Blood samples of patients with vivax and knowlesi malaria were collected from three endemic areas: the Republic of Korea (ROK), Myanmar, and Malaysia. To evaluate the longevity of immune response, blood samples from patients with vivax malaria were collected from Shwegyin Township in Myanmar on the day of the patient’s visit (D0, acute infection), day 28 after proper treatment (D28, sub-acute) ( Supplementary Table S1 ), and 1 year later, without reinfection (1 y). According to the National Malaria Treatment Guidelines, all participants received the appropriate treatment. This study was conducted concurrently with the therapeutic efficacy study of chloroquine in uncomplicated vivax malaria, using the template protocol provided by the WHO. A history of no malarial episodes (except for the relevant ones for D28 and 1 y samples) and current malarial infection status were confirmed from the data in the patient record files, microscopy, and species-specific polymerase chain reaction (PCR). For the negative control, blood samples of healthy individuals from non-endemic areas of the ROK were collected after microscopy- and PCR-based confirmation of malaria-free status. Sera were isolated from whole blood samples and stored at -80°C until a protein array was performed. All experiments complied with relevant guidelines and regulations, and all protocols involving human samples were approved by the Institutional Ethical Committee of the Department of Medical Research, Myanmar (Approval number 49/Ethics-2014), the Kangwon National University Hospital Ethical Committee (IRB No. 2014-08-008-002), University of Malaysia Medical Ethics Committee (Ref No. 817.18), and the Medical Research Ethics Committee (MREC), Ministry of Health, Malaysia (NMRR-16-2840-33769). Informed consent was obtained from all participants.

In this study, sequences of MTRAP proteins of divergent Plasmodium species were retrieved from PlasmoDB (www.plasmodb.org), which included P. vivax (PVX_111290), P. knowlesi (PKNH_0613400), P. falciparum (PF3D7_1028700.1), P. ovale curtisi (PocGH01_06021700.1), and P. cynomolgi (PcyM_0617400-t36_1); P. malariae (XP_028862173.1) was obtained from GenBank. Sequence alignment was performed by the CLUSTAL-W program in MegAlign Lasergene ver. 7.0 (DNASTAR, Madison, WI). The primary structures of PvMTRAP and PkMTRAP were analyzed using the Simple Modular Architecture Research Tool (http://smart.emblheidelberg.de/). To obtain the predicted three-dimensional structures of PvMTRAP and PkMTRAP, the protein sequences were submitted to SWISS-MODEL (Biasini et al., 2014) and AlphaFold2 webserver (Jumper et al., 2021); the selected structures were then visualized and analyzed by UCSF Chimera (Huang et al., 2014).

Recombinant PvMTRAP and PkMTRAP were produced using a mammalian HEK293E cell-expression system as described elsewhere (Patel et al., 2022). For plasmid preparation, GeneArt (https://www.thermofisher.com/order/geneartgenes/projectmgmt) was used to codon-optimize a cDNA sequence encoding the full-length ectodomain of PvMTRAP and PkMTRAP for expression in mammalian cells, and was followed by chemical synthesis with a Twist Bioscience tool (South San Francisco, CA, USA). The cDNAs were cloned into a pTT5 vector containing a leader sequence of mouse variable-k-light chain (mIgGk) as an exogenous signal peptide, and 8× His C-terminal tags were then transfected into HEK cells for protein production. In accordance with the manufacturer’s protocol, the His-tagged proteins were purified using Ni-NTA agarose (QIAGEN, Hilden, Germany) and Poly-Prep® Chromatography column (Bio-Rad, Hercules, CA, USA).

For producing polyclonal antibodies against PvMTRAP and PkMTRAP, we used 6- to 8-week-old female BALB/c mice (Daehan Biolink Co., Eumsung, ROK). Twenty microliters of purified recombinant proteins mixed with complete Freund’s Adjuvant (Sigma-Aldrich, St. Louis, MO) was injected intraperitoneally into each mouse. At weeks 3 and 5 following the initial immunization, a mixture of the same quantities of antigens with incomplete Freund’s Adjuvant (Sigma-Aldrich) was administered as booster doses. Mouse sera were collected 2 weeks after the final boost. For producing rabbit antibodies against PvMTRAP and PkMTRAP, 250 μg protein mixed with complete Freund’s Adjuvant was administered subcutaneously to Japanese white rabbits. Two booster immunizations with 250 μg protein mixed with incomplete Freund’s Adjuvant were given every 3 weeks, and rabbit sera were obtained 2 weeks after the final injection. In accordance with the manufacturer’s instructions, IgG purification from rabbit sera was performed using a protein G HP column (GE Healthcare Life Sciences, Buckinghamshire, UK). The elution fractions were then subjected to buffer exchange for incomplete RPMI1640 (Invitrogen, Carlsbad, CA, USA) and concentrated at 30 kDa using centrifugal devices (Millipore, Billerica, MA, USA) for downstream assay.

Protein lysates from P. knowlesi A1-H.1 schizont-stage parasites were obtained as previously described (Muh et al., 2018a). After treatment with a reducing agent to induce a reducing condition, P. knowlesi schizont lysates or His-tagged recombinant proteins were separated by 13% SDS-PAGE gel. A gel for SDS-PAGE analysis was stained with Coomassie brilliant blue (Sigma-Aldrich). Concurrently, another gel was used for protein blotted onto PVDF membranes (Millipore) and blocked with 5% skim milk at room temperature (RT) for 1 h. The membrane containing His-tagged protein was then incubated with either mouse anti-penta-His antibody (1:2,000) or animal immune sera (1:1,000) against PvMTRAP or PkMTRAP, followed by incubation with a secondary goat anti-mouse IRDye^® 800 (1:10,000) or goat anti-rabbit IRDye^® 680 (1:20,000) antibody. Data analysis was performed using an Odyssey infrared imaging system and software (LI-COR^® Bioscience).

Amine-coated slides were prepared and used as described elsewhere (Chen et al., 2010). To evaluate the humoral immune response to PvMTRAP and PkMTRAP, 1 µL of each purified recombinant protein was spotted onto duplicate wells of the array at 100 ng/μL. After blocking with 5% bovine serum albumin in phosphate-buffered saline (PBS) with 0.1% Tween 20, the slides were incubated with serum samples from patients with vivax or knowlesi malaria or unexposed healthy individuals (1:25 dilution) at 37 °C for 1 h. Non-binding human IgG was washed from the slides, whereas bound IgG were labeled with Alexa Fluor 546 goat anti-human IgG antibody (10 ng/µL; Invitrogen) for detection using a fluorescence scanner (InnoScan, Carbonne, France). A cutoff value was established at the mean fluorescence intensity (MFI) plus two standard deviations (SD) of healthy individuals, which was then used to normalize individual MFI (Muh et al., 2017).

Plasmodium knowlesi A1-H.1 parasites were fixed with a mixture of 4% paraformaldehyde and 0.075% glutaraldehyde and used for further immunostaining procedures. The fixed parasites were permeabilized with 0.1% Triton X-100 in PBS for 30 min at RT. The slides were then co-stained with mouse serum against either PvMTRAP or PkMTRAP as well as each rabbit serum against P. knowlesi merozoite surface protein 1–19 (PkMSP1–19) (Lee et al., 2023), P. knowlesi Duffy binding protein (PkDBP) (Hart et al., 2023), P. vivax rhoptry neck protein (PvRON2) (Arévalo-Pinzón et al., 2011), and P. vivax rhoptry-associated membrane antigen proteins (PvRAMA) (Lu et al., 2014) as subcellular localization makers for merozoite surface, microneme, rhoptry neck, and rhoptry body, respectively. After washing, the IgG binding was probed with Alexa Fluor 568-conjugated anti-mouse IgG (H+L) and Alexa Flour 488-conjugated anti-rabbit IgG (H+L) antibody (1:500, Invitrogen), whereas nuclei were stained with 4′,6-diaminidino-2-phenylindole (DAPI) (1:1000, Invitrogen). The slides were mounted with ProLong Gold antifade reagent (Invitrogen) and visualized under oil immersion using a Nikon C2 confocal microscope (Nikon Instrument Inc., City, NY, USA) equipped with a 60× oil objective. The acquired images were prepared using ImageJ.

The P. knowlesi A1-H.1 strain was cultured with human erythrocytes in RPMI 1640-based medium (Invitrogen/Life Technologies, Grand Island, NY, USA), as described previously (Moon et al., 2013). The Percoll solutions were freshly prepared by mixing nine parts of Percoll density gradient media (Cytiva, Marlborough, MA, USA) with one part 10× PBS. Subsequently, this stock solution was diluted with 1× PBS to the desired concentrations. The parasite culture with predominant schizont stage was pelleted and carefully overlayed on the 40%/70% Percoll gradient before centrifugation at 900 ×g for 12 min without brake. The schizonts collected from the lower interphase were washed three times with RPMI prior to use in downstream experiments. A schizont enrichment assay using magnetic-activated cell sorting (MACS) (Miltenyi Biotec, Bergisch Gladbach, Germany) with an LD column (Meltenyi Biotec), well-established for P. knowlesi A1-H.1, was performed in parallel to Percoll to compare the proliferation feature of enriched schizonts (Amir et al., 2016).

The invasion inhibition assay was performed as previously described (Muh et al., 2018b). Briefly, purified schizonts of P. knowlesi A1-H.1 were sub-cultured in a 96-well plate at 2% hematocrit and 1.0–1.5% parasitemia. Purified rabbit IgG against PvMTRAP or PkMTRAP at serial concentrations were added to duplicate wells. Cultures without antibodies or with Duffy antigen receptor for chemokines (DARC) monoclonal antibody (2C3, Absolute antibody, Oxford, UK) at 25 µg/mL served as controls for normal erythrocyte invasion and antibody-mediated invasion inhibition, respectively. The sub-culture was maintained in a humidified chamber under mixed gas (90% N₂, 5% CO₂, 5% O₂) at 37°C until newly invaded ring-stage parasites were detected on Giemsa staining (approximately 10 h). Parasites were fixed with 0.05% glutaraldehyde in PBS, then stained with SYBR Green (1:10,000; Invitrogen). A total of 200,000 events were counted per sample using a FACS Accuri™ C6 Flow Cytometer (Becton-Dickinson Co., Mansfield, MA, USA). Data were analyzed using FlowJo (Treestar, Ashland, OR, USA).

The data were analyzed using GraphPad Prism (GraphPad Software, San Diego, CA, USA) and Microsoft Excel 2018 (Microsoft, Redmond, WA, USA). During the comparison of experimentally measured values across groups, the Student’s t- and Mann–Whitney U-tests were used. Spearman analysis was utilized to examine the correlation of nonparametric data; p<0.05 indicated significant differences.

The PvMTRAP and PkMTRAP proteins consisted of 387 and 375 amino acids, respectively, and both were predicted as transmembrane type 1 proteins with a signal peptide followed by two TSR domains at the N-terminal ( Figure 1A ). The repeated sequence of “WxxWxxC~G~R~C” (“x” could be substituted by any amino acid) was largely conserved among TSR domains of MTRAP from human-infected malarial parasites and human thrombospondin-1 protein ( Supplementary Figure S1A ). PvMTRAP and PkMTRAP sequences shared 77.7% identity, and thus exhibited the closest phylogenetic relationship among malarial parasites that infect humans ( Figures 1B, C ; Supplementary Figure S1B ).

Given the unsuccessful attempt to find an experimentally determined template for PvMTRAP and PkMTRAP, the full protein sequences were submitted to an AlphaFold 2 webserver for predicting their 3D structures. Only TSR domains of MTRAP proteins exhibited the sufficient per-residue confidence score (high and very high, per-residue model confidence score >70) and were selected for further comparison with those of human thrombospondin-1 (HsTSP-1; PDB identification 1LSL) (Uchime et al., 2012). The TSR domains of PvMTRAP and PkMTRAP shared 81.8% sequence identity, and their 3D structures were predicted to be extensively overlapped ( Figures 2A, B ). As predicted, each TSR domain of PvMTRAP and PkMTRAP folded into a long, thin, antiparallel, three-stranded domain that was similar to that of HsTSP-1 (Tan et al., 2002); however, significantly more dense structures were noticed in MTRAP-TSR domains ( Figure 2C ), which likely resulted from the prediction that a single TSR domain of MTRAP was approximately half the length of the HsTSP-1 TSR domain (~27.76 Å compared to ~55 Å, respectively), and linkers between the two TSR domains of MTRAPs had eight fewer amino acids than that of HsTSP-1 ( Figures 2A, C ). The PvMTRAP/PkMTRAP TSR domains included four disulfide bridges, which is less than the six found in HsTSP-1 ( Figures 2A, C ). The disulfide bonds between a cysteine at the N terminal of A-loop and a cysteine at the strand B (the first TSR domain) or in the B-C loop (the second TSR domain) were considered specific for MTRAP as it was not found in HsTSP-1 ( Figures 2A, C ) or TRAP-TSR domains (Tan et al., 2002).

The full-length ectodomains of PvMTRAP and PkMTRAP proteins were expressed and purified as single bands of approximately 57.8 and 51.5 kDa, respectively ( Figure 3A ). Western blot analysis using anti-His antibody confirmed the migration pattern of recombinant PvMTRAP and PkMTRAP observed by SDS-PAGE ( Figures 3B, C , lane a-His). At 32.4 and 32.5 kDa for PvMTRAP and PkMTRAP, respectively, the product weight exceeded the predicted molecular weight; this may have resulted from post-translational modifications and/or the structure of MTRAP proteins, as the same phenomenon was found in PfMTRAP (Baum et al., 2006). Antibodies from rabbits and mice immunized with recombinant PvMTRAP and PkMTRAP were able to recognize the respective antigens similarly to an anti-His antibody ( Figures 3B, C , lane R and M); furthermore, anti-PvMTRAP IgG cross-reacted with recombinant PkMTRAP ( Figure 3B , lane a-Pv) and vice versa ( Figure 3C , lane a-Pk). Interestingly, antibodies against PkMTRAP and PvMTRAP explicitly bound to the native protein from P. knowlesi schizont lysate (approximately 51.0 kDa; Figure 3D ), giving an additional indication of cross-species immunoreactivity between PkMTRAP and PvMTRAP. The extra bands detected by anti-MTRAP IgGs may represent processing fragments of MTRAP proteins during parasitic development ( Figure 3D ).

The fluorescence signal of the antibody against PkMTRAP presents as a punctured pattern, indicating that PkMTRAP localizes to an apical part of the merozoite ( Figure 4A , in red). Unlike PfMTRAP (Baum et al., 2006), PkMTRAP does not reside in the microneme, and this inference is drawn from the distinct signal pattern observed with the anti-PkMTRAP antibody, which differs from that of the anti-PkDBPα antibody, the latter serving as a marker for P. knowlesi merozoite micronemes (Hart et al., 2023).Antibodies to PvRAMA, a P. vivax rhoptry body protein, and PvRON2, a P. vivax rhoptry neck protein, were both able to recognize their orthologous in rhoptry of P. knowlesi merozoite (Muh et al., 2018a), suggesting that they could be used as subcellular localization makers for rhoptry region of P. knowlesi. Given the lack of evidential subcellular markers for P. knowlesi rhoptries and our attempt to further localize PkMTRAP between the bulb and neck region of merozoite rhoptry, antibodies against PkMTRAP were co-stained with antibodies against PvRAMA and PvRON2. The signals from PkMTRAP mostly overlapped with those from PvRAMA (r ^{2 =} 0.85), but not PvRON2 (r ^{2 =} 0.47), ( Figure 4A ) revealing that (i) P. knowlesi orthologous of PvRAMA and PvRON2 might localize to distinct parts of the rhoptry, and (ii) the rhoptry body was likely the localization of PkMTRAP.

Extensively overlapping signals between rabbit antisera against PvMTRAP and mouse antisera against PkMTRAP (r ^{2 =} 0.95, Figure 4A , bottom line) and in reverse orientation (r ^{2 =} 0.92, Figure 4B , bottom line) indicated that anti-PvMTRAP could cross-react with native PkMTRAP antigens. Moreover, replacing anti-PkMTRAP antibodies with anti-PvMTRAP antibodies using the same subcellular markers replicated identical fluorescence signal patterns ( Figure 4B ). This confirmed the cross-binding of anti-PvMTRAP to the PkMTRAP antigen and reaffirmed the subcellular localization of PkMTRAP. Pre-immune animal IgG, used as a negative control, did not recognize the parasite antigens ( Supplementary Figure S2 ).

Recombinant PvMTRAP or PkMTRAP proteins were reacted with sera from P. vivax patients (n=70), P. knowlesi patients (n=70) collected from endemic areas, and non-exposed healthy individuals (n=30) ( Figure 5A ). The IgG responses of P. vivax and P. knowlesi patients’ sera to the respective MTRAP antigens were significantly higher than those of non-exposed human sera (p<0.001). Furthermore, sera from P. vivax patients produced a significant IgG response to PkMTRAP ( Figure 5A ), with an extensive positive correlation with the response to PvMTRAP (r ^{2 =} 0.91, 95% CI: 0.85–0.94; Supplementary Figure S3A ). PvMTRAP induced a high seropositivity of 75.71% (95% CI: 63.99–85.17%), with 96.67% (95% CI: 82.78–99.92%) specificity, in P. vivax–infected individuals ( Table 1 ). To evaluate the longevity of immune responses, we further conducted the assay with three different sets of sera obtained from vivax patients at D0_acute (n=32), D28_sub-acute infection (n=32), and at least 1 year after infection without reinfection (n=32) ( Figure 5B ). The results demonstrated that the IgG responses to PvMTRAP in the sera of P. vivax patients were higher than those in the healthy human sera, regardless of the post-infection duration (p<0.001). Furthermore, compared to that in acute and sub-acute infection, the IgG response did not significantly change after at least 1 year of infection. Related information on patients with malaria, including age, sex, and parasitemia, is presented in Supplementary Table S1 . Overall, the protein array data suggest that PvMTRAP is a potential vaccine candidate against P. vivax, particularly when coexisting with P. knowlesi.

The enrichment of human erythrocyte-infected P. knowlesi A1-H.1 schizont using Percoll gradient is considered a reasonable method, based on successful experiments on other strains of P. knowlesi maintained with either monkey or human red blood cells (Ginsburg et al., 1987; Moraes Barros et al., 2017; Armistead et al., 2018). Various concentrations of Percoll were assessed before choosing the 70% cushion for efficiently purifying late-stage parasites. Additionally, a 40% Percoll cushion effectively eliminated dead cells and debris (data not shown). Using the 40%/70% Percoll gradient ( Figure 6A ), a majority of late-stage parasites were concentrated within the lower interface layer ( Figure 6B ), thereby effectively separating most of the rings, early trophozoites, and uninfected human erythrocytes in the pellet ( Figure 6C ). The enriched late-stage parasites were re-cultured, and subsequently developed into new daughter rings within approximately 10 h ( Figure 6D ), but without significantly impacting parasite growth ( Figure 6E ). This confirmed that the 40%/70% Percoll gradient is appropriate for preparing P. knowlesi A1-H.1 schizonts for downstream invasion inhibition assays.

A growing body of practical evidence supports the use of in vitro culture P. knowlesi-based invasion inhibition assays for screening potential asexual blood-stage vaccine candidates from P. vivax (Muh et al., 2018a; Han et al., 2019; Muh et al., 2020; Ndegwa et al., 2021). In the present study, we employed the previously established method using SYBR Green I as a specific DNA staining to evaluate the ability of purified rabbit IgG against PkMTRAP or PvMTRAP to prevent the parasite from invading erythrocytes (Muh et al., 2018b). The healthy ring-form morphology of the parasite was obtained from schizonts cultured without any intervention as well as antibody-treated schizonts, suggesting that the function of antibodies (if extant) was related to the erythrocyte invasion rather than to other steps of the schizont-to-ring stage development ( Figure 7A ). The percentage of daughter rings was evaluated by flow cytometry using manual gating strategies ( Figure 7B ). Both purified anti-PkMTRAP and anti-PvMTRAP rabbit IgG suppressed the erythrocyte invasion of P. knowlesi parasites in a concentration-dependent manner. At IgG concentration ≥1.5 mg/mL, the anti-PkMTRAP and anti-PvMTRAP antibody invasion inhibition activity was significantly higher than that of pre-immune IgG. Within the tested range, the best inhibitory performance of antibodies against PkMTRAP and PvMTRAP was 35.02 ± 2.48% and 43.08 ± 3.28%, respectively, without significant difference between the two antibodies ( Figure 7C ).