Study sites
This study was conducted at two different regions in Tanzania: the urban Dar es Salaam and the rural village within the Kilombero valley in the Morogoro region. Dar es Salaam is the largest City of Tanzania, situated at 6° 51′S, 39° 18′E along the Indian Ocean with an estimate of 5 million people according to the national census of 2012 [35]. A detailed description of the study area has been previously published elsewhere [36, 37]. The main malaria vectors are An. gambiae s.s. and An. arabiensis, but An. merus and An. funestus s.s. are also available, though existing in very low numbers throughout the year [38]. An. gambiae s.s. which is often regarded as the most anthropophagic vector (rely feeding heavily upon human blood), feeds predominantly at the midnight [36, 39]. In contrast, its sibling species, An. arabiensis, which is commonly referred to as zoophagic (prefer feeding on cattle) mosquitoes throughout SSA [22], starts actively feeding in the early evening and mainly outdoors, time which coincides with the period when most residents of this city are still outside [36, 39]. This overlaps overtime, and outdoor space between mosquito and human activity potentially increases the risk of human exposure to malaria transmission, which cannot be effectively addressed by using indoor-targeted interventions such as LLINs [39]. During this study, human Plasmodium falciparum malaria infection was around 10% among residents in all age groups [37], and with the strong reduction in malaria vectors densities of An. gambiae complex and An. funestus group [40]. This was achieved due to the scaling-up of larvicides [41] and LLINs [36, 37]. The scaling-up of larvicides and LLINs coincided spontaneously with the wide use of window screening across the city of Dar es Salaam [40]. The average annual rainfall ranging from 800mm to 1300mm with a 25ºC annual temperature [42].
The second study site was at Kilombero valley, Lupiro village (8°23’03.8” S, 36°40’26.7” E), which is located 40 km south of Ifakara town within the Kilombero Valley, south-eastern Tanzania [43]. The detail of an area can be found elsewhere [13, 43]. The area is located at 300 meters above sea level on the floodplains of Kilombero valley. The average annual rainfall ranges between 1200 to 1800 mm between December to May, and the temperature is recorded at ranges from 20 to 32.6°C. The most resident lives on subsistence farming of rice, fishing, and sparse livestock keeping. An. arabiensis and An. funestus group are the primary malaria vectors in the area, but the latter exist in relatively very low numbers throughout the year [12]. The historically-important malaria vector An. gambiae s.s. had been virtually eliminated, following the widespread use of LLINs [13].
Experimental design
Eight Ifakara Tent Trap version C (ITT-C) [39] baited with either humans or calves were simultaneously used to catch wild malaria vectors in urban Dar es Salaam and rural Kilombero Valley. In each site, an open field ground measuring more than 500m long was selected. Four (human versus calf) pairing catching stations, spaced about 50m apart from each other, were established within these field grounds. Within each pair, the host was spaced 5m apart, allowing for a competitive host preference assay. A Latin square design involving the movement of trap-host combinations between positions was implemented to minimize possible biases associated with each position and natural variations in individual hosts' attractiveness to mosquitoes [44, 45]. Each pair was rotated after each experimental night through four stations. Four nights were required to make a complete round of experimentation (Figure 1). After each round of four nights, the actual human volunteers and calves were replaced. The calf within each ITT-C was tethered to lure the mosquito entry inside the trap. Each morning, calves were taken out of the tent for daily grazing. There was no exchange of host between traps (calf-baited versus human-baited) because it was not acceptable to expect human participants to sleep in traps soiled by a calf. Trapping was conducted from 19:00 hours to 06:00 hours, and trapped mosquitoes were emptied from the trap every morning using a mouth aspirator. The details on how to empty mosquitoes inside the ITT-C can be found in the previous article [39]. In urban Dar es Salaam, 104 (60 nights between May to August 2009 and 44 nights between March and June 2010), experimental nights were conducted. In rural Kilombero Valley, only 16 nights (from August to September 2010) was conducted. It took longer in Urban Dar es Salaam due to the limited number of malaria vector densities.
Position of Figure 1
Mosquito identification
Every morning, trapped adult mosquitoes from each trap were collected by mouth aspirator, placed in a respective paper cup prior labelled according to the host, and killed using chloroform. Morphological identification was conducted based on the keys of Gillies and Coetzee [46]. All collected An. gambiae s.l. were stored individually in Eppendorf tubes (1.5 ml) with silica gel desiccant and cotton before transport for Polymerase chain reaction (PCR) assay for species identification. The field-collected data were recorded and linked with laboratory results using the designated forms adapted from Kiware et al. [47].
Statistical analysis
Statistical analyses were carried out using the R statistical software version 3.6.1, augmented with the matrix, lattice, and lme4 packages. To test the effect of species-specific on attacking human host, only PCR confirmed individuals from the An. gambiae complex (An. gambiae s.s and An. arabiensis) were used. Because the response variable for each species is binary (that is, an individual mosquito can only attack a single host at a time and not both), a Generalized Linear Mixed Effect Models (GLMMs) [48], using binomial distribution and logit link function, was applied. The proportion of mosquitoes caught attacking humans was treated as the response variable, with a variable combination of PCR confirmed species and sites (PCR confirmed species*Sites) as a fixed effect. The experimental night and stations were fitted as a random effect. The model was run first without fitting an intercept so that the absolute proportion of mosquitoes attacking the human for each species and from each site can be estimated and compared. This was followed by fitting models that included intercept to obtain the contrast in human feeding preference between species with An. gambiae s.s in urban Dar es Salaam treated as a reference species in the model. This detailed statistical analysis on the effect of species on the propensity of attacking upon human host species was restricted to An. arabiensis and An. gambiae s.s., partly because of their importance in driving malaria transmission in these settings, and their number captured was sufficient to detect the effect.