Concentration methods were first tested using P. falciparum gametocytes (NF54) that were cultured in a tipper system, ¹⁷ as previously described. ¹⁸ Mosquito feeding experiments with P. falciparum were performed on day 14 post seeding, by spiking pelleted cultured gametocytes at a concentration of 0.75 million/mL into 10 mL fresh heparinized blood (ref 367526, BD). P. vivax gametocytes were the primary focus of this study and obtained during the RaViCHMI1 clinical trial (ID NL-OMON57011) that will be published separately. Briefly, study participants were inoculated intravenously with the blood-stage P. vivax W1 challenge inoculum (PvW1), provided by the University of Oxford. ¹⁹ PvW1 is a clonal parasite line from Thailand that has been cryopreserved in 2021. ²⁰ P. vivax mosquito feeding experiments were conducted using blood samples from eight participants with a positive thick blood smear at a parasitemia between 2–38 parasites/μL between day 18 to 24 post infection (ID NL-OMON57011). Heparinized blood was obtained from study participants by venipuncture and immediately transferred to a water bath set at 37 °C until sample processing. All additional blood products required were supplied by the national blood bank (Sanquin Nijmegen, The Netherlands).

Anopheles stephensi mosquitoes of the Sind-Kasur strain were reared at 30 +/− 1 °C and ~ 80% relative humidity, in a room with a reversed light period (12 h shifts). ²¹ P. vivax infected mosquitoes received an extra uninfected blood-meal on day 6 post-infection to harmonize oocyst and sporozoite development ²² and were housed at 26 +/− 1 °C with ~80% relative humidity. For all membrane feeding assays An. stephensi mosquitoes were allowed to feed for 10 min in the dark on blood in water-jacketed membrane feeders. For the membrane feeding assay with un-enriched gametocytes, about 100 female mosquitoes between 1–3 days old were collected in cages and fed using midi glass feeders (approximately 1.2 mL). ²³ For gametocyte enriched feeds, about 50 female mosquitoes (1–3 days old) were collected in cups and fed using mini glass feeders that contained approximately ~300 μL material. ²³ Unfed and partially fed mosquitoes were removed. Fully fed mosquitoes were maintained on 5% glucose (D+ Glucose Monohydrate, ref 49159, Sigma) and dissected 5 to 8 days post feeding to count the number of developed oocysts on midguts by microscopy after 1% mercurochrome staining.

All steps until mosquito feeding, including centrifugation, were performed at 37 °C to prevent early activation of gametocytes and thereby a potential loss of infectivity. Blood meals were prepared with lithium heparinized blood (ref 367526, BD) for P. falciparum cultured gametocytes, or with blood obtained by venipuncture (again in lithium heparin tubes) from participants enrolled in a P. vivax CHMI. The non-enriched blood meal (control) for P. falciparum was prepared as previously described. ²⁴ For P. vivax-infected blood, plasma for the non-enriched feed was replaced with malaria-naive serum (group AB; Sanquin blood bank, The Netherlands).

Prior to gametocyte-enrichment, both P. falciparum cultured gametocytes and P. vivax gametocyte-infected in whole blood were leukodepleted by Plasmodipur (Europroxima) filtration. After flushing the filter with 2 mL pre-warmed RPMI medium, 10 mL blood material was gently filtered into a 15 mL tube. Tubes were spun for 5 min at 760  g after which plasma was removed without disturbing the top layer of erythrocytes. Blood was then brought to a hematocrit of 50% by resuspension with pre-warmed RPMI medium. ¹⁸

The MACS procedure was performed as previously described, ¹² with minor modifications. Briefly, columns were attached to the QuadroMACS™ separator (Miltenyi Biotech) in a 37 °C incubator, a hypodermic needle (25G) was attached, and the column was equilibrated with 1 mL pre-warmed RPMI medium. Subsequently, 3 mL of blood diluted in RPMI (50% hematocrit) was loaded onto the column; this was repeated on the same column until a total volume of 10 mL blood was processed. Columns were subsequently washed two times with 1 mL RPMI medium to remove unbound erythrocytes (referred to as wash 1 and wash 2). Columns were then detached from the magnet and placed in 15 mL tubes. Twice 1 mL RPMI medium was added to collect the bound fraction; the plunger that was provided with the column was used after the second 1 mL medium addition to expel all the retained liquid. Collection tubes were spun at 760  g for 1 min. The supernatant was carefully removed and approximately 1 mL was left on top of the pellet.

Percoll density gradient centrifugation was performed as previously described, ¹³ with minor modifications. A 90% isotonic Percoll (SIP) solution (Cytiva, 17089101) was prepared with PBS (10x) on the day of use. The SIP solution was subsequently used to prepare 65% and 70% Percoll working solutions (v/v) in PBS (1x). 6 mL aliquots were prepared in 15 mL tubes and pre-warmed at 37 °C. The 10 mL blood sample was divided over 3 Percoll tubes as follows: a maximum of 3.3 mL blood (50% hematocrit) was gently layered via the wall of the tube onto the Percoll, using a 3 mL syringe with blunt needle. Tubes were spun at 760  g or 1500  g for 15 min at 37 °C without break. A 3 mL syringe with blunt needle was used to collect and transfer the layer on top of the gradient to a new 15 mL tube. Samples were topped up to 12 mL with pre-warmed RPMI and spun at 760  g for 5 min at 37 °C. The supernatant was carefully removed without disturbing the pellet, and the washing procedure was repeated two times. After the final wash, 0.5 mL of supernatant was left on top of the pellet. Subsequently, pellets from three tubes (from one study participant) were resuspended and combined.

The resuspended pellet obtained by MACS or DGC was transferred to a 1.5 mL Eppendorf tube in a heating block set at 37 °C. 180 μL of pelleted erythrocytes from a universal donor (for P. falciparum) or from the same clinical trial participant were added. Tubes were spun in a benchtop centrifuge for 20 s at 10,000 rpm. The supernatant was very carefully removed, and the pellet was resuspended in 150 μL serum (group AB). The complete protocols can be found at protocol.io (Extended data).

A subset of mosquitoes was dissected between day 11 and 20 post-blood feeding to collect the salivary glands, collecting 2 glands with 3 lobes each, into a glass pestle grinder containing 1 mL PBS (1x). After tissue homogenization, 10 μL from the suspension was loaded into a Bürker-Türk counting chamber that was left for 20 min at RT to allow sporozoites to settle. Sporozoites were counted using a phase contrast light microscope with a final magnification of 400x.

To quantify P. vivax parasites in material collected during MACS processing, 100 μL sample was collected in a 1.5 mL Eppendorf tube and mixed with 400 μL of RNA protect Cell Reagent (Qiagen). Samples were stored at −20 °C and extracted within 10 weeks post-collection. Total nucleic acid (NA) was extracted with the automated MagNAPure 96 instrument (Roche) using the MagNA Pure 96 DNA and Viral NA Large Volume Kit (Roche, Basel, Switzerland) and eluted in 50 μL. The number of P. vivax parasites was quantified by 18S rRNA quantitative PCR (qPCR) as described previously, ²⁵ using the forward- 5′-GCTTTGTAATTGGAATGATGGGAAT-3′ and reverse primer 5′- ATGCGCACAAAGTCGATACGAAG −3′. A probe was used for amplicon detection 6FAM-AGCAACGCTTCTAGCTTA-MGBNFQ. P. vivax gametocytes were quantified by qRT-PCR, amplifying pvs25 as described previously, ²⁵ using the forward- 5′- ACACTTGTGTGCTTGATGTATGTC -3′ and reverse primer 5′- ACTTTGCCAATAGCACATGAGCAA −3′ set. A probe was used for amplicon detection 6FAM-TGCATTGTTGAGTACCTCTCGGAA-BHQ1. The number of parasite copies was calculated with a standard curve, obtained using 10-fold dilutions of plasmids as previously described by Tadesse et al. ²⁶ The MACS-column binding efficiency (%) was calculated by dividing copies in the bound fraction against the the total number of copies in the sample (total copies in bound + flow-through + wash 1 and 2 fractions).

Statistical analyses were performed using R (version 4.3.2). For oocyst density, values were analysed on the log10 scale. We fitted a linear model with condition (direct vs each enrichment method) as a categorical predictor and tested each enrichment method against the direct control using planned pairwise contrasts, reporting estimated mean differences with 95% confidence intervals and two-sided p-values. For ease of interpretation, these estimated differences on the log10 scale were back-transformed to mean fold changes (ratios) by exponentiation, yielding the enrichment-to-control oocyst density ratio. We analyzed the percentage infected mosquitoes using a linear model with condition as predictor, again testing each enrichment method versus the direct non-enriched control. Data are presented as violin plots with 95% confidence intervals. Correlations in proportion of infected mosquitoes and oocyst density between control and enrichment methods were visualized as scatter plots with dashed lines indicating a 1:1 reference. The relationship between oocyst prevalence and oocyst density was assessed using a logistic regression model, and 95% confidence intervals were calculated for the model fit. The significance level was set at a two-tailed p < 0.05.

The enrichment protocols were first examined for P. falciparum gametocytes. Methods tested were MACS with LS columns and two DGC gradients with two centrifugation speeds: Percoll 65% ¹⁴ (P65) and 70% ¹⁵ (P70) spun at a centrifugation speed of 760  g ¹¹ or 1500  g ¹⁴ ( Figure 1). All gametocyte enrichment methods increased the oocyst density and the proportion of infected mosquitoes ( Figure 2A, B) relative to the non-enriched sample (control). Mean oocyst densities were as follows: control, 3.0 (range 0.6–6.2); MACS, 11.0 (range 4.9–21.2); DGC65-760  g 16.1 (7.9–29.4); DGC65–1500  g, 10.0 (6.3–16.1); DGC70–760  g, 17.4 (11.2–23.6); DGC70–1500  g, 21.2 (18.6–23.8) ( Extended data). Mean infection rates were control, 66% (76/116); MACS, 88% (105/119); DGC65-760  g, 98% (55/56); DGC65-1500  g, 85% (51/60); DGC70-760  g, 90% (36/40); DGC70-1500  g, 97% (36/37) ( Extended data). DGC70 enrichment resulted in the highest oocyst densities ( Figure 2B) compared with control. DGC70-760  g and 1500  g increased oocyst density 7.5-fold (95% CI 2.0–29.1; p = 0.007) and 9.8-fold (95% CI 2.5–37.6; p = 0.004), respectively. The proportion of mosquitoes that became infected similarly increased with absolute increases of 24.8% (95% CI −1.8 to 51.4; p = 0.06) and 31.9% (95% CI 5.2–58.5; p = 0.02) for DGC-760  g and 1500  g, respectively ( Figure 2A). DGC65–760  g achieved a 6.3-fold increase in oocyst density compared with control (95% CI 1.9–21.1; p = 0.006) ( Figure 2B). This was accompanied by an absolute increase in the proportion of infected mosquitoes of 33.1% (95% CI 9.3–57.0; p = 0.01) ( Figure 2A). In comparison, DGC65–1500  g and MACS resulted in smaller increases in oocyst density of 4.3-fold (95% CI 1.3–14.2; p = 0.02) and 4.4-fold (95% CI 1.3–14.8; p = 0.02), respectively. Their effects on the proportion of infected mosquitoes were comparable to those observed with DGC70–760  g, with absolute increases of 19.8% (95% CI 4.0–43.6; p = 0.09) for DGC65–1500  g and 23.1% (95% CI −0.7 to 46.9; p = 0.06) for MACS.

Although enrichment of P. falciparum gametocytes with DGC70 was achieved in two experiments, blood occasionally failed to penetrate the gradient after centrifugation (data not shown). Because this was observed repeatedly, independent of centrifugation speed, this condition was not considered in further evaluations. Both DGC65 and MACS were further assessed for their applicability in enriching P. vivax gametocytes.

Next, we evaluated P. vivax gametocyte enrichment by DGC65 (spun at 760  g or 1500  g) or MACS (LS columns); both were examined relative to the non-enriched control ( Figure 1). Blood was drawn from eight study participants in a P. vivax blood-stage CHMI. Paired observations were collected for un-enriched versus enriched gametocytes with a total of 27 observations for DGC65-760  g, 28 for DGC65-1500  g, and 8 for MACS. For each condition, 11 mosquitoes were dissected to determine oocyst counts. DGC65-1500 g achieved the highest oocyst density and proportion of infected mosquitoes ( Figure 3A, B). Compared with control, DGC65-1500 g increased the oocyst density 86-fold (95% CI 14–525; p < 0.0001) and increased the proportion of infected mosquitoes by 58.3% (95% CI 42.6–74.0; p < 0.0001). More modest effects were observed with DGC65-760  g that increased oocyst density 11-fold (95% CI 2.9–42.6; p < 0.001) and the proportion of infected mosquitoes by 28.0% (95% CI 12.1–43.9; p < 0.001) ( Figure 3A, B). In contrast to all DGC conditions, MACS enrichment did not improve P. vivax mosquito infection, showing no increase in oocyst density (density ratio 0.68; 95% CI 0.09–5.28; p = 0.71) or in the proportion of infected mosquitoes (−5.4%; 95% CI −29.6 to 18.9; p = 0.66) compared to control ( Figure 3A, B).

Gametocyte enrichment using DGC65 improved transmission significantly ( Figure 3C). This was most apparent when mosquitoes were infected with blood containing low parasite densities, with 567 parasites/mL being the lowest parasitemia at which we observed transmission. DGC at a speed of 1500  g was superior and resulted in a higher oocyst intensity compared to 760  g ( Figure 3D).

A subset of mosquitoes that fed on DGC65-1500  g gametocyte enriched blood were salivary gland dissected between day 11 and 20 post-blood feeding for quantification of sporozoites. Mean sporozoite load increased approximately proportionally with oocyst density in a weighted log–log regression model, adjusting for days post-infection relative to day 14 (β = 0.91, 95% CI [0.51, 1.30], p = 0.0003). The model-predicted relationship between oocyst density and sporozoite load shows that oocyst burden and sporozoite load were strongly associated ( Figure 4).

A subset of 7 Percoll-enriched gametocyte samples was processed with and without filtration to explore whether leukocyte depletion by Plasmodipur filtration is an essential component to increase P. vivax transmission. For the proportion of infected mosquitoes, absolute differences between filtered and unfiltered conditions were small and not statistically significant: 9.09% (95% CI: −47.30 to 65.49; p > 0.99) for 1500  g, and − 0.00% (95% CI: −89.17 to 89.17; p > 0.99) for 760  g. Differences in oocyst intensity were similarly negligible: 0.29 (95% CI: −1.00 to 1.57; p = 0.96) for 1500  g, and 0.16 (95% CI: −1.87 to 2.19; p = 0.96) for 760  g.

As detailed above, P. vivax gametocyte enrichment by MACS did not result in increased mosquito infection prevalence and intensity compared to the control conditions. To explore whether this was due to inefficient column binding, we molecularly quantified enriched parasites. Blood from two study participants was processed in parallel on LS and LD columns and the following fractions were collected: starting, flow-through, wash 1, wash 2, and bound (as described elsewhere ¹¹ ). Parasites were quantified by qPCR targeting 18S (total parasites) and Pv25 (gametocytes) transcripts. ²⁴ The proportion of P. vivax gametocytes in the bound fraction was 0 and 0.1% when using LS columns and 17.7 and 33.6% when using LD columns ( Extended data).