Ivermectin-treated cattle reduces blood digestion, egg production and survival of a free-living population of Anopheles arabiensis under semi-field condition in south-eastern Tanzania

The study was conducted within the semi-field systems (SFS) at the Ifakara Health Institute (IHI) in the Kilombero Valley, south-eastern Tanzania [64]. Although three major African malaria vectors (An. gambiae s.s, An. funestus and An. arabiensis) are found in this valley [65‐69], An. arabiensis that composes >90% of the population in most of the villages contributes to outdoor transmission of malaria [3, 51]. Throughout this valley, majority of cattle shelters are kept at 2.5–30 m from human dwelling houses (Lyimo et al., unpublished data). The previous analysis of blood fed mosquitoes revealed that 80% of An. arabiensis obtain blood from cattle and rest outside houses [51], but this exophily and zoophagy has genetic basis [61].

Experiments were conducted using a self-sustaining population of An. arabiensis surviving within a large SFS (21 × 9.1 × 7.1 m) at the IHI since 2008 [62, 70]. This population was established from individuals of wild mosquitoes collected from a nearby Sagamaganga village (~15 km from IHI). Within this system, mosquitoes have access to a range of natural habitat including cattle blood sources, larval habitat, vegetation and animal shelters. These mosquitoes express similar patterns of larval development, mating, feeding and resting behaviour [62]; genetic and phenotypic diversity; and reduced inbreeding as the wild population in the field [60, 62].

Livestock owners from nearby local communities surrounding IHI compound provided their cattle for these experiments. Cattle were collected after the purpose of the experiments was explained to livestock owners, and those accepted for their animals to participate in the trial were required to fill the written informed consent. Majority of cattle (>88%) in Kilombero Valley are sprayed with irritant, repellent pyrethroids, especially Alphacypermethrin (Lyimo et al., unpublished data). Therefore, cattle that had no history of being sprayed with any insecticide for the past 2–3 months were chosen for these experiments. To avoid residues of pyrethroids on cattle, all cattle were transferred to the IHI compound, and washed with water for 3 days consecutively prior to the start of experiments.

A total of 6 cattle were divided into two groups of 3 individuals: untreated (control) and ivermectin-treated group. The weight of individuals in treated group was estimated by veterinarians using girth tape, and then subcutaneously injected with commercially available 1% ivermectin solution (IVOMEC^®) at a therapeutic dose of 0.2 mg/kg body weight. Another group of three individuals remained untreated as control during these experiments.

The effects of ivermectin-treated cattle, the post-treatment time, and their interaction on the continuous (i.e. mass of haematin defecated, and number of eggs), and binomial response variable (proportion laid eggs) of female An. arabiensis were respectively analysed by fitting generalized linear mixed effect model (glmer) with Poisson (log link) and binomial (logit link) errors in the lme4 statistical package in R version 3.1.1 [81]. Firstly, the effects of ivermectin-treated cattle on the response variables were assessed by using model with cattle ‘individuals’, and ‘experimental night’ as random effects, and ‘Treatment’ as the main effects. The null model built with ‘random effects’ was compared with full model composed of both ‘random effect’ and ‘main effects’ using ‘ANOVA’ to identify statistical significant effects of ‘Treatment’. Lastly, the effects of ivermectin-treated cattle ‘Treatment’ on the response variables across time were analysed by sequential addition of the main effects including ‘Treatment’, ‘post-treatment time’ and their interaction into a null model (forward selection). Then likelihood ratio test (LRT) was used to identify if the addition of main effects into a null model lead into a statistical significant improvement of the model. If the interaction between ‘treatment’ and ‘post-treatment time’ was significant, then the main effects of treatment was analysed for each post-treatment time to establish the time point with statistical significant impact on response variables.

The continuous response variable of survival of mosquitoes (number of days survived) after blood feeding on untreated or ivermectin-treated cattle was analysed using survival package in the statistical software of R version 3.1.1 [81]. The Cox Proportional Hazards Model (coxph) in the survival package compares between survival curves using Hazard Ratio (HR). In this model, a frailty function was used to incorporate the random effect of ‘individual cows’ or ‘replicate’ to form null model. Therefore, the main effect of ‘Treatment’ was added into a null model, and tested which random effect has statistical significant improvement of the model. The full model was used to estimate hazard ratio (risk of death) to identify statistical significant differences between Kaplan–Meier survival curves. These survival curves were used to generate median survival times of mosquitoes after blood feeding on ivermectin-treated or untreated cattle. Further analysis of survival data was conducted to investigate how the impact of ivermectin on the survival of mosquitoes changes over time. The Cox Proportional Hazard Model (coxph) was composed of frailty function to incorporate the random effect of ‘individual cows’ or ‘block’, ‘treatment, ‘post-treatment time’, and their interactions (treatment*post-treatment time) were fit as main effects in R statistical software. These terms were sequentially added into the null model, and tested if there is statistical significant improvement. When the interaction term of ‘treatment*post-treatment time’ was statistically significant, the Cox Proportion Hazard Models (coxph) were fit with main effect of treatment and random effects of ‘individual’ for each post-treatment time to identify statistical significant differences between survival curves of mosquitoes after blood feeding on untreated or ivermectin-treated cattle. Then the Kaplan–Meier survival function was used to generate survival curves, and to estimate median survival times for each post-treatment time.

Blood meal digestion as estimated by the mass of haematin defecated by An. arabiensis was significantly influenced by the treatment of cattle with ivermectin (\(\upchi_{1}^{2}\) = 27.26, P < 0.01, Fig. 2), post-treatment time (\(\upchi_{6}^{2}\) = 2077.16, P < 0.01, Fig. 2), and their interactions (ivermectin treatment*post-treatment time: (\(\upchi_{7}^{2}\) = 594.45, P < 0.001, Fig. 2). The efficiency of blood meal digestion in mosquitoes fed on ivermectin-treated cattle was significantly reduced by half compared to the mosquitoes fed on control cattle at 6 days post-treatments (\(\upchi_{1}^{2}\) = 4.06, P = 0.04, Fig. 2), and the other post-treatment time shown a non-significant decreasing trend (P > 0.05, in all cases, Fig. 2).

The probability that An. arabiensis laid eggs after blood feeding was significantly affected by ivermectin-treated cattle (\(\upchi_{1}^{2}\) = 9.35, P < 0.001, Fig. 3a), post-treatment time (\(\upchi_{6}^{2}\) = 30.82, P < 0.001, Fig. 3b), and their interactions (treatment*post-treatment time: \(\upchi_{8}^{2}\) = 36.99, P < 0.001, Fig. 3b). The oviposition rates of An. arabiensis were significantly reduced by 64.61% after blood feeding on ivermectin-treated cattle relative to those fed on control cattle, but it changed across time (Fig. 3a, b). The oviposition rates of An. arabiensis fed on ivermectin-treated cattle was reduced relative to those mosquitoes fed on control cattle by 54.64% at day 3 (\(\upchi_{1}^{2}\) = 5.58, P = 0.02), 74.14% at day 6 (\(\upchi_{1}^{2}\) = 5.96, P = 0.01), 76.87% at day 9 (\(\upchi_{1}^{2}\) = 5.49, P = 0.02) and 81.62% at day 12 (\(\upchi_{1}^{2}\) = 6.95, P < 0.01) post-treatments (Fig. 3b), but the effects gradually decreased up until 15 days post treatment (P > 0.05, Fig. 3b).

Similarly, the fecundity of An. arabiensis was significantly influenced by the treatment of cattle with ivermectin (\(\upchi_{1}^{2}\) = 13.25, P < 0.001, Fig. 4a), post-treatment time (\(\upchi_{6}^{2}\) = 18.28, P < 0.01, Fig. 4b), and their interactions (treatment*post-treatment time: \(\upchi_{8}^{2}\) = 23.60, P < 0.01, Fig. 4b). The ivermectin-treated cattle reduced fecundity of mosquitoes by 62.98% relative to mosquitoes that fed on control cattle (Fig. 4a). Over time, the ivermectin-treated cattle reduced fecundity of An. arabiensis relative to control cattle by 60.89% at day 3 (\(\upchi_{1}^{2}\) = 5.88, P = 0.01), 74.71% at day 6 (\(\upchi_{1}^{2}\) = 6.75, P < 0.01), 75.20 at day 9 (\(\upchi_{1}^{2}\) = 6.20, P = 0.01) and 79.67% at day 12 (\(\upchi_{1}^{2}\) = 6.98, P < 0.01) post treatments, but the reduction declined slowly up until 15 days post-treatments (P > 0.05, in each case, Fig. 4b).

The long-term survival of An. arabiensis was significantly influenced by the ivermectin-treated cattle (\(\upchi_{1}^{2}\) = 142.54, P < 0.001, Fig. 5), post-treatment time (\(\upchi_{7}^{2}\) = 101, P < 0.001, Fig. 5), and their interactions (treatment*post-treatment time: (\(\upchi_{8}^{2}\) = 87.67, P < 0.001, Fig. 5). The long-term survival of An. arabiensis was significantly reduced by 52.53% after blood feeding on ivermectin-treated relative to mosquitoes that fed on control cattle (\(\upchi_{1}^{2}\) = 62.87, P < 0.001, Fig. 5a). The proportion of surviving An. arabiensis and their median survival times after blood feeding on ivermectin-treated cattle was significantly reduced relative to mosquitoes that fed on control cattle (Fig. 5a–c; Tables 2, 3), but they remained lower than on control cattle until at day 21 post-treatments (Fig. 5d–h; Tables 2, 3). Furthermore, the daily mortality rates of mosquitoes that fed on ivermectin-treated cattle increased from 50 to 80% relative to mosquitoes that fed on control cattle, and then slowly declined up until 21 days post-treatment. The ivermectin-treated cattle increased risk of death of mosquitoes by five-folds relative to control cattle (P < 001 in most cases, Tables 2, 3; Figs. 5, 6), but such risk gradually decreases with time up until 21 days (P > 0.05, Tables 2, 3; Figs. 5, 6).