Clinicopathological Analysis and Multipronged Quantitative Proteomics Reveal Oxidative Stress and Cytoskeletal Proteins as Possible Markers for Severe Vivax Malaria

In Plasmodium vivax malaria, mechanisms that trigger transition from uncomplicated to fatal severe infections are obscure. In this multi-disciplinary study we have performed a comprehensive analysis of clinicopathological parameters and serum proteome profiles of vivax malaria patients with different severity levels of infection to investigate pathogenesis of severe malaria and identify surrogate markers of severity. Clinicopathological analysis and proteomics profiling has provided evidences for the modulation of diverse physiological pathways including oxidative stress, cytoskeletal regulation, lipid metabolism and complement cascades in severe malaria. Strikingly, unlike severe falciparum malaria the blood coagulation cascade was not found to be affected adversely in acute P. vivax infection. To the best of our knowledge, this is the first comprehensive proteomics study, which identified some possible cues for severe P. vivax infection. Our results suggest that Superoxide dismutase, Vitronectin, Titin, Apolipoprotein E, Serum amyloid A, and Haptoglobin are potential predictive markers for malaria severity.

). In Q-Exactive analysis, a total of 430 proteins were identified, out of which 176 were with ≥ 2 unique peptides. Significant increase (fold-change ≥ 1.5) in the abundance of 24 proteins was observed in NSVM and 84 proteins in SVM, whereas abundance of 23 and 19 proteins was found to be reduced in NSVM and SVM, respectively, compared to HC (Fig S3; Table S5). S-curve distributions of the differentially abundant proteins in vivax malaria and scatter plots exhibiting correlations among the different iTRAQ data sets are represented in Fig. S2. iTRAQ data obtained from the Q-TOF and Q-Exactive were compared to evaluate overlaps between the two data sets (Fig. 2E). Comparative analysis of the findings obtained from these two different mass spectrometers indicated overlap of 73 proteins, out of which 65 were found to be with ≥ 2 peptides. 59 candidates including Serum amyloid A (SAA), C-reactive protein (CRP), Alpha-1-antichymotrypsin (ACT), Apolipoprotein A-I Apo A-I, HP and Serum albumin (ALB) showed similar trends of differential abundance in malaria patients (compared to healthy subjects) in both Q-TOF and Q-Exactive analyses (Table S4; Table S5).
Interestingly, many serum proteins such as SAA, Apo E, TTN, NEB, Alpha-1-acid glycoprotein 1, Apolipoprotein A-IV and Transthyretin were found to have different serum abundances in SVM compared to NSVM (Table 2). A comparison of 2D-DIGE and iTRAQ analysis indicated that iTRAQ provides more comprehensive proteomic coverage as compared to 2D-DIGE and almost all the candidates (except DNA2-like helicase) identified in gel-based proteomics analysis were also identified in iTRAQ-based quantitative proteomics analysis.

Modulation of essential physiological pathways and networks in malaria. Bioinformatics analysis
of the differentially abundant proteins (p-value < 0.05) using Ingenuity Pathway Analysis (IPA) identified six and eight overlapping interaction networks in NSVM and SVM, respectively. (Table S6). The most significant related functions derived from these overlapping networks included, lipid metabolism, inflammatory response, skeletal and muscular disorder, immune cell trafficking and cellular movement (Table S6). According to this functional pathway analysis, SVM leads to the alteration of multiple serum proteins involved in diverse essential physiological pathways including acute phase response, LXR/RXR activation, atherosclerosis signalling and primary IL-12 signaling and production in macrophage ( Fig. S4; Table S6).
PANTHER (Protein ANalysis THrough Evolutionary Relationships) analysis also indicated involvement of the differentially abundant proteins in blood coagulation system. In addition, plasminogen activating cascade, apoptosis signaling, inflammation mediated by chemokine and cytokine signaling pathways were identified as the related physiological pathways with statistical significance (p < 0.05) (Table S6). In DAVID (Database for Annotation, Visualization and Integrated Discovery) analysis complement and coagulation cascades, hemostasis, plasminogen activating cascade, metabolism of lipids and lipoproteins were identified as the significant pathways (Table S6). Taken together, bioinformatics analysis indicated that both severe and non-severe P. vivax infections lead to alteration of serum levels several proteins involved in diverse essential physiological pathways including acute phase response signaling, complement cascades, lipid transport and metabolism and blood coagulation (Table 3; Fig. 3).
Measurement of serum abundances of some potential marker proteins in NSVM and SVM by ELISA and surface plasmon resonance (SPR). Compared to HC, both NSVM and SVM patients were found to have lower serum levels of HP, Apo A-I, and retinol binding protein 4 (RBP4) (p < 0.05) ( Fig. 4A; Table S7). HP and Apo A-I exhibited more than two times lower mean value in SVM patients compared to HC, whereas elevated levels of HP were observed in DF and LEP patients. Serum levels of SAA, Apo-E, HPX, Ceruloplasmin (CP), and Plasminogen (PLS) were found to be significantly higher in both NSVM and SVM patients, compared to HC ( Fig. 4A; Table S7). Intriguingly, similar serum levels of Apo E, RBP4 and HPX were observed in DF and LEP patients compared to HC. Serum levels of SOD, VTN and TTN were found to be significantly higher only in SVM patients compared to HC. PLS exhibited slightly higher serum level in SVM compared to NSVM, but the difference in its serum abundance between the severe and non-severe stages was found to be statistically insignificant. Interestingly, SAA, TTN, SOD and CP exhibited differential abundance (> 1.5-fold up-regulated) between SVM and NSVM, indicating their potentiality as the predictive markers for malaria severity. In receiver operating characteristic (ROC) curve analysis, SAA, HP, and Apo A-I were found to be efficient (area under the curve (AUC) > 0.8) in discrimination of both NSVM and SVM patients from HC (Fig. 4B). However, the efficiency of these candidates was found to be comparatively lower for discrimination of DF or LEP from HC (Fig. S5). Importantly, SOD, VTN, HPX and TTN exhibited adequate prediction accuracy (AUC > 0.70) in discrimination of SVM from NSVM ( Fig. 4B; Table S8). Calibration-free concentration analysis (CFCA) using SPR, which allows the measurement of functionally active target proteins, also indicated a progressive increase in serum abundance of SAA with increasing disease severity (Fig. 5A). SAA concentration measured in pooled serum of HC and DF patients was observed to be in   Table 2. List of the differentially abundant serum proteins identified in NSVM and SVM (a) . (a) This is a partial list for some selected differentially abundant serum proteins (fold-change ≥ 1.4 in NSVM/HC or SVM/ HC with a p-value ≤ 0.05; NS indicates p value > 0.05) identified in iTRAQ-based quantitative proteomics analysis (identified with ≥ 2 unique peptides) using ESI-Q-TOF instrument (complete lists are provided in Table S4). (b) Median value for the identified unique peptides in different replicates is represented. (c) Associated pathways obtained from Uniprot database. (d) Differential abundance for these candidates is also identified in iTRAQ analysis using Q-Exactive (details are provided in Table S5). (e) Differential abundance for these candidates is also identified in 2D-DIGE analysis (details are provided in Table S3). (f) These candidates are validated by ELISA (details are provided in Table S7). * Fold-change value for differential abundance derived after immunodepletion of albumin and IgGs. a very comparable range (Table S9A). Interestingly, the measured concentrations of SAA in the pooled serum of both the NSVM and SVM patients were found to be significantly higher compared to HC (p < 0.005). Analysis of individual samples also presented similar results as obtained with the pooled samples (Table S9B). Nearly similar serum level of SAA was observed in the individual samples of HC and DF. In contrast, a significant increase in SAA level was observed in NSVM and SVM (p < 0.005) ( Table S9B).

Measurements of increased oxidative stress in vivax malaria patients.
In ELISA analysis, nearly 2-fold up-regulation in serum level of SOD was observed in SVM as compared to NSVM and HC. Further analysis of SOD activity in HC, NSVM and SVM sera also indicated an elevated level (more than 2-fold) of this antioxidant enzyme in severe infection (Fig. 5B). However, nearly equal serum level of SOD has been observed in HC and NSVM study cohorts (Table S10). Induction of such antioxidant defense in SVM patients is an apparent indicator of oxidative stress due to the severe infection by the parasite. Serum level of thiobarbituric acid reactive substances (TBARS), which serves as an indicator of lipid peroxidation and oxidative stress, was also found to be around 1.5-fold higher in SVM compared to NSVM and HC ( Fig. 5C; Table S11).

Discussion
In plasmodial infections, mortality occurs mainly due to the severe disease syndromes leading to cerebral, renal, pulmonary or hepatic involvement and often multi-organ dysfunctions. Although, P. falciparum infection represents the major cause of malaria associated fatality, P. vivax can also cause severe infections with deadly complications 18,19 . P. vivax can cause severe and fatal manifestations even at a very low parasite biomass 5 and there is no precise, systematic global assessment of endemicity for vivax malaria 6 . Consequently, diagnosis, treatment and control of vivax malaria are even more difficult than falciparum malaria. Pathogenesis of severe vivax malaria is still largely a black box since the mechanisms that trigger the transition of uncomplicated malaria into severe-complicated manifestations are merely unidentified 19 . Sequestration of infected erythrocytes in brain capillary endothelia leads to cerebral malaria in P. falciparum infection, while in case of vivax malaria the cause for coma is not precisely known. Since P. vivax does not sequester, neurological syndromes could be due to the release of toxic molecules in the circulation 20 . High-throughput "omics" technologies are capable of fast and sensitive screening of thousands of biomolecules. Therefore, such integrated approaches are effective for studying various alterations in physiological system under diseased conditions and are valuable for investigation of underlying mechanisms of disease pathogenesis and identification of next-generation diagnostic and therapeutic targets. In this multidisciplinary prospective study we performed an inclusive analysis of clinicopathological parameters and serum proteome profiles of NSVM and SVM patients to decipher pathogenesis and identify surrogate markers of severity. However, due to the lack of reliable standard case definitions for severe vivax malaria, parameters for severe falciparum malaria defined by WHO were extrapolated during the selection of vivax malaria patients. Consequently, it is expected that some of the clinicopathological and proteomics features identified in the severe vivax malaria patients in our study will mimic the alterations observed in severe falciparum malaria in earlier studies.
In this study, hepatic dysfunction, renal dysfunction, severe anemia, hypoglycaemia, acute respiratory distress syndrome (ARDS), cerebral manifestation and multiple organ involvement were observed in severe vivax malaria. Comparative proteomics analysis of HC, NSVM and SVM patients indicated alteration in serum levels of diverse classes of proteins associated with the coagulation pathway and hemostasis, complement cascade, muscle contraction and cytoskeletal regulation, signaling in immune system and acute phase response and lipid metabolism and transport in P. vivax infections. Aberrations in multiple essential physiological processes can provide a glimpse of the pathogenesis of severe manifestations of vivax malaria (Table 3). Serum proteins which exhibited alterations in their abundance equally in SVM and NSVM compared to HC, could be indicators of the febrile status of the subjects, but may not provide any direct correlation with the malaria pathogenesis; whereas proteins showing sequential changes in their abundance with respect to the increased disease severity can serve as good indicators for disease progression and complications.
Clinicopathological analysis and proteomics profiling indicated alterations in different hematological parameters and serum abundance of some components of the coagulation system in malaria patients. Thrombocytopenia is observed almost invariably in malaria infection and can be used as a sensitive but non-specific marker of acute infection 21 . In our study, platelet count was found to be consistently lower in NSVM and SVM patients in comparison to HC. Platelets are sequestrated by macrophages in the spleen due to immune mediated injury led by immune complexes of malarial antigen 22 . Additionally, there could be platelet mediated 'clump' formation of parasite infected erythrocytes, which requires the platelet glycoprotein CD36 23 . The severity of anemia is generally found to be more in P. falciparum infection, which could be due to the invasion of P. falciparum parasites to red cells of all ages even as early as orthochromatic erythroblasts, whereas P. vivax has specific predilection to infect only young RBC's (reticulocytes) limiting the parasitemia level and RBC lysis 24 . Reduced levels of hemoglobin in malaria is well established; however, changes in hemoglobin level depend on multiple factors such as high level of malaria endemicity, nutritional and socio-demographic status, existing hemoglobinopathies and immunity from malaria 25 . Our analysis also indicated decrease in hemoglobin concentration in all the disease groups, but mild anemia was observed only in SVM patients (Hb < 10 g/dL). It could be due to the increased hemolysis and decreased rate of erythrocytes production during the severe infection 26 .
Severe malaria is associated with systemic inflammation, increased capillary permeability, hypoxia, acidosis, endothelial activation and microvascular coagulopathy. Plasmodial infections influence blood coagulation by various pathobiological mechanisms which could be the critical components of malaria pathogenesis 27 . Strikingly, in our study majority of the blood coagulation cascade components were found to be unaltered in severe vivax malaria. To this end, activation of coagulation cascade is often observed in falciparum malaria Scientific RepoRts | 6:24557 | DOI: 10.1038/srep24557 patients which leads to thrombocytopenia 28 . Intriguingly, thrombocytopenia in vivax malaria is generally mediated by an immune mechanism in the absence of blood coagulation activation 29 . This indicates a notable difference in the pathogenesis of these two plasmodial infections.
Elevated serum levels of various muscle proteins in circulation indicate the possibilities of muscle damage and microvasculature lesions in severe malaria. A recent study by Bachmann et al. demonstrated increased levels of multiple muscle-specific proteins including creatine kinase, carbonic anhydrase III and myosin XVA in children with cerebral falciparum malaria. Interestingly, we have identified elevated serum levels of NEB, VTN, and TTN in SVM patients, while in non-severe infection, serum levels of these candidates were found to be nearly similar compared to the healthy controls. This finding may raise the possibility of endoplasmic reticulum stress and cytoskeletal involvement in malaria. Earlier studies reported significant alterations of the red cell cytoskeleton in falciparum malaria patients 30 . Previous studies support the involvement of apoptotic pathways, ER stress and mitochondrial toxicity in malaria 31 . However, most of the earlier studies have shown adverse alterations in the serum abundance of diverse classes of proteins associated with cytoskeleton system and muscle contraction in falciparum malaria. We are anticipating that the elevated serum levels of cytoskeletal proteins could be one of the major cues for the severe P. vivax infection. A previous study by Baker et al. presented evidences for the involvement of circulating neutrophil extracellular traps (NETs) in a protective mechanism against falciparum malaria 32 . To this end, it would be informative to measure the blood levels of NETs in NSVM and SVM patients because these extracellular fibers are released by rupturing plasma membrane, and thereby may lead to some cytoskeletal dysregulation.
Oxidative damage of platelets plays a crucial role in the pathogenesis of thrombocytopenia found in P. vivax malaria 33 . Plasmodial infection develops oxidative stress in liver, which triggers the induction of apoptosis by activating mitochondrial pathways 34 . In our study, HPX and CP were found to be up-regulated in both NSVM and SVM patients. An increase in the serum level of CP could be one of the protective mechanisms to trap the reactive oxygen species (ROS) against lipid peroxidation and could serve as a marker for the acute phase response in malaria 35 . Sohail et al. demonstrated that measurement of Glutathione-S-transferase, lipid peroxidation and Catalase could be implemented as reliable biochemical markers for vivax malaria 36 . Interestingly, we found an elevation in SOD-1 abundance and activity in the severe malaria patients. SOD plays a vital role in circumventing the ROS (mainly superoxide anions) produced during plasmodial infections. Plasma SOD-1 has been reported earlier as a surrogate marker of vivax malaria severity 37 . Lipid peroxidation is also considered as an effective indicator of oxidative stress, while measurement of TBARS is often used for screening and monitoring lipid peroxidation 38,39 . Thus the elevated level of TBARS observed in SVM patients is eventually providing a glimpse regarding the higher echelon of oxidative stress in severe form of the infection. Taken together, differential serum abundance of multiple proteins, free-radical scavenging enzymes, anti-oxidative enzymes and advanced oxidation protein products (AOPP) cumulatively represent the oxidative stress and anti-oxidative status of the patients suffering from malaria and reflect the severity level of the infection.
In summary, results obtained from this comprehensive proteomics analysis revealed activation of the oxidative stress and counteractive pathways, as well as elevated serum levels of cytoskeletal proteins as the possible cues contributing towards SVM. Noticeably, blood coagulation cascade was not affected adversely in P. vivax infection, Activation of pro-inflammatory responses and cytokine alterations have been observed in cases of malaria, especially severe malaria with an increase in levels of TNF-α , IL-1β , IFN-γ and other mediators during the initial stages of infection. The upregulation of a large number of acute phase reactants such as C-reactive protein, serum amyloid A and P, has also been observed, suggesting the role of a strong inflammatory response as the host's major defence mechanism against infection. Previous findings related to the haemostatic changes associated with malaria, that seemed to correlate strongly with parasitemia, have provided sufficient evidence for the implication of coagulation cascades in the pathogenesis of the disease. Alterations in the red cell membrane composition as well as the endothelial cell destruction associated with malaria have been found to activate blood coagulation, thereby leading to the activation of the complement system. However, it still remains unclear whether these alterations reflect the result or cause of the pathogenic process, since recent evidence has suggested their critical roles in cerebral and placental malaria.

27,50-53
3 Oxidative stress c 12 Oxidative stress in response to malaria infection is induced due to liberation of heme upon haemoglobin breakdown by parasites within erythrocytes leading to the generation of reactive oxygen species, decreased antioxidant activity within erythrocytes and platelets and increased lipid peroxidation. The resulting oxidative stress has been suggested to be one of the major mediators of erythrocyte damage, anemia, thrombocytopenia and hepatic dysfunction in malaria.  Table 3. Modulation of various physiological pathways in vivax malaria. a Identified in IPA, PANTHER, and DAVID. b Identified in IPA and DAVID. c Identified in IPA only. # Some representative studies have been included to restrict the total number of references within the maximum allowed limit.
which is often observed in falciparum malaria. Serum abundance of proteins involved in cytoskeletal system and apoptotic pathways were highly modulated in severe P. vivax infection, which was not observed in the non-severe cases. Additionally, serum levels of a few proteins identified in this study including SAA, CRP, SOD, HP, Apo E, Apo A-I and TTN exhibited sound correlation with disease severity, and thereby could serve as potential indicators for severity of malarial infection. Hitherto, there is a serious lack of a "gold standard" definition for severe vivax malaria. Parameters for severe falciparum malaria defined by WHO cannot be merely extrapolated to define severe P. vivax infection, since there are considerable differences between the pathogenesis of these two plasmodial infections. We speculate that apart from clinicopathological parameters, serum/plasma levels of protein candidates identified in this study such as SOD, Apo E, VTN, TTN, SAA, and HP are potential predictive markers for malaria severity and could be studied further for improving case definitions for severe malaria.  Subject recruitment and criteria for inclusion and exclusion. This study was conducted by recruiting malaria patients and controls from both the urban and rural populations from three different malaria endemic regions of India: Mumbai, Kolkata and Bikaner. Recruitment of the subjects was carried out during 2010 to 2013 at different study sites. Patients (adult, of either gender) with SVM and NSVM were diagnosed by clinical manifestations and microscopic examination of peripheral smear, and results were confirmed through RDT. Patients were classified as "severe malaria" based on the clinical description of severe falciparum malaria as per the WHO guidelines for malaria 40 . Blood specimens were collected from age and sex matched HC and patients with two non-malarial febrile infectious diseases; DF and LEP for performing a comparative analysis. Malaria patients with a diagnosis of mixed-species infection (infected with both P. vivax and P. falciparum) or co-infection(s) with any other infectious disease were excluded from this study. Female patients who were pregnant at the time of enrolment and participants below 18 and above 65 year age were not enrolled in this study. Subjects with a history of significant systemic diseases such as autoimmune disorders, chronic liver diseases and bleeding disorders or Blood collection and serum separation. Blood samples (5.0 mL) were collected from the antecubital vein of the patients (during the febrile stage of infection) and HC subjects using serum separation tubes (BD Vacutainer ® ; BD Biosciences). Serum separation and storage was performed as described previously 16 .

Methods
Analysis of clinicopathological parameters. After evaluation of clinical manifestations and confirmation of diagnosis, hematological and biochemical parameters including complete blood count, total bilirubin, creatinine, ALT, AST and ALP were measured in the study groups. Hematological investigations were carried out using a fully automated cell counter and biochemical measurements were performed using automated chemical analyzers. ESR was measured by using the Westergren method 41 . For each parameter, normality of the data distribution was assessed using the Kolmogorov Smirnov test and the data of study and control groups were compared by using Mann Whitney U test. A p-value < 0.05 was considered as statistically significant.
Sample processing, protein extraction and 2D-DIGE. Depletion of the high abundant serum proteins, protein extraction, and 2D-DIGE were performed as described previously 16 . In brief, protein extraction from depleted serum samples was performed using TCA/acetone precipitation method. 50 μg protein samples (NSVM/ SVM and HC) were labeled with 400 pmol of Cy3 or Cy5 following the manufacturer's instructions (GE Healthcare, Uppsala, Sweden). Dye-swapping was performed while labeling the malaria and control samples to exclude any type of labeling effects. A mixture of equal amounts of all the samples to be analyzed in this study, regarded as an internal standard, was labeled with Cy2. CyDye labeled protein samples were focused on linear pH 4-7 IPG strips (18 cm) and then separated on 12.5% polyacrylamide gels. After electrophoretic separation of the proteins, 2D gels were scanned using a Typhoon 9550 Variable Mode Imager (GE Healthcare). DIGE experiments were performed in three technical replicates. Image acquisition and data analysis were performed as described previously 16 .

MALDI-TOF/TOF analysis and protein identification.
In-gel digestion of the statistically significant (p < 0.05) differentially abundant protein spots identified in 2D-DIGE experiment and subsequent enrichment of the digested peptides using Zip-Tip C18 pipette tips (Millipore, USA) were performed as described previously 16 . In-solution digestion, iTRAQ labeling and OFFGEL fractionation. Serum samples from each of the study cohorts (HC, NSVM and SVM) were divided into three separate pools (n = 10) for iTRAQ-based quantitative proteomics analysis. Protein samples (in rehydration solution) were exchanged to TEAB buffer using Amicon Ultra 0.5 mL centrifugal 3 kDa filters (Millipore, Watford, UK). After buffer exchange, quantification of protein content in each sample was performed using QuickStart Bradford reagent (BioRad, USA). In-solution digestion (75 μg proteins from each sample) and subsequent iTRAQ labeling of the digested peptides were performed following the manufacturer's instructions (AB Sciex, USA). HC samples were labeled with the 114 iTRAQ reagent, while NSVM and SVM samples were labeled with 115 and 116 iTRAQ labels, respectively. OFFGEL fractionation of the labeled peptides was performed using a 3100 OFFGEL fractionator (Agilent Technologies, Santa Clara, CA) with high resolution (pH 4-7, 24 cm) IPG strips. In-solution digestion, iTRAQ labeling, and OFFGEL fractionation protocols are described in details elsewhere 42 .

LC-MS/MS analysis for the protein identification and quantitation.
iTRAQ-based quantitative proteomics analysis was performed using two mass spectrometry platforms; Agilent 6550 Q-TOF and Thermo Scientific Q-Exactive. Agilent 6550 iFunnel Q-TOF LC MS/MS instrument (Agilent Technologies, USA) equipped with a Chip-Cube controlled by the Mass Hunter acquisition software was set to perform data acquisition in a positive ion mode as described previously 42 . Data files were processed by Spectrum Mill Protein Identification software (Agilent Technologies, USA) using the Paragon algorithm and Mascot version 2.2 (Matrix Science, London, UK). Searches were performed against UniProt database specifying Homo sapiens taxonomy (Proteome ID: UP000005640; Organism ID: 9606; Protein count: 70225). Data were extracted between MH + 600 and 4000. IAA for cysteine and iTRAQ (N-term, K) were selected as fixed modifications and oxidized methionine was specified as the variable modification. 20 ppm precursor mass tolerance and 50 ppm fragment mass error tolerance were specified. Peptides identified with confidence interval (CI) values above 95% were used for protein identification and quantification. iTRAQ labeled samples (a single pool containing all the samples) were also analyzed using Q-Exactive mass spectrometer (Thermo Fisher Scientific, MA, USA). Details for the LC conditions and MS data acquisition and analysis parameters have been described previously elsewhere 43 . In brief, the vacuum dried iTRAQ labeled samples were re-suspended in 0.1% formic acid in water and injected into nano flow HPLC pump coupled online with the Q-Exactive Orbitrap mass spectrometer with a nano-electrospary ion source. The samples were analysed using a 90-min linear gradient of buffer B (95% acetonitrile and 0.1% formic acid) at a flow rate of 300 nL/min. Full MS scans were done in the range of 350-1800 m/z at a resolution of 70000 with a target value of 1.00 + E 6 and an allowed ion accumulation time of 60 ms. All the raw .msf files were processed using proteome discoverer 1.4 (Thermo Fisher Scientific); Mascot 2.2.4 and SEQUEST were used for database searching against the Uniprot Homo sapiens FASTA. The database searching parameters included precursor ion mass tolerance of 5 ppm and fragment mass tolerance of 0.02 Da.

Statistical analysis of iTRAQ-based quantitative proteomics data. Normalization and statistical
analysis of the Q-TOF datasets was performed using the Perseus workstation (version 1.5.2.6). Reverse and contaminant database hits were filtered out during data processing in Perseus. Categorical annotation was applied to group reporter ion intensities, values were log2 transformed, and were normalized by "subtract (mean)" followed by Z score normalization. Proteins groups were filtered for valid values, and p-values obtained from a paired t-test were used to evaluate significance of differences in the protein abundances between HC and malaria (NSVM/ SVM) study cohorts. P-values (adjusted) < 0.05 were considered to be statistically significant. Venny 2.0.2 was used to generate the Venn diagrams (http://bioinfogp.cnb.csic.es/tools/venny/index.html) 44 .

Proteins networks and bioinformatics analysis. Differentially abundant serum proteins (HC vs.
NSVM/SVM) satisfying the threshold of statistical significance (p-value < 0.05) in 2D-DIGE and/or iTRAQ (Q-TOF) analysis were subjected to bioinformatics analysis for determining their molecular functions, cellular component annotations, and possible connections with various biological processes, physiological pathways and networks. Differentially abundant proteins (with > 1.2-fold change in NSVM/HC and/or SVM/HC) identified in iTRAQ-based quantitative proteomics profiling using the Q-Exactive mass spectrometer were also included in pathway analysis. Bioinformatics analysis and functional clustering of the differentially abundant proteins was performed using IPA version 9.0 (Ingenuity ® Systems, www.ingenuity.com), PANTHER system, version 10 (http:// www. pantherdb.org) 45 and DAVID database version 6.7 (http://david.abcc.ncifcrf.gov/ home.jsp) 46  . Selection of the proteins for immunoassay-based validation was carried out on the basis of their fold-change, correlation with disease severity, association with malaria pathogenesis and commercial availability of the required ELISA kits. Quantitative direct/ competitive enzyme assay was carried out where standard and serum samples (malaria and controls) at a specific dilution were subjected to a microplate pre-coated with a polyclonal antibody specific for the target proteins. Optical densities were measured by using a SpectraMax M2e scanner (Molecular Devices, USA). Statistical significance of the average ratio of abundance was analyzed by Mann-Whitney U test (p < 0.05). Efficiency of the classifier proteins for prediction of NSVM, SVM and other two infectious diseases (DF and LEP) was analyzed by plotting ROC curves using GraphPad Prism software package (version 6.02).

SPR-based quantification of active protein concentration in serum samples. Quantification of
active serum protein concentration for one of the differentially abundant candidates, SAA, was performed by CFCA method using a Biacore T200 system (GE Healthcare) following the same methodology as described earlier 47 . HBS-EP + (10 mM HEPES pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.05% (v/v) P20) was used as the running buffer for the immobilization and interaction analysis. Anti-SAA was immobilized covalently on the surface of CM5 sensor chip using amine coupling chemistry to prepare the reaction surface appropriate for CFCA. CFCA for SAA protein was performed on serum samples of HC, vivax malaria (NSVM and SVM) and DF patients. Initially, three different pools (each pool containing 10 samples) from each study cohorts were analyzed. Results obtained from the pooled samples were further validated by performing CFCA on 7 individual serum samples from each group. For performing the concentration analysis, each serum sample was serially diluted to 100 and 200 folds (in running buffer) and injected in duplicates at flow rates of 5 μL/min and 100 μL/min for 120 sec over the active and reference flow cells at 25°C. Data obtained from the reference flow cell were automatically subtracted from experimental measurements to yield the specific signal. Protein concentration in each sample was then determined from the binding data using the CFCA evaluation feature of the software. SOD activity assay. SOD activity in the serum samples of HC, NSVM and SVM was measured using a commercially available kit (Cayman Chemical, USA) following the manufacturer's instructions. In brief, 10 μL of standard solution and samples (in duplicates) were added in the designated wells on a 96-well plate along with 200 μL of the diluted radical detector. Reactions were initiated by addition of 20 μL of diluted xanthine oxidase to all the wells. The plate was carefully vortexed for a few seconds and incubated on a shaker for 30 minutes at room temperature. The absorbance was measured at 440-460 nm using a SpectraMax M2e scanner (Molecular Devices, USA). TBARS assay. Serum levels of TBARS in HC, NSVM and SVM were measured using a commercially available kit (Cayman chemicals, USA). Reagents and colorimetric standards were prepared according to manufacturer's instructions. 100 μL of samples and standards (in duplicates) were added to labelled 5 mL vials containing 100 μL of SDS solution. 4 mL of color reagent was added and the vials were placed in boiling water for an hour. The vials were then immediately incubated on ice for 10 mins to stop the reaction. 150 μL from each vial was loaded onto clear plates at room temperature. The absorbance was measured at 530-540 nm using a SpectraMax M2e scanner (Molecular Devices, USA).