Experimental design
Mosquito collection
1583 larvae used in this study were obtained from 17 field caught gravid Anopheles arabiensis female collected via mouth aspiration from Kigoche village (00° 34′ S, 34° 65′ E) in the Ahero irrigation scheme, Kenya and transported to the International Centre of Insect Physiology and Ecology (ICIPE)-Duduville campus in Nairobi, Kenya. During collection, a torch was used to locate Anopheles gambiae s.l. indoors on the walls of muddy houses. This was guided by identification protocol illustrated in67; the morphological traits used were resting position, characteristics of wings and abdomen. Abdomens of the collected gravid females were morphologically examined and those observed as engorged, and dark were considered gravid. MB+ females used in these experiments were selected over three different collection timepoints. In September 2022, 1123, gravid field collected female An. arabiensis were screened for presence of Microsporidia MB, 180 were positive resulting in a 16.03% prevalence in the field (5 MB+ females were used for this experiment , offspring n = 404). In November 2022, 399 mosquitoes were screened, 142 were positive for Microsporidia MB, this recorded a prevalence of 35.58% of the symbiont in the field (5 MB+ females, offspring n = 511). In July 2023, 565 mosquitoes were screened, 75 were positive for the symbiont resulting to 13.27% prevalence in the field (7 MB+ females, offspring n = 624). The gravid females were placed in 1.5 ml micro-centrifuge tubes containing 1 cm by 1 cm Whatman filter paper to allow egg laying following the methods described in11,12,13,32,43. After oviposition, they were screened for species ID64 and the presence of Microsporidia MB11,12,32 using PCR.
Larval rearing
Eggs from Microsporidia MB positive and negative female An. arabiensis were separated into larval trays with around 300 ml of deionised water to hatch. In three replicates, stage one (L1) larvae from the same MB+ female An. arabiensis were randomly and in equal number split into four temperature treatments: A total of 444, 515 and 624 L1 larvae were used to set up replicates one, two and three of the experiments. L1 larvae from each MB+ female An. arabiensis were divided into four equal proportions and put in four different larval trays. L1 larvae from MB− female An. arabiensis was also put in a separate larval tray for each of the temperature regimes. We, therefore, had four larval trays per each MB+ and MB− female An. arabiensis, each tray per temperature regime. One MB− female An. arabiensis was used per replicate. This was due to limited space in the incubators. We set temperature 22 °C using insect growth chamber since it supported low temperature settings. Trays for temperature 27 °C were put in an isolated room with control ambient room temperature of 27 °C. We used small incubators to set experiments for temperatures 32 °C and 37 °C, this is because these incubators could only support temperature settings above 30 °C. The number of larvae per tray in the for the MB+ female An. arabiensis were dependent on the amount of offsprings produced by each female An. arabiensis (the data of larvae per tray in each temperature regime has been attached for reference). In MB− female An. arabiensis, 23, 25 and 18 L1 larvae were used per each tray in each temperature regime for replicates one, two and three respectively. The larvae were fed on a pinch of Tetramin baby fish food throughout their development until pupation. We monitored daily larval mortality, rate and date of pupation of each pupa.
Quantification of Microsporidia MB
The ammonium acetate protein precipitation method was used for DNA extraction from offsprings of MB+ female An. arabiensis68,69. Whole pupae were homogenised in 50 µl of phosphate buffered saline (PBS), incubated at 56 °C for 1 h in 300 µl of cell lysis buffer then we precipitated out proteins using 100 µl protein precipitate while incubating the samples in ice for 30 min. The supernatant was centrifuged for 20 min at 14,000 revolutions per minute then transferred to 300 µl of isopropanol, the samples were inverted 100 times to allow the reagents mix before centrifuging at 14,000 revolutions per minute for 1 h to remove excess salt. To obtain a clean DNA, we poured out the resulting supernatant then added 300 µl of ice cold 70% ethanol, inverted the samples 50 times then centrifuged at maximum speed of 14,000 revolutions per minute for 30 min to remove excess salts. The resultant DNA was air dried under the biosafety cabinet overnight before elution in 60 µl of nuclease free water11,12,32,43.
All pupae collected from the experimental group (offspring of MB+ female An. arabiensis) were screened to identify those infected with Microsporidia MB using conventional PCR11,12,32. We measured the Microsporidia MB infection rate in the collected G0 female An. arabiensis and offspring as well and quantified Microsporidia MB density through relative quantification using qPCR. Partial Microsporidia MB 18 s gene region from each DNA sample was amplified using specific 18 s primers (MB18SF: CGCCGG CCGTGAAAAATTTA and MB18SR: CCTTGGACGTG GGAGCTATC)11,12,13,32,43. The gene was then amplified in an 11 µl reaction volume of a mixture containing 0.5 µl of 5 pmol/µl reverse and forward primers, 2 µl HOTFirepol Blend Master Mix Ready-To-Load (Solis Biodyne, Estonia), 6 µl of nuclease-free PCR water and 2 µl of DNA template. The amplification was achieved under the following conditions: initial denaturation at 95 °C for 15 min, denaturation at 95 °C for 1 min for 35 cycles, annealing at 62 °C for 30 s, a further extension for 30 s at 72 °C, and finally, final elongation for 5 min at 72 °C. To quantify the level of infection, samples positive for Microsporidia MB were subjected to relative qPCR analysis using MB18SF/MB18SR primers normalised with the reference host-keeping gene for the Anopheles ribosomal s7 gene (S7F: TCCTGGAGCTGGAGATGAAC and S7R: GACGGGTCTGTACCTTCTGG). Since the ribosomal protein S7 is a highly conserved gene in Anopheles mosquitoes, its expression levels are stable across different conditions and tissues, making it a reliable internal control for qPCR experiments70. The qPCR reaction mixture consisted of 11 µl reaction volume containing 0.5 µl of 5 pmol/µl reverse and forward primers, 2 µl HOT FIREPol® EvaGreen® 416 HRM no ROX Mix Solis qPCR Master mix (Solis Biodyne, Estonia), 6 µl of nuclease-free PCR water and 2 µl of DNA template. The amplification was achieved under the following conditions: initial denaturation at 95 °C for 15 min, denaturation at 95 °C for 1 min for 35 cycles, annealing at 62 °C for 60 s, and a further extension for 45 s at 72 °C. The PCR was carried out in a proflex cycler, and the qPCR was carried out in a MIC qPCR cycler (BioMolecular Systems, Australia). The MB18SF/MB18SR primers were used to confirm samples with the characteristic Microsporidia MB melt curve11,12,13,32,43.
Statistical analysis
We analysed the pupation rate and age at death using Mixed-Effects Cox Models and the R “coxme” package71. The mean development time for the pupated larvae was analysed using the linear mixed-effects model using the “lme4” package. We analysed the infection rate and Microsporidia MB intensity using binomial and gaussian logistic mixed-effect model (GLMMs) and glmmTMB package. In all models, the temperature treatments, the G0 female An. arabiensis‘ infection status, and their interactions were included as fixed terms, and the time of capture in the field was included as a random effect. In addition, the development time model also looked at the interaction between temperature treatments and infection status in offspring (Microsporidia MB negative offspring coming from un-infected colonized female An. arabiensis, Microsporidia MB positive offspring coming from field-collected infected G0 female An. arabiensis and Microsporidia MB negative coming from field collected infected G0 female An. arabiensis). Individuals that pupated were excluded from the age-at-death analysis. Individuals who died were excluded from the development time and infection status analysis. The Microsporidia MB intensity analysis (log transformed for better data visualisation) excluded uninfected pupae, and we used temperature treatments and transmission groups (0–33%, 33–66%, or 66–99% transmission from mother to offspring) as interaction terms in the model. We used the Tukey post-hoc test and “means” function to perform multiple comparisons among the infection status and temperature treatments72. P values for comparisons among treatments have been stated before the overall p values for each model done. Statistical analysis was performed using R statistical software version 4.1.2 and R Studio73.
Modelling the Microsporidia MB dissemination potential
After obtaining experimental data on infection rates, development, and survival, we used these parameters to develop a mathematical model predicting Microsporidia MB dissemination in Anopheles arabiensis populations under different temperature conditions. To express the probability that an L1 offspring coming from MB+ female An. arabiensis is infected, survives to age x, and pupates at age x given temperature T, we combined the conditional probabilities:
P(infected (cap) survives to age × (cap) pupates at x| T) = P(infected| T). P (survives to age x |infected, T). P (pupates at x| infected, T).
$$P(infected| T) = frac{{{text{Number of infected larvae at temperature }}T{ }}}{{{text{Total number of larvae at temperature }}T}}$$
$$P(survives to age x |infected, T) = frac{{{text{Number of infected larvae that survive to age }} times {text{ at temperature }}T{ }}}{{{text{Total number of larvae at temperature }}T}}$$
$$P(pupates at x| infected, T) = frac{{{text{Number of infected larvae that pupate to age }} times {text{ at temperature }}T{ }}}{{{text{Total number of larvae that survive to age }} times {text{ at temperature }}T}}$$
Using the Gaussian function, the probability is given by:
$${mathbb{P}}left( {{text{T}},{text{x}}} right) = P(infected cap survives to age x cap pupates at x| T) = Ae^{{ – frac{{left( {x – mu } right)^{2} }}{{2sigma^{2} }}}} ,$$
(0 < x < infty).
This formula considers the conditional dependencies based on infection status and temperature, providing a logical path to estimate the combined probability.
A continuous logistic model was chosen to provide a smooth and accurate representation of mosquito population growth, reflecting natural, gradual changes without the constraints of fixed time intervals required by discrete models. This continuous approach allows precise population estimates at any point in time, making it ideal for understanding temporal growth rates and incorporating stochastic variability to reflect environmental influences on fecundity. The logistic growth equation:
$$frac{{{text{dN}}left( {text{t}} right)}}{{{text{dt}}}} = {text{F}}.{text{r}}.{mathbb{P}}left( {{text{T}},{text{x}}} right) cdot {text{N}}left( {text{t}} right)left( {1 – frac{{{text{N}}left( {text{t}} right)}}{K}} right)$$
was used to model the population growth of infected individuals, where N(t) is the number of MB+ individuals at time t, F represents the fecundity, r the sex ratio, ({mathbb{P}}left( {T,x} right)) the probability of infection, survival, and pupation under temperature T, and K the carrying capacity74,75. The carrying capacity was set to 1000 to simulate real-world limitations such as resource and space constraints, establishing a stable population maximum that aligns with natural conditions. Additionally, targeting a population of 1000 MB+ offspring provides a measurable endpoint for assessing the spread of Microsporidia MB within mosquito populations. The solution to this equation,
$${text{N}}left( {text{t}} right) = frac{K}{{1 + left( {frac{{K – {text{N}}_{0} }}{{{text{N}}_{0} }}} right){text{e}}^{{ – {text{F}}.{text{r}}.{mathbb{P}}left( {{text{T}},{text{x}}} right).t}} }}$$
enabled us to estimate the rate at which the population of MB+ offspring increases from an initial population of 10 MB+ female An. arabiensis, with the goal of reaching a target population of 1000 MB+ individuals.
In our deterministic simulation, parameters such as: F (fecundity), r (sex ratio), K (carrying capacity), and (N_{0}) (initial population) remained constant. Fecundity was set at three fixed rates (33, 66, or 99 viable eggs per female An. arabiensis) based on observed averages, providing a baseline for population growth under stable conditions. The sex ratio male: female was considered to be 1:1. Details of the stochastic simulation are provided in the supplementary material.
To implement this methodology, we used Python for all data processing, simulations, and statistical computations. Python’s libraries, including numpy for numerical operations, scipy for probability computations and fitting, and matplotlib for visualization, were integral to generating plots, calculating probabilities, and fitting model parameters.