Identification of the asymptomatic Plasmodium falciparum and Plasmodium vivax gametocyte reservoir under different transmission intensities

Background Understanding epidemiological variables affecting gametocyte carriage and density is essential to design interventions that most effectively reduce malaria human-to-mosquito transmission. Methodology/Principal findings Plasmodium falciparum and P. vivax parasites and gametocytes were quantified by qPCR and RT-qPCR assays using the same methodologies in 5 cross-sectional surveys involving 16,493 individuals in Brazil, Thailand, Papua New Guinea, and Solomon Islands. The proportion of infections with detectable gametocytes per survey ranged from 44–94% for P. falciparum and from 23–72% for P. vivax. Blood-stage parasite density was the most important predictor of the probability to detect gametocytes. In moderate transmission settings (prevalence by qPCR>5%), parasite density decreased with age and the majority of gametocyte carriers were children. In low transmission settings (prevalence<5%), >65% of gametocyte carriers were adults. Per survey, 37–100% of all individuals positive for gametocytes by RT-qPCR were positive by light microscopy for asexual stages or gametocytes (overall: P. falciparum 178/348, P. vivax 235/398). Conclusions/Significance Interventions to reduce human-to-mosquito malaria transmission in moderate-high endemicity settings will have the greatest impact when children are targeted. In contrast, all age groups need to be included in control activities in low endemicity settings to achieve elimination. Detection of infections by light microscopy is a valuable tool to identify asymptomatic blood stage infections that likely contribute most to ongoing transmission at the time of sampling.

Introduction A variety of malaria control interventions aim to reduce the transmission of parasites from the human to the mosquito host. Vector control tools such as bed nets and indoor residual spraying [1] lower the risk for infection in humans, and for onward transmission. Additional public health interventions primarily aimed at reducing human-to-mosquito transmission are currently being applied or developed, e.g. mass screening and treatment [2], mass drug administration [3], transmission blocking vaccines [4], and ivermectin administration [5].
Interventions that reduce human-to-mosquito transmission are most effective when they target individuals within a population who contribute most to transmission. Not all individuals with blood-stage parasitemia are equally infectious to mosquitos. Only a small fraction of all parasites in the human host develop into sexual stages termed 'gametocytes'; parasite development in the mosquito relies on uptake and subsequent mating of male and female gametocytes [6].
Developing P. falciparum gametocytes are sequestered in extravascular sites such as bone marrow for 1-2 weeks [7,8]. Gametocytes appear in the blood stream after the first wave of asexual parasites, and, in the case of symptomatic malaria cases, are detectable by microscopy often only 1-2 weeks after presentation with fever [9]. Mature gametocytes infective to mosquitos circulate in peripheral blood for a period of a few days to up to three weeks [10][11][12]. Most commonly used antimalarials used to treat symptomatic cases do not clear sequestered or mature gametocytes. This is also the case for artemisinin combination therapy (ACT), the first-line drug in most P. falciparum endemic countries. As a result of the continued release of sequestered gametocytes after treatment, and their circulation for days to weeks, individuals can carry gametocytes for several weeks after treatment [12][13][14]. Primaquine is the only approved drug that clears P. falciparum gametocytes. Low-dose primaquine has been shown to reduce duration of gametocytemia after treatment [15,16].
In most malaria endemic regions the vast majority of infections are asymptomatic, i.e. not associated with fever. 50-80% of infections are not detectable by microscopic inspection of blood smears [23,24]. Since gametocytes account for only a small proportion of all parasites in peripheral blood, they are more difficult to detect by light microscopy (LM) than asexual parasites. The development of molecular methods to detect gametocyte-specific RNA transcripts by nucleic acid sequence-based amplification (NASBA) or reverse-transcriptase quantitative PCR (RT-qPCR) has allowed for detection of submicroscopic gametocytemia [25,26]. However, few studies have reported gametocyte carriage in asymptomatic individuals in non-African settings, especially with respect to P. vivax [27][28][29][30][31][32][33]. It is not known whether gametocyte densities differ across regions of different transmission intensities, and the distribution of gametocyte carriers among various demographic groups within a community is not well understood. Knowledge of these epidemiological variables is needed to target transmissionreducing interventions to those at highest risk of gametocyte carriage and to understand the long-term impact of these interventions on progress towards malaria elimination.
To advance our understanding of P. falciparum and P. vivax gametocyte carriage across a range of transmission settings, five cross-sectional surveys involving a total of 16,493 individuals were conducted in Brazil, Thailand, Papua New Guinea (PNG) and Solomon Islands [27][28][29][30][31]. The surveys included endemic areas where transmission was moderate to high or had recently decreased (PNG in 2010 and 2014, Solomon Islands) and regions where transmission was low, with P. vivax being the predominant parasite (Brazil, Thailand). Blood samples were collected from the general population irrespective of symptoms of malaria illness. Total bloodstage parasites and gametocytes were quantified by LM and sensitive qPCR and RT-qPCR assays using the same methodology across all studies, allowing for the first direct comparison across transmission intensities. ). Prior to sample collection, the aims of the study were explained to all individuals and informed written consent was obtained from participants or, in case of minors, from their guardians.

Study sites and sample collection
Details on study sites are given in Table 1. Community sensitization took place 1-2 weeks prior to sample collection. Convenience sampling was applied to select households for the surveys. All members of the selected households were noted on a list, and individuals above 6 months of age were invited to participate. Sampling started in the morning and continued throughout the day. As children might be in school during this time and adults away for work, efforts were made to sample school-aged children and working adults after they returned home. The age and gender distributions of the study participants for each survey is given in S1 Table, and is expected to be representative for the population. Overall, the distributions were similar across surveys, with the exception of a lower proportion of all sampled being small children in Brazil and Thailand.
The study sites experience little or moderate seasonality in transmission. In PNG, the rainy season is from December to April. Both surveys were conducted in May to July, i.e. after the rainy season. There is minimal seasonal variation in Solomon Islands; samples were collected in May to June. In Thailand, the peak transmission season is from April to July, and samples were collected in September and October. In Manaus, Brazil, highest incidence occurs from May to September. Half of the samples were collected in September to early January, and the other half in August to September.
Form each participant, 250 μL blood was collected by finger prick into 2 mL EDTA microtainers (BD). Hemoglobin levels were determined using the HemoCue handheld meter. For RNA extraction and gametocyte detection, 50 μL blood was transferred into tubes containing 250 μL of RNAprotect (Qiagen) in the field. Samples in RNAprotect were kept on ice packs in the field and transferred to -80˚C storage every evening and kept there until RNA extraction. The remaining 200 μL blood were kept in the EDTA microtainer, also kept on ice packs, and transferred to -20˚C storage until DNA extraction.

Molecular methods
DNA was extracted from 200 μL whole blood kept in EDTA microtainers and eluted in 200 μL elution buffer. Parasites were quantified by qPCR using the 18S rRNA gene as target [34]. The assay used detects one copy of the gene. 4 μL DNA was screened by qPCR, thus the limit of detection was 0.25 parasites/μL blood (i.e. 1 genome per 4 μL DNA). RNA was extracted from 50 μL whole blood kept in RNAprotect, and eluted in 50 μL elution buffer. 2 μL RNA was screened by RT-qPCR. Gametocytes were quantified by RT-qPCR of the female gametocytespecific transcripts pfs25 and pvs25 [35].
Procedures for sample collection, qPCR, and RT-qPCR were standardized between all sites. Results of individual cross-sectional surveys have been published previously [28][29][30][31]35]. For qPCR and RT-qPCR, standardized plasmids were distributed to all laboratories and run along samples for relative quantification and estimation of sensitivity. Sensitivity of all assays was 0.5-1 copies/uL DNA or RNA. For absolute measurements of copy numbers, a subset of samples from each laboratory was quantified by droplet digital PCR [36]. Due to different procedures of RNA sample collection in Solomon Islands, only positivity, but not pvs25 and pfs25 copy numbers, were included in the analysis. Expert microscopy was conducted in PNG, Solomon Islands, and for parts of the survey in Brazil. To determine multiplicity of infection (MOI), P. falciparum infections were genotyped by msp2 [37], and P. vivax infections were genotyped by msp1F3 and MS2 [38].

Data analysis
The following definitions are used: 'proportion gametocyte-positive infections' describes the number of gametocyte carriers divided by the number infected with asexual parasites and/or gametocytes; 'population gametocyte prevalence' is the prevalence of gametocytes among all individuals, infected and non-infected [39].
Multivariable regression models were used to predict factors associated with the proportion gametocyte-positive infections and gametocyte density. Parasite densities were log 10 transformed for all calculations. To correct for imperfect detection of gametocytes and include lowdensity infections without gametocytes detected in multivariable models, +0.1 was added to all gametocyte density values prior to log 10 -tranformation. For multivariable analyses, individuals were grouped into age classes �6, >6-12, >12-20, and >20 years. Survey and age group were included as fixed effects. The ratio of pfs25 or pvs25 transcripts per parasite genome was assessed, representing the proportion of gametocytes among all parasites. Infections (Pf: n = 12, Pv: n = 5) with densities at the technical limit of detection of 0.25 copies/μL blood (i.e. 1 DNA/RNA template per PCR) were excluded from correlation analysis as quantification is imprecise at very low densities, and including them at a set density of 0.25 copies/μL would artificially increase correlation. All data is available in supplementary file S1 Data.
To assess whether the proportion of gametocytes among all blood-stage parasites differs between infections of different parasite density, the number of pfs25 or pvs25 transcripts per P. falciparum or P. vivax genome among gametocyte-positive samples was plotted (Fig 3A and  3B, infections with �5 DNA copies excluded). For P. falciparum, a 0.39-fold decrease in the proportion gametocytes per 10-fold increase in parasite density was observed (n = 255, P<0.001), while no significant change was observed for P. vivax (n = 243, 1.19-fold increase per 10-fold increase in parasite density, P = 0.057). A febrile episode and/or antimalarial treatment in the preceding 2 weeks increased the proportion gametocytes significantly for P.  Fig 3C), but had no impact on P. vivax gametocyte proportions (n = 288, 4 episodes, P = 0.4319, Fig 3D).

Multivariate risk factors of gametocyte positivity and density
In multivariate analysis across all surveys, parasite density was a strong predictor for the probability that a sample was gametocyte-positive (Table 3). Each 10-fold increase in P. falciparum parasite density resulted in a 1.59-fold increase in the odds of gametocyte positivity, and in a 1.9-fold increase in gametocyte densities ( Table 3). The correlation was much stronger for P. vivax, where each 10-fold increase in parasite density resulted in a 3.15-fold increase in the odds of gametocyte positivity and a 3.9-fold increase in gametocyte density (Table 3). Concordance between P. vivax genome and gametocyte density was even higher for infections above 5
Among P. falciparum positive individuals, the odds to detect gametocytes was 54% lower (P<0.0001) and gametocyte densities were 69% lower (P<0.0001) in individuals co-infected with P. vivax. Reported malaria or anti-malarial treatment in the past 2 weeks was associated with higher P. falciparum gametocyte prevalence and densities (Table 3). Parasite densities strongly decreased with increasing age. Even when including parasite density as confounder, P. falciparum gametocyte positivity and density decreased with increasing age, i.e., gametocyte densities decrease to a greater extent than blood-stage parasitemia (Table 3). Among P. vivax positive individuals, gametocyte densities, but not positivity decreased with age (Table 3). Apart from DNA copy numbers, no other significant associations were observed.
Factors affecting gametocyte density were assessed independently in PNG 2010 (moderatehigh transmission), PNG 2014 (recently decreased transmission), and Thailand and Brazil pooled (long-time low transmission) ( Table 4). For PNG 2010 results were very similar to

PLOS NEGLECTED TROPICAL DISEASES
pooled data from all surveys. In PNG 2014, for P. falciparum, gametocyte densities decreased with age and were lower in individuals co-infected with P. vivax, but were not significantly associated with parasite density (P = 0.085). In Thailand and Brazil, for both species, and for P. vivax in PNG 2014, no other factors than parasite density were significantly associated with gametocyte density. As a result of the correlation of parasite and gametocyte density and the decreasing parasite densities with age in both PNG surveys, for both species the majority of individuals with detectable gametocytes were children. 48-78% of gametocyte carriers were <12 years (Fig 4). In contrast, 65-67% of gametocyte carriers were >20 years in Thailand and Brazil.
When including parasite density in multivariable analysis to correct for differences in mean density among surveys, the proportion gametocyte positive infections, and gametocyte densities of both species differed significantly between surveys (P<0.0001, Table 3). The probability to detect gametocytes increased in all surveys with increasing genome density. An interaction analysis did not reveal a significant difference of this increase between surveys (Pf: n = 676, P = 0.471; Pv: n = 1501, P = 0.512). Likewise, P. falciparum and P. vivax gametocyte densities were not affected by an interaction between DNA copies and survey (Pf: n = 375, P = 0.877, Pv: n = 415, P = 0.132).
Hemoglobin measurements and multiplicity of infection data were available from the surveys in PNG and Solomon Islands; neither affected gametocyte positivity or density in multivariate analysis (S2 and S3 Tables).   Table 2), and appeared to be particularly high when prevalence was low. For example, 4/4, and 7/10 P. falciparum gametocyte carriers were LM-positive in Solomon Islands and Brazil, respectively ( Table 2). P. vivax gametocytes were detected by RT-qPCR in 64.9% (235/362) of LM-positive individuals, but in only 16.7% (163/976) of LM-negative individuals. Mean gametocyte densities in LM-positive individuals were almost twice as high compared to LM-negative individuals (6.02 vs. 3.44 transcripts/μL, P<0.001). Thus, approximately 50% of P. falciparum and 59% of P. vivax gametocyte carriers were positive by light microscopy.

Discussion
We observed substantial differences in the proportion of P. falciparum and P. vivax infections carrying detectable gametocytes in 5 cross-sectional surveys representing distinct malaria-epidemiological contexts. The proportion of gametocyte-positive infections in community surveys is heavily impacted by the sensitivity of the assays used for parasite and gametocyte detection [39]. The use of the identical methodology and external reference standards for all surveys allowed, for the first time, direct comparisons between regions of different transmission intensity.

PLOS NEGLECTED TROPICAL DISEASES
Blood-stage parasite densities were a strong predictor for gametocyte positivity and could largely explain differences between surveys. In most surveys, the majority of gametocyte carriers of both Plasmodium species (as determined by RT-qPCR) were positive by LM for asexual parasites. A lower proportion of gametocyte carriers was LM-positive in Madang 2014 (both species) and Solomon Islands (P. vivax), where transmission had declined in the years prior to the surveys [28,30], and parasite densities were very low. In these surveys, a large proportion of gametocyte carriers could not be diagnosed by microscopy. In contrast, in the sites where transmission had been reduced for longer (Brazil, P. falciparum in Solomon Islands), expert LM or other tools such as RDT remain sufficiently sensitive to identify the majority of gametocyte carriers.
Pronounced age trends in gametocyte carriage were evident in PNG and Solomon Islands. Prevalence of infection peaked in children or adolescents, and parasite densities decreased rapidly with increasing age, most likely due to the acquisition of immunity. As a result, the vast majority of gametocytes were detected in children below 6 years, especially for P. vivax. This contrasts findings from P. falciparum in Africa, where school-age children were proposed to contribute most to transmission densities [40,41]. In moderate-high transmission settings and in regions of steep decline in transmission in recent years, gametocyte densities decreased even faster than parasite densities with age. Thus, changes in parasite prevalence and density with age might not appropriately reflect changes in transmission potential. In Brazil and Thailand, the risk of infection increased with increasing age, age trends of parasite densities were moderate, and as a result no age trends in gametocyte densities were evident.
Apart from parasite density, limited effects of transient factors on gametocyte densities were observed. Multiple clone infection or hemoglobin levels did not affect gametocyte carriage of either species. For P. vivax, a constant proportion of gametocytes among all parasites was observed irrespective of parasite density. In the case of P. falciparum, high proportions of gametocytes were observed in a subset of infections with low-to-moderate densities The 2-week sequestration of developing P. falciparum gametocytes results in a temporal lag of peak gametocytemia following peak parasitemia [6]. Thus, infections with low parasite but high gametocyte densities might have experienced a recent wave of asexual parasitemia [42]. This is corroborated by the fact that self-reported febrile illness in the two weeks prior to sample collection resulted in higher gametocyte densities (Fig 3C and 3D). Conversion of a large proportion of all parasites into gametocytes when parasite densities drop to very low levels has also been described in a rodent malaria model [43]. Longitudinal studies with frequent sampling will be needed to assess how closely P. falciparum gametocyte density reflects parasite density in the preceding 2 weeks.
As an exception to the limited impact of transient factors, a lower proportion P. falciparum gametocyte positive infections and lower gametocyte densities were observed in individuals co-infected with P. vivax. It is not known whether co-infection results in an adjustment of the gametocyte conversion rate, or whether multi-species infection is a surrogate marker for higher exposure, and thus, higher levels of gametocyte-specific immunity.
Mosquito feeding assays have repeatedly shown a correlation between parasite density and infectivity. Few studies have included asymptomatic individuals and individuals negative by LM for gametocytes and asexual stages [44][45][46][47][48][49][50]. With few exceptions, e.g., one study on P. vivax infectivity in Brazil [48], individuals with either asexual parasites or gametocytes detected by LM were far more infective than submicroscopic infections (Fig 5). For P. vivax, in Thailand a steep increase in infectivity was found at densities (by LM) of 10-100 parasites/uL, with little effect if densities increased further [47]. This density closely matches the limit of detection of expert LM. The finding in the present study of a majority of gametocytes concentrated in LM-positive individuals, together with the results from mosquito-feeding studies, corroborate that LM-positive individuals likely are the main infectious reservoir.
Studies assessing gametocyte densities and infectivity over time will be required to determine what proportion of low-density infections will rise in density and become highly infective. The importance of male gametocyte densities to predict infectivity is increasingly recognized. In low density infections, male gametocytes might be the limiting factor for onward transmission [51]. Measuring male gametocyte densities in addition to female densities, as measured by pfs25 and pvs25 RT-qPCR, is expected to allow for better predictions of infectivity [51]. Due to the non-linear relationship between gametocyte density and infectivity, age trends in gametocyte density might not fully reflect infectivity. Lastly, while higher gametocyte densities result in increased oocyst numbers in mosquitos [51], it is unclear whether oocyst numbers have an effect on the efficiency of onward transmission.
Even after correcting for different mean parasite densities between surveys, a higher proportion gametocyte positive samples were found for both species in Thailand and for P. falciparum in Brazil and Solomon Islands compared to PNG. Recent malaria control activities have resulted in significant changes in vector composition and biting behavior in the study sites [52,53]. A reduction in the number of mosquito bites or shift towards a less competent vector might select for parasites with higher gametocyte conversion rates. Such a selection has been suggested by recent genome and transcriptome studies. Expression of the AP2-G transcription factor and additional epigenetic factors involved in gametocytogenesis were adjusted to transmission levels in P. falciparum populations in East Africa [54], and the gametocyte development gene 1 (gdv1), which is essential for early gametocyte development, was found to be under strong selection in P. falciparum populations in regions of different endemicity in West Africa [55]. In Cambodia, control efforts resulted in strong selection of the AP2-G homolog in P. vivax [56]. The present study, for the first time, found differences in the proportion of gametocytes among all parasites between regions of different transmission intensity.

Conclusions
The probability to detect gametocytes was closely correlated to blood-stage parasitemia in different transmission settings. The vast majority of all infections with high gametocyte densities (as determined by RT-qPCR) could be diagnosed by microscopy. Pronounced age trends of gametocyte carriage in areas of moderate to high transmission was observed. Even though the contribution to transmission is influenced by vector exposure and transmission blocking '% Mosquitos infected' shows the combined proportion of mosquitos infected including individuals that did not infect any mosquitos. n/N show the proportion of individuals that infected at least one mosquito. Infectivity varied widely, but with the exception of the study on P. vivax in Brazil, microscopy-positive individuals were 5-10-fold more infective that those with submicroscopic infection. Data from [45,[47][48][49][50].
https://doi.org/10.1371/journal.pntd.0009672.g005 PLOS NEGLECTED TROPICAL DISEASES immunity in addition to gametocyte prevalence and density, the age trends suggest that interventions to reduce transmission will have the greatest effect when targeted towards children. In contrast, in order to achieve elimination in low transmission settings individuals of all ages need to be protected from vector contact, which is often not the case [57].