Total parasite biomass but not peripheral parasitaemia is associated with endothelial and haematological perturbations in Plasmodium vivax patients

  1. João L Silva-Filho  Is a corresponding author
  2. João CK Dos-Santos
  3. Carla Judice
  4. Dario Beraldi
  5. Kannan Venugopal
  6. Diogenes Lima
  7. Helder I Nakaya
  8. Erich V De Paula
  9. Stefanie CP Lopes
  10. Marcus VG Lacerda
  11. Matthias Marti  Is a corresponding author
  12. Fabio TM Costa  Is a corresponding author
  1. Laboratory of Tropical Diseases – Prof. Luiz Jacintho da Silva, Department of Genetics, Evolution, Microbiology and Immunology, Institute of Biology, University of Campinas, Brazil
  2. Wellcome Centre for Integrative Parasitology, Institute of Infection, Immunity & Inflammation, University of Glasgow, United Kingdom
  3. Post-Graduation in Medical Pathophysiology, School of Medical Sciences, University of Campinas, Brazil
  4. School of Pharmaceutical Sciences, University of São Paulo, Brazil
  5. Hospital Israelita Albert Einstein, Brazil
  6. Department of Clinical Pathology, School of Medical Sciences, University of Campinas, Brazil
  7. Institute Leônidas & Maria Deane, Fiocruz, Brazil
  8. Tropical Medicine Foundation Dr. Heitor Vieira Dourado, Brazil

Abstract

Plasmodium vivax is the major cause of human malaria in the Americas. How P. vivax infection can lead to poor clinical outcomes, despite low peripheral parasitaemia, remains a matter of intense debate. Estimation of total P. vivax biomass based on circulating markers indicates existence of a predominant parasite population outside of circulation. In this study, we investigate associations between both peripheral and total parasite biomass and host response in vivax malaria. We analysed parasite and host signatures in a cohort of uncomplicated vivax malaria patients from Manaus, Brazil, combining clinical and parasite parameters, multiplexed analysis of host responses, and ex vivo assays. Patterns of clinical features, parasite burden, and host signatures measured in plasma across the patient cohort were highly heterogenous. Further data deconvolution revealed two patient clusters, here termed Vivaxlow and Vivaxhigh. These patient subgroups were defined based on differences in total parasite biomass but not peripheral parasitaemia. Overall Vivaxlow patients clustered with healthy donors and Vivaxhigh patients showed more profound alterations in haematological parameters, endothelial cell (EC) activation, and glycocalyx breakdown and levels of cytokines regulating different haematopoiesis pathways compared to Vivaxlow. Vivaxhigh patients presented more severe thrombocytopenia and lymphopenia, along with enrichment of neutrophils in the peripheral blood and increased neutrophil-to-lymphocyte ratio (NLCR). When patients’ signatures were combined, high association of total parasite biomass with a subset of markers of EC activation, thrombocytopenia, and lymphopenia severity was observed. Finally, machine learning models defined a combination of host parameters measured in the circulation that could predict the extent of parasite infection outside of circulation. Altogether, our data show that total parasite biomass is a better predictor of perturbations in host homeostasis in P. vivax patients than peripheral parasitaemia. This supports the emerging paradigm of a P. vivax tissue reservoir, particularly in the haematopoietic niche of bone marrow and spleen.

Introduction

Malaria remains a heavy burden across endemic regions worldwide. In 2018, Plasmodium vivax infection accounted for 41% of all malaria cases outside of Sub-Saharan Africa, with a total of 6.5 million cases and more than 2 billion people in 90 countries at risk (World Malaria Reports, 2019). There are concerns that P. vivax elimination will be significantly more difficult than P. falciparum as the current measures for malaria control are less effective for P. vivax than for Plasmodium falciparum, with the elimination of the former presenting a major challenge in areas that successfully reduced P. falciparum burden. This persistence is due to some unique biological features complicating treatment and elimination, including low peripheral parasitaemia and presence of dormant liver stages (hypnozoites) which relapse weeks or months after blood infection has been cleared.

P. vivax infection is associated with low peripheral parasitaemia (<2%) as a result of a strict host cell tropism to immature reticulocytes (Malleret et al., 2015; Mayor and Alano, 2015) that are exceedingly rare in peripheral blood (<2%) but highly prevalent in the haematopoietic niche of bone marrow (BM) and spleen (Klei et al., 2017; Rhodes et al., 2016). Because of limited microvascular adherence in vivo and endothelial cell (EC) binding in vitro (Lacerda et al., 2012; Valecha et al., 2009), it was generally assumed that peripheral parasitaemia reflects the majority of P. vivax parasites during infection. However, discrepancy of parasite biomass based on systemic biomarkers such as Plasmodium lactate dehydrogenase (pLDH) compared to peripheral parasitaemia supports existence of a major P. vivax reservoir outside of circulation, particularly in patients with complicated outcomes (Barber et al., 2015). In support of this hypothesis, studies have demonstrated that late asexual blood stage P. vivax parasites (i.e. schizonts) are capable of cytoadhering to endothelial host receptors (Carvalho et al., 2010; De las Salas et al., 2013) and present at reduced abundance compared to the other blood stages in the blood of P. vivax patients (Obaldia et al., 2018; Lopes et al., 2014). In experimentally infected non-human primates (NHPs), a significant enrichment of sexual stages (gametocytes) and schizonts in BM sinusoids and parenchyma has been observed (Obaldia et al., 2018), supporting previous evidence from multiple case reports that identified P. vivax in BM and spleen (Yx et al., 2009; Wickramasinghe et al., 1989; Wickramasinghe and Abdalla, 2000; Salutari et al., 1996; Baro et al., 2017; Machado Siqueira et al., 2012; Lacerda et al., 2008; Brito et al., 2020). A series of recent studies in acute and chronic human P. vivax infection have meanwhile provided direct evidence that BM and spleen represent the major reservoir of parasite biomass in P. vivax infection (Baro et al., 2017; Brito et al., 2020; Kho et al., 2021a; Kho et al., 2021b).

P. vivax parasites can elicit a potent host response, including inflammation and EC activation, and cause severe and fatal manifestations at significantly lower peripheral parasitaemia than the more virulent species, P. falciparum (Barber et al., 2015; Yeo et al., 2010). However, the pathogenic mechanisms underlying these alterations in host homeostasis and their relationship with P. vivax biomass are not fully understood (Lacerda et al., 2011; Naing and Whittaker, 2018; Rodriguez-Morales et al., 2005; Tangpukdee et al., 2008).

Here we systematically investigated host responses in a cross-sectional cohort of uncomplicated P. vivax patients from Manaus, in the Brazilian Amazon region. Our analysis revealed an association between alterations in host homeostasis, including EC activation, damage, and haematological disturbances, such as thrombocytopenia, lymphopenia, and increased neutrophils turnover, with total parasite biomass but not peripheral parasitaemia. These findings are in line with the emerging paradigm of a clinically relevant parasite reservoir outside of circulation and merit systematic investigations of this reservoir in vivax malaria.

Results

Uncomplicated P. vivax patients present with haematological changes

We have conducted a cross-sectional study with uncomplicated P. vivax malaria patients seen at FMT-HVD in Manaus, Brazil. We included 79 adult patients (median age of 36 years) with confirmed P. vivax infection (smear and PCR positive) and 34 age- and sex-matched uninfected healthy donors (controls; Table 1). All individuals within the study including controls were from the state of Amazonas, in the Amazon region of Brazil. Blood was collected at enrolment for determination of haematological parameters, peripheral parasitaemia by Giemsa staining of blood smears, and PCR to determine genome copy numbers. Preparation of poor platelet plasma (PPP) was done within 15 min of sampling. The median peripheral parasitaemia was 4290 infected red blood cells (iRBCs)/μL of blood (25–75 interquartile range 1860–6620 parasites/μL) and parasite load of 26,642 copies of 18S RNA/μL (25–75 interquartile range 9253–522,297). We also measured total parasite biomass independently of peripheral parasitaemia by quantifying levels of P. vivax lactate dehydrogenase (PvLDH) in plasma (Table 1).

Table 1
Demographic, parasite, and multiplexed microbead-based immunoassay (Luminex) data obtained from the plasma of a representative subset of 31 P. vivax patients and 9 healthy donors (controls).
ParametersHealthy donors (n = 36)Symptomatic Pv patients (n = 79)p-Value(Pv vs. control)
Median [IQ 25–75]Median [IQ 25–75]
Age32 (23–49)36 (28–45)0.06
Parasitaemia (103/mL)-4.29 [1.86–6.62]
Parasitaemia (%)-0.76 [0.57–1.25]
Parasite load (copies 18S RNA/mL)-26,642 [9253-522,297]
PvLDH (O.D.)-0.18 [0.005–0.34]
Plasma biomarkersHealthy donors (n = 9)Symptomatic Pv patients (n = 31)p-Value(Pv vs. control)
TNF-α (pg/mL)17.2 [11.0–22.3]38.4 [30.0–69.6]<0.0001
IL-1α (pg/mL)11.9 [10.0–19.5]25.4 [19.8–33.5]0.0004
IL-1β (pg/mL)12.0 [8.0–12.8]21.4 [14.5–27.6]<0.0001
IL-6 (pg/mL)3.0 [2.5–3.7]33.4 [7.6–133.1]<0.0001
IL-8 (pg/mL)2.2 [0.6–2.4]6.4 [2.7–19.9]0.0005
IL-10 (pg/mL)–*314 [169–562]
G-CSF (pg/mL)9.485 [9.485–9.485]101.5 [33.49–239.6]<0.0001
L-selectin (ng/mL)326 [287–391]481 [386–579]0.0019
ICAM-1 (ng/mL)323 [260–464]634 [456–849]0.0026
VCAM-1 (ng/mL)819 [623–959]2875 [1753–5108]<0.0001
E-Selectin (ng/mL)26.4 [22.5–33.7]56.7 [41.5–74.1]0.0001
P-selectin (ng/mL)17.0 [15.4–20.6]22.2 [17.6–25.7]0.0621
Angiopoietin-1 (ng/mL)0.4 [0.3–0.6]0.5 [0.2–0.9]0.8874
Angiopoietin-2 (ng/mL)1.8 [1.5–2.1]4.3 [2.7–5.3]0.0003
Ang-2:Ang-1 ratio4.2 [2.7–5.6]12.14 [2.7–40.2]0.03
VWF-A2 (pg/mL)126 [120–150]218 [199–277]<0.0001
ADAMTS13 (ng/mL)1110 [483–1740]776 [572–1328]0.5485
PAI-1 (pg/mL)78.9 [62.4–96.4]112 [69.3–242]0.1541
CD40L (ng/mL)0.5 [0.4–0.7]1.0 [0.7–1.3]0.0001
Syndecan-1 (ng/mL)1.8 [1.6–2.4]3.7 [2.9–6.0]0.0003
IL-11 (ng/mL)3.5 [2.9–4.3]5.7 [4.7–6.4]<0.0001
TPO (ng/mL)2.0 [1.7–2.2]3.0 [2.6–3.4]<0.0001
CXCL4 (ng/mL)0.8 [0.6–1.2]1.4 [0.7–2.8]0.1236
CXCL7 (ng/mL)0.4 [0.4–0.5]0.73 [0.4–1.7]0.1958
SCF (pg/mL)47.61 [37.34–89.34]45.68 [36.22–61.39]0.1594
  1. PvLDH: P. vivax lactate dehydrogenase.

  2. * = under detection limit.

Analysis of haematological parameters revealed significantly reduced haemoglobin levels and haematocrit across P. vivax patients compared to controls, with anaemia in 38% of the patients (Figure 1A). Similarly, leukocyte numbers were significantly decreased (mean ± SD: 4.36 ± 1.74 × 103/μL vs. 5.72 ± 1.34 × 103/μL, p=0.0004), with 54.5% of the patients presenting with leukopenia (defined as a leukocyte count <4000 cells/μL). In contrast, neutrophil counts were not significantly different, and only 8.3% of P. vivax patients were presenting with neutropenia (neutrophil counts < 1500 cells/μL) (Figure 1B). Other myeloid cell populations, however, such as monocytes, basophils, and eosinophils (MXD), were significantly reduced. We also observed a significant reduction in lymphocyte and platelet counts in this cohort (Figure 1C), with 80% presenting with lymphopenia (lymphocyte counts < 1000 cells/μL) and 87% with thrombocytopenia (platelet counts < 150,000 cells/μL), many of them with severely reduced levels (Figure 1C). Alterations in platelet counts were accompanied by the release of mega platelets in the peripheral circulation as a significant increase on mean platelet volume was observed (Figure 1C).

Figure 1 with 1 supplement see all
Clinical data of P. vivax patients (Pv) and healthy donors (HDs).

(A) Red blood cell parameters. Shown are red blood cell counts, haemoglobin levels, and haematocrit. (B) Other blood cell parameters. Shown are numbers of leukocytes, neutrophils, and monocytes, basophils, and eosinophils (MXD). (C) Number of lymphocytes, platelets, and mean platelet volume (MPV). Parameters are depicted as box plots showing each individual value and the median with maximum and minimum values. Dashed lines in black mark the minimum threshold for normal reference values, while lines in red mark threshold for severe lymphopenia and thrombocytopenia, respectively. Two-tailed Student’s t-test was used to compare variables with normally distributed data, and Mann–Whitney test was used to compare variables with non-normal distributions; p-value is indicated above the graph when p<0.05. HDs: healthy donors (controls, n = 34); Pv: P. vivax-infected patients (n = 79).

In summary, patients in our cohort presented with a wide range of parasitaemia and uncomplicated clinical signs of P. vivax infection at medical consultation. However, significant haematological abnormalities were present in the majority of patients during early onset of disease, in line with previous findings (Barber et al., 2015; Lacerda et al., 2011; de Mast et al., 2009; de Mast et al., 2007; Gomes et al., 2014; Park et al., 2003; Punnath et al., 2019).

Unsupervised clustering reveals two P. vivax patient subgroups that differ in parasite biomass: Vivaxhigh vs. Vivaxlow

To determine whether the observed changes were associated with specific host signatures, particularly circulating biomarkers of haematological and endothelial changes, we applied a multiplexed microbead-based immunoassay (Luminex) in a representative subset of 31 P. vivax patients and 9 controls, as explained in the Materials and methods section (Figure 1—figure supplement 1). We selected a series of circulating biomarkers associated with haematological changes, including cytokines altering thrombopoiesis (TPO and IL-11), myelopoiesis, and lymphopoiesis (TNF-α, IL-1α, IL-1β, IL-6, IL-8, G-CSF) (Boiko and Borghesi, 2012; Chiba et al., 2018; Kovtonyuk et al., 2016). In addition, we selected markers of EC and platelet activation, coagulation (ICAM-1, VCAM-1, E-selectin, P-selectin, Angiopoietin-1 and -2, CD40L, VWF-A2, ADAMTS13, PAI-1, CXCL4, CXCL7), and EC glycocalyx breakdown (Syndecan-1).

We observed significant upregulation of multiple cytokines associated with haematological changes in the P. vivax patients compared to control (Table 1). In addition, patient samples exhibited a strong phenotype of increased EC activation, glycocalyx breakdown and coagulation. The high interquartile range in parasitaemia and host signatures (Table 1) suggested a heterogenous phenotype across the patient population. In order to identify possible stratification of patients into distinct subgroups, we further analysed the clinical data (Figure 1), parasite parameters, and Luminex data (Table 1) from the same 31 P. vivax patients and 9 controls as above. After z-score normalization, principal component analysis (PCA) was performed for data dimensionality reduction, considering the large number of variables in our dataset. Next, we ran K-means clustering (k) followed by bootstrapping (Figure 2A and B, Figure 2—figure supplement 1, Figure 2—figure supplement 2, Figure 2—figure supplement 2—source data 1) to identify possible subclusters of individuals. This analysis revealed consistent separation of samples into two clusters, one of them including all controls (cluster 1a) and a subset of 14 patient samples (cluster 1b) and a second one representing the remaining 17 patient samples (cluster 2) (Figure 2A and B). In order to visualize covariables of the observed patient distribution (PCA) and clustering (K-means), we plotted the correlation (loading score) of each input variable with a principal component (PC; Figure 2C, Figure 2—source data 1). This analysis demonstrated covariation of lymphopenia and thrombocytopenia, on the one hand, and markers of EC changes, platelet production, activation, and parasite parameters (PvLDH and peripheral parasitaemia), on the other hand, as major contributors to the PCs (Figure 2C). Direct comparison of the two patient subgroups revealed significant higher total parasite biomass but not peripheral parasitaemia or parasite load (Figure 3A). In agreement with previous findings (Barber et al., 2015; Fonseca et al., 2017; Silva-Filho et al., 2021), z-score comparison further demonstrated that total parasite biomass was higher than and not correlated with peripheral parasitaemia levels or parasite load, particularly in patients of cluster 2 (Figure 3B and C). In addition, PvLDH was the input parasite variable with the highest loading score (correlation = 0.59) and lowest p-value (0.0000917) in the first PC dimension when compared with peripheral parasitaemia and parasite load (Figure 2C, Figure 2—source data 1). Indeed, using a best-fit classification tree model and a random forest machine learning model defining K-means clusters as categorical outcome, PvLDH is the best parasite predictor attribute segregating patients into clusters 1b and 2 (Figure 3D and E). After both models were trained in a set of 30 individuals, randomly selected by the training algorithm set, they were tested in the 10 remaining individuals, where all cluster 1a (control) individuals and 80% of P. vivax patients were correctly classified into either cluster 1b or cluster 2. Based on these observations, we designated cluster 1a as Control cluster (representing the healthy donors), cluster 1b as Vivaxlow (representing patients with low P. vivax biomass), and cluster two as Vivaxhigh (representing patients with high P. vivax biomass).

Figure 2 with 2 supplements see all
Characterization of heterogeneity in symptomatic P. vivax patients defines clusters of individuals.

(A, B) Clustering of patients and healthy controls. After z-score normalization, principal component analysis (PCA) was performed for data dimensionality reduction. K-means clustering using k = 2 followed by bootstrapping (1000 times) in a PCA plot was performed and produced the most stable clusters regardless of the starting point (ln 1000/1000): cluster 1 = 23 individuals comprising 9 healthy donors and 14 P. vivax patients and cluster 2 comprising 17 P. vivax patients. The jaccard_index measures cluster similarity across bootstrap samples (jaccard_index ranges from 0 to 1, an index <0.6 hints at a weak, unreliable cluster while >0.85 means generally reliable). As indicated in the PCA plot, k = 2 gives stable clusters for all configurations (jaccard_index 0.9 and 0.86) and withinss (wss) = 1122. Open ovals represent 95% confidence interval ellipses around group mean points. PCA was performed for data dimensionality reduction, in parallel with K-means clustering (k) followed by bootstrapping (1000 times). Open ovals represent 95% confidence interval ellipses around group mean points. (B) The resulting clusters represent healthy controls (1a) and patients (1b, 2). (C) Contribution of variables to clustering. In the circular plot, the correlation between each input variable and principal components is used as coordinates (loading score). Plots show how covariables determine patient distribution in the PCA plot.

Figure 2—source data 1

Correlation (loading score) of variables to principal components.

https://cdn.elifesciences.org/articles/71351/elife-71351-fig2-data1-v2.xlsx
Unsupervised clustering analysis reveals two P. vivax patient subgroups that differ in parasite biomass.

(A) Parasite parameters vs. patient clusters. Comparison of the two patient clusters (clusters 1b and 2) across parasite parameters reveals significant differences with total parasite biomass (P. vivax lactate dehydrogenase [PvLDH]) but not peripheral parasitaemia or parasite load (copies of 18S rRNA/μL of blood). (B) Parasite biomass vs. parasitaemia across clusters. Heatmap represents z-scores of PvLDH with peripheral parasitaemia or parasite load, respectively. Black boxes highlight patients with relatively lower peripheral parasitaemia compared to PvLDH levels, indicating the underestimation of total parasite biomass based on peripheral parasitaemia values. (C) Correlation between parasite biomass and parasitaemia. Scatter plot showing lack of correlation between PvLDH and peripheral parasitaemia or parasite load, respectively. Regression line in red, with 95% confidence interval shown in shaded grey. (D, E) Predicting parasite clusters. (D) Top parameters prioritized by random forest analysis ranked by the mean decrease in accuracy. (E) Best-fit decision trees and random forest machine learning models corroborate PvLDH value as the most important parasite signature in segregating patients into clusters 1b and 2. Cut-off values of the attribute that best divided groups were placed in the root of the tree according to the parameter value. The total of classified registers for each class and the percentage of observations used at that node are given in each terminal node.

Different levels of haematological alterations between Vivaxhigh and Vivaxlow patients

The three clusters were not significantly different in patient age (median: 33; IQ 25–75: 22–57), gender (80% male; 20% female in each cluster), average days of symptoms when samples were collected, haemoglobin levels, haematocrit, or RBC counts, indicating that these parameters are not confounders accounting for the differences observed between the clusters (Figure 4A). However, systematic analysis of haematological parameters between Vivaxhigh and Vivaxlow patients revealed significant differences. Vivaxhigh patients showed a more intense reduction in platelet counts when compared to Vivaxlow patients (Vivaxhigh 63,000 ± 6413 vs. Vivaxlow: 100,700 ± 9381; p=0.002), with a higher frequency of patients with severe thrombocytopenia (Vivaxhigh 47% vs. Vivaxlow 8%) (Figure 4B). Although not significant, there was a trend in the reduction of lymphocyte counts in Vivaxhigh patients when compared to Vivaxlow, with 88% of Vivaxhigh patients presenting lymphopenia versus 64% in Vivaxlow patients. In addition, we observed a fourfold increase in the frequency of patients with severe lymphopenia in the Vivaxhigh cluster compared to Vivaxlow patients (Figure 4B). While there was no change in the number of circulating neutrophils in the different clusters of individuals, mixed cell counts (MXD), a parameter representing monocytes, basophils, and eosinophils numbers, were significantly reduced in Vivaxhigh patients. As a result, there was a significant enrichment of neutrophils in the leukocyte fraction in the blood of Vivaxhigh patients as well as a higher NLCR (Figure 4B).

Figure 4 with 1 supplement see all
More severe haematological alterations in Vivaxhigh compared to Vivaxlow patients.

(A) Patient data and haematological parameters. Comparison of patient age, average days of symptoms when samples were collected, haemoglobin levels, haematocrit, or RBC counts across patient clusters (Control: n = 9; Vivaxlow : n = 14; Vivaxhigh: n = 17). Data are depicted as plots showing individual values and the median (black lines) and the interquartile range. (B) Blood cell counts. Comparison of differential haematological counts across clusters. Shown are numbers of platelets, lymphocytes, neutrophils, and monocytes, basophils, and eosinophils (MXD), neutrophil to total leukocyte ratio, and neutrophil to lymphocyte ratio (NLCR). Top dashed lines mark the minimal threshold for normal reference values, while bottom dashed lines mark the threshold for severe lymphopenia and thrombocytopenia, respectively. Parameters are depicted as plots showing individual values and the median (black lines) and the interquartile range. One-way analysis of variance with Bonferroni-corrected multiple comparisons test was performed. p-Value is indicated above the graph when reached significance of p<0.05. (C) Cytokine response and neutrophil activation across clusters. Heatmap represents z-scores obtained by centering values of Luminex data. Shown are thrombopoiesis-inducing cytokines, myelopoiesis-inducing cytokines, and neutrophil activation markers. Biomarker concentrations were normalized (scale function in R), and the average scaled value is showed in blue and yellow scales. Blue shading represents the highest average scaled value, and yellow shading represents the lowest average scaled value. Each column (i.e. individual) in the heatmap is matched with colour-coded cluster assignment: Cluster Control – green bar; Cluster Vivaxlow – blue bar; and Cluster Vivaxhigh – red bar.

In parallel to more severe thrombocytopenia in Vivaxhigh patients, plasma levels of cytokines inducing megakaryocytic differentiation in the BM, thrombopoietin (TPO), and IL-11 were significantly increased in this cluster (Figure 4C, Figure 4—figure supplement 1). In accordance with the pattern of immune cell fractions in the peripheral blood of P. vivax patients, the Vivaxhigh cluster showed a significant increase in the levels of proinflammatory cytokines associated with induction of myeloid-biased haematopoietic stem cell (HSC) differentiation and inhibition of lymphopoiesis in BM (e.g. TNF-α, IL-1α, IL-1β, IL-6, IL-8; Figure 4C, Figure 4—figure supplement 1; Boiko and Borghesi, 2012; Chiba et al., 2018; Kovtonyuk et al., 2016). In addition, Vivaxhigh patients had increased circulating levels of G-CSF, a major mediator of HSC-biased myelopoiesis and the neutrophil activation marker, L-selectin (Figure 4C, Figure 4—figure supplement 1; Soehnlein et al., 2017; Ivetic, 2018; Crockett-Torabi et al., 1995). Together, these Luminex data support the haematological measurements, suggesting that a compensatory response is mounted in the BM to counterbalance the massive decrease of platelets in periphery. Upregulation of cytokines inducing myelopoiesis, while inhibiting lymphopoiesis (Boiko and Borghesi, 2012; Chiba et al., 2018; Kovtonyuk et al., 2016), might also explain the decrease of lymphocyte counts and enrichment of activated neutrophils in the peripheral circulation of P. vivax patients.

Elevated circulating markers of EC activation and damage in Vivaxhigh compared to Vivaxlow patients

Patient clustering indicated that Vivaxhigh patients have increased levels of EC markers in the plasma compared to Vivaxlow patients (Figure 2C). Previous studies indicate that EC activation and damage might contribute to thrombocytopenia and inducing haematopoiesis, resulting in HSC differentiation directed towards myelopoiesis (Lacerda et al., 2011; de Mast et al., 2009; de Mast et al., 2007; Boiko and Borghesi, 2012; Chiba et al., 2018; Kovtonyuk et al., 2016; Graham et al., 2016; Dos-Santos et al., 2020; Lazzari and Butler, 2018). In our cohort, circulating levels of EC adhesion molecules (ICAM-1, VCAM-1, E-selectin, and P-selectin) and other EC activation markers and procoagulant molecules (Ang-2, VWF-A2, CD40L, and PAI-1) were significantly increased in the plasma of Vivaxhigh patients compared to Vivaxlow patients and healthy controls (Figure 5A, Figure 5—figure supplement 1A and B). Likewise, Syndecan-1, a marker of EC glycocalyx breakdown (i.e. damage of EC plasma membrane; Yeo et al., 2019; Pillinger and Kam, 2017), was significantly increased in Vivaxhigh but not in Vivaxlow patients (Figure 5A, Figure 5—figure supplement 1C).

Figure 5 with 2 supplements see all
Elevated circulating markers of endothelial cell (EC) activation and damage in Vivaxhigh compared to Vivaxlow patients.

(A) Endothelial changes across clusters: Luminex. Heatmap represents z-scores obtained by centering values of Luminex data. Shown are markers of EC activation, procoagulation, and glycocalyx damage. Each column (each individual) in the heatmap is matched with colour-coded cluster assignment: Cluster Control – green bar; Cluster Vivaxlow – blue bar; and Cluster Vivaxhigh – red bar. (B) Endothelial changes across clusters: qRT-PCR. Transcriptional response of human umbilical vein endothelial cells (HUVECs) incubated for 6 hr with 30% v/v pooled plasma from different clusters. Heatmap reflects relative mRNA expression intensity (average scaled value) after results were normalized to GAPDH housekeeping gene expression and untreated condition (mock). Data shown represent the mean of three independent experiments. For each experiment, two technical replicates were performed for each condition. (C) Endothelial changes across clusters: impedance changes. Endothelial monolayer integrity was measured during 20% v/v of pooled plasma incubation. Each line represents the mean ± SD of normalized resistance of HUVECs measured by electric cell-substrate impedance sensing (ECIS) at 4000 Hz. Data shown are representative of three independent experiments. For each experiment, two technical replicates were performed for each condition. (D) Endothelial changes across clusters: imaging and flow cytometry. HUVECs were incubated for 18 hr with 30% v/v of pooled plasma of individuals in the different clusters or left untreated (mock). Percentage of cells expressing EC activation markers (adhesion molecules) ICAM and VCAM as well as quantification of protein expression was determined by flow cytometry and immunofluorescence analysis (scale bar = 33 μM). Isotype antibodies were used as control to define positive populations. Significance was calculated for comparisons between conditions at the corresponding time point . One-way analysis of variance statistical test with Tukey’s corrected multiple comparisons test was performed. p-Value is indicated above the graph when p<0.05. Data shown are representative mean ± SEM of three independent experiments.

To independently test whether host factors and/or parasite products present in the plasma of the different patient groups can directly induce changes in ECs, we stimulated primary human umbilical vein endothelial cells (HUVECs) with pools of plasma from either Vivaxhigh patients, Vivaxlow patients, or healthy controls. These experiments demonstrated that only pooled plasma from Vivaxhigh patients induces significant transcriptional upregulation of EC activation markers ICAM-1, IL-1α, and IL-8 along with downregulation of Ang-1, ADAMTS13, and NOS3 (eNOS) in HUVECs (Figure 5B, Figure 5—figure supplement 1D). In contrast, expression of Syndecan-1 and VEGF, two indicators of vascular damage, was not affected by either treatment (Figure 5B, Figure 5—figure supplement 1D). Similarly, electric cell-substrate impedance sensing (ECIS) assays did not detect differences in functional perturbations in the endothelial cellular monolayer upon incubation with P. vivax pooled plasma when compared to control pooled plasma (Figure 5C). In contrast, flow cytometry and immunofluorescence assays performed with stimulated HUVECs revealed increased prevalence and protein expression levels of EC activation markers ICAM-1 and VCAM-1 upon exposure with Vivaxhigh pooled plasma (Figure 5D, Figure 5—figure supplement 1E), in support of qRT-PCR data. These data indicate that local EC activation, mediated by direct or indirect interactions with parasitized RBCs, can be measured systemically.

Indirect evidence for parasite-induced changes in deep tissues

To further investigate the interplay between host biomarkers and associated cellular responses as well as parasite parameters, we constructed a network of interactions based on Pearson’s correlations with absolute correlation coefficient above 0.5 and p-value<0.05 (Figure 6A). In addition, we also performed hierarchical clustering on matrices of Pearson’s correlations (p-value<0.01) with selected modules of parasite and host parameters (Figure 6B). Data from Vivaxlow and Vivaxhigh patient subgroups were combined for this analysis as they similarly contribute to the associations we found so far (Figure 6—figure supplement 1).

Figure 6 with 3 supplements see all
Network analysis and clustering of parasite and host signatures indicate parasite-induced changes in deep tissues.

(A) Network analysis. Networks of the Pearson’s correlations (absolute coefficient above 0.5 and p-value<0.05) between parasite biomass (P. vivax lactate dehydrogenase [PvLDH]) and host signatures in healthy donors (left graph) and in P. vivax-infected patients (right graph), using a force-directed layout. The symbols of the nodes represent biological functions: triangle represents markers of platelet activation and thrombopoiesis-inducing cytokines; V shape represents haematological parameters (neutrophil, lymphocyte, and platelet counts); circles represent endothelial cell activation markers; squares represent myelopoiesis-inducing cytokines and neutrophil activation markers. The colours in the nodes represent the fold change in relation to control levels. Because healthy donors do not have parasitaemia, PvLDH node is represented in black. Each connecting line (edge) represents a significant interaction detected by the network analysis using R. Correlation strength is represented by edge colour transparency and width. Positive correlations are represented by red edges, and negatives correlations are represented by blue edges. (B, C) Correlation matrix and heatmap. (B) Representative image of Pearson’s correlation matrix calculated for all P. vivax patients. Only correlations with p-value<0.01 are shown, and hierarchical clustering was applied. Red circles highlight positive correlations in the functional modules depicted in (A), and blue circles highlight negative correlations in the functional modules also depicted in (A). (C) Heatmap showing p-values of the correlations between different parasite parameters, parasite biomass (PvLDH), and peripheral parasitaemia and host signatures (haematological and Luminex parameters). (D) Decision tree model. Best-fit classification tree model generated with the C4.5 algorithm showing Syndecan-1, IL-6, and platelet counts are the dominant variables capable of predicting total parasite biomass in P. vivax patients. Cut-off values of the attribute that best divided groups were placed in the root of the tree according to the parameter value (pg/mL for soluble markers or number of cells × 1000/μL of blood for platelet counts). The total of classified registers for each class is given in parentheses for each terminal node with the k-fold cross-validation (k-fold CV) accuracy indicated.

Similar to a previous study with P. vivax patients and healthy donors from an endemic area in Brazil (Mendonça et al., 2013), our analysis revealed a dense network of interactions with homogenous and centralized topology among the biomarkers in healthy donors (Figure 6A, Supplementary file 1). The network topology is drastically altered in symptomatic P. vivax patients, largely due to the introduction of parasite parameters in the patient graph (Figure 6A, Supplementary file 1). The network analysis revealed a decentralized topology, lower complexity and connectivity between the edges with data from P. vivax patients compared to the highly dense, homogenous and centralized network graph of healthy donors (91 edges vs. 166 edges, respectively). Of note, the network pattern described in our study is similar to protein-protein-associated networks described previously in P. vivax malaria and in other clinical contexts (Mendonça et al., 2013; Frankenstein et al., 2006). Interestingly, due to its decentralized and heterogenous patterns of interactions, the network graph of P. vivax patients is separated into three modules of strong interactions, with closely related biological functions. Module 1 is formed by markers of EC activation and damage, together with lymphocyte, platelet, and neutrophil counts in addition to the megakaryocyte differentiation-inducing cytokines (TPO and IL-11) (Figure 6A). In support of the role of EC activation and damage in the haematological changes observed in this cohort, hierarchical clustering revealed a positive correlation between adhesion molecules VCAM-1 and E-selectin and EC glycocalyx breakdown (Syndecan-1) (Figure 6B). In addition, VCAM-1, E-selectin, Ang-2 and VWF-A2, and Syndecan-1 are negatively correlated with platelet and lymphocyte counts, while ICAM-1 is positively correlated with neutrophil counts (Figure 6A and B). Module 2 is formed by proinflammatory cytokines with myelopoiesis-inducing effects and molecules associated with platelet activation and coagulation cascades (Figure 6A and B). Interestingly, EC activation markers and Syndecan-1 (EC damage) from module 1 also display positive correlations with myelopoiesis-inducing cytokines from module 2 (Figure 6B). Finally, module 3 is formed by Ang-2 and the proinflammatory cytokine IL-1β negatively associated with haemoglobin, haematocrit, and RBC numbers (anaemia markers) (Figure 6A and B). Most notably, PvLDH connects the two main functional modules 1 and 2 (Figure 6A and B). Accordingly with Figures 2, 6A and B, the biological significance of total parasite biomass, but not peripheral parasitaemia or parasite load, in affecting host response is also corroborated by the high significant and positive associations of PvLDH with multiple host parameters, including Syndecan-1 (EC damage), VCAM-1, VWF (EC activation and platelet pooling), and IL-6, IL-8, and TNF-α (inflammation and myelopoiesis-inducing cytokines) (Figure 6C). Meanwhile, platelet, lymphocyte, and neutrophil counts are negatively correlated with high significance (p-value<0.0001) with total parasite biomass, but not with peripheral parasitaemia or parasite load (Figure 6C). The association between endothelial activation, Syndecan-1, and parasite biomass (PvLDH) indicates a positive feedback loop between glycocalyx breakdown, activation of endothelial receptors such as ICAM-1 and VCAM-1, and parasite accumulation in deep tissues (Carvalho et al., 2010; Lopes et al., 2014). Similar to Figure 2E, application of a best-fit classification tree model demonstrated that Syndecan-1, IL-6, and platelet counts are the most dominant predictor attributes capable of classifying P. vivax patients based on total parasite biomass levels (Figure 6D). Using this model, all P. vivax patients were correctly classified into either low (Vivaxlow) or high (Vivaxhigh) total parasite biomass (PvLDH). In turn, PvLDH is a relevant predictor attribute (high information gain) in predicting thrombocytopenia severity, and it is associated with increased severity of thrombocytopenia and lymphopenia in our cohort (Figure 6—figure supplement 2, Figure 6—figure supplement 3). Together, these data further support the hypothesis that a parasite population outside of circulation, as represented by total parasite biomass, is driving the host response including EC activation and damage as well as haematological disturbances (i.e. lymphopenia, thrombocytopenia, and anaemia) in P. vivax patients (Figure 6—figure supplement 2, Figure 6—figure supplement 3).

Discussion

In this study, we performed a comprehensive analysis of host and parasite signatures detected in the plasma of a cross-sectional cohort of uncomplicated P. vivax malaria. Initial analysis of a series of circulating host biomarkers revealed significant levels of thrombocytopenia, lymphopenia, and anaemia, as well as EC activation and damage across P. vivax patients compared to healthy controls. Deconvolution of heterogeneity across patients revealed two patient subgroups (Vivaxhigh and Vivaxlow) characterized by differences in total parasite biomass (based on circulating PvLDH levels) but not peripheral parasitaemia (based on blood smears). We observed a significant correlation between total parasite biomass (but not peripheral parasitaemia) and systemic levels of markers of EC activation and damage and haematopoietic perturbations. In addition, by applying a supervised machine learning tree-structured model, we were able to associate EC damage and thrombocytopenia with parasite biomass. In agreement with a previous study (Barber et al., 2015; Silva-Filho et al., 2021), our observations further suggest that total parasite biomass as measured by PvLDH is a better predictor of P. vivax host responses and pathogenesis than peripheral parasitaemia. Furthermore, these findings support the emerging paradigm of a major P. vivax parasite reservoir outside of circulation, particularly in the haematopoietic niche of BM and spleen (Silva-Filho et al., 2021).

The existence of a significant P. vivax reservoir outside of circulation was first predicted by disproportionately high PvLDH levels in peripheral circulation compared to parasitaemia by blood smear (particularly in patients with complicated outcomes) and by modelling using experimental Plasmodium cynomolgy infections in NHPs (Barber et al., 2015; Fonseca et al., 2017). Recent studies provide direct evidence that BM and spleen represent the major reservoir of parasite biomass in P. vivax infection (Obaldia et al., 2018; Baro et al., 2017; Brito et al., 2020; Kho et al., 2021a; Kho et al., 2021b). PvLDH is produced by viable or recently killed parasites and hence considered a proxy for ongoing rather than past infection (Barber et al., 2015; Druilhe et al., 2007). PvLDH antigen capture ELISA established a direct relationship between pLDH levels and P. vivax parasitaemia in ex vivo experiments, demonstrating that pLDH reflects total P. vivax parasite biomass (Druilhe et al., 2007). Our study further explores the relevance of PvLDH as a prognostic marker of host perturbations and disease severity, with a particular focus on markers of changes in the haematopoietic niches of BM and spleen. A major observation in the network graph of P. vivax patients is the central position of the total parasite biomass marker PvLDH due to its equally strong interactions with the two main functional modules 1 and 2. Given that the haematopoietic niches of the BM and the spleen are the major reservoir of parasite biomass, interactions of PvLDH with these two main modules indicate an interplay between parasite infection in these niches and endothelial activation/damage as well as the proinflammatory response that results in myeloid-biased differentiation, thrombocytopenia, and lymphopenia. Furthermore, the highly significant and positive associations between endothelial activation, Syndecan-1, and parasite biomass (PvLDH) indicate a positive feedback loop between glycocalyx breakdown, activation of endothelial receptors such as ICAM-1 and VCAM-1, and parasite accumulation in deep tissues. VivaxHigh patients show higher plasma levels of all these markers. Consistent with previous reports (Yeo et al., 2019; Barber et al., 2021), we propose that elevated EC activation and glycocalyx damage increases the exposure of adhesion molecules, which in turn favours endothelial cytoadherence of P. vivax-infected RBCs, particularly in the splenic red pulp cords and in the BM (Kho et al., 2021a; Introini et al., 2018; Hempel et al., 2017; Toda et al., 2020). Accordingly, application of a best-fit classification tree model identifies Syndecan-1 as a putative host biomarker (EC glycocalyx breakdown marker) predicting total parasite biomass in P. vivax patients. We hypothesize that elevated endothelial activation and damage in VivaxHigh patients results in increased cytoadherence of P. vivax iRBCs and hence accumulation and growth in deep tissues, thus reducing the fraction of the parasite biomass in circulation.

In contrast to P. falciparum-infected individuals, a wide range of complicated clinical syndromes occur in P. vivax patients even at low or subpatent parasitaemia (Baird, 2013), thus indicating that peripheral parasitaemia is a poor predictor of clinical outcomes. Two lines of evidence support our conclusion that severity of infection is dependent on parasite biomass instead. First, the discrepancy between PvLDH levels and peripheral parasitaemia determined by blood smears is more evident in P. vivax-infected patients with complicated outcomes: the ratio of plasma pLDH to peripheral parasitaemia is sixfold higher than in non-severe patients. The same comparison between severe and non-severe P. falciparum patients reveals only a 1.4-fold difference (Barber et al., 2015). Second, although thrombocytopenia and lymphopenia are not included in the World Health Organization (WHO) criteria for defining severe malaria, it has been associated with severe manifestations and the need for blood and platelet transfusions in severe vivax malaria. This points out their clinical relevance in malaria diagnosis and treatment (Lacerda et al., 2011; Naing and Whittaker, 2018; Gerardin et al., 2002; Kochar et al., 2010; Kochar et al., 2005), suggesting that these haematological complications could be explored as markers of severity for this species. Both severe thrombocytopenia and lymphopenia were more frequent in patients in cluster 2 (Vivaxhigh) in our study. By integrating these clinical perturbations with host biomarker measurements and parasite parameters, we demonstrated the high attribute value of total parasite biomass in predicting the severity of thrombocytopenia and lymphopenia and highly significant correlations with endothelial activation, glycocalyx breakdown, and other markers of inflammation.

Thrombocytopenia, lymphopenia, and anaemia are the most frequent P. vivax- and P. falciparum-associated haematological complications (Lacerda et al., 2011; Naing and Whittaker, 2018; Rodriguez-Morales et al., 2005; Tangpukdee et al., 2008). In our cohort, 34, 85, and 87% of patients exhibited anaemia, lymphopenia, and thrombocytopenia, respectively. Various mechanisms have been proposed to explain the damage or excessive removal of platelets in P. vivax infection, including oxidative stress, platelet phagocytosis, IgG binding to platelet-bound malaria antigens, spleen pooling, or increased circulating nucleic acids levels (Lacerda et al., 2011; Naing and Whittaker, 2018; Kochar et al., 2010; Andrade et al., 2010). EC activation and damage also plays a role in intravascular platelet agglutination and increased platelet clearance from the circulation (Park et al., 2003; Punnath et al., 2019). Our data also demonstrate that thrombocytopenia is associated with an increase in IL-1, IL-6, IL-8, IL-10, and TNF-α. We also observed elevated levels of cytokines inducing megakaryocyte differentiation, TPO, and IL-11, suggesting that a compensatory response is mounted in the BM to counterbalance the massive decrease of platelets in the periphery. In contrast, the relatively large drop in peripheral lymphocyte numbers we observed in the P. vivax patients is likely non-specific effect, for example, pooling in the enlarged spleen rather than a response by Plasmodium-specific lymphocytes (Hviid and Kemp, 2000). Corroborating the potential role of total parasite biomass, rather than peripheral parasitaemia, in haematological disturbances (i.e. lymphopenia, thrombocytopenia, and anaemia), Figures S7 and S8 show that total parasite biomass increases accordingly with thrombocytopenia and lymphopenia severity. Patients with severe thrombocytopenia also show more severe leukopenia, lymphopenia, and mega platelets (higher MPV). In addition, plasma levels of cytokines – such as TNF-α, IL1-β, IL-8, IL-10; EC activation/damage markers, VCAM-1, E-selectin, VWF-A2, Ang-2, Ang-2:Ang1 ratio; Syndecan-1; thrombopoiesis-inducing cytokines, TPO and IL-11; platelet activation marker, CD40L; and neutrophil activation marker, L-selectin – follow the increase in thrombocytopenia severity (Figure 6—figure supplement 2). A similar pattern is observed when stratifying patients based on lymphopenia severity (Figure 6—figure supplement 3). Interestingly, a tree-structured model demonstrated that PvLDH, along with VCAM-1 and Syndecan-1, is a relevant predictor attribute (high information gain) in predicting thrombocytopenia severity in our cohort (Figure 6—figure supplement 2).

Our data support previous studies suggesting a role for EC activation and damage in increased leukocyte adhesion, intravascular platelet agglutination with increased platelet clearance from the circulation and skewing of haematopoiesis towards the myeloid lineage (likely at the expense of lymphopoiesis) in the BM (de Mast et al., 2009; de Mast et al., 2007; Gomes et al., 2014; Boiko and Borghesi, 2012; Chiba et al., 2018; Kovtonyuk et al., 2016; Graham et al., 2016; Dos-Santos et al., 2020; Pillinger and Kam, 2017). P. vivax elicits a stronger inflammatory response and more pronounced endothelial activation when compared with other Plasmodium infections with similar or higher peripheral parasitaemia (Yeo et al., 2010); however, the role of EC activation in P. vivax pathogenesis is not yet understood. Damage of the EC plasma membrane, as represented by glycocalyx breakdown, has been associated with poor prognostic outcome in P. falciparum (Yeo et al., 2019), but there is no data available for P. vivax. In our cohort, soluble EC activation biomarkers (e.g. ICAM-1, VCAM-1, E-selectin, Ang-2, CD40L, vWF-A2) and the EC damage product, Syndecan-1, are positively correlated with thrombocytopenia, lymphopenia, anaemia, and neutrophil enrichment in the peripheral blood. In addition, these biomarkers are positively correlated with increased circulating levels of cytokines inducing megakaryocyte differentiation (e.g. IL-11 and TPO) and with cytokines inducing myeloid-biased HSC differentiation (e.g. TNF-α, IL1-α, IL6, IL-8, and G-CSF), suggesting both direct and indirect links between EC activation and damage and haematological perturbations. Total parasite biomass-inducing EC activation might act synergistically with inflammatory changes potentially leading to splenic platelet pooling and platelet clumping in the vasculature without DIC (Lacerda et al., 2008; Pillinger and Kam, 2017; Becker et al., 2015). Likewise, increased activation-induced cell death (AICD) in T cells, splenic T-cell accumulation (Hviid and Kemp, 2000), or decreased lymphopoiesis due to myeloid-biased HSC differentiation induced by inflammatory cytokines and EC activation in the BM (Boiko and Borghesi, 2012; Chiba et al., 2018; Silva-Filho et al., 2021) might explain the severe lymphopenia and neutrophilia in vivax patients. Together, such mechanisms could explain the link between parasite biomass and EC activation/damage with haematological changes observed in vivax patients that might contribute to pathogenesis and disease severity.

In a second series of experiments, we performed ex vivo stimulation of HUVECs with the plasma of the P. vivax cohort demonstrating that the mixture of parasite and host factors can directly induce EC activation in the absence of parasitized RBCs. Of note, functional differences between HUVECs and adult vascular endothelium, including lack of ABO blood group antigen expression, have been reported (O’Donnell et al., 2000; Tan et al., 2004). Hence, EC stimulation with patient plasma may be further evaluated using primary vascular ECs.

ECs are capable of responding to pathogens by sensing pathogen-associated molecular patterns (PAMPs) through pattern-recognition receptors (PRRs), which might play a key role in inducing EC activation when detecting P. vivax molecules enriched in the tissues where the parasite accumulates. ECs also express specific cytokine/chemokine receptors to detect proinflammatory signals released systemically or locally by activated immune cells in response to infection (Bernardo et al., 2004; Bevilacqua, 1993). As a result, EC activation induces exocytosis of secretory granules known as Weibel–Palade bodies that leads to the release of Ang-2 and VWF, as well as transcriptional programmes that activate expression of adhesion molecules such as ICAM-1, VCAM-1, E-selectin, and secreted cytokines and chemokines (de Mast et al., 2007; Bernardo et al., 2004; Bevilacqua, 1993). However, EC pathophysiology is complex, and changes represent a heterogenous spectrum ranging from simple perturbation to activation and even EC damage (de Mast et al., 2009). Our Luminex data clearly confirm such heterogeneity in the spectrum of EC changes due to P. vivax infection, with systemic increase of markers of EC activation and damage only detected in Vivaxhigh patients. The ex vivo data show that increased systemic host proinflammatory factors and/or parasite products can alter EC properties, including activation of adhesion molecules and proinflammatory cytokines and downregulation of ADAMTS13. In contrast, vascular integrity was not affected. These data indicate that systemic inflammatory responses in P. vivax patients can lead to local EC activation but not vascular damage, central events in malaria pathogenesis. It is likely that other circulating factors that we have not directly measured in our study are also contributing to EC activation and vascular permeability. In particular, extracellular vesicles (EV) originating from ECs, platelets, and RBCs are present during malaria infection and are known to modulate the host immune response to the parasite (Toda et al., 2020; Mantel et al., 2016; Mantel et al., 2013). In P. falciparum, infected RBCs release EVs containing immunogenic parasite antigens, which activate macrophages, induce neutrophil migration, and alter endothelial barrier function (Mantel et al., 2016; Mantel et al., 2013). In P. vivax, plasma-derived EVs from iRBCs are taken up by human spleen fibroblasts (hSFs). This event signals NF-kB translocation and upregulation of ICAM-1 expression, facilitating cytoadherence of P. vivax-infected reticulocytes (Toda et al., 2020).

Although our study lacks longitudinal information, the findings might have clinical implications during and after treatment of vivax malaria. Several case reports demonstrate progressive clinical deterioration after commencement of treatment in P. vivax patients, associated with parasite killing that result in haemolysis of iRBCs and intravascular inflammation and oedema in response to the products released from these cells (Anstey et al., 2007; Anstey et al., 2002; Tan et al., 2008; Val et al., 2017). Patients presenting with a strong host response during acute infection might therefore be at increased risk of deteriorating and developing severe symptoms after commencement of treatment (Figure 5—figure supplement 2). Thus, identification of unique biological signatures in P. vivax patients might help to build rational approaches to the diagnosis, prognosis, and individualized treatment to modulate the host response to vivax malaria.

Altogether, our data indicate that changes in clinical parameters and biomarkers detected in the plasma of P. vivax patients are the result of both systemic host responses and local infection in tissue reservoirs such as BM and spleen. Our analysis shows that measuring a combination of host parameters (e.g. Syndecan-1, IL-6, platelet levels) and total parasite biomass (PvLDH) could predict the extent of parasite infection outside of circulation. Our data also instigate future investigations of systemic signatures with parallel analysis focused on tissue responses, particularly in reservoirs such as the haematopoietic niche of BM and spleen, which has great potential to advance better diagnosis and treatment of P. vivax.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Biological sample (Homo sapiens)Human umbilical vein endothelial cells (HUVEC)Hematology Center, University of Campinas, Campinas, SP, BrazilPrimary cells isolated from the umbilical vein
AntibodyICAM-1 (mouse monoclonal)AbcamCat. #ab2213;RRID:AB_302892Antibody clone MEM-111IF (1:100)
AntibodyVCAM-1 (mouse monoclonal)AbcamCat. #ab212937;RRID:AB_2892824Antibody clone 1.4C3IF (1:500)FC (1:100)
AntibodyIgG1 isotype control (mouse monoclonal)DakoCat. #X0931;RRID:AB_2892825IF (1:10)
AntibodyICAM-1 (mouse monoclonal)BioLegendCat. #322714;RRID:AB_535986Antibody clone HCD54FC (1:100)
AntibodypLDH Vivax-specific (mouse monoclonal)Vista Diagnostics International LLC, WACat. #3h8;RRID:AB_2892826Antibody clone 3h8ELISA (1 μg/mL)
AntibodypLDH detection antibody (goat monoclonal)Vista Diagnostics International LLC, WACat. #6c9;RRID:AB_2892827Antibody clone 6c9ELISA (1:4000)
Sequence-based reagentqRT-PCR OligonucleotidesThis studySee Supplementary file 2
Commercial assay or kitCustomized multiplex suspension detection systemR&D Systems
Commercial assay or kitAccutase Cell Detachment SolutionBioLegendCat. #423201
Chemical compound, drugFixable Viability Dye eFluor 506ThermoFisherCat. #65-0866-14
Software, algorithmFlowJo software (v10)Ashland, ORhttps://www.flowjo.com
Software, algorithmRStudio software (v1.4.1106)RStudio, Boston, MAhttps://www.rstudio.com
Software, algorithmCytoscape software (v3.8.1)NIGMS, Bethesda, MDhttps://cytoscape.org
Software, algorithmGraphPad Prism 9 (v9.1.1 (223))GraphPad Software, San Diego, CAgraphpad.com
Software, algorithmImageJ softwareNIH, Bethesda, MDimagej.nih.gov

Patients

Peripheral blood and plasma samples were collected from 79 patients infected with P. vivax, as diagnosed by light microscopy, seen at FMT-HVD and 34 healthy donors (controls). Patients and healthy donors were age and sex-matched, with a frequency of 30% female and 70% male individuals in both groups. All individuals within the study were from a local vivax malaria epidemic area in the Amazon region of Brazil. All patients included were outpatients that did not meet WHO criteria for severe malaria. Diagnosis was further confirmed by quantitative PCR (qPCR) for both P. vivax and P. falciparum, using previously published nucleotide sequences (Rosanas-Urgell et al., 2010). Excluding other coinfections could have been of interest. However, the differential diagnosis for an acute febrile illness is very broad and it would be impractical to track all other possible diseases. In addition, the patients included in the present work had mild disease, and therefore were discharged from hospital after a positive malaria diagnosis. No further investigation on other infections was done. The main coinfection to be considered for an acute febrile illness with no localizing signs in our context is dengue fever. Although dengue coinfection in our cohort is possible, the incidence at the hospital is only 2.8% (P. vivax/dengue coinfection; Magalhães et al., 2014Magalhães et al., 2014). Thus, it is unlikely that such a coinfection would have a major impact on our findings. Exclusion criteria were (1) under 18 years of age, (2) pregnancy, (3) use of antimalarials, (4) chronic disease, (5) medication known to interfere with platelet count/function, and (6) smoking.

Anaemia is defined as haemoglobin <12.5 g/dL; haematocrit <37%; RBCs counts <4.45 × 106/μL. Thrombocytopenia is defined as a decrease in platelet counts to <150,000/μL. Based on platelet levels, patients were grouped into (1) non-thrombocytopenia (NT: platelet counts >150,000/μL), (2) mild thrombocytopenia (MT: platelet counts 100,000–150,000/μL), (3) moderate thrombocytopenia (MDT: platelet counts 50,000–100,000/μL), and (4) severe thrombocytopenia (ST: platelet counts <50,000/μL). Lymphopenia was defined as a lymphocyte count of less than 1000 cells/μL. Neutropenia was defined as a neutrophil count of less than 1500 cells/μL and neutrophilia as a neutrophil count of more than 7000 cells/μL Punnath et al., 2019; van Wolfswinkel et al., 2017.

Preparation of poor platelet plasma

Request a detailed protocol

After signing the informed consent, 20 mL of venous blood were drawn by venepuncture in a syringe with 15% acid citrate dextrose as anticoagulant to minimize in vitro platelet activation. Complete blood counts were done within 15 min of blood sampling with a Sysmex KX21N counter. Whole blood was centrifuged at 180 g for 18 min at room temperature, without brake for gradient formation, to obtain the platelet-rich plasma (PRP). PRP was centrifuged at 100 g for 10 min for removal of residual leukocytes, and subsequently centrifuged at 800 g for 20 min to obtain the platelet pellet. Prostaglandin E1 at 300 nM was used to minimize platelet aggregation. The supernatant was centrifuged at 1000 g for 10 min to obtain platelet-poor plasma (PPP).

Multiplex bead array assay

Request a detailed protocol

The biomarkers were analysed in thawed plasma with a customized multiplex suspension detection system (R&D Systems) for quantification of the following biomarkers:

  • (1) Proinflammatory and myelopoiesis-inducing cytokines: TNF-α, IL-1α, IL-1β, IL-6, IL-8, and G-CSF.

  • (2) EC activation and coagulation markers: ICAM-1, VCAM-1, E-selectin, P-selectin, Ang-1, Ang-2, von Willebrand factor (vWF-A2), CD40L, PAI-1, and ADAMTS13.

  • (3) Glycocalyx breakdown and EC damage marker: Syndecan-1

  • (4) Platelet activation markers: CXCL4 and CXCL7

  • (5) Megakaryocyte differentiation-inducing cytokines: TPO and IL-11; and other proteins such as IL-10, L-selectin, and SCF.

    A representative set of 31 P. vivax patients were selected for the multiplex assay (Figure 1—figure supplement 1). These patients were selected to encompass the wide range of peripheral parasitaemia present in the cohort (260–25,150 infected RBCs/μL) and to match age, gender, and other haematological parameters to those that were not selected. Nine healthy donors matched for age and sex were also selected.

PvLDH ELISA

Request a detailed protocol

To measure PvLDH in patient plasma samples, ELISA was performed using a matching pair of capture and detection antibodies (Vista Diagnostics International LLC, Greenbank, WA). Briefly, 96-well microtiter plate was coated with monoclonal anti-pLDH Vivax-specific (clone 3H8, Vista Diagnostics International LLC; RRID:AB_2892826) at a concentration of 1 μg/mL in PBS (pH 7.4) and incubated overnight at 4°C. The plate was washed and incubated with blocking buffer (reagent diluent) at room temperature for 1 hr. After washing, samples were added and incubated for 2 hr. Next, plates were washed and biotinylated anti-PvLDH detection antibody (clone 6c9, Vista Diagnostics International LLC; RRID:AB_2892827), diluted 1:4000 in blocking buffer, was incubated for 2 hr at room temperature, followed by streptavidin-HRP incubation for 20 min at room temperature. Plates were washed and incubated for 20 min with substrate solution. Optical density was determined at 450 nm. Cut-off of positivity was defined by correcting absorbance values generated in the plasma samples from healthy donors (controls) by blank values (plate controls), with both values being in the same range. Absorbance values higher than controls were considered positive. In parallel, we used schizont extracts to perform standard curves and lower absorbance values were in the range of O.D = 0.03–0.04. All positive patient samples gave O.D. values equal to or higher than 0.05.

PCA and K-means hierarchical clustering

Request a detailed protocol

Haematological parameters (haemoglobin levels, haematocrit, differential blood cell counts), parasite parameters (peripheral parasitaemia by blood smear, parasite load by qPCR, and parasite biomass PvLDH ELISA), and Luminex data (24 biomarkers) from the selected 9 healthy donors and 31 P. vivax patients were normalized to avoid variable-specific bias and z-score values were determined. Since the host response is complex and multidimensional (one dimension per Luminex biomarker), we applied dimension reduction and clustering for ease of downstream analysis. For this, all variables were used as input for PCA to reduce the dimensionality of data using the PCA function in the FactoMineR package in R. For visualization of PCA results, ggplot2, factoextra, and corrplot packages were used. For each PC, we determined which variables are better represented and the contribution (correlation or loading score) of each variable for each (PC). Investigation of eigenvalues and the percentage of explained variances retained by the PCs demonstrated that the first 10 PCs accounted for the variance of the data (Figure 2—figure supplement 1). However, variables were well represented by the first two PCs (Dim 1 and Dim 2), which were therefore retained for further analysis. In parallel, we performed K-means clustering (k) followed by bootstrapping, which produced the most stable clusters with k = 2 (cluster 1 = 21 individuals; cluster 2 = 18 individuals), which seemed to be the most consistent with the data (Figure 2A, Figure 2—figure supplement 1, Figure 2—figure supplement 2, Figure 2—figure supplement 2—source data 1). Figure 2—figure supplement 2—source data 1 contains the numerical data representing cluster stability via bootstrapping. The metrics of interest is jaccard_index, which measures the cluster similarity across bootstrap samples. Similar to the above, k = 2 gives stable clusters for all configurations (jaccard_index 0.9 and 0.86). Using Monte Carlo reference-based consensus clustering (M3C) analysis (M3C function in the M3C package in R) indicated that k = 2 is the optimal number of clusters when using K-means clustering (Figure 2—figure supplement 2C and D), but when determining spectral clusters, different from elliptical k-means clusters, k = 3 gives the best number of clusters (Figure 2—figure supplement 2E–G).

Correlation plots and heatmap visualization

Request a detailed protocol

Heatmaps were created to visualize variable values using R function Complex Heatmap. They represent z-scores using row scaling obtained by centring represented variables with the scale function, followed by column clustering using average cluster method and Euclidean distance metric in R. The same software was used to determine pairwise Pearson’s correlation coefficients between variables by running the function cor in the ggcorrplot package and visualized as a correlogram using R function corrplot in the Hmisc package displaying positive correlations in red and negative correlations in blue using p≤0.01 as a cut-off.

Recursive partitioning decision tree classification and machine learning models

Request a detailed protocol

We used recursive partitioning decision tree classification models to evaluate dominant signatures (attributes) predicting a specific outcome. For decision tree construction, we applied the C4.5 algorithm, using the RWeka, caret (Classification and Regression Training) and e1071 packages or the rpart package in R. First, the library caret is used to create a 10-fold training set to train the model. Then, the algorithm implements decision trees (using the J48 method, which is an open-source Java implementation of the C4.5 algorithm) starting with all instances in the same group, then repeatedly divides the data based on attributes until each item is classified. The attribute on which to divide is selected by information gain, a statistical technique for determining which attribute split will most cleanly divide the data. To avoid overfitting, sometimes the tree is pruned back. In parallel, the algorithm performs k-fold cross-validation to measure the performance of a given predictive model and indicates which one has the higher accuracy. Here, we used k = 10 to yield test error rate estimates that suffer neither from excessively high bias nor from very high variance (James et al., 2013). In parallel, features with mean decrease accuracy larger than six were used for random forest. In the random forest analysis, a thousand trees were built using R package randomForest (version 4.6.14). The normalized additive predicting probability was computed as the final predicting probability. Those selected important features were used for the random forest analysis on the test cohort for model validation.

Stimulation of HUVEC with patients’ plasma pools

Request a detailed protocol

After standardization procedures, primary HUVEC were stimulated or not (mock control) in culture media for 6 hr – to evaluate mRNA expression – or for 18 hr – to evaluate protein expression – with complete EGM-2 medium (Lonza) containing 30% (v/v) plasma pools generated from the three subgroups – healthy control, Vivaxlow and Vivaxhigh – and 3 U/L heparin.

Real-time quantitative RT-PCR

Request a detailed protocol

After 6 hr stimulation, total RNA was isolated from the cell lysate using the miRVana miRNA Extraction kit (Ambion) according to the instructions of the manufacturer. cDNA was synthesized with TaqMan Reverse Transcriptase (Applied Biosystems, Foster City, CA) and mRNA expression of genes were determined by qRT-PCR. Real-time qRT-PCR was performed on an ABI-Prism 7000 PCR cycler (Applied Biosystems) or on the CFX96 Real-Time PCR Detection System (Bio-Rad). Cycling parameters were 95°C for 1 min and then 35 cycles of 95°C (15 s) and 60°C (1 min), followed by a melting curve analysis. The median cycle threshold (Ct) value and 2-ΔΔCt method were used for relative quantification analysis, and all Ct values were normalized to the GAPDH mRNA expression level. Results expressed as means and SEM of biological replicates are shown. The mock sample (HUVECs incubated with culture media only) was used as reference. The oligonucleotides used are described in Supplementary file 2.

Endothelial cell flow cytometry (FC) and immunofluorescence analysis (IFA)

Request a detailed protocol

For IFA, cells were grown in eight-well chambered coverslips (IBIDI) until confluence. After 18 hr stimulation with plasma pools, cells were washed with PBS and fixed/permeabilized with ice-cold 100% methanol for 5 min at –20°C. In brief, cells were incubated with 10% goat serum (ThermoFisher) to avoid secondary antibody nonspecific binding for 1 hr at room temperature and then incubated with specific primary antibodies to human ICAM-1 (mouse monoclonal clone MEM-111; Abcam; Cat. # ab2213; RRID: AB_302892; used at a dilution of 1:100 in 10% goat serum); VCAM-1 (mouse monoclonal clone 1.4C3; Abcam; Cat. # ab212937; RRID: AB_2892824; used at a dilution of 1:500 in 10% goat serum); and mouse IgG1 isotype control (Dako; Cat. # X0931; RRID: AB_2892825; used at a dilution of 1:10 in 10% goat serum) overnight at 4°C. After washing, wells were overlaid for 1 hr with AF488-conjugated secondary antibody (used at a dilution of 1:500 in 10% goat serum) and Hoechst (diluted at 1:2000) at room temperature. For controls, primary antibodies were omitted from the staining procedure and were negative for any reactivity. The chambers were placed at 4°C until use for immunofluorescence assay (IFA). Percentage of positive cells and expression profiles for ICAM-1and VCAM-1 were then determined using ImageJ software (NIH, Bethesda, MD).

In flow cytometry, after 18 hr stimulation with 30% plasma pools, cells were washed 2× with DPBS and treated with Accutase Cell Detachment Solution (BioLegend) at room temperature for up to 3 min or until the cells are detached. Cell count and viability with trypan blue dye were determined and cells were resuspended in ice-cold DPBS without calcium/magnesium, 0.5 mM EDTA, and 10% foetal bovine serum (FBS; Gibco). Cells were incubated with FcBlock (BD Biosciences, San Jose, CA), followed by incubation with unconjugated anti-VCAM (mouse monoclonal clone 1.4C3; Abcam; Cat. # ab212937) or AF488-conjugated anti-ICAM-1 (mouse monoclonal clone HCD54; BioLegend; Cat. # 322714; RRID:AB_535986) or unconjugated mouse IgG1 isotype control (Dako; Cat. # X0931) for 1 hr at 4°C. Cells were then washed and incubated for 1 hr at 4 °C with secondary antibody AF488-conjugated anti-mouse IgG (ThermoFisher). Finally, cells were incubated with Fixable Viability Dye eFluor 506 (ThermoFisher) in DPBS without calcium/magnesium, 0.5 mM EDTA for 30 min at 4°C. Cells were washed and resuspended in buffer and acquired using a BD FACSCelesta cytometer (100,000 events/sample). Percentage of positive cells and expression profiles for ICAM-1and VCAM-1 were then determined by the mean fluorescence intensity using FlowJo software (v10; Ashland, OR).

Ex vivo evaluation of endothelial cell monolayer function

Request a detailed protocol

EC monolayer function was measured using ECIS, an electric cell-substrate impedance sensing system (ECIS Zθ, Applied Biophysics, Troy, NY), as previously described (Santaterra et al., 2020). The system then applies weak alternating currents through the electrode array and continuously measures the ability of the cell monolayer to impede the movement of electrons between adjacent EC (resistance). Briefly, cells were seeded at 2.5 × 105 cells/well on fibronectin-coated (10 µg/mL) eight-well arrays (8WE10, Applied Biophysics) containing interdigitated gold electrodes. ECs were seeded 48 hr before experiments and the resistance started to be recorded after 48 hr. Only wells with resistance >1500 ohms and stable impedance/resistance readings were used. Before stimulation, resistance was continuously monitored for 2 hr to confirm monolayer stability represented by a plateau in the resistance curve. Stimuli (20% v/v pooled plasma in complete medium) was then added to wells under continuous impedance/resistance monitoring for 12 hr. A baseline resistance value was recorded immediately prior to the addition of each stimuli, and results are expressed as a ratio from baseline resistance (normalized resistance).

Network analysis

Request a detailed protocol

The values of each circulating factor measured in the plasma samples, as well as haematological parameters and parasite biomass in healthy donors and P. vivax malaria patients, were input in the RStudio software (version 1.4.1106, 2021) to determine pair-wise Pearson’s correlation coefficients to generate correlation networks and the p-value to test for non-correlation was evaluated using p≤0.05 as a cut-off. In order to analyse the structure of the networks, edges list was generated in R using the functions melt (reshape2 package), graph_from_edgelist (igraph package). Graphs were customized in the Cytoscape software (version 3.8.1) using the force-directed layout, which simulates a system of forces, determined by the correlation strength. In the equilibrium state, edges tend to have uniform length, and nodes that are not connected by an edge tend to be drawn further apart. Network topology and module analysis were performed using the NetworkAnalyzer, jActiveModules, and MCODE plugins in Cytoscape (Cline et al., 2007; Doncheva et al., 2012). Supplementary file 1 shows the results for all parameters quantified in the comparative network topology analysis between the graphs for healthy donors and P. vivax patients.

Statistical analysis

Request a detailed protocol

Fisher’s exact test was used for categorical data. Data normality was checked by Shapiro–Wilk test. Student’s t-test was used to compare means between groups with normally distributed data, and data sets with non-normal distributions were compared using Mann–Whitney test. All tests were performed two-sided using a nominal significance threshold of p<0.05 unless otherwise specified. When appropriate to adjust for multiple hypothesis testing, Tukey’s or Bonferroni corrected multiple comparisons test significance at the p-value<0.05 threshold was performed unless otherwise specified. Data are presented as scatter plots with median and 25–75% interquartile range, box plots showing minimum to maximum range or means and SEM, unless otherwise stated. Analyses were performed and the graphs generated in GraphPad Prism 9 (version 9.1.1 [223], 2021) and RStudio software (version 1.4.1106; 2021). To ensure that differences observed between P. vivax- infected patients and controls, as well as between the clusters, were due to disease status and not confounded by age or sex, the clinical parameters were fitted as response variables in a linear model with sex and/or age fitted as explanatory variables. Age and sex were included in the model if their coefficients were estimated as different from zero with p-value<0.05. The residuals from the linear model were then used as age- and/or sex-corrected parameters in subsequent analyses.

Study approval

Request a detailed protocol

All subjects enrolled in the study were adults. Written informed consent was obtained from all participants, and the study was conducted according to the Declaration of Helsinki principles. The study was approved by the local Research Ethics Committee at Fundação de Medicina Tropical Dr. Heitor Vieira Dourado (FMT-HVD, Manaus, Brazil), under #CAAE: 54234216.1.0000.0005.

Data availability

All data generated or analysed during this study are included in the manuscript and supporting files. Numerical tables and source data files have been provided. Table 1, Figure 2—source data 1 and Figure 2—figure supplement 2—source data 1 contain the numerical data used to generate the figures.

References

    1. Crockett-Torabi E
    2. Sulenbarger B
    3. Smith CW
    4. Fantone JC
    (1995)
    Activation of human neutrophils through L-selectin and Mac-1 molecules
    Journal of Immunology 154:2291–2302.
    1. de Mast Q
    2. Groot E
    3. Asih PB
    (2009)
    ADAMTS13 deficiency with elevated levels of ultra-large and active von Willebrand factor in P
    Falciparum and P. Vivax Malaria. Am J Trop Med Hyg 80:492–498.
    1. James G
    2. Witten D
    3. Hastie T
    4. Tibshirani R
    (2013)
    An introduction to statistical learning with applications in R introduction
    Springer Texts Stat 103:1–14.
    1. Salutari P
    2. Sica S
    3. Chiusolo P
    (1996)
    Plasmodium vivax malaria after autologous bone marrow transplantation: an unusual complication
    Bone Marrow Transplantation 18:805–806.
    1. Wickramasinghe SN
    2. Abdalla SH
    (2000) Blood and bone marrow changes in malaria
    Bailliere’s Best Practice & Research. Clinical Haematology 13:277–299.
    https://doi.org/10.1053/beha.1999.0072
  1. Software
    1. World Malaria Reports
    (2019)
    World Malaria Reports
    World Malaria Reports.

Decision letter

  1. Urszula Krzych
    Reviewing Editor; Walter Reed Army Institute of Research, United States
  2. Dominique Soldati-Favre
    Senior Editor; University of Geneva, Switzerland
  3. Rays HY Jiang
    Reviewer; University of South Florida, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

[Editors' note: this paper was reviewed by Review Commons.]

Acceptance summary:

In contrast to Plasmodium falciparum, which is the deadliest of all malaria parasites, Plasmodium vivax which causes nearly 40% of malaria cases outside of the sub Saharan Africa has been the less studied parasite. The unique feature of P. vivax is its dormant phase, not detectable in the periphery, and which allows the parasite to emerge, causing a recurrent episode of malaria infection. In this study, the authors observed profound differences in immune responses amongst subjects with low versus high total biomass of P. vivax, which points to a reservoir of Plasmodium parasites outside of the peripheral circulatory system as having a profound affect on disease severity.

Decision letter after peer review:

Thank you for resubmitting your work entitled "Total parasite biomass but not peripheral parasitaemia is associated with endothelial and haematological perturbations in Plasmodium vivax patients" for further consideration by eLife. Your revised article has been evaluated by Dominique Soldati-Favre (Senior Editor), a Reviewing Editor and 1 reviewer. The following individual involved in review of your submission have agreed to reveal their identity: Rays Jiang (Reviewer #4).

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

Please respond to the reviewers queries numbered 1 – 4.

Reviewer #4:

The manuscript by by Silva- Filho et al., entitled: "Total parasite biomass but not peripheral parasitaemia is associated with endothelial and haematological perturbations in Plasmodium vivax patients" reports the importance of total parasite mass of P. vivax infection, and its relationship with a variety of host hematological and immunological markers. The study classifies infection based on clinical features, parasitemia, and host features. I have the following questions:

1. The use of PvLDH in patient plasma as a proxy for parasite biomass has been established before. But the use of the biomass marker in combination with other host markers are novel. Can the authors explains more in-depth why the suspected deep tissue mass contribute to the PvLDH levels from different host tissues? Because the conclusion of this study relies on this parameter, an in-depth understanding of this marker is necessary.

2. One of the important conclusion is the severity of the infection is parasite-biomass dependent. While the current study does not include severe vivax cases, are there studies support the author's conclusion?

3. Results Page17. The authors found that the total parasite biomass was higher and not correlated with parasitemia, particularly in patients of cluster 2. Why is cluster 2 strong in this aspect?

4. Network and module analysis. The results seem to show that the healthy donor and the patients have different network structure and module numbers? Can the authors explain more on it?

https://doi.org/10.7554/eLife.71351.sa1

Author response

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

This study focused on P. vivax, which is an important neglected human malaria killer. The reported evidence will have a significant impact on diagnosing infectious diseases. The language in the manuscript is very good. However, some typos were reported. Some paragraphs might need particular attention to punctuation. Overall, the work is very good. The statistics are straight forward. However, there are a couple of major points that must be addressed before publication. Some of my comments are just recommendations to clarify some sections of the text.

Major comments:

The statistical methods can be improved by using generalised mixed models (GLMM).

1. PCA graphs need to be organised in more descriptive ways. Dim1 and Dim2 in each axis need to be defined clearly in the figures. PCA in Figure 2 c is very difficult to follow, and it needs to be organised.

Figures have been amended to be more self-explanatory and clearer to the reader.

2. In this study, patients were male and female, and we know already male and female haematological parameters are hugely different, specially Hb level, and so on. My question is how the sex variable is treated in this study? Did your control group were from both sexes? Sex could be treated as a random variable in all studies if GLMM models were used.

Information in how the sex variable was treated in the study has been added to the methods section. In our cross-sectional study with uncomplicated P. vivax malaria patients seen at FMT-HVD in Manaus, Brazil, patients and healthy donors (controls) were matched by age and sex. In both groups, frequency of female individuals was 30% and male individuals 70%.

We think sex is better fitted as fixed effect since only two levels for this factor are possible. Thus, we used linear models with age and sex as fixed variables for statistical testing and to ensure that the differences observed between P.vivax- infected patients and controls, as well as between the clusters, were only due to disease status. This analysis showed that red blood cells count, hemoglobin, hematocrit, MXD and neutrophils counts (this parameter only when comparing the clusters) needed to be corrected only due to sex influence. For these parameters, estimates of predicted sex influence were subtracted from the raw parameter values and residuals were used for statistical testing. We have added this information in the Methods section as indicated below:

“Patients and healthy donors were age and sex-matched, with a frequency of 30% female and 70% male individuals in both groups.”

“To ensure that differences observed between P. vivax – infected patients and controls, as well as between the clusters, were due to disease status and not confounded by age or sex, the clinical parameters were fitted as response variables in a linear model with sex and/or age fitted as explanatory variables. Age and sex were included in the model if their coefficients were estimated as different from zero with p-value < 0.05.”

The residuals from the linear model were then used as age and/or sex corrected parameters in subsequent analyses.

3. Why 6h and 18h used for the HUVEC evaluation?

We ran several optimization experiments with individual plasma samples where we observed maximal mRNA expression changes after 6h of stimulation. For experiments detecting protein expression (IFA and flow cytometry), we increased the stimulation time to 18h. Preliminary experiments suggested this to be the optimal duration without compromising cellular viability.

4. It is mentioned only neutrophil enriched in this study, if myelopoiesis is affected, why the other granulocytes were not showed significant enhancement?

Our data reveal no change in the number of circulating neutrophils in the different clusters of individuals. However, mixed cell counts (MXD), a parameter representing monocytes, basophils and eosinophils numbers, was significantly reduced in Vivaxhigh patients. As a result, there was a significant enrichment of neutrophils in the leukocyte fraction in the blood of Vivaxhigh patients as well as a higher Neutrophil:Lymphocyte count ratio (NLCR) (Figure 4). In hematopoietic progenitors, stochastic changes in each factor’s concentration could result in one factor’s becoming more abundant and committing a hematopoietic progenitor to a particular lineage. To generate each mature granulocyte population (e.g. basophils, eosinophils and neutrophils), common myeloid precursor cells (CMPs) and later precursors for granulocytic and monocytic lineages (GMPs) follow in the BM different lineage commitment programs, tightly-regulated or instructed by a specific set of soluble factors, cell-cell interactions and transcription factors, that define cell fate decisions and lineage restrictions. For instance, differential PU.1 activity can specify different cell fates during haematopoiesis regulating monocyte and neutrophils differentiation. Genetic and biochemical analyses have shown that G-CSF can direct granulocyte differentiation by changing the ratio of C/EBPα to PU.1 (Zhu et al., Oncogene 2002; Friedman Oncogene 2002; Dahl et al., Nat Immunol 2003). High expression levels of PU.1 and C/EBPα, stimulated by G-CSF, promote GMP differentiation to neutrophils and inhibits monocyte differentiation, while only PU.1 expression, IRF-8 and lower expression/activity of C/EBPs induce GMP differentiation to monocytes (Zhu et al., Oncogene 2002; Friedman Oncogene 2002; Dahl et al., Nat Immunol 2003). Meanwhile, a combination of PU.1, C/EBPβ and low levels of GATA-1 differentiates GMPs to eosinophil lineage (Kulessa et al., 1995; McDevitt et al., 1997; Yamaguchi et al., 1999) and PU.1 must also cooperate with GATA2 to direct mast cell differentiation (Walsh et al., Immunity 2002). In addition, eosinophil and basophil differentiation are induced by a different set of cytokines, usually produced in prevalent T-helper 2 response, such as IL-5, which should be inhibited in the strong Th1 environment evidenced by our and previous Luminex data in Pv patients. The enrichment of activated neutrophils in the peripheral circulation of P. vivax patients could be due to a response that specifically enhances neutrophil production and release from the bone marrow (BM). This hypothesis is supported by emerging evidence for enrichment of P. vivax parasites in the hematopoietic niche of BM, our Luminex data showing significant increase in pro-inflammatory cytokines associated with emergency myelopoiesis (e.g., TNF-α, IL-1α, IL-1β, IL-6, IL-8), and increased circulating levels of G-CSF, the major inducer of neutrophils production in the BM. Likewise, increased activation-induced cell death (AICD) in T cells, splenic T-cell and platelet accumulation or decreased lymphopoiesis due to myeloid-biased HSC differentiation induced by inflammatory cytokines and EC activation in the BM (refs 36,37,39) might explain the neutrophil enrichment in vivax patients.

5. I would also ask the authors to speculate a bit on, What could be the mechanism behind the different function of P. vivax compared to P. falciparum? From an evolutionary perspective, the parasite should rather become softer and keep the host alive for its own benefit.

One of the characteristics of P. vivax that could play an important role in immunity is its restriction to invade immature reticulocytes. For example, the infected reticulocyte could play a role in the presentation of parasite antigens as reticulocytes (but not mature RBCs) express MHC-I and are capable to process and present antigens on their surface for recognition by T cells. Indeed, it has been shown that reticulocytes act directly as an antigen-presenting cell, emphasizing the importance of erythrocyte surface antigens both in the induction as well as the target of a protective immune response (Burel et al. 2016, Junqueira et al. 2018). Recent investigations comparing P. vivax and P. falciparum controlled human infection models (CHMIs) also revealed marked differences in the immune profiles generated following infection with the two species and postulated that protective immune responses to Plasmodium are species-specific. It has been hypothesized that this difference is due to strict P. vivax tropism for MHC-I-expressing reticulocytes that, unlike mature red blood cells, can present antigen directly to CD8+T cells. Specifically, P. vivax but not P. falciparum infection led to the expansion of a specific subset of CD38+CD8+ T cells which were associated with an activated phenotype and cytotoxic potential. Corroborating Burel et al. findings in the CHMI model, Junqueira et al. showed that P. vivax–infected reticulocytes express HLA-I. In P. vivax-infected patients, CD8+ T cells in the peripheral blood express high levels of cytotoxic proteins, recognize and form immunological synapses with P. vivax–infected reticulocytes in HLA–dependent manner. Next, it was showed that P. vivax-specific CD8+ T cells release their cytotoxic granules to kill both host cell and intracellular parasite, which prevented reinvasion (Junqueira et al. 2018). Although these data indicate a protective role of cytotoxic CD8+ T cells during P. vivax blood-stage malaria, it is not clear whether these lymphocytes would always be beneficial because they might contribute to anemia, inflammation or other pathological sequelae of infection, which needs to be further investigated.

Minor comments:

1. It is important to have a reference, version, and date for the R software, packages and GraphPad.

We have added version and date for the R and GraphPad software.

2. In Figure 5, E missed to report. This figure can be better organised. It is very hard to read and follow.

There is no E in Figure 5. We will organize the figure to make it easier to read and follow.

Reviewer #2 (Significance (Required)):

P. vivax remains endemic in 51 countries across Central and South Americas, the Horn of Africa, Asia and the Pacific islands. In most areas it is co-endemic with P. falciparum, which has been the priority species to address for national malaria control programmes. Malaria related deaths are mostly attributable to the more pathogenic P. falciparum, but over the last decade these have declined, however there has been a consistent rise in the proportion of malaria cases due to P. vivax. However, because it is difficult to diagnose resistant strains, strategies to detect and track drug resistant P. vivax are limited. In this context it is vital to develop better tools to assess diagnostic, antimalarial efficacy and drug susceptibility so that emerging drug resistance can be tracked, and novel treatment strategies explored.

From my viewpoint, despite some statistical problems to understand the complex nature of data (mixed interactions among multiple variables), these findings seem to be very interesting and (after a major revision) worth to be published. As said before, the story told by the authors could become interesting.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

The manuscript titled: "Total parasite biomass but not peripheral parasitaemia is associated with endothelial and haematological perturbations in Plasmodium vivax patients" by Silva-Filho et al., reinforce the original observation and data by the group of Nicholas Anstey and coworkers, who first proposed the use of plasma parasite lactate dehydrogenase and PvLDH as a marker of parasite biomass. In that work, it was already demonstrated that P. vivax biomass is related to plasma concentration of LDH levels. As such, the present work cannot be considered of high novelty. Yet, through a meticulous approach including clinical data, computational approaches, machine learning, LDH measurement, multiplex analysis and quantitave RT-PCR, the authors here have extended the original observations that a large biomass of P. vivax parasites is out of blood circulation. In contrast, unlike the original observations of Anstey´s group, a correlation between total parasite biomass and systemic levels of markers of endothelial cells activation, was observed. The manuscript is very well written and the discussion brings new knowledge in this key topic for elimination of malaria. This manuscript is therefore recommended for publication after the following comments are addressed.

Major comments:

1. The vascular endothelium plays a pivotal role in malaria. Therefore, to test whether cell and/or parasite factors affect the vascular endothelium, HUVEC cells were used in this study. This is of major concern as endothelial cells from the bone marrow, where most hematological disturbances, notoriously thrombocytopenia, occur, were not used instead. HUVEC cells seems the only endothelial cell that does not express ABO blood group antigens, thus suggesting that surface expression on these cells is highly altered (O´Donnell et al., 2000 J Vasc Res). Moreover, significant functional differences between HUVEC cells and adult vascular endothelium have been reported (Chan et al., 2004). Together, this indicates that results obtained with HUVEC cells might not reflect responses of the bone marrow vascular endothelium. As one of the corresponding authors have ample experience with working with human bone marrow endothelial cells (Mantel et al., 2016 Nat Comm), it is suggested to perform some experiments with these cells to assure extrapolation of the results obtained with HUVEC cells.

We agree with the reviewer that performing ex vivo assays with primary human bone marrow endothelial cells would be an excellent alternative. However, we would like to argue that HUVECs are also suitable for our purposes. HUVECs are widely used to study endothelial barrier function, for example in the context of angiogenesis and inflammatory responses/barrier disruption. To emphasise this point, we have now referenced examples where HUVECs were used in the context of endothelial barrier biology and in different inflammatory conditions (see also lists a, b, c below).

A) Papers showing the use of HUVECs in studies yielding important insights about endothelial barrier function

– Krispin S et al. Growth Differentiation Factor 6 Promotes Vascular Stability by Restraining Vascular Endothelial Growth Factor Signaling. Arterioscler Thromb Vasc Biol. 2018.

– Aranda JF et al. MYADM controls endothelial barrier function through ERM-dependent regulation of ICAM-1 expression. Mol Biol Cell. 2013.

– Orsenigo F et al. Phosphorylation of VE-cadherin is modulated by haemodynamic forces and contributes to the regulation of vascular permeability in vivo. Nat Commun. 2012.

B) Papers that used HUVECs in studies about endothelial barrier function in inflammatory conditions

– Dickinson CM et al. Leukadherin-1 ameliorates endothelial barrier damage mediated by neutrophils from critically ill patients. J Intensive Care. 2018.

– Kuck JL et al. Ascorbic acid attenuates endothelial permeability triggered by cell-free hemoglobin. Biochem Biophys Res Commun. 2018.

– Tramontini Gomes de Sousa Cardozo F et al. Serum from dengue virus-infected patients with and without plasma leakage differentially affects endothelial cells barrier function in vitro. PLoS One. 2017.

– Fox ED et al. Neutrophils from critically ill septic patients mediate profound loss of endothelial barrier integrity. Crit Care. 2013.

– Rahbar E et al. Endothelial glycocalyx shedding and vascular permeability in severely injured trauma patients. J Transl Med. 2015.

C) Papers showing that HUVECs behave similarly to other endothelial cell types in regard to barrier function, except when the comparison is with blood brain barrier models

– Totani L et al. Mechanisms of endothelial cell dysfunction in cystic fibrosis. Biochim Biophys Acta Mol Basis Dis. 2017, Dec;1863(12):3243-3253.

– Gündüz D et al. Effect of ticagrelor on endothelial calcium signalling and barrier function. Thromb Haemost. 2017 Jan 26;117(2):371-381.

– Deitch EA et al. Mesenteric lymph from rats subjected to trauma-hemorrhagic shock are injurious to rat pulmonary microvascular endothelial cells as well as human umbilical vein endothelial cells. Shock. 2001 Oct;16(4):290-3.

Importantly, we were able to reproduce in the HUVEC ex vivo assays a phenotype of endothelial perturbations that is inferred based on the in vivo Luminex data using the same plasma sample. These data also support our hypothesis that patients with higher parasite biomass present higher endothelial cell perturbations, corroborating the associations between parasite accumulation in deep tissues (total parasite biomass represented by PvLDH levels) and endothelial cell activation as demonstrated in the Figure 6.

2. Strikingly, the authors stated that "P. vivax infection results in different ranges of EC alterations without massive cytoadhesion". This statement has no data supporting it. In fact, their own flow cytometry data convincingly demonstrated that exposure of HUVEC cells to plasma of vivax-high patients significantly increased the surface expression of ICAM-1 and VCAM. ICAM-1 expression is a well know receptor for cytoadhesion in malaria and Dr. Costa first demonstrated the importance of this receptor in cytoadherence of P. vivax (Carvalho et al., 2010). Moreover, these data are in some contradiction with the original observations of Anstey and collaborators who demonstrated that parasite LDH concentration did not correlate with markers of endothelial activation (Barber et al., 2015 PLoS Path). Therefore, this sentence should be modified to accommodate the alternative possibility of cytoadherence, deleted from the manuscript or binding functional assays should be performed to sustain it.

We agree with the reviewer and have removed this statement.

“The association between endothelial activation, Syndecan-1 and parasite biomass (PvLDH) indicates a positive feedback loop between glycocalyx breakdown, activation of endothelial receptors such as ICAM-1and VCAM-1 and parasite accumulation in deep tissues9,12.”

3. Extracellular vesicles are key players in pathology of malaria and this includes P. vivax where concentration of circulating microparticles were associated with acute infections (Campos et al., 2010 Mal J). Moreover, Dr. Marti has pioneered this field since the original manuscript describing the role of EVs in malaria as intercellular communicators (Mantel et al., 2013 Cell). More recently, his group also demonstrated that interaction of EVs with bone marrow endothelial cells induce expression of IL-6 and IL-1 as well as vascular endothelium perturbations after trans-endothelial electrical resistance experiments (Mantel et al., 2016 Nat Comm). Furthermore, another recent report showed the physiological role of EVs in vivax malaria by demonstrating that EV uptake by human spleen fibroblast induced nuclear translocation of the NF-κB transcriptional factor, concomitant with surface expression of ICAM-1, thus facilitating cytoadherence of infected reticulocytes from P. vivax patients (Toda et al., 2020 Nat Comm). This growing evidence indicates that plasma circulating EVs are key communicators in malaria infections potentially explaining some of the findings reported in this work. Neglecting the importance of EVs in the discussion of this article is not reasonable and weakens this manuscript. Including a paragraph on EVs and accurate references in the discussion is thus strongly recommended.

We agree with the reviewer that extracellular vesicles are key communicators in malaria infection. We have not measured them in our study, however, and therefore can only speculate about their impact on our observations. We have added a phrase in the discussion:

“It is likely that other circulating factors that we have not directly measured in our study are also contributing to EC activation and vascular permeability. In particular, extracellular vesicles (EV) originating from ECs, platelets, and RBCs are present during malaria infection and are known to modulate the host immune response to the parasite54-56. In P. falciparum, infected RBCs release EVs containing immunogenic parasite antigens, that activate macrophages, induce neutrophil migration and alter endothelial barrier function54,55. In P. vivax, plasma-derived EVs from iRBCs are taken up by human spleen fibroblasts (hSFs). This event signals NF-κB translocation and upregulation of ICAM-1 expression, facilitating cytoadherence of P. vivax-infected reticulocytes56.”

Minor comments:

1. The lack of a group including severe vivax malaria patients is a drawback of this article as this group would have firmly validated the predictor of severe disease.

This study was investigating a cohort of uncomplicated P. vivax malaria compared to controls. We agree that it will be important to extent our analysis to severe vivax malaria in future studies.

2. In the selection criteria of the patients to be included in the study, no information on other co-infections were mentioned. Is this information available? If so, this should be mentioned.

As described in the Methods sections, Page 6, line 132, mono infection by P. vivax was confirmed by analysis of blood smears and quantitative PCR (qPCR) for both P. vivax and P. falciparum. We agree that excluding other coinfections could have been of interest. However, the differential diagnosis for an acute febrile illness is very broad and it would be impractical to track all other possible diseases. In addition, the patients included in the present work had mild disease, and therefore were discharged from hospital after a positive malaria diagnosis. No further investigation on other infections was done.

The main coinfection to be considered for an acute febrile illness with no localizing signs in our context is Dengue Fever. Although Dengue coinfection in our cohort is possible, the incidence at the Hospital is only 2.8% (P. vivax/Dengue coinfection) (Magalhães et al., Plos NTD 2014). Thus, it is unlikely that such a coinfection would have a major impact on our findings.

3. This work determined the levels of PvLDH in a cohort of uncomplicated P. vivax patients as well as healthy volunteers using a double-sandwich ELISA assay: (i) are the clones to determine PvLDH values freely available to facilitate similar studies by independent groups? (ii) How was the cut-off of positivity defined? This is not evident, neither in the Materials and methods, nor in the results.

Clones are commercially available and were purchased from Vista Diagnostics International LLC, WA, USA. Information has been amended to the text in the Methods section.

Cut-off of positivity was defined by correcting absorbance values generated in the plasma samples from healthy donors (controls) by blank values (plate controls), with both values being in the same range. Absorbance values higher than controls were considered positive. In parallel, we used schizont extracts to perform standard curves and lower absorbance values were in the range of O.D = 0.03-0.04. All positive patient samples gave O.D. values equal or higher than 0.05.”

4. It is not clear why varying percentages of pooled plasma (30% for imaging and flow cytometry, and 20% for impedance changes) from the different clusters were used for the functional EC assays. Moreover, no information about the concentration of plasma used for transcriptional analysis is available. Please clarify.

The concentration of 30% pooled plasma was also used for transcriptional analysis, as indicated in the Methods section, page 11, line 250. This information was also added in the legend of Figure 5B. We had run several optimisation time-course and titration experiments with individual plasma samples, testing concentrations of plasma varying from 10% up to 30% v/v and we did not observe differences in mRNA expression between 20% and 30% v/v plasma conditions.

As for the ECIS, our collaborators (Erich V de Paula group) have optimised this assay and they use a range of 15 to 20% (Santaterra et al. 2020). Higher concentrations of plasma reduces the reproducibility, probably to fibrin formation.

5. Reference 9 is a nonhuman primate study where no LDH is used. Please remove it.

Reference 9 has been removed following the reviewer suggestion.

6. Reference 39 is a review on the subject and cannot be included in the sentence on line 556 In agreement with a previous study8,39, where reference 8 is accurate. Please remove reference 39 from here.

The text has been amended as suggested.

Reviewer #3 (Significance (Required)):

This paper further contributes to explain the conundrum of low peripheral blood parasitemia and clinical severity in P. vivax. Moreover, by including new human markers and solidly applying computational tools, this paper further contributes to advance clinical research in P. vivax.

Clinical diagnosis of hematological disorders including anemia, lymphopenia and thrombocytopenia, are routinely obtained from a complete blood count. Therefore, I believe the major significance of this work is to raise public health awareness of including in these clinical examinations, the determination of PvLDH levels. They might prognose, as suggested by the authors, better diagnosis and treatment of P. vivax,

My main expertise is the biology of host-pathogen interactions with a focus on P. vivax.

Reviewer #4 (Evidence, reproducibility and clarity (Required)):

The study evaluates P. vivax biomass (serum LDH) versus peripheral parasitemia with multiple variables. From the biomass Vivax high vs. Vivax low, they compare multiple determination in patients with uncomplicated P. vivax. This raises questions about disease and the presence of parasites in various organs. The question is if P. vivax sequesters and the answer is yes in the bone marrow and spleen. Does it sequester like P. falciparum that causes disease by sequestration by binding endothelium in various organs. That is less clear. As P. vivax is rarely fatal, the sequestration has not been studied. The presence of parasites in organs of P. vivax infected splenectomized squirrel and Aotus monkeys has been found in bone marrow and liver (note: splenecotomized monkeys so parasitemia can rise to higher levels than in non-splenectomized monkeys). There are studies of binding of schizonts infected red cells to lung endothelium in vitro does not answer the question of whether sequestration occurs in vivo.

The most important complication of P. vivax is generally anemia. This did not correlate with vivax biomass, but this raises the question of the length of infection and the possibility that parasite biomass may vary at different times of infection. Anemia was seen in P. vivax infected patients, but it did not relate to biomass at the time of study. Note the caveat mentioned in the previous sentence on long term effects of infection on anemia.

The finding of biomass with reduced platelet counts and endothelial effects that may be related to a serum factor and not sequestration. This is the main limitation of the paper besides the unknown long term effect infection. If one could identify an effect of P. vivax infected human serum, this may be worth a study in the future on what is in serum causing the effects.

Reviewer #4 (Significance (Required)):

This study is unique with the caveats mentioned above. It has a good review of the literature.

We appreciate the reviewer comments. In our cohort, the frequency of anaemia was not as high or severe as the frequency of thrombocytopenia and lymphopenia. However, we still find associations between endothelial cell activation marker Ang-2 and the pro-inflammatory cytokine IL-1 IL-1 negatively associated with several markers of anaemia, such as haemoglobin, haematocrit and RBC numbers. Although we did not further investigate this association, it may indicate indirect effects of parasite biomass on anaemia mediated by inflammation and EC activation, which will be further investigated in other current longitudinal cohort studies.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #4:

The manuscript by by Silva- Filho et al., entitled: "Total parasite biomass but not peripheral parasitaemia is associated with endothelial and haematological perturbations in Plasmodium vivax patients" reports the importance of total parasite mass of P. vivax infection, and its relationship with a variety of host hematological and immunological markers. The study classifies infection based on clinical features, parasitemia, and host features. I have the following questions:

1. The use of PvLDH in patient plasma as a proxy for parasite biomass has been established before. But the use of the biomass marker in combination with other host markers are novel. Can the authors explains more in-depth why the suspected deep tissue mass contribute to the PvLDH levels from different host tissues? Because the conclusion of this study relies on this parameter, an in-depth understanding of this marker is necessary.

Low peripheral parasitaemia of P. vivax infections (generally below 2%) compared to P. falciparum has been attributed to the strict tropism of P. vivax to infect mostly immature reticulocytes, a red blood cell (RBC) population that it is largely confined to the erythropoietic niche of the bone marrow (~0.016% of all immature reticulocytes are in the circulation) [1, 2]. Such low peripheral parasitaemia has long been considered to indicate a low total parasite biomass in vivax malaria. More recently, existence of a significant P. vivax reservoir outside of circulation was predicted by disproportionately high PvLDH levels in peripheral circulation compared to parasitemia by blood smear (in particular in patients with complicated outcomes), and by modelling using experimental P. cynomolgy infections in non-human primates [3, 4]. A series of studies, including from our labs, have meanwhile provided direct evidence that bone marrow and spleen represent the major reservoir of parasite biomass in P. vivax infection [5-9].

PvLDH is produced by viable or recently killed parasites and hence considered a proxy for ongoing rather than past infection [3, 10]. PvLDH antigen capture ELISA established a direct relationship between pLDH levels and P. vivax parasitemia in ex vivo experiments, demonstrating that pLDH reflects total P. vivax parasite biomass [10]. Our study further explores the relevance of PvLDH beyond determination of parasite biomass, establishing it as a prognostic marker of host perturbations and disease severity, with a particular focus on markers of changes in the hematopoietic niches of bone marrow and spleen.

We have amended the text in the Introduction and Discussion sections as indicated below:

“A series of recent studies in acute and chronic human P. vivax infection have meanwhile provided direct evidence that bone marrow and spleen represent the major reservoir of parasite biomass in P. vivax infection 17,20-22.”

“The existence of a significant P. vivax reservoir outside of circulation was first predicted by disproportionately high PvLDH levels in peripheral circulation compared to parasitemia by blood smear (in particular in patients with complicated outcomes), and by modelling using experimental P. cynomolgy infections in non-human primates8,36. Recent studies provide direct evidence that bone marrow and spleen represent the major reservoir of parasite biomass in P. vivax infection11,17,20-22. PvLDH is produced by viable or recently killed parasites and hence considered a proxy for ongoing rather than past infection8,48. PvLDH antigen capture ELISA established a direct relationship between pLDH levels and P. vivax parasitemia in ex vivo experiments, demonstrating that pLDH reflects total P. vivax parasite biomass48. Our study further explores the relevance of PvLDH as a prognostic marker of host perturbations and disease severity, with a particular focus on markers of changes in the hematopoietic niches of bone marrow and spleen.”

2. One of the important conclusion is the severity of the infection is parasite-biomass dependent. While the current study does not include severe vivax cases, are there studies support the author's conclusion?

This is a key question. In contrast to P. falciparum-infected individuals, a wide range of complicated clinical syndromes occurs in P. vivax patients even at low or subpatent parasitemia [11] – thus indicating that peripheral parasitemia is a poor predictor of clinical outcomes. Two lines of evidence support our conclusion that severity of infection is dependent on parasite biomass instead. First, the discrepancy between PvLDH levels and peripheral parasitaemia determined by blood smears is more evident in P. vivax-infected patients with complicated outcomes: the ratio of plasma pLDH to peripheral parasitaemia is 6-fold higher than in non-severe patients. The same comparison between severe and non-severe P. falciparum patients reveals only a 1.4-fold difference [3]. Second, severe thrombocytopenia (platelet counts under 50,000/μL) and lymphopenia (lymphocyte counts under 500/μL) have been associated with severity in vivax patients, suggesting that these haematological complications could be explored as a marker of severity for this species [12-15]. Both thrombocytopenia and lymphopenia were more frequent in patients in the Cluster 2 (Vivaxhigh) in our study. By integrating these clinical perturbations with host biomarker measurements and parasite parameters, we demonstrated the high attribute value of total parasite biomass in predicting the severity of thrombocytopenia and lymphopenia (Figures S7 and S8), and highly significant correlations with endothelial activation, glycocalyx breakdown and other markers of inflammation (Figure 6). Our ongoing studies aim to directly investigate the role of host parasite interactions in the hematopoietic niches in P. vivax pathogenesis and severity.

The following text has been added in the Discussion section:

“In contrast to P. falciparum-infected individuals, a wide range of complicated clinical syndromes occurs in P. vivax patients even at low or subpatent parasitemia53 – thus indicating that peripheral parasitemia is a poor predictor of clinical outcomes. Two lines of evidence support our conclusion that severity of infection is dependent on parasite biomass instead. First, the discrepancy between PvLDH levels and peripheral parasitaemia determined by blood smears is more evident in P. vivax-infected patients with complicated outcomes: the ratio of plasma pLDH to peripheral parasitaemia is 6-fold higher than in non-severe patients. The same comparison between severe and non-severe P. falciparum patients reveals only a 1.4-fold difference8. Second, although thrombocytopenia and lymphopenia are not included in the World Health Organization (WHO) criteria for defining severe malaria, it has been associated with severe manifestations and the need for blood and platelet transfusions in severe vivax malaria. This points out their clinical relevance in malaria diagnosis and treatment 24,25,54-56, suggesting that these haematological complications could be explored as markers of severity for this species. Both severe thrombocytopenia and lymphopenia were more frequent in patients in the Cluster 2 (Vivaxhigh) in our study. By integrating these clinical perturbations with host biomarker measurements and parasite parameters, we demonstrated the high attribute value of total parasite biomass in predicting the severity of thrombocytopenia and lymphopenia and highly significant correlations with endothelial activation, glycocalyx breakdown and other markers of inflammation.”

3. Results Page 17. The authors found that the total parasite biomass was higher and not correlated with parasitemia, particularly in patients of cluster 2. Why is cluster 2 strong in this aspect?

The highly significant and positive association between endothelial activation, Syndecan-1 and parasite biomass (PvLDH) indicates a positive feedback loop between glycocalyx breakdown, activation of endothelial receptors such as ICAM-1and VCAM-1 and parasite accumulation in deep tissues. Patients in Cluster 2 show higher plasma levels of all these markers. Consistent with previous reports [16, 17], we propose that elevated EC activation and glycocalyx damage increases the exposure of adhesion molecules, which in turn favours endothelial cytoadherence of P. vivax-infected RBCs, in particular in the splenic red pulp cords and in the BM [18-20]. Accordingly, application of a best-fit classification tree model identifies Syndecan-1 is a putative host biomarker (EC glycocalyx breakdown marker) predicting total parasite biomass in P. vivax patients (Figure 6D). We hypothesise that elevated endothelial activation and damage in patients of Cluster 2 results in increased cytoadherence of P. vivax iRBCs and hence accumulation and growth in deep tissues – thus reducing the fraction of the parasite biomass in circulation.

The following text has been added in the Discussion section:

“Furthermore, the highly significant and positive associations between endothelial activation, Syndecan-1 and parasite biomass (PvLDH) indicates a positive feedback loop between glycocalyx breakdown, activation of endothelial receptors such as ICAM-1and VCAM-1 and parasite accumulation in deep tissues. VivaxHigh patients show higher plasma levels of all these markers. Consistent with previous reports 44,49, we propose that elevated EC activation and glycocalyx damage increases the exposure of adhesion molecules, which in turn favours endothelial cytoadherence of P. vivax-infected RBCs, in particular in the splenic red pulp cords and in the BM 21,50-52. Accordingly, application of a best-fit classification tree model identifies Syndecan-1 is a putative host biomarker (EC glycocalyx breakdown marker) predicting total parasite biomass in P. vivax patients. We hypothesise that elevated endothelial activation and damage in VivaxHigh patients results in increased cytoadherence of P. vivax iRBCs and hence accumulation and growth in deep tissues, thus reducing the fraction of the parasite biomass in circulation.”

4. Network and module analysis. The results seem to show that the healthy donor and the patients have different network structure and module numbers? Can the authors explain more on it?

We have included in the Methods and Results sections details about the network topology analysis using the NetworkAnalyzer, jActiveModules and MCODE plugins in Cytoscape. We have also added the supplementary Table S4 containing all parameters for the comparative network topology analysis between the graphs for healthy donors and P. vivax patients. Detailed descriptions of all analyzed parameters can be found in the references 33 and 34 in the main manuscript text.

The observed difference in the network structure is largely due to the introduction of parasite parameters in the patient graph. A major observation in the network graph of P. vivax patients is the central position of the total parasite biomass marker PvLDH, due to its equally strong interactions with the two main functional modules 1 and 2 (Figure 6A). Given that the hematopoietic niches of the BM and the spleen are the major reservoir of total parasite biomass, interactions of PvLDH with these two main modules indicate an interplay between parasite infection in these niches and endothelial activation/damage as well as the proinflammatory response that results in myeloid-biased differentiation, thrombocytopenia and lymphopenia.

Similar to a previous study with P. vivax patients and healthy donors from an endemic area in Brazil [21], our analysis revealed a dense network of interactions with homogenous and centralized topology among the biomarkers in healthy donors (see values for network diameter; heterogeneity; density and connectivity in Table S4). The network topology is drastically altered in symptomatic P. vivax patients, with a decentralized topology and lower complexity including reduced number of significant interactions between parameters (91 pairwise connections in P. vivax patients vs 166 pairwise connections in healthy donors; p<0.0001) (Figure 6A, Table S4). Interestingly, due to its decentralized and heterogenous pattern of interactions, the network graph of P. vivax patients results in clear functional modules and more closely connected biomarkers. Of note, the network pattern described in our study is similar to protein-protein associated networks described previously in P. vivax malaria and in other clinical contexts [21, 22].

The following text has been added in the Methods, Results and Discussion sections:

“Network topology and module analysis were performed using the NetworkAnalyzer, jActiveModules and MCODE plugins in Cytoscape 72,73. Supplementary File1 shows the results for all parameters quantified in the comparative network topology analysis between the graphs for healthy donors and P. vivax patients.

Page 12, line 285: Similar to a previous study with P. vivax patients and healthy donors from an endemic area in Brazil 46, our analysis revealed a dense network of interactions with homogenous and centralized topology among the biomarkers in healthy donors (Figure 6A, Supplementary File 1). The network topology is drastically altered in symptomatic P. vivax patients, largely due to the introduction of parasite parameters in the patient graph (Figure 6A, Supplementary File 1).”

“Of note, the network pattern described in our study is similar to protein-protein associated networks described previously in P. vivax malaria and in other clinical contexts46,47.”

“A major observation in the network graph of P. vivax patients is the central position of the total parasite biomass marker PvLDH, due to its equally strong interactions with the two main functional modules 1 and 2. Given that the hematopoietic niches of the BM and the spleen are the major reservoir of parasite biomass, interactions of PvLDH with these two main modules indicate an interplay between parasite infection in these niches and endothelial activation/damage as well as the proinflammatory response that results in myeloid-biased differentiation, thrombocytopenia and lymphopenia.”

References

1. Kanjee, U., et al., Molecular and cellular interactions defining the tropism of Plasmodium vivax for reticulocytes. Curr Opin Microbiol, 2018. 46: p. 109-115.

2. Ovchynnikova, E., et al., DARC extracellular domain remodeling in maturating reticulocytes explains. Blood, 2017. 130(12): p. 1441-1444.

3. Barber, B.E., et al., Parasite biomass-related inflammation, endothelial activation, microvascular dysfunction and disease severity in vivax malaria. PLoS Pathog, 2015. 11(1): p. e1004558.

4. Fonseca, L.L., et al., A model of Plasmodium vivax concealment based on Plasmodium cynomolgi infections in Macaca mulatta. Malar J, 2017. 16(1): p. 375.

5. Obaldia, N., et al., Bone Marrow Is a Major Parasite Reservoir in Plasmodium vivax Infection. MBio, 2018. 9(3).

6. Baro, B., et al., Plasmodium vivax gametocytes in the bone marrow of an acute malaria patient and changes in the erythroid miRNA profile. PLoS Negl Trop Dis, 2017. 11(4): p. e0005365.

7. Brito, M.A.M., et al., Morphological and Transcriptional Changes in Human Bone Marrow During Natural Plasmodium vivax Malaria Infections. J Infect Dis, 2020.

8. Kho, S., et al., Evaluation of splenic accumulation and colocalization of immature reticulocytes and Plasmodium vivax in asymptomatic malaria: A prospective human splenectomy study. PLoS Med, 2021. 18(5): p. e1003632.

9. Kho, S., et al., Hidden Biomass of Intact Malaria Parasites in the Human Spleen. N Engl J Med, 2021. 384(21): p. 2067-2069.

10. Druilhe, P., et al., Improved assessment of plasmodium vivax response to antimalarial drugs by a colorimetric double-site plasmodium lactate dehydrogenase antigen capture enzyme-linked immunosorbent assay. Antimicrob Agents Chemother, 2007. 51(6): p. 2112-6.

11. Baird, J.K., Evidence and implications of mortality associated with acute Plasmodium vivax malaria. Clin Microbiol Rev, 2013. 26(1): p. 36-57.

12. Gerardin, P., et al., Prognostic value of thrombocytopenia in African children with falciparum malaria. Am J Trop Med Hyg, 2002. 66(6): p. 686-91.

13. Kochar, D.K., et al., Thrombocytopenia in Plasmodium falciparum, Plasmodium vivax and mixed infection malaria: a study from Bikaner (Northwestern India). Platelets, 2010. 21(8): p. 623-7.

14. Kochar, D.K., et al., Plasmodium vivax malaria. Emerg Infect Dis, 2005. 11(1): p. 132-4.

15. Andrade, B.B., et al., Severe Plasmodium vivax malaria exhibits marked inflammatory imbalance. Malar J, 2010. 9: p. 13.

16. Yeo, T.W., et al., Glycocalyx Breakdown Is Associated With Severe Disease and Fatal Outcome in Plasmodium falciparum Malaria. Clin Infect Dis, 2019. 69(10): p. 1712-1720.

17. Barber, B.E., et al., Endothelial glycocalyx degradation and disease severity in Plasmodium vivax and Plasmodium knowlesi malaria. Sci Rep, 2021. 11(1): p. 9741.

18. Introini, V., et al., Endothelial glycocalyx regulates cytoadherence in Plasmodium falciparum malaria. J R Soc Interface, 2018. 15(149): p. 20180773.

19. Hempel, C., et al., Binding of Plasmodium falciparum to CD36 can be shielded by the glycocalyx. Malar J, 2017. 16(1): p. 193.

20. Toda, H., et al., Plasma-derived extracellular vesicles from Plasmodium vivax patients signal spleen fibroblasts via NF-κB facilitating parasite cytoadherence. Nat Commun, 2020. 11(1): p. 2761.

21. Mendonça, V.R., et al., Networking the host immune response in Plasmodium vivax malaria. Malar J, 2013. 12: p. 69.

22. Frankenstein, Z., U. Alon, and I.R. Cohen, The immune-body cytokine network defines a social architecture of cell interactions. Biol Direct, 2006. 1: p. 32.

https://doi.org/10.7554/eLife.71351.sa2

Article and author information

Author details

  1. João L Silva-Filho

    1. Laboratory of Tropical Diseases – Prof. Luiz Jacintho da Silva, Department of Genetics, Evolution, Microbiology and Immunology, Institute of Biology, University of Campinas, Campinas, Brazil
    2. Wellcome Centre for Integrative Parasitology, Institute of Infection, Immunity & Inflammation, University of Glasgow, Glasgow, United Kingdom
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing – original draft, Writing – review and editing
    Contributed equally with
    João CK Dos-Santos
    For correspondence
    joao.dasilvafilho@glasgow.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4762-2205
  2. João CK Dos-Santos

    1. Laboratory of Tropical Diseases – Prof. Luiz Jacintho da Silva, Department of Genetics, Evolution, Microbiology and Immunology, Institute of Biology, University of Campinas, Campinas, Brazil
    2. Post-Graduation in Medical Pathophysiology, School of Medical Sciences, University of Campinas, Campinas, Brazil
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing – original draft, Writing – review and editing
    Contributed equally with
    João L Silva-Filho
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5916-9845
  3. Carla Judice

    Laboratory of Tropical Diseases – Prof. Luiz Jacintho da Silva, Department of Genetics, Evolution, Microbiology and Immunology, Institute of Biology, University of Campinas, Campinas, Brazil
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1839-053X
  4. Dario Beraldi

    Wellcome Centre for Integrative Parasitology, Institute of Infection, Immunity & Inflammation, University of Glasgow, Glasgow, United Kingdom
    Contribution
    Data curation, Formal analysis, Methodology, Software, Validation, Writing – review and editing
    Competing interests
    No competing interests declared
  5. Kannan Venugopal

    Wellcome Centre for Integrative Parasitology, Institute of Infection, Immunity & Inflammation, University of Glasgow, Glasgow, United Kingdom
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
  6. Diogenes Lima (deceased)

    School of Pharmaceutical Sciences, University of São Paulo, São Paulo, Brazil
    Contribution
    , Data curation, Formal analysis, Investigation, Supervision, Validation
    Competing interests
    No competing interests declared
  7. Helder I Nakaya

    1. School of Pharmaceutical Sciences, University of São Paulo, São Paulo, Brazil
    2. Hospital Israelita Albert Einstein, São Paulo, Brazil
    Contribution
    Conceptualization, Formal analysis, Supervision, Validation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5297-9108
  8. Erich V De Paula

    Department of Clinical Pathology, School of Medical Sciences, University of Campinas, Campinas, Brazil
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Supervision, Validation, Writing – review and editing
    Competing interests
    No competing interests declared
  9. Stefanie CP Lopes

    1. Department of Clinical Pathology, School of Medical Sciences, University of Campinas, Campinas, Brazil
    2. Institute Leônidas & Maria Deane, Fiocruz, Manaus, Brazil
    3. Tropical Medicine Foundation Dr. Heitor Vieira Dourado, Manaus, Brazil
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Supervision, Validation, Visualization, Writing – review and editing
    Competing interests
    No competing interests declared
  10. Marcus VG Lacerda

    1. Institute Leônidas & Maria Deane, Fiocruz, Manaus, Brazil
    2. Tropical Medicine Foundation Dr. Heitor Vieira Dourado, Manaus, Brazil
    Contribution
    Conceptualization, Data curation, Funding acquisition, Project administration, Supervision, Validation, Writing – review and editing
    Competing interests
    No competing interests declared
  11. Matthias Marti

    Wellcome Centre for Integrative Parasitology, Institute of Infection, Immunity & Inflammation, University of Glasgow, Glasgow, United Kingdom
    Contribution
    Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review and editing
    For correspondence
    matthias.marti@glasgow.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1040-9566
  12. Fabio TM Costa

    Laboratory of Tropical Diseases – Prof. Luiz Jacintho da Silva, Department of Genetics, Evolution, Microbiology and Immunology, Institute of Biology, University of Campinas, Campinas, Brazil
    Contribution
    Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review and editing
    For correspondence
    fabiotmc72@gmail.com
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9969-7300

Funding

Fundação de Amparo à Pesquisa do Estado de São Paulo (2019/01578-2)

  • João L Silva-Filho

Fundação de Amparo à Pesquisa do Estado de São Paulo (2017/18611-7)

  • Fabio TM Costa

Wellcome Trust (104111)

  • Matthias Marti

Fundação de Amparo à Pesquisa do Estado de São Paulo (2016/12855-9)

  • João L Silva-Filho

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We would like to thank all patients enrolled in this research and the support of the malaria diagnosis and field team at field team in the Fundação de Medicina Tropical Dr. Heitor Vieira Dourado (FMT-HVD) in Manaus, Brazil. The authors also gratefully acknowledge the help and assistance provided by the Central Laboratory of High-Performance Technologies (LaCTAD, University of Campinas) and the Institute of Infection, Immunity and Inflammation Flow Core Facility in the generation of some of the data reported in this manuscript. MM was supported by a Wolfson Merit Award from the Royal Society and Wellcome Trust Center award (number 104111). JLSF was supported by the Sao Paulo Research Foundation (FAPESP grant 2019/01578-2 and 2016/12855-9), and FTMC was supported by the Sao Paulo Research Foundation (FAPESP grant 2017/18611-7). MVGL and FTMC are CNPq research fellows.

Ethics

All subjects enrolled in the study were adults. Written informed consent was obtained from all participants and the study was conducted according to the Declaration of Helsinki principles. The study was approved by the local Research Ethics Committee at Fundação de Medicina Tropical Dr. Heitor Vieira Dourado (FMT-HVD, Manaus, Brazil), under #CAAE: 54234216.1.0000.0005 and by the Research Ethics Committee at University of Campinas (UNICAMP, Campinas, Brazil), under #CAAE: 54234216.1.3001.5404.

Senior Editor

  1. Dominique Soldati-Favre, University of Geneva, Switzerland

Reviewing Editor

  1. Urszula Krzych, Walter Reed Army Institute of Research, United States

Reviewer

  1. Rays HY Jiang, University of South Florida, United States

Publication history

  1. Preprint posted: March 20, 2021 (view preprint)
  2. Received: June 17, 2021
  3. Accepted: September 28, 2021
  4. Accepted Manuscript published: September 29, 2021 (version 1)
  5. Version of Record published: October 22, 2021 (version 2)

Copyright

© 2021, Silva-Filho et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 781
    Page views
  • 163
    Downloads
  • 5
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. João L Silva-Filho
  2. João CK Dos-Santos
  3. Carla Judice
  4. Dario Beraldi
  5. Kannan Venugopal
  6. Diogenes Lima
  7. Helder I Nakaya
  8. Erich V De Paula
  9. Stefanie CP Lopes
  10. Marcus VG Lacerda
  11. Matthias Marti
  12. Fabio TM Costa
(2021)
Total parasite biomass but not peripheral parasitaemia is associated with endothelial and haematological perturbations in Plasmodium vivax patients
eLife 10:e71351.
https://doi.org/10.7554/eLife.71351

Further reading

    1. Microbiology and Infectious Disease
    Cecile Berne, Sebastien Zappa, Yves V Brun
    Research Article

    In their natural environment, most bacteria preferentially live as complex surface-attached multicellular colonies called biofilms. Biofilms begin with a few cells adhering to a surface, where they multiply to form a mature colony. When conditions deteriorate, cells can leave the biofilm. This dispersion is thought to be an important process that modifies the overall biofilm architecture and that promotes colonization of new environments. In Caulobacter crescentus biofilms, extracellular DNA (eDNA) is released upon cell death and prevents newborn cells from joining the established biofilm. Thus, eDNA promotes the dispersal of newborn cells and the subsequent colonization of new environments. These observations suggest that eDNA is a cue for sensing detrimental environmental conditions in the biofilm. Here we show that the toxin-antitoxin system (TAS) ParDE4 stimulates cell death in areas of a biofilm with decreased O2 availability. In conditions where O2 availability is low, eDNA concentration is correlated with cell death. Cell dispersal away from biofilms is decreased when parDE4 is deleted, probably due to the lower local eDNA concentration. Expression of parDE4 is positively regulated by O2 and the expression of this operon is decreased in biofilms where O2 availability is low. Thus, a programmed cell death mechanism using an O2-regulated TAS stimulates dispersal away from areas of a biofilm with decreased O2 availability and favors colonization of a new, more hospitable environment.

    1. Epidemiology and Global Health
    2. Microbiology and Infectious Disease
    James A Watson, Robert J Commons ... Nicholas J White
    Research Article Updated

    Tafenoquine is a newly licensed antimalarial drug for the radical cure of Plasmodium vivax malaria. The mechanism of action and optimal dosing are uncertain. We pooled individual data from 1102 patients and 72 healthy volunteers studied in the pre-registration trials. We show that tafenoquine dose is the primary determinant of efficacy. Under an Emax model, we estimate the currently recommended 300 mg dose in a 60 kg adult (5 mg/kg) results in 70% of the maximal obtainable hypnozoiticidal effect. Increasing the dose to 7.5 mg/kg (i.e. 450 mg) would result in 90% reduction in the risk of P. vivax recurrence. After adjustment for dose, the tafenoquine terminal elimination half-life, and day 7 methaemoglobin concentration, but not the parent compound exposure, were also associated with recurrence. These results suggest that the production of oxidative metabolites is central to tafenoquine’s hypnozoiticidal efficacy. Clinical trials of higher tafenoquine doses are needed to characterise their efficacy, safety and tolerability.