Transmission-blocking activities of artesunate, chloroquine, and methylene blue on Plasmodium vivax gametocytes

ABSTRACT

Plasmodium vivax is now the main cause of malaria outside Africa. The gametocytocidal effects of antimalarial drugs are important to reduce malaria transmissibility, particularly in low-transmission settings, but they are not well characterized for P. vivax. The transmission-blocking effects of chloroquine, artesunate, and methylene blue on P. vivax gametocytes were assessed. Blood specimens were collected from patients presenting with vivax malaria, incubated with or without the tested drugs, and then fed to mosquitos from a laboratory-adapted colony of Anopheles dirus (a major malaria vector in Southeast Asia). The effects on oocyst and sporozoite development were analyzed under a multi-level Bayesian model accounting for assay variability and the heterogeneity of mosquito Plasmodium infection. Artesunate and methylene blue, but not chloroquine, exhibited potent transmission-blocking effects. Gametocyte exposures to artesunate and methylene blue reduced the mean oocyst count 469-fold (95% CI: 345 to 650) and 1,438-fold (95% CI: 970 to 2,064), respectively. The corresponding estimates for the sporozoite stage were a 148-fold reduction (95% CI: 61 to 470) and a 536-fold reduction (95% CI: 246 to 1,311) in the mean counts, respectively. In contrast, high chloroquine exposures reduced the mean oocyst count only 1.40-fold (95% CI: 1.20 to 1.64) and the mean sporozoite count 1.34-fold (95% CI: 1.12 to 1.66). This suggests that patients with vivax malaria often remain infectious to anopheline mosquitos after treatment with chloroquine. Use of artemisinin combination therapies or immediate initiation of primaquine radical cure should reduce the transmissibility of P. vivax infections.

INTRODUCTION

Plasmodium vivax is a major cause of malaria worldwide: approximately one-third of the global population is at risk of infection, and it is estimated that there are about 10 million symptomatic cases each year (1). Vivax malaria has been relatively neglected because it rarely causes acute death (2, 3), although it is associated with indirect morbidity, poor pregnancy outcomes, and, in highly endemic areas, recurrent infections do contribute to anemia-related mortality (4, 5). P. vivax is associated with repeated malaria relapses arising from persistent liver stages (hypnozoites) and is particularly difficult to control and eliminate (6, 7). Treatment of symptomatic malaria with effective antimalarials reduces transmission (8). This plays a central role in malaria control and elimination in the low-transmission settings where P. vivax is prevalent (8). In P. falciparum infections, gametocytogenesis is delayed, so prompt effective treatment with schizonticidal drugs (artemisinin-based combination therapies) reduces transmissibility (9, 10). In addition, a single low dose of primaquine is usually co-administered to kill the mature transmissible gametocytes of P. falciparum that are insensitive to schizonticidal drugs (11). The treatment of vivax malaria is more complex. Schizonticidal drugs (chloroquine- or artemisinin-based combination therapies) active against the pathogenic asexual blood stages are used to clear parasitemia and obtain clinical remission, but to achieve radical cure (i.e., killing the hypnozoites in the liver and thereby preventing subsequent relapses), treatment with a higher dose of 8-aminoquinoline (7 or 14 days primaquine or single-dose tafenoquine) is required in addition (12). In contrast to P. falciparum, gametocytogenesis in P. vivax infections occurs together with asexual stage development, so symptomatic patients are usually infectious to vector anopheline mosquitos.

The inability to cryopreserve and then conduct long-term culture of P. vivax (13) compromises laboratory assessment of transmission-blocking activity outside endemic areas. These assessments therefore require the proximity of an insectary, a laboratory, and parasitemic patients. As a result, few studies have been performed, and the effects of antimalarial drugs on P. vivax gametocytes are not well characterized (14). Chloroquine is currently considered active against Plasmodium gametocytes, except for the mature stages of P. falciparum (15–17). However, the transmission of P. vivax to mosquitos has been observed for up to 72 hours after starting treatment which suggests that chloroquine may lack parasiticidal activity against mature P. vivax gametocytes (18). With the exception of the 8-aminoquinolines, artemisinins are more active against mature Plasmodium gametocytes than other antimalarials (16, 19). In vivax malaria, the artemisinin combination therapy (ACT) dihydroartemisinin-piperaquine was reported to have a superior transmission-blocking effect compared with chloroquine, but the individual effects of the two drugs in the combination were not studied (20). Similarly, the artesunate-mefloquine ACT regimen co-administered with primaquine was reported recently to have a superior transmission-blocking effect compared with chloroquine co-administered with primaquine or tafenoquine (21). This supports earlier observations that artemisinin derivatives had greater activity than chloroquine in reducing gametocyte carriage in vivax malaria (22, 23). High concentrations of methylene blue, which has a potent gametocytocidal activity in P. falciparum (24, 25), were shown recently to block transmission of P. vivax gametocytes in membrane-feeding experiments, but the sample size was very small (only five patients were recruited in this study) (26).

The aim of this study was to compare the transmission-blocking activities of artesunate, chloroquine, and methylene blue on P. vivax gametocytes. Anopheles dirus mosquitos (a major vector in most of Southeast Asia) from a laboratory-adapted colony were fed on blood specimens collected from vivax malaria patients and incubated with or without drug. Drug effects on oocyst and sporozoite counts in mosquito samples were analyzed under a Bayesian multi-level negative binomial model, accounting for assay variability and heterogeneity of Plasmodium development in the mosquito (27).

RESULTS

Overall 38 adult patients with P. vivax malaria provided blood samples, and 342 Anopheles dirus mosquito batches were fed on these samples (Appendix, Table S1). Baseline sample characteristics are shown in Table 1. Overall, the median parasite densities were 13,161 asexual parasites per microliter [inter-quartile range (IQR): 6,981 to 27,798] and 1,092 gametocytes per microliter (IQR: 473 to 2,009), respectively. All but one of the blood samples were infectious to mosquitos (the one sample that was not infectious at baseline became infectious after 24 hours of incubation). The median oocyst index (i.e., the proportion of mosquitos harboring malaria oocysts per batch) was 0.96 (IQR 0.84 to 0.98), and the median oocyst count in infected mosquitos was 63.8 oocysts per mosquito (IQR 4.1 to 124.9). The corresponding figures for the sporozoite stage were 0.7 (IQR 0.35 to 0.90) and 211 sporozoites per infected mosquito (IQR 5 to 4,321). The median ratio of the median parasite count in the controls after 24 hours of incubation without drug to the median baseline parasite count (on the day of sample collection) was 0.93 (IQR: 0.51 to 3.68) and 0.9 (IQR: 0.01 to 6.56) for the oocyst and sporozoite stages, respectively (Appendix, Fig. S1).

TABLE 1.

Characteristics of the blood samples at baseline

Characteristic	Value in the experimental group
	Median	IQR
No. of blood samples
Artesunate	9	–a
Chloroquine	21	–
Methylene blue	8	–
Asexual parasitemia (no. of asexual parasites per microliter of blood)
Artesunate	12,109	8,258 to 16,608
Chloroquine	12,740	3,610 to 21,652
Methylene blue	17,718	10,116 to 28,675
Gametocytemia (no. of asexual parasites per microliter of blood)
Artesunate	767	524 to 1,350
Chloroquine	898	195 to 1,849
Methylene blue	1,821	1,309 to 4,063
Oocyst index
Artesunate	0.98	0.94 to 0.98
Chloroquine	0.96	0.68 to 0.98
Methylene blue	0.96	0.92 to 0.98
Median oocyst count (no. of oocysts per mosquito)
Artesunate	95	48.5 to 127
Chloroquine	18.5	2 to 92
Methylene blue	114.5	54.6 to 183.9
Sporozoite index
Artesunate	0.60	0.10 to 0.70
Chloroquine	0.65	0.27 to 0.86
Methylene blue	0.95	0.87 to 1.00
Median sporozoite count (no. of sporozoites per mosquito)
Artesunate	33	0 to 211
Chloroquine	122	2.5 to 2,377
Methylene blue	4,604	2,402 to 10,876

^a –, not applicable.

Chloroquine exhibited little transmission-blocking activity on P. vivax gametocytes despite the high concentrations used (Fig. 1). Of all dissected mosquitos in the chloroquine-treated samples, 2,974/4,036 (74%) carried oocysts in the treated replicates compared with 3,299/4,026 (82%) in the controls [relative risk (RR): 0.85 (95% CI: 0.82 to 0.88), P < 0.0001], and 701/1,177 (60%) carried sporozoites in the treated replicates compared with 785/1,228 (64%) in the controls [RR: 0.89 (95%CI: 0.81 to 0.97), P = 0.005]. In contrast, artesunate and methylene blue almost completely interrupted gametocyte transmission. For artesunate, only 207/1,798 (12%) of the mosquitos carried oocysts in the treated replicates compared with 1,591/1,797 (89%) in the controls [RR: 0.057 (95% CI: 0.044 to 0.075), P < 0.0001], and 1/360 (0.3%) dissected mosquitos carried sporozoites in the treated replicates vs 152/360 (42%) in the controls [RR: 0.005 (95% CI: 0.001 to 0.035), P < 0.0001]; for methylene blue, only 76/1,599 (5%) carried oocysts in the treated replicates vs 1,470/1,592 (92%) in the controls [RR: 0.039 (95% CI: 0.028 to 0.054), P < 0.0001], and only 5/320 (1.6%) carried sporozoites in the treated replicates vs 267/320 (83%) in the controls [RR: 0.01 (95% CI: 0.003 to 0.027), P < 0.0001]. The lower sporozoite index observed in the controls of the artesunate group in comparison to the sporozoite index at baseline was probably explained by the detrimental effect of artesunate wash off on sporogony (see model coefficient estimate below).

Fig 1.

Effects of chloroquine, methylene blue, and artesunate on the development of P. vivax in An. dirus mosquitos. (A–C) Median oocyst count, (D–F) oocyst index, (G–I) median sporozoite count, and (J–L) sporozoite index. Values in the control and treated replicates were collated by assay run.

As observed previously, there was considerable heterogeneity in the count data across mosquitos within a batch and considerable variability in the count data across blood samples and experimental batches. To account for this heterogeneity, we estimated the drug effects under a Bayesian multi-level model (mixed effects), whereby the count data were modeled as negative binomial with the dispersion parameter as a parametric function of the mean count (Table 2; see Materials and Methods) (27). In contrast to the previous data describing proportions of mosquitos with parasites, the model parameterized the drug effect as a reduction in the mean number of parasites per mosquito, accounting for variability across blood samples and mosquito batches. Under this model, gametocyte exposure to chloroquine decreased the mean oocyst count by only 1.40-fold [95% credible interval (CrI): 1.20 to 1.65; observed mean of 100 oocysts per mosquito in the controls vs 69 in the treated replicates], and it decreased the mean sporozoite count by 1.34-fold (95% CrI: 1.12 to 1.66; observed mean of 14,414 sporozoites per mosquito in the controls vs 11,132 in the treated replicates). In contrast, artesunate reduced the mean oocyst count by 469-fold (95% CrI: 345 to 650; observed mean of 60 oocysts per mosquito in the controls vs 0.22 in the treated replicates), and methylene blue reduced the oocyst count by 1,438-fold (95% CrI: 970 to 2,064; observed mean of 107 oocysts per mosquito in the controls vs 0.08 in the treated replicates). For the sporozoite counts, the estimates were a 148-fold reduction for artesunate (95% CrI: 61 to 470; observed mean of 1,303 sporozoites per mosquito in the controls vs 0.1 in the treated replicates) and 536-fold for methylene blue (95% CrI: 246 to 1,311; observed mean of 13,914 sporozoites per mosquito in the controls vs 1.8 in the treated replicates). The model fitted the data well as shown by the inferred relationship between the mean parasite count and proportion of Plasmodium-infected specimens in mosquito samples (Fig. 2). As expected, both inter- and intra-experiment variability were large, and inter-experiment variability was larger than intra-experiment variability. The median fold variation in the mean parasite count across blood samples was 1.07-fold (IQR: 0.62 to 2.08, range: 0.16 to 4.00) and 1.32-fold (IQR: 0.47 to 2.40, range: 0.12 to 6.83) for the population means for the oocyst and sporozoite stages, respectively (Fig. 3). The median fold variation in the mean parasite count across technical replicates was 1.00-fold (IQR: 0.77 to 1.30, range: 0.005 to 4.05) and 0.99-fold (IQR: 0.94 to 1.05, range: 0.62 to 1.88) the patient means for the oocyst and sporozoite stages, respectively (Fig. 4). One sample with abnormally high intra-experiment variability in the mean oocyst count was detected, but no obvious explanation for this outlier was identified. Inclusion or exclusion of this sample from the analysis did not significantly change the results (data not shown). Moreover, the development of sporozoites mirrored that of the oocysts: a 10-fold increase in the mean oocyst count was associated with a 3.52-fold (95% CI: 2.15- to 4.90-fold) increase in the mean sporozoite count (Appendix, Fig. S2). To explain variation in blood meal infectiousness to mosquitos across blood samples, the log₁₀[mean oocyst count], log₁₀[asexual parasitemia], and log₁₀[gametocytemia] assessed on admission (i.e., on the collection day before the 24-hour incubation time with or without drug) were introduced as linear predictors of the mean parasite count in mosquito samples of the experimental replicates (i.e., after 24 hours of incubation with or without drug). A 10-fold increase in the mean oocyst count and gametocytemia at baseline was associated with a 1.76- (95% CrI: 1.20 to 2.48) and a 6.47-fold increase (95% CrI: 3.04 to 13.37), respectively, in the mean oocyst counts in the experimental replicates; there was no significant association between the mean oocyst count in experimental replicate and baseline asexual parasitemia [model coefficient estimate: 1.08 (95% CrI: 0.56 to 2.11)] or artesunate wash off [model coefficient estimate: 0.84 (95% CrI: 0.41 to 1.71)]. A 10-fold increase in baseline gametocytemia and artesunate wash off was associated with a 3.56-fold increase (95% CrI: 1.21 to 9.20) and a 3.85-fold decrease [model coefficient estimate: 0.26 (95%CrI: 0.10 to 0.61)], respectively, in the mean sporozoite counts in the experimental replicates. There was no significant association between the mean sporozoite count in the experimental replicates and the mean oocyst count [model coefficient estimate: 1.21 (95% CrI: 0.72 to 1.96)] or asexual parasitemia [model coefficient estimate: 1.22 (95% CrI: 0.48 to 2.96)] at baseline.

TABLE 2.

Parameter estimates given by the output of transmission-blocking activity model

Parameter	Oocysts		Sporozoites
	Median of posterior draws	95% CrI	Median of posterior draws	95% CrI
μ population	44.6	31.3 to 63.7	4906.6	2,875 to 8158.8
σ patient	0.92	0.74 to 1.15	1.22	0.98 to 1.54
σ batch	0.46	0.41 to 0.53	0.31	0.23 to 0.41
a 0	0.41	0.37 to 0.45	0.0007	0.0004 to 0.0012
a 1	0.19	0.16 to 0.21	0.61	0.56 to 0.68
Artesunate effect	0.0021	0.0015 to 0.0029	0.0068	0.0021 to 0.0165
Chloroquine effect	0.71	0.61 to 0.84	0.74	0.60 to 0.90
Methylene blue effect	0.0007	0.0005 to 0.0010	0.0019	0.0008 to 0.0041
Log10[baseline oocyst count]	1.76	1.20 to 2.48	1.21	0.72 to 1.96
Log10[baseline parasitemia]	1.08	0.56 to 2.11	1.22	0.48 to 2.96
Log10[baseline gametocytemia]	6.47	3.04 to 13.37	3.56	1.21 to 9.20
Wash off	0.84	0.41 to 1.71	0.26	0.10 to 0.61

Fig 2.

Relationship between the mean parasite count and proportion of Plasmodium-infected mosquitos in the assay. (A) Paired mean number of oocysts per mosquito and oocyst index determined in the control and treated replicates; (B) paired mean number of sporozoites per mosquito and sporozoite index. The black line and shaded area show the model-fitted relationship plotted using a₀ and a₁ estimates given by the model output and the corresponding 95% credible interval, respectively.

Fig 3.

Inter-experiment variability. Mean estimated oocyst (top panel) and sporozoite (bottom panel) counts (in the untreated state) under the Bayesian multi-level model; points and error bars show the median and 95% credible interval of posterior draws, respectively.

Fig 4.

Temporal trend in intra-assay variability. Random variation of the mean oocyst (top panel) and sporozoite (bottom panel) count under the Bayesian multi-level model; points and error bars show the median and 95% credible interval of posterior draws, respectively.

DISCUSSION

This assessment of the transmission-blocking effects of antimalarial drugs on P. vivax gametocytes revealed that chloroquine has little activity against P. vivax gametocytes. Its activity is probably limited to the immature sexual stage parasites having a food vacuole, i.e., the pre-macrogametocytes originally described by Boyd (28). In contrast, high doses of artesunate and methylene blue have potent P. vivax gametocytocidal and thus transmission-blocking effects.

The data also confirm previous observations on the relationship between intensity (number of oocysts or sporozoites per mosquito) and prevalence (oocyst or sporozoite index) of Plasmodium infections in artificially infected mosquitos (27). Transmission-blocking drugs primarily reduce the intensity (number and viability) of oocyst development, and the resulting effect on prevalence varies with the mean parasite count in mosquito samples, being less in high-intensity than in low-intensity infections. This is well described by a negative binomial model with a dispersion parameter as a function of the mean count (27). As oocysts arising from gametocytes exposed to an antimalarial drug may fail to produce viable sporozoites, the primary outcome of transmission-blocking assays should therefore be the reduction in sporozoite carriage.

The results of this study highlight important differences in the intrinsic susceptibility of P. vivax and P. falciparum gametocytes to antimalarial drugs. Unlike other human malaria parasite species, Plasmodium falciparum gametocytes’ emergence is delayed with respect to asexual parasite densities. Their maturation takes longer, and mature stage V gametocytes are intrinsically resistant to most antimalarial drugs, except methylene blue and the 8-aminoquinolines (29). Artesunate, which kills young circulating sexual stages but fails to kill mature P. falciparum gametocytes (19), exhibited a potent transmission-blocking effect on P. vivax. This has important implications for the choice of treatment.

This study had several limitations. The characteristics of P. vivax gametocyte maturation and the determinants of gametocyte infectiousness to mosquitos are not well characterized (30). The sex- and stage-specific effects of drugs on the gametocytes were not assessed. Sex- and stage-specific gametocytocidal effects in P. falciparum were characterized previously (19, 31). This has never been assessed in P. vivax, probably because these aspects of Plasmodium biology are less well understood in P. vivax than in P. falciparum. The experiment was designed to maximize the power to detect a drug effect, and so a single high concentration was investigated. Concentration-response relationships were not evaluated. The correlation between the exposures of drugs in vitro and in vivo is also not well characterized, and the concentrations tested in this study may not represent drug activity at therapeutic doses. Testing of lower concentrations of the active drugs and assessment of dose-response relationships would be informative. The 8-aminoquinolines are considered potent gametocytocides, but they are pro-drugs, and the absence of ex vivo metabolism precluded their investigation. The bioactivation of primaquine is complex, generating unstable bioactive intermediates and compromising quantitative assessment of exposure-response relationships (32–34). Interestingly, primaquine, which has potent effects against P. falciparum gametocytes (11, 35), was shown in one study to be less effective in killing P. vivax gametocytes (36). Assessment of the gametocytocidal effects of biotransformed primaquine and other 8-aminoquinolines on P. vivax gametocytes will require further research. Sporozoite viability was not assessed in this study and may lead to underestimation of the effect of chloroquine. However, successful invasion of the mosquito salivary glands is already an indication of their viability. This limitation could be addressed by assessing the development of liver stages inoculated with sporozoites detected in the assay (37). Susceptibility of asexual parasites to the drugs was not determined. It could be argued that the observed low-transmission-blocking activity of chloroquine against P. vivax gametocytes results from parasite resistance rather than intrinsic lack of gametocyte susceptibility to the drug. However, high-level resistance is largely confined to Oceania and Indonesia, so significant resistance is unlikely given the good treatment efficacy and the reported data on P. vivax asexual blood stages susceptibility to antimalarial drugs in this study area (38).

Using gametocytocidal drugs (artemisinin combination treatments) for the first-line treatment of vivax malaria may reduce infection transmissibility, but it is important to consider the timing of gametocyte development and transmission in vivo. P. vivax gametocytes can arise directly from exo-erythrocytic schizonts and become detectable in the peripheral circulation as early as the asexual blood stages (37, 39). In addition, the lower limit of gametocyte density for transmission to mosquito is lower in P. vivax than for the other human malaria parasite species: successful transmission to vector mosquitos can occur with densities of gametocytes as low as five gametocytes per microliter (40). These densities are below the limit of routine microscopy detection. Previous exposure increases the pyrogenic threshold (circa 10 parasites per microliter in naive individual vs approximately 200 parasites per microliter in the immune subject) (41, 42), and infected individuals often carry transmissible densities of gametocytes without any symptom. Therefore, in endemic areas, the majority of patients are infectious to mosquitos before diagnosis and treatment of the symptomatic infection (22, 43). Nevertheless, if artemisinin-based combination therapies are indeed superior to chloroquine in preventing P. vivax transmission as this study suggests, this is an argument in favor of a unified treatment for all malarias (44), particularly if radical treatment is delayed or not given.

MATERIALS AND METHODS

Participants and sample collection

Patients with vivax malaria attending outpatient consultation at the clinics of the Shoklo Malaria Research Unit in Wang Pha and Maw Ker Tai (Northwest border of Thailand) were invited to participate in the study by giving a single 10-mL blood sample drawn into a sterile sodium heparin tube before receiving antimalarial drug treatment. The sample was kept into a Thermos bottle filled with water warmed at 37°C until processing (typically within 1 hour after collection).

In order to estimate parasite densities on admission, a thin smear and a thick film of participant blood sample were prepared on a glass slide, stained with 5% Giemsa for 35 minutes, and examined under a microscope at a 1,000 magnification using standard procedures (45), and a complete blood count was performed. The proportion of red blood cells infected with malaria parasites was estimated by recording the total parasite count in 2,000 red cells in the thin smear. If no parasite was detected in 2,000 red cells (3/38 samples), the count was determined for 500 white cells in the thick film. Then, gametocyte and asexual parasites were counted separately in a subset of 100 parasites, and the proportions were used to estimate gametocytemia and asexual parasitemia from the total parasite count per 2,000 red cells or per 500 white cells and the concentration of red cells or white cells in participant blood sample, as appropriate. All slides were read independently by two blinded microscopists, and discrepant results were resolved by a third microscopist. The mean values of the two concordant readings were used in the analysis.

Compounds

Chloroquine and artesunate were supplied by the WorldWide Antimalarial Resistance Network. Chloroquine diphosphate (Sigma-Aldrich, catalog no. C6628) stock solution was prepared at a concentration of 97 mmol/L in water. Artesunate (Sigma-Aldrich, catalog no. 88495-63-0) stock solution was prepared at a concentration of 52 mmol/L in 100% ethanol. Methylene blue (Poveblue, methylthioninium chloride trihydrate solution at 5 mg/mL or 13 mmol/L) was kindly provided by Provepharm (Marseille, France) and used as a stock solution. All stock solutions were kept at −80°C, used within 6 months, and thawed only once before being used in the assay.

Parasite handling

The blood sample was transferred into a 50-mL conical tube and centrifuged at 500 g for 5 minutes at 37°C. The serum and buffy coat were discarded, and the cell pellet was washed twice with 45 mL of incomplete culture medium warmed at 37°C using the same centrifugation conditions. Incomplete culture medium was composed of RPMI-1640 (Sigma-Aldrich, catalog no. R6504) supplemented with 2 g/L of NaHCO₃ (Sigma-Aldrich, catalog no. S6014), 5.7 g/L of HEPES (Sigma-Aldrich, catalog no. H4034), and 18 mg/L of hypoxanthine (Sigma-Aldrich, catalog no. H9636). The cell pellet was resuspended into a complete culture medium warmed at 37°C in a total volume of 20 mL. The complete culture medium was composed of incomplete medium supplemented with 10% of heat-inactivated AB serum. The serum was inactivated by heating at 56°C for 30 minutes, and aliquots were kept at −80°C and thawed only once before performing the assay. Eight culture flasks containing 8 mL of complete culture medium were warmed at 37°C without (control flasks, n = 4) or with a spike of the test drug (treated flasks, n = 4) and were inoculated with 2 mL of the blood cell suspension (total volume of 10 mL). The flasks were incubated with 5% CO₂ at 37°C for 24 hours. An additional wash-off step was added for assay runs carried out with artesunate to mimic the rapid elimination of this drug in vivo. After 4 hours, the contents of all flasks (both control and treated states) were transferred into 15-mL conical tubes and washed twice with 12 mL of complete culture medium using the same centrifugation conditions, then resuspended into 10 mL of complete culture medium, and then incubated for a further 20 hours.

Assay design, sample size, and power

Mosquitos from a laboratory-adapted colony of An. dirus were artificially infected with P. vivax by carrying out membrane-feeding experiments using the vivax malaria blood samples. The mosquito colony was maintained as described previously (46). Before the test feed, the blood specimen was incubated with artesunate (1 µmol/L for 4 hours, followed by 20 hours of incubation without drug), chloroquine (5 µmol/L for 24 hours), or methylene blue (1 µmol/L for 24 hours). The same specimen incubated without drug was used as the control. This design was chosen to maximize the power to detect a drug effect. In assay runs carried out with artesunate, all control and treated flasks were washed to control the effects of washing steps on sample infectiousness to mosquitos. Four technical replicates were performed for each group (treated and control), yielding eight mosquito batches per assay run. The artesunate and methylene blue concentrations each of 1 µmol/L were chosen to represent the high concentrations typically used for in vitro drug screening; chloroquine was tested at a concentration of 5 µmol/L because a concentration of 1 µmol/L did not exhibit evident transmission-blocking activity during preliminary experiments in the initial assay setup (data not shown). Drugs were assigned to blood samples in sequential order: the assay was repeated 18 times with chloroquine, 8 times with methylene blue, and 9 times with artesunate. The assay was then repeated three times with a different batch of chloroquine to exclude assessment bias relating to compound quality. The development of oocysts was assessed in samples of 50 mosquitos per batch 7 days after the feed, yielding a total of 450 oocyst counts per experiment: 50 for the baseline feed, 200 in the controls, and 200 in the treated replicates. Similarly, the development of sporozoites was assessed in samples of 5 mosquitos per batch 14 and 15 days after the feed (10 mosquitos per batch in total), yielding a total of 90 sporozoite counts per experiment: 10 for the baseline feed, 40 in the controls, and 40 in the treated replicates. The sporozoite count could not be determined in three assay runs because of the laboratory shutdown during a COVID-19 outbreak. To estimate the required sample size, a multi-level Bayesian model was fitted to a data set of oocyst counts in 97 artificial mosquito infections carried out at the same facility, and the model output was used to perform simulation experiments. Power to detect a 10% reduction in the mean oocyst count was calculated at varying numbers of dissected mosquitos per technical replicate, number of technical replicates per assay run, and number of independent assay runs. Using this power calculation, the study was powered to detect a 10% reduction in the mean oocyst count with seven independent assay runs for each drug.

Membrane-feeding assay

Membrane feeds were carried out as described previously with some modifications (46). At the end of incubation, the content of the flasks was transferred into 15-mL conical tubes and centrifuged at 500 g for 5 minutes at 37°C. The supernatant was discarded, and the cell pellet (approximately 500 µL) was resuspended into 500 µL of heat-inactivated AB serum warmed at 37°C. The suspension was then fed to the An. dirus mosquitos with a Hemotek membrane-feeding system (Blackburn, United Kingdom) using 1 mL reservoirs covered with stretched Parafilm (Bemis, USA). The assay was carried out with 5- to 7-day-old nulliparous female imagoes starved by removing the wet towel covering the cage and the sugar source for 4 to 6 hours before the feed. Mosquitos were transferred into 750 cm³ plastic containers at a density of 150 specimens per cup and left undisturbed for 30 minutes before the feed; eight cups were prepared in total (one per replicate), and the same mosquito batch was used for a given assay run. The feed was carried out by putting the Hemotek insert on top of the corresponding mosquito container and regularly blowing through the net every 5 minutes for 1 hour. Fully engorged mosquitos were transferred into 4,500 cm³ plastic containers at 15-minute intervals until 100 fully engorged mosquitos per replicate were collected (typically about 1 hour). Engorged mosquitos were kept at 25°C and provided with 10% sugar solution ad libitum until dissection.

Assessment of oocyst and sporozoite development

Oocyst and sporozoite counts in mosquitos were performed as described previously with some modifications (46, 47). Dissected mosquito midguts were stained with 2% mercurochrome solution for 5 minutes, observed under a microscope at a 40 magnification, and the number of oocysts per midgut was recorded. Pairs of salivary glands were crushed in 1 µL of 1× PBS using the corner of a glass slide. The crushed salivary glands were rinsed with 20 µL of 1× PBS. The mixture (approximately 15 uL) was then transferred into 1.5-mL plastic tubes and kept on wet ice until determination of the sporozoite concentration with a hemocytometer (typically within 4 hours after the dissection). If no sporozoite was detected in the hemocytometer, the dried slide was examined under a microscope at a 40 magnification to identify mosquito specimens that carried few sporozoites, below the detection limit of the hemocytometer. The sporozoite count in such specimens was arbitrarily set to 10 sporozoites per mosquito.

Data analysis

The proportion of Plasmodium-infected mosquitos was analyzed under a multi-level logistic regression model, including group allocation as a linear predictor and a random effect across participant blood samples to account for correlation in mosquito Plasmodium infection between experimental replicates of the same sample. The relative risk was then calculated using odds ratio estimate and proportion of infected specimens in the controls. Parasite count data (the number of oocysts and the sporozoites per mosquito) were analyzed under a Bayesian multi-level model, taking into account intra- and inter-experiment variability as per Medley et al. (27). In order to consider heterogeneity of Plasmodium development in the mosquito, the likelihood function was a Negative Binomial distribution parameterized by its mean μ and the dispersion κ for integer parasite counts y; y ~ Negative Binomial(μ, κ), with κ set as a function of the mean:

$$k=a_0\times {\mu }^{a1}.$$

Under the Negative Binomial model, the prevalence of infection P (oocyst or sporozoite index, defined as the number of Plasmodium-infected specimens divided by the number of dissected specimens) varies as a function of the mean infection intensity:

$$P\left(\mu ,\mathrm{K}\right)=1-{\left(1+\mu /\mathrm{k}\right)}^{-}k.$$

For each patient i and technical replicate k, the model-predicted mean log parasite count μ_i,k was expressed as the sum of a patient-dependent random effect ʎ_i and a batch random effect ʎ_k; ʎ_i ~ Normal(μ_i , σ_patient) and ʎ_k ~ Student-t(7, 0, σ_batch). The Student-t distribution with 7 degrees of freedom was chosen to accommodate the observed heterogeneity across batches (48). Thus, μ_i,k = ʎ_i + ʎ_k, where μ_i is the mean log parasite count in mosquito samples fed on blood from patient i, such as μ_i ~ Normal(μ_population, σ_patient), with μ_population being the mean log parasite count in mosquito samples fed on blood specimens from the overall patient population, σ_patient the standard deviation of individual patient mean log counts around the population mean, and σ_batch the standard deviation of batch effects.

For a given blood sample, the drug treatment effect in treated replicates β_T[i] was parameterized in the model as a proportional decrease in the mean number of counts on the log scale. Thus, the likelihood of the count data y_i,k,T[i] (patient i, technical replicate k, treatment assignment T[i]), given the parameters is

$${Y_{\mathrm{i},\mathrm{k},\mathrm{T}\left[\mathrm{i}\right] }}^{~}\mathrm{Negative} \mathrm{Binomial}\left(e^{\mu \left(i,k,\mathrm{T}\left[i\right]\right)},k_{i,k,T\left[i\right]}\right)$$

where μ_i,k,T[i] = ʎ_i + ʎ_k + β_T[i] + β_cov[i], and where κ_i,k,T[i] is a function of μ_i,k,T[i] as above. The additional model coefficients β_cov[i] account for the differences in experimental conditions for the artesunate samples (washing) and baseline characteristics of the sample (asexual parasitemia, gametocytemia, and oocyst count assessed on the day of sample collection, before incubation with or without the test drug). Continuous covariables were log transformed with a logarithm of base 10, meaning that a 10-fold increase in the covariable of interest was associated with a fold variation in the parasite count equal to the exponent of the coefficient estimate.

We used weakly informative priors to help computational convergence. These were μ_population ~ Normal (5, 5) and μ_population ~ Normal (5, 9) in the model fitted to oocyst and sporozoite data, respectively. The priors for other parameters were the same in both fits: σ_patient ~ zero-truncated Normal (1, 0.25), σ_batch ~ zero-truncated Normal (0.5, 0.25), log(a₀) ~ Normal (−1, 1), log(a₁) ~ Normal (−1, 1), β_T[i] ~ Normal (0, 1), and β_condition[i] ~ Normal (0, 1). The model was run with four independent chains each consisting of 4,000 iterations. Convergence of the chains was assessed by examining the values of effective sample size and Rhat and the traceplots (Appendix, Figure S3 to 6).

DATA AVAILABILITY

All analysis code and data are available via an accompanying GitHub repository: https://github.com/victorSMRU/transmission-blocking-plasmodium-vivax.

ETHICS APPROVAL

The study was approved by the Oxford Tropical Research Ethics Committee (reference 12-19), the Tak Public Health Office Ethics Committee (reference 23/2562), and the Tak Province Border Community Ethics Advisory Board (reference TCAB201901) (49). All participants provided their written informed consent to participate in the study.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/aac.00853-24.

Supplemental material. aac.00853-24-s0001.docx.

Table S1; Figures S1 to S6.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.