Information of materials used is available in Supplementary Table 1 .

The sample collection and all experiments were conducted under guidelines approved by the Institutional Biosafety and Biosafety Committee (IBBC) of Universiti Malaya (UMIBBC/PA/R/FOM/PARA-025/2022) and University of Malaya Medical Centre (UMMC) Medical Research Ethics Committee (MRECID#2024312-13526). All experiments in this study were conducted with at least six biological replicates unless stated otherwise, where biological replicates of parasite were defined as individual parasite cultures of different culture batches that contained distinct individual IRBC (either with or without distinct individual URBC). The biological replicates of endothelial cell lines were defined as independent batches of cultured cell lines seeded in distinct culture chambers of fixed dimensions and culture conditions.

Laboratory-adapted P. falciparum and P. knowlesi were thawed using the sodium chloride method and cultured with group O RBC at 2% hematocrit in RPMI-1640 media enriched with AlbuMAX II and 20% (v/v) heat-inactivated human serum. Cultures were maintained under standard in vitro cultivation conditions: 37°C, humidity exceeding 90%, and gas mixture of 7% CO₂, 5% O₂, 90% N₂ (as previously described in Lee et al., 2014 and Lee et al., 2022c). Late-stage parasites were purified using a magnetic-activated cell sorting (MACS) system.

For the cultivation of human endothelial cell lines, culture flask was treated with 0.5% gelatin solution for 6 hours at 37°C prior to the inoculation of thawed cell lines. The cell lines used were human cerebral microvascular endothelial cell line (hCMEC/d3), human renal glomerular endothelial cell line (HRGEC), human pulmonary microvascular endothelial cell line (HPMEC) and human umbilical vein endothelial cell line (HUVEC)]. They were cultured with the complete endothelial cell medium (ECM).

Recombinant human periostin protein (henceforth known as “OSF-2”) was reconstituted with 1X phosphate-buffered saline (PBS) (stock protein concentration: 100 μg/ml) according to the manufacturer’s manual. The stock was aliquoted and stored at 4°C until its subsequent use.

The experiment was conducted when approximately 70% of the parasite population had reached late-stage development (late trophozoite to schizont stages). The parasite culture suspension was exposed to various working concentrations of OSF-2. A separate aliquot of parasite culture suspension, which was unexposed to OSF-2, served as the control. Notably, the selected OSF-2 concentrations were within the reported pathophysiologic range of serum OSF-2 levels ( Supplementary Table 2 ) (Okamoto et al., 2011; Yamaguchi et al., 2013; Kou et al., 2014; Caswell-Smith et al., 2016; Yang et al., 2016; Zhu et al., 2016; Thuwajit et al., 2017; Ding et al., 2018; Gadermaier et al., 2018; Yildiz et al., 2021). The plate was incubated for one hour under standard in vitro cultivation conditions before the rosetting assay (Lee et al., 2013; Lee and Rénia, 2020). Briefly, the culture suspension was stained with Giemsa, transferred to a clean glass slide, and mounted with a glass cover-slip. The wet mount was then examined with a light microscope using immersion oil magnification. The rosetting rate was calculated as the percentage of IRBCs that formed rosettes, by recruiting 200 IRBC. After determining the optimal working concentration of OSF-2 for subsequent experiments, the importance of human serum availability to the OSF-2-mediated rosetting was tested by repeating the assay with the selected working concentration of OSF-2 in culture media supplied with different levels of human serum enrichment, where the original culture medium (20% serum-enriched medium) was removed via centrifugation (300 g for 2 minutes), and the pelleted cells were suspended with experimented culture media.

In a separate experiment, trypsinization was performed based on previous trypsin sensitivity profiling of P. falciparum cytoadherence ligands (Kyes et al., 2000; Niang et al., 2014). Briefly, purified late-stage IRBC were divided into three groups. One treated with 10 µg/mL of trypsin, another with 1mg/ml of trypsin, and a third untreated control group. Following a ten-minute incubation under in vitro culture conditions, the enzymatic reaction was stopped with human serum-enriched culture medium. Each group was then divided into two parts; one was supplemented with OSF-2 (200 ng/ml) and the other served as a control. The rosetting assay was conducted after one hour of incubation under in vitro cultivation conditions.

An additional experiment was conducted, where MACS-sorted late-stage IRBC were divided into four groups. Each group of purified IRBC was mixed with URBC of A, B, O, and AB groups, respectively, to create a suspension of 1% parasitemia with 2% hematocrit. Each group was further divided into two experiment fractions (OSF-2-exposed and unexposed control), and incubated under in vitro cultivation conditions for one hour before the rosetting assay.

When the cultures of endothelial cell lines reached approximately 70% confluency, they were transferred into the gelatin-coated eight-well LABTEK chamber slides (each well was seeded with 1 × 10⁵ cells). On the following day, a mixture of ECM and parasite antigen suspension in a 2:1 ratio was added to the seeded cell lines and incubated under in vitro cultivation conditions for 24 hours, to ‘prime’ the cells with parasite antigens (Lee et al., 2022c). After cell priming, each well of the seeded cells was allocated parasite culture suspension (1% parasitemia, 2% hematocrit) and OSF-2 suspension (200 ng/ml). An aliquot of suspension from each well was taken for rosetting assay. The endothelial cell-parasite mixture was incubated for two hours under in vitro cultivation conditions. Subsequently, the suspension from each well was collected into separate microcentrifuge tubes for rosetting assay to evaluate the before-after effect. For the seeded endothelial cells, the unbound IRBCs were gently washed away with human serum-enriched medium for three times. This was followed by fixation with ice-cold absolute methanol for ten minutes. The chambers were then removed from the chamber slides, and the slides were stained with 5% Giemsa for 20 minutes before being examined under a light microscope using immersion oil magnification. The IRBC–endothelial cell line cytoadherence rate was determined by counting the number of IRBCs attached to endothelial cells per 100 fields (equivalent to coverage of approximately 8,000 cells). The IRBC-endothelial cytoadherence assay was repeated with a separate set of “primed” cell lines that were not exposed to OSF-2 (as control).

GraphPad Prism 9.5.1 software was used for data analyses. Normality of the data was evaluated using Shapiro-Wilk test. For normally distributed datasets, Welch’s t-test was used for the comparison of two datasets with unequal variances. For paired comparisons, paired t-test was used. For multi-group comparison against a control group, One-way ANOVA with Dunnett’s test was performed. For non-normally distributed datasets, Mann-Whitney test was used to compare two datasets, and Wilcoxon matched-pairs signed rank test was used for paired comparison. For the comparison of multiple datasets, Kruskal-Wallis with Dunn’s test was conducted. For the comparison of multiple sets of matched non-parametric data, Friedman with Dunn’s test was performed. P values smaller than 0.05 were considered as statistically significant.

A preliminary experiment was conducted to examine the basal rosetting rates of the parasites across seven cycles of cultivation post-thawing. Both species of parasites demonstrated persistent but fluctuating rosetting rates. The fluctuation of P. knowlesi rosetting rates across the monitored period was of insignificant difference (Kruskal-Wallis H(2) = 7.15; P = 0.41) ( Supplementary Figure 1A ). However, significant difference in rosetting rates of P. falciparum across the monitored cycles was found (Kruskal-Wallis H(2) = 16.11; P = 0.01) ( Supplementary Figure 1B ). Thus, culture cycle-matched statistical comparison was done for data collected from subsequent experiments.

The rosetting rates of P. falciparum and P. knowlesi responded positively to OSF-2 (Friedman with Dunn’s test X²(7) = 38.94; P < 0.0001 for P. falciparum; and X²(7) = 35.37; P < 0.0001 for P. knowlesi). For P. falciparum, significant rosette-stimulation was observed at OSF-2 working concentration of 200 ng/ml (P = 0.0016) and 250 ng/ml (P < 0.0001) ( Figure 1A ). For P. knowlesi, significant rosetting rate increment was recorded at OSF-2 working concentrations of 150 ng/ml (P = 0.0027), 200 ng/ml (P = 0.0007), and 250 ng/ml (P = 0.0002) ( Figure 1B ). Of note, there was no significant difference in rosette-stimulation between OSF-2 of 200 ng/ml and 250 ng/ml for spP. falciparum. Similarly, no significant difference in P. knowlesi rosette-stimulation was found at OSF-2 of 150 ng/ml, 200 ng/ml and 250 ng/ml. Subsequent experiments were then conducted with 200 ng/ml OSF-2. The OSF-2-mediated rosetting was human serum-dependent for P. falciparum (Friedman with Dunn’s test X²(7) = 28.40; P < 0.0001) ( Figure 1C ) and P. knowlesi (Friedman with Dunn’s test X²(7) = 27.76; P < 0.0001 for P. knowlesi) ( Figure 1D ), where significant increment of rosetting rates was found at conditions supplied with 15% serum (P = 0.0028 and 0.005 for P. falciparum and P. knowlesi, respectively) and 20% serum (P = 0.0001 and 0.0015 for P. falciparum and P. knowlesi, respectively), as compared with the serum-free setting. For both species, the OSF-2-mediated rosette stimulation under conditions supplied with 15% and 20% serum were of insignificant difference. In short, OSF-2 stimulated P. falciparum and P. knowlesi to form more rosettes in a human serum-dependent manner.

For P. falciparum, treatment with either low (10 µg/ml) or high (1 mg/ml) concentration of trypsin to IRBC abrogated the OSF-mediated rosette-stimulating effect ( Figure 1E ). Similarly, the OSF-2-mediated rosette-stimulation on P. knowlesi was abolished by the IRBC treatment with both concentrations of trypsin ( Figure 1F ). We also investigated the role of human ABO blood groups in OSF-2-mediated rosetting. The OSF-2-mediated rosetting in P. falciparum ( Figure 1G ) and P. knowlesi ( Figure 1H ) was independent of human ABO blood groups, where the OSF-2-mediated rosette-stimulation on IRBC purified from the same batch of parasite culture was insignificantly different when subjected to URBC of A, B, O and AB groups (Friedman with Dunn’s test X²(7) = 1.68; P = 0.64 for P. falciparum; and X²(7) = 3.16; P = 0.37 for P. knowlesi). Succinctly, the OSF-2 mediated rosetting by P. falciparum and P. knowlesi was trypsin-sensitive and independent of human blood ABO groups.

We evaluated the effect of OSF-2 on the P. falciparum-IRBC cytoadherence dynamics using endothelial cell lines hCMEC/d3 ( Figures 2A, B ), HRGEC ( Figures 2C, D ), HPMEC ( Figures 2E, F ), and HUVEC ( Figures 2G, H ). The protein demonstrated similar effects on the P. falciparum-IRBC cytoadherence dynamics with these endothelial cells. Following the parasite-endothelial cell line incubation, OSF-2 significantly increased the rosetting rates of P. falciparum as expected (Wilcoxon matched-pairs signed rank test W = 21; P = 0.03 for hCMEC/D3, HRGEC, HPMEC and HUVEC) ( Figures 2A, C, E, G , respectively). The IRBC-endothelial binding was significantly lower when OSF-2 was supplied (Paired t-test: t = 12.60, df = 5, P < 0.0001 for hCMEC/D3; t = 20.17, df = 5, P < 0.0001 for HRGEC; t = 7.652, df = 5, P = 0.0006 for HPMEC; t = 8.92, df = 5, P = 0.0003 for HUVEC) ( Figures 2B, D, F, H , respectively).

Similar OSF-2-mediated effect was observed with P. knowlesi when tested with hCMEC/D3 ( Figures 3A, B ), HRGEC ( Figures 3C, D ), HPMEC ( Figures 3E, F ), and HUVEC ( Figures 3G, H ). The rosetting rates by the zoonotic parasite were significantly increased by OSF-2 (Wilcoxon matched-pairs signed rank test W = 21; P = 0.03 for hCMEC/D3, HRGEC, HPMEC and HUVEC) ( Figures 3A, C, E, G , respectively), whereas the P. knowlesi-IRBC-endothelial cytoadherence was significantly reduced under OSF-2-supplied settings (Wilcoxon matched-pairs signed rank test W = 21; P = 0.03 for hCMEC/D3, HRGEC, HPMEC and HUVEC) ( Figures 3B, D, F, H , respectively). Of note, at basal level, the P. falciparum IRBC-endothelial binding rate was the highest with the brain-derived hCMEC/D3, and the P. knowlesi IRBC-endothelial binding rate was the highest with the umbilical cord-derived HUVEC. To sum up, following the coincubation of OSF-2, IRBC, URBC, and endothelial cells, the protein significantly increased the rosetting rates while reducing the IRBC-endothelial cytoadherence with endothelial cell lines derived from the brain, kidney, lung, and umbilical cord.