Mosquito host preferences affect their response to synthetic and natural odour blends

The semi-field experiments utilized laboratory colonies of the Mbita strain of An. gambiae sensu stricto and An. arabiensis. Aquatic stages of the mosquitoes were separately reared under ambient atmospheric conditions in screen-walled greenhouses at the Thomas Odhiambo Campus Odhiambo (TOC) of the International Centre of Insect Physiology and Ecology (ICIPE), Mbita, Kenya. Adult mosquitoes were placed in a holding room under ambient conditions with a scotophase of 12:12 h. Female adult mosquitoes were fed three times a week on a human arm [18]. Eggs were laid on moist filter paper and dispensed into plastic trays containing filtered water from Lake Victoria. Newly hatched larvae were transferred into plastic basins and fed on Tetramin® baby fish food (Melle, Germany) three times a day. Collection of pupae until adult emergence is described in Mukabana et al. [10]. Female mosquitoes used for semi-field experiments were placed in mosquito netting covered plastic cups [10]. They had no prior access to a blood meal but were fed only on water, provided on wet cotton towels placed on top of mosquito-holding cups during starvation [10]. All semi-field experiments were carried out at night (20:00–06:30 h) inside a 7 × 11 m screenhouse [16]. Two-hundred females of An. gambiae and 200 An. arabiensis aged three to eight days old were painted with either pink or yellow fluorescent dyes (FTX Series, Astral Pink, Swada, London) ten hours before the experiments, as described before [19]. Mosquitoes were starved for eight hours and simultaneously released at the centre of a screen-walled greenhouse.

Semi-field experiments were conducted between February and April 2013 in a 7×11 m screenhouse constructed on the grounds of the TOC of ICIPE, Kenya (00⁰25¹S, 34°13¹E). Field studies were conducted between May and June 2013 at Kigoche village, situated near Ahero town, in the Kano plains of Kisumu County, Kenya (00°34′S, 34°65′ E) [10,20]. The area receives between 1,000 and 1,800 mm of rainfall annually with annual temperature and relative humidity (RH) ranges of 17-32°C and 44-80%, respectively. The long rainy season occurs between March and August while short rains are common in October to November. The main economic activity is rice farming which creates numerous mosquito larval habitats resulting in high malaria transmission. Indigenous goats, cattle, poultry, and sheep are also kept in Kigoche [18]. During the night, domestic animals are tethered outdoors adjacent to houses occupied by humans. Many houses in the area are mud-walled with roofs made of corrugated iron sheets or thatch, or without ceiling. Eaves of most houses are open due to the high daytime temperatures [21]. Previous studies reported that the annual Entomological Inoculation Rate (EIR) was 416 and An. arabiensis and An. funestus s.l. the main malaria vectors [10,22,23].

Human foot odour previously shown to be moderately attractive to mosquitoes [24] was collected from nylon socks worn by a Kenyan male (age 31) (Additional file 1: Figure S1). The socks were worn for 24 hours before they were used in the experiment [25]. The volunteer did not smoke, use alcohol, spicy food, perfumes and the last shower was without soap [24,26]. Animal odours were collected from the same individual throughout the experiments by wrapping a clean nylon sock above the knee of a cow or around the leg of a chicken for 24 hours (Additional file 1: Figure S1). For the cow odour sample, a piece of cloth was wrapped over the sock to prevent dirt or faeces from contaminating the odour sample. Clean latex gloves were worn to avoid contamination by other odours. Henceforth, human, cow and chicken skin emanations collected on nylon socks will be termed “human odour”, “cow odour” and “chicken odour”, respectively.

Sugar and molasses were used to produce CO₂ in semi-field and field experiments respectively. Sugar-produced CO₂ was prepared by mixing 250 g sugar (Mumias Sugar Co Ltd, Kenya), 17.5 g yeast (Angel®Company, China) and 2 L water in 5-L containers which would result in an average CO₂ production of 242.3 ± 74.1 ml/min [27]. Molasses-produced CO₂ was obtained by mixing 2 L water, 250 g molasses (Mumias Sugar Co Ltd, Kenya) and 17.5 g dry instant yeast in 5 L containers [18]. Tap water was used during semi-field experiments while all field bioassays were conducted using clean water from Kigoche village. Released CO₂ was delivered through a 60-cm long silicon tubing (0.5 cm diameter) into individual MM-X traps (American Biophysics, North Kingstown, RI, USA) [18]. The MB5 and the SBs used in the current study were separately prepared following protocols described before [10,13]. Socks containing cow, chicken and human odour, and synthetic blends were separately hooked on a wire ring and hung inside the plume tube of a MM-X trap and always supplied with CO₂ from either molasses or sugar. Control traps were baited with CO₂ alone unless indicated specifically. The lower end of the plume tube was suspended 15 cm above ground level [28]. Socks and synthetic blends were placed in glass jars, and stored in a freezer until and between experiments and replaced after four experiments.

All MM-X traps were operated using a 12-V battery. Vaseline pure petroleum jelly was applied on suspension wire bars, electrical cables and CO₂ tubing to prevent ants from preying on mosquitoes caught in the MM-X traps. To terminate experiments, a plug was inserted into the outer tube of the MM-X trap, the CO₂ supply was cut off, and the power disconnected [18]. Traps containing mosquitoes were placed in a refrigerator at −4°C for 10 min. Immobilized mosquitoes were collected, counted, and recorded. Traps were cleaned between experiments using 70% ethanol (to remove residual odours). A manual, hand-held aspirator was used to collect untrapped, free-flying mosquitoes from the screenhouse. The sand-filled floor of the greenhouse was moistened daily to enhance survival of mosquitoes. Latex gloves were worn during experiments to avoid contamination with human volatiles or other odorant compounds.

MM-X traps were placed in all four corners of the screenhouse, and rotated with identical treatments placed at opposite corners of the house. A total of 8 replicates (for a total of 4 nights) were carried out. The treatment combination included: (i) CO₂ vs no stimulus; (ii) cow odour + CO₂ vs clean sock + CO₂; (iii) chicken odour + CO₂ vs clean sock + CO₂; and, (iv) human odour + CO₂ vs clean sock + CO₂.

Randomized 4 × 4 Latin square experimental design was adopted. MM-X traps were placed in all four corners of the screenhouse and treatments rotated for 4 consecutive nights. A total of 16 replicates were carried out. The treatment combination included: (i) only CO₂ and clean sock (control); (ii) cow odour + CO₂; (iii) chicken odour + CO₂; and, (iv) human odour + CO₂.

Randomized 4 × 4 Latin square experimental design was adopted. MM-X traps were placed in all four corners of the screenhouse and treatments rotated for four consecutive nights. A total of 12 replicates were carried out. The treatment combination included: (i) only clean nylon strips without CO₂ (control); (ii) clean nylon strips + CO₂; (iii) Simple Blend (SB: NH3 + Lactic acid + C14, [15]) + CO₂; and, (iv) Mbita blend (MB5: NH3 + Lactic acid + C14 + 3-methyl-1-butanol + Butan-1-amine [13,14]) + CO₂.

Five village houses were selected and experiments were carried out from 18.30 to 06.30 h each night. Randomized 5 × 5 Latin square experimental design was adopted. One MM-X trap was placed at each house and treatments rotated for five consecutive nights. A total of 25 replicates were carried out. The treatment combination included a MM-X trap with CO₂ produced by molasses fermentation and (i) clean sock; (ii) sock with cow odour; (iii) sock with chicken odour; (iv) sock with human odour; and, (v) MB5 blend.

The houses were mud-walled, had open eaves, and corrugated iron sheet roofs and had owner occupants throughout the night sleeping under untreated bed nets. The houses were located at least 25 m apart [29] to exclude the potential interaction of treatments placed in any two adjacent houses. All the baited MM-X traps were hung outside the bedroom window, under the eaves at 15 cm high [17].

Adult mosquitoes were identified morphologically [30] and abdominal status was recorded (Empty (E), blood fed (F), gravid (G)) [31]. Female An. gambiae s.l. and An. funestus s.l. were preserved in 2-ml Eppendorf tubes containing 80% ethanol and a subset (215 fully blood-fed An. gambiae s.l. and 92 unfed An. funestus s.l.) was selected for DNA extraction (Qiagen DNeasy kit) and molecular analysis. Anopheles gambiae species were identified using a multiplex PCR approach as previously described [32], while An. funestus species were determined by PCR amplification, sequencing and phylogenetic sequence analysis of a 380–704 bp fragment of the rDNA gene using primers designed to amplify coding regions flanking the internal transcribed Spacer Region 2 (ITS2) domain [33].

Blood meals were identified using two PCR-based approaches. The first method utilized species-specific primers targeting a fragment (132–680 bp amplicon) of mammalian cytochrome b (cytb) [34]. To ensure sensitive detection of mixed blood meals using this method, DNA from each blood fed mosquito was amplified in individual reactions, containing either a human-, cow-, goat-, pig-, or dog-specific forward primer and a universal reverse primer [34]. PCR amplicons were sequenced and subjected to phylogenetic analysis to verify blood meal origin. Second, to ensure sensitive detection of human blood meals, DNA from each mosquito was amplified using primers designed to target the hypervariable D-loop region of ape mitochondrial DNA [35], sequenced, and subjected to phylogenetic analysis.

To screen for the presence of Plasmodium parasites in field caught mosquitoes, DNA extracted from whole mosquitoes was subjected to nested PCR targeting a 956 bp cytb fragment of the Plasmodium mitochondrial genome [36,37]. All PCR reactions used previously reported cycling conditions and the Roche Expand Long Template PCR system.

Scientific and ethical approval of the present study was granted by the Kenya Medical Research Institute (KEMRI/RES/7/3/1). Consent for houses used in the study was obtained from the household heads and the local administration prior to the start of the study.

A generalized linear model (GLM assuming a Binomial distribution with logit link function) was used to investigate the relative attractiveness of each combination of odours tested in the traps in the semi-field and field experiments, expressed as the number of mosquitoes of one species caught in one of the traps divided by the total number of mosquitoes of that species trapped in all traps during each experimental night [38,39]. The effects of treatment, position of trap or house on mosquito catches were fitted in the model and the non-significants dropped. Models were compared by the Pearson Chi-square value divided by the degrees of freedom. Differences between treatments were tested by pair-wise comparisons with least square differences (LSD) correction [40]. Effects were considered significant at P <0.05. All analyses were performed using IBM SPSS statistical software, version 22.