In this study, asynchronized culture was employed. The purpose of using asynchronized culture was to capture small metabolite molecules that might be commonly released by all stages of parasites. The laboratory-adapted 3D7 strain of P. falciparum was used. The asynchronized blood stage parasites were cultured as described [14] in RPMI 1640 medium supplemented with 10% heat-inactivated O+ human serum, 1 ug/ml gentamicin, 36 uM hypoxanthine, 31 mM HEPES and 25 mM sodium bicarbonate. Four flasks of parasite culture with 3% haematocrit and 0.5% starting parasitaemia were prepared at the same time using red blood cells from different donors. At the same time, four flasks of culture without parasite inoculation but containing the same culture medium and haematocrit were also prepared. Culture materials, including supernatants and cell pellets, from all the infected and not infected flasks were collected at 12, 24, 36, and 48 hours without changing and adding culture medium. In total, 16 supernatant samples obtained from the infected flasks and the same numbers of supernatant samples from non-infected flasks were used for this study. All the samples used in this study were mycoplasma free. Parasite densities in the infected flasks were recorded as % count by blood smear reading. They increased with time, 0.83%±0.1%, 1.05%±0.06%, 1.45%±0.20% and 2.38%±0.30% (mean±SD) at 12, 24, 36 and 48 hours, respectively.

All the samples were run in duplicate. One hundred μl aliquots of supernatant samples were treated with acetonitrile (2:1, v/v), spiked with 2.5 μl internal standard mix, and centrifuged at 14,000 × g for 5 minutes at 4°C to remove protein as described previously [15]. Then LTQ-FTMS (Thermo, hybrid linear ion trap-Fourier Transform Ion Cyclotron Resonance mass spectrometry, MA, USA) coupled with C18 liquid chromatography was run on those collected samples. HRM offers an important advantage in analysis of highly complex metabolite mixtures, such as biological extracts, because detection of mass/charge (m/z) with 5 ppm or better mass resolution and mass accuracy substantially decreases the demand for physical separation prior to detection. Detection of m/z of ions from 85 to 850 with 50,000 resolution over 10 min LC runs with data extraction using apLCMS [16] provides a minimum of 3,000 reproducible features, many with sufficient mass accuracy to allow prediction of elemental composition. A m/z feature is defined by m/z, RT (retention time), and ion intensity (integrated ion intensity for the peak). The Kyoto Encyclopedia of Genes and Genomes (KEGG) database [17] was utilized to map the features distribution on both human and Plasmodium metabolic pathways. Examination of m/z of metabolites in the KEGG [18,19] human and Plasmodium metabolomics pathways shows that less than 10% of metabolites are redundant with others in terms of elemental composition [18,19]. Identified metabolites are annotated using Metlin Mass Spectrometry Database [20,21]. Direct examination by MS/MS of selected accurate mass m/z features of human plasma showed that for many m/z, the ion dissociation patterns matched those of authentic standards with identical elution times. For such metabolites, quantification relative to stable isotope internal standards has a coefficient of variation (5 to 10%) and sensitivity (low nanomolar to picomolar), which is similar to other methods and sufficient in allowing targeted analysis of selected chemicals within the context of an information-rich non-targeted profiling of all m/z detected within biological samples.

Analyses were performed based on the results from both biological and technical replicates. Total features of culture supernatant were collected after processing mass spectral data with apLCMS. The features from duplicate LCMS analyses were averaged, log2 transformed and quantile normalized for subsequent statistical and bioinformatics analyses including univariate analysis, Manhattan plot, and false discovery rate (FDR) [22] to determine the significant metabolites between infected and non-infected cultures. Furthermore, the metabolic profiles were discriminated using Limma-hierarchical cluster analysis to separate two groups in association with metabolites. Limma is originally a package of Linear Models for Microarray to analyse the gene expression data arising from microarray or RNA-Seq technologies from Bioconductor. A core capability is the use of linear models to assess differential expression in the context of multifactor designed experiments. Limma provides the ability to make comparisons between many targets simultaneously including metabolites [23,24].

The database of Kyoto Encyclopedia of Genes and Genomes (KEGG) was utilized to map the features distribution on both human and plasmodium metabolic pathways. Detected m/z features matching known human and Plasmodium intermediary metabolites are shown in black dots in pathway maps; most human and known Plasmodium metabolic pathways are represented.

3-methylindole (M51458) and succinylacetone (D1415) were purchased from Sigma-Aldrich (MO, USA). A standard curve was made by known concentrations (0.1-0.5nmole) of the reagent grade compounds in cell culture media. Areas from the samples were plotted against concentrations and a standard curve with an r²>0.99. 3-methylindole and succinylacetone in supernatants from parasite culture were treated using 100 μl in microcentrifuge tubes. These tubes were centrifuged at 14,000× g at 4 degrees Celsius for 5 minutes. 10 μl of the supernatant was injected in mass spectrometry and resulting areas were noted. Concentrations of the compounds were calculated using the standard curve.

MWAS was used to identify changes in supernatants from non-infected and Plasmodium-infected cultures at all points (12, 24, 36, and 48 hrs). A Manhattan plot (Figure 1) combines a statistical test (e.g., p-value, ANOVA) with the magnitude of change and enables visual identification of statistically significant data-points (metabolites) that display large-magnitude changes. Multiple testing corrections like FDR adjusts p-values (q-values) derived from multiple statistical tests to correct for the occurrence of false positives. The Y axis represents the -log₁₀ of the raw p-value comparing supernatants of culture system between non-infected and Plasmodium-infected red blood cells. The X axis indicated m/z ranging 85-850 m/z. The dotted line was shown as the FDR significant level, therefore, any m/z above this line were significantly different between two groups at FDR q=0.05. The total number of significant features was 1025 out of the 3270 detected.Figure 1

Metabolome-wide association study (MWAS). This is the snapshot of metabolites between supernatants from Plasmodium, non-infected and infected cultures at all time points using Manhattan plot on 3270 features. The features from duplicate run were averaged, log2 transformed, and quantile normalized to find out the significant features using false discovery rate (FDR). The dotted line represents FDR q=0.05. The metabolites over this line were the significant metabolites (n=1025) between two groups.

HRM platform provides precise metabolic phenotypes to determine the possible pathways altered by P. falciparum. The schematic representation in Additional files 1 and 2 shows that FDR significant 1025 features covering malaria parasite waste chemicals and human metabolites from supernatants were mapped through KEGG both human and Plasmodium metabolic pathways. Results show that these metabolites matched 439 metabolic compounds in both human and Plasmodium metabolic pathways. The remaining 586 of 1025 chemicals which were not matched, however, might be either waste products of the parasite or could be utilized by unidentified Plasmodium pathways.

Two way HCA was performed on combined sample classification with metabolites clustering to identify which metabolites are the most important for sample grouping. In this study, HCA was performed using 1025 metabolites at FDR q=0.05, which were the key components to separate two groups using all four time points (Figure 2). HCA determines similarity measures using Euclidean distance and Pearson linear correlation. The top panel showed that two main clusters separating supernatant of non-infected from infected cultures. The sample name was listed at the bottom panel. The right panel included 1025 metabolites which contributed to discrimination of samples according to malaria infection. In addition, Figure 3 shows a broad increasing trend in significant features at FDR q=0.05 during 48 hours.Figure 2

Two way hierarchical cluster analysis (HCA) on FDR significant features of supernatants between non-infected and infected cultures. HCA was performed using the 1025 metabolites at FDR q=0.05. The analysis utilized the samples from all time points to separate two groups in top bar (red clusters for control samples, green clusters for infected samples).

In order to validate methodology and analytical approach used for the identification of candidate biomarker molecules, the intensity of arginine and isoleucine known to be consumed and critical for malaria parasite growth were analysed. Amino acids are the building blocks of proteins. As parasites grow within host red blood cells, they utilize large quantities of amino acids. The malaria parasite derives most of its amino acid requirements from the proteolysed haemoglobin in the host blood cells [25]. In Figure 4A, arginine decreased to zero in a time dependent fashion during 48 hours. Most importantly, hemoglobin lacks one important amino acid, isoleucine, and the parasite, therefore, has to source this from culture system. In Figure 4B, isoleucine was reduced significantly in supernatants after the 48 hour incubation period.Figure 4

Decrease of arginine and isoleucine concentration during 48 hours’ culture. Figure 4 A: Arginine, and Figure 4 B: Isoleucine. White bars indicated supernatants from Plasmodium non-infected culture and black bar represented as supernatants from Plasmodium infected culture.

Among the remaining 586 of 1025 significant metabolites (FDR q=0.05) determined by FDR, the ion intensities of several metabolites were increased with culture time in infected culture supernatants but not in non-infected culture supernatants, suggesting a positive association between the quantity of these molecules released and level of parasitaemia. These metabolites were: i) 3-methylindole (Figure 5A), ii) succinylacetone (Figure 5B), iii) S-methyl-L-thiocitrulline (Figure 5C), and iv) O-arachidonoyl glycidol (Figure 5D). The compound 3-methylindole has been shown to stimulate an odorant receptor to attract malaria mosquito vector [26] while succinylacetone has been identified as an inhibitor of haem biosynthesis [27,28]. S-methyl-L-thiocitrulline has been identified as a potent nitric oxide synthase (NOS) inhibitor to reduce nitric oxide production and endothelial dysfunction [29]. On the other hand, O-arachidonoyl glycidol was reported to be an inhibitor of fatty acid amide hydrolase [30,31]. In addition, a similar pattern was observed for phosphocholine (Additional file 3), the molecule that was reported in previous study using infected erythrocytes [11] and is involved in the phosphobase methylation for phosphatidylcholine production [32]. The observed increase in phosphocholine with culture time further validated our methodology and analytical approach for identifying the four molecules reported above.Figure 5

Increase in concentrations of four molecules during 48 hours’ culture. Figure 5 A: 3-methylindole, Figure 5 B: succinylacetone, Figure 5 C: S-methyl-L-thiocitrulline, and Figure 5 D: O-arachidonoyl glycidol. White bars indicated supernatants from Plasmodium non-infected culture and black bar represented as supernatants from Plasmodium infected culture.

The production of 3-methylindole and succinylacetone were measured at each time point during 48 hours in the 50 μl of supernatants from 3% haematocrit cultures. At 36 hours, the amount of 3-methylindole was highest and the concentration was 178±18.7 pmoles (Figure 6A). The generation of succinylacetone increased over the time. The amounts were 2±2 pmoles at 12 hours culture, with the highest reading at 157±30.5 pmoles at 48 hours culture (Figure 6B). Supporting information can be seen with the MS/MS data of these compounds found in Additional files 4, 5, 6 and 7.Figure 6

The quantification of 3-methylindole and succinylacetone at each time point during 48 hours’ culture. Figure 6 A: Boxplot of 3-methylindole, and Figure 6 B: Boxplot of succinylacetone.