Functional Roles for C5a and C5aR but Not C5L2 in the Pathogenesis of Human and Experimental Cerebral Malaria

Abstract

INTRODUCTION

Malaria is a leading cause of global morbidity and mortality, with an estimated 243 million cases and 1.24 million deaths occurring worldwide annually (1, 2). The majority of malaria deaths are due to severe and cerebral malaria (CM) caused by Plasmodium falciparum infection (2). Currently, the first-line therapy for CM is intravenous artesunate, based on the superiority of artesunate over quinine in large randomized trials in adults and children with severe malaria (3, 4). However, despite treatment with parenteral artesunate, the case fatality rate of CM remains high, indicating that antimalarial therapy alone may be insufficient to prevent fatality and long-term neurocognitive deficits in survivors (3, 4).

The pathogenesis of CM is multifactorial and involves both host and pathogen determinants. These include dysregulated host immune responses to infection and the sequestration of parasitized erythrocytes (PE) in cerebral microvasculature contributing to vessel occlusion, hypoxia, endothelial activation, and blood-brain barrier (BBB) dysfunction (5,–7). Elevated levels of inflammatory cytokines, such as gamma interferon (IFN-γ) and tumor necrosis factor (TNF), are often observed in human CM compared to uncomplicated malaria (8, 9). While it is unclear whether inflammatory cytokines have a causal role in human CM, it is postulated that these cytokines contribute to endothelial activation and sequestration of PEs in the brain microvasculature by upregulating the expression of cell adhesion molecules such as ICAM-1 on cerebral endothelium to which PEs bind (7, 10, 11). This hypothesis is supported, at least in part, by observations in the murine model of experimental CM (7, 12, 13). A detailed understanding of the molecular basis of pathological host responses in CM may facilitate the identification of novel interventions to improve clinical outcome over that achievable with antimalarial therapy alone.

The complement system represents a key component of the innate immune response and consists of three major pathways—classical, lectin, and alternative—which all converge at the cleavage of C5 into C5a and C5b (14). Additional pathways can generate C5a independently of C3 via thrombin or serine proteases (15, 16). C5a is a potent inducer of inflammatory mediators and antiangiogenic factors (17,–20). C5a signaling primarily through its major receptor, C5aR, has been shown to contribute to a variety of pathological processes, including sepsis, ischemia-reperfusion injuries, age-related macular degeneration, autoimmune disorders, altered angiogenesis, and neurodegenerative disorders (21, 22).

Increased complement activation has been consistently observed in human malaria infections (23). However, in human studies, it has been difficult to establish a causal role for complement activation in the pathogenesis of CM due to the challenges of performing mechanistic studies in human populations. Using murine models of experimental CM (ECM) to investigate causality, previous studies have reported that genetic deficiency of C5 (24, 25) or antibody blockade of C5a or C5a receptor (C5aR) (25) significantly protects against ECM. C5a has an alternative receptor, C5L2, whose role in malaria has not been previously reported. C5L2 is similar to C5aR in structure, with 60% sequence homology. However, it lacks the key amino acid motif (known as the DRY motif) required for binding G proteins and is, therefore, unable to induce classic G protein-dependent signaling functions, such as the release of intracellular Ca²⁺ (26, 27). C5a and its degradation product, C5a_desArg, display high affinity for both C5aR and C5L2 (28). Because C5L2 is not coupled to G proteins, it was hypothesized that C5L2 may be a nonsignaling decoy receptor for C5a or that it may have an anti-inflammatory role (26, 29). In contrast, other studies have implicated a proinflammatory role for C5a-C5L2 signaling in different disease contexts, such as sepsis (30).

In this study, we investigated the roles of C5a, C5aR, and C5L2 in the pathogenesis of CM. Here we show that circulating levels of C5a are elevated in children with CM compared to the levels in children with uncomplicated malaria (UM). Moreover, C5a contributed to the pathogenesis of ECM via C5aR, as evidenced by decreased inflammation, decreased endothelial activation, and improved blood-brain barrier integrity and survival in C5aR^−/− but not in C5L2^−/− mice. These results support a role for C5a and C5aR but not C5L2 in the pathogenesis of ECM.

MATERIALS AND METHODS

RESULTS

Deletion of C5aR but not C5L2 improves survival in a mouse model of ECM.

We investigated whether C5a and its receptors, C5aR and C5L2, play causal roles in ECM by examining survival following P. berghei ANKA infection of C5aR^−/− and C5L2^−/− mice compared to their wild-type (WT) counterparts. In these experiments, ∼90% of the WT mice succumb to ECM, displaying various degrees of difficulty in righting, ataxia, seizures, and coma, as described previously (36). Importantly, survival was significantly improved in C5aR^−/− mice (Fig. 1A) (log rank test, P = 0.004). In contrast, C5L2^−/− mice were fully susceptible to ECM, with survival similar to that of WT mice. No significant difference was observed in levels of parasitemia among the three groups (Fig. 1B). To confirm that our findings were robust, we repeated survival studies in 8 additional independent experiments (totaling n = 79 to 90 mice per group), and in each experiment, significantly improved survival was observed in C5aR^−/− mice compared to C5L2^−/− and WT mice (log rank test for the combined results, P < 0.0001) (data not shown).

Open in a separate window

FIG 1

Deletion of C5aR, but not C5L2, significantly improves survival in ECM. Wild type (WT) (black circles, solid line), C5aR^−/− (blue diamonds, dashed lines), and C5L2^−/− (red circles, dotted lines) mice were infected with 1 × 10⁶ P. berghei ANKA-parasitized erythrocytes (PEs) and were monitored for survival (A) (pooled result from four independent experiments; n = 30 to 35/group) and peripheral parasitemia (B) (representative of four independent experiments; n = 8 to 10 in each experiment). Survival differences were assessed by the log rank test. Differences between the WT and C5aR^−/− mice were significant (P = 0.004) (A).

Deletion of C5aR but not C5L2 reduces systemic inflammation.

Dysregulated systemic inflammation, characterized by high circulating levels of proinflammatory cytokines, such as tumor necrosis factor (TNF), IFN-γ, and the chemokine, monocyte chemoattractant protein 1 (MCP-1) (CCL2), has been associated with poor clinical outcomes in human infections and in mouse models of malaria (7, 9, 37, 38). Mice susceptible to P. berghei ANKA develop a marked systemic proinflammatory response during infection (7, 37, 39). To determine if improved survival in C5aR^−/− mice was associated with decreased levels of malaria-induced inflammation, serum levels of proinflammatory cytokines were assessed on day 5 postinfection (Fig. 2). Baseline levels of all cytokines measured were below the limit of detection in uninfected mice from the three groups (data not shown). Compared to P. berghei ANKA-infected WT mice, infected C5aR^−/− mice had significantly lower serum levels of TNF, IFN-γ, and MCP-1 (Fig. 2A to ​toC)C) (P < 0.05). These data suggest that a reduced inflammatory response can contribute to the improved survival observed in C5aR^−/− mice and are in agreement with previous studies showing that C5a-C5aR signaling contributes to the synergistic induction of proinflammatory cytokines and chemokines in response to malaria parasite products (17).

Open in a separate window

FIG 2

Deletion of C5aR but not C5L2 reduces systemic inflammation in ECM. (A to E) Levels of TNF, IFN-γ, MCP-1 (CCL2), IL-6, and IL-10 were measured from serum collected on day 5 postinfection by using the mouse inflammation cytometric bead array. Box plots represent median and interquartile range (IQR), and whiskers denote range. *, P < 0.05, Kruskal-Wallis test for comparing all groups, followed by Mann-Whitney test to compare WT and C5aR^−/−. Data are representative of 3 independent experiments (n = 8 to 10/group).

HMGB-1 is associated with progression of ECM.

High-mobility group protein B-1 (HMGB-1) is an inflammatory mediator associated with tissue injury and sepsis (40). In a previous study examining the role of C5a and its receptors in the pathogenesis of sepsis, Rittirsch et al. reported that the plasma level of HMGB-1 was significantly increased in experimental sepsis and that the release of HMGB-1 required C5L2, suggesting that HMGB-1 is a key mediator of C5a downstream of C5L2 in a murine model of sepsis (19). We investigated whether the same relationship existed between C5L2 and HMGB-1 in ECM by determining circulating HMGB-1 levels on day 5 postinfection in WT, C5aR^−/−, and C5L2^−/− mice. Due to the lack of reliable mouse HMGB-1 ELISA reagents, Western blotting was employed to measure the expression of HMGB-1, as described previously (32). In order to account for the semiquantitative nature of Western blotting, quantitative comparisons of the HMGB-1 band densities were made only among the bands blotted on the same membrane (Fig. 3). All blots were repeated twice using plasma samples collected from at least two independent infection studies. No significant difference was found at the basal level of HMGB-1 expression among three groups (data not shown). In agreement with a previous study, the plasma level of HMGB-1 significantly increased after P. berghei ANKA challenge in WT mice (Fig. 3A and ​andD)D) (41). Of note, C5L2-deficient mice showed a similar pattern of induction of HMGB1 upon P. berghei ANKA infection (Fig. 3C) (P < 0.05), suggesting that C5L2 is not required for HMGB1 release in the ECM model. In contrast, induction of HMGB1 after P. berghei ANKA challenge was not observed in C5aR^−/− mice (Fig. 3B).

Open in a separate window

FIG 3

HMGB-1 is associated with progression of ECM. Expression levels of HMGB-1 were measured by Western blotting from plasma collected on day 5 (D5) postinfection. (A to C) Band intensities for day 0 (gray bars) and day 5 (white bars) in each group were compared on the same blot by using the NIH Image J software. *, P < 0.05 (paired t test). Histograms represent means ± standard errors of the means (SEM) obtained from pooled results from two independent Western blot analyses from two independent P. berghei ANKA infection studies (n = 15 to 20/group). Representative Western blots are shown in panel D.

Deletion of C5aR but not C5L2 reduces endothelial cell activation.

In addition to enhanced production of proinflammatory cytokines, endothelial activation is a central feature of CM (10, 42, 43). The angiopoietin (Ang)-Tie2 pathway is a key regulator of endothelial activation and vascular integrity (44). Constitutive interaction between angiopoietin 1 (Ang-1) and Tie2 functions to maintain endothelium quiescence, while Ang-2 antagonizes the Ang-1–Tie2 interaction, promoting vascular permeability (44). Several studies have previously reported that elevated levels of Ang-2 and sICAM-1 and reduced levels of Ang-1 are correlated with severe disease and fatal outcome in patients with CM (5, 34, 45, 46). Similar observations have been reported in ECM after P. berghei ANKA challenge (47, 48). Consistent with these studies, we observed that P. berghei ANKA infection was associated with decreased levels of Ang-1 and increased levels of sICAM-1 in all three groups (Fig. 4). Importantly, endothelial activation was significantly attenuated in the C5aR-deficient mice infected with P. berghei ANKA, which displayed higher levels of Ang-1 and lower levels of sICAM than their WT counterparts (P < 0.05). Due to a lack of ELISA reagents to measure circulating mouse Ang-2, we assessed the changes in Ang-2 by measuring Ang-2 mRNA using the quantitative real-time PCR in brain homogenates of infected mice (Fig. 4C and ​andD).D). Transcript levels of Ang-2 and ICAM-1 were significantly lower in the C5aR^−/− mice on day 5 postinfection than those in WT mice (P < 0.05). Collectively, these data support the hypothesis that C5aR deficiency can contribute to the improved survival in ECM by promoting endothelial quiescence.

Open in a separate window

FIG 4

C5aR deletion but not C5L2 is associated with reduced endothelial activation in ECM. (A and B) Serum samples collected from on days 0 and 6 postinfection (D0 and D6, respectively) were analyzed for the levels of Ang-1 and sICAM-1 by ELISA. (C and D) Quantitative real-time PCR was performed on RNA extracted from the brains collected on days 0 and 5 postinfection to measure the mRNA levels of Ang-2 and ICAM. Copy numbers of Ang-2 and ICAM were normalized by the copy numbers of GAPDH and β-actin. Box plots represent the median and IQR, and whiskers denote range. The Mann-Whitney test was performed to compare WT mice on D0 and D5/D6 postinfection as a control (denoted as “*”) and to compare WT and C5aR^−/− mice on D5/D6 (denoted as “**”). * and **; P < 0.05. Data are representative of two independent experiments (n = 8 to 10/group).

Deletion of C5aR promotes blood-brain barrier integrity.

One of the hallmarks of CM is the loss of BBB integrity (6, 10, 43, 49). Based on our observation that C5aR deletion can improve endothelial quiescence, we hypothesized that BBB integrity would be better preserved in C5aR^−/− mice after P. berghei ANKA infection. Therefore, we compared Evans blue extravasation in parasitemia-matched WT versus C5aR^−/− mice following P. berghei ANKA infection. Consistent with our hypothesis, Evans blue leakage into brain parenchyma was significantly reduced in C5aR^−/− mice compared to that in WT mice with similar parasitemia (Fig. 5) (P < 0.05).

Open in a separate window

FIG 5

Deletion of C5aR improves blood-brain barrier integrity in ECM. (A) Representative photographs of brains from WT mice (left panels) and C5aR^−/− mice (right panels) after Evans blue infusion on day 6 postinfection. Peripheral parasitemia is indicated in the bottom right corner of each panel. (B) Quantification of total Evans blue extravasation in brain as an indication of vascular leak. (C) Peripheral parasitemia of WT mice and C5aR^−/− mice used for Evans blue extravasation. Box plots represent the median and IQR, and whiskers denote range. *, P < 0.01, Mann-Whitney test (n = 6 to 8/group). Data are representative of two independent experiments.

C5a is elevated in children with CM.

In order to extend these murine observations to an informative patient population, we investigated a role for C5a in the pathobiology of human CM by comparing plasma levels of C5a in Ugandan children with CM to those in matched children with UM in a nested case-control design (Table 1). Circulating C5a levels were significantly higher in children with CM versus those in children with UM as assessed by the Mann-Whitney test (median [interquartile range] for CM, 43.7 [22.9 to 59.5], versus UM, 22.4 [15.7 to 32.2]; P = 0.0004) (Fig. 6). All other patient characteristics were comparable between the two groups as assessed by the Mann-Whitney test. The median parasitemia was higher among the children with CM, but this difference did not reach statistical significance.

TABLE 1

Patient characteristics^a

Characteristic	Value for patient group
	UM	CM
No. of patients	33	31
Age (yr)	3.0 (1.7–6.8)	3.1 (1.6–4.4)
Parasitemia (no. of parasites/μl)	3.9 × 104 (6,120–1.4 × 105)	1.1 × 105 (1.8 × 104–2.8 × 105)
Wt (kg)	13.0 (10.0–21.0)	13.0 (10.0–15.0)
No. of days reported sick prior to presentation	3.0 (2.0–4.0)	3.0 (2.0–3.0)
C5a level (ng/ml)	22.4 (15.7–32.2)	43.7 (22.9–59.5)b

Open in a separate window

^aAll variables except the number of patients are presented as medians (interquartile ranges). A Mann-Whitney test was performed to assess a difference between the UM and CM groups for each variable.

^bP = 0.0004.

Open in a separate window

FIG 6

C5a levels at presentation are significantly increased in children with cerebral malaria. C5a levels were measured in plasma of Ugandan children with uncomplicated malaria (UM; n = 33) and cerebral malaria (CM; n = 31) at clinical presentation. The Mann-Whitney test was performed for the comparison of the groups.

DISCUSSION

The complement system plays an important role in the host response to infectious processes. However, complement activation is tightly regulated, and the loss of this control can contribute mechanistically to the pathogenesis of multiple conditions, including sepsis, asthma, transplant rejection, and neurodegenerative disorders (18, 19, 21, 22, 50). Although previous studies have reported activation of complement via several pathways during malaria infection (23), it has been difficult to determine whether excessive complement activation plays a causal role in the pathogenesis of severe disease in humans. In this study, we present several lines of evidence supporting a causal role for C5a-C5aR signaling in the pathogenesis of CM. Using an experimental model of CM, we extend our previous findings that deficiency of C5 confers protection as does antibody blockade of C5a or C5aR in CM-susceptible wild-type mice (25). Providing further mechanistic insights, our data suggest that the protective mechanisms by which C5aR deficiency improves outcome are likely to involve reduced systemic inflammation, decreased endothelial activation, reduced PE sequestration in the brain, and improved vascular integrity: features previously associated with improved clinical outcomes in human malaria (5, 34, 43, 45, 51, 52). Importantly, our analysis revealed divergent roles for C5aR and C5L2 in the context of malaria infection. Deletion of C5L2 failed to reduce inflammation or endothelial activation or improve survival. We extended these observations to human CM and show in a nested case-control study that C5a levels are associated with disease severity and are significantly higher in African children presenting with CM than in children with uncomplicated malaria, which is consistent with our previous finding in ECM (25). Collectively these data implicate C5a in human CM and provide direct evidence for a role for C5a acting through C5aR in the pathogenesis of ECM.

Reduced levels of proinflammatory mediators (TNF and IFN-γ) and markers of improved endothelial stabilization (decreased levels of sICAM-1 and increased levels of Ang-1) were observed in C5aR^−/− mice following P. berghei ANKA infection. This is consistent with the role of C5a as a potent inducer of inflammation and with our previous in vitro findings that C5a and Plasmodium falciparum bioactive product, pfGPI, can synergistically induce cytokines (interleukin-6 [IL-6], TNF, IL-1β, and IL-10), chemokines (IL-8, MCP-1, macrophage inflammatory protein 1α [MIP-1α], and MIP-1β) and the anti-angiogenic factor sFlt-1 (17). Dysregulation of proinflammatory cytokines and chemokines is likely to contribute to aberrant accumulation of mononuclear cells and PEs in the cerebral vasculature in the CM-susceptible mice, as previously reported (25). Importantly, mice that are naturally resistant to ECM do not express C5a, underscoring the involvement of C5a-C5aR in leukocyte trafficking to the brain, and are consistent with the known role of C5a-C5aR-mediated induction of inflammatory cytokines and chemokines (25).

In addition, our observations of endothelial activation and dysfunction following P. berghei ANKA infection are in agreement with accumulating evidence that increased levels of sICAM-1 and Ang-2, and decreased levels of Ang-1 are associated with both adverse clinical outcomes in human malaria infections and progression of ECM in mice (5, 34, 45,–48, 53). Of note, in the same cohort of Ugandan children with CM we studied with increased C5a levels, we also observed significantly increased levels of sICAM-1 and Ang-2 compared to those in children with uncomplicated malaria (34). The Ang-Tie2 pathway plays a pivotal role in the transition from a quiescent endothelium to an activated endothelial phenotype, which in addition to severe malaria is associated with a number of prothrombotic states and inflammatory disorders (44). Infection of C5aR^−/− mice was associated with decreased endothelial activation and the preservation of BBB integrity, supporting the hypothesis that C5a-C5aR signaling contributes to endothelial dysfunction and the loss of BBB integrity that characterize ECM (54, 55). Interestingly, C5a and C5aR have been shown to mediate BBB dysfunction in experimental lupus, and blockade of C5a or C5aR significantly improved BBB dysfunction in this model, suggesting that the role of C5a in vascular leakage in the brain may extend to a variety of disease states (56, 57). Taken together, these data support a mechanistic model whereby C5a-C5aR contributes to the ECM pathogenesis by multiple mechanisms, including endothelial activation and cytokine and chemokine induction, resulting in aberrant leukocyte accumulation in the brain and breakdown of the BBB.

In contrast to C5aR^−/− mice, levels of proinflammatory cytokines did not differ significantly between WT and C5L2^−/− mice after P. berghei ANKA infection, suggesting that inflammation induced by C5a is mediated primarily via C5aR in this model. The role of C5L2 has been examined in animal models of several human diseases, including allergic asthma, lung injury, sepsis, and peritonitis (19, 29, 30, 58). Data from these studies suggest both a positive and a negative role of C5L2 in complement-mediated inflammation, depending on the disease context and the animal models used, and the precise role of C5a-C5L2 signaling in the pathogenesis of complement-mediated diseases remains controversial. In ovalbumin (OVA)-induced allergic asthma, C5L2 plays a positive regulatory role much like C5aR, and C5L2 deficiency is associated with reduced levels of airway hyperresponsiveness, serum IgE, and mucus production (30, 58, 59). Similarly, in the midgrade cecal ligation puncture model of sepsis, deficiency or blockade of C5aR or C5L2 was associated with improved survival (19). In contrast, C5L2 was shown to have an anti-inflammatory role in the lipopolysaccharide (LPS)-induced shock (58) and in an immune complex-mediated lung injury model (29). In the LPS-induced shock model, C5L2 deficiency was associated with increased susceptibility to lethality, and it was associated with increased levels of neutrophil infiltration, serum TNF, and IL-6 in the immune complex-mediated lung injury model (29, 58).

In our hands, C5L2 deficiency did not accelerate disease progression as would be predicted if this receptor played an anti-inflammatory role as suggested by others (26, 29). Alternatively, we also did not observe a proinflammatory role for C5L2 as reported in the cecal-ligation puncture model of sepsis or in an experimental model of allergic asthma (19, 30). Instead, both the median survival and the changes in the inflammatory mediators (including HMGB-1) were similar in WT and C5L2^−/− mice. Lack of a functional role for C5L2 in ECM is consistent with observations in models of experimental placental malaria in which C5aR deficiency and antibody-mediated blockade of C5aR but not C5L2 significantly reduce fetal growth restriction and improve fetal viability (60). Collectively, these data support the view that, in the context of malaria pathogenesis, C5a appears to function predominantly through C5aR rather than C5L2.

We examined whether the C5L2-dependent regulation of HMGB-1, which has been previously observed in experimental sepsis, can be extended to ECM (19). Our data suggest that it does not occur in ECM and that C5L2-dependent regulation of HMGB-1 is context specific. Higgins et al. recently reported that the elevated plasma HMGB-1 levels at presentation were significantly associated with severe malaria and a subsequent fatal outcome in pediatric patients with P. falciparum infection (41). However, antibody-mediated neutralization of HMGB-1 in ECM failed to improve survival (41). These findings suggest that HMGB-1 may be a marker of severe malaria but is unlikely to be a critical mediator of the ECM pathogenesis. Lack of a significant induction observed in the C5aR^−/− mice in our study is consistent with this hypothesis.

Our findings would appear to differ somewhat from those of Ramos et al. who reported that while C5 deletion conferred protection from ECM, the mechanism was via inhibition of C5b and the membrane attack complex (MAC) rather than C5aR (24). In a subsequent study, the authors demonstrated that the terminal complement pathway is activated independently of C5 convertase and that it is likely to be activated by coagulation enzymes of the extrinsic protease pathway (61). Coagulopathy has been consistently associated with falciparum malaria, with patients displaying prolonged bleeding times, prolonged prothrombin and partial prothrombin times, and impaired aggregation responses to defined stimuli (62,–64). Moreover, the complement and coagulation pathways are intimately linked. The coagulation pathway can activate complement pathway and vice versa (15, 65), and both pathways can be activated by parasite components (66).

Nonetheless, if C5 is cleaved, both C5a and C5b would be generated as well as the MAC. In eight independent P. berghei ANKA challenge experiments, we consistently observed a significant survival advantage in C5aR^−/− mice. While our data do not preclude a role for C5b as a potential contributor to ECM, our data indicate that C5a-C5aR may contribute to the pathogenesis of CM via multiple pathways, including dysregulated cytokine induction, endothelial activation, aberrant accumulation of leukocytes in the brain, and ultimately BBB dysfunction. This hypothesis is supported by our data from the present and previous in vitro and in vivo studies (17, 25) and by observations in human CM (5, 8,–10, 17, 34, 43, 45, 46, 52). Furthermore, our observations are consistent with a growing body of compelling evidence implicating a causal role for C5a-C5aR in the induction of proinflammatory mediators and antiangiogenic factors in multiple disease conditions that share features with cerebral malaria, including sepsis, ischemia-reperfusion injuries, pathological angiogenesis, and neurocognitive and neurodegenerative disorders and most recently in studies of human and experimental placental malaria demonstrating a causal role for C5a-C5aR in altering placental angiogenesis and mediating adverse birth outcomes associated with malaria infection (19, 22, 60).

In summary, we provide the first evidence of elevated C5a in human CM and using an experimental model demonstrate a causal role for C5aR but not C5L2 in the enhanced cytokine (e.g., TNF and IFN-γ) and endothelial (e.g., decreased Ang-1, increased Ang-2) activation and BBB dysfunction associated with poor clinical outcomes in human malaria infections.

ACKNOWLEDGMENTS

Footnotes

REFERENCES