Comparison of infectivity of Plasmodium vivax to wild-caught and laboratory-adapted (colonized) Anopheles arabiensis mosquitoes in Ethiopia

Background Mosquito-feeding assays that assess transmission of Plasmodium from man-to-mosquito typically use laboratory mosquito colonies. The microbiome and genetic background of local mosquitoes may be different and influence Plasmodium transmission efficiency. In order to interpret transmission studies to the local epidemiology, it is therefore crucial to understand the relationship between infectivity in laboratory-adapted and local mosquitoes. Methods We assessed infectivity of Plasmodium vivax-infected patients from Adama, Ethiopia, using laboratory-adapted (colony) and wild-caught (wild) mosquitoes raised from larval collections in paired feeding experiments. Feeding assays used 4–6 day-old female Anopheles arabiensis mosquitoes after starvation for 12 h (colony) and 18 h (wild). Oocyst development was assessed microscopically 7 days post-feeding. Wild mosquitoes were identified morphologically and confirmed by genotyping. Asexual parasites and gametocytes were quantified in donor blood by microscopy. Results In 36 paired experiments (25 P. vivax infections and 11 co-infections with P. falciparum), feeding efficiency was higher in colony (median: 62.5%; interquartile range, IQR: 47.0–79.0%) compared to wild mosquitoes (median: 27.8%; IQR: 17.0–38.0%; Z = 5.02; P < 0.001). Plasmodium vivax from infectious individuals (51.6%, 16/31) infected a median of 55.0% (IQR: 6.7–85.7%; range: 5.5–96.7%; n = 14) of the colony and 52.7% (IQR: 20.0–80.0%; range: 3.2–95.0%; n = 14) of the wild mosquitoes. A strong association (ρ(16) = 0.819; P < 0.001) was observed between the proportion of infected wild and colony mosquitoes. A positive association was detected between microscopically detected gametocytes and the proportion of infected colony (ρ(31) = 0.452; P = 0.011) and wild (ρ(31) = 0.386; P = 0.032) mosquitoes. Conclusions Infectivity assessments with colony and wild mosquitoes yielded similar infection results. This finding supports the use of colony mosquitoes for assessments of the infectious reservoir for malaria in this setting whilst acknowledging the importance of mosquito factors influencing sporogonic development of Plasmodium parasites.


Background
With the move towards malaria elimination and eradication, new tools and strategies to reduce onward transmission of Plasmodium infections, including transmission-blocking interventions (TBI), are considered highly beneficial [1,2]. An increasing number of drugand vaccine-based TBI are in the pipeline [3] and will require monitoring tools for efficacy. Additionally, it is considered highly beneficial to characterize the human infectious reservoir for malaria in low endemic settings approaching elimination, to better target and monitor TBI [4,5]. Both TBI evaluation and infectious reservoir characterization require robust tools to measure human infectivity to mosquitoes. Mosquito-feeding assays can directly assess Plasmodium transmission from man-tomosquitoes and play central role to estimate efficacy of TBI and the assessment of the infectious reservoir [6].
Maintenance of insects in artificial breeding conditions favors accumulation of traits that favor survival in the new environment, resulting in a change in genetic make-up over generations [25]. Parasite-mosquito combinations and their susceptibility to malaria infection are regulated at multiple steps during the development of the parasites [26] and numerous factors may modulate this interaction. These factors range from mosquito genetics [27,28] and immune system [29] to parasite polymorphisms that allow evasion of the mosquito immune system [30]. Environmental factors such as midgut microbiota [31,32], mosquito larval diet [33,34], and temperature to support sporogony [35] are also implicated. These findings emphasize that infectivity studies from colony mosquitoes might not represent the infectivity in natural settings and therefore, assessment of the relative permissiveness of colony and wild mosquitoes could assist in the interpretation of mosquito-feeding assays to the local context. In this study, the relative permissiveness to Plasmodium vivax infection of colony and wild Anopheles arabiensis mosquitoes was assessed in paired experiments.

Study site, immature mosquito stages collection and rearing
Data was collected from September 2018 to February 2019 in Adama, Ethiopia (formerly called Nazareth), a city located within the Great Rift Valley, with an average elevation of ~ 1624 meters above sea level. Extensive irrigation activities characterize the area surrounding Adama with an annual peak malaria transmission season occurring between September and November [5,36]. Both P. falciparum and P. vivax are endemic; the latter contributes towards ~ 60% of the cases [5,37].
Immature mosquito stages (larvae/pupae) were collected by standard dipping method from potential breeding sites located at ~ 35 km from the city, close to a hot spring resort (Sodere, 8°24′N, 39°23′E, at an altitude of 1360 meters above sea level) [38]. The breeding habitat is located at a publicly accessible site where there are temporary/permanent puddles made of rock pool or pools in a grassy area (Additional file 1: Figure S1) emanating from a natural hot spring sources which exist throughout the year and form a marshy area. The collected larvae, transported in plastic jars to the field laboratory, were maintained in plastic trays in the original water collected from the breeding sites and provided with fish food (Cichlid Sticks; Tetra, Maidenhead Aquatics, Leicester, UK). Pupae were picked in glass beakers containing sedimented water from the breeding sites and kept in cages until emergence to adults. Adult female An. arabiensis mosquitoes were identified morphologically using standard keys [39,40]. Colony mosquitoes (> 800th generation) were reared to adulthood as described previously [5]. Mosquitoes of both sources were maintained at the same laboratory settings; developmental stages were Conclusions: Infectivity assessments with colony and wild mosquitoes yielded similar infection results. This finding supports the use of colony mosquitoes for assessments of the infectious reservoir for malaria in this setting whilst acknowledging the importance of mosquito factors influencing sporogonic development of Plasmodium parasites. Keywords: Wild mosquito, Anopheles arabiensis, Plasmodium vivax, Membrane-feeding, Infectivity, Relative permissiveness reared using fish food (Cichlid Sticks, Tetra) and adult mosquitoes were maintained on sucrose solution (10%) at ambient conditions at temperatures of 26-30 °C and a relative humidity of 60-80% before and after feeding.

Membrane-feeding assays
Venous blood samples (5 ml) were collected after obtaining informed written consent from patients with microscopy-confirmed P. vivax infection attending the Adama Malaria Clinic. Blood collected in lithium heparin tubes (Vacutainer; BD, Oxford, UK) was offered to colony and wild An. arabiensis mosquitoes in parallel using membrane-feeding apparatus as detailed previously [7]. Briefly, 5-6 day-old female mosquitoes were starved for 12 h (colony) and 18 h (wild) before feeding. This timing was decided upon following pilot experiments where aggressiveness was unfavorable for wild mosquitoes after 12 h starvation. We have observed a positive association between starvation time and feeding efficiency (ρ (41) = 0.352; P = 0.024); 18 h was considered appropriate for wild-caught mosquitoes with sufficient numbers of fully fed mosquitoes and minimal mortality. Feeding was performed in the dark for 25 min using water-jacketed glass-feeders (mini-feeder; Coelen Glastechniek, Arnemuiden, the Netherlands) that were covered with an artificial membrane (parafilm) and connected to a circulating water bath (Julabo GmbH; Seelbach, Germany) maintained at 38 °C. Unfed and partially-fed mosquitoes were removed from the holding cages, leaving fully-fed mosquitoes undisturbed. Fully-fed mosquitoes were maintained for 7 days under the same laboratory condition using 10% sucrose solution. At least 10 mosquitoes were dissected, and oocyst presence was assessed microscopically after staining with 1.0% mercurochrome (Sigma-Aldrich, Taufkirchen, Germany). This minimum number was mainly determined by the feeding efficiency and availability of wild mosquitoes. Asexual parasite and gametocyte densities were quantified in thick blood films, screening against 1000 leukocytes.

Mosquito genotyping
A representative set of wild and colony mosquitoes were genotyped using multiplex polymerase chain reaction targeting the intergenic spacer gene of the ribosomal DNA of all cryptic species in the An. gambiae complex as described previously [41], with a few modifications. All conditions, including primers, were as per the original protocol except that the MgCl 2 concentration was increased to 2 mM and the amplification time (at 72 °C) was raised to 40 s. Two microliters of eluate of whole mosquito body crushed in phosphate buffer saline was run in a final reaction volume of 25 µl without prior DNA extraction. In every reaction round, negative (non-template and An. stephensi mosquitoes) and positive controls (An. arabiensis colony mosquitoes) were included.

Statistical analysis
All analyses were performed in STATA version 13 (Stata-Corp., TX, USA) and GraphPad Prism 5.3 (GraphPad Software Inc., CA, USA). Feeding efficiency (proportion of fully-fed mosquitoes) was compared in matched experiments using the Wilcoxon matched-pairs signedrank test. Proportions were compared by Chi-square and Fischer's exact tests. Differences between median parasite densities between single-species infections and co-infections were assessed using Wilcoxon rank-sum test. The bias between wild and colony mosquitoes was compared using the Bland-Altman test. The correlation between mosquito infection prevalence and gametocyte density as continuous variable was determined by Spearman's rank correlation coefficient for colony and wild mosquitoes separately. Logistic regression was performed to compare infection status between colony and wild mosquitoes using individual mosquito data. A fixed effect for human participant was included thus taking into account the number of mosquito experiments and adjusting for correlations between mosquito observations from the same blood donor.

Discussion
In recent years, there is increasing interest in transmission assays to evaluate TBI and assess the human infectious reservoir for malaria. More and more laboratories are establishing mosquito colonies to examine infectivity among natural infections [42]. Whilst established colonies offer some advantage in terms of feeding efficiency [43], it is generally assumed that locally relevant mosquitoes are important to allow inference to the local transmission situation. We evaluated the permissiveness of An. arabiensis mosquitoes raised from wild-collected larvae in comparison with colony mosquitoes maintained for over 800 generations in 36 paired MFA. Whilst Fig. 2 Comparison of the proportion of infected colony vs wild mosquitoes. a The proportion of infected wild mosquitoes (Y-axis) is plotted against colony mosquitoes (X-axis) for P. vivax single-species infections with at least 10 mosquitoes dissected. The dotted line is the line of perfect agreement. b The differences between the proportion of infected colony and wild mosquitoes plotted against the averages of the two mosquito sources. The average of the proportion of infected colony and wild mosquitoes for each paired infection is indicated in the X-axis vs excess infections in wild mosquitoes (differences between proportions of infected wild mosquitoes vs colony mosquitoes) in the Y-axis. The limits of agreement are indicated as the mean difference (middle dotted line) and the 95% confidence interval of the limit of agreement (mean ± 1.96 SD of differences) with horizontal dotted lines. Unfilled dots indicate P. falciparum + P. vivax co-infections mosquito feeding rates were markedly higher in colony mosquitoes, we found no evidence for epidemiologically meaningful differences in infection prevalence or infection burden between mosquito sources.
In our experiments, we encountered challenges with the aggressiveness of wild mosquitoes, exemplified by roughly two-fold lower feeding rates on the membrane for wild versus colony mosquitoes, which is not surprising given the selection over several hundred generations in the latter. Colony mosquitoes were maintained using rabbits as source of blood for generations in the present study. Fewer mosquito observations were available for wild mosquitoes on the day of dissection for some of the infections. This may have contributed to the borderline higher proportion of MFA resulting in at least one infected colony mosquito, simply reflecting the higher number of mosquito observations [7]. Despite this, we observed a similar proportion of infected mosquitoes among infectious feeds between colony and wild mosquitoes, in line with several other studies [24,44]. This holds true when mosquitoes were of the same [24] or different [44] species. The F1 progeny of wild-caught An. funestus compared with colonized An. coluzzii mosquitoes [44] and similarly, colonized An. stephensi mosquitoes compared with their field counterpart raised from wildcaught larvae and pupae [24] were equally susceptible, when the end point was oocyst detection in the midgut.
Importantly, oocyst density was high and similar between the colony and wild mosquitoes in our study in line with previous studies on P. vivax that used mosquitoes of different species [11,12,45]. Lower oocyst densities are typically observed in P. falciparum [46][47][48]. Earlier studies also examined sporozoite prevalence and load in feeding experiments; most reporting similar levels between colonized and wild mosquitoes [15]. Similar prevalence but higher sporozoite density (but only at higher sporozoite loads) was detected in the wild mosquitoes in one of the studies [24]. Given the strong association between oocyst prevalence and intensity [49] and the strong association between oocyst density and sporozoite densities [6,8,50], it seems intuitive that highly similar oocyst burden, as observed in our study, precludes large differences in sporozoite density. Furthermore, variations in insectary and natural conditions that allow sporogony might potentially explain some of the differences observed [15,24]. Mosquito innate immune responses can abrogate infections through melanization [51]. We have not observed any evidence for melanization in the present study. In addition, we also examined mosquito guts for pathogens that may influence parasite development such as microsporidia [43] and found no evidence for this. Future studies may nevertheless benefit from examining sporozoites, a limitation of the present study. Investigation of effects of environmental factors on sporogony with a specific focus on midgut microbiota that can influence transmission efficiency by stimulating the mosquito innate immune system and production of metabolites directly impairing parasite survival will also be informative [32]. In addition, mosquito blood-meal size, a poorly studied parameter that may be higher in colony-and membrane-adapted mosquitoes, needs to be considered in future evaluations. We have reared wild collected and colony developmental stages to adults at the same laboratory conditions using the same larval food to minimize the chance this could contribute to a larger body size [52] and subsequently to higher oocyst prevalence and density as a function of larger volume of blood ingested (and therefore more gametocytes) [53,54]. Future studies would benefit by including wing length measurement as an indication of mosquito body size.
To the best of our knowledge, our findings are the first of its kind with African vivax malaria which is commonly referred to as a major cause of malaria outside sub-Saharan Africa [55]. Ethiopia forms an exception with vivax malaria, contributing towards three-quarters of the global burden together with India and Pakistan [56]. One of the unique features of P. vivax is the earlier generation of gametocytes, i.e. within 3-4 days after the first appearance of asexual parasites [57]. As a result, most patients start infecting mosquitoes before the onset of symptoms [58]. Despite a limited number of studies reporting a lack of association between microscopically determined gametocyte density and infectivity to mosquitoes [59], a very strong association was observed in the likelihood of infectivity between gametocyte densities and both colony and wild mosquitoes in our study. This is concordant with previous reports that used colonized An. dirus [9] and An. arabiensis mosquitoes [5] as well as An. stephensi [60] and An. darligi wild mosquitoes [61] raised from wild-collected immature stages and F1 generations, respectively.
One relevant limitation of our study was the limited sample size, relying on 36 blood donors but a total of 1755 colony and 2303 wild mosquitoes were used for the feeding experiments. We can thus not rule-out subtle differences between colony and wild mosquitoes. It would, however, be questionable whether small differences would render colony mosquitoes less suitable for assessments of the human infectious reservoir or the evaluation of interventions.