Larval diet affects mosquito development and permissiveness to Plasmodium infection

The larval stages of malaria vector mosquitoes develop in water pools, feeding mostly on microorganisms and environmental detritus. Richness in the nutrient supply to larvae influences the development and metabolism of larvae and adults. Here, we investigated the effects of larval diet on the development, microbiota content and permissiveness to Plasmodium of Anopheles coluzzii. We tested three fish diets often used to rear mosquitoes in the laboratory, including two pelleted diets, Dr. Clarke’s Pool Pellets and Nishikoi Fish Pellets, and one flaked diet, Tetramin Fish-Flakes. Larvae grow and develop faster and produce bigger adults when feeding on both types of pellets compared with flakes. This correlates with a higher microbiota load in pellet-fed larvae, in agreement with the known positive effect of the microbiota on mosquito development. Larval diet also significantly influences the prevalence and intensity of Plasmodium berghei infection in adults, whereby Nishikoi Fish Pellets-fed larvae develop into adults that are highly permissive to parasites and survive longer after infection. This correlates with a lower amount of Enterobacteriaceae in the midgut microbiota. Together, our results shed light on the influence of larval feeding on mosquito development, microbiota and vector competence; they also provide useful data for mosquito rearing.

Scientific RepoRts | 6:38230 | DOI: 10.1038/srep38230 and metabolism of adult mosquitoes 22 ; habitat type also affects the adult permissiveness to malaria parasites 14 , but whether this is mediated by nutrition is as yet unclear.
In a laboratory context, we hypothesized that such effects of larval nutrition and environment may translate in rearing conditions, and more specifically larval diet, affecting the biology of larvae and adult mosquitoes. As diet is known to influence the microbiota composition in metazoans 23,24 , we were particularly interested in assessing whether it affects the microbiota of colonised mosquitoes and thus potentially contributes to the high difference in microbiota composition between laboratories 10,[25][26][27] . In this study, we analysed the effect of larval diet on the development, the microbiota and the vector competence of A. coluzzii. Focusing on three fish diets commonly used in insectaries, we show that these diets differentially influence the speed of larval development, the size of adult mosquitoes and the larval and adult microbiota. Larval diet also significantly impacts prevalence and intensity of infection by P. berghei, likely with a microbiota-dependent mechanism.

Results
Larval diet influences mosquito development, physiology and permissiveness to Plasmodium.
As the success and rate of larval development are critically affected by the amount of food present in the water 28,29 , we hypothesised that the diet offered to mosquito larvae had an impact on their growth rate and time to pupation. Using Tetramin fish-flakes ('FF'), commonly used in our laboratory, as a control, we tested two other dietary treatments commonly used for mosquito rearing: Dr. Clarke's Pool Pellets ('D') and Nishikoi fish-pellets ('FP'). We split 2-day old larvae into three pans, provided them with comparable quantities of food, determined according to the weight of fat and proteins, and monitored their growth by measuring their length 7 days later. Larvae were 22% and 15% longer when reared on the D and FP pelleted foods, respectively, than on the FF flaked diet, indicating that these larvae elongated faster when reared on the D and FP foods (Fig. 1a). Larval development was also faster on D and FP than on FF, as quantified with a 2-to-3-day shorter time to pupation (Fig. 1b). As diet richness during larval development influences the size of the adults, we quantified the wing length as a read-out of the adult size. Larvae fed on both pelleted diets gave rise to significantly bigger adults than larvae fed on flakes (Fig. 1c).
Next, we investigated whether larval diet influences the permissiveness of adults to Plasmodium. Females fed as larvae with the three different diets were infected with P. berghei, and their infection rate was monitored 12 days later by counting the number of oocysts per gut (intensity) and the proportion of mosquitoes carrying oocysts (prevalence). Females reared on FP were significantly more permissive to P. berghei than their siblings reared on FF, as shown by their significantly higher intensity and prevalence of infection (Fig. 2a,c,e; + 360% in median intensity, p < 0.001, + 33% prevalence, p < 0.001), whilst no significant difference was observed between D and FF (Fig. 2a,d,e; + 125% in median intensity, p = 0.84, + 6% prevalence, p = 0.76). Interestingly, the proportion of females surviving until the day of dissection was higher in the group reared on FP than in the control (Fig. 2b,f).
Larval nutrition influences larval development and adult physiology through the target-of-rapamycin (TOR) signalling in Aedes 22 . TOR signalling is induced after blood feeding and leads to the production of the nutrient transporter Vitellogenin; starved larvae produce smaller mosquitoes, which show a reduced Vitellogenin expression 22 . As rearing on FP and D diets produced bigger mosquitoes, we examined whether the Vitellogenin response to blood feeding was increased in these mosquitoes. We quantified Vitellogenin at its expression peak, 24 h after blood feeding, by quantitative PCR 30 . Indeed, Vitellogenin transcripts were 2.8-fold more abundant in the midguts of mosquitoes reared on D (D vs FF, p = 0.044), but they were at similar levels in the midguts of FP-reared and  FF-reared mosquitoes ( Supplementary Fig. S1). This suggests that larval diet indeed affects mosquito metabolism, but in a way that does not directly correlate with permissiveness to Plasmodium. As the microbiota is another factor affecting mosquito development 15,17,18 and vector competence 31 , we went on to analyse whether the observed effects of larval diet correlated with any change in the microbiota.
Larval diet impacts the microbiota of larvae and adult mosquitoes. We quantified the microbiota of whole larvae by quantitative PCR on the 16S rRNA bacterial gene seven days after the first feed. The bacterial load in larvae increased by 18-fold when fed with pelleted food (Fig. 3a-d vs FF: 20-fold increase, p = 0.048; FP vs FF: 5.5-fold increase, p = 0.037). We previously showed by 16S sequencing that over 95% of the microbiota of our colony is represented by the Enterobacteriaceae and Flavobacteriaceae families 25 . Therefore, we monitored the bacterial load of these families by family-specific quantitative PCR as a proxy of the microbiota composition. We observed an increase in Enterobacteriaceae when larvae were fed on both pelleted diets ( Fig. 3a-d vs FF: 44-fold increase, p = 0.044; FP vs FF: 8.5-fold increase, p = 0.067), whilst the difference in Flavobacteriaceae was not significant. When expressing these families as a ratio of the total bacterial load, Flavobacteriaceae were present in a lower proportion in pellet-fed larvae than in flake-fed ones (Fig. 3b, 8.5-fold decrease, p = 0.017). As Asaia sp. are known to accelerate development when added to the larval breeding water 18 , we also quantified Acetobacteraceae but their amount was too close to the detection threshold to allow any reliable quantification. We also quantified bacteria in each diet to investigate whether diets may be the source of larval microbiota. We found more bacteria in the daily amount of food given to larvae in FP than in FF (Fig. 3c, p = 0.0042, Tukey-corrected ANOVA) and marginally significantly more in D than in FF (Fig. 3c, p = 0.066, Tukey-corrected ANOVA). However, the proportions of each bacterial family did not reflect the differences observed in the larval microbiota (Fig. 3d), suggesting that while diet may participate in providing bacteria to larvae, it also influences the microbiota composition by other means. As the adult mosquito microbiota is known to be influenced by the larval breeding site and the larval microbiota 10,17,21 , we next investigated the impact of larval diet on the midgut microbiota of adult females prior to or after a blood meal (referred to as sugar-fed, SF, and blood-fed, BF, respectively). As shown previously 31 , we observed that the bacterial load in mosquito midguts significantly increased after blood feeding, regardless of the larval diet ( Fig. 4a; BF vs SF, 730-fold, p < 0.001). This increase was observed for both bacterial families (Fig. 4c,d; Enterobacteriaceae in BF vs SF, 360-fold, p = 0.046; Flavobacteriaceae in BF vs SF, 100-fold, p < 0.001), but the relative contribution of Flavobacteriaceae to the total bacterial load decreased after the blood meal (Fig. 4f, BF vs SF, 1.8-fold, p = 0.001). Larval diet did not significantly affect the total bacterial load in adult midguts (Fig. 4a, all diets, p = 0.54), but Enterobacteriaceae were present at lower levels in the midguts of sugar-fed and blood-fed mosquitoes that received FP as larval diet ( Fig. 4c; Sugar-fed FP vs FF, 2.8-fold, p = 0.047; blood-fed FP vs FF, 2.0-fold, p = 0.048). The proportion of Enterobacteriaceae within the bacterial microbiota also decreased in sugar-fed midguts of FP-reared mosquitoes (Fig. 4e, FP vs FF, 19-fold, p = 0.022). The amount of Flavobacteriaceae decreased in the midguts of blood-fed mosquitoes reared on D (Fig. 4d, D vs FF, 2.6-fold, p = 0.043), but was not affected when reported as a proportion of the total microbiota (Fig. 4f). Again, Acetobacteraceae were present at too low levels to allow any comparison.
We also quantified the microbiota in the ovaries, to identify any a vertically-transmissible effect of larval diet on the microbiota. Blood-fed mosquitoes reared on pelleted larval diets appeared to have a lower bacterial load in their ovaries (Fig. 4b, pelleted diets vs FF, 2.9-fold, p = 0.019). In blood-fed mosquitoes, rearing on both pelleted diets also correlated with a lower Flavobacteriaceae load and a non-significantly reduced Enterobacteriaceae load ( Supplementary Fig. S2).
Finally, we looked for statistical interactions between larval diets and tissues or blood-feeding statuses. Considering tissues, we only detected a significant interaction with larval diets on the proportion of Enterobacteriaceae in sugar-fed mosquitoes ( Table 1, p = 0.04), where the significant difference between FF and FP in the midgut is not observed in ovaries (Figs 4e, S2c). For all other observations, significant effects of diets in one tissue correlate with similar trends in the other tissue, and interaction factors are non-significant (Table 1). Similarly, we found that blood-feeding status only statistically interacts with larval diet when considering the proportion of Enterobacteriaceae to the total bacterial load in midguts (Table 1, p = 0.0099). These interactions suggest that the effect of larval diet on the proportion of Enterobacteriaceae in sugar-fed midguts is specific, differing to the effect on blood-fed midguts and on ovaries.

Discussion
In this study, we show that the diet of Anopheles larvae influences their rate of growth and development, their microbiota and, after reaching adulthood, their permissiveness to Plasmodium infection. Larvae feeding on D or FP were found to grow and develop faster than larvae fed on FF, while they carried a higher bacterial load with a smaller proportion of Flavobacteriaceae. Previous studies have shown positive correlations between the presence of a microbiota and the success of larval development 15,17 , and between a diet supplemented with Asaia sp. and the speed of development 18 . This suggests that the effect of larval diet on larval development that we report here may be mediated by the observed differences in microbiota load and/or composition. Although the exact mechanisms underlying the role of the microbiota on mosquito development are currently unknown, genomic analyses of the bacterial strains and functional studies of other insects suggest that microbiota participate in macromolecule breakdown 32 , synthesise nutrients such as vitamins and essential amino acids 33 and promote growth and development via the insulin pathway 34,35 .
The three diets were matched in protein and fat content in this study, but we cannot rule out that differences in growth and development may be linked with a higher nutrient supply when feeding on D and FP. Indeed, the composition of all three commercial diets was not fully available, in particular energetic values and vitamin contents were missing. Nucleic acids are known to stimulate a feeding response; sterols, which are specific to eukaryotes, are essential for mosquito development 12 . These compounds might be involved in the observed differences in larval growth. As higher food concentration is reported to reduce time to pupation 13 , larvae coming in contact with the pellets may have developed faster due to ingesting a more concentrated food than their siblings fed with flakes, which were more evenly distributed throughout the water surface. In this case, one could expect a higher variance in time to pupation and larval lengths in pellet-fed larvae due to the more uneven accessibility to the pellets, which was not the case.
We also observed a significant effect of larval diet on larva and adult microbiota composition. Elizabethkingia, a genus of Flavobacteriaceae, is known to dominate the microbiota in colonies of laboratory-reared mosquitoes, including our own 10,25 . This is not representative of what is observed in field-collected mosquitoes, where Elizabethkingia is indeed prevalent but in low abundance 10 . Our results indicate a lower proportion of Flavobacteriaceae in larvae when feeding on both pelleted diets, suggesting that these diets may promote a microbiota that is more relevant to the study of mosquitoes. Although the larval microbiota, and more particularly Elizabethkingia, is known to be partially transmitted to adults 17,19,20 , the proportion of Flavobacteriaceae was not affected in adult mosquitoes in this study. More generally, although the larval diets significantly affected the adult mosquito microbiota, the effects observed in adult mosquitoes were not consistent with the effects in larvae. This may be due to differences in the efficiency of transstadial transmission between bacterial strains. For instance, it was previously documented that Elizabethkingia is much more efficiently transmitted from pupae to adults than Pseudomonas 19 . Some strains of Enterobacteriaceae efficiently colonising larvae of D and FP may not be efficiently transmitted to adults, restoring in proportion Flavobacteriaceae to the same level as found in adults reared on FF. Rearing mosquitoes on pelleted diets for several generations may lead to more substantial differences in the adult mosquito microbiota.
As previously reported, blood feeding affected the composition of the mosquito microbiota. While both Enterobacteriaceae and Flavobacteriaceae grew after blood feeding, the proportion of Flavobacteriaceae to the total bacterial load decreased. This may be due to kinetics of growth, or to the fact that Elizabethkingia is efficiently digested by A. gambiae 36 , potentially reducing its growth speed in the midgut. When quantifying bacteria in the midguts and ovaries of mosquitoes, we observed an overall correspondence between the effects on both tissues, where significant differences in microbiota observed in a tissue correlated with a similar trend in the other tissue. This is in agreement with a recent sequencing study of the midgut, ovary and salivary gland microbiota, showing a clustering of the microbiota composition by feeding history rather than by tissue 37 . The effect of diets however differed both between blood-feeding statuses and between tissues in the relative contribution of Enterobacteriaceae to the adult microbiota, sugar-fed midguts appearing to be affected differently to blood-fed guts or ovaries. This may reflect an effect of larval diet on Enterobacteriaceae strains that are not fast growing after the blood meal and do not cross the gut epithelium.
Our data also indicate that larval diet influences mosquito permissiveness to Plasmodium infection. FP-fed larvae produced highly permissive mosquitoes that survived longer after Plasmodium infection than the controls fed on FF, whereas feeding on D did not affect adult permissiveness to Plasmodium infection or survival. Although an increase in survival might be surprising with an increased infection load, Plasmodium infection is known not to inflict much of a fitness cost on mosquitoes in the absence of nutritional stress 38 , as is the case for our mosquitoes fed with fructose ad libitum. Thus, our results point to FP mosquitoes being fitter, thus either allowing a higher Plasmodium infection or surviving it better. These effects on mosquito infection correlate with a significantly lower amount of Enterobacteriaceae in mosquitoes reared on FP. We previously reported a similar combination of a decrease in Enterobacteriacae, a higher survival rate and a higher permissiveness to Plasmodium when adding a penicillin-streptomycin antibiotic cocktail to the blood meal 25 . As already reported for several strains of Enterobacteriaceae including Enterobacter EspZ and Serratia marcescens HB3 8,39 , we can hypothesize that a strain of the Enterobacteriaceae present in our mosquito colony may have a direct or indirect anti-parasitic action. These data might appear contrary to the previously reported positive correlation between the relative contribution of Enterobacteriaceae in the mosquito microbiota and the success of parasite infection 10,40 . However, the effect of bacteria from the microbiota is known to be highly strain-specific.
In conclusion, our results provide new insights into the effect of larval diet on mosquito development, microbiota and permissiveness to malaria parasites. They are also of technical interest in mosquito rearing, showing that faster development can be achieved when using the pelleted diets D and FP, and that FP produces mosquitoes with a higher Plasmodium infection rate.

Methods
Ethics Statement. Animal use was carried out in accordance with the UK Animals (Scientific Procedures) Act 1986. The protocol for infecting mosquitoes with P. berghei by blood feeding on parasite-infected mice was carried out under the UK Home Office License PPL70/7185 awarded in January 2010. The procedures are of mild to moderate severity and the number of animals used are minimised by incorporation of the most economical protocols. Protocols were implemented following approval by the Imperial College Ethical Review Committee.
Mosquito Colony and Maintenance. Experimental work was performed on the An. coluzzii Ngousso colony established in 2006 from field-collected mosquitoes from Cameroon, maintained on human blood at 27 °C and 70% humidity, with 12 h:12 h light/dark cycle. Adult mosquitoes were fed with a 5% fructose solution and larvae were kept in distilled water. Larval generations prior to those obtained for dietary treatments were fed Tetra FF ground to 1-2 mm fragments, the control diet used in this study. Pupae were collected daily, placed in glass beakers with distilled water and transferred to cages.
Larval feeding. For each dietary treatment, 400 (± 10%) 48 h-old unfed An. coluzzii first instar larvae were counted manually or using a COPAS larvae sorter (Union Biometrica, USA) and put into a plastic pan. Each larval pan was fed on one of three conventional fish-based diets -Tetra Tetramin fish-flakes ('FF'), Dr. Clarke's Pool Pellets ('D'), and Nishikoi Fish Pellets ('FP'). Nutritional summaries of these diets are shown in Table 2. As reported in Table 2, diets were provisioned in quantities providing comparable levels of nutritional constituents in each feed. FF were slightly blended to reduce the size of the flakes, as commonly performed in our laboratory for larval rearing.
Larval length measurements. Twenty 9-day-old larvae (from the time of egg collection) were collected from each treatment of 3 independent experiments, and placed into individual wells of a 96-well plate for imaging at 20x magnification. The whole body length was measured from the top of the head to the tip of the abdomen using Fiji 41 . Wing length measurements. Twenty adults were collected from each treatment of 4 independent experiments. One wing of each individual was removed and taped onto a sheet of paper. After scanning the paper, wing lengths were measured using Fiji 41 .
Microbiota and gene expression analysis. Twenty individuals (9-day-old larvae, 3-5 day-old females prior to or 24 h after blood feeding) were collected from each treatment of 4 independent experiments, suspended in 70% ethanol for 3-5 minutes, then washed 3 times in PBS as performed previously 25 . Adult midguts and ovaries were dissected and sampled in 2 groups of 10 tissues. Forceps and slides were cleaned with distilled water and ethanol between each sample. To quantify bacteria in food, a daily (d7+ ) amount of food was homogenized in 5 mL sterile water, then 1 mL was sampled, spun at 800 g for 5 min and the supernatant was discarded. 4 replicates were monitored for each diet. In order to normalise the amount of RNA, a suspension of Sf9 cells from a confluent culture of a 175 cm 2 flask was split into all 12 samples. RNA was extracted from whole larvae, adult tissues and pelleted water contents with Trizol and chloroform, precipitated using isopropanol and washed twice in 70% ethanol. Reverse transcription (Takara) was performed using PrimeScript RT Enzyme Mix I with 650 ng RNA following the manufacturer's instructions. The bacterial 16S rRNA gene was used for quantification by qRT-PCR performed on a LightCycler 480 (Roche), using SYBR Premix Ex Taq kit (Takara) following the manufacturer's instructions. The bacterial load and composition was assessed using 16S rRNA gene primers general to all bacteria and specific to bacterial families known to colonise mosquitoes in the insectary insectary, respectively (Table S1 and ref. 25). Bacterial load and Vitellogenin expression are presented as ratios of the target gene to the reference Anopheles house-keeping gene 40S ribosomal protein S7 (S7), while bacterial load in food was quantified as a ratio of the GAPDH (glyceraldehyde-3-phosphate-dehydrogenase) house-keeping gene in Sf9 cells.
Plasmodium infections. [8][9][10][11][12]week-old CD1 female mouse was intraperitoneally injected with 100 μ l 5O7 P. berghei-infected blood and kept for two days to allow parasite growth and gametocyte differentiation. After checking that parasitemia was between 3 and 8%, the mouse was intramuscularly injected with 4 μ L. g −1 of an anesthetic solution (Rompun 0.33%, ketamine 33 mg. mL −1 in PBS) and used to blood feed three cups of mosquitoes in turns of 5-10 min. Mosquitoes were kept for 12 days at 20 °C to allow parasite development until the oocyst stage. Then, dead mosquitoes were removed and scored and all surviving mosquitoes were dissected for oocyst counts using a fluorescent microscope. For each experiment, at least 50 mosquitoes per condition were offered a blood meal and non-engorged mosquitoes were removed 24 h later. On average, the oocysts of 36 (± 3.6 s.e.m.) mosquitoes per condition per experiment were counted. Differences in median intensities and prevalence are mentioned in the text as averages of the effect from each replicate; that is, (data(test diet) -data(control diet))/ data(control diet).
Statistics. Statistical analyses were performed by generalised linear mixed models (GLMM) in R (version 3.2.3). An ANOVA X 2 -test on a logistic regression (glmer) was performed on factorial data (prevalence, survival). A Wald Z-test on a zero-inflated negative binomial regression (glmmADMB) was performed for oocyst counts. An ANOVA X 2 -test on a common linear regression (lmer) was performed on qPCR data (16S, gene expression), larva length, time to pupation and wing length. GLMM analyses decompose data into a fixed component (here larval diets) and a random component (here replicate experiments). Odds ratios presented in Fig. 2 are the exponential of the regression coefficients. Sample size was chosen according to the standard protocols (oocyst counts, gene expression, microbiota analyses) or estimated according to mosquito rearing experience (larval development). Mosquitoes were excluded if they did not blood feed, if they burst during dissection (microbiota analyses, gene expression) or if they died (as larvae or adults) during the experiment.   Table 2. Summary of nutritional constituents of larval diets and of feeding regimes. Ash consists of inorganic matter. Foods contained vitamins and provitamins although these were of unknown quantity and not consistently specified. Feeding regime was designed to provide comparable levels of proteins and fats between diets in each feed. Larvae were not fed throughout weekends and instead were given a 2-day worth feed on Fridays.