, was scored, it is unknown how these odours interacted. In a previous study on A. aegypti we demonstrated that what would be, based on the end result, commonly classified as ‘synergism’ between CO2 and skin odour was more properly described as CO2-induced sensitization of the skin-odour response (Dekker et al., 2005).

Similarly, it may be that some skin odours are primarily important near the host and involved in selection of biting sites or induction of landing (which seem in part odour mediated and independent of CO2) (De Jong and Knols, 1995; Dekker et al., 1998), and/or act for instance as ‘arrestants’ more than ‘attractants’, or both. Arrestants may increase trap capture through inducing the mosquito to remain in the vicinity of the odour source, increasing its likelihood of entering the trap's effective region of suction. Further studies on the effect of plume structure and composition, and interaction of odours on host finding and trap capture of A. aegypti and other mosquito species may aid development of traps that have a high efficacy of capture of mosquitoes that have been lured to its vicinity (e.g. Cooperband and Cardé, 2006a; Cooperband and Cardé, 2006b; Lacey and Cardé, 2011).

Finally, we observed that the dynamics (response time, duration and speed) of CO2-induced upwind surging were highly similar across an enormously wide range of concentrations from 100% down to 0.05%. The surprisingly consistent response across concentrations underscores the fundamental importance of CO2 in A. aegypti host orientation. However, differences between concentrations were also observed: reverting to casting was significantly faster at higher concentrations. This would be of ecological relevance as increasing concentrations in general indicate proximity to a host, and localizing manoeuvres may be critical to prevent loosing or overshooting a plume source, especially if the host is small (such as birds and rodents). Conversely, longer sustained flights more due upwind may be more favourable strategy at low concentrations of CO2 as the potential host must be located at greater distances, and casting may not necessarily increase the probability and efficiency of recontacting the lost plume. In contrast to the response of A. aegypti to CO2, males of several moth species reduced the track angles and ground speed at higher pheromone concentrations (Kuenen and Baker, 1982; Charlton et al., 1993; Mafra-Neto and Cardé, 1995). This disparity may reflect differences between insect lineages or differences in the source size, but it could also reflect differences in the analysis. In the moth studies, angles and speeds were lumped across tracks irrespective of plume encounters, whereas here we aligned tracks with respect to intercepting a single CO2 filament, which allows for more detailed moment-to-moment analysis.

Fig. 9.

Activation and source finding in response to broad turbulent and homogeneous plumes. (A) Activation of mosquitoes over experimental time in 30 bins of 10 s displayed in grey scale in an intensity graph. The scale ranges from 100% black (100% activation) to 100% white (no activation). The fractions of non-responders (NR) at the end of experimental time are similarly expressed as intensity graphs. The shape parameter α was in all cases <1 (between 0.5 and 0.6), which implies a higher activation rate at the beginning of the experiment. (B) Activation rate expressed as the proportion of activation per unit time (s–1). (C) Source finding expressed as a percentage of the total number of mosquitoes that were activated. Bars within the same graph labelled with a different letter were significantly different from each other. N-values are indicated in brackets.


CO2 induced a tightly arranged zigzagging pattern of horizontal and vertical turns across the windline, the frequency of which was slightly concentration dependent. Temporally regular zigzags were absent in clear air and after brief encounters with skin odour. What causes zigzagging flights in insects has been much disputed. Some authors argue that zigzags result from delayed torque responses to drift, others have proposed that moths maintain a preferred ratio between longitudinal and drift-induced flow, which in an odour plume would produce zigzagging flights. Kennedy viewed pheromone-induced zigzagging as a self-steered, central nervous system programme which he called counterturning (Kennedy, 1983). In this view, casting is an expression of the counterturning programme without upwind progress. Behavioural evidence supports the last view, as zigzagging continued in still air, i.e. without any transverse image flow. In addition, some moths can be induced to fly straight upwind, without discernible zigzagging (Mafra-Neto and Cardé, 1994; Vickers and Baker, 1994). Chow and Frye demonstrated that vinegar odour induced enhanced responses to visual rotation, permitting straighter flights (Chow and Frye, 2008). If increased visual saliency also follows odour contact in mosquitoes, one would have expected increased flight accuracy and reduced zigzagging under the premise that flight imprecision underlies zigzagging, which differs from our observations. Finally, Kanzaki and Shibuya found premotor neurons in the lateral protocerebrum of moths that showed a flip-flopping activity pattern reminiscent of counterturning, which may be involved in generating the zigzagging flights (Kanzaki and Shibuya, 1992).

Fig. 10.

Ground speed, flight speed and airspeed in response to skin odour and CO2 at various concentrations (±s.e.m.). (A) Homogeneous skin-odour series (left panel) and homogeneous CO2 series (right panel); (B) turbulent skin-odour series (left panel) and turbulent CO2 series (right panel). Significant differences within each graph are denoted by different letters above the bars. N-values are indicated on the bars.

In mosquitoes, regular zigzagging is odour-triggered, with detection of an above-ambient CO2 filament inducing a highly regular pattern of crosswind zigzagging in both the horizontal and vertical planes. Its frequency remained constant for many seconds, also when casting, and was similar for all CO2 concentrations. This supports the notion that, in mosquitoes, CO2 induces a motor programme that dictates zigzagging and casting. Perhaps more intriguing is the observation that the vertical counterturning frequency was approximately 1.6- to 1.9-fold higher than the horizontal counterturning frequency. Apart from the fact that this supports the idea that a motor programme dictates counterturning, the offset frequency in Z relative to Y caused the mosquito to transect repeatedly the YZ area in which the plume was previously detected (see Fig. 7C, right panel, YZ projection), and suggests that counterturning is adaptive in increasing the chance of recontacting a lost plume.

Flight control in A. aegypti: angular velocity and Coriolis forces

Rapid turning and surging upwind, such as induced by brief CO2 encounters, may negate drift information due to opposing, torque-induced optic flow. This implies that, during surging, the mosquito transiently gauges upwind by means other than optomotor feedback. Interestingly, in spite of the rapid changes in displacement across all three axes following CO2 contact, both the angular velocity and the rotational Coriolis forces remained very constant.

There are two obvious possibilities for how a mosquito may steer upwind during brief upwind surging. First, it could keep the overall visual flow constant. However, a fixed image flow seems unsustainable when turning occurs across all three axes with concordant changes in the visual flow. A second possibility is that the mosquito could keep the Coriolis force input constant, consistent with our observations. Organs ideally suited to balance flight are the halteres (Tammero and Dickinson, 2002), perhaps in conjunction with the Johnston organs. The latter are required for flight in both mosquitoes and moths (Bässler, 1958; Sane et al., 2007), are used by locusts as velocity (Gewecke, 1970) and heading detectors (Arbas, 1986), and have been suggested in detection of Coriolis forces in moths and fruit flies (Sane et al., 2007; Budick et al., 2008). Thus a mechanism relying transiently on non-visual cues would be superimposed on optomotor anemotaxis, only briefly relieving the insect from using optomotor cues. The turn would to some degree resemble saccade-type movements in Drosophila. Optomotor cues are not negated during saccades, but do not actively steer these (Tammero and Dickinson, 2002). Without detailed information regarding the position of the head and the body axis it is hard to judge by what means the mosquito controls its heading during upwind surges. Further experiments using, for instance, a flight simulator (Budick et al., 2008) with magnetically tethered mosquitoes or recording free flight with a high-speed camera that enables determination of the alignment of the body axis and antennal deflections with respect to wind and flight direction would be helpful in understanding how a mosquito controls its course.

Mosquito host finding and mosquito control: future prospects

A suite of odours that evoke peripheral responses in mosquitoes is currently being identified (e.g. Ghaninia et al., 2008; Carey et al., 2010). Besides the interest in understanding how mosquitoes recognize a prospective host, another impetus for these studies is their use of these odours in traps for capturing mosquitoes. Such traps can be useful for assessing mosquito biting habits, population densities and their dynamics, and may be essential in the concurrent evaluation of mosquito control programs (Okumu et al., 2010b). Odour-baited traps may even become an intervention tool themselves by intercepting host-seeking mosquitoes, with some recent progress (see Okumu et al., 2010a), although the importance of CO2 in the orientation process may diminish their potential in mass trapping considering the infrastructural and cost issues associated with using CO2 sources in tropical countries. Regardless, the feasibility of any mosquito trap depends largely on how effectively mosquitoes are captured. Although the correct composition of the odour blend is pivotal to the success of future traps, here we demonstrate that plume structure also influences orientation success, and in an odour-specific manner.

Fig. 11.

Mean change in track angle over time of mosquitoes flying into (right half of the panels) or exiting (left half of the panels) the plume consisting of CO2 or skin odour, presented in either a turbulent or homogeneous fashion. (A) Skin-odour series; (B) CO2 series. The vertical grey line indicates the plume boundary. The homogeneous plume data are from Dekker et al. (Dekker et al., 2005). Error bars are ±s.e.m. N-values are indicated in brackets in the keys.

Mosquito capture may not be just a matter of having the right host-odour blend at hand, and the odours themselves may not neatly fit the descriptors of either ‘attractant’ or ‘repellent’. Such terms may not be informative of the navigational processes influencing host finding, as the actual ‘result’ is not due to an inherent property of the stimulus and its detection. Moreover, successful odour source location can be dependent on many other factors including the availability of visual cues, light levels, odour plume structure, wind speed and the mosquito's ability to detect wind direction (Cardé and Gibson, 2010). Deciphering the cues used in host finding by mosquitoes may require an improved understanding of the behavioural sequences that support mosquito upwind flight and, ultimately, capture. These findings on A. aegypti need to be complemented with studies on other important mosquito vectors, such as A. gambiae and Culex quinquefasciatus (Lacey and Cardé, 2011).


We thank Kris Justus and Joseph Bau for assistance with PID measurements and track analysis, respectively. The helpful comments of two anonymous reviewers are acknowledged.


  • Present address: Division of Chemical Ecology, Department of Plant Protection Biology, Swedish University of Agricultural Sciences, PO Box 102, 230 53 Alnarp, Sweden

  • This work was supported by the University of California Systemwide Mosquito Research Program, the ONR-DARPA Plume Tracing Program (R.T.C.), the ICE3 Linnaeus grant and FORMAS (T.D.).


View Abstract