Uneven Plasmodium falciparum Transmission and Cryptic ovale Transmission From the Asymptomatic Reservoir in Bagamoyo, Tanzania

Abstract

Background.

Asymptomatic malaria carriers often harbor low parasite densities missed by rapid diagnostic tests (RDTs), yet they contribute to transmission. Direct skin feeding (DSF) assays can sensitively measure their infectiousness to mosquitoes.

Methods.

To characterize human-to-mosquito transmission from the asymptomatic reservoir in Bagamoyo, Tanzania, DSFs were performed in persons >5 years of age positive for Plasmodium falciparum by RDT or real-time PCR. Fifty colony-reared Anopheles gambiae were fed on the posterior calves. Successful mosquito infection was defined as ≥1 oocyst-positive mosquito midgut among 25 dissected eight days postskin feeding.

Results.

Among 491 participants with median parasite density of 5.1 parasites/μL who underwent DSF, 22% were infectious to mosquitoes. RDT-positive participants infected roughly twice as many mosquitoes compared to RDT-negative/PCR-positive persons. However, up to 21% of infectious carriers were PCR-negative at the time of skin feeding, after screening PCR-positive a few days earlier. Overall, 9.1% (342/3741) of mosquitoes fed on infectious carriers were parasite-positive at dissection. Half of infectious individuals infected a single mosquito, while the top 16 transmitters (3% of those undergoing DSF) cumulatively infected 57% of infected mosquitoes. RDT-positive school-age children (6–15 years), 27% of the DSF cohort, contributed to 58% of infected mosquitoes. Unexpectedly, mosquito midguts from 39 DSFs (44% of oocyst-positive feeds analyzed) tested positive for Plasmodium ovale.

Conclusions.

Parasites circulating at the limit of PCR detection commonly infect mosquitoes. However, a small proportion of highly infectious carriers contribute disproportionately to transmission, offering potential for targeted interventions. P. ovale was frequently co-transmitted with P. falciparum to mosquitoes.

The last two decades have seen both a celebrated reduction in malaria morbidity and mortality globally, but also a sobering resurgence as areas of high burden persist and malaria cases rebound in areas once considering malaria elimination. In sub-Saharan Africa, substantial asymptomatic Plasmodium falciparum carriage persists even in areas where malaria transmission has declined and symptomatic cases have decreased.

The central role of asymptomatic malaria carriers as a source for ongoing transmission in sub-Saharan Africa has been established in field studies spanning different transmission settings [1-3]. Asymptomatic infections, especially those due to chronic carriage, are more likely to harbor transmissible gametocytes that remain undetected [4, 5]. School-age children, in particular, may not have acquired the antiparasite immunity that prevents efficient transmission to biting mosquitoes [1, 4, 6] and may be more exposed or attractive to biting mosquitoes [7]. Many asymptomatic infections are low density, often detectable only by PCR—what proportion of these are infectious, and how best to target them with surveillance and control interventions remains an ongoing question and challenge.

Most studies investigating human-to-mosquito transmission of malaria over the last decade have used mosquito direct membrane feeding assays (DMFAs) to measure individuals’ infectiousness to mosquitoes, whereby mosquitoes feed on fresh blood through an artificial feeding system. DMFAs have contributed to our understanding of the human infectious reservoir for malaria and the investigation of transmission-blocking vaccine candidates. However, likely due to their limited ability to recapitulate the natural physiology of skin feeding, which involves sampling sexual-stage gametocytes in dermal capillaries, they may fail to detect the infectious potential of parasite carriers with very low parasite and gametocyte densities [8, 9].

We sought to determine the infectiousness of low-density infections using direct skin feeding (DSFs) assays in coastal Tanzania, where malaria transmission has declined but a substantial asymptomatic reservoir remains. We investigated the relative transmissibility of asymptomatic carriers readily identified by point-of-care rapid diagnostic tests (RDTs) versus molecular diagnostics, and investigated whether there is a threshold parasite/gametocyte density below which human-to-mosquito transmission of malaria does not occur.

METHODS

Study Design and Participants

Rural Bagamoyo District lies on the eastern coast of Tanzania, approximately 40 km north of Dar es Salaam. This region historically experienced high malaria transmission but has undergone an epidemiologic transition to lower transmission [10]. Residents are mostly farmers and fishermen. Transmission occurs throughout the year, with peaks during the long (March to May) and short (October to December) rainy seasons. Project TranSMIT (Transmission of Submicroscopic Malaria in Tanzania) prospectively screened and enrolled asymptomatic persons ≥6 years of age from four primary schools (Mwavi, Mtakuja, Mkenge, and Fukayosi) and two health centers (Yombo and Fukayosi clinics) (Supplementary Figure 1). Parental/guardian consent and child assent to participate were obtained for those <18 years of age. Screening was conducted by malaria RDT (SD Bioline with Pf/Pan bands), microscopy, and real-time PCR as previously described [11]. Participants positive by any test were eligible for enrollment. The study was approved by institutional review boards at the University of North Carolina, Tanzania National Institute for Medical Research, Muhimbili University of Health and Allied Sciences, and Ifakara Health Institute.

Enrollment and Mosquito Feeding Procedures

At enrollment, finger pricks and venipuncture were performed to make capillary and venous dried blood spots (DBS, 50 μL each) on Whatman 3MM filter paper and thick and thin blood smears. Single DBS underwent Chelex extraction and real-time 18S rRNA PCR (qPCR) for Plasmodium falciparum detection onsite [12, 13]. Plasmodium ovale and Plasmodium malariae 18S rRNA qPCR were performed later [13-15]. Capillary and venous blood, stored in DNA/RNA shield (Zymo Research, Irvine, California), were used to measure gametocytemia via RT-qPCR [16-18].

DSFs were performed as previously described [13]. Fifty female An. gambiae s.s. (IFAKARA strain) 4–7 day old mosquitoes were starved 4–6 hours prior to feeding, then placed in 2 cups of 25 mosquitoes each to feed on the right and left posterior calves of each participant for 15 minutes. DSFs were immediately stopped if participants reported any discomfort. Antihistamine cream was applied postfeeding and provided for home use. Blood-fed mosquitoes were kept in climate-controlled incubators with 10% sucrose until day 8 post-DSF, when midguts were dissected, stained with 0.1% Mercurochrome, and scored for oocysts. Positive midguts were individually stored in either 95% ethanol or DNA/RNA Shield for Plasmodium screening by qPCR.

In a subset of participants, mosquitoes were maintained until day 14 post-DSF instead of dissected on day 8. Mosquito abdomens were removed, and heads/thoraxes were pooled in groups of five for qPCR detection of sporozoites. Additionally, parallel DMFAs were completed in a subset of participants, either with autologous plasma or with replacement with malaria-naive sera (US volunteers). DMFAs were conducted within 15 minutes of venipuncture, using 2 mL of blood to feed up to 200 An. gambiae using a water-jacketed glass feeder [9].

After blood collection and mosquito feeding, RDT-positive participants were treated with artemether-lumefantrine. Those who tested PCR-positive only were offered treatment if they declined to participate in a longitudinal phase of the study (results not reported here).

Molecular Analysis

Gametocytemia was measured by RT-qPCR via two female markers, Pfs25 [16] and PF367_030000 [17], and one male marker, PfMGET (PF3D7_1469900) [18]. Plasmid constructs were used for quantitation of study samples, for which a conversion factor of 87.5 Pfs25 transcripts/gametocyte [16], 9.4 PF367_030000 transcripts/gametocyte (derived in similar fashion as [16]) and 12.5 PfMGET transcripts/gametocyte [19] were used.

For Plasmodium species detection from oocyst-positive midguts, DNA was extracted using DNAzol Reagent (Invitrogen, Massachusetts) [13] and analyzed by nested PCR, first undergoing Plasmodium genus PCR [20] followed by species-specific 18S qPCR using both undiluted first-round PCR product as well as a 1:50 dilution. Any qPCR run with cycle threshold (Ct) <40 cycles was considered positive. Plasmodium species detection from pooled mosquito heads/thoraxes was performed on DNA extracted using the Quick-DNA Tissue/Insect Kit (Zymo Research) via 18S species-specific qPCR [13].

Statistical Analysis

Clinical, entomologic, and molecular data were entered into a REDCap database. We determined the proportion of participants with asymptomatic P. falciparum infection, stratified by RDT versus PCR-positive only status that was infectious to mosquitoes by DSF. Infectiousness to mosquitoes was defined as detection of at least one microscopically detected oocyst-positive midgut out of 25 dissected mosquitoes. To limit the inclusion of artefactual results [21], we reviewed oocyst morphology using images of mercurochrome-stained tissues scored as positive midguts that were available for roughly half the cohort. This led to the exclusion of all midguts that were scored as containing >100 oocysts, and those scored with >50 oocysts that were the only positive midgut in the DSF, due to instances found of a single dissector failing to remove mosquito ovaries and scoring developing eggs as collections of oocysts.

We compared gametocyte densities in capillary versus venous blood using Pearson’s correlation coefficient and evaluated the proportion of male gametocytes in relation to total gametocytemia within individuals. To compare factors among participants that infected ≥5 mosquitoes by DSF versus those who infected fewer mosquitoes, or were not infectious at all, we estimated prevalence ratios and 95% confidence intervals using the Wald method. Finally, we assessed the relative contribution of participants to the number of mosquitoes infected from the total DSF cohort based on their test positivity status (smear or RDT, PCR). Sensitivity analyses excluded DSFs without PCR-confirmed mosquito infection. Comparisons between groups used Wilcoxon rank-sum (two groups) or Kruskal–Wallis (≥3 groups) tests. Hypothesis tests were two-sided at a significance level of 0.05. Analyses were performed using R 4.2.1.

RESULTS

Study Population in Coastal Tanzania

Among 6511 persons ≥6 years of age screened for P. falciparum infection from October 2018 to December 2021 [11, 12], 14% (888) were RDT-positive and 30% (1950) were PCR-positive (Figure 1). Out of 2092 who tested positive by either test, 606 were enrolled for mosquito DSF. RDT-positive participants (n = 260) underwent DSF on the same day as screening. Participants who were RDT-negative, but found to be PCR-positive (RDT−/PCR+) (n = 346), were called to return for enrollment and DSF. This occurred a median of 2 days (IQR 2–5) later, with nearly half (153/346, 44%) of those enrolled based on PCR-positivity at screening no longer PCR-positive on the day of DSF. Those PCR-negative at enrollment had lower parasite densities at screening (Supplementary Figure 2). The median parasite density of the RDT-positive and RDT−/PCR+ participants at enrollment was 118 (IQR 4.4–790) and 6.2 (IQR 1.7–31) p/μL, respectively. Enrolled participants ranged in age from 6 to 70 (median 16 years). The majority were female (382, 63%), as women tended to accompany their children for screening or be recruited when bringing their children to clinic. Age, sex, and PCR-positivity were similar among those screened positive for P. falciparum (n = 2092), enrolled (n = 606), and included in the mosquito infectivity analysis (n = 491) (Supplementary Table 1).

Figure 1.

Study schematic and mosquito DSF results. Among 6511 asymptomatic persons ≥6 years of age, 2092 screened positive for P. falciparum by rapid diagnostic test (RDT) or PCR. Of 606 who were subsequently enrolled for mosquito DSF assays, 43% (260) were RDT-positive, while 32% (193) were PCR-positive only at the time of enrollment. The remaining 25% (153) of participants were PCR-negative at the time of enrollment, but had been PCR-positive at the time of screening a median of 2 days (IQR 2–5) prior. Mosquito DSF assays were successfully completed in 590 participants. Mosquito infectivity was evaluated for 491 participants in which at least 25 blood-fed Anopheles gambiae survived to dissection. 5.9% of these 491 participants who underwent skin feeding were gametocytemic by microscopy, while half had gametocytes detectable by RT-qPCR. Approximately one-fifth (22%, n = 106) were infectious to mosquitoes, based on microscopic detection of at least one midgut with oocyst/s at midgut dissection eight days after mosquito feeding. At least one oocyst-positive midgut was confirmed with P. falciparum-specific qPCR in 30% (32) of positive skin feeds.

Male-biased Gametocytemia at Low Gametocyte Densities

Microscopic gametocytemia was present in 5.6% (34/604) of enrolled participants, while half (49%, 296/601) harbored submicroscopic gametocytes detectable by qRT-PCR targeting the male marker, PfMGET [22]. In the first two seasons of the study, when both male and female gametocyte markers were investigated (PfMGET and PF367_030000), male and female gametocytes were detected in 74% (101/137) and 43% (59/137), respectively of RDT-positive participants, and in 54% (38/70) and 26% (18/70) of RDT−/PCR+ participants. Gametocyte densities measured using the intron-spanning female marker, PF367_030000, were concordant with Pfs25-based densities (Supplementary Figure 3). Among 76 participants where both male and female gametocytes were detected, gametocyte sex ratio was female biased overall, but less so at submicroscopic densities, with an increased proportion of male gametocytes at gametocyte densities <30/μL (Figure 2A and 2B). Parasite and gametocyte densities did not differ in concurrently collected capillary and venous blood samples (Figure 2C and 2D).

Figure 2.

Gametocyte sex ratio and density in different blood compartments. Comparison of male and female gametocytemia among 76 participants in which both male and female gametocytes were detected (A) and the proportion of male gametocytes versus total gametocytes (sum of male and female gametocytes) (B). DNA and RNA extracted from equal volumes of dried blood spots (50 μL blood) made with concurrently collected venous (venipuncture) or capillary (finger prick) blood samples were subjected to qPCR targeting 18S rRNA to measure parasite densities (C) and qRT-PCR targeting Pfs25 and PF3D7_0630000 to measure gametocyte densities (D).

Uneven Transmission From Asymptomatic Parasite Carriers to Mosquitoes

DSFs were well-tolerated as only 7/606 (1%) were halted due to skin irritation. In total, 582 participants successfully underwent skin feeding with 50 colony-reared An. gambiae ss, with 491 (84%) yielding at least 25 blood-fed mosquitoes surviving to dissection (Figure 1). These 491 participants had a median parasite density of 5.1 parasites/μL. Twenty-two percent (106) were infectious to mosquitoes, with at least one oocyst-positive midgut identified at dissection. This rate of parasite transmission via DSFs (22%) was similar in complementary investigations, namely (1) sporozoite-positivity (29%) in 17 contemporaneous DSFs and (2) parallel DMFAs for 58 participants (Supplementary Results 3.1).

Confirmation of mosquito infection by PCR detection of P. falciparum DNA in oocyst-positive midguts proved difficult. Among the 106 DSFs containing 342 oocyst-positive midguts, some midguts were lost at dissection (n = 114 midguts from 38 DSFs) while many failed to amplify (n = 115 midguts from 76 DSFs), attributed to a small amount of DNA in midguts often containing a single oocyst, and/or the failure to isolate the midgut for extraction, as it was not visible once lysed in DNA/RNA Shield. Overall, P. falciparum DNA was successfully detected in 32 DSFs—40% (32/89) of oocyst-positive feeds where midguts were available for analysis, and 58% (30/52) of oocyst-positive feeds with multiple oocyst-positive midguts. Given these difficulties, we use oocyst-positive midguts to measure mosquito infection rate in subsequent analyses, but provide sensitivity analyses using only PCR-confirmed DSFs.

The rate of infectivity was greater among RDT-positive participants: 25% (54/218) of RDT-positive versus 19% (52/273) of RDT−/PCR+ persons were infectious to mosquitoes (P = .16). When restricting to PCR-confirmed mosquito infection, 11% (21/185) versus 5% (11/232), respectively, were infectious (P = .02). RDT-positive participants also infected roughly twice as many mosquitoes (median 5.8% vs 3.3% of mosquitoes (P < .01)) and accounted for 74% (252/342) of infected mosquitoes in the cohort, despite making up less than half (44%, 218/491) of those undergoing DSF (Supplementary Figure 5). These relative proportions were similar when analyzing only PCR-confirmed mosquito infections (Supplementary Results 3.3). Overall, infectious individuals infected 9.1% (342/3741) of mosquitoes, while 1.9% (342/18 071) of all dissected mosquitoes were parasite-positive.

Half of all positive DSFs (53/106, 50%) resulted in just one infected mosquito, often harboring a single oocyst (Figure 3A and 3C). Feeds resulting in two infected mosquitoes represented the next quarter of positive DSFs (25/106, 24%). Meanwhile, the most infectious parasite carriers contributed disproportionately to the number of infected mosquitoes. The top 15% (16/106) infectious persons infected 5–35 mosquitoes each, and cumulatively infected 196 mosquitoes, or 57% of all infected mosquitoes. Oocyst burden was high in these highly infectious feeds (Figure 3B and 3D), with 15 oocysts on average across infected midguts. If considering only PCR-confirmed infectious feeds, 13 highly infectious persons (41% of all infectious persons) contributed to 75% (144/192) of infected mosquitoes.

Figure 3.

Parasite oocysts at midgut dissection. Out of 106 positive skin feeds, half (53, 50%) led to infection of a single mosquito midgut, often containing a single oocyst (A, C). In 16 skin feeds (15% of positive feeds, 3.3% of all evaluable DSFs), 5 or more mosquito midguts were infected (C), containing up to as many as 97 oocysts (B). Mean oocyst count per DSF in those with at least one mosquito infection are presented (D). The dotted line represents the hypothetical situation where all infected mosquitoes in a DSF harbored exactly one oocyst. DSFs in which at least one oocyst-positive midgut was confirmed by PCR are depicted in magenta. Additional selected midgut photos are presented in Supplementary Figure 6.

The top 16 highly infectious individuals were more likely to be RDT-positive, smear-positive, and gametocytemic compared to the 475 individuals who were noninfectious (n = 385) or less infectious (n = 90) (Figure 4). School-age children were also slightly more prevalent in the highly infectious versus noninfectious group (prevalence ratio 1.5 (95% CI 1.0, 2.3)).

Figure 4.

Factors associated with high infectiousness to mosquitoes. Sixteen participants exhibited high infectivity to mosquitoes (infected ≥5 mosquitoes out of ≥25 dissected, median infection rate of 22%, IQR 18–45%). The characteristics of these highly infectious “super-transmitters” were compared to all other transmitters (“less infectious,” n = 90) as well as those who failed to infect mosquitoes at all (“noninfectious,” n = 385). A crude prevalence ratio >1.0 indicates that higher parasite and gametocyte burden is more prevalent in the highly infectious cohort. School-age children (6 to 15 years) were compared to those 16 years and older. High village-level prevalence was compared to low village-level prevalence, previously defined in reference [10]. The wet season was defined as March–May and October–December, with a 6 weeks lag from DSF date, with all other dates falling in the dry season. The table on the right depicts prevalence of each factor by infectivity category.

Evidence of Cryptic Plasmodium ovale Transmission

Due to limited PCR amplification of P. falciparum from infected midguts, we explored the possibility of P. ovale transmission in the cohort. Among DSFs with P. falciparum-positive midguts identified by PCR, 24/32 (75%) were also found to have P. ovale-positive midguts, with either the same midgut testing positive for both P. falciparum and P. ovale (n = 20), or another oocyst-positive midgut from the same DSF testing positive for P. ovale (n = 4) by PCR (Figure 5). P. ovale was additionally detected in 15/57 (26%) of P. falciparum PCR-negative but oocyst-positive DSFs. In these 39 DSFs with P. ovale-positive midguts, P. ovale was detected in the blood of the skin-fed participant on the day of mosquito feeding in only 4/37 (11%) participants. Overall, among 89 oocyst-positive DSFs with midguts available for analysis, 39 (44%) showed evidence of cryptic P. ovale transmission (Supplementary Results 4.1).

Figure 5.

Oocyst burden and mosquito infection rate. The dotted line represents the hypothetical situation if all infected mosquitoes in a DSF harbored exactly one oocyst. Oocyst-positive midguts from 89 positive skin feeds were successfully saved at dissection and underwent parasite species detection via a nested PCR approach. Midguts from 42 feeds failed to amplify Plasmodium DNA, usually in the context of a low mosquito infection rate (single oocyst-positive midgut). Dots are color coded to indicate PCRs with at least one midgut amplifying DNA of either P. falciparum (red), P. ovale (blue), or both species (orange).

Plasmodium falciparum Transmission From Parasite Densities at the Limit of PCR Detection

Figure 6 depicts the relationships between P. falciparum parasite/gametocyte density and mosquito infection rate in the cohort. Twenty-one percent (32/152) of individuals without parasitemia detectable by qPCR at the time of DSF were nonetheless infectious to mosquitoes (Figure 6B). When we confine the analysis to PCR-confirmed mosquito infection, we still find that 5% (6/126) of PCR-negative individuals infected mosquitoes (Supplementary Figure 7). Because the PCR-negative individuals were PCR-positive at screening, a median of 2 days earlier, we presume they harbored parasite densities at the limit of PCR detection. Our onsite qPCR has a limit of detection of approximately 1 parasite/μL from DBS (Supplementary Figure 8), with the equivalent of 3 μL blood assessed across duplicate PCR reactions for each participant [12]. When we applied an ultra-sensitive qPCR to leftover DBS, targeting the multicopy varATS gene that detects down to 0.1 parasite/μL [23], P. falciparum was detected in 19% of (6/32) of PCR-negative infectious individuals. Similarly, 20% (47/241) of participants without detectable gametocytes by RT-qPCR at enrollment were nonetheless infectious to mosquitoes (Figure 6C), with this figure decreasing to 4% (9/203) when confining to PCR-confirmed midguts (Supplementary Figure 7). Concomitant Pfsbp1 expression data in five of six (83%) gametocyte-negative participants who were 18S PCR-positive makes it unlikely that RNA extraction or amplification failed in all of these participants.

Figure 6.

The relationship of parasite and gametocyte density to mosquito infection. In each plot, parasite density and gametocyte densities were measured by 18S rRNA qPCR and Pfmget RT-qPCR, respectively (A). PCR-negative participants are depicted as having 0.1 parasites and 0.01 gametocytes/μL, respectively, below the limit of detection of each assay. In plots B and C, the size of the circles correlates with the number of mosquitoes dissected, while data derived from DSFs performed in RDT-positive participants are represented by filled circles. Parasite-negative and gametocyte-negative participants are indicated within rectangular boxes.

Infectivity of School-Age Children

Ten individuals in the highly infectious group (n = 16) were school-age children (6–15 years) (Figure 4), including three participants aged 11–15 who each infected >50% of mosquitoes (Figure 7A). In total, school-age children were 44% of those evaluated for mosquito infectivity, 56% of those found to be infectious to mosquitoes, and responsible for infecting 67% of the mosquitoes in the cohort (Figure 7B). RDT-positive school-age children (6–15 years) represented 27% of those assessed for transmissibility, but contributed to 58% of all infected mosquitoes. Results were similar when considering only PCR-confirmed mosquito infections (Supplementary Figure 9).

Figure 7.

Age-related contribution to mosquito infections. Mosquito feeding data showing the relationship of gametocyte density to mosquito infection prevalence is stratified by age of the skin-fed participant (A). The distribution of age among those evaluated for mosquito infectivity, those found to be infectious to mosquitoes, and number of mosquitoes infected based on age of the skin-fed participants is shown (B).

DISCUSSION

We performed over 500 DSF assays on low-density infections in a region in sub-Saharan Africa where malaria transmission has declined but a sizable asymptomatic reservoir remains. We found that P. falciparum transmission to mosquitoes occurs even when parasites are circulating at the limit of detection of real-time PCR. We also discovered that P. ovale species are frequently being transmitted under the radar. However, transmission was highly uneven, with a small proportion of infectious carriers responsible for the majority of infected mosquitoes, many of them identifiable by conventional RDTs, suggesting strategies to tackle the infectious reservoir may exist.

Studies that used skin feeding for xenodiagnosis have long highlighted the low densities at which P. falciparum is transmitted, but many were done before the widespread use of PCR. Compared to more contemporary studies in Uganda, Kenya, Ethiopia, and Tanzania that relied on membrane feeding studies to assess the infectious reservoir [1, 2, 24-26], we describe a similar relationship between increasing gametocyte densities and higher mosquito infection rates, but uncover more low-level transmission from low-density infections, including in up to 20% of those without gametocytes detected by RT-qPCR. Parasites in these individuals were likely fluctuating at the limits of molecular detection [11].

The sensitivity of xenodiagnosis for detecting transmissible gametocytemia may be explained when one considers that 25 mosquitoes collectively sample 25 × 2–3 μL of blood, compared to <5 μL blood sampled in molecular assays, offering a greater chance that rare gametocytes are ingested. Because a male and female gamete must fertilize to form a zygote in the mosquito midgut, a single oocyst among 25 dissected midguts reflects the capacity for <0.1 circulating gametocytes/μL to find their way into the same mosquito host. The mechanisms behind this remarkable efficiency at the human-to-mosquito interface are yet to be elucidated. Similar to others, we did not find differences in gametocyte density in capillary versus venous blood to suggest gametocyte clustering in dermal capillaries [27, 28]. Neither have skin biopsies nor mosquito blood meal analysis led to evidence that gametocyte sampling differs in skin versus membrane feeding [27, 29]. Yet modeling of this biological efficiency suggests a role for gametocyte clustering in the human host [30].

Unexpectedly, we observed a high rate of cryptic co-transmission of P. ovale parasites (44% of DSFs with oocyst-positive midguts available for analysis). We have previously detected significant P. ovale prevalence in the study area, ranging as high as 15% during the long rainy season as well as efficient mosquito infectivity from mixed P. falciparum/P. ovale parasite carriers [13, 31]. Efficient transmission could help explain the persistence of P. ovale in areas where P. falciparum transmission has declined [15, 32]. Mosquito feeding studies in areas where P. ovale is known to be coendemic that use molecular methods for species detection can contribute to a better understanding of the true burden of P. ovale. The use of PCR for oocyst detection may overcome the challenge of microscopic detection of oocysts of malaria species that differ in their oocyst development time, with P. ovale oocysts taking about 3 days longer to mature compared to P. falciparum [33].

Despite these levels of cryptic transmission, it is encouraging that RDTs still identify a greater proportion of those infectious to mosquitoes relative to their ability to identify parasite carriers. We found that RDT-positive children (6–15 years) represented 27% of those assessed for transmissibility, but contributed to 58% of infected mosquitoes in the study cohort. This is consistent with other studies showing the importance of microscopically detectable infection [1, 24] and school-age children playing a key role in transmission [1, 6, 34]. Compared to microscopy, RDTs, the default malaria diagnostic throughout much of Africa, may more sensitively identify infectious carriers due to their ability to detect previous parasitemia, implicating some degree of chronic infection necessary to generate gametocytes. Additionally, the outsized contribution of a few highly infectious transmitters [1] may point to the potential for targeted strategies, or biology that can be harnessed to develop transmission-blocking interventions.

The primary limitation of this study was our inability to assess infectiousness of children <6 years who were not eligible for skin feeding. Unlike other epidemiologic studies, ours was not designed to be population-representative sampling to measure the relative components of the human infectious reservoir for malaria. Rather, our goal was to identify subpatent parasite carriers and study transmission from low-density infections. For example, females were overrepresented in our cohort (Supplementary Table 1). And we did not attempt to measure or account for disparities in mosquito exposure [7, 24]. Colony-reared mosquitoes adapted through many generations may not maintain the same susceptibility to infection as natural mosquito populations. Finally, molecular data for confirmation of P. falciparum mosquito infection were incomplete. However, our main findings are supported by sensitivity analyses conducted using only PCR-confirmed midgut infections, and limited parallel DMFAs and sporozoite infection assessments supported our overall estimates for infectivity.

Much remains to be learned about the nature of the malaria transmission reservoir. Ongoing analysis of this cohort is focused on other determinants of transmissibility. As malaria vectors evolve, infectivity data from wild-caught mosquitoes and other anopheline species, especially An. funestus and emerging An. stephensi, are crucial to identify the challenges ahead. The transmissibility of parasitemia post-treatment is also crucial to understanding the spread of drug resistance. Finally, a greater understanding of factors underlying the unevenness of mosquito infectivity and exposure can lead to new strategies to interrupt transmission, all the while understanding the limits of our ability to find and treat every member of the infectious reservoir.

Supplementary Material

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Data availability.

Datasets for the enrolled participants and mosquito direct skin feeding assays are available in the UNC Dataverse repository at https://doi.org/10.15139/S3/LLSKVX and https://doi.org/10.15139/S3/UU9KN6, respectively.