Identification of circulating biomarkers in sera of Plasmodium knowlesi-infected malaria patients – comparison against Plasmodium vivax infection

Background Plasmodium knowlesi was identified as the fifth major malaria parasite in humans. It presents severe clinical symptoms and leads to mortality as a result of hyperparasitemia in a short period of time. This study aimed to improve the current understanding of P. knowlesi and identify potential biomarkers for knowlesi malaria. Methods In the present study, we have employed two-dimensional gel electrophoresis-coupled immunoblotting techniques and mass spectrometry to identify novel circulating markers in sera from P. knowlesi-infected patients. Specifically, we have compared serum protein profiles from P. knowlesi-infected patients against those of healthy or P. vivax-infected individuals. Results We identified several immunoreactive proteins in malarial-infected subjects, including alpha-2-HS glycoprotein (AHSG), serotransferrin (TF), complement C3c (C3), hemopexin (HPX), zinc-2-alpha glycoprotein (ZAG1), apolipoprotein A1 (Apo-A1), haptoglobin (HAP), and alpha-1-B-glycoprotein (A1BG). However, only TF and HPX displayed enhanced antigenicity and specificity, suggesting that they might represent valid markers for detecting P. knowlesi infection. Additionally, six P. knowlesi-specific antigens were identified (K15, K16, K28, K29, K30, and K38). Moreover, although HAP antigenicity was observed during P. vivax infection, it was undetectable in P. knowlesi-infected subjects. Conclusions We have demonstrated the application of immunoproteomics approach to identify potential candidate biomarkers for knowlesi malaria infection.


Background
Malaria has been known to be caused by major human malaria parasites -Plasmodium falciparum, P. vivax, P. malariae and P. ovale. P. knowlesi which was initially found in only long-tailed (Macaca fascicularis) and pigtailed (Macaca nemestrina) macaques [1]. It has been recognized as the fifth species of Plasmodium that causes malaria infection in humans [2,3]. Several cases of P. knowlesi infection in humans were reported in Malaysian Borneo, Myanmar, Philippines, Singapore, and Thailand [1,4], supporting the notion that P. knowlesi represents a major cause of malaria in Southeast Asia. P.
knowlesi is the only malaria species that has a 24-hour asexual reproduction cycle (quotidian). Therefore, patients infected with P. knowlesi can easily reach lethal parasite densities in a relatively short period of time [5]. Respiratory distress, renal dysfunction, jaundice, hypoglycaemia and severe anaemia are the general clinical manifestations of severe malaria infection [6]. In addition, the severity of knowlesi malaria was found to be associated with hyperparasitemia and this has been reported in Malaysia and in other Southeast Asia regions [5,6].
Like other malaria species, identification of P. knowlesi infection is achieved through examination of thick and thin blood films followed by Giemsa microscopy. This method uses specific morphological characteristics to differentiate parasites. However, accurate diagnosis of P. knowlesi by microscopy is often limited by the fact that P. knowlesi bares strong morphological resemblance to P. falciparum (early trophozoite stage) and P. malariae (erythrocytic stages) [7]. Thus, diagnosis of P. knowlesi usually requires molecular detection methods, which are performed in reference laboratories. In this regard, polymerase chain reaction (PCR) and molecular characterization currently represent the most reliable detection methods for P. knowlesi infection. Nevertheless, PCR-based techniques are not suitable for routine identification, since this method requires parasite DNA and is time consuming. In addition, P. knowlesi is frequently misdiagnosed as P. malariae through PCR and conventional microscopy [5]. Currently, there are no commercially available malaria rapid diagnostic tests are designed specifically for P. knowlesi detection. Although Plasmodium lactate dehydrogenase (pLDH) assay is more reliable in detecting P. knowlesi infection, the crossreactivity of P. knowlesi with P. falciparum-specific and P. vivax-specific pLDH has been shown [8]. Therefore, as stated by Cox-Singh et al. [5], there is currently a fundamental need for effective and practical diagnostic methods, which will not only contribute to reduced malariaassociated complications and mortality, but also facilitate global malaria control.
Notably, host biomarkers can be used to assess the risk of infection, examine protection against active diseases, or to determine therapeutic responses. Indeed, employing gene profiling, Schaecher et al. [9] has identified host biomarkers that can differentiate between lethal and non-lethal blood stages of murine malaria. Serum samples collected during the course of infection were utilized to analyze differential protein expression patterns, which were found to correlate with the degree of infection. Such data can provide insight into the cell regulatory mechanisms that participate in pathogenesis, immune responses and host recovery. Moreover, the identified proteins could be measured by serological tests to detect variants of malaria parasites while conducting epidemiological studies or implementing control programs. For example, serum angioprotein I and angioprotein 2/L have been suggested as diagnostic and prognostic biomarkers as well as potential therapeutic targets in cerebral malaria [10]. In the present study, we have examined the proteomic profiles of serum from P. knowlesi-infected patients to identify distinctive immunological protein features. Specifically, we have compared serum protein profiles from P. knowlesi-infected patients against those of healthy or P. vivax-infected subjects. Our findings have the potential to enhance our understanding of P. knowlesi and might contribute to the development of novel diagnostic approaches.

Clinical samples
From 200 serum samples previously screened for malaria parasites [11], we selected 15 samples for the current study. Notably, these samples corresponded to patients who were newly diagnosed with either P. knowlesi (n = 9, parasitemia range: 0.04-22.80%, age range: 29-55 years, Male, Malaysian) or P. vivax (n = 6, parasitemia range: 0.10-0.50%, age range: 29-50 years, Male, non-Malaysian Asian). Additionally, 23 serum samples were collected from normal healthy individuals and were used as a control group (age range: 29-50 years, Male, Malaysian). All samples were obtained with patients' written consent, and this study was approved by the University of Malaya Medical Centre Ethical Committee in accordance with ICH-GCP guidelines for good clinical practice and the Declaration of Helsinki (PPUM/MDU/ 300/04/03).

Two-dimensional electrophoresis (2-DE)
Two-dimensional electrophoresis (2-DE) was performed as previously described by Chen et al. [12]. Briefly, 10 μl of unfractionated whole human serum (either individual serum or pooled sera) was subjected to isoelectric focusing using 13-cm rehydrated precast immobilized dry strips (pH 4-7) (GE Healthcare Bio-Sciences, Uppsala, Sweden). For the second dimension, focused sample within the strips was subjected to electrophoresis using an 8-18% gradient polyacrylamide gel in the presence of sodium dodecyl sulphate (SDS). All samples were analyzed in duplicate. The 2-DE gels were silver stained as described by Heukeshoven and Dernick [13]. For mass spectrometric analysis, gels were stained using Coomassie Brilliant Blue (CBB) or a modified mass spectrometry (MS) silver staining method, as described by Shenvchenko et al. [14]. CBB gel plugs were used instead of silverstained gel plugs when a higher peptide concentration was expected from in-gel digestion.

Mass spectrometry analysis and database search
Selected spots were excised and subjected to in-gel tryptic digestion using the commercially available Proteo Extract™ All-in-One Trypsin Digestion Kit (Calbiochem, Darmstadt, Germany). Mass spectrometry (MS) analysis was performed at the Faculty of Biological Sciences Proteomic Centre, National University of Singapore. After digestion, the resulting peptides were mixed with CHCA matrix solution (5 mg/ml of cyano-4-hydroxycinamic acid in 0.1% trifluoroacetic acid [TFA] and 50% acetonitrile [ACN]) in 1:2 ratio and spotted onto a matrixassisted laser desorption/ionization (MALDI) target plate. Peptide mass spectra were obtained using an ABI 4800 Proteomics Analyzer MALDI-TOF/TOF Mass Spectrometer (Applied Biosystems, Foster City, CA, USA). For MS analysis, 1,000 shots were accumulated for each sample. MS data were automatically obtained with the five most intense ions for MS/MS. Peptides were subsequently subjected to MS/MS analyses using air with collision energy of 2 kV and a collision gas pressure of~1×10 −6 Torr. The stop conditions were set to accumulate approximately 2,000 to 3,000 shots, depending on the quality of the spectra. The Mascot search engine (version 2.1; Matrix Science, London, UK) was used to analyze all of the tandem MS results. Also, GPS Explorer™ software (version 3.6; Applied Biosystems) was employed in combination with the Mascot search engine for peptide identification. The search parameters allowed for Nterminal acetylation, C-terminal cysteine carbamidomethylation (fixed modification), and methionine oxidation (variable modification). The peptide and fragment mass tolerance were set to 100 ppm and ±0.2 Da, respectively. Moreover, peptide mass fingerprinting (PMF) parameters for the data search were as follows: one missed cleavage allowed in trypsin digest; monoisotropic mass value; ±0.1 Da peptide mass tolerance; and 1+ peptide charge state. Initial protein identification was determined by comparing peptide masses to a database of tryptic peptides from known proteins (ProteinPilot proteomics software [4800] Proteomic Analyzer; Applied Biosystems), and a score was assigned based on similarity to theoretically and experimentally determined masses. Analyses were conducted using International Protein Index (http://www. ebi.ac.uk/IPI/), NCBI, Unigene (version 3.38), and Plas-moDB (version 8.0; http://plasmodb.org/plasmo/) databases for human proteomics. There were a total of 10,719 entries included in the database search. Search scores of >82 (Mascot NCBI database) or >30 (Mascot search engine using PlasmoDB) were considered as significant.

Immunoblotting
For immunoblotting, we analyzed 10 μl of pooled, unfractionated human serum. Following electrophoresis, the 2-DE gels for the pooled serum samples were grouped into five categories: (a) normal pooled sera probed with normal pooled sera, (b) normal pooled sera probed with P. knowlesi pooled sera, (c) P. knowlesi pooled sera probed with normal pooled sera, (d) P. knowlesi pooled sera probed with P. knowlesi pooled sera, (e) P. knowlesi pooled sera probed with P. vivax pooled sera. Each of the gels was transferred onto nitrocellulose membrane using the Multiphor II Novablot semi-dry system (GE Healthcare, Sweden). The blotted nitrocellulose membranes were then blocked with SuperBlock (Pierce, USA) and washed three times with Tris-buffered saline-Tween 20 (TBST). The membranes were subsequently incubated overnight (4°C) with the indicated primary antibodies, which corresponded to P. knowlesi infection, P. vivax infection, or normal healthy controls (all diluted at 1:50). After another washing step, the membranes were incubated with monoclonal antihuman Immunoglobulin M (IgM) conjugated to horseradish peroxidase (HRP) (Invitrogen, USA) at a dilution of 1:6,000 for 1 h at room temperature. The resulting immunocomplexes were visualized using chemiluminescent blotting reagent (Pierce, USA) and X-ray film (18 × 24 cm; Kodak).

Differential image and data analysis
We utilized LabScan image scanner (version 5; Amersham, Germany) to capture and store the 2-DE gel images. PD-Quest™ 2-D gel analysis software (Version 8.0.1, Bio-Rad) was used to detect, match, and quantify the distinct protein spots. In order to identify differentially expressed serum proteins, the percentages of volume contributions (vol%) were calculated (i.e., the spot volume of a specific protein as a percentage of the total spot volume of all proteins in the gel, including unresolved peptides).

Statistical analysis
Protein levels in the gels are presented as mean vol% ± SD (standard deviation). The variance ratio test (F) was used to analyse differences between control subjects and patients. Correlations between the variables were examined using GraphPad Prism 5 software. A p-value of less than 0.05 was considered as statistically significant.

Functional enrichment and protein interaction analysis
Functional enrichment and protein interactions were analysed using web-based bioinformatics tools. DAVID v6.7 (Database for Annotation, Visualization and Integrated Discovery) was employed for protein functional enrichment analysis [15,16]. DAVID bioinformatics provide a comprehensive biological knowledgebase of functional annotation tools for understanding the biological meaning behind large lists of genes or proteins. The functional categorization is considered significant when the p-value is less than 0.05. The identified host-specific proteins were further evaluated using STRING v9.1 (Search Tool for the Retrieval of Interacting Genes), which is an application that aggregates available databases of known and predicted protein-protein associations [17].

Differential expression of antigenic proteins
We have performed a 2-DE assay that allowed for rapid detection and differentiation of P. knowlesi from other Plasmodium species. We first separated unfractionated normal control sera samples with 2-DE and observed high-resolution profiles, which were made up of several distinct protein clusters (Figure 1a). Sera samples from P. knowlesi-infected patients, presumably containing P. knowlesi antigens, were also separated by 2-DE ( Figure 1b). Subsequently, the 2-DE profiles from these patients were compared with those obtained from normal individuals, revealing key differences in the expression of several serum proteins. Notably, knowlesi malaria patients displayed protein spots/clusters that appeared to undergo up-or downregulation ( Figure 1b).

Identification of expressed biomarkers by mass spectrometry
We analysed the protein clusters using PD-Quest TM 2-DE gel analysis software (version 8.0.1, Bio-Rad), which indicated that there were a total of 200 and 124 protein spots detected by 2-DE for the P. knowlesi and normal samples, respectively. Therefore, 76 protein spots were differentially observed in malaria knowlesi sera compared to normal controls.
The above results were confirmed through the use of MALDI-TOF/TOF and database searches. The Plas-moDB database was employed to blast MS/MS-identified peptides against the hypothetical P. knowlesi H strain proteome, which is only comprised of computationally predicted sequences [18]. This database contains the genomes of various mammalian Plasmodium species (P. falciparum, P. knowlesi, P. vivax, and P. yoelii), which range from 23 to 27 Mb across 14 chromosomes and comprise approximately 5,500 genes (http://plasmodb.org/plasmo/). A remarkable 77% of these genes represent orthologous between these four species, and almost one-half of the genes encode conserved hypothetical proteins of unknown function. However, although PlasmoDB contains the most up-to-date annotation on P. knowlesi sequences, it is incomplete. Thus, it cannot be ruled out that some genes may be missing or incorrectly annotated in the PlasmoDB database. For this reason, experimental verification will be needed to assess whether current data offer a comprehensive view of the Plasmodium proteome, especially in the case of P. knowlesi. Nevertheless, our findings provide evidence to support the existence of some of these predicted sequences. The Mascot accession numbers, isoelectric points (pI), and molecular mass (Mr) values associated with the identified proteins are listed in Tables 4 and 5.

Detection of immunocomplexed biomarkers by 2-DE immunoblotting
In order to confirm the 2-DE image analysis and MS results, we performed immunoblotting with pooled patient sera (anti-P. knowlesi antibodies). Specifically, our immunoblotting analysis involved five distinct conditions to allow  Fold change measures the degree of change in the protein of the P. knowlesiinfected individuals (n = 9), compared to normal controls (n = 23). This is measured by dividing the average spot intensity in the infected persons by the average spot intensity in the controls.
direct comparison of P. knowlesi-infected sera against normal sera and sera from P. vivax-infected patients. The following conditions were tested (categories a-e): (a) normal pooled sera probed with normal pooled sera, (b) normal pooled sera probed with P. knowlesi pooled sera, (c) P. knowlesi pooled sera probed with normal pooled sera, (d) P. knowlesi pooled sera probed with P. knowlesi pooled sera, (e) P. knowlesi pooled sera probed with P. vivax pooled sera are displayed in Figures 2 and 3. Immunoblotting of 2-DE membranes revealed significant variations in the control (Figure 2a, b) and patient (Figures 2c, d and 3) sera profiles. In Figure 2, only categories 'a' (negative control) and 'c' (positive control) failed to show immunogenic spots. However, the remaining categories (Figures 2b, d, and 3) revealed many immunogenic host-specific and malaria antigens. Strikingly, we could specifically distinguish differences when immunoblotting vivax malaria patients with sera from knowlesi-infected patients. Immunogenic host-specific proteins are displayed in Table 2. Some of these malaria serum-reactive proteins could be detected in both non-infected and infected serum, including AHSG, A1BG, and Apo-AI (Figure 2b, d). However, ZAG1 and C3 appeared prominently for both individual patients in categories d and e. In addition, the A1BG signal was observed in categories d and e. Notably, HAP was the only protein that showed immunogenicity in category 'e' but not in category 'd'. TF was detected in category 'd'. HPX only appeared in the category 'd' immunoblot, whereas HRG was exclusively observed in category 'e'.

Functional enrichment and protein interaction analysis
In order to extend the above results, we employed DAVID v6.7 (http://david.abcc.ncifcrf.gov/) to perform functional ontology enrichment analyses (i.e., biological processes, molecular functions, cellular components, and pathways) for 11 significantly identified proteins ( Table 6). This functional analysis revealed that AHSG, C3, and histidine-rich glycoprotein precursor (HRG) were involved in regulating responses to external stimuli, whereas HAP, HPX and TF participated in iron homeostasis. In addition, molecular function analysis showed that AHSG, C3 and HRG play a role in endopeptidase and peptidase inhibitor activity. Also, the majority of the proteins were located in the extracellular region. However, cellular components analysis revealed that Apo-A1, fibrinogen γ chain (FGG), HRG, and TF could be found in secretory granules and membrane-bound vesicles. In addition, KEGG pathway analysis indicated significant participation of C3 and FGG in complement and coagulation cascades pathways (p = 0.026957).

Discussion
The global impact of malaria has sparked interest in developing effective diagnostic strategies, which are not only essential for resource-limited areas where malaria is a substantial societal burden, but also for developed countries [18,20]. In theory, the detection of malaria parasites or antigens in patient blood should be simple. However, diagnostic efficacy is subject to many important variables, including the various stages of erythrocytic schizogony, species endemicity, interrelationship between levels of transmission, population movement, parasitemia, immunity, and species-specific symptoms [11]. Among the Plasmodium species, detection of P. knowlesi is the most challenging due to morphological similarities and physical co-localization of genetic loci on the same chromosome within an individual or species. In this respect, Carlton et al. was the first to generate and compare a synteny map of P. falciparum, P. vivax, P. knowlesi, and the rodent malaria parasites (P. yoelii, P. berghei, and P. chabaudi) [19]. Overall, high synteny was observed between P. vivax and P. knowlesi chromosomes, with the exception of microsyntenic breaks in species-specific genes (e.g., P. knowlesi kir and SICAvar) [21]. These breaks were recognized as foci involved in the evolution of host-parasite interaction genes [22]. Proteomic analysis by 2-DE has been widely used to identify specific bacterial or viral antigens during vaccine and/or serodiagnostic test development [23]. Nevertheless, there have been few proteomic studies focused at elucidating pathogenic mechanisms or identifying potential diagnostic markers in malaria [24][25][26]. In the present study, we have analysed serum samples from infected individuals to identify and characterize novel markers of malaria infection using 2-DE coupled with immunoblotting techniques and mass spectrometry analysis. It is known that the immune response is relative with the presence of parasite density. Therefore, the immunoproteomics approach used in this study can be applied to identify antigens targeted by the immune system that response to infection. In addition, several studies have also reported on the protective role of IgM in the immune response during malaria infection [27,28]. Thus, both the host proteins and P. knowlesi antigens, represents the potential candidate biomarkers that could be used in the development of future diagnostic tools for P. knowlesi.
It has been suggested that invasion of erythrocytes by malaria parasites involves specific interactions between parasite receptors and erythrocyte ligands [29,30]. For this process, glycoproteins on the parasite receptors bind to hydrophobic peptides on the surface of human erythrocytes. Thus, hydrophobic peptides, which can be found in the form of sialic acid-rich regions, could make cells susceptible to infection by creating a negative charge at the surface [29]. Based on our findings, TF, HPX and HAP constitute very selective biomarkers for malaria infection. These three proteins are known as iron-binding glycoproteins which are involved in the  regulation of iron homeostasis that plays a key role in the innate immune response [31]. Iron is essential for the development of malaria parasite. It is known that Plasmodium parasites synthesize their own TF receptors, which can become localized on the surface of infected cells [32][33][34]. The delivery of extracellular iron from TF to infected erythrocytes is the source of ferric ions for malaria parasites. The alteration of TF level may influence the balance between inhibiting and promoting the survival of malaria parasite [35]. Thus, it is possible that P. knowlesi and P. vivax directly secrete TF to gain access to iron through receptormediated endocytosis.
Notably, it was reported that iron deficiency induces TF receptor expression and doubles the number of HPX surface receptors (subsequently increasing HPXmediated heme uptake in vitro) [36]. Under homeostasis, HPX can scavenge most of the free heme to form heme-HPX complex which prevent the onset of malaria [37]. Interestingly, some bacterial species, such as Haemophilus influenzae, Campylobacter jejuni, and Yersinia pestis [38][39][40][41], possess specialized iron acquisition systems for survival in hosts and are capable of heme uptake through heme-HPX complexes. Although the heme uptake activity is not parasite specific, it is possible that a similar survival system might be utilized by Plasmodium species. The expression of HPX has also been identified in several malaria studies, where this protein provides the support line of defence against haemoglobinmediated oxidative damage during intravascular haemolysis [30,42].
Additionally, our results have demonstrated that HAP is downregulated in malaria patients. However, HAP was only antigenic in sera from P. vivax-infected individuals and not from P. knowlesi-infected patients. High HAP expression can reduce symptoms associated with malaria by causing toxicity to Plasmodium parasites [43] and by removing free haemoglobin (Hb) following Plasmodiumtriggered haemolysis. In this regard, evidence has indicated that a higher peak of parasitemia and/or parasite burden was found in P. berghei (ANKA)-or Plasmodium chabaudi-infected mice compared to that of wild type [44]. In short, decreased expression of HAP could contribute to the life-threatening levels of parasitemia observed in P. knowlesi-infected patients.

Conclusions
In summary, we have demonstrated the application of immunoproteomics approach to understand the immune response and identify potential candidate biomarkers for knowlesi malaria infection. Taken together, we have specifically identified TF, HPX and HAP as antigenic markers in P. knowlesi. A further investigation on the functional roles of the identified potential biomarkers in larger clinical samples will be valuable to enhance our current understanding of P. knowlesi and to develop effective diagnostic tools to detect knowlesi malaria. Figure 4 Interaction networks of identified host specific proteins using STRING v9.1. STRING database is a curated knowledge database that relies on various evidences from highthroughput proteomic, genomic and co-expression studies. The predicted functions of proteins are shown.