Blood samples were collected at the Fundação de Medicina Tropical Dr. Heitor Vieira Dourado (FMT-HVD), located in the city of Manaus, Amazonas state, Brazil. 107 adults’ patients (aged >18 years) with P. vivax mono-infection, microscopically diagnosed with parasitemia higher than or equal 1,000 parasites/µL and no current antimalarial treatment were included in the study. Before the treatment, a sample of 9 mL of peripheral blood was withdrawn using heparin-coated vacutainer tubes. Patients were treated following the Brazilian Health Ministry malaria treatment guidelines (Ministério da Saúde do Brasil, 2021).

An. aquasalis were reared in a well-established colony in the Medical Entomology Laboratory at FMT-HVD. Colonies were kept at 24–26 °C and relative humidity of 70–80% on a 12:12 light-dark cycle. Larvae were hatched at room temperature in water containing salt at a final concentration of 2g/L, and a fish food TetraMin^® (Madison, Wisconsin, USA) was provided daily. Larvae were allowed to pupate and become adults in an enclosed mesh-covered cage with water provided and fed ad libitum with a 10% sucrose solution until two days before being given the infective blood meals (Baia-da-Silva et al., 2018).

An aliquot of total blood was reserved to direct membrane feeding assay (DMFA) (control group). The percentage of P. vivax gametocytes-were determined by counting the blood smears stained with Giemsa. The blood was immediately processed to obtain enriched P. vivax - parasitized red blood cells (Pv-pRBC). Blood was centrifuged, separating plasma from RBCs. Plasma was stored at 4°C for rosette assay. The RBC were washed twice with RPMI-1640 medium (Sigma-Aldrich^®, USA) and filtered in a cellulose column (Sigma-Aldrich^®, USA) to remove leukocytes and platelets as previously described (Russell et al., 2011). A Percoll (GE Healthcare, USA) cushion (45%) was performed to separate the mature asexual parasites and gametocytes from the younger asexual forms and non-infected erythrocytes as previously described (Vera et al., 2015). The total number of gametocytes (gPv-pRBC) after the purification process was determined by counting the percentage of gametocytes in thin blood smears from the Percoll interface and determining the quantity of cells per ml using a Neubauer Chamber.

In parallel to the DMFA (next section) gPv-pRBC were adjusted to 10% parasitemia and 5% hematocrit in a rosetting medium (McCoy5A medium supplemented with 20% native autologous plasma); after that, 50 µL of this suspension was plated in a 96 well plate (u- bottom), followed by incubation for 40 minutes at 37°C. Then, they were stained with 50µL of acridine orange (50 µg/mL) and incubated for 10 minutes at 37°C. Subsequently, one drop of the stained suspension was pipetted onto a glass slide and immediately covered with a glass cover slip. Finally, the rosette formation was examined by direct light and fluorescence microscopy (Nikon Eclipse 50i, filter 96311 B-2E/C) (Udomsanpetch et al., 1995). The proportion of gPv-pRBC in rosettes was measured by counting 100 gPv-pRBC in each experiment in duplicate. A rosette was determined by the adhesion of two or more uninfected erythrocytes to a gPv-pRBC ( Supplementary Figure S1 ). Giant erythrocyte rosettes surrounding a P. vivax infected cell were also counted (Marin-Menendez et al., 2013). The isolates were classified into rosette (rosette formation rate higher than or equal to 10%) and non-rosette-forming (rosette forming rate less than 10%).

After the incubation to evaluate the rosette formation, the cell suspension from 4 wells of the 96 well plates was collected in a 1.5 mL microtube per group (200 µL in total). One microtube were maintained in static condition at room temperature (intact rosettes) and the other was exposed to constant vortex agitation for 5 min after completely disrupting rosettes using a needle (disrupted rosettes). Subsequently, membrane feeding assay was performed with these samples (intact and disrupted rosettes) following the parameters described in P. vivax DMFA.

The DMFA was conducted according Fabbri et al., 2021 with few modifications. Three different sets of DMFA experiments were performed. The first one was performed in order to verify if the procedure of parasite purification interfere with An. aquasalis infection, and therefore DMFA were performed comparing two groups: total blood and gPv-pRBC. For this first set of experiments, 500 µL of the total blood were separated and maintained at 37 °C until be used in the DMFA. The remaining blood was processed for gametocyte purification as described above. The pellet of cells obtained was suspended in non-infected O-type blood and AB serum to a hematocrit of 40%. Both suspensions were given to fed mosquitos in the DMFA.

The second set of experiments was performed using 47 isolates enriched to gPv-pRBC to verify the impact of P. vivax rosette at Anopheles infection. For this, part of the purified sample was used to set the rosetting assay as described above and around 1x10⁵ gPv-pRBC were suspended in non-infect O-type blood and autologous plasma to a hematocrit of 40% and given to fed mosquitoes immediately.

The last set of experiments was performed to evaluate the rosette effect in Anopheles infection in the same isolate and for this purpose a mechanical rupture of rosettes were performed. Briefly, after the identification of positive samples for the rosette phenomenon (rosette formation rate higher than or equal to 10%), these were divided into two aliquots for mosquito feeding. One aliquot was kept at room temperature in static conditions (intact rosettes), and the other was subjected to mechanical rupture (disrupted rosettes). After that, the samples were centrifuged and suspended in non-infect O-type blood and autologous plasma to a hematocrit of 40% and given to fed mosquitoes immediately.

The blood suspension in different conditions was offered to groups of 120-150 An. aquasalis females between 90-120 minutes through membrane feeders at 37 °C. After the blood-feeding experiments, only fully engorged mosquitoes were transferred to rearing containers; they were maintained in the insectary at 26 °C, 70–80% relative humidity, and fed daily on a 10% sugar solution. Seven days post-infection, the surviving mosquitoes from each experimental group were dissected. The midguts were stained with 2% commercial Mercurochrome (Merbromin, Sigma- Aldrich, USA), placed under a cover glass, and examined for the presence of oocysts in an optical microscope. The infection rate (percentage of infected mosquito midguts) and the infection intensity (mean number of oocysts per midgut) on each mosquito were recorded and compared among the experimental groups.

Shapiro-Wilk test was evaluated to verify the normality of the distribution. A non-parametric Kruskal-Wallis test or Wilcoxon test were used for infection intensity analysis and for the infection rate, Mann Whitney parametric distribution. Correlations were analyzed using Pearson test. Statistical significance was defined as p < 0.05. All statistical analyses were conducted using GraphPad Prism version 8 software (GraphPad Software, Inc., San Diego, CA). The statistical tests applied are indicated in the figure’s caption, and (n) means the number of samples analyzed.

The P. vivax gametocytes’ capacity to develop rosetting was assessed in 107 isolates obtained from patients with non-severe vivax malaria blood. The rosetting assay was developed with native autologous plasma, and 77.5% (83/107) of the tested samples presented rosette formation capacity ( Figure 1A ). The percentage of rosettes in gPv-pRBC in the samples evaluated presented a mean of 12.97% ± 13.61, ranging from 0 to 50% ( Figure 1B ). Additionally, we observed that the average of gPv-pRBC in rosette formation was higher in isolates from patients who had contracted malaria previously than in the primo infection ones (14.08 and 10.59% respectively); however, it was not statistically significant (Student’s t-test, p = 0.1091, n = 48) ( Figure 1C ). Moreover, no correlation between the number of previous episodes and rosettes rate was observed (Spearman’s test, p = 0.749; Pearson test) ( Figure 1D ). Finally, correlation analysis with the hematocrit, parasitemia and gametocytemia levels of the patients, showed no correlation with the rosetting capacity of the isolates analyzed in the present study. ( Supplementary Figures S2A–C ).

Initially, we evaluated whether the purification process of gPv-pRBC could affect mosquito infectivity to verify if this process does not interfere with the parasites’ ability to infect the mosquitoes. No changes were observed in the An. aquasalis infection rate (p = 0.0986; Student’s t-test) and in the infection intensity (p = 0.5075; Mann Whitney test) ( Supplementary Figure S3 ). Subsequently, among the 107 samples assessed for rosetting ability, 47 were used to evaluate the role of rosetting in Anopheles aquasalis infection by DMFA. Twenty-nine isolates of gPv-pRBC (61.7%) were rosette producers and showed 10 to 50% of rosette formation rates. In these isolates, infection rate was 49.25%, while the non-rosette forming isolates (n=18) had a 33.98% (p = 0.0114; Student’s t-test) ( Figure 2A ). Moreover, a positive correlation between the rosette and infection rate was found (p = 0.001; Pearson test) ( Figure 2B ). However, no difference between rosette forming and non-rosette forming group in infection intensity was observed ( Figure 2C ) but a positive correlation between the rosette rate and intensity of infection for these isolates was evidenced (p = 0.0387; Pearson test) (( Figure 2D ).

Once rosetting seems to be correlated with infection rate and intensity, 23 rosette rupture assays were performed. The mechanical rosette rupture process was effective, showing a decrease in rosette rate, which ranged from 0 to 5% after the procedure; and it was significantly different compared to the same untreated sample (p < 0.0001, Wilcoxon paired test) ( Figure 3A ). Therefore, the investigation of the ability to infect mosquitoes were performed by DMFA comparing the two groups: intact and disrupted rosettes. It was observed that all isolates assessed were able to infect An. aquasalis, presenting a mean infection rate of 37.6%, ranging from 7.14 to 72.2%, and a mean infection intensity of 28.67 ± 33.26, ranging from 0.30 to 32.42. The paired comparison between groups showed that the infection rate was higher in the intact (37.95 ± 19.20) than in the ruptured (26.23 ± 17.73) rosette groups (p < 0.0001; Wilcoxon test) ( Figure 3B ). Moreover, the infection intensity was also higher in the intact rosettes (4.49 ± 7.83) than in the rupture group (2.93 ± 4.91) (p= 0.0003; Wilcoxon paired test) ( Figure 3C ).