Rosettes integrity protects Plasmodium vivax of being phagocytized

Plasmodium vivax is the most prevalent cause of malaria outside of Africa. P. vivax biology and pathogenesis are still poorly understood. The role of one highly occurring phenotype in particular where infected reticulocytes cytoadhere to noninfected normocytes, forming rosettes, remains unknown. Here, using a range of ex vivo approaches, we showed that P. vivax rosetting rates were enhanced by plasma of infected patients and that total immunoglobulin M levels correlated with rosetting frequency. Moreover, rosetting rates were also correlated with parasitemia, IL-6 and IL-10 levels in infected patients. Transcriptomic analysis of peripheral leukocytes from P. vivax-infected patients with low or moderated rosetting rates identified differentially expressed genes related to human host phagocytosis pathway. In addition, phagocytosis assay showed that rosetting parasites were less phagocyted. Collectively, these results showed that rosette formation plays a role in host immune response by hampering leukocyte phagocytosis. Thus, these findings suggest that rosetting could be an effective P. vivax immune evasion strategy.


Scientific RepoRtS
| (2020) 10:16706 | https://doi.org/10.1038/s41598-020-73713-w www.nature.com/scientificreports/ In the present study, sets of freshly P. vivax isolates were collected from Manaus, Brazil, and a variety of controlled ex vivo experiments were performed to mechanistically study host and parasite factors that may contribute to P. vivax rosetting.

Results
Plasmodium vivax rosettes are enhanced by plasma from infected patients. Rosetting was assessed in a total of 81 P. vivax isolates and all samples formed rosettes at some extended in presence of autologous plasma. Rosettes were evaluated at late stages (late trophozoites/early schizont stages, 30-44 h post infection), and a mean rosetting of 38.78% ± 27.66 per isolate was found with a range from 1.30 to 95.51% (Supplementary Fig. S1). Rosettes were also evaluated in presence of native autologous plasma and heat-inactivated autologous plasma and it was found to be higher in the first condition (Student's t test, p < 0.05, n = 23) (Fig. 1A). www.nature.com/scientificreports/ When comparing rosetting in the presence of autologous or a nonimmune heterologous plasma from a donor without history of malaria and no detectable antibodies to PvAMA-1 or PvMSP-1, rosetting rates were higher in the first group (Fig. 1B). Few rosettes were observed in presence of heterologous nonimmune plasma ranging from non-rosettes to 28.1% rosetting (Fig. 1B). Surprisingly, when these isolates were evaluated in the presence of autologous plasma, rosetting was enhanced by eightfold (Student's t-test, p < 0.05, n = 20) (Fig. 1B). However, no significant differences were observed when a subset of isolates were assessed for rosetting in the presence of a heterologous immune plasma from a highly rosetting isolate (rosetting rate of 91.1%) (Student's t-test, p = 0.95, n = 9) (Fig. 1C). Rosettes were not formed in plasma free medium containing Albumax II (Wilcoxon test, p < 0.05, n = 6), indicating that plasma components are important (Fig. 1D). Hematological parameters as platelet levels and hematocrit were also evaluated and it was found to be independent of rosetting rates (Person correlation coefficient, Supplementary Fig. S2A,B).
Rosetting correlates with total immunoglobulin M but not IgG in patients infected with P. vivax. Immunoglobulins are among the plasma components that influence rosetting in P. falciparum 11 .
Genes related to phagocytosis pathway are differentially expressed in peripheral blood mononuclear cells of patients infected with P. vivax forming rosettes. Transcriptome analysis was performed on peripheral leukocytes from P. vivax infected patients with very low or moderate rosetting (Table 1). Five genes that showed a log2 fold change higher than 2 were identified (Table 2, Fig. 5). Three out of five genes were in the Fc gamma receptor (FCGR)-dependent phagocytosis pathway. Immunoglobulin kappa constant (IGKC) and immunoglobulin heavy constant gamma 1 (IGHG1) were upregulated and actin-related protein 2/3 complex subunit 2 (ARPC2) was downregulated in individuals with moderate rosetting compared to patients with low rosetting.

Rosetting protects P. vivax parasites from phagocytosis. Because transcriptional analysis indi-
cated that a phagocytosis pathway may be affected by rosetting, functional assays were conducted to investigate the role of this adhesive phenomenon. Phagocytosis assay was performed using THP-1 human monocytic cell line and P. vivax rosette isolates at two different conditions: intact or mechanically disrupted. For THP-1 cells the higher phagocytosis index was observed when rosettes were mechanically disrupted. The mean difference of phagocytosis index was − 11.8 (95% confidence interval (CI) − 20.2; − 5.6) (Wilcoxon test, p < 0.05, n = 7), (Fig. 6A). To confirm these results, phagocytosis assays were also performed with peripheral blood mononuclear cells (PBMCs) isolated from individuals infected with P. vivax. As observed to THP-1, phagocytosis index for PBMC were also higher when rosettes were mechanically disrupted. The mean difference of phagocytosis index was − 10.8 (95% CI − 16.3; − 6.2) (Wilcoxon test, p < 0.05, n = 9) (Fig. 6B).

Discussion
Although it has been more than two decades that rosetting in P. vivax was reported 7 , its role in parasite pathobiology remains unknown. The present study showed that P. vivax are less phagocyted when forming rosettes rather than when these rosettes are disrupted. The results suggest rosetting as an evasion mechanism. The ability of all P. vivax isolates to form rosettes at late stages suggested an important role of this adhesion phenomenon. Rosetting was revealed to be a gain for the parasite as confirmed by the positive correlation between rosetting and parasitemia. While the data presented here suggest that rosetting provides a biological advantage to P. vivax, the nature of this mechanism is not completely elucidated.
Rosettes does not facilitate merozoite invasion since it was showed by Lee et al. that P. vivax rosettes are preferentially formed by mature red blood cells which are refractory to P. vivax merozoite invasion 9 . It has also been speculated that rosetting may protects P. vivax from being eliminated by the host immune system, however no data has been provided to support this hypothesis 9,13 . Here, experimental data supported this latter idea that P. vivax uses rosetting to avoid a key aspect of the host immune response.
In malaria caused by P. falciparum, the rosetting success is highly dependent on plasma components 11,14 . In the present study, P. vivax parasites did not form rosettes in the presence of Albumax II and rosetting rate was eight times higher when autologous plasma was present. Rosetting rates were similar when in presence of autologous plasma or when in presence of a heterologous immune plasma, from an infected donor with high rosetting isolate, suggesting that a component, yet to be identified, in infected patients could favour P. vivax rosetting. IgM levels in plasma from P. vivax infected patients correlated with rosetting frequency. In falciparum malaria, some but not all rosetting strains are dependent on IgM 11 . A potential role of IgM in falciparum rosetting is to stabilize rosettes, acting as a bridge between infected and noninfected cells 15 . IgM may also assist P. falciparum in the host immune evasion as natural IgM binding to iRBCs have been shown to mask IgG epitopes 16 , thus limiting opsonization and phagocytosis. Although the present data suggested that IgM is also important in P. vivax rosettes the role of Table 2. RNAseq analysis of PBMCs from patients infected with P. vivax. Graph of log2 fold change of genes from PBMC of P. vivax infected patients between moderate and low rosetting isolates data analysis. Gene lists using p-value < 0.05, q-value < 0.3 and log2 (fold change) > 2 cut-offs obtained from RNAseq differential gene expression analysis.  Here, the correlation between hematocrit and rosetting was not observed. P. vivax rosetting has been previously associated with anemia during pregnancy 21 . Even though the mechanisms that are driving anemia in vivax malaria are not completely elucidate, there are evidences that the presence of autoantibodies are increased during acute P. vivax infection [22][23][24][25] .
A cytokine storm is frequently observed in vivax malaria which is a result of an excessive activation of innate immune cells 26 . Elevated levels of pro-inflammatory cytokines have been reported in acute P. vivax infection 27,28 . Besides that, the anti-inflamatory cytokine IL-10 has been also detected at high concentration in symptomatic malaria 29,30 . In this current study, rosetting was correlated with IL-6 and IL-10 levels, however the mechanism behind cytokine-aided P. vivax rosetting remains unknown. One of the attributes of IL-6 is to promote early production of parasite-specific IgM antibodies 31 . Here, no direct association between IL-6 and IgM levels was detected, but both parameters were correlated with rosetting. IL-10 levels have been correlated with higher parasitemia in vivax malaria 32,33 , whereas asymptomatic patients have lower levels of IL- 10 34 . No direct association between IL-10 and parasitemia was found. However, parasitemia and IL-10 correlated with rosetting frequency. IL-10 has also been associated with previous malaria infection 12,29 , which could indicate that rosetting is also more frequent in individuals that were previously infected.
The host transcriptomic profile directly compared patient isolates with low versus moderate rosetting, which were quantified immediately after blood sample collection, when most parasites are young 6 . Rosetting is mostly formed by mature parasites, this constituted an experimental limitation of the present study. Additionally, a low quantity of host sample input effectively transcribed resulted in few statistically significant pools of differentially expressed genes. Nevertheless, three out of five genes showing more than twofold expression change are annotated as part of Fc gamma phagocytic pathway. Two of the genes up regulated in the moderate rosetting group are related to antibody production. These results suggest that patients with higher rosetting are producing higher amounts of antibodies but as phagocytosis assays indicate, rosettes are protecting parasites of being phagocyted. Besides that, one of the genes differentially expressed was the actin related protein (ARPC2), which is important for the remodelling of macrophages membrane during phagocytosis 35 . This gene was down regulated in moderate rosetting isolates, indicating that phagocytosis might be compromised in these individuals.
Finally, parasites that were in a rosetting formation were less likely phagocytized. When rosettes were mechanically disrupted, rates of phagocytosis were increased more than twofolds. The most likely explanation for this observation is that the noninfected erythrocytes in the rosette formation shielded the iRBCs from antibodies and provided a physical barrier to restrict contact with phagocytes and other immune effector cells.
Taken together, these data suggest that vivax malaria rosetting is an evasion mechanism that allows the parasite to escape from the host immune system. The fact that rosetting is a frequent feature in P. vivax late stages www.nature.com/scientificreports/ indicates that this phenotype could be an advantage for the parasite by conferring significant protection from the host immune system. Therefore, the understanding of P. vivax rosettes may help in the development of new strategies for malaria control.

Material and methods
Ethics statement. All  Parasite isolation and enrichment. Once microscopic diagnosis of uncomplicated vivax malaria was made and before the treatment was initiated, 8 mL of blood was collected into citrate-coated Vacutainer tubes (BD). Blood was immediately processed to obtain enriched Pv-iRBCs. Immediately after collection, the RBCs containing trophozoites and schizonts were separated from the younger forms on a 45% Percoll (GE Healthcare) gradient as previously described 5 .
Rosetting assay. IRBCs (20 µL) at 2.5-5% parasitemia and 2.5-5% hematocrit were initially washed three times with McCoy's 5A medium (400 g, 5 min., at room temperature). After that, iRBC were incubated for 40 min at 37 °C in rosetting medium (McCoy's 5A medium supplemented with 20% of autologous plasma). Duplicated samples were stained with 45 µg/ml acridine orange and examined by direct light and fluorescence microscopy (Nikon Eclipse 50i, filter 96311 B-2E/C). Rosetting was assessed by counting 200 iRBCs, in duplicate. Slides were counted at diagonal vision, to balance for a possible irregular distribution of rosettes in the slide. A rosette was determined by the binding of two or more uninfected erythrocytes to an iRBC ( Supplementary  Fig. S5). To assess the involvement of plasma factors in the rosette formation, the plasma in rosetting medium was substituted for 0,5% of Albumax II. For heat-inactivated plasma, samples were heat inactivated for 30 min at 56 °C. Nonimmune heterologous plasma evaluated was from a donor without history of malaria and no detectable antibodies to PvAMA-1 or PvMSP-1 (reactivity index < 1).

Measurements of plasma cytokines and IgG and IgM levels.
The levels of IL-2, IL-4, IL-6, IL-10 TNF-α and IFN-γ were quantified in cryopreserved plasma samples using the Cytometric Bead Array system (CBA; BD Biosciences) following the manufacturer's instructions. Samples were analyzed using a BD FACSCalibur (BD Biosciences, USA). Standard curves were derived from cytokine standards for each cytokine analyzed. Total IgG and IgM were measured from plasma using an ELISA quantitation kit (Bethyl Laboratories) following the manufacturer's recommendations.
IgG and IgM red blood cell binding. Briefly, erythrocytes (A+) at 0.8% hematocrit in RPMI medium were incubated with plasma from P. vivax-infected patients at 1:10 dilution for 1 h. After three washing steps with RPMI, erythrocytes were incubated for 30 min with anti-IgG or anti-IgM Alexa Fluor 488 conjugate (Ther-moFisher Scientific). Erythrocytes were washed thrice and suspended in PBS solution, and 100.000 events were acquired using a BD FACSCanto II (BD Biosciences, USA). Flow cytometry results were analyzed using FlowJo software.
Naturally acquired IgG antibody subclasses of antibodies to PvAMA-1 and PvMSP-1 19 merozoite antigens. Specific IgG subclasses of antibodies (IgG1, IgG2, IgG3 and IgG4) to PvAMA-1 and PvMSP-1 19 were detected in plasma by enzyme-linked immunosorbent assays (ELISAs). PvAMA-1 and PvMSP-1 19 recombinant proteins were expressed and purified as previously described 36,37 . ELISAs were conducted as previously described 38 . The results for IgG subclasses were expressed as reactivity indices (RIs), which were calculated by dividing the mean optical density (OD) values of tested samples by the mean OD values plus three standard deviations of 20 non-exposed individuals living in nonendemic areas of malaria. RI values > 1 were considered positive.
RNAseq of peripheral blood mononuclear cells. mRNA from PBMCs isolated from total blood of patients infected with P. vivax with high (n = 2) or low (n = 2) rosetting was analyzed by RNAseq. For this specific approach, moderate rosetting isolates were considered as 20% rosettes, and low rosetting isolates were considered as less than 10% rosettes. RNA  phagocytosis assay with P. vivax. THP-1 cells were grown on 8-well culture slides. Maturation was induced by incubation with 60 ng/mL phorbol 12-myristate 13-acetate (PMA) (CalbiochemH, San Diego, CA) for 24 h at 37 °C. The supernatant was then removed, and cells were washed twice with cell culture medium. In parallel, parasites were kept at room temperature in static conditions (intact rosettes) or agitated using a vortex for 30 min after complete disruption of rosettes using a needle (disrupted rosettes), in the presence of autologous plasma. P. vivax isolates (1 × 10 6 ) at 5-8% parasitemia were added to each well of THP-1 cells. After 30 min of incubation in 5% CO 2 atmosphere at 37 °C, THP-1 cells were harvested and washed three times in cell culture medium, and the slides were fixed in methanol and stained with panoptic stain. At least 200 THP-1 cells were counted in each well, and all samples were tested in duplicated. THP-1 cell culture was checked for Mycoplasma contamination weekly. Twenty mL of total blood were used to separate PBMC using Ficoll-Paque-PLUS following the fabricant instructions. After isolation, cells were counted at optical microscopy and 1 × 10 6 cells were added to a Labtek slide and kept for 4 h at 37 °C. Non-adherent cells were removed and phagocytosis assay was performed. Immediately before phagocytosis assay, parasites collect from the same patients were kept at room temperature in static conditions (intact rosettes) or agitated using a vortex for 30 min after complete disruption of rosettes using a needle (disrupted rosettes), in the presence of autologous plasma. P. vivax isolates (2 × 10 6 ) at 5-8% parasitemia were added to each slide of PBMC. After 1h30min of incubation at 37 °C, PBMCs were harvested and washed three times in cell culture medium, and the slides were fixed in methanol and stained with panoptic stain. At least 200 PBMCs were counted in each well, and all samples were tested in duplicated.
Statistics. Data were analyzed using R software version 3.6.2. For quantitative variables, we analyzed the frequency distribution, measures of central tendency (mean or median) and variability (standard deviation or interquartile ranges), and searched for outlier observations (those values above [75 percentile value + 1.5 * IQR] or below [25 percentile value -1.5 * IQR], IQR: interquartile range, represented by percentile 75 value minus percentile 25 value). Additionally, for IL-6 and IL-10 variables, we applied logarithm transformation to stabilize its high variability and for IFN-γ we have dichotomized its values in not detect (0 values) and detected (above 0 values) IFN-γ. For immunoglobulin reactivity indexes joint distribution presentation, we have scaled these variables (subtraction of the mean and division by standard deviation) since it mean and standard deviation were markedly different.
In bivariate analysis, for paired data comparisons of rosetting rate between autologous plasma and the following conditions-heat-inactivated plasma, heterologous nonimmune plasma, heterologous immune plasma and Albumax-and for comparisons of phagocytose index between integrated and disrupted rosette groups we applied paired Student t-test or Wilcoxon test, depending on the distribution of the difference of the variable of interest between paired observation. For comparisons of rosetting rate between detected and not detected IFN-γ groups, we applied t-test. To investigate possible relationships between quantitative variables, we used Pearson (r) in case distribution approximately normal, and Spearman's rank correlation coefficient (rho) otherwise, with correspondent 95% confidence intervals.
A p-value < 0.05 was considered to be statistically significant. All statistical tests applied are indicated in figures caption, as well as the number of samples analyzed.

Data availability
All data generated or analyzed during this study are included in this published article (and its Supplementary Information files). Deep sequencing data was deposit in Array Express, accession number E-MTAB-8385.