Comparative assessment of  An. gambiae  and  An. stephensi  mosquitoes to determine transmission-reducing activity of antibodies against  P. falciparum  sexual stage antigens

There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.

Abstract

Background

With the increasing interest in vaccines to interrupt malaria transmission, there is a demand for harmonization of current methods to assess Plasmodium transmission in laboratory settings. Potential vaccine candidates are currently tested in the standard membrane feeding assay (SMFA) that commonly relies on Anopheles stephensi mosquitoes. Other mosquito species including Anopheles gambiae are the dominant malaria vectors for Plasmodium falciparum in sub-Saharan Africa.

Methods

Using human serum and monoclonal pre-fertilization (anti-Pfs48/45) and post-fertilization (anti-Pfs25) antibodies known to effectively inhibit sporogony, we directly compared SMFA based estimates of transmission-reducing activity (TRA) for An. stephensi and An. gambiae mosquitoes.

Results

In the absence of transmission-reducing antibodies, average numbers of oocysts were similar between An. gambiae and An. stephensi. Antibody-mediated TRA was strongly correlated between both mosquito species, and absolute TRA estimates for pre-fertilisation monoclonal antibodies (mAb) showed no significant difference between the two species. TRA estimates for IgG of naturally exposed individuals and partially effective concentrations of anti-Pfs25 mAb were higher for An. stephensi than for An. gambiae.

Conclusion

Our findings support the use of An. stephensi in the SMFA for target prioritization. As a vaccine moves through product development, better estimates of TRA and transmission-blocking activity (TBA) may need to be obtained in epidemiologically relevant parasite-species combination.

Electronic supplementary material

The online version of this article (10.1186/s13071-017-2414-z) contains supplementary material, which is available to authorized users.

Related collections

Most cited references 43

Record: found
Abstract: found
Article: not found

Implication of the Mosquito Midgut Microbiota in the Defense against Malaria Parasites

Yuemei Dong, Fabio Manfredini, George Dimopoulos (2009)

Introduction The malaria parasite has to go through series of complex developmental transitions within the mosquito vector before it can be transmitted to the human host. The major bottleneck for Plasmodium's development occurs during the ookinete invasion of the midgut epithelium, prior to the development of oocysts on the basal lamina [1]. The factors that are believed to contribute to parasite losses at this stage are digestive enzymes, the mosquito's immune defenses and the intestinal microbial flora [2]–[4]. Large communities of diverse microorganisms reside in insects with a major concentration in the intestinal sections [5]. While much research has been focused on the microbiota of the mammalian intestine and its role in defense against pathogenic microorganisms [6], studies of insect gut microbiota have mainly concentrated on the contribution of microbial endosymbionts to the host's nutritional homeostasis [5]. However, the microbiota of the insect gut has also been shown to play a pivotal role of preventing development of pathogens. Studies have reported the wide spread of various species of Gram-negative bacteria in the midguts of both laboratory-reared and field derived mosquitoes, and some of this flora has been associated with an inhibitory activity on the sporogonic development of the Plasmodium parasites in the mosquito midgut [7]–[11]. However, these studies have not identified the causal mechanisms through which the presence of bacteria negatively impacts on malaria parasite development. Bacteria within the midgut lumen may directly interact with, and adversely affect, the different malaria parasite stages within the bloodmeal through the production of various enzymes and toxins or physical barriers that hinder the interaction between Plasmodium ookinetes and the midgut epithelium (reviewed in [12]). Alternatively, the effect of bacteria on parasite development may occur indirectly through alterations in the physiology of the mosquito host itself, possibly through induction of immune responses that are cross-reactive between bacteria and malaria parasites, and/or changes of host metabolism that would affect the composition of mosquito derived molecules that are essential for Plasmodium development. Some studies have indicated that some of the mosquitoes' immune factors induced by bacterial challenge are involved in the killing of parasites at the pre-oocysts development stages [13]–[15]. Indeed, a great overlap, at the functional level, between antibacterial and anti-Plasmodium immune responses has been observed and suggests that mosquitoes lack highly specific mechanisms for defense against malaria parasites, but are using their anti-bacterial mechanisms to limit Plasmodium infection [14],[16],[17]. A reasonable hypothesis is that the presence of bacteria activates the mosquito's antimicrobial immune responses and the synthesized antimicrobial peptides and other immune factors will act against co-infecting Plasmodium parasites. Indeed, a complex interplay between the mammalian immune system and the intestinal microbiota is essential for protection from infectious pathogenic microorganisms [18]. Some intestinal microbial species induce innate immune effector molecules which can kill competing bacterial species, including pathogens (reviewed in [19]). The composition of mosquito midgut microbiota is much less complicated than that of mammalian intestine microbiota which makes it as a good model for dissecting the dynamics between the host innate immune system, natural bacterial flora, and the pathogenic microorganisms. Besides, mosquitoes transmit a broad range of human parasitic and viral diseases, within which malaria is still one of today's most devastating infectious diseases. A better understanding of the roles of microbiota in the exploiting host immunity in defending against pathogens could potentially lead to the development of new malaria control strategies. We have examined the influence of the mosquito's midgut microbial flora and the derived antibacterial immune responses on malaria parasite infection through a series of infection assays in conjunction with functional genomics analyses. Results/Discussion Composition of microbiota in A. gambiae mosquitoes To gain a better understanding of the potential fluctuations in microbial load and species composition between laboratory reared mosquitoes of different generations and within the same generation, we monitored the bacterial loads and species composition in individual five-day-old female A. gambiae mosquitoes of five consecutive generations. In accordance to previous studies our results showed a great variability in both parameters [8], [9], [20]–[22]. Interestingly, these variations were also observed between mosquitoes originating from the same generation and cage (Figure 1). This intriguing pattern may in some way relate to the equally broad distribution of Plasmodium infection intensities among mosquitoes that have fed on the same gametocyte culture. On average, individual mosquitoes carried around 40,000 colony forming units (CFU). Similarly to previous studies, the majority of the isolated bacteria were Gram-negative suggesting that the midguts of mosquitoes have more optimal growth conditions for this type, especially those from the Enterobacteriaceae family. This strong bias is also likely to have been attributed, to some degree, to the LB agar– based aerobic culturing method that was used for these assays. Sequence analyses of the 16s ribosomal genes from morphologically distinct bacteria colonies identified the following five different species as dominant in all assayed generations: Enterobacter asburiae (98%), Microbacterium sp. (98%), Sphingomonas sp. E-(s)-e-D-4(2) (100%), Serratia sp. (99%) and Chryseobacterium meningosepticum (100%). The C. meningosepticum and Serratia sp. species were dominant within all five generations and the former was the most abundant, especially within the second generation. Other bacteria identified from different generations were: Asaia bogorensis (99%), Bacillus subtilis (99%), Enterobacter aerogenes (98%), Escherichia coli (91%), Herbaspirillum sp. (99%), Pantoea agglomerans (98%), Pseudomonas fluorescens (99%), Pseudomonas straminea (99%), Phytobacter diazotrophicus (97%) and Serratia marcescens (99%). Interestingly, when C. meningosepticum became the dominant bacterium of the midgut flora, the growth of other bacterial species, that could be cultured on LB agar, was usually limited suggesting that this species may possess some competitive advantages in the gut environment. 10.1371/journal.ppat.1000423.g001 Figure 1 Distribution of bacterial loads and major species composition of midguts microbiota in 5 individual laboratory-reared 5-d-old female A. gambiae mosquitoes from 5 consecutive generations. The bacteria species were determined to be closely related to Enterobacter asburiae, Microbacterium sp., Sphingomonas sp., Serratia sp., and Chryseobacterium meningosepticum. G1 to G5 denotes generation 1st to 5th. Our LB agar –based culture assays have some limitations in providing the complete picture of the composition of the mosquito midgut microbiota since a large fraction of bacteria are likely to be un-culturable, similarly to the human intestinal microbiota [23]. Future high throughput sequencing -based metagenomics approaches are likely to provide comprehensive information on the composition of the midgut microbiota. Nevertheless, as a proof of principle, our approach shows the great variations in both load and composition of the microbiota between different individuals and generations of insectary-reared mosquitoes. The mosquitoes' natural microbiota can influence their permissiveness to Plasmodium infection We assessed the impact of the mosquito's natural microbial flora on P. falciparum's capacity to establish infection through the removal of bacteria with antibiotic treatment, according to the established methodology [24],[25]. Provision of antibiotic through the sugar meal effectively eliminated all detectable bacteria from mosquitoes fed on either sugar or human blood (Figure 2A). The average bacterial load of sugar fed mosquito midguts was 104 CFU, and those fed on blood contained as many as 106 CFU (Figure 2A). After antibiotic treatment mosquitoes became aseptic and are referred as aseptic mosquitoes, while untreated mosquitoes are referred as septic. Aseptic mosquitoes were significantly more susceptible to P. falciparum infection, as a measure of oocysts numbers on the midgut, compared to the septic mosquitoes (p 0.05), compared to the PBS injected controls (Figure 3B). This result further supports that the anti-Plasmodium activity of bacteria is indirect and involves a response by the mosquito vector since the injected bacteria are unlikely to directly interact with the parasites that are confined within the midgut epithelium or under the basal lamina. It is more likely that the systemic infection will induce a battery of defense molecules in the hemolymph, from where they can attack the midgut-stage parasites on the basal side of the gut, or even within the epithelium by diffusion through the basal labyrinth. Indeed our previously published studies showed that injected bacteria induced a battery of anti-Plasmodium immune factors [14]. The stronger anti-Plasmodium effect of either injected or co-fed live bacteria, compared to heat inactivated bacteria, suggest that a factor which is more specific for live bacteria may be responsible for the inhibitory effect. Alternatively, the stronger effect of live bacteria may simply reflect their proliferative capacity which resulted in multiplication of their numbers to induce a much stronger immune response from the mosquito host. Mosquito genome-wide responses to microbial exposure Mosquitoes, as all other higher organisms, are continuously exposed to a variety of microbes. And we have shown that this exposure, whether it originates from the midgut lumen or the hemolymph, can influence the mosquito's permissiveness to P. falciparum infection. We have also shown that this effect is likely to be mediated through a mosquito response to the bacterial exposure. To better understand this response we have performed a series of genome-wide expression analyses to assess the regulation of the mosquito transcriptome upon microbial exposure. We used a microarray-based genome-wide gene expression strategy to compare transcript abundance between septic and aseptic adult female mosquitoes that had been fed on either sugar or non-infected blood (Figure 5 and Tables S2, S3). The presence of the endogenous bacteria flora in sugar fed mosquitoes resulted in the differential regulation of some 185 genes; 121 genes were up-regulated and 64 genes were down-regulated compared to antibiotic treated aseptic mosquitoes. A similar number of 195 genes were regulated by the presence of the endogenous microbial flora after feeding on non-infected blood; 137 genes were up-regulated and 58 genes were down-regulated (Figure 5A). The relatively small number of genes that were regulated as a consequence of the presence of the endogenous microbial flora most likely indicates a symbiotic relationship that has led to the adaptation of the mosquito to this flora. This hypothesis is strengthened by subsequent experiments that investigated the effect of ingested non-natural bacteria on the mosquito's transcriptome (see below). The mosquitoes' responses to natural microbiota when fed with either sugar or non-infected blood were quite divergent with only limited overlap in gene expression (Figure 5A), that comprised 21 induced and 1 repressed gene, corresponding to approximately 6.5% of the total regulated genes. However, one third of the commonly induced genes belonged to the immunity class. 10.1371/journal.ppat.1000423.g005 Figure 5 Global gene regulation at the different conditions of infection. (A) Comparison of transcript abundance between whole septic and aseptic mosquitoes after feeding on sugar (SF) or uninfected blood (BF), and in the midguts (Bac-Gut) or carcass tissues (Bac-Carc.) of mosquitoes 12 hrs post ingestion of uninfected blood supplemented with E. coli and S. aureus (substitution of bacteria with PBS as control). Colored arrows indicate genes that are up- or down- regulated in the corresponding treatment group. (B) Proportions and numbers of genes belonging to distinct functional groups which were up- or down- regulated in the corresponding treatment group. SF Whole: sugar-fed whole septic mosquitoes compared to aseptic ones; BF Whole: uninfected blood-fed whole septic mosquitoes compared to aseptic ones; Bac-Gut: mosquitoes midgut tissues 12 hrs post ingestion of experimental bacteria; Bac-Carc.: mosquitoes carcass tissues 12 hrs post ingestion of experimental bacteria (E. coli and S. aureus); I/A: putative immunity and apoptosis; R/S/M: oxidoreductive, stress-related and mitochondrial; C/S: cytoskeletal, structural; MET: metabolism; R/T/T: replication, transcription, translation; P/D: proteolysis, digestion; TRP: transport; DIV: diverse; UKN: unknown functions; gene functions were predicted based on Gene Ontology data and manual sequence homology searches. (C) Same as in (B), but also including genes of diverse functions (DIV) and unknown functions (UKN). The regulated genes represented a variety of functional classes with a general strong bias and over-representation of innate immunity genes (Figure 5B and 5C). Several of these immune genes have been previously shown to be transcriptionally-induced during malaria parasite infection, and to mediate anti-Plasmodium activity (Tables S2, S3). The septic mosquitoes displayed elevated expression of genes code for the antimicrobial peptides Cecropins 1 (Cec1) and 3 (Cec3), Defensin 1 (Def1) and Gambicin; the signal transducing serine proteases SP5, ClipA9, ClipA7 and ClipB8, and various pattern recognition receptors including AgMDL8, CTLMA4, FREP7 and FBN51, Tep4 and Tep5, Galectin 8, and PGRP-LB, PGRP-LC2 and PGRP-S3 [14], [27]–[31]. Surprisingly, the expression of the anti-Plasmodium factors FBNs 6, 9, and 36 were decreased in the septic mosquitoes (Tables S2, S3). The immune responsive Lysozyme c-1 (LYSC1) which previously has been linked to melanization reactions [32]–[34], was up-regulated in septic sugar-fed mosquitoes; lysozymes are key antibacterial factors. These results suggest that the natural microbiota play an important role in stimulating a basal immune activity which in turn is likely to contribute towards the determination of the mosquito's susceptibility to various pathogens, and hence their vectorial capacity. In fact a recent study has established that Plasmodium development is significantly more influenced by the mosquito's basal level immunity rather than the induction of immune responses upon parasite infection [35]. Of particular interest was the elevated expression of the peritrophic matrix protein gene Ag-Aper1 in septic mosquitoes that had fed on either sugar or uninfected blood, and several other genes encoding proteins with peritrophin-like, laminin-EGF-like and chitin-binding like domains (Tables S2, S3) [36]. Ag-APer1 and proteins containing chitin-binding domains may function as structural components of the insect cuticle, the peritrophic matrix and/or as pattern recognition receptors. The elevated expression of Ag-Aper1 in septic mosquitoes may indicate a role of the peritrophic matrix in protecting the epithelium from the infection of midgut flora bacteria. The natural microbial flora also stimulated expression of several metabolic genes involved in glycolysis, gluconeogenesis and sugar transport and this may relate to digestion of midgut bacteria that function as a food source for the mosquitoes [37] (Tables S2, S3). The genes exhibiting the greatest fold-differences in expression between septic and aseptic mosquitoes were of unknown function (Figure 5C). To investigate the mosquito's global transcriptional response to exposure to non-natural midgut flora we compared transcript abundance between mosquitoes that had fed on blood supplemented with a mixture of both Gram-negative (E. coli) and Gram-positive (Staphylococcus aureus) bacteria and control mosquitoes that had fed on uninfected blood with PBS. These treatments resulted in a much broader response. The ingestion of these bacteria triggered the regulation of as many as 656 and 520 genes in the midgut and carcass, respectively (Figure 5 and Tables S4, S5). In the midgut, 458 genes were up-regulated and 198 genes were down-regulated. As expected, fewer genes were regulated in the carcass compared to midgut tissue which was in direct contact with the ingested bacteria; 224 genes were up-regulated and 296 were repressed. Among the immune genes exhibiting differential expression between sterile-blood-fed and bacteria-blood-fed mosquitoes were several that have previously been shown to mediate anti-Plasmodium immune responses and to be transcriptionally up-regulated during Plasmodium parasite infection (Tables S4, S5). The ingestion of bacteria stimulated an elevated expression of genes code for the antimicrobial peptide IRSP1, the signal transducing serine proteases ClipB16, and inhibitor SRPN6 and SRPN7, and various pattern recognition receptors including AgMDLs 4, 6, and 7, CTL, CTLGA1, CTLGA3, and CTLMA6, FBNs 9, 20, 21, and 51, LRRD8, PGRP-LB, PGRP-LC2 and PGRP-S3, Tep11 and Toll6 [14], [38]–[42]. Only four immune genes, SP5, TPX4, DCCE2, and FBN51 were induced by both the natural flora and the ingested non-natural bacteria, while Tep-like, PGRP-LD, and FBN9 displayed an opposite pattern of regulation (Figure 5A and Tables S4, S5). As mentioned above, these differences are likely to reflect an adaptation of the mosquito to its natural microbial flora. Potential differences in the dosage of bacterial exposure may however also have influenced the quite different outcome. The natural microbiota stimulates basal immune activity that controls its proliferation Depletion of several immune factors through RNAi-mediated gene silencing has been shown to result in a proliferation of bacteria in the hemolymph as a result of a compromised immune system [43],[44]. To test whether the immune genes that are induced by the natural flora are indeed implicated in defending against opportunistic bacterial infections, we assayed the proliferation of the mosquito midgut flora upon their silencing. We subjected 12 genes to this test of which Cec1, Cec3, Def1, ClipA9, Gambicin, PGRP-LB, and FBN9 were induced by the presence of the natural bacterial flora, and the remaining LRRD7, LRRD19, TEP1, Rel1 and Rel2 genes represented anti-Plasmodium pattern recognition receptors or immune signaling pathway factors [35], [45]–[47]. Depletion of Cec3, Gambicin, PGRP-LB, LRRD7, TEP1, and Rel2 resulted in the significant proliferation of the natural bacterial flora in the mosquitoes' midguts. Gene silencing of Cec1, Def1, ClipA9, FBN9, and LRRD19 also resulted in some increase of bacterial loads in the midgut; however these effects were not statistically significant (Figure 6). The lack of significant bacterial proliferation in these knock-down mosquitoes could also be explained by the lower efficacy of gene silencing in the midgut tissue compared to the abdominal and thoracic compartment (Figure S3, Table S1). These results show that the mosquito's innate immune system is actively involved in controlling the bacterial load in the midgut lumen in a constitutive fashion, and that exposure to increased bacteria will result in increased production of some of these anti-Plasmodium factors. We believe that this is the mechanistic basis of how the mosquito's endogenous flora is important in priming an anti-Plasmodium defense. 10.1371/journal.ppat.1000423.g006 Figure 6 Immune gene-silencing influenced the bacterial loads of the mosquito midguts. Bars represent the mean values of total CFUs (log10 transformed) from 10 midguts, and standard error bars are included. *, p<0.05; **, p<0.01. The anti-Plasmodium action of immune genes can be modulated by the presence of the mosquito's endogenous microbiota The dual role of anti-Plasmodium factors in defending against both the parasite and bacteria, and the influence of bacteria on Plasmodium development, suggests the existence of complex interactions and relationships between the parasite, the microbiota and the mosquito's innate immune system. For example, the anti-Plasmodium activities of certain genes might be modulated by their parallel activities against bacteria. To assess such complexities and interactions, we studied the effect of various immune genes on P. falciparum's capacity to establish infection in the midgut tissue of both septic and aseptic mosquitoes through RNAi gene silencing approach (Figure 7). 10.1371/journal.ppat.1000423.g007 Figure 7 The depletion of PGRP-LB, Cec3, and ClipA9 through RNAi gene silencing resulted in the changes of P. falciparum oocyst intensity in the septic (untreated) and aseptic (antibiotic-treated) mosquitoes. Points indicate the absolute value of oocysts counts in individual mosquitoes, and horizontal black bars in each column represent the median value of oocysts from three replicates where the narrow black bars above or below the median values indicate the standard errors. p-values were calculated through a Kruskal-Wallis test. (A) P. falciparum oocyst intensity increased in aseptic mosquitoes (Antibiotic) when Cec3 or PGRP-LB was silenced. dsGFP injected mosquitoes (GFP) were used as controls. (B) P. falciparum oocyst intensity decreased in septic mosquitoes (Untreated) when ClipA9 was silenced. RNAi-mediated depletion of the antimicrobial peptides Cec1, Def1, and Gambicin had no statistically significant effect on the levels of P. falciparum oocyst infection in either mosquito groups (data not shown), while gene silencing of Cec3 and PGRP-LB resulted in an increased susceptibility to P. falciparum only in the aseptic mosquitoes (Figure 7A). This result may suggest that the depletion of these two immune genes in septic mosquitoes resulted in a proliferation of the microbial flora which in turn may have counteracted, or masked, the potential decrease of anti-Plasmodium immune responses. Another striking example of how important the microbial flora is in regulating anti-Plasmodium activity of immune genes is represented by the serine protease ClipA9. When this factor was depleted in septic conditions, the mosquitoes became significantly less susceptible to P. falciparum infection (p<0.05). In contrast, when ClipA9 was silenced in aseptic mosquitoes it had no significant effect on susceptibility to the parasite (Figure 7B). This observation suggests that the ClipA9-mediated anti-Plasmodium defense is exerted through the microbial flora and not directly against the parasite. ClipA9 is likely to be more specific for antibacterial defense and its depletion, under septic conditions, will hence result in the proliferation of bacteria which will exert strong anti-Plasmodium activity. Alternatively, it may mediate some direct Plasmodium protective activity which is abolished in the absence of bacteria. Interestingly the malaria parasite infection phenotype of ClipA9 gene silencing is opposite to that observed for the serine protease inhibitor SRPN6, suggesting that SRPN6 may function in the same cascade as an inhibitor of ClipA9 [30],[39]. In conclusion, similarly to humans, the mosquito intestine harbors a natural microbiota which is necessary for maintaining normal physiological functions including host metabolism and immune homeostasis. Accordingly, we have shown that the mosquito's natural bacterial flora show great variability between mosquitoes originating from the same colony and that it is an important regulator of mosquito permissiveness to Plasmodium. The mosquito's natural microbiota and artificially introduced non-natural bacteria negatively affected malaria parasite development through a mechanism that appears to implicate in the innate immune system, and not a direct killing of Plasmodia by the bacteria. The natural bacterial flora is essential in inducing a basal level immunity that in turn enhances the mosquito's ability in defending against the infection from the malaria parasites [35]. Interestingly, the effect of certain immune genes on Plasmodium infection is dependent on the presence of the microbial flora, suggesting that their mode of action is complex. This finding suggests that future studies on gene specific anti-Plasmodium action should also consider the complex interplay between the microbiota and the mosquito's immune defenses against the Plasmodium parasite. This relationship is further corroborated by observations from Dr. Barillas-Mury's group, where RNAi gene silencing of one immune gene facilitated the proliferation of microbial flora but reduced the Plasmodium infection. The natural bacterial flora has also been shown to be involved in the suppression of other pathogenic organisms in other mosquito species. Tetracycline treatment of Culex bitaeniorhynchus rendered this mosquito more susceptible to the Japanese encephalitis virus [48] and the Aedes aegypti mosquito microbial flora has been shown to stimulate a basal-level immunity which suppresses dengue virus infection [25]. Materials and Methods Ethics statement All animals were handled in strict accordance with good animal practice as defined by the relevant national and/or local animal welfare bodies, and all animal work was approved by the appropriate committee. Mosquito rearing, antibiotic treatments, and RNA sample preparation A. gambiae Keele strain mosquitoes were maintained on a 10% sugar solution in laboratory culture at 27°C and 70% humidity with a 12 hrs light/dark cycle according to standard rearing procedures [49]. A single cohort of adult female mosquitoes were collected immediately after eclosion, and either maintained under normal, non-sterile insectary conditions or placed into a sterile environment. Following, adult female mosquitoes were daily given fresh filtered sterilized 10% sucrose solution containing 15 µg gentamicin sulphate (Sigma) and 10 units/10 µg of penicillin-streptomycin (Invitrogen) per ml, respectively. Each cohort of mosquitoes was simultaneously membrane-fed freshly washed human erythrocytes resuspended to 40% haematocrit using human serum. As far as possible, every care was taken to maintain the sterility of the blood and membrane-feeding apparatus used to feed the mosquitoes, in order to prevent the antibiotic-treated mosquitoes acquiring bacterial infection during the process of membrane-feeding. The mosquitoes were starved for 8 hrs before feeding to encourage engorgement, and sugar solution was replaced once blood feeding had finished. At 24 hrs after blood feeding, 20 mosquitoes from each replicate of each cohort was collected and dissected on ice. RNA was extracted from dissected tissues at the assayed time points using the RNeasy kit (Qiagen). The quantification of RNA concentrations was performed using a Spectrophotometer (Eppendorf). Microarray hybridization and data mining Probe sequence design and microarray construction were kept the same as described in [14]. Probe preparation and microarray hybridizations were performed essentially as previously described with some modifications [14]. Briefly, Cy3-labeled control cRNA probes and Cy5-labeled treatment cRNA probes were synthesized from 2–3 µg of RNA using the Agilent Technologies low-input linear amplification RNA labeling kit according to the manufacturer's instructions. Hybridizations were performed with the Agilent Technologies in situ hybridization kit according to the manufacturer's instructions with 2 µg of cRNA probes and 16 hrs after hybridization the microarray slides were washed and dried with compressed air. Microarrays were scanned with an Axon GenePix 4200AL scanner using a 10 µm pixel size (Axon Instruments, Union City, California, United States). Laser power was set to 60%, and the photomultiplier tube (PMT) voltage was adjusted to maximize effective dynamic range and minimize the saturation of pixels. Scanned images were analyzed by using GenePix software, and Cy5 and Cy3 signal and ratio values were obtained and subjected to statistical analysis with TIGR MIDAS and MeV software [50]. The minimum signal intensity was set to 100 fluorescent units, and the signal to background ratio cutoff was set to 2.0 for both Cy5 and Cy3 channels. Three or four biological replicates were performed for each experimental set. The background-subtracted median fluorescent values for good spots (no bad, missing, absent, or not-found flags) were normalized according to a LOWESS normalization method, and Cy5/Cy3 ratios from replicate assays were subjected to t tests at a significance level of p<0.05 using cutoff value for the significance of gene regulation of 0.7 and 0.8 in log2 scale, for septic mosquitoes and mosquitoes co-fed with experimental bacteria respectively, according to previously established methodology [51]. Microarray-assayed gene expression of 6 genes was further validated with quantitative RT-PCR and showed a high degree of correlation with the Pearson correlation coefficient (p = 0.84), the best-fit linear-regression analysis (R2 = 0.70), and the slope of the regression line (m = 1.247) demonstrated a high degree of correlation of the magnitude of regulation between the two assays (Figure S2). Primers design and qRT–PCR Primers' sequences for validation of microarray hybridization data were as described in [14]. And new primers for RNAi gene silencing and verification were designed with Primer 3 Program on a web-based server (http://frodo.wi.mit.edu/). All the primer sequences were listed in Table S1. Real-time quantitative PCR (qRT–PCR) to check the efficiency of gene silencing were done essentially according to [14]. The quantification was performed using the QuantiTect SYBR Green PCR Kit (Qiagen) and ABI Detection System ABI Prism 7300. All PCR reactions were performed in triplicate. Specificity of the PCR reactions was assessed by analysis of melting curves for each data point. The ribosomal protein S7 gene was served as internal control for normalization of cDNA templates. RNAi gene silencing and P. falciparum infection assays Sense and antisense RNAs were synthesized from PCR-amplified gene fragments using the T7 Megascript kit (Ambion). The sequences of the primers are listed in Table S1. dsRNA mediated gene silencing was done according to [14],[28]. About 80 4-d-old female mosquitoes were injected, in parallel, with GFP dsRNA as a control group or with target gene–specific dsRNA for the experimental group. Gene silencing in the whole mosquitoes was verified 3 to 4 d after dsRNA injection by qRT-PCR, done in triplicate, with the A. gambiae ribosomal S7 gene as the internal control for normalization. Gene silencing efficiency were listed in Table S1 with standard errors shown (KD%±SE). RNAi gene silencing in the midguts was verified by RT-PCR, 10 midguts were used for each replicate and at least two replicates were included with only one replicate shown (Figure S3). At least 50 control (GFP dsRNA–injected) and 50 experimental (gene dsRNA–injected) mosquitoes were fed on the same P. falciparum NF54 gametocytes culture at 3–4 d after the dsRNA injection. 24 hrs post blood feeding (pbf), the unfed mosquitoes were removed and the fed-mosquitoes were dissected at 7–8 d after feeding and midguts were stained with 0.2% mercurochrome [43]. Oocyst numbers per midgut were determined using a light-contrast microscope (Olympus). The median number of oocysts per midgut was calculated for each tested gene and for GFP dsRNA–injected control mosquitoes. The results for equal numbers of midguts from all three independent biological replicates were pooled. The dot plots of the oocysts number in each midgut within each treatment were presented by MedCalc software with the median value of the oocysts indicated. The Kruskal-Wallis (KW) test and Mann-Whitney test were used to determine the significance of oocysts numbers (p<0.05). Co-feeding and pre-injection of bacteria and P. falciparum infection assays About 80 4-day-old mosquitoes were first injected with PBS as control, or a mixture of live bacteria with approximately 30,000 E. coli and 60,000 S. aureus, or a mixture of heat-inactivated bacteria with the same number as the live ones. 24 hrs or 48 hrs after injection, mosquitoes were fed with P. falciparum NF54 gametocytes culture which were carried out according to our establish protocols [14]. For the co-feeding assay, the same sets of control PBS or bacteria were mixed in the blood meal to result in the same amount of either bacterium in the mosquito midguts. Unfed mosquitoes were removed, and the rest were kept in 26°C for 8 days before the oocysts counts. The infection phenotypes were determined as described above. Endogenous bacteria enumeration from mosquitoes' midguts Isolation and colony forming units (CFU) enumeration of bacteria from midguts of untreated control, antibiotic-treated mosquitoes and gene-silenced mosquitoes were done essentially according to [43] with modifications. The midguts from surface sterilized mosquitoes were dissected with sterilized PBS 4 d after dsRNA injection, and CFU were determined by plating the homogenate of the midguts with series dilutions on LB agar plates and incubating the plates at 27°C for 2 days. Each assay was performed with one midgut and at least 10 independent replicates were included for each gene. The species of the isolated bacteria were determined by amplifying a region of the 16s rDNA as described by using primers 27f and 1492r [52]. PCR products were sequenced and blasted against Nucleotide collection (nr/nt) database to verify the species. Immuno-fluorescent microscopy of ookinetes from bloodmeal and oocysts from midgut epithelium The early stages of P. falciparum development within untreated, antibiotic-treated and bacteria co-feeding mosquito midguts were compared by using the immuno-staining of ookinetes with anti-Pfs25 antibody (MRA-28, provided by MR4). Preparation of samples for immuno-fluorescence microscopy of malaria parasite within the bloodmeal was carried out based on [53] with substantial modifications. Sterile 0.5 ml “non-stick” low retention hydrophobic tubes (Alpha Laboratory Supplies) and sterile “non-stick” low retention hydrophobic pipette tips (Alpha Laboratory Supplies) were used to minimize malaria parasite loss during sample preparation due to their adhesion to plastic surfaces. The midguts including the entire bloodmeal contents were individually homogenized and diluted in 280 µl of PBS. 10 µl was then spotted, in duplicate, onto Teflon®-printed microwell glass slides (VWR International) previously coated with 3-aminopropyltriethoxysilane (APES) according to the supplier's instructions (Sigma). The sample slides were then air-dried, fixed in ice cold acetone for 2 mins and subjected to blocking in 10% goat serum for 1 hr, followed by the incubation with primary antibodies at 1∶400 dilutions for 2 hrs. After three PBS washes, sample slides were incubated with secondary antibodies (Molecular Probes, 1∶1000) for 2 hrs with Alexa Fluor 488-conjugated (green) goat anti-mouse antibody (1∶500 dilution). After another three PBS washes, sample slides were analyzed under a Nikon E800 upright microscope with epi-fluorescence. The total number of round forms, retort-forms and mature ookinetes in each spotted sample was counted. Average values for the densities of each malaria parasite stage present within each midgut examined were calculated from the three replicates. For checking the ookinetes and early oocysts in the midgut epithelium cells, at 24–30 hrs or 8 d after blood feeding, the midguts were dissected in 1% paraformaldehyde and washed with 3 times of PBS to remove the blood content and were subjected to the fixation in 4% paraformaldehyde (in PBS) for 1 hr and followed with 2 PBS washes. The midguts were then subjected to blocking and immune staining with primary antibody and secondary antibody as mentioned above. Midguts stained with pre-immune of anti-Pfs25 antibody were used as control. Midgut samples were mounted using the ProLong Antifade Kit (Molecular Probes) with DAPI staining of the cell nuclei and analyzed with same microscopy set as described above. Supporting Information Figure S1 Survival rates of A. gambiae Keele mosquitoes after P. falciparum infection. At least 40 mosquitoes were in each replicate, and three replicates were included with standard errors shown. Non-treated: septic mosquitoes harbor natural microbiota; Antibiotic-treated: mosquitoes treated with antibiotics, referred as aseptic mosquitoes; Rechallenged: aseptic mosquitoes co-fed with bacteria and P. falciparum infected blood. (0.02 MB PDF) Click here for additional data file. Figure S2 Validation of microarray-assayed gene expression with qRT-PCR. The values for the expression data obtained by microarray analysis (log2 ratio) for six genes were plotted against the corresponding expression values obtained with qRT-PCR (also log2 transformed) from two biological replicates of each experiment. Only the comparisons between the whole septic and aseptic mosquitoes which fed on sugar or uninfected blood were shown here. (0.01 MB PDF) Click here for additional data file. Figure S3 Verification of gene silencing in the mosquito midgut tissue 4-d post dsRNA injection. dsGFP-injected mosquito midguts were used as controls, and 10 midguts were included in each replicate and at least two replicates were done with only one replicate shown here. Def1: defensin 1, Gam: gambicin; Cec: cecropin. (0.09 MB PDF) Click here for additional data file. Table S1 Primers used for dsRNA synthesis, qRT-PCR validation of RNAi-mediated gene silencing and the efficiency of gene silencing. (0.01 MB PDF) Click here for additional data file. Table S2 List of genes identified from microarray analysis exhibiting significant differential expression between untreated septic and antibiotic-treated aseptic adult female A. gambiae Keele mosquitoes fed with sugar (7-day-old whole mosquitoes minus head). (0.06 MB XLS) Click here for additional data file. Table S3 List of genes identified from microarray analysis exhibiting significant differential expression between septic and aseptic adult female A. gambiae Keele mosquitoes fed on uninfected blood (7-day-old whole mosquitoes minus head). (0.06 MB XLS) Click here for additional data file. Table S4 List of genes identified from microarray analysis exhibiting significant differential expression in the midguts of 7-day-old female A. gambiae Keele mosquitoes 12 hrs after feeding on uninfected blood supplemented with E. coli and S. aureus, PBS substitution of bacteria as control. (0.17 MB XLS) Click here for additional data file. Table S5 Table S5. List of genes identified from microarray analysis exhibiting significant differential expression in the carcass of 7-day-old female A. gambiae Keele mosquitoes 12 hrs post feeding on uninfected blood supplemented with E. coli and S. aureus, PBS substitution of bacteria as control. (0.15 MB XLS) Click here for additional data file.

0 comments Cited 313 times – based on 0 reviews      Review now

Bookmark

Record: found
Abstract: found
Article: found

Is Open Access

Reducing Plasmodium falciparum Malaria Transmission in Africa: A Model-Based Evaluation of Intervention Strategies

Jamie Griffin, T. Hollingsworth, Lucy C Okell … (2010)

Introduction Over the past five years, dramatic declines in malaria disease caused by Plasmodium falciparum have been reported across a range of settings within sub-Saharan Africa. These declines are associated with increased distribution of long-lasting insecticide-treated nets (LLINs) and with the switch from a failing drug regimen to artemisinin-based combination therapies (ACT) as first-line therapy [1]–[4]. Whilst this pattern of reducing disease is encouraging, there remain many countries within Africa that continue to have a high burden of disease and hence malaria remains a leading cause of mortality in children under five years of age [5]. Thus control of the disease, and ultimately elimination of the parasite in this continent, remain major public health goals. Eradication of malaria was attempted in the 1950s under the auspices of the World Health Organization-led Global Malaria Eradication Program (GMEP) [6]. Notably, Africa was not formally included in this program despite clear evidence of the large disease burden within the continent at that time. However, elimination campaigns were subsequently undertaken on a smaller scale within Africa, most prominently in two areas of moderate to high transmission in Nigeria (the Garki project [7],[8]) and on the Kenyan/Tanzanian border (the Pare-Taveta project [9]), but also periodically in areas of lower transmission including the Kenyan highlands [10] and the island of Madagascar [11]. These campaigns included frequent insecticide spraying of houses to reduce the vector populations and rounds of mass treatment to reduce the human infectious reservoir. Whilst substantial declines in infection and disease were observed in all of these campaigns, the control measures were not sufficient to eliminate the parasite on a short time scale, and failure to sustain control programs inevitably led to rebound of infection and disease in later years. This under-performance was perceived as a lack of success by past eradication attempts, which may in part be attributed to over-optimism about what could have been achieved with the tools then available [12]. Two years ago, following a renewed commitment to malaria control from donor organizations, the focus shifted again to malaria eradication as an ultimate goal. Previously, many countries had already intensified their own malaria control programs with much success in reducing both the burden of disease and ongoing transmission [1]–[4],[13],[14]. However, Africa poses the biggest challenge to a global eradication initiative, given the heterogeneous yet ubiquitous nature of P. falciparum transmission across much of the continent. Levels of transmission in Africa range from absent or low in many urban areas, through epidemic outbreaks in the highlands, to highly seasonal or perennial transmission in rural areas [15],[16]. This variable transmission pattern is further complicated by local variation in the major Anopheles vector populations that sustain transmission (principally An. gambiae s.l. and An. funestus, although approximately 70 relevant species have been identified worldwide [17]). Of the 47 countries within sub-Saharan Africa, the majority are currently classified by WHO/Roll-Back Malaria as being in the control stage and thus need to scale up interventions to sustain control and reduce the burden of disease via a reduction in transmission [18]. On the northern borders of the continent, transmission is already low, with Egypt and Algeria in the elimination phase and Morocco and Mauritius having interrupted local transmission. Similarly, in the southernmost countries, a sustained move towards local control and potentially elimination in border areas has been agreed upon via cooperation with neighbouring countries (the “elimination eight”) [19]. On the island of Zanzibar, a highly successful control program has reduced transmission to very low levels. However, a recent assessment of the feasibility of moving to elimination concluded that, whilst it is technically feasible to reduce local transmission to zero in this setting, the resources, both financial and operational, required to sustain elimination in the face of repeated reintroduction from mainland Africa make this a difficult prospect [20]. Compared to the past campaigns in the 1950s, additional tools are now available which, combined with sustained policy commitment, may make local elimination achievable in some settings and can aid control of disease by dramatically reducing malaria prevalence in countries with high rates of ongoing transmission. These include new LLINs, which have increased killing effects on the vectors compared to traditional nets and are more durable [21]–[25], and ACTs, which, through their gametocytocidal effect, can impact transmission from humans to vectors [26],[27]. In addition, a pre-erythrocytic malaria vaccine, RTS,S, has shown promising results in Phase II trials [28]–[34] and could soon contribute to elimination programs. National control agencies have varying levels of resources but can rarely implement all major control interventions at a given time. Understanding how to choose policy that is appropriate to the local setting is therefore key to effective control. Whilst the efficacies of most interventions have been individually evaluated in the field, the impact of different combinations of these is not clear. Field trials will be important to inform control policies but will be able to test only a few of the combinations of interventions in a limited number of settings. Mathematical models provide a tool with which to explore the expected impact of different interventions against malaria, both individually and in combination, on a range of program endpoints [26],[35]–[40]. Whilst simple models can provide important general insights, the heterogeneity in transmission intensity [41]–[43], the variability in vector species composition and associated bionomics [17],[35], and the seasonality in vector populations [44] are all important factors that affect the transmission potential of a site and the likely impact of intervention packages. We therefore constructed an individual-based simulation model which captures these key factors while remaining sufficiently mathematically tractable to enable the baseline model parameters to be rigorously fitted to data within a Bayesian framework. The model includes the suite of current tools most often employed by (or likely to be employed by) National Malaria Control Programs—namely, LLINs, IRS, ACTs in case treatment and in mass treatment campaigns, and a vaccine with characteristics similar to the RTS,S/AS01 vaccine now in Phase III trials. The principal aim of the modelling presented here is to explore the potential for current control measures to reduce parasite prevalence to a low level (defined here as below a threshold of 1% prevalence across all age groups detected through microscopy which represents a level below which surveillance would likely switch to case detection) as laid out in the control phase of the global elimination framework [45]. We illustrate our results by applying our model to six well-characterized transmission sites which represent the full range of transmission intensity–vector species combinations and seasonality patterns most commonly observed across Africa. Methods Simulation Model for Malaria Transmission Dynamics We developed a stochastic simulation model for P. falciparum transmission dynamics in which people are represented as individuals while vectors are represented as aggregated populations, stratified by species. The model builds on an earlier compartmental model which incorporates the acquisition and loss of immunity to disease and to detectable parasitaemia [46],[47], but is extended to incorporate infection-blocking immunity and heterogeneity in biting rates. Full technical details are given in Protocol S1 and the flow diagram is presented in Figure 1A. Briefly, individuals begin life susceptible (S) to infection but with partial maternal immunity determined by the level of immunity in women of childbearing age. Maternal immunity decays in the first six months of life, thereby increasing susceptibility to disease. Individuals become infected at a rate determined by the force of infection in the population (Λ), which is determined by the ratio of vectors to humans, the biting rate per mosquito on humans, the proportion of infectious mosquitoes in the vector population, and the person's level of anti-infection immunity. On infection, they pass through the liver (pre-patent) stage and then either develop clinical disease (with a probability φ determined by their current level of anti-disease immunity) or develop patent (detectable under microscopy) asymptomatic infection (1−φ). Those who develop clinical disease have a fixed probability (fT ) of being treated successfully (T), in which case they will clear infection and, depending on the drug, enter (with rate rT ) a period of prophylactic protection (P) before returning (rP ) to being susceptible to new infection. Those who fail treatment (1−fT ) are assumed to eventually clear disease (D) and become patently asymptomatic (A) with rate rD . From patent asymptomatic infection, individuals will eventually move to a sub-patent stage (U) which can be an important component of the infectious reservoir [48], at a rate (rA ) that depends on their current level of anti-parasite immunity. Sub-patent infection is eventually cleared (rU ) and individuals return to being fully susceptible. From all infected states, acquiring a new infection in the presence of an existing infection (superinfection) is possible. Rather than explicitly tracking mixed infections, we assume that the new infection dominates and thus individuals move to either the clinical disease or asymptomatic states dependent on their level of anti-disease immunity. Individuals become infectious to vectors, at differing rates, in the clinical disease, patent and sub-patent asymptomatic stages—the states that compose the human infectious reservoir (Figure 1D). Four types of human immunity are included and are modelled dynamically. Maternal immunity, which protects against clinical disease, is assumed to decay exponentially from birth. Anti-disease immunity, which reduces the probability of developing clinical symptoms on infection, and infection-blocking immunity, are both exposure-driven whilst anti-parasite immunity, in which individuals control parasite densities and thus leave the patent infection state more quickly, is assumed to develop with age, conditional on having been exposed. 10.1371/journal.pmed.1000324.g001 Figure 1 Transmission model; EIR, prevalence and seasonality; and infectious reservoir. (A) Flow diagram for the human component of the model. S, susceptible; T, treated clinical disease; D, untreated clinical disease; P, prophylaxis; A, asymptomatic patent infection; U, asymptomatic sub-patent infection. (B) The relationship between EIR and parasite prevalence in children under 15 y. Solid line: fitted relationship; filled circles: data representative of this age group; open circles: data from other age groups (mostly younger) used in the model fitting. (C) The relationship between transmission intensity characterized by EIR and seasonality, defined as the proportion of EIR over a single calendar year that occurs within the peak three months of transmission. The colours of the markers indicate the different transmission settings and the shapes the species. (D) The estimated age-specific infectious reservoir for the different transmission settings defined in (C), with the same colours as (C). This is defined as the product of the age-specific biting rate, age-specific prevalence states (T, D, A, and U), state-specific onward infectivity to mosquitoes and the size of the population at this age. Three Anopheles vector species (An. gambiae s.s., An. funestus, and An. arabiensis) are modelled explicitly as the predominant vectors in the transmission sites that we consider. Vectors begin susceptible and on taking an infectious bite move into a latent state. From this they become infectious to humans, with infectivity determined by their human blood index (HBI) and biting rate and are assumed never to recover before death. Vector density is assumed to follow a seasonal pattern as determined by fitting an appropriate functional form to entomological data from the areas considered (see Table 1 and Protocol S4). 10.1371/journal.pmed.1000324.t001 Table 1 Summary of the six malaria transmission settings considered here. Country Location Population Type of Transmission Reported Annual EIR (ibppy) Fitted Annual EIR (ibppy) Anopheles Species Relative Abundance Reference Cameroon Nkoteng Rural Moderate, perennial 94 81 72% An. funestus; 28% An. gambiae s.s. [97] Democratic Republic of Congo Kinkole Rural Moderate, perennial 48 43 Nearly 100% An. gambiae s.s. [98] Ghana Kassena-Nankana District Rural High, seasonal 630 586 60% An. gambiae s.s.; 40% An. funestus [99] Mozambique Matola, Maputo Coastal suburb of capital Moderate, perennial 28 46 42% An. arabiensis; 46% An. funestus (additional 12% An. coustani are not considered here) [100] Tanzania Matimbwa Rural High, seasonal 703 675 85% An. gambiae s.s.; 10% An. funestus; 5% An. arabiensis [101] Uganda Kjenjojo Kasiina Rural Low 7 3 65% An. gambiae s.s.; 35% An. funestus [102] Model Parameterization Model parameterization was undertaken in several stages. First, a literature search was undertaken to formulate prior distributions for all model parameters. Where there was no information in the literature, vague priors were used or parameters were fixed if they could not be identified from subsequent model fitting. The human model parameters were estimated by fitting the equilibrium model conditional on EIR using Bayesian Markov Chain Monte Carlo (MCMC) methods to data on the stationary distributions of parasite prevalence (by both microscopy and PCR) by age from 34 locations across a wide range of transmission intensities from Africa (see Protocol S3) and of clinical disease incidence from two settings in Senegal [49]. Site-specific prior distributions for EIR were used based on published data ([50] and Protocol S3). By fitting the model to these data we were able to characterize the relationship between EIR (ibppy, the number of infectious bites per person per year) and parasite prevalence (Figure 1B). The parameters determining the onward transmissibility of the human infectious stages (clinical disease, patent and sub-patent infection) to mosquitoes were obtained by model fitting to data from human feeding studies and the Garki project [7],[51]–[54]. These parameters combined with parasite prevalence determine the age profile of the infectious reservoir (Figure 1D) [55],[56]. Only age-targeted strategies are sensitive to this profile. Parameters for the vector model were taken from the literature. A full listing of model parameters, their prior and posterior medians, and literature sources are given in Table S3.1 in Protocol S3. To run the model in specific settings, data on vector species composition, their seasonal profile, and the intensity of transmission (EIR) were extracted from the literature (Table 1, Figure 1C, Figure 2, and Protocol S4). A functional form was fitted to monthly data on either EIR or vector density to enable a single seasonal driver input (emergence of vectors) into the model. Full details of the settings and the seasonal profile fitting are in Protocol S4. 10.1371/journal.pmed.1000324.g002 Figure 2 Fitted seasonal profile of EIR for the six transmission settings by vector species. The fitted seasonal profiles of EIR per day and fitted annual EIR were obtained by fitting a transformed sinusoidal function to reported time series of either EIR or mosquito densities in the settings (see Protocol S4). Grey, total; red, An. gambiae s.s.; blue, An. funestus; green, An. arabiensis. (A) Nkoteng, Cameroon; (B) Kinkole, DRC; (C) Kassena-Nankana District, Ghana; (D) Matola, Maputo, Mozambique; (E) Matimbwa, Tanzania; (F) Kjenjojo Kasiina, Uganda. Interventions The implementation of each intervention is described briefly below. Full mathematical details and tables of parameter values are provided in Protocols S2 and S3. Long-lasting insecticide-treated nets We adapted an existing model [36] to our individual-based framework. Nets are assumed to have four effects: direct killing of a mosquito that lands on them, repellency which results in a longer gonotrophic cycle and possible diversion to a non-human blood host, a direct protective effect for the individual sleeping under the net, and a reduction in transmission from infected individuals sleeping under the net to susceptible mosquitoes. The degree of indoor-biting (endophagic) behaviour for the different species is incorporated into the model when assessing the LLIN effect. These behaviours are assumed to remain constant throughout the intervention. Indoor residual spraying IRS was added to the LLIN model as an additional intervention which can kill mosquitoes as they rest within the house or repel them before they feed. In the model the repellency effect extends the duration of the gonotrophic cycle in the same way as the repellency effect of LLINs. For IRS the killing effect depends primarily on the indoor-resting (endophilic) nature of the species as well as its HBI. Simulations assumed a DDT-like insecticide with a half-life of 6 mo which acts by repelling and killing mosquitoes [57]. Switch to ACT as first-line treatment Effective treatment (i.e., treatment which fully clears infection) was assumed to be given to a proportion of those developing clinical disease. Treatment failures were not explicitly modelled but are assumed to follow the same infection path as untreated infections. The half-life of the drugs pre-ACT (where we assume sulphadoxine-pyrimethamine [SP] was first-line therapy) and following ACT introduction determine the period of prophylaxis. In addition, the gametocytocidal effect of ACTs was incorporated as a reduction in onward infectiousness as in a previous model, based on data from human-to-mosquito transmission experiments involving treated patients [27]. Mass drug administration We considered the impact of a mass screening and treatment approach (MSAT) using a single dose of an ACT. We assumed that a rapid diagnostic test (RDT) would have approximately the same sensitivity as microscopy and thus all those in the clinical disease or asymptomatic patent infection stages would receive the drug, but that the uninfected and sub-patent infected individuals would not. The ACT was assumed to clear any infection present and provide a period of prophylactic protection (25 d, corresponding to an artemisinin coupled with a drug such as SP). The coverage level refers to the number of individuals screened. Pre-erythrocytic vaccine A pre-erythroyctic vaccine was assumed to reduce the probability of transmission from mosquitoes to people. It remains unclear whether this lower exposure to infection will affect the development of anti-disease immunity. Here we assume that it does but that it has no effect on the development of anti-infection immunity. Individual vaccine efficacy was assumed to decay exponentially with a half-life of 3 y. The vaccine is delivered through the Expanded Program for Immunization (EPI) and given at ages 3–5 mo, or as a mass vaccination program across all ages every 3 y. Transmission Settings We considered the impact of these interventions, individually and in combination, in six different settings that characterize the spectrum of transmission patterns of P. falciparum across Africa. These settings range in transmission intensity from measured EIRs of approximately 5 to over 500, translating in our model to parasite prevalence in 2- to 10-year-olds of 14% to 85%. In 2007, 80% of Africa's population was estimated to reside in an area with parasite prevalence in 2- to 10-year-olds of >5% and 50% in an area with prevalence >40% [16]. These specific settings, summarized in Table 1, Figure 1C, and Figure 2, were chosen because of the large number of both entomological and clinical studies undertaken in these areas and to represent patterns of perennial/seasonal transmission of varying intensity and with different mixes of Anopheles species. We fitted the model to data from these settings, which we take as our baseline scenarios. For each scenario, we present the mean of ten simulation runs in a population of 10,000 individuals, which was sufficient to approximate the dynamics in a larger population. The population size was assumed to be static over time, with age structure based on data from Tanzania. After introducing infection, the model was run for 50 y to reach equilibrium representing the situation in the year 2000. Between 2000 and 2010 we increased the distribution of LLINs from a baseline of zero coverage to a maximum of 20% coverage [58] and implemented a switch to ACT as first-line therapy in the year 2000. Combinations of interventions were then introduced from 2010 onwards. Note that this does not necessarily reflect the true intervention programs in place in these settings in these years, and hence model outputs do not directly predict expected patterns in these settings; rather they give an indication of the likely effectiveness of the modelled intervention packages in different setting types. Intervention Package Scenarios Coverage here is defined as the proportion of individuals receiving an intervention (for LLINs ownership, for vaccination those receiving the vaccine, for IRS those that reside in houses where spraying occurs, and for MSAT the number of individuals screened). We separately consider the impact of adherence/usage for LLINs, which is assumed to decay over time. This proportion of people using LLINs is termed effective coverage. For IRS we assume no loss of adherence. For MSAT we assume that all those who are screened and positive on microscopy take the drug. Similarly, for the vaccine we assume that those offered it accept. Finally, for all interventions there is a decay in protective efficacy over time for those who have received and use the intervention. For LLINs this is due to wear-and-tear and loss of insecticidal effect. For IRS we model the loss of insecticidal effect. For vaccines we assume that efficacy declines through waning protection. Unless stated otherwise we assumed that IRS and MSAT were given at 80% coverage (the maximum achievable in well-managed control programs [59]) and the vaccine at 90% coverage (based on EPI distribution statistics). For the roll-out of LLINs we considered two realistic scenarios. In the first, distribution was increased gradually to a maximum of 80% within 5 y and a new net was distributed to individuals every 5 y. In the second, we assumed almost immediate distribution at 80% coverage, redistribution every 5 y, plus delivery of a net to 80% of newborn infants and an average of 0.75 adults for every infant who receives a net (Figure 3A). These coverage levels are similar to the targets set for 2010 for scaling up for impact in the Global Malaria Action Plan [18]. In both scenarios, we assume that LLIN use wanes over time so that effective coverage is lower. Here we assumed an exponential decay at a rate 0.2 per year so that after 5 y effective coverage is approximately 37% of the baseline level. We also considered the impact of a theoretical (unachievable) maximum of 100% coverage with LLINs coupled with no decay in usage over time (Figure 3A). Protective efficacy of the nets due to decaying insecticide efficacy and wear-and-tear was assumed to decay exponentially with a half-life of 2.64 y (Protocol S3 and [60]). We did not consider any decay in effective coverage of IRS as we assumed that coverage remained constant at each round (i.e., people do not refuse to have their house sprayed as the intervention goes on). The protective efficacy of DDT was assumed to decay exponentially with a half-life of 6 mo (Protocol S3 and [57]). Adherence to LLINs given receipt was assumed to be independent of IRS acceptance. 10.1371/journal.pmed.1000324.g003 Figure 3 Impact on parasite prevalence of LLINs alone. (A) Example of the ways in which coverage of LLINs is considered to increase in various model scenarios. Baseline (blue): our baseline scenario in which 80% coverage is achieved over five years but adherence also decays between net distribution rounds; rapid (brown): as baseline but with more rapid scale-up to 80% coverage; rapid, no drop-out (green): rapid scale-up to 80% coverage with no decay in adherence; 100% (yellow): 100% coverage with no decay in adherence (theoretical maximum effect). (B) Model-predicted impact on parasite prevalence over calendar time of four scenarios for LLIN scale-up combined with an earlier switch to ACT as first-line therapy in Kjenjojo Kasiina, Uganda. (C) Final parasite prevalence and (D) absolute reduction in parasite prevalence after 15 years of a sustained intervention program in the six transmission settings with the baseline scenario for LLIN distribution, the rapid scenario for LLIN distribution, and the rapid scenario with no loss of adherence for LLIN distribution. We undertook preliminary runs for IRS and MSAT to identify the optimal time of year for annual programs. The optimal time was defined on the basis of providing the maximum reduction in mean prevalence of parasitaemia across all age groups in year 10 of the intervention campaign. We found that in those settings which have a clear seasonal peak in the EIR, it is always optimal to spray just before the upward trend in EIR. In settings with less seasonality, there is less difference in impact, but spraying at the start of the main transmission season tends to remain optimal. In contrast, across most settings, the optimal time of year to mass treat in terms of reducing overall prevalence of asexual parasitaemia as an endpoint is at the beginning of the period of lowest EIR (also shown in [61]), which generally occurs approximately 2 mo after peak slide prevalence. For scenarios in which IRS and MSAT were undertaken every 6 mo, they were implemented at the optimal time of year as defined above, plus 6 mo thereafter. Effective coverage and protective efficacy do not alone determine intervention effectiveness, as they also depend on whether the same individuals receive multiple interventions or whether interventions are randomly distributed across the population. We therefore allowed correlations between repeat distribution for each individual intervention (where a correlation of 0 means that redistribution is completely random and of 1 that redistribution always occurs to those who had previously received the intervention). We also allowed correlations between receiving LLINs, IRS, and MSAT. Here a positive correlation means that individuals who receive one intervention are also more likely to receive the other (which could reflect access to interventions) whilst a negative correlation means that those who receive one intervention tend not to receive the other (which would reflect a propensity not to use multiple interventions). As our focus is to consider intervention packages aimed at reducing transmission, our primary outcome was the annual mean prevalence of asexual parasitaemia as measured by microscopy in the whole population up to 25 y following the start of the intervention program. We chose this rather than prevalence restricted to children as it enables us to correctly compare age-targeted interventions. We specifically do not focus on short-term “predictions” or timelines, as our sensitivity analysis shows that these are highly dependent on parameters relating to the loss of acquired immunity (which impact the fitted duration of infection). Currently these parameters are not well-estimated from the available data (see Section 5.2.1 in Protocol S5, and Box 1). Furthermore, time scales of impact will inevitably depend on the speed with which scale-up of interventions occurs and so cannot be reliably predicted without detailed assessment of local situations. Box 1. Uncertain Parameters Whilst models can be useful tools in setting realistic expectations for intervention programs, some key parameters in our current model are based on limited data. Further empirical work in these areas could improve future models. These include: The duration of natural infection and the extent to which super-infection prolongs this duration or increases infectivity; The rate of acquisition of immunity at different transmission intensities, and the rate of loss of immunity when transmission is reduced; The bionomics of the principal vector species and the impact of vector-targeted interventions on them; Detailed data on the speed with which coverage of interventions is scaled up, heterogeneity in coverage levels achieved, and the degree of adherence to the interventions over time. Software A user-friendly software package for reproducing the simulations presented here, as well as other potential combinations of the interventions included in this paper, is freely available to download from our Web site (http://www1.imperial.ac.uk/medicine/about/divisions/publichealth/ide/research_groups/malaria/). A short summary of the interface is provided in Protocol S6. Results Continued Scale-Up of LLINs Continued scale-up of LLINs from the baseline assumption of 20% coverage could potentially reduce transmission across all six transmission settings, given that the dominant vector species in these settings are primarily endophagic and their peak biting times coincide closely with normal sleeping hours [62] (provided changes in mosquito behaviour in response to the interventions are not dramatic). However, the magnitude of the effect will depend not only on the intensity of transmission in each setting but also how roll-out is achieved, the final level of coverage, adherence to LLIN use, and the decay in insecticide effectiveness over time. Figure 3A shows four potential scenarios for scale-up if nets are redistributed every 5 y. Theoretically, the greatest impact is achieved with rapid deployment, 100% coverage, and perfect adherence. However, even at this unrealistically high level, the efficacy will be less than its maximum due to decaying effectiveness of the insecticide. Even at the target coverage levels of 80%, with gradual roll-out and realistic adherence, effective coverage levels can, on average, be as low as 50% (Figure 3A). The additional decay in insecticide efficacy over time can result in protective coverage levels as low as 30%. This is even without the additional limitation of an interrupted supply chain, which is likely to reduce effective coverage further [63]. In the low-transmission setting of Kjenjojo Kasiina, Uganda, the basic reproduction number (R 0) is already close to 1 in the absence of additional interventions. Thus, parasite prevalence can be reduced to below the 1% threshold over a 15 y time horizon with LLIN use alone (Figure 3B). However, even in this relatively low-transmission setting, high levels of coverage and adherence are required. Furthermore, with decaying adherence in their use it is likely that transmission will be sustained, albeit at a low level. Furthermore, if LLINs have a lower killing effect than that assumed here, our model would predict sustained transmission in this setting (Figure S5.6 in Protocol S5). In contrast, in the moderate-transmission settings of Nkoteng (Cameroon), Kinkole (Democratic Republic of Congo or DRC), and Maputo (Mozambique), and in the high-transmission settings of Kassena Nankana District (KND) (Ghana), and Matimbwa (Tanzania), scale-up of LLINs alone does not reduce parasite prevalence to below 1%, even over longer time periods (Figure 3C). We can, however, expect to see dramatic declines in the first five years of the program followed by an increase to new endemic levels as levels of immunity in the population change (Figure S5.1 in Protocol S5). The time scale of this rebound is difficult to ascertain from current data due to uncertainty in the rate of loss of acquired immunity (see section 5.2.1 in Protocol S5, and Box 1). In high-transmission settings, with continued scale-up of LLINs to 80% coverage within five years, the absolute drop in prevalence is between 5% and 10%. If rapid scale-up occurs and adherence is sustained, drops in prevalence of 20%–25% can be expected (Figure 3D). However, despite the smallest relative impact occurring in the high-transmission settings, because most cases of infection and disease occur in these settings, the absolute impact in terms of numbers of infections averted will be greater. Thus, in terms of reduction in infections per net distributed, impact will be greatest in these high-transmission settings. Additional Use of IRS and MSAT Whilst continued scale-up of LLINs is predicted to reduce transmission substantially, under realistic assumptions about the level of coverage and adherence to LLIN use, additional tools will be necessary in many settings. In Kjenjojo Kasiina, Uganda, yearly rounds of IRS with DDT combined with continued scale-up of LLINs to 80% coverage is predicted to locally eliminate transmission (Figure 4A). Yearly rounds of MSAT as an alternative to IRS tend to have less impact although this would also achieve a reduction below the 1% parasite prevalence threshold. 10.1371/journal.pmed.1000324.g004 Figure 4 Impact of combining LLINs with IRS and MSAT. (A–F) Impact of intervention scenarios incorporating IRS and MSAT on parasite prevalence in the six transmission settings. All scenarios include the earlier switch to ACT as first-line therapy. “LLIN only” uses the baseline scale-up for coverage. All other scenarios include LLIN scale-up using the baseline scenario except where noted. (G and H) Final parasite prevalence and absolute reduction in prevalence after 15 years of a sustained intervention program in the six transmission settings with baseline scenario for LLIN distribution; baseline LLIN + yearly MSAT; baseline LLIN + yearly IRS; baseline LLIN + yearly MSAT + yearly IRS. In the moderate-transmission setting of Kinkole, DRC, more intensive rounds are required. Thus, in this setting, twice yearly IRS and MSAT are required to reduce parasite prevalence below the 1% threshold (Figure 4B). In contrast, in the slightly higher-transmission setting of Nkoteng, Cameroon, this is not sufficient in itself and additional faster scale-up of LLINs is needed to achieve this threshold (Figure 4D). In Maputo, Mozambique, in which transmission intensity as measured by EIR is similar to Kinkole, DRC and lower than Nkoteng, Cameroon, even these more intense programs are unable to reduce prevalence below the 1% threshold (Figure 4C). This is due to the high proportion of transmission that occurs via An. arabiensis in this setting, whose more exophilic behaviour reduces the impact of IRS on transmission. Assuming a lower degree of exophilic behaviour of this species compared to our baseline assumption, this conclusion continues to hold (section 5.2.2 in Protocol S5). In all three moderate-transmission settings, IRS with an insecticide similar to lambdacyhalothrin (which is less repellent and hence more lethal but has a shorter half-life than DDT) is predicted to have a lesser effect on transmission than DDT (Figure S5.7 in Protocol S5). In both high-transmission settings (KND, Ghana and Matimbwa, Tanzania), current tools are insufficient to reduce parasite prevalence below the 1% threshold (Figure 4E and 4F; see also higher levels of adherence and coverage in Figure S5.3 in Protocol S5, and higher frequency of MSAT in Figure S5.8 in Protocol S5). However, in both settings, an intense program involving rapid scale-up of LLINs with sustained adherence and twice-yearly rounds of MSAT and IRS could result in marked declines in prevalence from 60% to 10% in the population as a whole (Figure 4E and 4F). However, in these settings, the interventions would need to be sustained indefinitely to maintain this new endemic level. Yearly IRS and MSAT combined with 80% coverage of LLINs is predicted to reduce parasite prevalence after 15 y to below 10% in moderate transmission settings and below 25% in high-transmission settings (Figure 4G). Again, the absolute reduction will be greatest in the latter, with a 40%–50% drop in parasite prevalence in these settings (Figure 4H). Targeting and Overlap in Intervention Coverage LLIN distribution programs initially focused on young children as one of the high-risk groups for developing severe disease. However, as shown in Figure 5A and elsewhere [23],[24], this strategy is unlikely to have an additional impact on transmission, because the youngest children tend not to be major contributors to the infectious reservoir (Figure 1D). However, if limited coverage is achievable, substantially greater reductions in prevalence could be obtained if, for a given level of distribution, nets were targeted towards those living in the local foci of transmission which impact strongly on sustaining transmission [38],[42],[43],[64]. Thus in Kinkole, DRC, in a program of LLIN distribution with a low 20% coverage, if distribution is prioritised to those at highest risk we could expect a reduction in prevalence after 15 y of approximately 6% compared to a reduction of 3% if the same number of nets were distributed randomly. A similar picture emerges for MSAT programs (Figure 5B), although the effect of targeting is greater for LLINs because in addition to their direct protective effect, they kill mosquitoes in proportion to the rate at which the protected person would have been bitten. 10.1371/journal.pmed.1000324.g005 Figure 5 The effect of non-random distribution of interventions. (A and B) Parasite prevalence after 15 years of an intervention program as a function of the target coverage of (A) LLIN distribution and (B) MSAT for Kinkole, DRC. Blue: if the intervention is distributed randomly; green: if the intervention is preferentially distributed to the youngest children; red: if the intervention is preferentially distributed to those who are bitten most frequently (excluding age dependency in biting rates). (C and D) Parasite prevalence after 15 years of a single intervention program as a function of the frequency of the intervention and whether successive rounds are given randomly (green) or to the same people (purple) for Kinkole, DRC. (C) IRS; (D) MSAT. (E and F) Parasite prevalence in all individuals (red), in 2- to 10-year-olds (blue) and EIR (green) after 15 years of a combined intervention program as a function of the correlation in receipt of the two interventions for KND, Ghana. A correlation of 0 represents random distribution at each round, 1 represents those receiving one intervention also receive the other and −1 represents those receiving one intervention do not receive the other. (E) IRS and LLIN; (F) IRS and MSAT. For (E) and (F) there is 50% coverage per round for IRS and MSAT and the baseline scenario for LLINs. With any intervention, it is likely that the same individuals or villages will tend to access the intervention at each distribution round. Thus for example, if 80% coverage of LLINs is achieved, but at each redistribution the same 80% receive the intervention, then after three rounds of redistribution the percentage of the population ever receiving an LLIN is 80%. However, if this 80% coverage reflects random distribution, then after three rounds the percentage of the population ever receiving an LLIN is 100×(1−0.2×0.2×0.2) = 99.2%. Figure 5C and 5D shows the predicted effect of rounds of IRS and MSAT between these two extreme (systematic versus random coverage) scenarios. In both cases, assuming random distribution results in an overestimate of the effect of the intervention, and this difference increases the more frequently IRS or MSAT is undertaken. Thus, to optimize program effectiveness it is necessary to ensure that as wide a proportion of the target population is reached by the intervention. In addition to correlations between those who receive an individual intervention, there is likely to be overlap in those who are offered different interventions. This is likely to be most strongly correlated for IRS and LLINs, given the perception of these interventions as providing direct protection to the individual or household. A positive correlation will occur if the same individuals access the interventions. Under these scenarios, we can expect the least impact of the intervention program (Figure 5E). However, if uptake is negatively correlated, for example if those who are offered IRS and LLINs choose only to have one, for the same overall coverage levels of the individual interventions total population coverage is increased over and above naïve expectations assuming both are randomly distributed. This increased total coverage results in the largest reductions in transmission (Figure 5E). Similar effects are observed for IRS and MSAT, although again, this is not as pronounced as for LLINs given that there is less redundancy between IRS and MSAT than between two antivectorial measures (Figure 5F). Additional Impact of RTS,S/AS01 Vaccine In the low-transmission setting of Kjenjojo Kasiina, Uganda, RTS,S (when it becomes available) could further reduce transmission and thus negate the need for additional rounds of IRS to speed declines. As found by others [65],[66], vaccination at birth under the EPI is expected to have relatively little impact either with or without additional rounds of MSAT (Figure 6A). If mass vaccination every 3 y is undertaken as an alternative alongside the baseline scale-up of LLINs to 80% coverage, prevalence is predicted to fall to under 1%. 10.1371/journal.pmed.1000324.g006 Figure 6 Impact of additional vaccination on parasite prevalence in the different transmission settings. All runs assume the RTS,S vaccine is 50% efficacious and has a half-life of 3 years. PEV at EPI denotes the pre-erythrocytic vaccine being given through the Expanded Program on Immunization, whilst mass PEV denotes a mass vaccination campaign. All runs include LLINs. (A) PEV at EPI with or without additional MSAT in Kjenjojo, Uganda (B) Mass PEV with or without additional MSAT in Kjenjojo, Uganda (C to F) MSAT and IRS with mass PEV in: (C) Kinkole, DRC, (D) Maputo, Mozambique, (E) Nkoteng, Cameroon and (F) KND, Ghana. In the moderate transmission settings of Kinkole, DRC (Figure 6C), Maputo, Mozambique (Figure 6D), and Nkoteng, Cameroon (Figure 6E), continuation of programs incorporating IRS and MSAT in addition to LLIN distribution will be needed even if a vaccine is available. However, with a mass vaccination program prevalence in all three sites can be driven below 5%. In Maputo especially, where IRS is predicted to be less effective, an additional vaccination program has a noticeable further impact on prevalence. In both high transmission settings (KND, Ghana, Figure 6F; and Matimbwa, Tanzania, results not shown), mass vaccination results in modest reductions in prevalence. Across all transmission settings, a more efficacious vaccine with a longer duration of protection would further reduce transmission (section 5.2.6 in Protocol S5). Discussion If deployed in combination, current interventions can result in substantial declines in malaria prevalence across a wide range of transmission settings. Our results show that in areas with relatively low transmission (EIR<10 ibppy), increased distribution and use of LLINs, coupled with the switch to an effective ACT as first-line therapy, could reduce transmission to very low levels if high levels of coverage and adherence are achieved. Defining low-transmission areas as those where parasite prevalence in 2- to 10-year-olds is under 25%, approximately 20%–50% of individuals living in areas of stable risk of P. falciparum transmission in Africa live in such settings [16]. Additional use of IRS and/or MSAT in these settings would speed this reduction and also allow overall parasite prevalence to be reduced to <1% even if adherence to LLIN use is not perfect. These results agree with recent observations made in a very low transmission setting in Western Kenya, in which the parasite appears to have been eliminated in an area in which ACT and LLIN usage have been coupled with IRS rounds [67]. Large reductions have also been achieved in Zanzibar, where the preintervention parasite prevalence was 9% in children aged 0 to 5 y and 12.9% in children aged 6 to 14 y [13]. After a switch to ACT as first-line therapy and high coverage of both LLINs and IRS rounds from 2003, parasite prevalence in all age groups is now well below the 1% threshold. The challenge in such settings is to sustain interventions at a sufficient level to maintain effective control in the face of reintroduction from neighbouring areas via human migration and travel. In some moderate-transmission settings it is also possible to reduce parasite prevalence below the 1% threshold with existing tools. In our example settings, this could be achieved in Kinkole, DRC where the endemic EIR was 48 ibppy if an intensive program of twice-yearly IRS and MSAT were combined with increasing LLIN coverage to 80% levels. In the slightly higher transmission setting of Nkoteng, Cameroon (EIR = 96 ibppy), current tools could reduce transmission below the 1% threshold but in this case (perhaps unrealistically) high levels of adherence to LLIN use would also be needed. Thus the first phase of elimination programs is achievable in many areas in which the LLIN and IRS in combination are effective (that is, in areas with primarily endophilic vectors). Additional use of MSAT, to date not considered by many programs, has the potential to speed further declines in prevalence. We considered one area, Maputo, Mozambique, in which the high proportion of An. arabiensis (exhibiting a high degree of exophilic behaviour), made elimination more difficult. Whilst the scale of the declines that our model predicts are similar to those observed in an IRS-based campaign in that area (that commenced in 2000 using Bendiocarb rather than DDT [4]), this study also demonstrated a greater impact on the population of An. arabiensis compared to that resulting from our model. This may be because our estimate of the degree of exophilic behaviour is too high (see Protocol S3) or because mosquito behaviour changes both with season and with setting [35],[68]–[71] and requires further exploration. In high transmission intensity settings, current tools can be used to substantially reduce transmission and the associated disease burden, but are insufficient to drive prevalence below the pre-elimination threshold. This finding is not surprising given the high basic reproduction numbers previously estimated in large parts of sub-Saharan Africa [43]. Such outcomes have been observed in the Bioko Island control program where, with intensive ongoing interventions, parasite prevalence in 2- to 5-year-olds fell from 42% to 18% between 2004 and 2008 [14]. Similarly, in the 1970s Garki project in Nigeria, an area of moderate to high transmission (annual EIRs in the range 20–130 ibppy), substantial declines in prevalence were recorded but elimination was not achieved [7]. In these settings, additional new tools are likely to be required if pre-elimination targets are to be achieved. Whilst a detailed comparison of the range of potential tools under development is beyond the scope of this paper, there are two broad areas of innovation that merit further consideration. The first aims to target the mosquitoes that are not reached by current interventions, particularly those on whom indoor-targeted interventions are least successful. Notably, this includes major species such as An. arabiensis, which preferentially rest outdoors after feeding and may also obtain blood meals from non-human animals. These mosquitoes could be targeted in a number of ways, including additional interventions that are applied on non-human hosts [72], killing adult females feeding or resting outdoors [73]–[75], or at source in the larval habitat [76],[77]. Secondly, our results on the levels of human adherence required in high-transmission settings suggest that interventions that do not strongly depend on human participation are likely to be needed. The methods outlined above are examples of such approaches. Our results confirm findings by others that the bionomics of the local vector species, including the degree of exophagy, exophily, and zoophagy [35],[36],[78], can potentially be a strong determinant of intervention success. Current tools, in particular LLINs and IRS, are focused towards species with strong endophagic, endophilic, and anthropophagic tendencies. Further data on the degree of endophilic behaviour of the different Anopheles species, coupled with information on how these parameters may change in response to interventions (we assumed here that they remain fixed),are critically needed to understand the longer-term impact of IRS and LLINs on transmission. Historically, there is some evidence of species replacement following the introduction of IRS in three different geographical locations [79]. More recently, a shift in species relative abundance (though not replacement or increased density) has been observed in Western Kenya following high coverage of LLINs [80],[81]. In addition, mapping of vector species distribution and proportional composition [17] is critical to the ability to predict program success outside of the well-studied research areas. Behavioural aspects of intervention programs are characterized in multiple ways. For example, the WHO report bed-net coverage as the number of nets distributed per person at risk [5], whilst Malaria Indictor Surveys collect data on the proportion of households owning a net or sleeping under a net [82]. Our results demonstrate that patterns of coverage and effective coverage are an important determinant of intervention success and may be one reason why simple models of LLIN impact have tended to appear highly optimistic [35]–[38]. Furthermore, it is unrealistic to assume perfect and uniform adherence. Indeed, rates of sleeping under LLINs tends to be highest in young children, but lower in school-aged children [83], who are important contributors to the infectious reservoir (Figure 1). Furthermore, whilst we did not explicitly consider reduction in adherence or take-up of IRS, this is likely to occur after repeated rounds as perceived risk declines, and will reduce the impact of the intervention. Receipt of interventions is also an important consideration in assessing impact, particularly if coverage levels are low. It is well recognized that malaria transmission is highly focal with some individuals at much higher risk than others [42],[43],[64],[84],[85]. Our results confirm other models' findings [43],[64] that, by targeting interventions at areas of intense transmission, substantially greater reductions in transmission are possible than by distributing them randomly or by focusing distribution towards younger children. However, little attention has previously been paid to the heterogeneous distribution of interventions within such target populations. In general, the impact of an intervention will be lower if the same individuals in the target population continually receive and adhere to the intervention than if distribution fully covers the target population. Thus data on repeat uptake of interventions would be useful to determine true target population coverage levels. Furthermore, health systems will need to be strengthened and laboratory capacity put in place to allow rapid identification of these foci. In addition, overall coverage levels can potentially be enhanced through consideration of a wide range of different delivery mechanisms appropriate to the local setting [86]–[88]. One aspect with the potential to hinder elimination campaigns not considered here is the development of resistance—either to drugs, to the insecticides used to treat nets or for indoor residual spraying, or to vaccines—and the potential for alterations in the behaviour of the vector in response to the interventions. Resistance to DDT was a particular problem during the GMEP and is credited with being a major reason for the abandonment of the program. DDT resistance at varying levels has now been reported in over 50 anopheline species [89]; thus, to reduce the further emergence of resistance, elimination campaigns should aim to reduce transmission as rapidly as possible. The recent emergence of partial drug resistance to artemisinin in Cambodia [90] has further highlighted the need to guard against and reduce the emergence and spread of resistance, particularly as access to treatment is scaled up. Our model is necessarily a simplification of the more complex dynamics underlying malaria transmission and control, so numerical results should be interpreted more as providing intuitive insight into potential scenarios than as firm predictions of what might happen in a given setting. Furthermore, whilst we give an indication of impact over a 25-year time horizon (including graphs that track expected trends over this period), given the uncertainty in some of the key parameters, it is not possible to give short-term indications of impact or timelines. Precise, accurate prediction remains challenging for a number of reasons. First, the mean duration of asymptomatic infection, and the dynamics of acquisition and loss of immunity, are key parameters determining the speed of decline in parasite prevalence once transmission is reduced [47]. These are both poorly understood in semi-immune populations. These parameters also determine the time scale for which interventions would need to remain in place to ensure that a rebound in infection and disease does not occur. Current best estimates of model parameters suggest that this is likely to be decades rather than months or years, but further data are needed to refine these estimates. Second, there are multiple model structures that can reproduce important characteristics of malaria epidemiology such as the age patterns of infection prevalence across different transmission settings. Whilst we have invested substantial effort in developing a modern statistical framework to better choose between model structures and to estimate associated model parameters, there are limited data to distinguish some aspects of the model. In the current exercise, we have focused on fitting the human model cycle to a wide range of datasets. This will be extended in future applications to fit the full cycle using explicitly seasonal models to more detailed data from specific research sites. In addition, the individual intervention models have not to date been validated by fitting to specific trial data. This process is underway. Such fitting will enable the addition of uncertainty bounds to model output through sampling of parameter posterior densities [91]. If feasible, this could be extended to incorporate model uncertainty using a Bayesian methodology [92]. Third, in our current model we use a relatively simple vector cycle in which the vector population is driven by a constant birth rate. This may underestimate the additional impact of interventions that increase vector mortality and thus reduce population-level fecundity. Vector models which incorporate capacity constraints and behavioural change are a natural extension that may better represent competition for larval habitats [93]. However, to date, such models have not been adequately validated against weather measurements and entomological data and thus further work is required to obtain a model that can reproduce entomological patterns from multiple transmission settings. Last, our current model has been developed and parameterized to be applied to single locations. It thus considers isolated areas and does not address the focal and heterogeneous nature of transmission on a wider spatial scale or the connectedness of local populations. As such, the current model cannot be used to assess the risk of reintroduction of the parasite from outside areas, which has been shown to be a major challenge in ongoing control [94]. However, it is possible to extend this framework to a fully spatial continental-scale simulator. The major challenge here is not in developing the software tool but in parameterising the model across settings. Basic requirements of such a model, e.g. human population size in each area, are not well known across parts of Africa, although synthetic data derived from satellite observations can be used as a proxy [95],[96]. In addition, such models require local-level information on vector species, seasonality patterns, intensity of transmission, and human movements to enable assessment of the risks of transmission spatially. Despite these limitations, mathematical models based on the biology of the transmission cycle provide an appropriate tool for a range of stakeholders to explore the potential impact of current and future interventions on malaria transmission and disease burden in a systematic manner. Further development of the models and approaches outlined here can help to identify optimal policies for the range of stages of malaria elimination programs from the consolidation phase outlined here, through the pre-elimination and elimination phases, to sustained elimination. By considering current tools and exploring potential future interventions, models can help us to understand the limits of current strategies and evaluate the potential for future products to achieve the ultimate goal of global eradication. Supporting Information Alternative Language Abstract S1 Abstract translated into French by Emilie Pothin. (0.03 MB DOC) Click here for additional data file. Alternative Language Abstract S2 Abstract translated into Spanish by MGB. (0.03 MB DOC) Click here for additional data file. Alternative Language Abstract S3 Abstract translated into Dutch by TB. (0.03 MB DOC) Click here for additional data file. Alternative Language Abstract S4 Abstract translated into Portuguese by Dr. Jose Sousa-Figueiredo. (0.03 MB DOC) Click here for additional data file. Protocol S1 Transmission model. (0.37 MB DOC) Click here for additional data file. Protocol S2 Intervention models. (0.88 MB DOC) Click here for additional data file. Protocol S3 Bayesian model fitting and parameter values. (0.92 MB DOC) Click here for additional data file. Protocol S4 Seasonal patterns and transmission settings. (0.49 MB DOC) Click here for additional data file. Protocol S5 Additional results and sensitivity analyses. (1.26 MB DOC) Click here for additional data file. Protocol S6 User-friendly software for model runs. (0.17 MB DOC) Click here for additional data file.

0 comments Cited 281 times – based on 0 reviews      Review now

Bookmark

Record: found
Abstract: found
Article: not found

A central role for P48/45 in malaria parasite male gamete fertility.

C Janse, J Braks, Ryan A. Waters … (2001)

Fertilization and zygote development are obligate features of the malaria parasite life cycle and occur during parasite transmission to mosquitoes. The surface protein PFS48/45 is expressed by male and female gametes of Plasmodium falciparum and PFS48/45 antibodies prevent zygote development and transmission. Here, gene disruption was used to show that Pfs48/45 and the ortholog Pbs48/45 from a rodent malaria parasite P. berghei play a conserved and important role in fertilization. p48/45- parasites had a reduced capacity to produce oocysts in mosquitoes due to greatly reduced zygote formation. Unexpectedly, only male gamete fertility of p48/45- parasites was affected, failing to penetrate otherwise fertile female gametes. P48/45 is shown to be a surface protein of malaria parasites with a demonstrable role in fertilization.

0 comments Cited 131 times – based on 0 reviews      Review now

Bookmark

All references

Author and article information

Contributors

Maarten Eldering: maarten.eldering@gmail.com

Anaïs Bompard: anais.bompard@gmail.com

Kazutoyo Miura: KMIURA@niaid.nih.gov

Will Stone: William.Stone@lshtm.ac.uk

Isabelle Morlais: isabelle.morlais@ird.fr

Anna Cohuet: anna.cohuet@ird.fr

Geert-Jan van Gemert: Geert-Jan.vanGemert@radboudumc.nl

Patrick M. Brock: Paddy.Brock@glasgow.ac.uk

Sanna R. Rijpma: Sanna.Rijpma@radboudumc.nl

Marga van de Vegte-Bolmer: Marga.vandeVegte-Bolmer@radboudumc.nl

Wouter Graumans: Wouter.Graumans@radboudumc.nl

Rianne Siebelink-Stoter: Rianne.Siebelink-Stoter@radboudumc.nl

Dari F. Da: dafrenick@yahoo.fr

Carole A. Long: CLong@niaid.nih.gov

Merribeth J. Morin: mmorin@path.org

Robert W. Sauerwein: Robert.Sauerwein@radboudumc.nl

Thomas S. Churcher: thomas.churcher@imperial.ac.uk

Teun Bousema: teun.bousema@radboudumc.nl

Journal

Journal ID (nlm-ta): Parasit Vectors

Journal ID (iso-abbrev): Parasit Vectors

Title: Parasites & Vectors

Publisher: BioMed Central (London )

ISSN (Electronic): 1756-3305

Publication date (Electronic): 17 October 2017

Publication date PMC-release: 17 October 2017

Publication date Collection: 2017

Volume: 10

Electronic Location Identifier: 489

Affiliations

[1 ]ISNI 0000 0004 0444 9382, GRID grid.10417.33, Department of Medical Microbiology, , Radboud University Medical Center, ; Nijmegen, The Netherlands

[2 ]ISNI 0000 0001 2113 8111, GRID grid.7445.2, MRC Centre for Outbreak Analysis and Modelling, Department of Infectious Disease Epidemiology, Imperial College London, ; London, UK

[3 ]ISNI 0000 0001 2164 9667, GRID grid.419681.3, National Institute of Allergy and Infectious Diseases, Laboratory of Malaria and Vector Research, National Institutes of Health, ; Rockville, MD USA

[4 ]ISNI 0000000122879528, GRID grid.4399.7, Institut de Recherche pour le Développement, UMR MIVEGEC UM-CNRS 5290-IRD 224, ; Montpellier, France

[5 ]ISNI 0000 0004 0564 0509, GRID grid.457337.1, Institut de Recherche en Sciences de la Santé, ; Bobo Dioulasso, Burkina Faso

[6 ]ISNI 0000 0001 2193 314X, GRID grid.8756.c, Institute of Biodiversity, Animal Health and Comparative Medicine, College of Medical, Veterinary and Life Sciences, , University of Glasgow, ; Glasgow, UK

[7 ]PATH Malaria Vaccine Initiative, Washington DC, USA

[8 ]ISNI 0000 0004 0425 469X, GRID grid.8991.9, Department of Immunology and Infection, London School of Hygiene and Tropical Medicine, ; London, UK

Article

Publisher ID: 2414

DOI: 10.1186/s13071-017-2414-z

PMC ID: 5646129

PubMed ID: 29041962

SO-VID: 59db030c-21e4-410c-a365-03f54a055659

License:

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License ( http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

History

Date received : 6 March 2017

Date accepted : 1 October 2017

Funding

Funded by: FundRef http://dx.doi.org/10.13039/501100003246, Nederlandse Organisatie voor Wetenschappelijk Onderzoek;

Award ID: 016.158.306

Award Recipient : Teun Bousema

Funded by: Seventh Framework Programme (BE)

Award ID: 242095

Award Recipient : Robert W. Sauerwein

Funded by: Malaria Vaccine Initiative

Custom metadata

ScienceOpen disciplines: Parasitology

Keywords: malaria,anopheles,transmission,vaccine,immunity

Data availability:

ScienceOpen disciplines: Parasitology

Keywords: malaria, anopheles, transmission, vaccine, immunity

Comments

Comment on this article

scite_

Cited by 9

See all cited by

Most referenced authors 896

See all reference authors

Comparative assessment of An. gambiae and An. stephensi mosquitoes to determine transmission-reducing activity of antibodies against P. falciparum sexual stage antigens

Read this article at

Abstract

Background

Methods

Results

Conclusion

Electronic supplementary material

Related collections

Malaria vaccine development

Most cited references 43

Implication of the Mosquito Midgut Microbiota in the Defense against Malaria Parasites

Reducing Plasmodium falciparum Malaria Transmission in Africa: A Model-Based Evaluation of Intervention Strategies

A central role for P48/45 in malaria parasite male gamete fertility.

Author and article information

Contributors

Journal

Affiliations

Article

History

Funding

Categories

Custom metadata

Comments

Comment on this article

Similar content 510

Cited by 9

Most referenced authors 896