This study was implemented in four phases in the field and semi-field environments. The field studies were done in five villages in Ulanga district, Kivukoni (– 8.2021°S, 36.6961°E), Minepa (– 8.1455°S, 36.4244°E), Tulizamoyo (− 8.3669°S, 36.7336°E), Kilisa (– 8.3721°S, 36.5584°E) and Ruaha (– 8.9068°S, 36.7185°E) and two villages in Kilombero district, Sululu (– 7.9973°S, 36.8317°E) and Ikwambi (– 7.9833°S, 36.8184°E), in south-eastern Tanzania (Fig. 1). The principal malaria vectors in the area are An. arabiensis, and An. funestus [40, 41, 46, 47]. Several other anophelines and some culicine mosquito species are also present in the study area. The semi-field experiments were conducted inside large screen house compartments at the Mosquito City facility maintained by Ifakara Health Institute in Kining’ina village (– 8.1080°S, 36.6668°E) in Kilombero district. These semi-field systems were designed to mimic the natural environment, and have in-build experimental huts for controlled mosquito studies .
The four study phases included: (i) assessment of indoor resting densities of male Anopheles mosquitoes relative to females in different house types, (ii) observations of insemination status of wild caught female mosquitoes at different times of night inside volunteer-occupied experimental huts in the village (iii) field observations to assess whether wild mosquitoes’ mate before or after entering local houses occupied by residents and (iv) semi-field observations of wild-caught mosquitoes inside large screen house chambers to verify and quantify the insemination rates inside and outside experimental huts under controlled settings.
Observations of male mosquitoes resting inside different types of houses
Mosquito collections were conducted in 80 houses in the four study villages (Fig. 1), targeting four common house types, namely: (i) 20 houses with thatched-roofs and mud-walls; (ii) 20 houses with thatched-roofs and un-plastered brick-walls; (iii) 20 houses with metal-roofs and un-plastered brick-walls; and (iv) 20 houses with metal-roofs and plastered brick walls. Inside the houses, mosquitoes were collected from various potential resting surfaces using Prokopack aspirators. Initially, mosquito collections were conducted from the morning to noon (6:00 a.m. – 12:00 noon), but additional collections were done in the mornings (7:00 a.m. – 8:30 a.m.), evenings (6:00 p.m. – 8:00 p.m.) and late nights (12:00 a.m. – 2:00 a.m.), as previously described by Msugupakulya et al .
Observations of insemination in wild mosquitoes caught inside volunteer-occupied experimental huts
This experiment was conducted in Tulizamoyo village using three volunteer-occupied experimental huts made (Fig. 2). These tent-styled huts consisted of easy-to-seal eave spaces and screened windows, and were located near other village houses used by residents (Fig. 2). During the study, the eave spaces remained open during the day time (7:00 a.m. – 6:00 p.m.) to allow mosquito entrance, and closed at 6 p.m. Trained volunteers entered each experimental hut just before the huts were closed at 6 p.m. and collected mosquitoes using a Prokopack aspirator from multiple indoor surfaces for a total of 5 minutes each time. Without re-opening the huts after the 6 p.m. closure, follow up collections were done at 11 p.m. and again at 6 a.m. the following morning, after which the huts were finally re-opened. At each of these time points, the sampling was done from multiple locations inside the huts. In between the collections, the volunteers slept under untreated nets inside the huts.
It was hypothesized that proportions of inseminated females would stay either similar (if no additional mating happened indoors after 6 p.m. hut closures) or increase (if there was additional mating after 6 p.m.). Collected mosquitoes were kept in labelled cups and transferred to the laboratory for morphological identification and assessment of their insemination status.
The sampled female mosquitoes were firstly identified using morphological keys for Afrotropical Anopheles mosquitoes [50, 51]. An. funestus and An. arabiensis were dissected under a stereo microscope. The seventh segment of immobilized mosquitoes was dissected to extract spermathecae, which was examined under a light microscope with a 10× magnification lens for insemination status. Female mosquitoes with filled long threads coiled brown spermatheca were considered as inseminated while those with clear and non-striated spermathecae were considered as non-inseminated . This study was conducted for 14 consecutive nights in the first round.
After the first round, this experiment was repeated for another ten consecutive nights using the same procedures as described above, except this time all mosquitoes were immobilized by freezing in portable cold-boxes immediately after collection to avoid any possible mating that could happen inside the holding cups.
Observation of insemination in wild female mosquitoes caught from local houses in the study village
An additional experiment to assess the insemination of malaria vectors was performed for five nights using natural houses where people live in the study village. Three thatched-roofed and three iron-roofed houses, all of which were occupied by natives, were used for this experiment. Mosquitoes were collected using the Prokopack aspirators and immobilized immediately in the cooler box with ice packs to prevent any mating potentially happening inside the holding cups and during transportation. These wild-caught mosquitoes were then transported to the insectary and females assessed for their insemination status as described above.
Observations of insemination in wild-caught mosquitoes maintained under semi-field conditions
The semi-field system consisted of large multi-chambered screen houses with netting walls, enclosing village-like ecosystems with vegetation and water puddles . Each chamber (9.6 m x 9.6 m) had an experimental hut constructed to mimic the design of typical local houses used in rural Tanzania (Fig. 3). Three chambers were used for this study. The experimental huts in the selected chambers were completely sealed with mosquito netting on eaves spaces, but the windows were fitted with window-exit traps to catch any mosquitoes attempting to exit the huts. This way, mosquitoes inside the huts could attempt exit (and be trapped in the window exit traps) but those outsides could not enter the huts (since all eave openings were screened, doors closed and windows covered with the exit traps). For additional control, the doors were fitted with overlapping net curtains to prevent mosquitoes from flying in or out whenever someone entered.
Field collection of 3rd and 4th instar Anopheles larvae were done using a combination of standard 350ml dippers (for small habitats) and ten-litre buckets (for large habitats) to maximize densities as previously described by Nambunga et al . The larvae were sorted and only Anopheles used for further observations. The larvae were maintained in rearing basins and fed daily on TetraMin® fish food until they reached the pupae stage.
For each batch of field-collected mosquitoes, the pupae were divided into two approximately equal-sized groups, one of which was put inside the experimental hut in the semi-field chamber, and the second group placed outside the hut. No sex separation of the mosquitoes was done, so each group of pupae yielded both male and female virgins. This experimental set-up was replicated in three different semi-field chambers, each with a similar-sized hut. The field collections were repeated weekly for twelve months, each time adding approximately half of the pupae inside and half outside the huts in each of the three chambers. Emergent adult mosquitoes were recaptured twice weekly from inside and outside the huts (using human landing catches) as well as in the exit traps fitted to the hut windows. The recaptured female mosquitoes were kept in labelled cups and immediately assessed for insemination by dissecting and observing the spermatheca as described above. The observations were done immediately after collections, to minimize the likelihood of mosquitoes mating inside the collection cups.
Molecular identification of An. funestus sibling species
A sub-sample of An. funestus collected from the field and those emerged adults in the semi-field experiments were packed individually in the microcentrifuge tubes with silica gel inside. The samples were transported to the molecular laboratory at Ifakara Health Institute for sibling species identification using polymerase chain reaction (PCR) as described by Koekemoer et al .
Data were analysed using open-source statistical software, R version 3.6.0 . Proportions and means were used for initial descriptive statistics of mosquitoes in different categories and mean outcomes calculated for each explanatory variable. In the first field survey, generalized linear mixed effect models (glmer), with a negative binomial distribution to account for over-dispersion, were used to assess proportions and variations of male mosquitoes collected inside different house types and at different collection times, i.e., early-morning (7 a.m. – 9 a.m.), late-morning (9 a.m. – 12 p.m.), evening (6 p.m. – 8 p.m.) and midnight (12 a.m. – 2 a.m.). Additionally, pairwise comparisons were performed using Tukey’s Honestly Significant Difference Test using the multcomp package in R.
To compare proportions of inseminated mosquitoes, glmer models with binomial distribution and logit functions were used. For the observations of wild mosquitoes in the village, the mean proportion of inseminated females was modelled as a function of time of mosquito sampling, i.e., evening (6 p.m.), night-time (11 p.m.) and early-morning (6 a.m.). The sampling date and hut identifier were included as random variables in the model. For observations of wild-caught mosquitoes observed in semi-field captivity, the mean proportion of females inseminated was modelled as a function of the location of mosquito recapture, i.e., inside hut, outside hut or in the window exit traps. Again, the sampling date and semi-field chamber identifiers were included as random variables. Results of the models are presented as relative rate ratios (RR) with 95% confidence intervals and their associated p-values.