A metabolite-based resistance mechanism against malaria

Abstract

Jaundice is a common presentation of Plasmodium falciparum malaria, which arises from the accumulation of circulating bilirubin. It is not understood whether it represents an adaptive or maladaptive response to Plasmodium spp. infection. We found that asymptomatic P. falciparum infection in humans was associated with a higher ratio of unconjugated over conjugated bilirubin and parasite burden compared with symptomatic malaria. Genetic suppression of bilirubin synthesis by biliverdin reductase A (BVRA) increased parasite virulence and malaria mortality in mice. Accumulation of unconjugated bilirubin in plasma, through genetic inhibition of hepatic conjugation by UDP glucuronosyltransferase family 1 member A1 (UGT1A1), was protective against malaria in mice. Unconjugated bilirubin inhibited P. falciparum proliferation in red blood cells by a mechanism that suppressed mitochondrial pyrimidine synthesis. Moreover, unconjugated bilirubin inhibited hemozoin crystallization and compromised the parasite’s food vacuole. Hence, jaundice appears to represent a metabolic response to Plasmodium spp. infection that limits malaria severity.

Graphical Abstract

Antimalarial effect of bilirubin. Malaria is associated with hemolysis and labile heme accumulation, which is catabolized into biliverdin by heme oxygenase-1 (HO-1) and converted into bilirubin by biliverdin reductase A (BVRA). Repression of bilirubin conjugation by UGT1A1 increases unconjugated bilirubin to target the Plasmodium mitochondrion and inhibit hemozoin crystallization, compromising the parasite’s food vacuole and amino acid (AA) acquisition. Δψ, mitochondrial membrane potential.

INTRODUCTION:

Jaundice arises when bilirubin, a yellow pigment, accumulates in plasma and gives a yellowish color to the skin and the sclera (the white portion of the eyeball). Bilirubin has long been considered as a “waste product” of heme catabolism. Because of its lipophilic nature, bilirubin excretion requires conjugation to glucuronic acid through a reaction catalyzed in hepatocytes by UDP glucuronosyltransferase family 1 member A1 (UGT1A1). The less toxic water-soluble conjugated bilirubin is then excreted via the bile into the intestinal lumen. Because bilirubin conjugation occurs in the liver, its accumulation in plasma is a reliable biomarker of liver dysfunction. Although accurate, this has led to the widespread perception of jaundice being a maladaptive and eventually pathogenic response. However, several investigators have shown that bilirubin participates in various activities, acting as a lipophilic antioxidant and presumably as a ligand of receptors involved in different metabolic functions.

RATIONALE:

Jaundice is a common presentation of malaria, the ancestral infectious disease caused by parasites from the Plasmodium genus. These parasites evolved to invade and proliferate inside the red blood cells of their hosts, causing hemolysis and the accumulation of extracellular hemoglobin in plasma. When the prosthetic heme groups of hemoglobin are detached from the globin chains of hemoglobin, there is an accumulation of labile heme in plasma, an independent risk factor for Plasmodium falciparum malaria severity. Survival from experimental malaria is contingent on the capacity of the infected host to catabolize heme into biliverdin, the substrate used by biliverdin reductase to produce bilirubin. This raised the hypothesis that bilirubin production by biliverdin reductase participates in a metabolism-based defense strategy against malaria.

RESULTS:

Using a highly specific approach to measure bilirubin in plasma, we found a correlation between the levels of circulating unconjugated bilirubin and the onset of symptomatic P. falciparum malaria in humans. We established that bilirubin is protective against malaria in an experimental model of malaria in mice, where repressing bilirubin production through genetic loss of function of biliverdin reductase precipitated malaria mortality. This lethal phenotype could be reversed by the administration of bilirubin, verifying that unconjugated bilirubin can be protective against experimental malaria. Repression of bilirubin conjugation by hepatic UGT1A1 was also protective against experimental malaria in mice, further supporting the protective effect of unconjugated bilirubin against malaria. Using several orthogonal approaches in vivo and in vitro, we found that unconjugated bilirubin targets Plasmodium inside the red blood cell to repress its proliferation and virulence. Bilirubin targets the parasite’s mitochondrion and simultaneously interferes with heme detoxification, disrupting the parasite food vacuole and therefore inhibiting the acquisition of essential amino acids from hemoglobin.

CONCLUSION:

The induction of bilirubin production and inhibition of its conjugation in response to Plasmodium spp. infection is an evolutionarily conserved resistance mechanism against malaria. Presumably, this metabolism-based defense strategy has a major evolutionary trade-off, namely, the insidious incidence of neonatal jaundice, which can potentially damage neurons in the brain. To what extent this defense strategy can be targeted therapeutically to overcome the enormous burden imposed by malaria on human populations remains to be established. □

Parasites from the Plasmodium genus proliferate in the red blood cell (RBC) compartment of their numerous vertebrate hosts, leading to intravascular hemolysis and release of hemoglobin into plasma (1, 2). Auto-oxidation of extracellular hemoglobin precipitates the detachment of its non–covalently bound prosthetic heme groups (1, 3–6). This produces labile heme (7), an independent risk factor for Plasmodium falciparum malaria severity that contributes to the pathogenesis of severe malaria in mice (3, 5, 8).

The pathogenic effects of labile heme are countered by heme oxygenase-1 (HO-1; encoded by gene HMOX1) (3, 4, 9). This stress-responsive enzyme catabolizes heme into biliverdin (10), which is reduced into lipophilic bilirubin by biliverdin reductase A (BVRA) (11) (fig. S1A). Unconjugated bilirubin circulates in plasma bound to albumin (12) (fig. S1A) and is conjugated to glucuronic acid by UDP glucuronosyltransferase family 1 member A1 (UGT1A1) in hepatocytes (13, 14) (fig. S1A). Water-soluble conjugated bilirubin is excreted via the bile or urine (15) (fig. S1A).

Severe presentations of P. falciparum malaria are often associated with the accumulation of bilirubin in plasma, a condition referred to as jaundice of malaria when associated with visible yellowing of the skin or white of the eyes (16–18). Jaundice in malaria develops once the rate of bilirubin production by BVRA exceeds that of bilirubin conjugation by UGT1A1 (19). In this work, we asked whether the accumulation of circulating bilirubin during Plasmodium spp. infection represents an adaptive or maladaptive metabolic response to malaria.

Results

Unconjugated bilirubin limits P. falciparum malaria symptoms

A retrospective analysis of a cohort of patients with P. falciparum infection and asymptomatic or symptomatic malaria showed a positive correlation between the presence of clinical symptoms and parasite burden (i.e., number of parasites per microliter of blood) (Fig. 1A), total circulating bilirubin (Fig. 1B), conjugated bilirubin (Fig. 1C), and unconjugated bilirubin (Fig. 1D). Although these observations are consistent with those of previous studies (16–18), the standard analytical methods used in clinical practice do not accurately quantify unconjugated bilirubin in plasma (12, 15, 20). The binding of unconjugated bilirubin to albumin in plasma (12) interferes with the colorimetric reaction used in these assays (12, 15, 20). Moreover, by-products of hemolysis, such as labile heme, can interfere with these colorimetric assays, adding to their inaccuracy in the context of Plasmodium infection (1, 6).

Fig. 1. Unconjugated bilirubin confers resistance to malaria.

(A) Parasite burden (P. falciparum–infected RBCs per μl of blood) and concentrations of (B) total (i.e., conjugated plus unconjugated), (C) conjugated, and (D) unconjugated bilirubin (measured using the Roche Diazo method) in plasma from P. falciparum–infected individuals, stratified according to disease severity as asymptomatic and symptomatic malaria. (E) Plasma concentration of unconjugated bilirubin measured by the UnaG-based assay in same patients as in (A) to (D). (F) Ratio of unconjugated bilirubin measured by the UnaG-based assay (E) to conjugated bilirubin (C). Data in (A) to (F) are shown as box plots; red lines correspond to median values, and error bars correspond to the interquartile range (IQR). (G) Concentration of unconjugated bilirubin in plasma from Blvra^+/+ and Blvra^−/− mice (n = 6 to 9 per genotype), before (day 0) and after Pcc infection (days 4, 7, 15, and 25) as measured by the UnaG-based assay. Data are shown as means ± SD, pooled from two independent experiments with a similar trend. (H) Survival of Pcc-infected Blvra^−/− and control Blvra^+/+ mice. Data are from n = 7 mice per genotype, pooled from two independent experiments with similar trend. (I) Survival of Pcc-infected Blvra^−/− mice receiving bilirubin [30 or 3 mg per kg of body weight (mg/kg) daily and intraperitoneally] or vehicle. Data are from n = 4 to 10 mice per treatment, pooled from two independent experiments with a similar trend. (J) Quantification of Ugt1a1 mRNA [quantitative polymerase chain reaction (qPCR); left] (n = 7 to 9 per genotype) and protein (Western blot; right) (n = 6 per genotype) expression in the liver at day 7 after Pcc infection (Pcc) or in noninfected (NI) C57BL/6J mice. Data are shown as means ± SD, pooled from two independent experiments with a similar trend. (K) Concentration of unconjugated bilirubin in plasma of adult DBA/2 mice transduced 2 to 4 days after birth with AAV8-gRNA-Ugt1a1 repressing hepatic Ugt1a1 or control AAV8-Cas9. Data are from n = 5 mice per group, shown as means ± SD from one experiment representative of three with a similar trend. (L) Survival of the mice from (K) infected with Pcc. Circles correspond to P. falciparum–infected patients in (A) to (F) and to individual mice in (G) to (L). The p values were determined using [(A) to (F)] Mann-Whitney U test; (G) two-way analysis of variance (ANOVA) with Bonferroni’s multiple comparison test (for genotypes) and ordinary one-way ANOVA with Tukey’s multiple comparison (for days after infection); [(H), (I), and (L)] log-rank (Mantel-Cox) test; [(J), right, and (K)] Student’s t test; and [(J), left] Mann-Whitney U test. NS is not significant; *p < 0.05; **p < 0.01; ***p < 0.001.

To obtain a precise measure of unconjugated bilirubin in plasma, we used a highly specific unconjugated bilirubin-inducible fluorescent protein (UnaG)–based assay (fig. S1B) (21, 22). We found that the concentration of unconjugated bilirubin in plasma from asymptomatic P. falciparum–infected individuals ranged from 8 to 50 μM (Fig. 1E), that is, 3.75-fold higher than that estimated using standard analytical methods (fig. S1C). The concentration of unconjugated bilirubin in plasma from symptomatic P. falciparum malaria patients ranged from 14 to 67 μM (Fig. 1E), 2.67-fold higher than that estimated using standard analytical methods (fig. S1C) (12, 15, 20).

Using the same UnaG-based assay, we found that asymptomatic P. falciparum infection was associated with a higher ratio of circulating unconjugated over conjugated (Fig. 1F) or total (fig. S1D) bilirubin as well as with a >10-fold higher ratio of circulating unconjugated bilirubin over parasite burden compared with symptomatic malaria (fig. S1E). Because unconjugated bilirubin inhibits P. falciparum proliferation in vitro (23), we hypothesized that its accumulation in plasma represents an adaptive response to P. falciparum malaria.

Bilirubin is protective in experimental malaria

To functionally test the antimalarial effect of unconjugated bilirubin, we infected C57BL/6J mice with Plasmodium chabaudi chabaudi AS (Pcc), a well-established nonlethal experimental model of malaria (24). Using the UnaG-based assay (fig. S1B) (21, 22), we found that the concentration of circulating unconjugated bilirubin increased abruptly from day 4 to day 7 after infection to reach 13 to 31 μM (Fig. 1G), in the range of that seen in P. falciparum malaria patients (Fig. 1E). The specificity of the UnaG-based assay was confirmed using Blvra-deficient (Blvra^−/−) mice (22), which do not express Blvra mRNA (fig. S2A) and BVRA protein (fig. S2, B and C) and do not produce bilirubin (Fig. 1G).

Compared with the nonlethal outcome of Pcc infection in C57BL/6J Blvra^+/+ mice, all littermate C57BL/6J Blvra^−/− mice succumbed to the infection within 7 to 10 days (Fig. 1H). This suggests that the production of bilirubin by BVRA confers protection against Plasmodium infection.

Using a biliverdin-inducible infrared fluorescent protein (iRFP)–based assay (fig. S2D) (22, 25) to quantify the concentration of biliverdin in plasma, we found that the Pcc-infected Blvra^−/− mice (Fig. 1H) accumulated relatively low levels of circulating biliverdin, in the range of 2.5 to 6 μM (fig. S2E). This suggests that the lethal outcome of Pcc infection in Blvra^−/− mice is not due to a putative accumulation of pathogenic levels of circulating biliverdin.

Administration of unconjugated bilirubin protected Blvra^−/− mice from succumbing to Pcc infection (Fig. 1I). The protective effect of bilirubin was dose dependent, that is, higher bilirubin dosage restored the survival of Pcc-infected Blvra^−/− mice to the same extent as Pcc-infected Blvra^+/+ mice (Fig. 1I). This suggests that the lethal outcome of Pcc infection in Blvra^−/− mice is attributable to a lack of bilirubin production.

We asked whether the production of bilirubin by BVRA is also protective against bacterial or viral infectious diseases. Disease severity and pathogen burden were indistinguishable in Blvra^−/− and control Blvra^+/+ mice subjected to polybacterial sepsis induced by cecal ligation and puncture (fig. S3, A to C) or to influenza A virus infection (fig. S3, D to F). This suggests that the protective effect of unconjugated bilirubin is specific to hemolytic conditions, such as malaria.

UGT1A1 is deleterious in experimental malaria

Expression of Hmox-1 mRNA was highly induced in different organs of Pcc-infected C57BL/6J mice compared with organs of noninfected C57BL/6J mice (fig. S4A), consistent with previous findings (9). However, expression of Blvra mRNA did not differ between Pcc-infected C57BL/6J mice and noninfected C57BL/6J mice (fig. S4B). This suggests that the induction of bilirubin production in response to Plasmodium infection is dependent on the induction of HO-1 to increase the amount of biliverdin that can be reduced into bilirubin by BVRA (fig. S1A).

Pcc infection was associated with a marked decrease in the relative level of hepatic Ugt1a1 mRNA (Fig. 1J and fig. S4C) and UGT1A1 protein (Fig. 1J and fig. S4D) expression 7 days after infection compared with noninfected controls. This suggests that the accumulation of circulating bilirubin in response to Plasmodium infection is sustained by the repression of hepatic UGT1A1 (fig. S1A).

To test whether inhibition of hepatic bilirubin conjugation is protective against malaria, we used Pcc infection in DBA/2 mice as a lethal experimental model of malaria (26). Hepatic UGT1A1 was repressed specifically in the liver of newborn DBA/2 mice, transduced with a recombinant adeno-associated virus serotype 8 (AAV8) encoding the Staphylococcus aureus (Sa) CRISPR associated protein 9 (Cas9) and a single guide RNA (gRNA) targeting Ugt1a1 (AAV8-gRNA-Ugt1a1), as previously described (27). Adult DBA/2 mice transduced with AAV8-gRNA-Ugt1a1 presented lower levels of hepatic UGT1A1 protein compared with controls transduced with a AAV8 encoding SaCas9 without the targeting gRNA (AAV8-Cas9) (fig. S4, E and F). This was associated with an increase in the concentration of unconjugated bilirubin in plasma (Fig. 1K) and with a major survival advantage against Pcc infection in these mice compared with control Pcc-infected DBA/2 mice transduced with AAV8-Cas9 (Fig. 1L). These observations suggested that the repression of hepatic UGT1A1 in response to Plasmodium infection is protective against malaria.

Bilirubin reduces Plasmodium burden

P. falciparum proliferation is inhibited by unconjugated bilirubin in vitro (23), suggesting that the accumulation of unconjugated bilirubin in plasma confers resistance to malaria (i.e., reduces the host parasite burden). In strong support of this hypothesis, Pcc-infected Blvra^−/− mice presented a clear increase in the percentage of circulating infected RBCs (parasitemia) 7 to 9 days after infection (Fig. 2A), accounting for a 10-fold increase in the number of infected RBCs (i.e., parasite burden), compared with Pcc-infected Blvra^+/+ mice (Fig. 2B). Consistent with this observation, the administration of unconjugated bilirubin to Pcc-infected Blvra^−/− mice reduced parasitemia (fig. S5A) and parasite burden compared with vehicle-treated Pcc-infected Blvra^−/− mice (fig. S5B). Moreover, repression of bilirubin conjugation also reduced the parasitemia (fig. S5C) and parasite burden (fig. S5D) of Pcc-infected DBA/2 mice transduced with AAV8-gRNA-Ugt1a1 compared with controls transduced with AAV8-Cas9. These observations suggest that the accumulation of unconjugated bilirubin in plasma confers resistance to Plasmodium infection.

Fig. 2. Unconjugated bilirubin regulates Plasmodium blood stage development and virulence.

(A) Percentage of infected RBCs and (B) parasite burden (infected RBCs/μl) in Pcc-infected Blvra^−/− and control Blvra^+/+ mice. Data are from n = 7 mice per genotype, pooled from two independent experiments with a similar trend. The same mice as in Fig. 1H are represented here. In (B), the right panel highlights day 8 from the left panel. (C) UMAP projection of single parasite transcriptomes, of circulating infected RBCs isolated by FACS from Blvra^+/+ (n = 1862; left) and Blvra^−/− (n = 3891; right) mice 7 days after Pcc infection. Colors and numbers identify different parasite developmental stages based on scmap projection to a P. falciparum atlas (see data tables S1 to S4). (D) Composition and (E) absolute number of parasite developmental stages. (F) UMAP projection as in (C), with arrows representing the relative change in transcriptional state based on RNA velocity analysis. (G) Survival of Blvra^−/− mice infected with Pcc-infected RBCs isolated from Blvra^−/− or Blvra^+/+ mice (n = 10 or 11 per genotype). (H) Percentage of infected RBCs (left) and parasite burden (infected RBCs/μl, right) for the same mice as in (G). Data are shown as means ± SD, pooled from three independent experiments with a similar trend. Circles in [(B), right] represent individual mice. The p values were determined using [(A), (B), left, and (H)] two-way ANOVA; [(B), right] Mann-Whitney U test; and (G) log-rank (Mantel-Cox) test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Bilirubin targets Plasmodium inside RBCs

Because unconjugated bilirubin diffuses across cellular membranes (28, 29), we tested whether unconjugated bilirubin targets Plasmodium inside the RBC. We performed single-cell RNA sequencing (scRNA-seq) of circulating Pcc-infected RBCs separated by fluorescence-activated cell sorting (FACS) from Blvra^−/− and Blvra^+/+ mice. Single parasite transcriptomes were assigned to specific developmental stages based on a single-cell malaria atlas (30). Visualization, using uniform manifold approximation and projection (UMAP), revealed a circular orientation of parasites throughout the asexual cycle: rings (clusters 1 and 2), trophozoites (clusters 3 and 4), schizonts (clusters 5 and 6), and developing gametocytes (cluster 7) (fig. S6, A and B, and data table S1), consistent with previous descriptions (30–32).

Circulating infected RBCs from Blvra^−/− mice exhibited a marked increase in the proportion (Fig. 2, C and D) and relative number (Fig. 2E) of parasite transcriptomes corresponding to metabolically active early trophozoites (cluster 3). This was associated with a loss of early schizonts (cluster 5) and late schizonts (cluster 6) as well as early rings (cluster 1) and with a decrease in late rings (cluster 2) and late trophozoites (cluster 4) (Fig. 2, C and D), as confirmed by morphological analyzes (fig. S7, A to C). These changes were not attributed to the modulation of parasite sequestration in different organs, as determined by shifting the “light cycle” of infection, an experimental approach that allows for the shifting of parasite sequestration at the time of analysis (fig. S7D).

Progression of Pcc infection in Blvra^+/+ mice was associated with the expected parasite developmental trajectory whereby late trophozoites (cluster 4) developed into early schizonts (cluster 5) and late schizonts (cluster 6) to give rise to early ring stages (cluster 1), as assessed by RNA velocity analyses (33) (Fig. 2F). Progression through late rings (cluster 2) and early trophozoites (cluster 3) was diffuse in Blvra^+/+ mice and contrasted sharply with the progression of the same parasite developmental stages in Blvra^−/− mice (Fig. 2F). This suggests that bilirubin reduces the fitness of early trophozoites from Blvra^+/+ compared with those from Blvra^−/− mice.

Late ring stages (cluster 2) from Blvra^−/− mice showed a marked increase in gene expression profiles associated with peptidase activity, heme binding, electron transport chain, proteosome activity, and catabolic processes compared with ring stages from control Blvra^+/+ mice (figs. S8 and S9A and data tables S2 and S3). Early trophozoites (cluster 3) in Blvra^−/− mice also showed a clear increase in gene expression associated with peptidase activity, adenosine triphosphatase (ATP)–dependent protein folding chaperone and unfolded protein binding, proteosome core complex, ubiquitin, and catabolic processes compared with early trophozoites from control Blvra^+/+ mice (figs. S8 and S9B and data tables S2 and S3). In addition, there was a concomitant decrease in gene expression profiles associated with chromatin structure (figs. S8 and S9B, and data tables S2 and S3). Late trophozoites (cluster 4) in Blvra^−/− mice showed an increase in gene expression profiles associated with ATP-dependent protein folding, RNA binding and translational activity, glycolytic, pyruvate and carbohydrate, adenosine diphosphate (ADP), and ATP metabolic processes (figs. S8 and S9C and data tables S2 and S3). Moreover, there was a concomitant decrease in gene expression profiles associated with structural constituents of chromatin, DNA replication, and chromosome organization and segregation (figs. S8 and S9C and data tables S2 and S3).

Bilirubin reduces parasite virulence

Among the 1019 differentially expressed genes in Pcc late rings and early and late trophozoites from Blvra^−/− compared with those from Blvra^+/+ mice (fig. S8), 118 (11.6%) (data tables S2 and S4) were previously linked to Pcc virulence (34). These included 53 up-regulated and 65 down-regulated genes in Blvra^−/− mice compared with Blvra^+/+ mice (data tables S2 and S4). The direction of gene regulation showed a significant association with previous studies (Fisher’s exact test, p < 0.0001; data tables S2 and S4) for 92% of the up-regulated genes (49 out of 53) and 49% of down-regulated genes (32 out of 65) (34). Among the up-regulated genes were four Plasmodium interspersed repeat (pir) genes (PCHAS_030190, PCHAS_041960, PCHAS_104230, and PCHAS_110030), which encode variant surface CIR proteins linked to parasite “immune evasion” (35, 36) and genes involved in parasite metabolic adaptation (34). Other virulence genes included 17 Pc-fam (or rodent malaria parasites; RMP-fam) genes (37) and eight genes encoding exported proteins (34), including the cir gene PCHAS_110030, a P. falciparum RIF ortholog linked to Pcc virulence (36, 38). These observations suggest that Pcc parasite virulence (i.e., the ability to cause disease or damage to the host) increases in Blvra^−/− mice compared with Blvra^+/+ mice.

We then asked whether the genetic fingerprint of higher virulence in parasites from Blvra^−/− mice versus Blvra^+/+ mice contributes to increased malaria mortality. In strong support of this hypothesis, the incidences of mortality (Fig. 2G), parasitemia, and parasite burden (Fig. 2H) were increased in Blvra^−/− mice infected with parasites isolated from Blvra^−/− mice compared with Blvra^−/− mice infected with those from Blvra^+/+ mice (Fig. 2, G and H). This suggests that the protective effect of bilirubin against malaria is mediated by a resistance mechanism that reduces Plasmodium burden and virulence.

Bilirubin accumulates in P. falciparum-infected RBCs

To explore the mechanism by which bilirubin reduces Plasmodium virulence, we asked whether bilirubin targets P. falciparum directly inside RBCs in vitro. We used scanning electron microscopy (SEM) to confirm that unconjugated bilirubin precipitates in culture medium (39) (fig. S10A), accounting for a 67% reduction of its expected concentration (fig. S10, B to D), as determined using the UnaG-based assay (fig. S1, B and C) (21, 22). There was a marked accumulation of bilirubin in P. falciparum 3D7 (Pf3D7; a commonly used strain in malaria research)–infected RBCs exposed in vitro to unconjugated bilirubin at a concentration in the range of P. falciparum seen in malaria patients (Fig. 3A and data table S5). By contrast, there was no detectable accumulation of bilirubin in noninfected RBCs, similar to vehicle-treated controls (Fig. 3A and data table S5). This suggests that unconjugated bilirubin accumulates specifically in Plasmodium-infected RBCs when present at a concentration in the range of that seen in P. falciparum malaria patients.

Fig. 3. Bilirubin inhibits P. falciparum proliferation and disrupts its mitochondrial integrity and function.

(A) Bilirubin accumulation in Pf3D7-infected RBCs (trophozoites), 8 and 12 hours after exposure to unconjugated bilirubin (41 μM) or vehicle. Data are shown as means ± SD, from one experiment with five technical replicates. (B) Percentage (left) of Pf3D7-infected RBCs, 24 to 72 hours after exposure of ring stages to increasing concentrations of unconjugated bilirubin. Data are shown as means ± SEM, pooled from six independent experiments with a similar trend, with four technical replicates per experiment. On the right are representative Giemsa-stained thin smears of Pf3D7-infected RBCs, 24 hours after exposure to unconjugated bilirubin (41 μM) or vehicle. The black arrowhead highlights parasite nuclear DNA fragmentation. Scale bars are 2 μm. (C) Relative levels of glucose, glucose 6-phosphate, succinate, Asp, fumarate, and AMP in Pf3D7-infected RBCs (trophozoites), as detected by mass spectrometry 12 hours after exposure to unconjugated bilirubin (41 μM) or vehicle. Data are shown as means ± SD, from one experiment with five technical replicates (see data table S5). (D) Schematic representation of differentially expressed genes involved in metabolic processes in Pcc-infected RBCs isolated from Blvra^−/− and Blvra^+/+ mice, as determined by scRNA-seq, in the different clusters identified in Fig. 2, C to F. (E) Representative histograms (left) and quantification of median fluorescence intensity (MFI; right) of mitochondrial volume (MitoTracker Green), 24 hours after exposure of Pf3D7-infected RBCs (rings) to increasing concentrations of unconjugated bilirubin. The gray histogram represents background staining in noninfected RBCs, and the dashed gray line (right) represents the average background signal from all replicates in noninfected RBCs. Data are shown as means ± SD from four replicates in one out of two independent experiments with a similar trend. (F) Relative level of dihydroorotate (DHO; left), orotate (middle), and uridine monophosphate (UMP; right), in Pf3D7-infected RBCs (trophozoites), as detected by mass spectrometry 12 hours after exposure to unconjugated bilirubin (41 μM) or vehicle (see data table S5). DMSO, dimethyl sulfoxide. (G) Percentage (left) of PfD10^TgDHODH-infected RBCs (rings) after exposure to increasing concentrations of unconjugated bilirubin. Data are shown as means ± SEM, pooled from three independent experiments with a similar trend, with four biological replicates per experiment. On the right are representative Giemsa-stained thin smears of Pf3D7-infected RBCs, 24 hours after exposure to unconjugated bilirubin (41 μM) or vehicle. The black arrowhead highlights pyknotic parasites. Scale bars are 2 μm. Dots represent technical replicates in (A), (E), and (F) and individual mice in (C). The p values were determined using [(A) and (E)] ordinary one-way ANOVA with Tukey’s multiple comparison test; [(B) and (G)] two-way ANOVA with Tukey’s multiple comparison test; and [(C) and (F)] Mann-Whitney U test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. The concentrations of unconjugated bilirubin in (A) to (C) and (E) to (G) were calculated according to the UnaG-based assay (21, 22) (see fig. S10).

Bilirubin arrests and kills P. falciparum

Unconjugated bilirubin inhibited the proliferation of Pf3D7 (Fig. 3B) as well as that of the P. falciparum multidrug-resistant Dd2 (fig. S11A) and IPC 5202 (fig. S11B) strains in vitro. This antiproliferative effect was dose dependent, that is, the higher the concentration of bilirubin, the lower P. falciparum proliferation (Fig. 3B and fig. S11, A and B). Importantly, the antiproliferative effect of bilirubin occurred in vitro at concentrations in the range of those seen in P. falciparum malaria patients, as assessed using the UnaG-based assay (Fig. 1E).

The antiproliferative effect of bilirubin was not associated with the induction of hemolysis, as assessed by light microscopy (Fig. 3B and fig. S11, A and B) and confirmed by lactate dehydrogenase release from lysing RBCs (fig. S11C). This suggests that unconjugated bilirubin targets Plasmodium specifically in RBCs without causing hemolysis.

A water-soluble bilirubin ditaurate derivative failed to inhibit the proliferation of Pf3D7 (fig. S11D), instead promoting ring stage proliferation at concentrations in the 40 to 80 μM range (fig. S11D). This suggests that the antiproliferative effect of bilirubin is restricted to its unconjugated form, which diffuses across cellular membranes (28, 29) and accumulates specifically in Plasmodium-infected RBCs (Fig. 3A and data table S5).

Biliverdin had no antiproliferative effect on the Pf3D7 (fig. S12A), Dd2 (fig. S12B), or IPC 5202 (fig. S12C) strains at concentrations up to 10-fold greater than those detected in Pcc-infected Blvra^−/− mice (fig. S2D). Of note, at a maximal concentration of 120 μM, there was a marginal effect of biliverdin on the proliferation of the Pf3D7 (fig. S12A) and IPC 5202 (fig. S12C) strains, consistent with previous observations (40). This suggests that circulating biliverdin does not exert antiproliferative effects on blood stages of Plasmodium at concentrations in the range of those detected in Pcc-infected Blvra^−/− mice (fig. S2E).

Bilirubin impairs Plasmodium’s energy metabolism

Unconjugated bilirubin caused a notable accumulation of glucose and, to a lower extent, glucose-6-phosphate in Pf3D7-infected RBCs compared with control Pf3D7-infected RBCs exposed to vehicle (Fig. 3C and data table S5). Moreover, unconjugated bilirubin reduced the relative levels of succinate, aspartate, fumarate, and adenosine monophosphate (AMP) in Pf3D7-infected RBCs compared with control vehicle-treated Pf3D7-infected RBCs (Fig. 3C and data table S5). This suggests that bilirubin compromises P. falciparum central carbon metabolism (41). Consistent with this notion, Pcc ring stages and trophozoites from Blvra^−/− mice showed an increase in the expression of a number of glycolytic genes, including PCHAS-121500 (enolase, putative; ENO); PCHAS-100970 (glucose-6-phosphate isomerase, putative), PCHAS-1329700 (phosphoglycerate kinase, putative), PCHAS-082370 (phosphoglycerate kinase; PGK, putative), and PCHAS-091620 (phosphoglycerate mutase; PGM1, putative) compared with control Blvra^+/+ mice (Fig. 3D and data tables S2 and S3). This suggests that unconjugated bilirubin compromises the central carbon metabolism of Plasmodium through a mechanism that targets the expression of the parasite’s glycolytic genes.

Moreover, Pcc ring stages and early trophozoites from Blvra^−/− mice also showed an increase in the expression of several genes involved in mitochondrial ATP metabolic processes, including PCHAS-041070 (ATP synthase F0 subunit d-like protein, putative), PCHAS-145260 (ATP synthase subunit beta, mitochondrial, putative), PCHAS-031590 (ATP synthase subunit alpha, mitochondrial, putative), and PCHAS-MIT0001 (cytochrome c oxidase subunit 3; COX3) (Fig. 3D and data tables S2 and S3). This suggests that unconjugated bilirubin compromises Plasmodium mitochondrial function in vivo.

Bilirubin disrupts Plasmodium’s mitochondrial function

Consistent with the notion that unconjugated bilirubin compromises Plasmodium spp. mitochondrion, Pf3D7 mitochondrial volume was severely reduced upon in vitro exposure to unconjugated bilirubin, as assessed by flow cytometry (Fig. 3E) and confirmed by live confocal microscopy imaging (fig. S13, A and B). This effect was dose dependent—that is, the higher the concentration of bilirubin, the lower the parasite’s mitochondrial volume (Fig. 3E)—at concentrations in the range of P. falciparum malaria patients, as assessed using the UnaG-based assay (Fig. 1E). Equimolar concentrations of biliverdin failed to reduce P. falciparum mitochondrial volume in vitro (fig. S13C).

Unconjugated bilirubin decreased Pf3D7 mitochondrial membrane potential (fig. S14A) and superoxide accumulation (fig. S14B) in vitro in Pf3D7-infected RBCs as compared with vehicle-treated control Pf3D7-infected RBCs (fig. S14, A and B). This effect was dose dependent, that is, the higher the concentration of bilirubin, the lower the parasite’s mitochondrial membrane potential (fig. S14A) and superoxide accumulation (fig. S14B). Equimolar amounts of biliverdin failed to reduce Pf3D7 mitochondrial membrane potential (fig. S14C) or superoxide accumulation (fig. S14D).

Bilirubin represses Plasmodium’s mitochondrial pyrimidine synthesis

Plasmodium spp. replication relies on de novo pyrimidine synthesis through the conversion of dihydroorotate (DHO) into orotate, catalyzed by the mitochondrial inner membrane dihydroorotate dehydrogenase (DHODH) (42). Exposure of Pf3D7-infected RBCs to unconjugated bilirubin caused a marked reduction in DHO, orotate, and uridine monophosphate (UMP), a downstream product of DHODH in the parasite’s de novo pyrimidine synthesis (Fig. 3F and data table S5). This suggests that the antiproliferative effect of unconjugated bilirubin is exerted through a mechanism that impairs mitochondrion-dependent pyrimidine synthesis.

Bilirubin acts beyond Plasmodium’s mitochondrion

To determine whether the antiproliferative effect of bilirubin relies exclusively on the inhibition of mitochondrial pyrimidine synthesis, we used a transgenic P. falciparum D10 strain that expresses a cytoplasmic Saccharomyces cerevisiae DHODH (PfD10^TgDHODH) (42). In contrast to the parental PfD10 strain, PfD10^TgDHODH parasites can support pyrimidine synthesis irrespectively of the mitochondrion (42). As expected (43), inhibition of the mitochondrial electron transport chain cytochrome bc₁ complex by atovaquone (ATQ) arrested the proliferation and killed the parental PfD10 strain but not the PfD10^TgDHODH strain (fig. S15A). By contrast, however, unconjugated bilirubin arrested the proliferation and killed both the parental PfD10 strain and the PfD10^TgDHODH strain (Fig. 3G and fig. S15B). This suggests that the antiproliferative effect of bilirubin acts beyond the inhibition of P. falciparum pyrimidine synthesis.

Bilirubin inhibits hemozoin crystallization

Exposure of Pf3D7 to unconjugated bilirubin in vitro was associated with dispersion of hemozoin (Hz) crystals in the infected RBCs, as visualized and quantified by live confocal microscopy imaging (Fig. 4A and fig. S16, A and B). Equimolar amounts of biliverdin had no effect on Hz, similar to vehicle controls (fig. S16, A to C). This suggests that bilirubin interferes directly or indirectly with the detoxification of the heme extracted from hemoglobin into Hz crystals.

Fig. 4. Bilirubin inhibits Hz formation and disrupts the P. falciparum food vacuole.

(A) Live confocal microscopy of Pf3D7-infected RBCs (trophozoites), 12 hours after exposure to unconjugated bilirubin (41 μM) or vehicle. Representative images (left) and quantification of Hz cellular intensity distribution (laser reflection mode, cyan; area, right) are shown. RBC membranes are stained with wheat germ agglutinin (RBC; red), and nuclei (N) with Hoechst (DNA; blue). Data are shown as means ± SD from two independent experiments (n = 10 to 15 parasites). Scale bars are 4 μm. (B) Relative inhibition of β-hematin crystallization by bilirubin versus chloroquine. Equimolar amounts of brivanib alaninate (a tyrosine kinase inhibitor) were used as control. The relative inhibition of β-hematin crystallization was inferred from heme accumulation measured at a wavelength of 405 nm (λ₄₀₅ nm) (95) and is represented as the mean of λ/λ₄₀₅ nm ± SD from two replicates in one out of two independent experiments with a similar trend. (C) Hz quantification in blood from Blvra^+/+ and Blvra^−/− mice 7 days after Pcc infection, at daily light (left) and dark (right) cycles. Data for Hz (i.e., nM heme) are shown as means ± SD, pooled from two independent experiments with a similar trend (n = 11 for light cycle and n = 9 or 10 for dark cycle per genotype). (D) Representative images (left) and quantification (right) of the food vacuole. RBC membranes are stained with wheat germ agglutinin (RBC; red), the parasite’s food vacuole (FV) with LysoTracker Green (green), and nuclei (N) with Hoechst (DNA; blue). Data are shown as means ± SD from two independent experiments (n = 10 to 13 parasites), as in (A). Scale bars are 4 μm. AU, arbitrary units; AUC, area under the curve. (E) Representative transmission electron microscopy images of Pf3D7-infected RBCs (trophozoites), 12 hours after exposure to unconjugated bilirubin (41 μM) or vehicle. The images on the right correspond to amplifications of the orange dotted and dashed areas highlighted in the images on the left. EV, endocytic vesicle; FV, food vacuole; Hz, hemozoin; MLB, multilamellar bodies; N, nucleus; RBC, red blood cell. Images are representative of three independent experiments with similar results. Scale bars are 500 nm. (F) Relative levels of amino acids in Pf3D7-infected RBCs (trophozoites), as detected by mass spectrometry, 12 hours after exposure to unconjugated bilirubin (41 μM) or vehicle (see data table S5). Circles represent individual parasites in (A) and (D), individual mice in (C), and technical replicates in (F). The p values were determined using [(A), (C), (D), and (F)] Mann-Whitney U test. NS is not significant; *p < 0.05; **p < 0.01; ****p < 0.0001. The concentrations of unconjugated bilirubin in (A) to (C) and (E) to (G) were calculated according to the UnaG-based assay (21, 22) (see fig. S10).

We then asked whether bilirubin interferes directly with Hz crystallization. Consistent with this hypothesis, computational simulation of a putative bilirubin docking in the deep groove on the fastest growing (i.e., 001) face of the crystal suggested that bilirubin interferes directly with Hz crystallization (fig. S17). The two top-scoring docking poses suggest that the adsorption of bilirubin to the crystal surface is facilitated by hydrogen bonding between its carboxylic acid moiety and the pyrrole groups of heme in the nascent Hz crystals (fig. S17). This suggested that bilirubin hinders heme incorporation in nascent Hz crystals, impairing the parasite’s capacity to neutralize cytotoxic labile heme (44, 45). This hypothesis was confirmed using a standard in vitro assay to quantify the crystallization of β-hematin, a synthetic heme adduct that is chemically and spectroscopically identical to Hz (46). Unconjugated bilirubin inhibited β-hematin formation (Fig. 4B), mediated by the lipid mimic Nonidet P-40 detergent, as monitored by the accumulation of heme as a bis-pyridyl complex (47). The median inhibitory concentration (IC₅₀) of bilirubin was 1 to 1.2 mM, that is, 50 to 60 times less potent than that of chloroquine (IC₅₀ = 20 μM) in the same assay.

Consistent with the notion that bilirubin inhibits Hz crystallization in vitro, Pcc infection in Blvra^−/− mice was associated with higher levels of Hz accumulation in circulating infected RBCs than in infected RBCs from control Blvra^+/+ mice, at the light and dark cycles of Pcc infection (Fig. 4C). This suggests that bilirubin exerts an antimalarial effect through Hz crystallization, akin to quinoline-based antimalarial drugs (45, 48).

Bilirubin disrupts the P. falciparum food vacuole

Inhibition of Hz crystallization by quinoline-based antimalarial drugs compromises the parasite’s digestive food vacuole (45, 48), suggesting that bilirubin might also disrupt the parasite’s food vacuole. In support of this hypothesis, unconjugated bilirubin disrupted the food vacuole of Pf3D7 trophozoites, as revealed by leakage of its acidic content, which was quantified by confocal microscopy live imaging (Fig. 4D and fig. S18, A and B). Equimolar amounts of biliverdin failed to disrupt the food vacuole of Pf3D7 trophozoites, similar to vehicle-treated control Pf3D7 trophozoites (fig. S18, A to C).

Disruption of the P. falciparum food vacuole by bilirubin was further confirmed by transmission electron microscopy (Fig. 4E and fig. S19), which revealed the formation of multilamellar bodies (Fig. 4E and fig. S19), reminiscent to those observed when targeting endosomal vesicle delivery to the food vacuole (49). This was associated with the accumulation of Hz crystals in the parasite’s cytoplasm (Fig. 4E and fig. S19), with a structural appearance consistent with the loss of the Hz crystals’ characteristic sharp rectangular morphology (Fig. 4E and fig. S19). These “rounded-edge” cytoplasmic Hz crystals were not observed in parasites exposed to equimolar amounts of biliverdin or to vehicle (Fig. 4E and fig. S19).

The Plasmodium food vacuole is vital for the acquisition of essential amino acids contained in hemoglobin (50). That bilirubin impairs this vital process is supported by the reduction in the relative amount of these essential amino acids in Pf3D7 trophozoites exposed to unconjugated bilirubin, as illustrated for leucine (Leu), phenylalanine (Phe), valine (Val), threonine (Thr), tryptophan (Trp), lysine (Lys), and proline (Pro) (Fig. 4F and data table S5). This was also observed for nonessential amino acids contained in hemoglobin, including aspartate (Asp) (Fig. 3C and data table S5), arginine (Arg), tyrosine (Tyr), and serine (Ser) (fig. S20A and data table S5). In sharp contrast, amino acids not contained in hemoglobin, which are not acquired via the food vacuole (50), were not affected by unconjugated bilirubin, as illustrated for methionine (Met) and isoleucine (Ile) (Fig. 4F). Moreover, bilirubin had no effect on the amino acid content of noninfected RBCs compared with vehicle-treated controls (fig. S20B and data table S5). This suggests that bilirubin specifically compromises the capacity of the P. falciparum food vacuole to extract essential amino acids from hemoglobin.

Several genes involved in hemoglobin digestion by the food vacuole were also increased in Pcc ring stages and trophozoites from Blvra^−/− mice compared with those from Blvra^+/+ mice (fig. S20C and data tables S2 and S3). These included PCHAS-083340 (M18 aspartyl aminopeptidase, putative), PCHAS-113650 (falcilysin; FLN, putative; highly active at acidic pH consistent with its critical role in hemoglobin degradation), PCHAS-131350 (M17 leucyl aminopeptidase, putative), and PCHAS-131310 (M17 leucyl aminopeptidase, putative) (fig. S20C and data tables S2 and S3), suggesting that bilirubin also acts in vivo to compromise the capacity of Plasmodium trophozoites to extract essential amino acids from hemoglobin, likely contributing to the antimalarial effects of unconjugated bilirubin.

Discussion

Our findings support the notion that the accumulation of unconjugated bilirubin in plasma during Plasmodium spp. infection (Fig. 1, B to E and G) represents a protective metabolic response to malaria (Fig. 1, H, I, and L). This is in keeping with growing evidence that suggests that bilirubin exerts major physiological functions (22, 51, 52), challenging the strongly held notion that unconjugated bilirubin is a final “waste product” that accumulates in plasma as a result of liver dysfunction (53).

The protective effect of unconjugated bilirubin against Plasmodium infection is propelled by BVRA (Fig. 1, G and H) and sustained by the inhibition of bilirubin conjugation through the repression of hepatic UGT1A1 (Fig. 1, J to L). Consistent with our findings, several UGT1A1 genetic hypomorphic variants are associated with mild unconjugated nonhemolytic hyperbilirubinemia in individuals of African ancestry (54, 55). These include the hypomorphic variant UGT1A1*28 (rs3064744) responsible for Gilbert’s syndrome (19, 56), which reduces UGT1A1 transcription by 70% in an estimated prevalence of 15 to 25% of individuals of African ancestry compared with 0 to 5% and 5 to 10% in Asian and Caucasian ancestries, respectively (54). However, an association between polymorphisms in or near the UGT1A1 locus (chr2:233,760,270-233,773,300) and P. falciparum malaria severity was not reported in previous genome-wide association studies (57–61). One possible explanation for this is that UGT1A1 genetic variants can increase the incidence and severity of neonatal jaundice (54, 55). Moreover, UGT1A1 genetic variants might cosegregate with balanced polymorphisms that confer protection against malaria through the induction of heme catabolism, such as sickle hemoglobin (5).

The targeting of Plasmodium inside RBCs by unconjugated bilirubin (Fig. 3A) is consistent with unconjugated bilirubin crossing cellular membranes (28, 29), despite its binding to albumin in plasma (12). Presumably this occurs through aqueous diffusion (62), suggesting that cellular membranes have higher affinities for unconjugated bilirubin than albumin (62).

The inhibition of Plasmodium spp. proliferation (Figs. 2A and 3B and fig. S5) and virulence (Fig. 2G and data tables S2 and S4) by unconjugated bilirubin is associated with repression of the parasite’s capacity to consume glucose through glycolysis (Fig. 3, C and D), presumably inhibiting mitochondrial tricarboxylic acid cycle function (Fig. 3, C and D). Moreover, unconjugated bilirubin disrupts the parasite’s mitochondrial structure (Fig. 3E and fig. S13, A and B) and function (fig. S14A), compromising de novo pyrimidine synthesis (Fig. 3F) and therefore inhibiting Plasmodium spp. proliferation (42). This, however, is not sufficient to fully explain the antimalarial effects of bilirubin (Fig. 3G).

Although Plasmodium spp. evolved to detoxify redox-active labile heme into redox-inert Hz crystals, the proliferation of these parasites inside RBCs is associated with intravascular hemolysis and the release of labile heme into plasma (1, 3–6). This induces the expression of HO-1 in the infected host, which catabolizes labile heme into biliverdin (3, 4, 9), fueling the production of bilirubin by BVRA. As it accumulates in plasma, unconjugated bilirubin inhibits Hz crystallization (Fig. 4, A to C, and fig. S16, A and B), similar to, although far less potent than, quinoline-based antimalarial drugs such as chloroquine (Fig. 4B) (45, 48). The inhibition of Hz crystallization leads to the disruption of the parasite’s food vacuole (Fig. 4, D and E, and fig. S18), compromising the extraction of essential amino acids from hemoglobin (Fig. 4F) while presumably leading to the accumulation of cytotoxic labile heme.

The induction of bilirubin production in response to Plasmodium spp. infection is a metabolite-based resistance mechanism (63) against malaria. We speculate that although evolutionarily conserved, this defense strategy carries, as an evolutionarily trade-off (64), the insidious prevalence of neonatal jaundice (65, 66), which can lead to encephalopathy (67). Considering the selective pressure exerted by malaria over the course of human evolution (68), it is conceivable that the antimalarial effects of unconjugated bilirubin outcompeted the fitness costs associated with the high incidence of neonatal jaundice in populations originating in endemic areas of malaria (69, 70).

Limitations

The association of asymptomatic P. falciparum malaria with a higher ratio of unconjugated over conjugated bilirubin compared with symptomatic P. falciparum malaria, suggests that unconjugated bilirubin counters the transition of asymptomatic to symptomatic P. falciparum malaria. Although consistent with the observation that repression of bilirubin conjugation is protective against malaria in mice, whether this is also the case for P. falciparum malaria remains to be established. Moreover, given the limited number of P. falciparum malaria patients analyzed, these observations should be confirmed by future independent studies.

The observation that Plasmodium infection is associated with the repression of bilirubin conjugation in mice suggests that the accumulation of unconjugated bilirubin in plasma from asymptomatic P. falciparum malaria patients is protective against the onset of disease. This, however, is a hypothesis that remains to be verified. If this were the case, one would expect future genome-wide association studies to determine that genetic polymorphisms that regulate bilirubin conjugation are associated with P. falciparum malaria incidence and/or outcome.

The molecular mechanism by which bilirubin represses P. falciparum proliferation was associated with disruption of the parasite’s mitochondrion and food vacuole. However, whether these are functionally linked is not clear.

Although our findings suggest that the antimalarial effect of bilirubin is exerted through a direct effect on the parasite, this should not exclude bilirubin from exerting additional protective effects that would act irrespectively of the parasite. These are likely to be exerted through its antioxidant effects, which should contribute to limiting tissue damage and possibly enforcing disease tolerance to malaria.

Materials and methods

Human data

Clinical patients’ data were collected from a prospective clinical cohort study, conducted between March and July 2022 at Centre de Recherches Médicales de Lambaréné (CERMEL), Gabon. The study was approved by the ethics committee of CERMEL (DEMIT-GAB: CEI 18/2021, and all patients provided written informed consent and were included in this analysis if they fulfilled the following criteria: a) age ≥18 years; b) microscopically confirmed asexual P. falciparum infection by positive blood smear; c) blood drawn before or ≤24 hours after initiation of antimalarial treatment. Patients were excluded in case of known liver (e.g., hepatitis, hepatic cancer) or hematologic disease, pregnancy or breastfeeding, mixed Plasmodium infection and with missing initial parasitemia data, hematology data, liver function data or bilirubin data. Patients were grouped as asymptomatic within 7 days before and 7 days after diagnosis or symptomatic malaria according to WHO criteria (71). Among 42 participants meeting the criteria for this analysis, the mean age was 33.1 (SD = 16.97) years (asymptomatic: 36.0 (SD = 19.6); symptomatic: 29.9 years (SD = 13.2). Overall, 48% (20/42) were male and 52% (22/42) female [males among asymptomatic: 45% (10/22); symptomatic: 50% (10/20)]. Median initial parasitemia was 576 infected RBC (iRBC) (IQR: 131 to 4,201) per μl (asymptomatic: 322,5 [110 to 851]; symptomatic: 4,289 [322 to 48,218]). Unconjugated bilirubin was measured exclusively by the UnaG-based assay (21, 22) in 4 asymptomatic and 1 symptomatic malaria patients out of the 42 patients. Parasitological analyses, hematology and biochemistry were performed by accredited laboratories at CERMEL and Charité - Universitätsmedizin Berlin. Study data were collected and managed using REDCap (Research Electronic Data Capture) tools hosted at CERMEL and Charité - Universitätsmedizin Berlin (72).

Mice

Mice were bred and maintained under specific pathogen-free (SPF) conditions at the Gulbenkian Institute of Molecular Medicine (GIMM), housed at standard vivarium temperature (22°C) on a regular light cycle, lights on from 8:00 a.m. (ZT 0) to 8:00 p.m. (ZT 12). When indicated, mice were kept on inverted light cycle, lights on from 8:00 p.m. (ZT 0) to 8:00 a.m. (ZT 12). Mice were maintained with free access to water and standard chow pellets (ad libitum). All experimental protocols were approved in a two-step procedure, by the Animal Welfare Body of the GIMM and by the Portuguese National Entity that regulates the use of laboratory animals in research (Direção Geral de Alimentação e Veterinária; DGAV). Experimental procedures followed the Portuguese (Decreto-Lei n^o 113/2013) and European (Directive 2010/63/EU) legislation. C57BL/6J and DBA/2 mice were obtained from the GIMM animal facility. C57BL/6J Blvra^−/− mice were generated at Ozgene (Australia), as described (22). Age-matched wild-type C57BL/6J mice were used as controls and co-housed at least 2 weeks with Blvra^−/− mice before infection.

Plasmodium chabaudi chabaudi AS infection and disease assessment

Mice (females and males, 8-14 weeks old) were infected with Plasmodium chabaudi chabaudi AS (Pcc AS) or transgenic GFP-expressing Pcc AS (Pcc AS-GFP_ML) (73). Infections were performed by intraperitoneal (i.p.) administration of freshly isolated blood (Passage 34-39; 2 × 10⁶ or 2 × 10⁵ infected RBC diluted in 200 μl PBS) collected from a previously infected C57BL/6J mouse. Mice were monitored daily from day 0 (day of infection) onwards for parasitemia (% infected RBC; iRBC), parasite burden (number of iRBC per μl of blood) and survival, essentially as described (9, 31). Briefly, the number of RBC per μl of blood was quantified by flow cytometry (LSR Fortessa X20 analyzer; BD Bioscience) using a standard concentration of reference latex beads (10 μm; Coulter CC Size Standard L10, Beckman Coulter, no. 6602796), gating on RBC, based on size and granularity and on bead population. For Pcc AS, parasitemia was determined manually by optical microscopy, counting the number of iRBC in at least 4 fields of Giemsa-stained blood smears (1000× magnification). For Pcc AS-GFP_ML, parasitemia was determined by flow cytometry, according to the percentage of RBC expressing GFP (GFP⁺ RBC). Parasite burden, expressed as iRBC/μl, was determined by multiplying the parasitemia by the number of RBC. Representative Giemsa-stained thin blood smear images were acquired suing a Zeiss Imager Z2/ApoTome.2, equipped with an Axiocam 105 color camera, using the 100× 1. 4NA Oil immersion objective, in a 5×5 tile stitched with Zeiss’s ZEN v3.1. The images were analyzed using the Fiji Software (ImageJ) and the background in parasite RGB images was corrected using the Color Correct plugin for Fiji, contributed by Gabriel Landini (https://github.com/landinig/IJ-Colour_Correct/blob/main/colour_correct.zip).

Bilirubin supplementation during Plasmodium infection

Blvra^−/− mice were infected with Pcc AS-GFPML (2 × 10⁵ iRBC, 200 μl, i.p.) and monitored as described above. Bilirubin (Frontier Scientific) stock solutions were prepared as described (74). Briefly, bilirubin was dissolved in 0.2N NaOH, buffered to pH 7.4 using 0.2N HCl, filtered (70 μm cell strainer) to remove precipitates and stored (−80°C). Absorbance was determined using a spectrophotometer (SmartSpec 3000) at 460nm. The concentration was calculated considering 53,846 cm⁻¹ M-1⁻¹ as molar extinction coefficient and following the Lambert-Beer law (A460nm = ε.C. l). Bilirubin was administered i.p. at 30 mg/kg or 3 mg/kg, once daily, from day 4 to day 15 after infection.

Repression of hepatic Ugt1a1 in vivo

A recombinant adeno-associated virus serotype 8 (AAV8) encoding the Staphylococcus aureus (Sa) CRISPR associated protein 9 (Cas9) and a single guide RNA (gRNA) targeting Ugt1a1 (AAV8-gRNA- Ugt1a1) was administered to 2-4 days old DBA/2 mice (3 × 10¹⁴ viral particles/kg body weight, in PBS; i.v. retro-orbitally), as described (27). Controls were transduced with a AAV8 encoding SaCas9 without the targeting gRNA (AAV8-Cas9). Female transduced mice were infected with Pcc (2 × 10⁶ iRBC, 200 μl, i.p.) at ~10 weeks after birth and monitored for survival, parasite burden and disease severity, as described above. Adult male transduced mice were sacrificed for organ collection to assess Ugt1a1 deletion efficiency by Western blot.

Plasmodium virulence assay

Pcc parasites were adoptively transferred from Blvra^+/+ vs. Blvra^−/− mice into female Blvra^−/− mice, essentially as described (31). Briefly, blood was collected from both Blvra^+/+ and Blvra^−/− mice 7 days after Pcc infection and the same number of iRBC (2x10⁶; 200 μl PBS) were passively transferred (i.p.) into recipient Blvra^−/− mice, monitored daily for disease assessment and survival, as described above.

Pcc-infected RBC sequestration

Blvra^+/+ and Blvra^−/− mice were infected with Pcc and kept on inverted light cycle, lights on from 8:00 p.m. (ZT 0) to 8:00 a.m. (ZT 12). Mice were sacrificed (CO₂ asphyxiation) 7 days after infection, perfused in toto via transcardiac infusion of ice-cold PBS (1X, 15 ml), organs were harvested, snap frozen in liquid nitrogen and stored at −80°C. The accumulation of Pcc AS 18S rRNA in different organs was quantified by qRT-PCR. RNA extraction and qRT-PCR were performed as described in the section “RNA extraction and qRT-PCR.”

Cecal ligation and puncture (sepsis)

Cecal ligation and puncture (CLP) was performed in Blvra^+/+ and Blvra^−/− mice (females and males, 8-14 weeks old), essentially as described (75). Briefly, mice were anesthetized (Ketamine: 12 mg/ml; and Xylazine: 1.6 mg/ml in 1x PBS; 180-200 μl i.p.) and subjected to a 15-20% cecum ligation and double puncture with a 23 gauge (G) needle. A small amount of feces was extruded, and the cecum was carefully placed back into the abdominal cavity. All animals received 0.9% saline (40 ml/kg, i.p.) and Imipenem/Cilastatin (25 mg/kg, i.p.), starting 2 hours after CLP and every 12 hours for 3 days. Mice were monitored daily from day 0 (day of infection) onward for body weight (Ohaus CS200 scaler, Sigma Aldrich), rectal temperature (Rodent thermometer; BIO-TK8851, Bioset), blood glucose concentration (Accu-CHECK Performa glucometer, Roche), and survival.

Bacterial load

Mice were sacrificed 24 hours after CLP and peritoneal fluid was obtained by peritoneal lavage (7 ml sterile PBS). Mice were perfused with sterile ice cold 1x PBS. Whole organs were harvested and homogenized under sterile conditions in 1 ml sterile ice cold 1x PBS using a dounce tissue grinder (Sigma Ref: D8939-1SET). Serial dilutions were plated onto TrypticaseSoy Agar II with 5% Sheep Blood plates (Becton Dickinson Ref:254053) and incubated (24 hours at 37°C) in air 5% CO₂ (aerobes) or in an airtight container equipped with the GasPak anaerobe container system (Becton Dickinson Ref 260678). Anaerobic conditions were confirmed in all experiments using BBL Dry Anaerobic Indicator Strips (Becton Dickinson Ref: 271051).

Influenza A virus infection

Blvra^+/+ and Blvra^−/− mice were maintained at the GIMM BSL-2 facility, and infected with influenza A/X-31, essentially as described (76, 77). Briefly, mice were anesthetized with isoflurane and 30 μl of inoculum (10³ PFU in sterile PBS), administered intranasally. Mice were monitored daily, as described above. Viral loads were determined at day 3 post-infection from the right lower lobes of influenza A virus-infected mice. Samples were homogenized in serum-free DMEM (Gibco) using tungsten carbide beads (Qiagen) in a TissueLyser II (Qiagen) at 20 s⁻¹ for 3 min and supernatants were collected after centrifugation. Titration by plaque assay was performed as previously described using Madin-Darby Canine Kidney cells (76).

UnaG protein synthesis and purification

UnaG was expressed in pMAL-6P2-6xHIS in BL21(DE3) cells (22). Starter cultures were grown to saturation in Luria Broth (LB) (overnight at 37°C), diluted 10-fold in LB and grown to an OD600 of 0.3 at 37°C, after which cultures were moved to 18°C. UnaG expression was induced at OD600 of 0.6, by the addition of 400 μM isopropyl β-D-1-thiogalactopyranoside (16 hours, 18°C). Cells were harvested by centrifugation (3,800 g, 4°C, 25 min), resuspended in 15 ml of resuspension buffer (50 mM HEPES, 300 mM NaCl, 0.5 mM TCEP, 10% glycerol, 1 mM PMSF, 2.34 μM leupeptin, 1.45 μM pepstatin at pH 7.4) per liter of culture. The sample was further lysed by sonication (QSonica Sonicator) with an amplitude of 5% in a pulse mode of 0.8sec ON and 0.5sec OFF for total 30 s. Lysate was clarified by centrifugation (26,000 g, 4°C, 30 min) and loaded onto an amylose column (MBPTrap HP Column, GE Healthcare). Protein was eluted with 20 mM maltose in protease buffer (50 mM Tris, 150 mM NaCl, 0.5 mM TCEP, 10% glycerol, 0.01% TritonX-100 at pH 7.4). MBP was removed by the addition (16 hours, 4°C) of PreScission Protease (GE Healthcare). UnaG was purified via a nickel column (HisTrap HP Column, GE Healthcare) to remove the cleaved tag and eluted with (30 mM Tris, 1M NaCl, 0.5 mM TCEP, 10% glycerol, 500 mM imidazole, pH 7.4). UnaG was further purified on a gel-filtration column (S-200, GE Healthcare) in gel filtration buffer (50 mM HEPES, 300 mM NaCl, 0.5 mM TCEP, 10% glycerol at pH 7.4), concentrated using 10 kDa amicon ultracentrifugal filter (Merck), and flash frozen in gel filtration buffer at 30% glycerol for storage at −80°C.

UnaG-based assay for quantification of unconjugated bilirubin

Blood from Blvra^+/+ and Blvra^−/− mice was obtained from the submandibular (facial) vein, before (Day 0) and after (Days 4, 7, 15 and 25) Pcc infection. Samples were collected in the dark, under a red light, immediately centrifuged (1,000 g, 15 min) and plasma was collected, frozen in liquid nitrogen and stored at −80°C until used for bilirubin quantification. Total plasma unconjugated bilirubin was determined using UnaG (78). Plasma sample from human patients or mice non-infected or infected with Pcc was diluted 1:100 in PBS. Then, 100 pM of UnaG was added to the diluted plasma sample and incubated for 10 min at RT. After incubation, fluorescence intensity was measured at excitation λ₄₈₀ nm and emission λ₅₃₀ nm using microplate reader (Promega GloMax). Unconjugated bilirubin (Frontier Chemicals) was used as standard.

Clinical quantification of unconjugated bilirubin

Clinical routine bilirubin measurements were performed at accredited laboratories at Charité – Universitätsmedizin Berlin and Centre de Recherches Médicales de Lambaréné, Gabon, using the modified colorimetric Diazo method (79) as implemented by Roche Diagnostics. In brief, bilirubin reacts with 3,5-dichlorophenyldiazonium salt to form azobilirubin. The red color intensity of azobilirubin is directly proportional to bilirubin concentrations. For measurement of total bilirubin, an accelerator is added to the sample to solubilize albumin-bound unconjugated bilirubin before adding the salt, allowing for photometric measurement of total bilirubin (i.e., conjugate and unconjugated bilirubin). Unconjugated bilirubin concentrations are calculated by subtracting conjugate from total bilirubin.

iRFP synthesis and purification

The iRFP gene (Addgene plasmid no. 31857) was subcloned into pGEX-6P-2 (GE Healthcare Life Sciences) expression vector (22) and subsequently transformed into BL21(DE3) cells for protein expression. Starter cultures were grown to saturation in Luria Broth (LB) (overnight at 37°C), diluted 10-fold in LB and grown to an OD600 of 0.3 at 37°C, after which cultures were moved to 18°C. iRFP expression was induced at OD600 of 0.6, by the addition of 400 μM isopropyl β-D-1-thiogalactopyranoside (16 hours, 18°C). Cells were harvested by centrifugation (3,800 g, 4°C, 25 min), resuspended (15 ml; 50 mM HEPES, 300 mM NaCl, 0.5 mM TCEP, 10% glycerol, 1 mM PMSF, 2.34 μM leu- peptin, 1.45 μM pepstatin; pH 7.4) per liter of culture. The sample was further lysed by sonication (QSonica Sonicator) with an amplitude of 5% in a pulse mode of 0.8 s ON and 0.5 s OFF for total 30 s. Lysate was clarified by centrifugation (26,000 g, 4°C, 30 min) and loaded onto GST column (GSTrap Column, GE Healthcare). Protein was eluted with 10 mM glutathione in protease buffer (50 mM Tris, 150 mM NaCl, 0.5 mM TCEP, 10% glycerol, 0.01% TritonX-100 at pH 7.4). GST was removed by the addition of PreScission Protease (GE Healthcare) for 16 hours at 4°C. iRFP was further purified on a gel-filtration column (S-200, GE Healthcare) in gel filtration buffer (50 mM HEPES, 300 mM NaCl, 0.5 mM TCEP, 10% glycerol at pH 7.4), concentrated using 10 kDa amicon ultracentrifugal filter (Merck), and flash frozen in gel filtration buffer at 30% glycerol for storage at −80°C.

iRFP-based assay for quantification of biliverdin

Blood was harvested from the submandibular (facial) vein of Blvra^+/+ and Blvra^−/− mice, before (Day 0) and after (Days 4, 7 and 15) Pcc infection. Plasma was collected in the dark, under a red light, immediately after centrifugation (1,000 g, 15 min), frozen in liquid nitrogen and stored at −80°C until used. Biliverdin concentration in plasma was quantified using biliverdin-inducible infrared fluorescent protein (iRFP) (25). Briefly, plasma samples were diluted (1:100 in PBS) and incubated with iRFP (87 nM; 15 min at RT). Fluorescence intensity was measured at excitation λ₆₉₀ nm and emission λ₇₁₃ nm using a microplate reader (BioTek Synergy H1). Biliverdin hydrochloride (Frontier Chemicals) was used as standard.

Bilirubin and biliverdin

For the in vitro studies, stock solution (34 mM) of high purity bilirubin and biliverdin (Frontier Scientific, Inc., Logan, UT, USA) were dissolved in DMSO (Sigma-Aldrich) and protected from light. Stock solutions were further diluted (1 mM) in RPMI 1640 supplemented with lipid-rich bovine serum albumin (0.5% AlbuMAXII; Invitrogen, Thermo Fisher Scientific). Albumin concentration in the culture medium was 75 μM. Unconjugated bilirubin was added to RBC or P. falciparum iRBC in culture medium (40, 80 and 120 μM) corresponding to 15, 25 and 41 μM of unconjugated bilirubin, as quantified within 2 hours using a UnaG-based assay (see above) (21, 22).

P. falciparum in vitro culture

P. falciparum 3D7-GFP (Pf3D7-GFP; MRA-1029, MR4, ATCC Manassas Virginia; chloroquine and artemisinin-sensitive), Dd2 (Pf-Dd2; MRA-150, MR4, ATCC Manassas Virginia; chloroquine-resistant), IPC 5202 (Pf5202; MRA-1240, MR4, ATCC Manassas Virginia; artemisinin-resistant) and Plasmodium falciparum D10 (PfD10; atovaquone-sensitive) and its transgenic derivative (PfD10^TgDHODH) (42) were co-cultured (37°C; 95% humidity, 5% of CO₂) with human RBC from healthy donors (5% hematocrit), replacing human serum by 0.5% AlbuMAXII (Invitrogen, Thermo Fisher Scientific), as described (31, 80). Cultures were synchronized using 5% D-Sorbitol (Sigma-Aldrich). After RBC reinvasion, the parasites at the schizont stage were suspended, layered onto 70% Percoll (Sigma-Aldrich), centrifuged (1,000g; 15 min; no brake), collected from the upper layer of Percoll cell suspension, washed (PBS 1X) and incubated in standard culture conditions until RBC reinvasion (81). Ring stage (approximately 10-12 hours after RBC invasion) or trophozoite (approximately 18-20 hours after RBC invasion) stage parasites were diluted to approximately 1% parasitemia and 3% hematocrit in RPMI 1640 complete medium, (75 μM bovine serum albumin), seeded on a 96-well plate and exposed to vehicle (DMSO; represented in figures as 0 μM bilirubin/biliverdin) or different concentrations of bilirubin (15-41 μM, diluted in DMSO), biliverdin or water soluble bilirubin ditaurate (40-120 μM; Frontier Specialty Chemicals Inc.) for 24, 48 and 72 hours. Parental PfD10 and its transgenic derivative PfD10^TgDHODH were also treated with atovaquone (10 nM; Sigma-Aldrich) for the same period of time. Final DMSO concentration was always kept below 0.5%. Slides were prepared at 24, 48 and 72 hours and stained with Giemsa solution. In parallel, parasites were suspended (PBS 1X) and analyzed by flow cytometry (Beckman Coulter CytoFLEX) to determine the percentage of iRBC. PfDd2 and Pf5202 parasites were stained with 0.5X SYBR green I (Invitrogen, Thermo Fisher Scientific) and PfD10 and PfD10^TgDHODH parasites with 0.5X SYTO Deep Red Nucleic Acid Stain (Invitrogen, Thermo Fisher Scientific) in PBS (45 min in the dark; 37°C) and washed in PBS prior to acquisition. Pf3D7-GFP parasites were acquired directly (GFP⁺ signal). Representative Giemsa-stained thin blood smear images were acquired as described for Pcc.

Hemolysis assay

Pf3D7 parasites cultured in 96-well plates, were synchronized and ring stages (~1% parasitemia, 3% hematocrit) were treated with different concentrations of bilirubin, as described above. At 72 hours after treatment, 96-well plates were centrifuged (300 g, 5 min), supernatants were collected and stored at 4°C and parasitemias were quantified by flow cytometry as described above. 2% Triton X-100 was used as a positive control to establish 100% RBC lysis. Lactate dehydrogenase (LDH) release into the medium was quantified using the LDH-GloTM Cytotoxicity assay (Promega, no. J2380). Briefly, culture medium was diluted in LDH Storage Buffer (200 mM Tris-HCl, pH 7.3, 10% Glycerol, 1% BSA) to reach a 5X dilution. LDH detection reagent (50 μl of LDH detection enzyme mix + 0,25 μl Reductase Substrate) was added to each sample (1:1 ratio), incubated (60 min; RT) and the luminescence was recorded using the GloMax microplate reader (Promega). The percentage of cytotoxicity was calculated using the following formula: (Experimental LDH release – medium background)/(Maximum LDH release control – medium background) × 100.

RNA extraction and qRT-PCR

Mice were sacrificed by CO₂ asphyxiation, transcardially perfused in toto with ice-cold PBS (1X, 15 ml) and organs were harvested, snap frozen in liquid nitrogen and stored at −80°C. Total RNA was extracted using tripleXtractor reagent (GRISP), chloroform, isopropanol, and ethanol, according to manufacturer’s instructions. cDNA was synthesized using the Xpert cDNA Synthesis Mastermix (GRiSP), followed by qRT-PCR using the iTaq Universal SYBR Green Supermix (Bio-Rad) on a QuantStudio 7 Flex Real-Time PCR System (Applied Biosystems). Transcript values were calculated from the threshold cycle (Ct) of each gene using the 2^−ΔΔCT method using Acidic ribosomal phosphoprotein P0 (Arbp0) as the housekeeping control gene. Primers for qPCR include: Arbp0, Fwd: 5′-CTTTGGGCATCACCACGAA-3′, Rev: 5′-GCTGGCTCCCACCTTGTCT-3′; Blvra-deletion confirmation, Fwd: 5′-AGCCGCTGGTAAGCTCC-3′, Rev: 5′-ACCAACCACTACCACACCAAA-3′; Blvra, Fwd: 5′-ATTCTGCCACCATGGAAA-3′, Rev: 5′-CTCCAAGGACCCAGATTTGA-3′; Hmox1, Fwd: 5′-TGACACCTGAGGTCAAGCAC-3′, Rev: 5′-TCTCTGC AGGGGCAGTATCT-3′; Ugt1a1, Fwd: 5′-TCTGGCTGATGAGAAGTGACT-3′; Rev: 5′-GAAAACAACGATGCCATGCT-3′; Pcc AS 18S rRNA, Fwd: 5′-AAGCATTAAATAAAGCGAATACATCCTTAT-3′, Rev: 5′- GGGAGTTTGGTTTTGACGTTTATGCG-3′.

Flow cytometry (P. falciparum)

Pf3D7-GFP cultures were synchronized as described before. Ring stage (approximately 10-12 hours after invasion) or trophozoite stage (approximately 18-20 hours after invasion) parasites (approximately, 3% parasitemia) were treated with vehicle (0.12-0.47% vol/vol DMSO) or different concentrations of unconjugated bilirubin (15-41 μM) or biliverdin (40-120 μM) for 24 hours, collected, washed with 1X PBS and stained as previously described (31). Cells were stained with MitoTracker Green FM (200 nM in 1X PBS; Invitrogen, Thermo Fisher Scientific), MitoTracker Deep Red FM (200 nM in 1X PBS; Invitrogen, Thermo Fisher Scientific), and MitoSOX Red Mitochondrial Superoxide Indicator (5 μM in 1X PBS; Invitrogen, Thermo Fisher Scientific) for 20 min, 37°C, 5% CO₂. Cells were washed once with 1X PBS and stained with Hoechst 33342 (10 μM in 1X PBS; Thermo Fisher Scientific) for 20 min, 37°C, 5% CO₂. Cells were again washed once with 1X PBS and analyzed in a FACSAria Ilu Cell Sorter (BD Biosciences). FACS data was analyzed with FlowJo V10.8.1. Gating strategy is illustrated in data S1 and S2.

Western blot

Mice were sacrificed by CO₂ asphyxiation, transcardially perfused in toto with ice-cold PBS (1X, 15 ml) and organs were harvested and stored at −80°C. Tissues were lysed in 2% SDS-PAGE sample buffer (100 mM Tris, pH 6.8, 20% glycerol, 4% SDS, 0.2% bromophenol blue, 100 mM DTT) supplemented with 1X protease inhibitor cocktail (cOmplete, Mini, EDTA-free Protease Inhibitor Cocktail; Roche) and homogenized in a TissueLyser II (Qiagen) with tungsten carbide beads (Qiagen). Supernatants were collected, boiled (5 min, 95°C) and total protein was quantified at λ₂₈₀ nm using the DS-11 FX Spectrophotometer (DeNovix). Proteins (50 or 150 μg) were resolved on a 12% SDS-PAGE and transferred to Polyvinylidene fluoride (PVDF) membranes. Membranes were blocked (1 hour at RT; 5% milk in 1X TBS-T), washed (1X TBS-T) and incubated (overnight at 4°C) with primary antibodies (5% BSA in 1X TBS-T): rabbit polyclonal anti-BVRA (Invitrogen, ThermoFisher Scientific, PA5-92059; 1:1000), rabbit polyclonal anti-HO-1 (Enzo Life Sciences, ADI-SPA-896F; 1:1000), rabbit monoclonal anti-GAPDH (Cell Signaling Technology, clone 14C10, #2118; 1:1000) and rabbit polyclonal anti-β-actin (Cell Signaling Technology, no. 4967; 1:1000). Membranes were washed (3 times in 1X TBS-T) and incubated (2 hours; RT) with the peroxidase-conjugated secondary antibody (HRP conjugated goat anti-rabbit IgGH+L; Invitrogen, no. 31460; 1:5000; 5% milk in 1X TBS-T). Membranes were washed (3 times in 1X TBS-T) and peroxidase activity was detected using SuperSignal West Pico PLUS Chemiluminescent Substrate (ThermoFisher Scientific). Blots were developed using Amersham Imager 680 (GE Healthcare), equipped with a Peltier cooled Fujifilm Super CCD. Western blot analysis was performed using ImageJ (Rasband, W.S., ImageJ, U.S. NIH, Bethesda, Maryland, USA), from images without saturated pixels. Uncropped Western blot membranes are shown in data S3. For hepatic UGT1A1 detection the procedure was modified as follows: tissues were lysed using radioimmunoprecipitation assay (RIPA) buffer (pH 7.5) (50 mM TrisHCl, 1% NP40, 0.25% deoxycholic acid, 150 mM NaCl, 1 mM EGTA, 1 mM Sodium Orthovanadate, 1 mM Sodium Fluoride and 1X protease inhibitor cocktail (cOmplete, Mini, EDTA-free Protease Inhibitor Cocktail; Roche) and homogenized in a TissueLyser II (Qiagen) with tungsten carbide beads (Qiagen). Samples were centrifuged (15000 g 15min, 4°C), the supernatants were collected, and the total protein concentration was quantified by the Bradford assay, according to the manufacturer instructions (Bio-rad Protein Assay Dye Reagent Concentrate, no. 5000006). Protein (40 μg) were resolved on a 10% SDS-PAGE and transferred to Polyvinylidene fluoride (PVDF) membrane. The membrane was blocked (2 hours at RT; 5% milk in 1X PBS-T) and incubated with rabbit polyclonal anti-UGT1A1 (Boster Biological Technology, A01865-1; 1:1000). The membrane was washed (3 times in 1X-PBS-T) and incubated (1 hour, RT) with the peroxidase-conjugated secondary antibody (HRP conjugated goat anti-rabbit IgGH+L; Invitrogen, no. 31460; 1:5000; 5% milk in 1X PBS-T). The membrane was washed (3 times in 1X-PBS-T) and the peroxidase activity was detected as described above.

High-pressure freezing, freeze substitution, and transmission electron microscopy

Pf3D7-GFP cultures were synchronized as described (81). Trophozoite stage parasites (approximately 18-20 hours after invasion, 4-5% parasitemia) were treated with vehicle (0.35% vol/vol DMSO) or bilirubin (41 μM) or biliverdin (120 μM) for 8 or 12 hours, collected and fixed overnight at 4°C with 1% (v/v) glutaraldehyde (Science Services) in 0.1 M cacodylate buffer (Sigma-Aldrich) with 1 mM CaCl₂ (Alfa Aesar) and 3 mM MgCl₂ (Alfa Aesar). Parasites were pelleted by centrifugation (300 g; 5 min), washed 3 times in 0.1M cacodylate buffer (Sigma- Aldrich) (pH 7.2), mixed in RPMI medium containing 20% (w/v) polyvinylpyrrolidone 40 (Sigma-Aldrich) as a cryoprotectant and high-pressure-frozen using a High-Pressure Freezer Compact 02 (Wohlwend Engineering Switzerland). The samples were freeze-substituted using a Leica AFS2 equipped with a Leica EM FSP Robot with 0.2% (w/v) uranyl acetate (Analar) in acetone during 1 hour at −140°C, followed by an increasing slope of 4°C/hour until −90°C and then samples were substituted for 6 hours at −90°C. The temperature was raised to −50°C at a slope of 5°C/hour and the samples were washed three times in acetone for 1 hour each step. Samples were then infiltrated in Lowicryl HM20 resin (Polysciences) at increasing concentrations of 33, 66 and 100% and polymerized with UV light at −30°C for 72 hours after a slope of 5°C/hour. Sections of 70 nm thickness were cut using a Leica EM UC7 ultramicrotome using an Ultra 45° diamond knife (Diatome) and mounted on palladium-copper 1×2mm slot grids coated with 1% (w/v) formvar (Agar Scientific) in chloroform (VWR). Sections were stained with 2% (w/v) uranyl acetate (Analar) in 70% methanol (VWR) and Reynold’s lead citrate (Sigma-Aldrich; 5 min each) and analyzed using a Tecnai G² Spirit BioTWIN Transmission Electron Microscope from FEI operating at 120 kV and equipped with an Olympus-SIS Veleta CCD Camera.

Scanning electron microscopy

Vehicle (0.35% vol/vol DMSO) or bilirubin (41 μM) was added to the culture medium used to culture PfD7-GFP for 12h. The samples were collected, centrifuged (500 g, 5 min) and washed (3X in 100 mM NaHCO₃, pH 9.0 and 2% SDS; 11,000 g, 30 min; and 5X in distilled H2O to remove the salts and detergents (4 times 11,000 g,10 min and the last wash 11,000 g, 30 min). Samples were stored in 200 μl of distilled H₂O at 4°C until processing. 20 μl of each sample were added to a silicon wafer and allowed to air dry. Before analysis, samples were coated with gold for 5 s. Samples were imaged in a Quanta 650 FEG Scanning Electron Microscope from FEI operating at 5 kV.

scRNA-seq

scRNA-seq was performed essentially as described (31, 32). Briefly, Blvra^+/+ and Blvra^−/− mice were sacrificed by CO₂ asphyxiation 7 days after Pcc AS-GFP infection (i.e., peak of infection) and the blood was collected by cardiac puncture. The blood was diluted 1:75 in 1X PBS supplemented with 1% FBS and iRBC (GFP⁺ cells) were sorted on a BD FACSAria Ilu. Cells were washed twice with 1X PBS supplemented with 1% FBS and adjusted to a concentration of 1200 cells/ml.

Single-cell gene expression library preparation and sequencing

Library preparation and sequencing were performed essentially as previously described (31, 32). Briefly, samples were processed with Chromium Single Cell Controller to generate barcoded single-cell gel bead emulsions (GEMs) following the Chromium Next GEM Single Cell 3’ protocol. Single-cell cDNA libraries were obtained after the GEM-RT (reverse-transcriptase) clean-up and cDNA amplification on a Bio-Rad C1000 Touch Thermal Cycler. The cDNA profiles were checked on a Fragment Analyzer System (Agilent Technologies) according to the HS NGS Fragment kit (Agilent Technologies) manual. The scRNA-seq libraries were generated and indexed (Dual Index Plate TT, Set A) on a Bio-Rad C1000 Touch Thermal Cycler. The final 3′ Gene Expression libraries were verified and quantified on a Fragment Analyzer System (Agilent Technologies) following the HS NGS Fragment kit (Agilent Technologies) and samples were sequenced on a NextSeq 2000 (Illumina) using the NextSeq 1000/2000 P2 kit 100 cycles (Read 1: 28 Cycles; Read 2: 90 cycles).

scRNA-seq data analysis

Single-cell analyses were performed with Seurat (v.4.3.0) (82–84) and scuttle (v1.2.1) (85). Cells expressing less than 200, more than 1,500 genes, or more than 4,000 UMIs were filtered out. Remaining cells were annotated by projection to the malaria cell atlas, P. falciparum 10x set 4 (https://www.malariacellatlas.org/data-sets/) (86) using scmap (v1.14.0) (87) through 1:1 orthologs downloaded on 18.01.2024 from PlasmoDB (https://plasmodb.org/plasmo/app/search/transcript/GenesByOrthologPattern). The atlas and query samples were subset to the 1:1 ortholog gene set and log1p normalized. A cell index of the atlas was built using 500 genes and the cells from the query samples were individually projected using the scmapCell() function. The Blvra^+/+ and Blvra^−/− samples were individually normalized with scran (v1.20.1) (88) using the quickCluster() method, and integrated using STACAS (v2.0.1) (89) with 750 genes and semi-supervised by the scmap projected stages. The integrated data was visualized by projection into UMAP (Uniform Manifold Approximation and Projection) using 8 principal components and the umap-learn method through Seurat. Analysis of differentially expressed genes of Blvra^−/− vs. Blvra^+/+ was performed individually for different stage clusters based on the scmap projection, using the Wilcoxon rank sum test through Seurat. Functional enrichment analysis of significantly (Bonferroni-adjusted p < 0.05) up- or down-regulated genes was performed using gprofiler2 (v.0.2.2) (90). The archived version of the gprofiler2 server - Ensembl 108, Ensembl Genomes 55 (database built on 2022-12-28) was used: https://biit.cs.ut.ee/gprofiler_archive3/e108_eg55_p17/gost. All the analyses were performed with the R programming language in a containerized docker image publicly available at Docker Hub (elolabfi/sctoolkit) running R (v.4.2.1) (R Core Team, 2021) (91), RStudio server (2022.07.2 Build 576), Seurat (v.4.3.0) and its dependencies. RNA velocity was performed with velocyto (v.0.17.17) (92) and scvelo (v.0.2.5) (33) as described in (32) with python (v.3.9.5) run under Jupyter lab (v.3.5.3). All data are available in data tables S1 to S4.

Live-cell imaging and confocal microscopy

Pf3D7 cultures were synchronized, as described above and ring (10-12 hours after RBC invasion) or trophozoite (18-20 hours after invasion) stage parasites (3% parasitemia) were incubated with vehicle (0.35% vol/vol DMSO), unconjugated bilirubin (41 μM) or biliverdin (120 μM). RBC were harvested 12 hours or 24 hours for food vacuole and Hz or mitochondrion analyses, respectively. Cells were washed (1X PBS; 400 g, 2 min) and stained (30 min, RT) with MitoTracker Green FM (200 nM in 1X PBS; Invitrogen) or LysoTracker Green DND-26 (75 nM in 1X PBS; Invitrogen), Hoechst 33342 (20 μM in 1X PBS) and Wheat Germ Agglutinin (WGA) Alexa Fluor 633 (3 μg/ml in 1X PBS; Thermo Fisher Scientific). Cells were washed twice (1X PBS; 400 g, 2 min) and the cell pellet was resuspended in 1 ml of PBS. Then, 200 μl of the cell suspension were placed in chambered coverslip with 8 wells and a no. 1.5H glass bottom (Ibidi GmbH, Munich, Germany). After this step, the cells were analyzed using a laser scanning confocal microscope (Leica TCS-SP5) with a continuous Ar-ion and HeNe and a Ti:sapphire laser (Spectra-Physics Mai Tai BB, 710–990 nm, 100 fs, 82 MHz). Both MitoTracker Green FM and LysoTracker Green DND-26 were imaged using the 488 nm Ar+ laser line (with emission set at 500-610 nm) and the 633 nm He–Ne laser line was used for imaging WGA-Alexa 633 dye (emission 644-736 nm). Hoechst 33342 images were recorded in the multiphoton mode under 810 nm excitation (420-550 nm), and hemozoin was visualized using the laser reflection mode. Images (512 × 512 pixels) were collected using a 63× 1.2 N.A. water immersion objective (HCX PL APO CS 63.0x 1.20WATERUV) at a scan rate of 100 Hz laser. To measure the relative area and fluorescence intensity corresponding to LysoTracker Green (i.e., food vacuole) or Hz, a line was drawn across the diameter of P. falciparum 3D7 infected RBC, intercepting the food vacuole, using ImageJ 1.53k software (Rasband, W.S., ImageJ, U.S. NIH, Bethesda, Maryland, USA. Relative Intensity and distance were measured using ImageJ. For LysoTracker Green, Area under the curve (AUC) was calculated from the above values. Values were calculated from 2 independent experiments analyzing 5-7 parasites per experiment. To measure the relative area corresponding to MitoTracker Green staining, images were analyzed using Imaris 10.0.0 software. The surface around the MitoTracker Green was created using the Surface Tool in Imaris and the area per parasite was measured. Values were calculated from 2 independent experiments analyzing 15-17 parasites per experiment.

Hz quantification

Hz was quantified essentially as described (93, 94), with the following adaptations. Briefly, blood was obtained by heart puncture, from Blvra^+/+ and Blvra^−/− mice 7 days after Pcc infection. Mice were housed at a regular light cycle: lights on from 8:00 a.m. (Zeitgeber time: ZT 0) to 8:00 p.m. (ZT 12), or at an inverted light cycle: lights on from 8:00 p.m. (ZT 0) to 8:00 a.m. (ZT 12). Blood (300-400 μl) was collected at ZT 3 and 15, hypotonically lysed (4 ml H₂O), centrifuged (11,000 g, 45 min). The supernatant fraction was removed, and pellets (Hz) were washed (3X in 100 mM NaHCO₃, pH 9.0 and 2% SDS; 11,000 g, 30 min). Hz was dissolved (1 ml of 100 mM NaOH, 2% SDS, 3 mM EDTA), sonicated (1 min Branson SLPe Digital Sonifier) and centrifuged (11,000 g, 30 min). The heme released from Hz (200 μl) was quantitated by spectrophotometry (405 nm) in a 96 well plate reader (MultiscanSky, ThermoFisher) with molar extinction coefficient of 5.7 × 10⁴. Heme concentration was calculated according to the Lambert-Beer law: A = ε.c.l (A – absorbance at l₄₀₅ nm; ε – extinction coefficient of hemin-57000; c – concentration in molar (M); l – path length) and normalized to the volume of blood used. Data are presented as Hz (i.e., nM heme).

β-Hematin inhibition assay

The Nonidet P-40 (NP-40) detergent-mediated assay (47) adapted for high-throughput screening in 96-well plate was used to determine inhibition of β-hematin formation. Following a period of 4-5 hours incubation at 37°C, the formation of a bis-pyridyl hemochromogen is detectable at a wavelength of 405 nm, and allows quantification of the free heme component that has not reacted to form β-hematin (95). Measurements were made using a Thermo Scientific Multiskan GO plate reader. Sigmoidal dose-response curves were plotted in GraphPad Prism to determine half maximal inhibitory concentration (IC₅₀) values. Measurements were performed twice, each in technical duplicate, and the values are reported together with standard deviation.

Metabolite extraction

Pf3D7-GFP cultures were synchronized as described (81). Trophozoite stage parasites (approximately 18-20 hours after invasion, 15% parasitemia, 5% HCT, 2 ml culture per sample) and treated with vehicle (0.35% vol/vol DMSO) or bilirubin (41 μM) for 8 or 12 hours. The chemicals used were LC-MS grade water, acetonitrile (ACN), methanol (MeOH), and isopropanol (IPA), which were obtained from Th. Geyer (Germany). High-purity methyl tert-butyl ether (MTBE), ammonium formate, formic acid, ammonium acetate, and acetic acid were purchased from Merck (Germany). Stable isotope labelled internal standards for metabolomics (MSK-A2-1.2; Cambridge Isotope Laboratories, MA, USA) were used at final concentrations of 1.0% (vol/vol). The culture medium was completely aspirated and the cells were washed with 1XPBS (500 g, 5 min). For biphasic extraction of lipids and polar metabolites, samples were initially quenched by incubation on dry ice with 400 μl of 75% (vol/vol) cold methanol for 20 min and the appropriate internal standards (0.5 μl each/sample) were added. After incubation, the samples were vortexed for 5 min at maximum speed, lysed using an ultrasonic bath to sonicate the sample for 5 min and vortexed again briefly after sonication. After addition of 1000 μl of cold MTBE, the monophasic mixture was vortexed for 60 s and incubated at −20°C for 20 min. For phase separation, 250 μl of cold water were added, followed by another vortexing and incubation step (see previous conditions). The biphasic solvent system was then centrifuged for 15 min at 14,000 g and 4°C. For metabolomics analysis, 400 μl of the bottom aqueous phase were transferred, dried under a stream of nitrogen, and reconstituted in 75 μl 80% MeOH (v/v). The final samples were vortexed for 10 min, centrifuged (see previous conditions) and the supernatants were transferred to analytical glass vials for LC-MS/MS analysis.

LC-MS/MS analysis

LC-MS/MS analysis was performed on a Vanquish Horizon UHPLC system coupled to an Orbitrap Exploris 240 high-resolution mass spectrometer (Thermo Scientific, MA, USA) in negative and positive ESI (electrospray ionization) mode. To perform untargeted metabolomics, chromatographic separation was carried out on an Atlantis Premier BEH Z-HILIC column (Waters, MA, USA; 2.1 mm × 100 mm, 1.7 μm) at a flow rate of 0.25 ml/min. The mobile phase consisted of water:acetonitrile (9:1, v/v; mobile phase phase A) and acetonitrile:water (9:1, v/v; mobile phase B), which were modified with a total buffer concentration of 10 mM ammonium acetate (negative mode) and 10 mM ammonium formate (positive mode), respectively. The aqueous portion of each mobile phase was pH-adjusted (negative mode: pH 9.0 via addition of ammonium hydroxide; positive mode: pH 3.0 via addition of formic acid). The following gradient (20 min total run time including re-equilibration) was applied [time (min)/%B]: 0/95, 2/95, 14.5/60, 16/60, 16.5/95, 20/95. Column temperature was maintained at 40°C, the autosampler was set to 4°C and sample injection volume was 5 μl. Analytes were recorded via a full scan with a mass resolving power of 120,000 over a mass range from 60 – 900 m/z (scan time: 100 ms, RF lens: 70%). To obtain MS/MS fragment spectra, data-dependent acquisition was carried out (resolving power: 15,000; scan time: 22 ms; stepped collision energies [%]: 30/50/70; cycle time: 900 ms). Ion source parameters were set to the following values: spray voltage: 4100 V (positive mode) / −3500 V (negative mode), sheath gas: 30 psi, auxiliary gas: 5 psi, sweep gas: 0 psi, ion transfer tube temperature: 350°C, vaporizer temperature: 300°C. All experimental samples were measured in a randomized manner. Pooled quality control (QC) samples were prepared by mixing equal aliquots from each processed sample. Multiple QCs were injected at the beginning of the analysis in order to equilibrate the analytical system. A QC sample was analyzed after every 5th experimental sample to monitor instrument performance throughout the sequence. For determination of background signals and subsequent background subtraction, an additional processed blank sample was recorded. Data was processed using MS DIAL 4.9.221218 (96) and raw peak intensity data was normalized via total ion count of all detected analytes (97). Level 1 feature identification was based on an in-house library for metabolomics (EMBL-MCF 2) (98) using accurate mass, isotope pattern, MS/MS fragmentation, and retention time information and a minimum matching score of 80%. All the raw data results for annotated metabolites are available in data table S5.

Statistical analysis

For clinical data we used Mann-Whitney U test. Analysis was done in JMP Pro version 16 (SAS Institute Inc, Cary, NC, USA). For all other analyses, statistically significant differences between two experimental groups were assessed using a two-tailed unpaired Mann-Whitney U test or t test (in samples tested positive for normal distribution) and comparisons between more than two groups were assessed using one-way ANOVA, two-way ANOVA (parasitemias and parasite burden) or two-way ANOVA with Bonferroni’s or Tukey’s multiple comparison test. Survival curves are represented by Kaplan-Meier plots and differences between the groups were assessed using the log-rank test. All statistical analyses were performed using GraphPad Prism software. Differences were considered statistically significant at a P value < 0.05. NS: not-significant, p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

For in vivo Plasmodium infection experiments, sample size was maintained no less than 5 mice per experimental condition, in at least 2 independent experiments, except when indicated in the figure legends. For in vitro experiments, sample size was maintained no less than 4 replicates per experimental condition, in a minimum of 3 independent experiments, except when indicated in the figure legends.

Randomization was used in mice monitoring after infection. Samples were analyzed blindly, often being processed by a different investigator from the one preparing/collecting the samples.

Inclusion/exclusion criteria for human studies are detailed in the “Human data” section. For in vivo Plasmodium infection experiments, mice presenting a >2-day delay in the initial dynamics of parasite growth and peak of parasitemia were considered as technical outliers and were excluded from further analysis.

Data and materials availability:

All data are available in the main text or the supplementary materials. Clinical patient data is available from F.K. (florian.kurth@charite.de) upon request. scRNA-seq data are available in the Gene Expression Omnibus under accession number GSE254821 (99), and metabolomics data are available in the Metabolomics Workbench data repository (100).