C5a Enhances Dysregulated Inflammatory and Angiogenic Responses to Malaria In Vitro: Potential Implications for Placental Malaria

Background Placental malaria (PM) is a leading cause of maternal and infant mortality. Although the accumulation of parasitized erythrocytes (PEs) and monocytes within the placenta is thought to contribute to the pathophysiology of PM, the molecular mechanisms underlying PM remain unclear. Based on the hypothesis that excessive complement activation may contribute to PM, in particular generation of the potent inflammatory peptide C5a, we investigated the role of C5a in the pathogenesis of PM in vitro and in vivo. Methodology and Principal Findings Using primary human monocytes, the interaction between C5a and malaria in vitro was assessed. CSA- and CD36-binding PEs induced activation of C5 in the presence of human serum. Plasmodium falciparum GPI (pfGPI) enhanced C5a receptor expression (CD88) on monocytes, and the co-incubation of monocytes with C5a and pfGPI resulted in the synergistic induction of cytokines (IL-6, TNF, IL-1β, and IL-10), chemokines (IL-8, MCP-1, MIP1α, MIP1β) and the anti-angiogenic factor sFlt-1 in a time and dose-dependent manner. This dysregulated response was abrogated by C5a receptor blockade. To assess the potential role of C5a in PM, C5a plasma levels were measured in malaria-exposed primigravid women in western Kenya. Compared to pregnant women without malaria, C5a levels were significantly elevated in women with PM. Conclusions and Significance These results suggest that C5a may contribute to the pathogenesis of PM by inducing dysregulated inflammatory and angiogenic responses that impair placental function.


Introduction
Placental malaria (PM) is a major determinant of maternal and infant health in the developing world. PM, especially in primigravidae, can have profound maternal and fetal health consequences, including anemia, stillbirth, premature delivery, intrauterine growth restriction (IUGR) and delivery of low birth weight (LBW) infants [1,2]. The accumulation of Plasmodium falciparum parasitized erythrocytes (PEs) within the placenta is believed to be an essential step in the pathogenesis of PM [3]. A subpopulation of PEs that express novel variant surface antigens (VSA-PM), specifically adhere to glycosaminoglycan chondroitin sulfate A (CSA) in the placental intervillous space [3][4][5].
The maintenance of regulated cytokine responses in the placenta is essential in order to prevent rejection of the semiallogenic fetoplacental unit and ensure appropriate immuno-logical responses to infection [17]. This involves a generalized physiological adaptation that results in a bias towards cytokines promoting humoral immunity at the expense of cell-mediated immunity [18]. PM is associated with heightened T H 1 -type responses that disrupt the balance of cytokines at the maternalfetal interface [14,16,19,20]. However, the molecular basis underlying PM-associated cytokine dysregulation, placental immunopathology, and adverse birth outcomes is incompletely understood.
The complement system is an essential component of the innate immune response to a number of infectious agents. The complement cascade can be activated by four distinct pathways, three of which converge at the level of the C3 component, leading to the cleavage of C3 and C5 to their activated forms, C3a and C5a, as well as the formation of the terminal membrane attack complex [21,22]. Recently, a pathway of C5a generation occurring independently of C3 was described [21,23]. Several lines of evidence have implicated excessive activation of the complement system, notably generation of the potent proinflammatory peptide C5a, in mediating deleterious innate host responses to bacterial and fungal infections and contributing to the development of sepsis (reviewed in [22]) [21][22][23][24][25][26]. Sepsis, similar to severe malaria, is a clinical syndrome characterized by systemic inflammation and endothelial activation in response to infection [27,28]. Elevated C5a levels in sepsis have been implicated in adverse clinical outcomes and death. Blockade of C5a-C5aR activity in animal models of sepsis prevents end-organ injury and improves survival [29][30][31].
Excessive C5a generation has also recently been identified as a critical mediator of placental and fetal injury in a non-infectious mouse model of spontaneous miscarriage and IUGR, via dysregulation of angiogenic factors required for normal placental development [32]. Activated complement components were shown to directly induce the release of soluble fms-like tyrosine kinase-1 (sFlt-1, sVEGFR-1), a potent anti-angiogenic factor that prevents signaling of vascular endothelial growth factor and placental growth factor [33]. Excessive levels of sFlt-1 inhibit placental differentiation and are thought to play a direct role in pre-eclampsia [34][35][36] and C5a-associated IUGR and pregnancy loss [32]. Of note, Muehlenbachs et al. have reported elevated plasma levels of sFlt-1 in primigravid women with PM [37].
Based on the hypothesis that C5a may play a role in PM by generating angiogenic factors that mediate IUGR and fetal loss, we determined whether malaria parasites can activate C5 and if C5a potentiates the induction of sFlt-1, inflammatory cytokines and chemokines associated with adverse pregnancy outcomes. Here we show that CSA-and CD36-binding PEs activate C5, that C5a and pfGPI cooperate to induce an amplified inflammatory and anti-angiogenic response to P. falciparum malaria in vitro, and that primigravid women with PM have elevated circulating C5a levels in vivo.

P. falciparum culture
The laboratory isolates CS2 (CSA-binding) and E8B (ICAM-1 and CD36 binding) were cultured in vitro as described [38,39]. Cultures were routinely treated with mycoplasma removal agent (ICN), and were tested and found negative for mycoplasma by PCR.

Mononuclear cell isolation
PBMCs were isolated from healthy volunteers using a Ficoll gradient as previously described [40]. PBMCs were counted using a haemocytometer and cell viability was determined by Trypan blue exclusion test. Cells were plated at a concentration of 2.0610 6 cells/mL (48 well plate) in RPMI 1640 medium supplemented with 10% heat inactivated FBS and gentamycin (R10G). Cells were cultured at 37uC in 5% CO 2 for the time indicated.
Isolation and purification of GPIs from P. falciparum HPLC-purified P. falciparum GPI (pfGPI) was isolated [41], conjugated to gold beads and used at a concentration of 300 ng/ ml as previously described [10]. Unconjugated gold beads alone were added as controls to wells containing media, or C5a (endotoxin-free; Biovision, USA). All preparations of purified pfGPIs were tested for endotoxin by Limulus amebocyte lysate assay prior to use.

Complement assays
PEs, uninfected red blood cells (uRBCs), or media control were incubated in the presence of 30% serum from malaria naïve donors. After 30 min, supernatants were collected and analyzed for C5 activation. C5a levels in culture supernatants and human plasma were measured by ELISA (R&D Systems).

Flow cytometry
PBMCs (2.0610 6 ) were blocked in 20% serum for 20 min at 4uC. Cells were then stained for CD14-APC (eBioscience, 61D3, 100 uL/mL) and CD88-PE or isotype IgG1-PE (BD Biosciences, D53-1473, 150 uL/mL and eBioscience, P3, 150 uL/mL) for 30 min at 4uC. Cells were fixed in 1% paraformaldehyde/PBS overnight and analyzed on a Becton Dickinson FACS Calibur flow cytometer. Data was analyzed using FlowJo software. A minimum of 50,000 events were collected. Positive background PE staining in isotype samples was subtracted from respective CD88-PE stained samples and then reported as specific CD88 staining.

Statistical analysis
Data are expressed as means6SEM, unless otherwise noted. For continuous variables a Student's t test or one-way ANOVA with Bonferonni post-tests was used to compare means. Nonparametric tests (Mann-Whitney or Kruskall-Wallis) were used to compare groups where parametric test conditions were not met. Synergism was determined by the interaction term of a two way ANOVA (C5a*GPI). Synergism in the time course was determined using the cumulative cytokine levels after 48 hours by performing a two way ANOVA and examining the interaction term. Significance in the C5a receptor blockade experiments was assessed by repeated measures ANOVA to determine whether levels of cytokines differed between cells treated with pfGPI alone, or anti-CD88 plus C5a and pfGPI, or student's t-test by comparing the isotype and anti-CD88 co-treated (C5a and pfGPI) groups after correcting for the background levels observed in media controls. In cases where C5a receptor blockade resulted in values of ''0'', a one sample t-test was used to determine whether the isotype C5a and pfGPI co-treated groups differed from a hypothetical value of ''0''. Correlation of peripheral parasitemia to placental parasitemia was determined using Spearman correlation. Experiments were performed in duplicate or triplicate and repeated as indicated in the figure legends. Statistical analysis of experimental replicates is found in the supporting information files (Supporting Information S1, S2, S3, S4). Differences with p,0.05 were considered significant. Statistical analysis was performed using GraphPad Prism and SPSS.

Study participants
Plasma samples were collected at time of delivery from the peripheral blood and placental blood of primigravid women, living in an area of holoendemic malaria transmission in Western Kenya, with and without placental malaria infection as described previously [7]. Malaria infection was evaluated by examination of Giemsa-stained peripheral and placental thick and thin blood smears by light microscopy. Percent parasitemia was estimated by counting a minimum of 1000 erythrocytes and calculated using the following formula: (# infected erythrocytes/total # erythrocytes) 6100. Women who tested positive for HIV were excluded. This study was approved by the institutional review boards of the University of Georgia, the Kenya Medical Research Institute and the Centers for Disease Control and Prevention, Atlanta, Georgia, United States. Written informed consent was obtained from all participants.

CSA-and CD36-binding PEs activate C5 and pfGPI induces C5a receptor expression
The pro-inflammatory anaphylatoxin C5a has been implicated in mediating deleterious host responses to bacterial and fungal agents [22,25]. To confirm that CSA-binding PEs were capable of activating the complement system and thus generating C5a, serum from malaria-naïve donors was added to mature stage CS2 PEs or uRBCS. Compared to uRBCs or controls, C5 activation was only observed in the presence of PEs ( Figure 1A)  In order to respond to activated C5, cells must express a C5a receptor. Previous studies report that C5a responsiveness increases following cellular stimulation [43]. To determine the effect of pfGPI on monocyte C5aR (CD88) expression, PBMCs were incubated as above with pfGPI or media control for 24 or 48 hours, and C5aR and CD14 expression (used to define monocytes) was determined by flow cytometry. Compared to media only control cells, a significant increase in the number of monocytes expressing C5aR ( Figure 1B)

C5a potentiates pfGPI-induced inflammatory responses
C5a has been shown to enhance TLR-mediated pro-inflammatory responses to microbial products such as LPS, contributing to sepsis syndromes [22,25,44,45]. The parasite product pfGPI is recognized by TLR2 and is believed to be a mediator of the inflammatory response that characterizes severe malarial syndromes [10,[44][45][46]. To assess whether C5a would amplify pfGPIinduced pro-inflammatory responses, we treated PBMCs with increasing concentrations of recombinant human C5a (0, 1, 5, 10, 50, and 100 nM) with or without pfGPI for 24 hours and determined the production of IL-6 and TNF by ELISA. We observed a dose-dependent potentiation of pfGPI-induced IL-6 and TNF production by C5a (Figure 2A and B; two-way ANOVA (C5a*GPI), p,0.0001).

C5a Receptor blockade inhibits enhanced cytokine responses
Elevated levels of T H 1 cytokines have been implicated in PM pathogenesis and adverse pregnancy outcomes [14,16,47]. In order to determine the effect of C5a on the kinetics of pfGPIinduced cytokine production, we treated PBMCs with C5a (50 nM) and pfGPI (HPLC-purified, 300 ng/mL) for 48 hours. Supernatants were collected at baseline, 1, 2, 4, 6, 8, 12, 24 and 48 hours and assayed for inflammatory cytokines using a cytometric bead array. To examine the role of C5a-C5aR interactions in the observed inflammatory responses, we also performed parallel experiments including monoclonal antibody blockade of the C5a receptor. pfGPI-stimulated cytokines, including IL-6, TNF, IL-1b and IL-10, were synergistically induced in the presence of C5a ( Figure 3A-D: two-way ANOVA on cumulative cytokine levels (C5a*GPI); p,0.0001 for IL-6, p = 0.0124 for TNF, p = 0.0418 for IL-1b and p,0.0001 for IL-10). A significant reduction in cytokine production was observed in experiments using anti-C5aR blocking antibody compared to its isotype control for all cytokines measured ( Figure 3E-H: repeated measures ANOVA). Further analysis (repeated measures ANOVA) was performed to determine whether C5aR blockade returned cytokine production to levels for cells treated with pfGPI alone. In the presence of anti-C5aR, production of IL-6, TNF, and IL-1b was returned to levels that were not significantly different from pfGPI alone (p.0.05). However, this was not true for IL-10 (p = 0.0101).

C5a is increased in PM and is associated with higher placental parasite burdens
In order to extend our observations to the in vivo setting, we examined C5a levels in malaria-exposed women ( Table 1 and 2). We analyzed C5a levels in the peripheral and placental plasma of HIV negative primigravid women with or without PM from Kisumu, Kenya. Peripheral C5a levels were significantly increased in women with PM ( Figure 5A:

Discussion
This study provides the first evidence implicating excessive activation of the complement component C5 and the induction of anti-angiogenic factors in the pathogenesis of PM. We demonstrate that PEs and parasite products (pfGPI) activated C5 and induced C5a receptor expression (Figure 1). Further, C5a and pfGPI induced synergistic production of inflammatory cytokines, chemokines and an anti-angiogenic factor associated with adverse pregnancy outcomes (Figures 2, 3&4). A role for C5a-C5aR was confirmed by inhibiting inflammation with C5aR blockade ( Figures 3&4). These findings were extended to the in vivo setting, where primigravid women with PM were shown to have elevated C5a levels in both peripheral and placental plasma ( Figure 5).
The generation of C5a is normally tightly controlled and, under physiological conditions, C5a enhances the effector function of macrophages and neutrophils, contributing to effective innate  . C5a potentiates pfGPI-induced inflammatory cytokines and this effect is abrogated by C5a receptor blockade. Human PBMCs were cultured with unconjugated gold beads (media control), unconjugated gold beads and C5a (50 nM), pfGPI (300 ng/mL), or a combination of C5a and pfGPI. (A-D) Supernatants were assayed for IL-6 (A), TNF (B), IL-1b (C), and IL-10 (D). A significant interaction effect between C5a and pfGPI was observed for all cytokines tested. p,0.0001 for IL-6, p = 0.0124 for TNF, p = 0.0418 for IL-1b and p,0.0001 for IL-10 (by two-way ANOVA on cumulative cytokine production over 48 h). Data are representative of three independent experiments. (E-H) Human PBMCs were treated with an a-CD88 mAb (5 mg/mL) to block the C5a receptor or an appropriate isotype control and exposed to either unconjugated gold beads (media control) or a combination of C5a (50 nM) and pfGPI (300 ng/mL) for 48 h. Supernatants were assayed for IL-6 (E), TNF (F), IL-1b (G) and IL-10 (H). Cytokine production was significantly reduced upon C5a receptor blockade; p = 0.0005 for IL-6, p = 0.0028 for TNF, p = 0.0479 for IL-1b and p = 0.0191 for IL-10 (repeated measures ANOVA comparing cytokine levels between isotype control mAb and a-CD88 mAb groups co-treated with C5a and pfGPI). Values shown are corrected for background levels observed in the media controls, and are presented as means6SEM. Data are representative of two independent experiments. doi:10.1371/journal.pone.0004953.g003 responses to microbial pathogens [22]. However, detectable systemic C5a suggests a loss of regulation of complement activation. Elevated C5a levels are associated with a number of deleterious impacts on host innate defence, including defects in phagocyte and endothelial cell function, potentiated chemokine and cytokine secretion to bacterial components such as LPS, and lymphoid apoptosis (reviewed in [22,25]).
There are several mechanisms by which malaria may induce excessive complement activation, thereby contributing to the pathophysiologic mechanisms underlying PM. We observed that mature stage PEs are capable of generating C5a and a bioactive product of parasite rupture, pfGPI can result in an increase in monocytes expressing the C5aR. These observations suggest the potential for an autocrine loop, resulting in augmented inflammatory and angiogenic responses that may act in concert to mediate or amplify placental and fetal injury.
Our observations of malaria-induced C5a are consistent with studies of severe malaria and experimental malaria challenge models that report activation of the complement system [50][51][52]. In experimental human malaria infection, complement activation was observed prior to microscopic detection of parasitemia [50]. Similarly, a recent study of P. berghei ANKA implicated C5a as an early mediator of experimental cerebral malaria [53]. In the latter study, C5a levels were elevated prior to the onset of symptoms suggesting a role for C5a as an initiating or amplifying factor in malaria pathogenesis. Our observations are further supported by the findings of a recent genome-wide analysis of PM, where selected complement genes as well as C5aR were shown to be upregulated [54].
In our in vitro model, C5aR blockade abrogated the potentiated inflammatory responses that are believed to be central to the pathogenesis of PM. However, a different response was observed for IL-10. While IL-10 production was reduced by C5aR blockade, the levels did not return to those observed with pfGPI alone. We speculate that this may be attributable to the expression of the alternate C5a receptor, C5L2, which has been shown to promote T H 2 skewing [57] as well as have inflammatory effects in experimental sepsis [58]. Future studies will be required to investigate the role of pfGPI and other parasite products on the expression of C5L2. When cells were treated with pfGPI and C5a, IL-10 was induced earlier than with pfGPI alone ( Figure 3D). This may have important implications in vivo since TGF-b and IL-10 regulate inflammatory responses. Early IL-10 may impair T H 1 mediated parasite clearance during malaria infection and increased IL-10 levels have been reported in PM [59].
Monocyte infiltration into the placental intervillous space has been associated with poor birth outcomes including the risk of low birth weight infants [15,48,49,60,61]. Chemokines function as specific chemoattractants for leukocytes. IL-8 plays a role in neutrophil and monocyte migration, while the b-chemokines are involved in macrophage migration. High levels of macrophagederived IL-8 during PM have been associated with IUGR [47]. Elevated levels of MIP-1a, MIP-1b and MCP-1 have been observed in PM and it is likely that these chemokines are derived from both maternal (leukocyte) [6,8] and fetal (syncytiotrophoblast) sources [11,62]. As with cytokines, C5a and pfGPI synergistically induced chemokines, and this effect was abrogated for the b-chemokines by blocking the C5aR ( Figure 4A-D). No significant difference for IL-8 production was found between cells treated with anti-C5aR versus an isotype control due to the variation in the latter group. However, there was no difference between cells treated with pfGPI alone or with anti-C5aR plus C5a and pfGPI, indicating that blocking the C5aR inhibits the effect attributable to C5a.
Peak prevalence of malaria infection during pregnancy occurs between 13-20 weeks gestation and gradually falls as gestation increases [63]. This coincides with the second wave of trophoblast invasion of the maternal spiral arteries (16-18 weeks). It is hypothesized that PM may impair transformation of the maternal vasculature leading to placental dysfunction [64,65]. One protein  that may play a role in impaired angiogenesis is the antiangiogenic factor sFlt-1. This alternatively spliced soluble variant of VEGFR-1 binds and sequesters placental growth factor and vascular endothelial growth factor. Excess sFlt-1 has been shown to inhibit placental cytotrophoblast differentiation and invasion [33]. Increased sFlt-1 levels have been implicated in the pathogenesis of placentation associated with pre-eclampsia and IUGR [34][35][36] and have been shown to predict the development of pre-eclampsia [35]. sFlt-1 can be produced extra-placentally by monocytes treated with C5a [32]. Therefore, we determined whether C5a and pfGPI would potentiate sFlt-1 production from mononuclear cells. We show that sFlt-1 was synergistically induced by co-stimulation with these molecules and that this effect was reduced but not consistently inhibited by C5aR blockade ( Figure 4E). We postulate this may be due to different PBMC donors, as FLT1 genotype has been associated with differing sFlt-1 levels in the peripheral blood of primigravid women, as well as PBMC responsiveness to LPS in vitro [66]. Furthermore, the study by Muehlenbachs et al., provides the first evidence that both fetal and maternal genotype may contribute to immune responses to PM [66]. Taken together, these data suggest a putative mechanism for impaired angiogenesis and placental dysfunction during PM. In order to extend these observations to human infection, we measured C5a levels in the peripheral blood of a cohort of malaria-exposed primigravid Kenyan women and we measured C5a levels in the placental blood of a similar cohort of malariaexposed primigravid women at time of delivery ( Table 1-2) Significantly increased levels of C5a were found in peripheral blood ( Figure 5A) and the placental blood ( Figure 5B) of pregnant women with PM compared to those without. These results provide the first evidence that C5a levels are elevated in women with PM. However, these findings will need to be confirmed in larger prospective clinical studies. It will be of interest to also examine how C5a levels change with respect to other disease parameters, including parasitemia and mononuclear cell infiltrates in larger populations. Figure 6. Proposed mechanism for C5a-mediated dysregulation of the placental environment in PM. Mature schizonts activate C5 and rupture releasing parasite components containing GPI that induce expression of C5aR and activate macrophages. The concomitant expression of C5aR and generation of C5a represent a potential autocrine mechanism by which C5a can amplify inflammation in P. falciparum malaria resulting in the excessive cytokine and chemokine responses that characterize PM. The dysregulated production of cytokines, chemokines and the antiangiogenic factor sFlt-1 contribute to placental dysfunction, intrauterine growth restriction and low birth weight infants. doi:10.1371/journal.pone.0004953.g006 In summary, we have demonstrated a role for C5a in the enhanced cytokine and chemokine responses that characterize PM and in the induction of sFlt-1, an anti-angiogenic factor associated with abnormal placental development and poor birth outcomes ( Figure 6). Our findings that disruption of the C5a-C5aR interaction inhibits inflammatory and anti-angiogenic responses to malaria suggest anti-C5a-C5aR strategies as potential therapeutic approaches for PM [67].

Supporting Information
Supporting Information S1 Experimental replicates and statistics Figure 1A