Metabolic Profiling Framework for Discovery of Candidate Diagnostic Markers of Malaria

Despite immense efforts to combat malaria in tropical and sub-tropical regions, the potency of this vector-borne disease and its status as a major driver of morbidity and mortality remain undisputed. We develop an analytical pipeline for characterizing Plasmodium infection in a mouse model and identify candidate urinary biomarkers that may present alternatives to immune-based diagnostic tools. We employ 1H nuclear magnetic resonance (NMR) profiling followed by multivariate modeling to discover diagnostic spectral regions. Identification of chemical structures is then made on the basis of statistical spectroscopy, multinuclear NMR, and entrapment of candidates by iterative liquid chromatography (LC) and mass spectrometry (MS). We identify two urinary metabolites (i) 4-amino-1-[3-hydroxy-5-(hydroxymethyl)-2,3-dihydrofuran-2-yl]pyrimidin-2(1H)-one, (ii) 2-amino-4-({[5-(4-amino-2-oxopyrimidin-1(2H)-yl)-4-hydroxy-4,5-dihydrofuran-2-yl]methyl}sulfanyl)butanoic acid that were detected only in Plasmodium berghei-infected mice. These metabolites have not been described in the mammalian or parasite metabolism to date. This analytical pipeline could be employed in prospecting for infection biomarkers in human populations.


Material and Methods
Parasite-Rodent Model. The current work was approved by the Swiss local and national regulations of laboratory animal welfare (permission no. 2081). Forty 3-week-old female NMRI mice (Charles River, Sulzfeld, Germany) were kept in groups of 8 in macrolon cages under standard environmentally-controlled conditions (temperature: 25°C, humidity: 70%, light/dark cycle 12/12 h). Mice had access to water and rodent food ad libitum (Rodent Blox from Eberle NAFAG, Gossau, Switzerland) and were acclimatized in the animal facility of the Swiss Tropical and Public Health Institute (Swiss TPH) for one week before the first sampling was conducted. Each group of eight mice was allocated to a different infection schedule ( Fig.   1). On day 0, two groups received 80 infective H. bakeri third stage larvae (L 3 ) which were administered orally in 150 µl water. Three groups of mice received 2 x 10 7 erythrocytes, parasitized with the green fluorescent protein (GFP)-transfected P. berghei ANKA strain in 0.2 ml red blood cell solution in RPMI medium intravenously 1 , on day 15. Urine and blood were collected from all mice one day before and after each infection timepoint (1 day preinfection and days 1, 14, and 16), during the maturing helminth single infection (day 8) and four days post P. berghei-infection (day 19). On each sampling day, mice were monitored for weight and PCV. All mice were euthanized on day 19 and parasitemia was evaluated by FACScan (Becton Dickinson; Basel, Switzerland) for P. berghei 1 and by counting of the worms via binocular (16x) after manual removal of the worms from the intestine. For PCV, weight, worm counts and P. berghei parasitemia, as well as inter-group median variation was analyzed using the Mann-Whitney U test with Bonferroni correction in StatsDirect (version 2.4.5; StatsDirect Ltd; Cheshire, UK), with a significance level of 5%.
Biofluid Collection, Weight and Packed cell volume. Urine (at least 40 µl) and tail blood (at least 50 µl) were collected one day before infection and 1, 8, 14, 16 and 19 days after the first infection timepoint on day 0, always between 8 and 11 AM (Fig. 1). Mice were individually placed into empty cages and monitored until they released a minimum of 40 µl urine which was immediately collected into 1.5 ml Eppendorf tubes and frozen on dry ice. All 3 samples were stored at -80°C prior to 1 H NMR acquisition. Approximately 50 µl tail blood was sampled from each mouse into a Na-heparinized hematocrit tube (1.55 mm Ø, BRAND GMBH + CO KG; Wertheim, Germany) and centrifuged at 11,000 x rpm for 5 min (microcentrifuge Sigma 1-15). The prepared samples were transferred into NMR microtubes (Bruker, diameter: 1.7 mm) shortly before measurement, and stored at 4°C prior to spectral acquisition. A standard 1 H NMR spectrum was acquired from each individual sample on a Bruker DRX 600 MHz spectrometer (Bruker Biospin, Rheinstetten, Germany), in a standard 1D experiment, using the standard solvent suppression pulse delay [recycle delay (RD)-90°-t l -90°-t m -90°-acquire free induction decay (FID)] 2 . The relaxation delay (RD) was typically 2 s long and t I at 3 µs, while the mixing time (t m ) was set to 100ms. Water irradiation was performed during the relaxation delay and also during the mixing time. Acquisition time for each sample was 2.73 s and spectral width was set to 20.022 p.p.m. A line broadening factor of 0.3 Hz was applied to the free induction decay and the FIDs were Fourier-transformed into a spectral resolution of 65.5 K data points. A second set of data was acquired, using a 1D Carr-Purcell-Meiboom-Gill (CPMG) pulse [RD-90°-(τ-180°-τ) n ] sequence 3 . The samples were scanned 256 times in each experiment, at a constant temperature of 300 K.

Sample Preparation and
Data Reduction and Multivariate Analysis. Plasma spectra were manually phased and baseline-corrected in Topspin (version 3.1, Bruker) and referenced to lactate at δ 1.33. The aliphatic region of the spectra (δ 0.5-4.6) was imported into MATLAB (version 7.12.0, R2011a) for processing and multivariate modelling in order to minimize the impact of the water-related baseline distortion. Spectral regions containing signals from ethanol and methanol were additionally removed (δ 1.15-1.22, 3.64-3.69 and 3.355-3.375). Further spectral pre-processing included probabilistic quotient normalization and peak alignment, using in-house developed scripts 4 . The formate integral, which was the only peak found between the water peak region and δ 9.00, was tested for discriminatory power between groups, with the Mann-Whitney U test (StatsDirect).
An enzymatic digestion with β-galactosidase (Jack Beans, Sigma) was conducted on a selected sample with high relative signals between 2.0-2.1 ppm in order to confirm the nature of the glycoprotein. A baseline sample was therefore acquired and compared to the sample immediately after addition of 0.5 U of the enzyme 6 and during the next 20 h window.
The assay was run on a 400 MHz spectrometer at an ambient temperature of 300 K and at a pH of ~ 7.4. The enzyme addition resulted in cleaving off of the acetyl group resulting in a marked increase of acetate and a subsequent decrease of the 2.04 signal proving signal contribution from N-acetyl glycoprotein, whereby α 1 -acid glycoprotein is likely to be the main contributor.

Results and Discussion
Plasma Metabolic Biomarkers. Plasmodium infection elicited a stronger signature in the plasma than in the urine profile in terms of the number of infection-related metabolic changes. Unlike the urine profile where endogenous changes manifested mainly at day 19, the plasma metabolite signature P. berghei showed a significant but inconsistent response over the course of malaria infection (Tables S2, S3). A comparison of plasma spectra from different infection groups obtained at day 19 is provided in Fig. S3, which shows samples from: (a) an uninfected control mouse (Ctr); (b) an animal with a P. berghei single infection (P); and (c) a mouse with a delayed co-infection (DC).
At day 16 postinfection the co-infected groups (SC and DC) respond more to P.
berghei infection than the single Plasmodium infection but by day 19 the differences between the single and co-infected P. berghei groups are largely resolved. However, individual metabolites demonstrate different time-dependent responses to the infection.
A relative decrease in plasma lysine concentration is found in group P compared to groups H and Ctr, whereas a relative increase in levels of the same amino acid occurs in groups H and SC, when compared to group Ctr on day 16 postinfection. Lysine can be catabolized by the gut microbiota to give rise to pipecolic acid 11 ; an increase in the urinary levels of pipecolic acid was observed subsequently on day 19. Moreover, a decrease in lysine levels can also follow acute stress 12 , which may be associated with the response to metabolic stress resulting from the P. berghei infection introduced on day 15. In addition, several amino acids such as leucine, valine, and alanine were found to be augmented in P.
berghei single and co-infections compared to uninfected controls on day 19.
The most notable differences are however detected in glucose, alanine, glycerophosphocholine (GPC), and 3-hydroxybutyrate expression, whereby the two former metabolites are present in relatively higher concentrations in groups DC and P, and the latter two were present in relatively lower levels in the P. berghei-infected mice.
On day 16, glucose levels decrease in group P compared to all other groups.
Interestingly, this difference is not significant by day 19. In the SC model, glucose levels are lower on day 19, compared to groups Ctr, H, and DC. Depletion of plasma glucose is one of the main findings in the P. berghei single infection group (P) on day 16, and is also observed in the simultaneous infection on day 19 when compared to control mice. Decreased plasma glucose concentrations have previously been reported by Li and colleagues 5 and is consistent with the fact that Plasmodium-infected erythrocytes consume higher amounts of glucose than normal cells, as the parasites rely on anerobic glycolysis to obtain energy.
Further manifestation of this phenomenon is the substantial increase in plasma lactate levels on day 19 in plasma in all P. berghei-infected groups. A relative increase of glucose is observed in group H compared to group Ctr on days 8 and 14 of the experiment and in group SC compared to group Ctr on day 14 (Table S3). Our results suggest that an established H.
bakeri infection compensates at least in part the P. berghei-induced plasma glucose depletion, as shown by the lack of difference between groups DC and Ctr as well as between SC and Ctr on day 16. The literature on hookworm-related changes in blood glucose levels is inconsistent.
The current study and a previous experiment in Necator americanus-infected hamsters conducted by Kaul and colleagues have shown relatively higher blood sugar levels in infected animals compared to uninfected controls 13 . Furthermore, malabsorption of sugars has been demonstrated in hookworm-infected patients 14 . Similarly lower blood glucose levels were observed in Necator americanus-infected hamsters in a previous metabolic profiling study 15 . However, it may not be appropriate to assume a linear relationship between helminth infection, gross physiological impact, and blood glucose levels, since co-infection state, genetic background of the mice, and the resulting immune-status (e.g., degree of T cell proliferation, macrophage activation, etc.) may exert additional influence on glucose consumption and physiological distribution.
Plasma 3-hydroxybutyrate depletion was observed in groups P and DC and may be indicative of ketosis. 3-Hydroxybutyrate can be used by the brain as source of energy, when blood glucose is low 16 . Metabolic acidosis has been associated with P. falciparum-infection in children where accumulation of plasma hydroxybutyrate and lactate were reported 17 . We are currently unable to explain this observation, since no weight change was observed between any of the infection-groups compared to the uninfected control baseline weight but, perhaps, as the infection progresses, the levels of 3-hydroxybutyrate are depleted by the excessive energy consumption inflicted by a Plasmodium infection.
On day 16, a decrease in GPC was observed in all infection groups, including single hookworm infection, compared to uninfected controls; on day 19 the same trend was observed but was less pronounced. Changes in the global lipid profile were noted in P.
berghei infected groups on day 19 (Table S3).         Table S4. The mean weights and packed cell volumes (PCV) are given for each group (n=8).

Tables
The numbers in brackets represent the standard deviations.