The sugar meal of the African malaria mosquito Anopheles gambiae and how deterrent compounds interfere with it: a behavioural and neurophysiological study

Plant-derived sugar is the basic food of adult mosquitoes (Foster, 1995), usually including sucrose, which is also the most common sugar in nectar, along with fructose and glucose (Wykes, 1952; Baker and Baker, 1982; Van Handel, 1984; Petanidou and Lamborn, 2005; Manda et al., 2007b). Carbohydrates are consumed by males to supplement the energy required for their intense swarming (Yuval et al., 1994; Gary et al., 2009). Females of anautogenous mosquito spp. feed on sugar to provide energy for flight and increase longevity, but need blood to produce eggs (Gary and Foster, 2001). The major source of sugar is in floral nectaries, although mosquitoes also feed on extrafloral nectaries, honeydew, tree sap, damaged fruits and leaves or discarded plant material like sugar-cane waste (Foster, 1995). Glucose can also be found in vertebrate blood, although its calorific content is negligible compared with that of blood proteins (Foster, 1995). In the field, mosquitoes are generally observed to feed on flowers in full bloom that are abundant, demonstrating opportunism (Magnarelli, 1979). However, in the African malaria mosquito Anopheles gambiae (Giles), some clear preferences exist for particular plants available in the vicinity of human hosts or oviposition sites (Manda et al., 2007b). To date, the physiological basis in A. gambiae for discriminating between plant species is still unknown. At medium to long range, plant volatiles probably inform mosquitoes of the presence of a sugar source (Foster and Takken, 2004) as well as those of unsuitable plants, as shown by the strong repellent activity of some plant extracts against female A. gambiae searching for a host (Odalo et al., 2005). After landing on the plant, mosquitoes use terminal pore sensilla on their legs and mouthparts to find the nectar source (Pappas and Larsen, 1978) and probably to taste meal quality, such as the sugar concentration or the presence of toxic and unpalatable substances. Behavioural avoidance of noxious allelochemicals is a mechanism of defence developed by insects against plant toxins. Deterrence (feeding inhibition) by bitter-tasting plant secondary metabolites such as alkaloids is known in phytophagous insects such as the moth Manduca sexta (Glendinning, 2002), in the scavenger fly Phormia regina (Dethier and Bowdan, 1992) and in the fruitfly Drosphila melanogaster (Meunier et al., 2003). Recently, it has been shown that the alkaloids quinine and caffeine affect the sucrose feeding response of Aedes aegypti (Ignell et al., 2010). Alkaloids act by activation of a taste neurone called the ‘deterrent receptor cell’ or via inhibition of a receptor cell that responds to phagostimulants (Schoonhoven, 1982).

In recent years, important progress has been made in the understanding of the cellular basis and role of gustatory receptors (Grs) involved in sweet and bitter sensing in Drosophila (Dahanukar et al., 2001; Moon et al., 2006; Dahanukar et al., 2007; Jiao et al., 2007; Slone et al., 2007; Weiss et al., 2011). A large family of putative Grs has been identified in the genome of A. gambiae (Hill et al., 2002) and the projection patterns of gustatory neurones have been described in detail in the suboesophageal ganglion and the tritocerebrum of this species (Ignell and Hansson, 2005). However, despite the potential importance of deterrent compounds in controlling arthropod-borne diseases, relatively little is known about taste in haematophagous insects. Here we describe feeding experiments with sucrose and its constituents fructose and glucose in A. gambiae, and provide insight into sugar perception in this species from electrophysiological recordings performed on a labellar trichoid sensillum containing sucrose- and water-sensitive gustatory receptor cells. We then present the feeding inhibition effects of denatonium benzoate, some quinoline (quinine and quinidine) and isoquinoline (berberine) alkaloids added to sucrose, and account for the inhibitory effect of these compounds on the labellar sucrose- and water-sensitive chemoreceptor cells.

The A. gambiae colony (16cSS strain, derived in 1974 from wild-caught adults originating from Lagos, Nigeria, West Africa) was maintained in a climate chamber [28°C, 80% relative humidity (RH)] under a 12 h:12 h light:dark cycle with 2 h simulated sunrise and sunset. Females fed on a guinea-pig once a week and eggs were recovered on wet filter paper. About 275 larvae were reared in trays in distilled water and fed with pulverized Tetramin fish food. This density provides optimal nutrition for larvae, synchronous eclosion and adults of homogeneous size (Timmermann and Briegel, 1993). Nine to 10 days after oviposition about 800 adult mosquitoes emerged into a rearing cage (350×350×550 mm high) and were provided with 10% (w/v) sucrose and water ad libitum.

d-(−)-fructose, d-(+)-glucose monohydrate, d-(+)-sucrose, berberine chloride, caffeine, quinine anhydrous, denatonium benzoate, NaCl, CuSO₄ and LaCl₃ were purchased from Sigma-Aldrich (Buchs, Switzerland), (+)-quinidine from Alfa Aesar (Karlsruhe, Germany) and KCl and CaCl₂ from Merck (Darmstadt, Germany). Puritiy of all products was ≥98%. Solutions were kept at 10°C, for less than 1 week.

Experiments were performed in a walk-in climate chamber (25°C, 80% RH) during the last 5 h of the scotophase. Four- to six-day-old female A. gambiae were individually deprived of water and sugar for 3–5 h in small plastic tubes (15 mm long, 15 mm diameter) closed with two perforated plastic stoppers (diameter 7 mm) with the aperture covered by stainless-steel netting (mesh size: 0.4 mm) to facilitate exchange of air kept in boxes at 95% RH. After this conditioning, the tubes were weighed using a M3 Microbalance (reading precision of ±1μg; Mettler, Greifensee, Switzerland). For this procedure, the mean standard deviation of weighing five different empty plastic tubes five times was 1 s.d.=2.64 μg. The mosquitoes were then transferred into glass vials (40 mm high, 20 mm diameter) providing adequate space for feeding. Vials were opened and upturned for 30 min on filter paper (55 mm diameter, no. 10 311 807, Whatman Schleicher and Shuell, Dassel, Germany) treated with 600 μl of a test solution in a Petri dish (modified from Arsic and Guerin, 2008). During feeding, a beige plastic container was upturned on the Petri dishes to avoid visual stimuli. Light did not exceed 1 lx under the container. After the feeding experiment, mosquitoes were individually anaesthetized with CO₂ and transferred into the initial plastic tubes and weighed again. The amount of solution ingested was established as the difference in weight before and after feeding. The time between weighings was 60–90 min. Four types of experiments were performed using this bioassay. Firstly, to establish the sucrose level that induces feeding by female A. gambiae, solutions of sucrose from 1 to 250 mmol l⁻¹ were dissolved in nanopure water. In a second experiment, to measure any phagostimulatory effect of the two hexose sugars that make up sucrose, fructose and glucose were tested individually at 146 and 292 mmol l⁻¹, mixtures of fructose and glucose at 73 mmol l⁻¹ each (the same molarity as in 146 mmol l⁻¹ sucrose) and at 146 mmol l⁻¹ each (i.e. the same mass of glucose and fructose that constitutes 146 mmol l⁻¹ sucrose), 146 mmol l⁻¹ sucrose (5% w/v) as a positive control and nanopure water as a negative control. To test for feeding inhibition by bitter products, 146 mmol l⁻¹ sucrose mixed with solutions of 1 mmol l⁻¹ caffeine, berberine, quinine, quinidine, denatonium benzoate and 2 mmol l⁻¹ berberine were compared with 146 mmol l⁻¹ sucrose as control. In a fourth set of experiments, feeding was quantified for quinine and quinidine at concentrations from 0.01 to 1 mmol l⁻¹ in 146 mmol l⁻¹ sucrose. Solutions were tested in a random order.

Feeding assays were performed under the same environmental conditions as described above from glass capillaries with 4- to 6-day-old female A. gambiae at low light (7 lx) to establish if surrounding the mouthparts alone, i.e. excluding tarsal chemosensilla, was sufficient to induce feeding and feeding inhibition by sucrose and quinine presented in sucrose. The mosquitoes were rendered quiescent on ice, and attached by the wings on a glass slide using double-sided adhesive tape 3–7 h before the start of the experiments. During this conditioning period mosquitoes had no access to sucrose or water but were kept in boxes at 95% RH. For these feeding assays, 5 μl glass capillaries (no. 7087 07, Blaubrand, Wertheim, Germany), previously shortened to remove the proximal part without calibration marks, were pulled with an Ealing vertical micropipette puller (no. 50-2012, Harvard Apparatus, Edenbridge, Kent, UK) into two 40 mm segments with a 180 μm internal diameter tip to accommodate the proboscis. Mosquitoes were allowed to feed for 5 min on solutions of 146 mmol l⁻¹ sucrose or 146 mmol l⁻¹ sucrose + 1 mmol l⁻¹ quinine in nanopure water. A blue food colorant (E131) was added to each solution (five drops of 30 mg to 10 ml) to permit visualization of the liquid levels in capillaries. Liquid levels were recorded as images with a scanner (Scanjet 4570c, Hewlett-Packard, Dübendorf, Switzerland) at 600 d.p.i. before and after experiments, and fluid consumption was measured using ImageJ version 1.44 (Rasband, 2011) as described in Sellier et al. (Sellier et al., 2011) using the 1 μl marks on the capillaries as references. To control for possible evaporation within the capillary, the level of liquid in a capillary filled with 146 mmol l⁻¹ sucrose solution and manipulated exactly as the test capillaries was likewise measured.

For transmission electron microscopy, the probosces of A. gambiae were fixed in Karnovsky fixative (pH 7.5) overnight at 4°C and rinsed three times in 0.2 mol l⁻¹ sodium cacodylate buffer with 4% sucrose. After post-fixation in 1% OsO₄ for 2 h and rinsing in the same buffer, the specimens were stained in block with 2% uranyl acetate (pH 4) in 15% ethanol for 30 min at room temperature, rinsed in 15% ethanol and dehydrated through a graded series of acetone and embedded in Spurr's resin. Ultrathin sections were obtained on a Reichert Ultracut S microtome (Leica, Vienna, Austria) and stained with uranyl acetate and lead citrate. Serial sections were collected every 1 μm from the tip of the first distal trichoid sensillum up to 60 μm of the proboscis from six female and two male A. gambiae and were examined in a Philips CM 100 electron microscope (Philips Electron Optics, Eindhoven, The Netherlands) at 60 kV. For scanning electron microscopy, heads of A. gambiae were mounted on stubs with Leit-tabs (Plano, Wetzlar, Germany), air-dried for 1 week and coated with a layer of gold in a Sputter coater (Baltec SCD 005; Oerlikon Balzers, Balzers, Lichtenstein). Specimens were examined at 10 kV using a Philips XL 30 scanning electron microscope.

Electrophysiological responses of chemoreceptor cells in a trichoid sensillum on the labellum of female A. gambiae (Fig. 6A, arrow) were recorded to test stimuli. After being rendered quiescent on ice, mosquitoes were attached by the back on plasticine with two curved tungsten wires, surrounded by moist cotton. After removing the maxillary palps, the mouthparts were fixed with double-sided adhesive tape on a Plexiglas stub. Contact chemosensilla on the mosquito mouthparts were visualized under a microscope (BX51WXIF, Olympus Schweiz, Volketswil, Switzerland) at a magnification of 1000.

Imagoes that had emerged the same morning (0-day-old) were used for electrophysiological recordings when not otherwise specified. Using the method first described by Hodgson et al. (Hodgson et al., 1955), sensilla were stimulated with a recording borosilicate electrode (33 mm long, 14 μm tip diameter) containing the test products diluted in nanopure water. A borosilicate electrode of the same size, mounted on plasticine, filled with haemolymph Ringer solution (Kaissling, 1995), grounded via a chlorinated silver wire and inserted into the mosquito head acted as a reference electrode. The recording electrode was connected via a chlorinated silver wire to a high impedance ‘non-blocking’ pre-amplifier (Taste Probe, Syntech, Hilversum, The Netherlands) (Marion-Poll and van der Pers, 1996) mounted on a hydraulic micromanipulator (M0-103, Narishige Scientific Instrument Laboratory, Tokyo, Japan). The electrophysiological signal was amplified ten times with an amplifier in conjunction with a 100–3000 Hz band-pass filter provided in an IDAC-4 intelligent data acquisition controller (Syntech). AC signals were recorded via the IDAC-4 analogue to digital card to a PC equipped with the spike analysis software AutoSpike (version 3.9; Syntech) at a 10 kHz sampling rate. Most concentrations of test products were tested two consecutive times for 2.5–6.5 s at intervals of approximately 2 s. An analysed signal was always the first to be recorded. In cases where a problem occurred during first stimulation (e.g. low signal/noise ratio, insect movement), spikes were quantified from a subsequent stimulation. A delay of 2–5 min was respected between each solution tested

At first, to evaluate the inhibitory effect of salts on the neurophysiological response to water, concentrations from 0.01 to 100 mmol l⁻¹ KCl and NaCl, from 0.01 to 1 mmol l⁻¹ CuSO₄, from 0.01 to 10 mmol l⁻¹ CaCl₂ and from 0.1 to 1000 μmol l⁻¹ LaCl₃ in nanopure water were tested in increasing concentration on the sensillum. Each stimulation series started and ended with nanopure water as controls. To check for the presence of a ‘sucrose receptor cell’, concentrations from 0.1 to 292 mmol l⁻¹ sucrose were tested in increasing order in 10 mmol l⁻¹ KCl, i.e. at a KCl concentration that inhibits the neurophysiological response to water. Then concentrations from 0 to 100 mmol l⁻¹ sucrose in pure water were tested in increasing order to check for the presence of a ‘sucrose receptor cell’ different from the ‘water receptor cell’. This was repeated with 5-day-old mosquitoes, an age that corresponds to the one used in the feeding assays. To study the sugar receptor site(s), fructose and glucose were tested separately at 146 and 292 mmol l⁻¹, mixtures of fructose and glucose (each at 73 mmol l⁻¹ and 146 mmol l⁻¹), and compared with 146 and 292 mmol l⁻¹ sucrose in 10 mmol l⁻¹ KCl as positive controls, and 10 mmol l⁻¹ KCl alone as a negative control. The action of bitter products on the responses to water and sucrose by the receptor cells within the trichoid sensillum was investigated by stimulating with concentrations from 0.01 to 1 mmol l⁻¹ denatonium benzoate, quinine, berberine or caffeine diluted in nanopure water and in 146 mmol l⁻¹ sucrose plus 10 mmol l⁻¹ KCl. Each stimulation series started and ended with nanopure water or 146 mmol l⁻¹ sucrose in 10 mmol l⁻¹ KCl as controls.

Spikes were quantified during the 2 s following stimulation. In cases where receptor cells were activated with a delay, spikes were quantified during the 2 s following activation of receptor cells or from a subsequent stimulation showing no delay. All products tested affected receptor cell responses over 2 s. Spikes were discriminated on the basis of amplitudes and shapes using the interactive AutoSpike procedures. Spike amplitudes were sorted using DataView 8.0.0 software (Heitler, 1999) omitting the first 30 ms to avoid the contact artifact. Moreover, to confirm the activation of more than one receptor cell, all recordings were visually inspected for doublets (as presented in Hiroi et al., 2004). Since spike amplitude varies in parallel with frequency in labellar chemosensory cells of A. gambiae as in Drosophila (Fujishiro et al., 1984), a careful analysis of several recordings was made for different concentrations of products on the same sensillum of the same mosquito (see Results) for the identification of active sensory units.

To establish differences between sugar phagostimulatory effects, the meal sizes of female A. gambiae exposed to filter papers treated with sucrose, fructose, glucose and the two mixtures of fructose and glucose were analysed using one-way ANOVA on the ranked values (discrete data; normal distribution, Shapiro–Wilk test, P>0.05; homogeneous variance, Bartlett's test, P>0.05). The model was simplified by aggregating non-significant factor levels in a stepwise a posteriori procedure. Similar analyses were performed on the meal sizes of female A. gambiae exposed to filter papers treated with 146 mmol l⁻¹ sucrose and 146 mmol l⁻¹ sucrose with bitter products added. To analyse feeding inhibition by quinine and quinidine as a function of concentration and to check for possible interactions between products and concentrations, an ANCOVA based on a linear model was made on the ranked meal sizes (continuous data; normal distribution, Shapiro–Wilk test, P>0.05; homogeneous variance, Bartlett's test, P>0.05) of female A. gambiae exposed to different concentrations of quinine and quinidine added to 146 mmol l⁻¹ sucrose. In the feeding assay with glass capillaries, meal sizes (normal distribution, Shapiro–Wilk test, P>0.05; homogeneous variance, Bartlett's test, P>0.05) of females exposed to 146 mmol l⁻¹ sucrose and 146 mmol l⁻¹ sucrose with 1 mmol l⁻¹ quinine added were compared using the t-test.

The calculated model indicates that starved 4- to 6-day-old female A. gambiae start to feed at concentrations between 50 and 75 mmol l⁻¹ with meal size estimated at 0.045 mg at 50 mmol l⁻¹, and 0.088 mg at 75 mmol l⁻¹ sucrose (Fig. 1). At 250 mmol l⁻¹, mosquitoes engorge 0.376 mg of sucrose solution, close to the asymptote of the model (0.378 mg), an amount corresponding to 30% of their weight prior to feeding. ED₅₀ is calculated from the model to occur at 111.09 mmol l⁻¹ sucrose with a meal size of 0.19 mg. At 1 and 10 mmol l⁻¹ sucrose, mosquitoes imbibe almost nothing (0.01 mg), comparable to the water control [median meal size (MMS), −0.007 mg]. Zero-day-old females can discriminate between water and 292 mmol l⁻¹ sucrose like the 4- to 6-day-old females in our assay (data not presented).

Following ANOVA on the ranked meal sizes on fructose, glucose, their binary mixtures and sucrose followed by a stepwise aggregation of non-significant factor levels, three significant contrasts are retained in the final model (F_2,159=35.09, P<0.001, Fig. 2). Meal sizes on water (MMS, −0.021 mg), on 146 or 292 mmol l⁻¹ glucose (MMS, −0.0025 and 0.027 mg, respectively) and on 73 mmol l⁻¹ glucose plus fructose (MMS, −0.0135 mg) are low and not significantly different. While there are no significant differences between 146 and 292 mmol l⁻¹ fructose (MMS, 0.046 and 0.0505, respectively), fructose appears to have a mild phagostimulatory effect on female A. gambiae as meal sizes are significantly higher than on water, on 146 and 292 mmol l⁻¹ glucose or on 73 mmol l⁻¹ glucose plus 73 mmol l⁻¹ fructose (P=0.002), but lower than on 146 mmol l⁻¹ glucose plus 146 mmol l⁻¹ fructose and on 146 mmol l⁻¹ sucrose (P<0.001). The mixture of 146 mmol l⁻¹ glucose plus 146 mmol l⁻¹ fructose (MMS, 0.133 mg) and 146 mmol l⁻¹ sucrose alone (MMS, 0.137 mg) have a similarly strong phagostimulatory effect on starved female A. gambiae.

Except for caffeine, all bitter products tested significantly reduced feeding in A. gambiae when added to a solution of 146 mmol l⁻¹ sucrose (Fig. 3). Following ANOVA on the ranked meal sizes followed by a stepwise aggregation of non-significant factor levels, three significant contrasts were retained in the final model (F_2,156=101.4, P<0.001). When 1 mmol l⁻¹ caffeine is added to 146 mmol l⁻¹ sucrose the meal size of starved female A. gambiae is not significantly reduced (MMS, 0.184 mg) compared with 146 mmol l⁻¹ sucrose alone (MMS, 0.204 mg). Although no significant difference in meal size was found between 1 mmol l⁻¹ berberine (MMS, 0.121 mg) and 2 mmol l⁻¹ berberine (MMS, 0.069 mg), both concentrations of berberine have an intermediate deterrent effect, with meal sizes significantly lower than 1 mmol l⁻¹ caffeine in sucrose and 146 mmol l⁻¹ sucrose alone (P<0.001), but significantly higher than 1 mmol l⁻¹ quinidine, quinine and denatonium benzoate (MMS, −0.028, −0.0425 and −0.047 mg, respectively, P<0.001) all of which almost completely suppress feeding on 146 mmol l⁻¹ sucrose. Although the deterrent effect of both quinidine and quinine clearly depends on concentration (Fig. 4), no significant interactions and no significant differences in the intercepts of the linear equations describing the responses occur between these two products (F_3,115=40.85, R²=0.5, P<0.001). Meal sizes on water are not significantly different from meal sizes of mosquitoes exposed to 5% sucrose mixed with 1 mmol l⁻¹ quinine, quinidine and denatonium benzoate (ANOVA on the ranked values, P=0.296), allowing us to conclude that mosquitoes do not significantly lose more liquid by diuresis or regurgitation in the presence of the bitter compounds tested.

Five minutes are sufficient for starved female A. gambiae to engorge on 146 mmol l⁻¹ sucrose from glass capillaries (mean ingested volume 0.34 μl; Fig. 5), a level approaching that recorded on filter paper. Quinine at 1 mmol l⁻¹ significantly reduced the volume ingested by female A. gambiae when added to 146 mmol l⁻¹ sucrose (mean ingested volume: 0.15 μl, t=3.13, d.f.=28, P=0.004; Fig. 5). The mean level of evaporated liquid from capillaries was negligible (0.029 μl). As tarsal contact chemosensilla were not in contact with the solution, this indicates that contact chemosensilla either on the mouthparts or in the cibarium are involved in the perception of sucrose and alkaloids. For this reason the electrophysiological recordings described below focused on mouthpart sensilla.

Long trichoid sensilla of between 30 and 35 μm occur on the external surface of the labellar lobes of both male and female A. gambiae (Fig. 6A). These are typical gustatory sensilla with a terminal pore (Fig. 6B) and two lymphatic spaces divided by a cuticular wall (Fig. 6C). The inner receptor lymph cavity (Fig. 6C) contains four chemosensory dendrites and the outer lymph cavity contains an electron-dense fluid (Fig. 6C). In all, five neurones innervate each sensillum (Fig. 6D). One dendrite terminates in a tubular body (Fig. 6D) at the base of the hair, while the other four extend into the inner lymph cavity to the tip of the hair close to the terminal pore (Fig. 6C). At the base, the five outer dendritic segments are surrounded by a sheath, secreted by the thecogen cell, which ends below the cuticle of the socket (Fig. 6D). Further down, the three enveloping cells (thecogen, trichogen and tormogen) encase the five ciliary roots (with 9×2+0 microtubules) and the inner dendritic parts of the sensory cells (Fig. 6E). The morphology of the trichoid sensillum described here on the labellum of female A. gambiae appears to be very similar to the long labellar trichoid type 1 sensilla (T1) described by Pappas and Larsen (Pappas and Larsen, 1976) in Culiseta inornata (Williston) and those described by Owen (Owen, 1971) and Hill and Smith (Hill and Smith, 1999) in Anopheles atroparvus (Van Thiel) and A. aegypti, respectively. About 10 T1 sensilla occur on the dorsal and ventral sides of each labellar lobe of female and male A. gambiae and no differences have been found between the sexes in either the placement or morphology of T1 sensilla in this species. One dendrite appears to be larger than the other three (measured midway along four different T1 sensilla of three females; Fig. 6C).

When the trichoid sensillum (indicated in Fig. 6A) is stimulated with nanopure water, one receptor cell is activated at a mean frequency of 64.12±1.97 spikes per 2 s. However, this receptor cell is inhibited by increasing concentrations of salts, with CuSO₄ and LaCl₃ being the most inhibitory (Figs 7, 8). Inhibition is already significant in comparison with the water control at 0.5 mmol l⁻¹ CuSO₄ (P<0.001) where the mean total number of spikes per 2 s (MTNoS) is reduced by 60% (Fig. 7D). LaCl₃ significantly inhibits the water response at 1 mmol l⁻¹ (P<0.001) with a drop of 92% in the MTNoS recorded compared with the first water control (Fig. 7F). The inhibitory effects of 1 mmol l⁻¹ KCl, NaCl or CaCl₂ are lower, with a drop in the MTNoS of 40, 50 and 59%, respectively, but significantly different from the water control (P=0.002, 0.002 and 0.02, respectively; Figs 7, 8). The increase (six spikes) in the MTNoS observed between 0 and 0.1 mmol l⁻¹ CaCl₂ is not significant (P=0.525). Between 10 and 100 mmol l⁻¹ KCl and NaCl the MTNoS increases slightly (10 and 15 per 2 s, respectively; Fig. 7) due to the appearance of predominantly negative-going spikes at these levels of salts (Fig. 7C; Fig. 8A,B). Such spikes could arise as an artifact at high salt levels or represent activation of another cell. All spikes quantified in these experiments could be readily discriminated from the spikes of the mechanoreceptor cell, which are readily visualized at a lower amplitude when the sensillum is pulled proximally (Fig. 8C). In 50% of the mosquitoes tested, the water receptor cell was activated after a delay of up to 2 s following electrode contact with the sensillum tip (Fig. 9A). In such cases a second stimulation was generally made not more than 2 s later and the receptor cell then responded immediately in a phasic–tonic manner (Fig. 9B).

The mean number of spikes per 2 s (MNS) of the sucrose receptor cell increases strongly from 0.1 mmol l⁻¹ (MNS 0.75±0.75) to 50 mmol l⁻¹ sucrose (MNS 62.33±3.73) in 10 mmol l⁻¹ KCl (Fig. 10; Fig. 11A–C). The frequency of action potentials emitted by the sucrose-sensitive neurone tends to an asymptote (57.35 spikes per 2 s) from 50 mmol l⁻¹ in 10 mmol l⁻¹ KCl. Half of the maximum frequency (ED₅₀) is estimated at 4.07 mmol l⁻¹ sucrose with 28.67 spikes per 2 s. The amplitude of spikes generated by the sucrose receptor cell increases slightly from 3.0±0.2 to 3.8±0.4 mV between 1 and 50 mmol l⁻¹, followed by a drop to 2.0±0.2 mV at 292 mmol l⁻¹, possibly due to the increased resistance caused by the high sucrose concentration in the electrode (Fig. 11D; supplementary material Fig. S1). At 0 mmol l⁻¹ sucrose in 10 mmol l⁻¹ KCl, no spikes are generated from the sucrose receptor cell. Only spikes of smaller amplitude with a predominantly negative-going shape are observed (Fig. 11A). These spikes could be generated by the water neurone, not totally inhibited by 10 mmol l⁻¹ KCl, or by a third sensory neurone sensitive to salt (see above). In Fig. 11B the two types of spike amplitudes are observed during stimulation with 2.5 mmol l⁻¹ sucrose in 10 mmol l⁻¹ KCl. The smallest spike amplitude class is inhibited as the concentration of sucrose increases and drops to a MNS of 3 at 50 mmol l⁻¹. The same delay, as observed for the activation of the water neurone, was observed for the activation of the sucrose receptor cell.

To determine whether different cells are activated by water and sucrose, increasing concentrations of sucrose between 0.01 and 100 mmol l⁻¹ were tested in nanopure water. The water receptor cell fires at a MNS of 61.92±5.35 when the sensillum of 0-day-old mosquitoes is stimulated by pure water. As the sucrose concentration increases, a second spike amplitude category appears from 0.1 mmol l⁻¹ with increasing amplitude in parallel with frequency (Fig. 12A,C). At 10 mmol l⁻¹ sucrose in water, the amplitude distribution is bimodal in 100% of the recordings (shown in the histograms on the right-hand side of each electrophysiological recording; Fig. 13). This confirms the presence of two activated sensory units. Two electrophysiological traces were removed from the analysis as two sensory units, whose presence was confirmed by the presence of doublets, could not be discriminated on the basis of amplitude, something never observed when sucrose was diluted in 10 mmol l⁻¹ KCl. Attributing spike amplitudes to the sucrose or water receptor cell is based on the fact that spike amplitudes generated by these cells vary in parallel with frequency (Fujishiro et al., 1984). This observation on labellar chemoreceptor cells of Drosophila is confirmed here for the labellar sucrose and water cells of A. gambiae (Fig. 12A,C). Thus at the sucrose concentration higher than the one where the two spike amplitudes are most similar, the attributions assigned to the two sensory cells is switched for purposes of counting, i.e. the spike amplitude previously assigned to the sucrose cell becomes the water cell and the spike amplitude previously assigned to the water cell becomes the sucrose cell. The concentration at which the mean amplitude is most similar is 5 mmol l⁻¹ (Fig. 12A) but this can vary between 1 and 50 mmol l⁻¹ depending on the mosquito tested.

This indicates that the asymptote for spikes generated by the sucrose cell of 5-day-old mosquitoes is some two times higher than that of 0-day-old mosquitoes (supplementary material Fig. S2). The model already tends towards the asymptote between 10 mmol l⁻¹ (93.05 spikes per 2 s) and 25 mmol l⁻¹ (96 spikes per 2 s), and the ED₅₀ is situated at 1.99 mmol l⁻¹, more than two times lower than that for 0-day-old mosquitoes.

The sucrose receptor cell is hardly activated when the labellar trichoid sensillum is stimulated with either 146 or 292 mmol l⁻¹ glucose (MNS 1.2±1.2 and 1.4±1.4, respectively) with 146 or 292 mmol l⁻¹ fructose (MNS 0±0 and 4.20±1.91, respectively) in 10 mmol l⁻¹ KCl or the control (10 mmol l⁻¹ KCl alone; MNS 0±0, P=0.11; Fig. 14). Nevertheless, a mixture of 73 mmol l⁻¹ glucose plus 73 mmol l⁻¹ fructose (MNS 37.00±3.94), 146 mmol l⁻¹ glucose plus 146 mmol l⁻¹ fructose (MNS 36.80±5.29), 146 mmol l⁻¹ sucrose (MNS 65.20±10.33) and 292 mmol l⁻¹ sucrose (MNS 42.25±11.09) activates the sucrose receptor cell (Fig. 14A,B). The number of action potentials recorded per 2 s tend only to be different (F_3,15=3.04, P=0.06) for 146 mmol l⁻¹ sucrose compared with 73 mmol l⁻¹ fructose plus 73 mmol l⁻¹ glucose, 146 mmol l⁻¹ fructose plus 146 mmol l⁻¹ glucose and 292 mmol l⁻¹ sucrose in 10 mmol l⁻¹ KCl (Fig. 14).

Except for caffeine, all the bitter products tested inhibit both the sucrose and water cells at concentrations between 0.01 and 1 mmol l⁻¹ (Figs 15, 16), and inhibition is accompanied by a decrease in spike amplitude (supplementary material Fig. S3A,B). A linear model fitted to the number of spikes per 2 s as a function of the log₁₀ of concentration of the bitter products proved significant for all inhibitory effects (F_9,186= 37.34, P<0.001, R²=0.63; Fig. 15). At the highest concentration tested (1 mmol l⁻¹), one which corresponds to the intercept in the model, nine action potentials less are generated for all bitter products tested in water than in 146 mmol l⁻¹ sucrose (P=0.003). The slopes fitted by the model are also significantly steeper for water than for sucrose (P=0.03), suggesting that bitter products are more efficient at inhibiting the water than the sucrose receptor cell. Denatonium benzoate at 1 mmol l⁻¹ is significantly more efficient at inhibiting both the sucrose and the water receptor cells (MNS 6.62±2.18; Fig. 15A,E) than quinine (MNS 13.71±3.24, P=0.02; Fig. 15B,F), berberine (MNS 16.42±4.22, P=0.02; Fig. 15C,G) and caffeine (MNS 51±4.94, P<0.001; Fig. 15D,H). The slope of the model is significantly less steep for berberine than for denatonium benzoate (P=0.006) due to the high number of action potentials recorded at 1 mmol l⁻¹ berberine in sucrose or water and the greater ability of berberine to inhibit both the sucrose and the water receptor cells at 0.01 mmol l⁻¹ in sucrose (Fig. 15C,G). Caffeine has no effect on the electrophysiological response of the sucrose and the water receptor cells with a slope not significantly different from 0 in water or sucrose (P=0.39; Fig. 15D,H). Both the phasic and tonic part of the response of the sucrose and water neurones is affected by bitter products, as shown in supplementary material Fig. S4A,B for inhibition by quinine. No receptor cell appears to be activated by the concentration range of the bitter products tested in the labellar sensillum on which the present study focuses.

In this study sucrose proved a strong phagostimulant for starved female A. gambiae. Mosquitoes start to feed at concentrations between 50 and 75 mmol l⁻¹ sucrose in solution. ED₅₀ is reached at around 111 mmol l⁻¹, the inflexion point of the feeding model. Small changes in the sucrose concentration at this dose where the slope of the response is at its steepest (0.625%) have the highest effect on meal size. However, the amount of ingested solution reaches a maximum between 146 and 250 mmol l⁻¹ sucrose. Since females engorge on a 146 mmol l⁻¹ sucrose solution from a glass capillary in a comparable manner, this permits us to conclude that the presence of receptor cells on the mouthparts (eventually within the cibarium) sensitive to sucrose are sufficient to elicit feeding. The glass capillary feeding experiments indicate that tarsal receptor cells are apparently not necessary to induce feeding on sucrose. It should be remembered that A. gambiae females feed on NaCl solutions only when heated as in blood and the meal is directed to the mid-gut, whereas a sucrose solution needs to be at room temperature and is directed to the crop (Arsic and Guerin, 2008).

Our electrophysiological study focuses on two labellar receptor cells: one is activated by sucrose in a concentration-dependent manner, and a second is activated by water and inhibited by sucrose and salts. The electrophysiological response of the sucrose receptor cell to sucrose diluted in water starts well below the levels of sugar necessary to induce engorgement. For both 0- and 5-day-old mosquitoes, the slopes of the two electrophysiological models are steepest at 801 and 3326%, respectively, at 0.01 mmol l⁻¹ sucrose in water, decreasing to 179 and 103%, respectively, at 10 mmol l⁻¹, and plateaus from approximately 25 mmol l⁻¹ for both ages. This shows that A. gambiae is most sensitive to small increments in sucrose concentration up to 10 mmol l⁻¹. An ED₅₀ is recorded at 1.99 mmol l⁻¹ sucrose for 5-day-old mosquitoes, 56 times lower than the ED₅₀ for feeding. For 0-day-old mosquitoes, an almost identical model was calculated for the response of the sugar cell to sucrose presented in 10 mmol l⁻¹ KCl and for sucrose presented in water. The sugar cell seems not to be affected by 10 mmol l⁻¹ KCl, contrary to its effect on the water cell. Whereas A. gambiae is apparently sensitive to sucrose concentrations between 0.1 and 10 mmol l⁻¹, it does not necessarily imbibe such solutions of a suboptimal energy level. Mosquitoes probably use other taste sensilla during their probing-walking responses (Pappas and Larsen, 1978; Sanford and Tomberlin, 2011) to reach a source of sugar at 50 mmol l⁻¹ or higher that activates the neurone near its maximum frequency.

In contrast to sucrose, solutions of glucose or fructose alone (the hexose sugars of sucrose) show, respectively, low and intermediate stimulatory effects on feeding. However, female A. gambiae engorge on glucose and fructose when combined at 146 mmol l⁻¹ each, just as on 146 mmol l⁻¹ sucrose, and the feeding response to this mixture is higher than on a 292 mmol l⁻¹ solution of either glucose or fructose alone. A direct correlation exists between this feeding response and the sensory input in that an equimolar 146 mmol l⁻¹ solution of glucose plus fructose activates the sucrose neurone to the same extent as either 146 mmol l⁻¹ or 292 mmol l⁻¹ sucrose, whereas no such activation is observed when the sensory neurone is stimulated with either hexose alone. This suggests a synergetic effect between glucose and fructose on the feeding response of female A. gambiae that is mediated by the sucrose taste neurone in the labellar trichoid T1 sensillum, a phenomenon already described for the blowfly Phormia regina (Dethier et al., 1956; Omand and Dethier, 1969) and the mosquitoes Culiseta inornata (Schmidt and Friend, 1991) and A. aegypti (Ignell et al., 2010). The fact that the sucrose labellar neurone of A. gambiae is activated by sucrose or by a mixture of glucose plus fructose but not by glucose or fructose alone suggests that it expresses distinct domains for glucose and fructose that both require activation to obtain a response from the neurone. It was proposed that the dendritic membrane of contact chemoreceptor cells in C. inornata is composed of pyranose sites specific for α-glucopyranose and fructopyranose, and fructofuranose sites (Schmidt and Friend, 1991). It is known that several Grs are co-expressed in insect contact chemoreceptor cells (Isono and Morita, 2010) and sugars are detected through multimeric receptors composed of two or more Grs in Drosophila (Dahanukar et al., 2007). Eight genes related to the Gr5a of Drosophila have already been identified as candidate sugar receptors in the A. gambiae genome (Hill et al., 2002; Kent et al., 2008). The low phagostimulatory effect of glucose and fructose on A. gambiae is not in agreement with what is known for other mosquito genera where fructose, glucose and sucrose all show strong phagostimulatory effects on their own on C. inornata, A. aegypti and Aedes taeniorhynchus (Wiedemann) (Salama, 1967; Nayar and Sauerman, 1971).

Although mosquitoes can drink water to maintain their water balance (Benoit and Denlinger, 2010), the amount of pure water drunk in our feeding assay by mosquitoes maintained at a high relative humidity was low. In our assay, negative values for meal size are probably related to body water loss when the mosquitoes did not feed at all (Benoit et al., 2010). Our results suggest that the sugar receptor cells are responsible for food acceptance rather than the water neurones when mosquitoes are exposed to sucrose diluted in water. In the blowfly Protophormia terraenovae (Robineau-Desvoidy) it was shown that labellar water neurones function as osmometers since saline solutions, hypertonic with respect to the intracellular medium, cause cell shrinkage contrary to hypotonic solutions, such as that of glycerol, which cause cell swelling (Solari et al., 2010). Cell swelling coupled to a calcium influx leads to a transient receptor potential and action potential generation (Liscia et al., 1997). The possibility that the water receptor cell functions as an osmometer in A. gambiae is indicated by the strong inhibitory effect on the labellar hair neurone found here for CuSO₄ at 1 mmol l⁻¹ (a supposed aquaglyceroporin AQP3 inhibitor) (Zelenina et al., 2004). A similar effect has been recorded for CuSO₄ on the water cell on the labellar lobes of P. terraenovae (Solari et al., 2010). Despite the strong inhibitory effect found here for LaCl₃ on the labellar water neurone of A. gambiae, a broad Ca²⁺-permeable channel inhibitor, no significant activation of the neurone was observed when the extracellular Ca²⁺ concentration was increased by stimulating the sensillum with CaCl₂, as observed in the tarsal water neurones of Drosophila (Meunier et al., 2009). Whereas 1 mmol l⁻¹ KCl is used as the reference to characterize the water cell in labellar and tarsal taste sensilla of D. melanogaster (Hiroi et al., 2004; Meunier et al., 2009), the labellar neurone of Anopheles appears to be more sensitive to KCl as this concentration reduced the number of spikes generated by the water cell of the T1 sensillar receptor cells by 40%. Water, salt and sugar neurones have been identified in labellar trichoid T1 of C. inornata (Owen et al., 1974; Pappas and Larsen, 1976). The functionality of the lateral labellar T1 sensillum neurones in A. gambiae on which this study focused seems to be similar as separate sugar and water neurones have been identified in the present study. However, the functionality of the other labellar T1 sensilla of A. gambiae still need to be characterized.

Anopheles gambiae feeds preferentially on nectars from Ricinus communis (Euphorbiaceae) and Tecoma stans (Bignoniaceae) in its natural habitat (Manda et al., 2007b). In the case of R. communis, sucrose, glucose and fructose are present at a concentration of about 30% (w/v, 876 mmol l⁻¹ for sucrose and 1665 mmol l⁻¹ for the two hexoses) each in extrafloral nectar (Baker et al., 1978), and fructose and glucose are present at a concentration of around 50% (2775 mmol l⁻¹) each in floral nectar of T. stans whereas its sucrose content is only 2% (Freeman et al., 1985). The ratio between sucrose and hexose sugars can vary in floral nectar but it is rare to find nectar in which only one sugar occurs (Baker and Baker, 1982). Of the five Kenyan endemic plant species on which A. gambiae were observed feeding most frequently, fructose or glucose were never detected alone in floral or extra-floral nectaries (Manda et al., 2007a). In the same study these authors have shown that longevity in days of female A. gambiae that had access to 6% (w/v) glucose increased by more than fivefold over females that had access only to water. If A. gambiae is to exploit glucose or fructose alone as a resource, then individuals need to be starved for more than the 3 h period used in this study to compensate for the lower phagostimulatory effect of these monosaccharides. This is quite a plausible scenario since a steady diet of 10% sucrose during the days prior to feeding assays described here allowed Anopheles to increase its carbohydrate and lipid reserves (Briegel, 1990). Some plant nectars are more suitable at increasing A. gambiae longevity and fecundity over others (Manda et al., 2007a; Manda et al., 2007b). Such differences among plant species in meal quality could be due to optimal sugar levels combined with other essential nutrients or to the absence of toxic compounds, whereas plant preferences could be related to absence of deterrents.

Except for caffeine, all bitter products tested significantly reduced sugar feeding in A. gambiae. The glass capillary feeding experiment including quinine allows us to conclude that proboscis and/or cibarial gustatory receptor cells are sufficient for the perception of bitter products. Ignell et al. (Ignell et al., 2010) found that the concentration of quinine that starts to inhibit feeding in A. aegypti is affected by the sucrose level presented, with only 0.001 mmol l⁻¹ quinine required at 10 μmol l⁻¹ sucrose but with 0.1 mmol l⁻¹ required at 1 mol l⁻¹ sucrose. The latter concentration of quinine is proportional to what starts to be effective at inhibiting feeding by A. gambiae in our no-choice assay, i.e. 0.01 mmol l⁻¹ quinine in 146 mmol l⁻¹ sucrose. However, whereas Ignell et al. (Ignell et al., 2010) found that quinine only affects feeding by female A. aegypti, quinine and quinidine were found to affect the feeding response of male and female A. gambiae in a comparable manner in our assays (data not shown). In most cases alkaloids inhibit feeding in insects (Dethier and Bowdan, 1989; Ramaswamy et al., 1992; Zhang and Mitchell, 1997; Meunier et al., 2003). Bitter products are not all equally toxic, and, to date, it is not clear if there is a direct correlation between bitterness and toxicity of these products for insects (Glendinning, 2002; Weiss et al., 2011).

The concentration-dependent inhibitory effects of denatonium benzoate, quinine and berberine on the electrophysiological responses of labellar sucrose and water receptor cells to water or sucrose confirms the critical role of sensory inhibition in the perception of deterrent compounds in A. gambiae as in other insects (Schoonhoven, 1982). These compounds also inhibit feeding, whereas caffeine, which fails to inhibit feeding even at 1 mmol l⁻¹ by A. gambiae on sucrose, shows no inhibitory effect on the electrophysiological responses of the neurones. Inhibition of the receptor cell responses is stronger when denatonium benzoate, quinine and berberine are presented in pure water than in 146 mmol l⁻¹ sucrose, indicative of a competitive phenomenon between sucrose and these bitter products. This suggests that the bitter compounds induce inhibition of the sucrose cell by acting on specific Gr(s) or on a specific ion channel type. It was suggested that the water receptor cells in Drosophila are inhibited by bitter products through an osmolarity-based mechanism (Meunier et al., 2003; Weiss et al., 2011). In our study we show that both the labellar sucrose and water neurones would appear to be involved in the detection of all products showing feeding deterrent activity in A. gambiae. Nevertheless, the numbers of bitter compounds we have tested is not exhaustive. To date, we have not succeeded in characterizing receptor cells in sensilla on the proboscis or on the tarsi of A. gambiae that are selectively activated by bitter products, as is known in Drosophila (Meunier et al., 2003; Hiroi et al., 2004; Weiss et al., 2011). It has recently been shown that labellar neurones responding to bitter products in Drosophila can be segregated into four functional classes, each possessing a specific response pattern to 16 bitter products and expressing a specific combination of 33 Grs (Weiss et al., 2011).

In summary, this study highlights how bitter-tasting products have a deterrent effect on sugar feeding by A. gambiae. The diversity of bitter compounds in nature has led to a wide diversity of Grs involved in the detection of unpalatable substances in Drosophila (Weiss et al., 2011). Despite this, we have identified two neurones on the labellar lobes of A. gambiae that are activated by palatable compounds and inhibited by unpalatable ones. These neurones could be the target of future tests to uncover novel deterrents for A. gambiae and to identify phagostimulants to better understand the ecology of feeding in this species.

The authors wish to thank all members of the Animal Physiology Laboratory at the University of Neuchâtel, particularly Martine Bourquin for her help in mosquito rearing and with feeding assays, and Charles Chappuis, Jérôme Frei, Alexandre Gurba, Caroline Joris, Thomas Kröber and Andrew McMullin for their worthwhile suggestions. Two anonymous reviewers are also acknowledged for their pertinent comments on the manuscript. Maarten Voordouw, Institute of Biology, University of Neuchâtel, and Radu Slobodeanu, Institute of Mathemantics, University of Neuchâtel are acknowledged for their help with statistics. Erich Städler, University of Basel, Switzerland, Frédéric Marion-Poll and Moutaz Ali Agha, INRA Versailles, France, are thanked for having communicated useful information on the tip recording method from insect sensilla. Christian Hêche, Daniel Varidel and all collaborators at the Technical Service, Faculty of Science, University of Neuchâtel are acknowledged for their support.

 
• ED₅₀

  effective dose that induces 50% of the maximum response

 
• MMS

  median meal size

 
• MNS

  mean number of spikes per 2 s generated by only one receptor cell

 
• MTNoS

  mean of the total number of spikes per 2 s counted without discriminating between the receptor cells of the sensillum