Short-term metabolic adjustments in Plasmodium falciparum counter hypoxanthine deprivation at the expense of long-term viability

Plasmodium falciparum Nf54^attB parasites (a gift from David Fidock, Columbia University) were cultivated in RBCs at 2% haematocrit (HCT) under a mixed gas atmosphere (94% N₂, 3% O₂, and 3% CO₂) at 37 °C. The parasite cultures were maintained in Roswell Park Memorial Institute (RPMI)-1640 medium (Gibco, Gaithersburg, MD, USA) supplemented with 25 mM HEPES, 90 µM hypoxanthine, 0.3% sodium bicarbonate, 25 µg/mL gentamicin, 0.5 µM R-lipoic acid, and 0.5% AlbuMAX II (Life Technologies, Carlsbad, CA, USA). This medium is referred to as the hypoxanthine-rich medium.

Parasite samples for transcriptomic and metabolomic analyses were prepared according to the methods described by Tewari et al. [10]. Briefly, for each IDC, synchronous parasite cultures were maintained by passing them through cell-sorting MACS columns for a period of 2 weeks, while scaling the culture volume up to 300 mL. After the final passage through a cell-sorting MACS column, the cultures were divided into four 75-mL volumes at 2% HCT using Percoll-purified RBCs. The cultures were pelleted and washed once in hypoxanthine-free medium within 0–2 h after they were infected by the parasite. The pellets were washed and re-suspended in 300 mL of hypoxanthine-deprived medium. They were divided again into four 75-mL cultures, and this time point was denoted as 0 h. Quadruplicate uninfected RBC cultures were also prepared in four 50-mL volumes to provide control samples for metabolomic analyses. Samples were collected at seven time points during the IDC (at 0, 8, 16, 24, 36, 40, and 48 h). The quadruplicate samples were sent to the Johns Hopkins Genomic Analysis and Sequencing Core Facility for transcriptomic analysis using Agilent microarray chip AMADID 037237 (Agilent Technologies, Inc., Santa Clara, CA, USA) and to Metabolon, Inc. (Durham, NC, USA) for metabolomic analysis.

To determine how long hypoxanthine-deprived parasites remain viable, parasites were cultured using the methods described above at hypoxanthine concentrations of 90 µM (hypoxanthine-rich) and 0.5 (hypoxanthine-deprived) for over a period of 120 h, and blood smears were prepared at 24, 40, and 48 h. A tenth of the hypoxanthine-deprived culture was transferred to hypoxanthine-rich medium at 40, 48, 60, 72, and 96 h, while the hypoxanthine-rich culture was transferred at 40 h to serve as a control. Twenty-4 h after transfer, the cultures were assessed for parasitaemia using a blood smear test. All growth experiments were carried out in triplicate wells for each condition and the procedure was repeated at least twice.

The present study used the latest version of a P. falciparum metabolic network model, which consists of 1025 metabolic reactions and 923 metabolites [10, 11]. This model has been annotated to include the PubChem IDs [12] of 273 metabolites found in metabolomic data collected from the P. falciparum Nf54^attB strain maintained in RPMI-1640 medium during the IDC, as determined by (1) ultrahigh-performance liquid chromatography (UPLC) tandem mass spectrometry (MS/MS) [10]; and, (2) metabolomic data obtained under isoleucine deprivation [9].

The gene-to-reaction mapping of the metabolic network model was used to compute r, the level of expression of an enzymatic reaction, which is a function of the gene(s) transcribing the enzyme(s) catalyzing a given reaction (Fig. 2b). Specifically, for a metabolic enzyme encoded by a single gene, r was assumed to be identical to the transcription of that gene. To compute r for such a metabolic enzyme at a given time point in the IDC, the expression level of the encoding gene at time t–τ was used (where ‘τ’ denotes the transcriptional delay, as described in Foth et al. [13]). For a metabolic enzyme encoded by more than one gene, Boolean rules (AND/OR gate) were used to compute r. These rules were implemented by taking the maximum expression value of the genes encoding an enzyme if they were connected by an ‘AND’ gate, or the minimum expression value if they were connected by an ‘OR’ gate [14].

Figure 1 shows that an RBC can synthesize hypoxanthine from inosine via purine nucleoside phosphorylase (PNP). However, in practice, the free energy of the PNP reaction favours the formation of inosine from hypoxanthine. Therefore, any alteration in RBC hypoxanthine levels could alter metabolic reactions within the RBC as well. The present study used a recently developed method [10] that predicts metabolic reaction fluxes from metabolomic data. This method was used to identify alterations in parasite-infected RBC metabolism in response to hypoxanthine deprivation. The optimization steps used to predict the metabolism of parasite-infected RBC under hypoxanthine-rich and -deprived conditions are provided in Additional file 1: Text S1.

A previously described computational framework was used for simulating hypoxanthine deprivation [15]. Briefly, alterations in transcriptomic data obtained under control and stressed conditions were used to simulate and describe the stress phenotype, and the approach was modified to include the effects of metabolite pool alterations, which may occur during the IDC under different conditions. Specifically, time-dependent metabolomic data obtained under hypoxanthine-rich and -deprived conditions were used as an additional constraints to the previous approach [15].

To obtain the metabolic flux distribution of P. falciparum during the IDC, the following equations were solved: where R denotes the set of all metabolic reactions of P. falciparum, M represents the set of metabolic reactions influenced by the metabolomic data, S represents a matrix containing the stoichiometry coefficients of all reactions, \(\overline{v}\) denotes a column vector containing all metabolic reactions of the network, \(\delta\) denotes the minimum of \(\mathop \sum _{i \in M} \left| {v_{i} - \widetilde{{m_{i} }} \cdot v_{i,nom} } \right|\) obtained separately, v_g represents the metabolic reaction governing the growth of the parasite, \(\mu\) denotes the nominal value of the parasite growth rate, which is set to 0.48 g/h gDW of the original merozoite [2], and \(\widetilde{{m_{i} }}\) denotes the minimum value of \(m_{i}\), where \(m_{i}\) is a vector containing the median values of each metabolite taking part in the ith metabolic reaction. \(v_{i,nom}\) denotes the ith value of \(v_{nom}\), which is obtained by solving the following: here, N denotes the set of reactions transporting nutrients across the parasite plasma membrane and \(v_{N}\) denotes a vector containing the optimal values of every reaction in N. The other variables are as defined above.

The optimization problems shown in Eqs. (1) and (2) yield v_ref, which were modulated using the time-dependent transcriptomic and metabolomic data to obtain the temporal profile of P. falciparum metabolism. The time-dependent transcriptomic and metabolomic data were incorporated by solving the following: here, G denotes the set of all intracellular reactions of P. falciparum, \({\text{r}}_{\text{j}}^{\text{t}}\) denotes the reaction expression of the jth reaction at time t, \(\overline{m}_{i}^{t}\) represents the influence of a metabolite at time t for the ith reaction, and the bar above m denotes normalization of metabolite abundance by its median over the IDC. \(v_{j,ref}\) and \(v_{i,ref}\) represent the jth and ith values, respectively, of \(v_{ref}\). As suggested by Fang et al. [2], the problem shown in Eq. (3) can have multiple solutions. Therefore, the following optimization problem was solved to obtain a solution closest to \(v_{ref}\):

The method to solve the optimization problems in Eqs. (2)–(4) was identical to the method previously described [15], except for the term incorporating the metabolomic data in Eq. (3).

GSEA was performed based on the method used in a previous study [16]. The absolute value of each gene’s log-transformed fold-change value was first calculated under control and stress conditions. The maximum absolute fold-change value of DNA probes that mapped to one gene were used. Next, the mean (μ) and standard deviation (σ) of the absolute values were calculated based on the absolute fold-change values of all genes. The average score of a given gene set (\(\overline{X}\)) was computed by taking the mean of the absolute fold-change values associated with a given gene set. The enrichment of a gene set was estimated by its p value, i.e., the probability of observing a score (\(\overline{X}\)) by chance. According to the Central Limit Theorem, the probability distribution of an average value is approximately normal with parameters μ and \(\sigma /\sqrt n\), where n is the number of genes in a given gene set. The p-value can be calculated from the z-transform of the average score of a given gene set, i.e., \(z = \frac{{\overline{X} - \mu }}{\sigma /\sqrt n }\). GSEA was performed using the gene ontology information for P. falciparum, which is available through the GO2MSIG platform [17].

Plasmodium falciparum parasites are normally cultured in the presence of 90–360 μM added hypoxanthine. Previous studies have shown that hypoxanthine concentrations as low as 2 μM can support P. falciparum IDC completion [18], but at least 5 μM is required for optimal growth. To determine how parasites respond to lower levels of hypoxanthine, synchronized parasites were cultured for 40 h at different hypoxanthine concentrations, after which they were transferred to a hypoxanthine-rich medium (90 μM hypoxanthine). The parasites cultured in 0.5 µM hypoxanthine during the first 40 h experienced approximately a 50% reduction in parasitaemia (assessed at 72 h) when compared with the parasites cultured in hypoxanthine-rich medium (Additional file 1: Fig. S1). This concentration was selected to further study the effect of hypoxanthine deprivation on the parasite development. The parasite numbers and morphology in hypoxanthine-rich and -deprived conditions were comparable for the first 24 h, after which vacuole-like structures became visible in the cytosol of hypoxanthine-deprived parasites. Figure 3a (lower panel) shows these vacuole-like structures (black arrows).

The parasites in the rich medium underwent schizogony between 40 and 48 h. By contrast, the deprived parasites were in the trophozoite stage at 40 h and about 18% of the total parasites were schizonts at 48 h. There was no re-infection of RBCs during the 48-h IDC. Figure 3b shows that hypoxanthine-deprived parasites only developed rings at 60 h. To determine how long these deprived parasites remain viable, they were transferred to a hypoxanthine-rich medium after various lengths of hypoxanthine deprivation (i.e., at 40, 48, 60, 72, and 96 h). Figure 3c shows that the parasites were able to establish a second IDC when they were transferred between 40 and 72 h, but not after 72 h.

Malaria parasites require hypoxanthine to synthesize purine-based nucleotides. Both the parasite and the RBC have PNP enzymes, which can produce hypoxanthine from inosine. Because the parasite relies on RBC metabolism for a number of nutrients during the IDC, the data were first analysed to assess the impact of hypoxanthine deprivation on parasite-infected RBC metabolism per se.

Parasite-infected RBC metabolism did not significantly change in response to hypoxanthine deprivation. In fact, the metabolic pathways of a parasite-infected RBC were perturbed to a similar degree regardless of the culture medium. Table 1 lists selected metabolic pathways in the RBC and the number of reactions within each pathway that were significantly altered (p < 0.01) in response to parasite infection under hypoxanthine-rich and -deprived conditions. A complete list of the RBC metabolic pathways perturbed under these two conditions is provided in Additional file 2.

In the culture conditions used here, malaria parasites had access to only hypoxanthine as their sole purine source, which was kept at a pre-determined low concentration of 0.5 µM. Figure 4 shows key purine metabolites that were used by the parasite to synthesize purine-based nucleotides, indicating that hypoxanthine levels under the deprivation condition only dropped after the 24-h time point. By contrast, Fig. 4a and d show that the levels of adenosine and inosine monophosphate (IMP) under the hypoxanthine-deprived condition were either higher than (or at least comparable to) those under the hypoxanthine-rich condition. Figure 4b depicts inosine levels, which resemble the pattern observed with hypoxanthine levels under deprivation. A comparison of Fig. 4c and d indicates that the parasite was able to maintain IMP levels during the second day of the 48-h IDC despite having relatively negligible quantities of hypoxanthine during this time period. A complete list of metabolites detected under hypoxanthine-rich and -deprived condition is provided in Additional file 3.

Hypoxanthine is a key precursor of purine bases for P. falciparum to synthesize DNA, RNA, and protein. The extent to which the proposed model could capture the time-dependent changes in DNA, RNA, and protein synthesis over the course of the IDC in the hypoxanthine-deprived condition was examined (Fig. 5a–c). Compared to the hypoxanthine-rich condition (black plots), the hypoxanthine-deprivation simulations showed a slight increase in DNA synthesis during the first 24 h, followed by a slight decrease between 30 and 42 h, while both RNA synthesis and protein synthesis slightly decreased between 12 and 32 h. By contrast, hypoxanthine deprivation markedly altered the synthesis or generation of cofactors, polyamines, and inorganic ions relative to the hypoxanthine-rich condition (Fig. 5d–f).

As noted above, vacuole-like structures within the parasite cytoplasm became apparent at 24 h into the IDC (Fig. 3a). To identify the metabolic processes of P. falciparum that were significantly altered before and after the appearance of these structures under the hypoxanthine-deprived condition, model simulations of hypoxanthine-rich parasites and hypoxanthine-deprived parasites were compared to identify significantly altered metabolic reactions. Table 2 lists the top reactions and pathways of P. falciparum metabolism that were altered before and after the appearance of these structures. The additonal material lists the genes of each metabolic pathway perturbed within 24 h (Additional file 4, Sheet 3) or after 24 of infection (Additional file 4, Sheet 4).

Thus far, the analysis has only focused on identifying alterations in the metabolism of the parasite. This section presents the results of GSEA, which includes analyses of all biological processes of the parasite. Figure 6a shows the enrichment of gene sets associated with the stress response pathways (SRPs) of the parasite. Similar gene sets were clustered together, as indicated by the grey lines showing the interactions among them. The black lines show gene sets that directly contributed to the enrichment of SRPs. To identify the metabolic changes critical for the appearance of vacuole-like structures observed during hypoxanthine deprivation, enrichment analysis was performed for transcriptomic data obtained at 24 h into the IDC. This revealed an enrichment of genes associated with P. falciparum RBC membrane protein 1 (PfEMP1), intracellular signalling, Maurer’s cleft, rosette formation, protein export, and mitosis.

To ascertain whether enrichment of these gene sets constitutes a general response or a response specific to hypoxanthine deprivation, GSEA was also performed on published data obtained under isoleucine deprivation at the 30-h mark, when parasites enter a hibernatory state [9]. Figure 6b shows the gene sets associated with SRPs under these conditions, with gene clusters colour-coded in a manner identical to the data obtained under hypoxanthine deprivation. As in the case of hypoxanthine deprivation, this analysis also revealed enrichment of genes associated with rosette formation, Maurer’s cleft and protein export. However, gene sets associated with PfEMP1, intracellular signalling and mitosis were not enriched as they were under hypoxanthine deprivation. A complete list of gene sets belonging to the biological process ontology of P. falciparum, along with their enrichment level at each time point and deprivation condition, is provided in Additional files 5 and 6.

Parasites deprived of hypoxanthine displayed morphological and developmental aberrations during the first IDC and could not be rescued after two IDCs in the deprived medium. The formation of vacuole-like structures was evident in many parasites at 24 h into the IDC. Such vacuole formation, which has previously been observed in parasites exposed to heat stress, is attributed to autophagic programmed cell death [18]. The hypoxanthine-deprived parasites studied here also exhibited a prolonged trophozoite stage, but those that did not succumb to death ultimately established another round of IDC. This type of delayed cell cycle progression is also observed in isoleucine-deprived parasites [9]. Interestingly, parasites deprived of isoleucine are viable up to 72 h, similar to those deprived of hypoxanthine, but cannot be rescued upon transfer to an isoleucine-rich medium after four IDCs [9]. These observations suggest that P. falciparum is capable of effectively responding to short-term nutrient stress.

Based on model simulations, it can be inferred that the hypoxanthine-deprived parasite invokes mechanisms to maintain a sufficient pool of purine bases, which thereby supports its growth under limited availability of hypoxanthine. Model-predicted reaction fluxes that support this hypothesis are presented below (Fig. 7). First, hypoxanthine phosphoribosyltransferase (HXPRT) activity is largely unchanged during hypoxanthine deprivation (Fig. 7a). HXPRT synthesizes IMP, which is a precursor for xanthine monophosphate (XMP) and adenylosuccinic acid (ASA). Second, the parasite maintains flux through adenylosuccinate synthase (ADSS) but not through IMP dehydrogenase (IMPD) (Fig. 7c). This makes sense because the precursor for IMP is hypoxanthine, which is present in substantially less quantity in the medium (90 µM in the rich medium vs 0.5 µM in the deprived medium). Together, these results suggest that, under limited concentrations of hypoxanthine, the parasite invokes pathways that synthesize inosine, which it then uses to synthesize hypoxanthine through the PNP enzyme (Fig. 7e).

Figure 7d depicts the cycle of production of inosine through a pathway that requires cleaving of adenosine (ADN) from S-adenosylhomocysteine (SAH). The cycle is futile because SAH requires S-adenosylmethionine (SAM), which cannot be synthesized without ATP. Therefore, although the parasite can synthesize hypoxanthine through this pathway, it must also redistribute (enhance or suppress) reactions associated with ATP to maintain the synthesis of purine-based nucleic acids. The deprived parasites suppressed metabolic fluxes through ATP-consuming reactions, such as those of phosphoenolpyruvate carboxykinase (PPCK), adenylate kinase and carbamoyl-phosphate synthase, while also increasing metabolic fluxes through certain ATP-consuming reactions, such as those of glutathione synthase and guanylate kinase (Additional file 1: Fig. S2).

Interestingly, the parasites can also produce hypoxanthine through methylthioadenosine (Fig. 1), which is produced during polyamine synthesis. However, there was no increase in metabolic flux through this pathway under the hypoxanthine-deprived condition. In fact, hypoxanthine-deprived parasites produced significantly less putrescine, which is captured in Fig. 5e. The strategy behind producing hypoxanthine via inosine instead of methyladenosine is not clear.

The futile cycle discussed above requires SAM, which is a co-substrate for enzymes (i.e., 3-demethylubiquinone-9, 3-O-methyltransferase and ubiquinone/menaquinone biosynthesis C-methyltransferase) and is essential for ubiquinone synthesis. Therefore, an increase in flux through SAM should also increase flux through reactions associated with ubiquinone synthesis and create a demand for the precursors necessary for ubiquinone synthesis, such as chorismate. One of the top pathways significantly altered under hypoxanthine deprivation in the simulations was that of shikimate biosynthesis (Table 2), of which chorismate is a by-product. Therefore, chorismate is also expected to increase during hypoxanthine deprivation (Fig. 8).

An increase in shikimate synthesis should also create a demand for its precursor, phosphoenolpyruvate (PEP), and, consequently, influence other reactions that require PEP. PEP is also an important precursor in the methylerythritol phosphate pathway (MEP) pathway of isoprenoid synthesis, and as with shikimate synthesis, the flux through the MEP pathway significantly increased under hypoxanthine deprived conditions (Table 2). Interestingly, the flux through PPCK (Additional file 1: Fig. S2), an ATP-consuming reaction that converts oxaloacetate (OAA) back to PEP, decreased, suggesting a buildup of OAA.

Figure 8 shows that all fluxes downstream of cytosolic OAA (red arrows in the mitochondrion; e.g., malate:quinone oxidoreductase, fumarate hydratase, etc.) increased under hypoxanthine-deprivation conditions. In fact, in simulations of hypoxanthine-deprived parasites, the tricarboxylic acid (TCA) cycle seemed to deviate from the cyclic mode; specifically, malate:quinone oxidoreductase and fumarate hydratase worked in the reverse direction under deprived conditions (Fig. 8, red arrows). This deviation suggests an increase in oxidized nicotinamide adenine dinucleotide (nad). The metabolic flux through the oxidative branch of the pentose phosphate pathway (PPP) also decreased significantly under hypoxanthine-deprivation conditions. This suggests that the increased haem biosynthesis, together with a reduction in PPP, causes an unfavourable increase in oxidative stress of the parasite under deprivation conditions. Notably, metabolic reactions associated with haem degradation and minimization of oxidized glutathione (glutathione reductase and glutathione efflux) also increased, suggesting an increase in oxidative stress under deprivation conditions (Additional file 1: Fig. S3).