The aim of this study was to identify predictive factors of death or survival during CM. To this end, 70 Beninese children with CM (50 who survived and 20 who died) were included in the present study, which is part of a larger project named NeuroCM [24]. An accurate differential diagnosis was achieved through thorough clinical examination, including fundoscopic examination and extensive blood biochemical analysis associated with the molecular diagnosis of microbial infections in the patients' cerebrospinal fluid. Fundoscopic examination was difficult to perform and could not be performed on the children with the most severe conditions; the data are rather indicative. This approach resulted in a homogeneous group of CM patients thanks to the exclusion of coinfections and nonmalarial comas. In a previous paper dealing with the NeuroCM study, a multivariate analysis revealed that high bilirubin (jaundice) and lactate levels were predictive of early death in Beninese children with nontraumatic coma, whereas the use of antibiotics before admission and yellow fever vaccination were protective factors [23]. In our study, similar associations were found in the univariate analysis, except for vaccination against yellow fever. The similar results obtained for jaundice, hyperlactatemia and previous antibiotic use suggest that such a clinical profile is more specific to coma, including CM-related coma, than to CM itself. Interestingly, a lower Blantyre score and splenomegaly were also identified as risk factors for CM death. A more pronounced coma was previously described as a risk factor for death during severe malaria in children [28], in adults presenting with CM [29], and in children presenting with CM [30, 31]. Splenomegaly is also a clinical manifestation classically described in cerebral malaria [32, 33], although rarely identified as a risk factor for death during CM. Through the analysis of haematologic factors, we observed higher numbers of leukocytes, among which lymphocytes and monocytes were associated with CM death, indicating an increased immune response related to the infection. Higher creatinine levels were also associated with death in CM children, suggesting that renal damage was involved in the death of patients.
We then analysed death risk factors among specific plasma and immune mediators of both inflammatory and oxidative responses. In plasma, we measured the levels of cytokines and other soluble factors usually associated with malaria severity (the growth factor angiopoietin-2, the enzyme granzyme B, and two soluble endothelial factors ICAM-1 and EPCR). We also measured lipid mediators from arachidonic acid metabolism in urine. We found by univariate analysis that higher plasma levels of TNF, IL-8, IL-1β, IL-10, CXCL9, granzyme B and angiopoietin-2 measured at the acute phase (D0) were associated with CM death. For urinary mediators, lower levels of PGEM at D0 were associated with CM death through univariate analysis. However, a higher initial level of plasma IL-8 alone was identified as a risk factor for death by multivariate analysis. Plasma and urinary levels measured during convalescence (D3 and D30) confirmed these trends and suggested two kinds of marker profiles: a disease severity profile for decreasing markers including plasma TNF, IL-8, IL-10, CXCL9, granzyme B, angiopoietin-2, along with plasma IL-6, CXCL10, CCL2, CCL3, CCL4, ICAM-1, and urinary creatinine, GSH and GSH:GSSG ratio; and a disease resolution profile for increasing markers including plasma CXCL5, CCL17, CCL22, and urinary PGEM, 15-F2t-isoprostane, LTB4, LXA4, and GSSG.
In line with our results, previous studies have reported associations between malaria severity or death and proinflammatory cytokines and the anti-inflammatory cytokine IL-10 [34–40]. During CM, proinflammatory cytokines such as TNF and IL-1β are thought to induce endothelial activation, as demonstrated in vitro [41, 42], and to play a major role in iRBC sequestration in cerebral microvessels [14, 43] through overexpression of adhesion molecules (ICAM-1, VCAM), accumulation of cytotoxic CD8 + T cells and alteration of the BBB [14, 44]. An increase in plasma IL-1β along with IL-10 and TNF has also previously been associated with CM outcome [37]. However, the implication of IL-10, an anti-inflammatory cytokine involved in the inhibition of proinflammatory responses and the development of regulatory responses, is discussed more. It has been shown in mice that the absence of IL-10 expression was associated with the development of severe malaria and excess mortality, suggesting a protective role [45, 46], whereas high plasma IL-10 levels in humans were correlated with malaria severity and death [14, 39, 47]. Our results are in line with most human studies suggesting a role for high plasma IL-10 in malaria severity and CM death.
Among the plasma markers measured, high levels of IL-8 constituted the most significant risk of death (odds ratio 14.2) in our study. Other studies also reported a significant association of high levels of plasma or serum IL-8 with severity or death among severe cases (severe malarial anaemia or cerebral malaria) [40, 48]. Concerning more specifically CM, Pappa et al. showed a positive and significant correlation between plasma IL-8 levels and brain volume in CM children [49]. Such a correlation was also found for TNF, CCL2 and IL-10, in line with our results. Interestingly, elevated cerebrospinal IL-8 levels, but not serum IL-8 levels, were identified as a marker of CM death among children who died of CM, SMA, or nonmalaria causes [50]. However, in this study, samples were collected after death. Finally, a recent study reported higher levels of IL-8 in retinopathy-positive CM children who died than in survivors and UM children, similar to our results [51]. The authors claim that retinopathy diagnosis improves clinical CM specificity, as it is related to iRBC sequestration in brain. In our case, we think that our diagnosis strategy was sufficiently thorough to ensure the selection of real CM cases. IL-8 is a chemokine whose main function is the promotion of neutrophil activation and migration to sites of inflammation; activated neutrophils also produce IL-8 themselves, reinforcing the inflammatory loop [13]. During CM, neutrophils can play both a beneficial role by killing blood-stage parasites [52] but also a deleterious role through ROS production and toxic mediator release, leading to an alteration of the brain endothelium and increased iRBC sequestration [53–55]. Although ECM studies report leucocyte accumulation in the brain [56], autopsies carried out in humans after CM death rarely reported neutrophil accumulation in brains [57–59], suggesting a putative effect of neutrophils through their activation and ability to produce soluble factors rather than their migration to inflammation sites. Thus, high IL-8 levels may mostly participate in neutrophil activation. In addition, a disease resolution profile was obtained for CXCL5, a chemokine implicated in neutrophil trafficking [13], with increasing levels from D0 to D30 in surviving children. This results reinforce our hypothesis of the importance of neutrophil activation rather than migration to inflammation sites. Interestingly, IL-8 has also been identified as a pharmacological target to reduce ischaemia-induced myocardial injury [60], suggesting that decreasing or neutralizing plasma IL-8 could help reduce neuroinflammation and ischaemia during CM.
In our study, among chemokines other than IL-8, only the level of plasma CXCL9 was related to CM death. This is a surprising result, as high levels of CXCL10 and CCL2 have been frequently reported as markers of severity and death during malaria and even CM [36, 39, 61, 62]. For CXCL10, the low number of doses carried out successfully may explain our result. To our knowledge, this is the first time that a study highlights a correlation between CXCL9 and death in CM patients. However, its deleterious implication during ECM has already been demonstrated [63], and an increase in its gene expression has been observed in the brains of infected mice [64]. CXCL9 is known for its involvement in CD8 T-cell migration to the brain and in their activation leading to the release of granzyme B and brain endothelium alteration, as demonstrated in ECM [27, 65]. Two recent studies have highlighted the involvement of CD8 T cells in human CM through their localization in the brains of Malawian children who died of CM [66, 67]. Kaminski et al. also observed higher plasma levels of granzyme B during malaria infection compared to healthy patients and an increased proportion of blood CD8+/GrzB+ T cells in severe malaria patients [68]. Such results combined with ours support the idea that CXCL9 and granzyme B are players in the neuroinflammatory response during CM. Conversely, CCL17 and CCL2 which are known for their implication in Th2 response and regulatory T cell migration [13], were found to be associated with disease resolution in surviving children.
Elevated plasma angiopoietin-2 and soluble ICAM-1 are two markers of endothelium activation and increased permeability that were previously associated with malaria severity and risk of cognitive impairment [69, 70]. In our study, elevated levels of angiopoietin-2, a marker of increased vascular permeability, were also found in patients who died compared to those who survived. However, the levels of plasma soluble ICAM-1 did not differ between groups, suggesting that the level of angiopoietin-2 is a better prognostic marker of death than the level of soluble ICAM-1.
In urine, we measured some metabolites of arachidonic acid known for their implication in the inflammatory response. Among the metabolites measured, only PGEM was related to CM death with a disease resolution profile, i.e., with lower initial urinary levels in deceased subjects than in surviving subjects. PGEM is the major metabolite of PGE2, which is unstable in vivo and produced via the activation of the enzyme COX-2 [71]. Other teams previously reported lower levels of urinary bicyclo-PGE2, another PGE2 metabolite, in severe malarial anaemia patients and CM patients than in those presenting uncomplicated or asymptomatic malaria [72, 73]. These authors also demonstrated that reduced PGE2 levels were related to downregulation of COX-2 and enhanced uptake of haemozoin by monocytes. In our study, the large number of patients (70 CM subjects in total, of which 20 died and 50 survived) probably helped in identifying a significant relationship between the occurrence of death during CM and low levels of urinary PGEM. In addition, it is important to emphasize that the urinary determination of biological parameters in CM children is an interesting approach because of its ease of implementation and its noninvasive character.
Finally, the evolution of 15-F2t-isoprostane levels and the GSH:GSSG ratio in urine during follow-up of CM children in our study reflects a more pronounced antioxidant balance at D0 and a re-establishment of a more prooxidant response over the course of the follow-up among CM children who survived. However, these lower levels observed at inclusion were not related to death outcome. This result is surprising, as oxidative damage is thought to contribute to malaria pathogenicity [74]. Additionally, higher levels of urinary F2-isoprostane metabolites were previously reported in patients presenting severe malaria and acute kidney injury than in patients with nonsevere malaria [75, 76]. However, our study is the first to analyse the evolution of oxidative markers in urine among CM children. A defect in the oxidative response by the leukocytes of these children could explain these results and provide a new avenue for research.