Serologic and Cytokine Profiles of Children with Concurrent Cerebral Malaria and Severe Malarial Anemia Are Distinct from Other Subtypes of Severe Malaria

ABSTRACT.

We used a protein microarray featuring Plasmodium falciparum field variants of a merozoite surface antigen to examine malaria exposure in Malian children with different severe malaria syndromes. Unlike children with cerebral malaria alone or severe malarial anemia alone, those with concurrent cerebral malaria and severe malarial anemia had serologic responses demonstrating a broader prior parasite exposure pattern than matched controls with uncomplicated disease. Comparison of levels of malaria-related cytokines revealed that children with the concurrent phenotype had elevated levels of interleukin (IL)-6, IL-8, and IL-10. Our results suggest that the pathophysiology of this severe subtype is unique and merits further investigation.

INTRODUCTION

In 2020, there were 241 million cases of malaria and 627,000 malaria-related deaths globally, 96% of which occurred in Africa.¹ Most cases of malaria present as uncomplicated disease, whereas most malarial deaths are caused by three distinct subtypes of severe malarial disease: cerebral malaria (CM), severe malarial anemia (SMA), and respiratory distress.² In sub-Saharan Africa, children younger than 5 years account for 80% of malaria deaths.¹ A previous study estimated that the mortality of the concurrent CM and SMA phenotype in this population may be similar to that of cerebral malaria alone.² In central Mali, a highly endemic region for Plasmodium falciparum malaria, the average age of a child with severe malaria is 38.7 months, and the average age of patients with uncomplicated malaria is 10 years.^3,4 One hypothesis for this discrepancy proposes that older children are more likely to have uncomplicated disease because natural immunity to P. falciparum infection develops through cumulative acquisition of antibodies to parasite antigens, with severe malaria associated with a limited set of these antigens, resulting in more rapid acquisition of protection against severe disease than uncomplicated disease. We have previously found that sera from Malian children with severe malarial anemia or cerebral malaria recognized fewer variant surface antigens than those with uncomplicated malaria, suggesting less prior exposure to these antigens.^5,6

In this study, we hypothesized that children with concurrent CM and SMA represent a unique immunological phenotype. To determine how immune responses in children with CM and SMA differ from other severe malaria subtypes, we examined serologic responses to apical membrane antigen-1 (AMA1). AMA1 is a merozoite surface antigen that plays a critical role in parasite invasion of host erythrocytes, and it serves as a surrogate for P. falciparum exposure. AMA1 is a focus of the host immune response and a marker of the extent of exposure to P. falciparum.⁷ Using a protein microarray populated with 268 AMA1 variants, we measured AMA1 reactivity of sera for 72 Malian children with severe malaria and compared it with that for matched controls with uncomplicated malaria. We also present analysis of a subset of previously published data describing serum levels of cytokines interleukin (IL)-1β, IL-6, IL-8, IL-10, IL-12 and tumor necrosis factor (TNF)-α in our study population.

METHODS

Study population.

We recruited Malian children with severe malaria and age-, region-, and ethnicity-matched controls with uncomplicated malaria in the central Malian town of Bandiagara from 1999 to 2003. Controls were enrolled from the walk-in clinic of the Bandiagara Malaria Project. The study population, clinical assessments, sample acquisition and processing have been described elsewhere ( Supplemental Figure 1).⁸ Study participants received standard malaria treatment regimens at the time of study, including intravenous quinine and intramuscular pyrimethamine-sulfadoxine for severe malaria and chloroquine for uncomplicated malaria. Blood transfusions were not available in this remote rural setting at the time. Here, we present a chi-square analysis of mortality and t-test analysis of parasitemia stratified by severe malaria subtype from this cohort.

We classified cases based on modified WHO criteria.^1,8 We defined uncomplicated malaria as the presence of parasitemia and axillary temperature of ≥ 37.5°C and/or symptoms leading to treatment-seeking behavior.¹ CM was defined as parasitemia and Blantyre Coma Score (BCS) ≤ 2 with no other obvious cause of coma. SMA was defined as parasitemia and hemoglobin ≤ 5 g/dL. Concurrent CM and SMA cases had parasitemia, BCS ≤ 2, and hemoglobin ≤ 5 g/dL. Study participants with severe respiratory distress were not included in this study, as we were unable to exclude pneumonia as the underlying cause. All participants had a serum glucose > 40 mg/dL, and none of the participants were jaundiced nor had respiratory distress at enrollment.

Serum samples were obtained from participants during acute illness, before initiation of treatment. Uncomplicated malaria cases from Bandiagara, Mali, served as matched controls for severe malaria in the microarray study. Sera from 11 malaria-naive North American adult blood donors served as negative controls for the microarray.

Microarray construction and analysis.

We used a protein microarray populated with 268 full-length AMA1 antigen variants to measure AMA1 reactivity of sera. Construction of this microarray has been previously described elsewhere.^7,9,10 Fluorescence intensity was defined as the raw signal intensity reduced by the mean for the no-DNA negative controls. Positive serorecognition for an individual sample was defined as two standard deviations above the mean fluorescence intensity of North American controls. Serorecognition of the 268 AMA1 antigens for each participant was assessed. Differences in serorecognition between severe cases and uncomplicated controls were compared among participant groups using a non-parametric Wilcoxon signed-rank test for matched group-wise comparisons. Differences in seroreactivity between severe malaria subtypes and matched uncomplicated controls were assessed using a Wilcoxon signed-rank test. If statistically significant differences were found, groups were considered more or less seroreactive based on mean comparison. Direct comparisons between the severe malaria subtypes were performed using the Wilcoxon rank-sum test.

Presented P values are two-sided without correction for multiple comparisons using an alpha of 0.05, per previously described approaches for microarray analyses.^6,10–13 Statistical analyses were performed using R (version 3.4.3), Stata (version 11.4), and GraphPad (version 5.04).

Cytokine analyses.

Our study population is a subset of a previously published cohort, which assessed cytokine levels of IL-1β, IL-6, IL-8, IL-10, IL-12, and TNF-α, via cytometric bead array (BD Biosciences, Franklin Lakes, NJ) and flow cytometry.⁸ Here we present a subset of those cytokine results for the matched samples we used in array analyses. Using a Wilcoxon-signed rank test for matched pairs, we compared cytokine levels for children with different severe malaria syndromes and their matched controls with uncomplicated malaria. Analyses were based on geometric means given in picograms per milliliter. Using a Kolmogorov-Smirnov test for unmatched pairs, we also directly compared the cytokine levels for different severe malaria syndromes.

Ethics statement.

The protocols were approved by institutional review boards of the University of Bamako Faculty of Medicine, the University of Maryland, Baltimore, and the U.S. Army Surgeon General. Written informed consent was obtained before screening and enrollment in the trials. Verbal consent of illiterate parents or guardians was provided and documented using thumbprints and verified by independent witnesses.

RESULTS

Study population.

The previously described cohort from Lyke et al. reported five deaths among 13 cases (38.5%) of concurrent CM and SMA.⁸ The CM alone population had six deaths among 45 cases (13.3%). The SMA alone population had six deaths among 24 cases (25.0%). This mortality difference between concurrent CM and SMA versus CM alone was statistically significant (P = 0.042). There was no statistically significant difference when SMA was compared with CM alone and to concurrent CM and SMA (P = 0.22 and 0.39, respectively). Although cases of concurrent CM and SMA had a lower average parasitemia than CM alone and SMA alone, these parasitemia differences were not statistically significant (P = 0.35 for both comparisons) ( Supplemental Table 1).

We performed microarray analyses on 72 cases of severe malaria and 72 matched uncomplicated controls. Of the severe malaria cases, 42 children had CM, 21 had SMA, and nine had concurrent CM and SMA. Within this subset, parasitemia levels were significantly higher in the severe cases compared with uncomplicated controls. Summary statistics for the study population are provided in Supplementary Tables 2 and 3. One North American, malaria-naive control was removed from the analysis as an outlier.

Cytokine expression analyses included the 72 severe malaria samples, along with six additional samples that lacked matched controls. This included 43 CM cases, 24 SMA cases, and 11 cases of children who had concurrent CM and SMA.

Serorecognition.

The number of serorecognized AMA1 variants did not significantly differ between cases and controls for any comparison. Sera from children with CM alone recognized on average 87.8% (235.2 ± 58.0) of AMA1 variants studied per child, whereas their matched pairs without severe disease recognized 90.4% (242 ± 47.9) (P = 0.54). Sera from children with SMA alone recognized on average 92.1% (246.9 ± 33.9) of antigens per child, whereas their matched pairs without severe disease recognized 90.1% (241.4 ± 53.3) (P = 0.65). Sera from children with concurrent CM and SMA recognized an average 97.5% of antigens (261.3 ± 6.3) per child, compared with 94.5% (253.4 ± 28.1) of antigens for matched counterparts with uncomplicated disease (P = 0.99).

Seroreactivity.

Sera of CM cases (N = 42) reacted less intensely to 39 of 268 AMA1 antigen variants than matched controls and did not react more intensely for any of the AMA1 variants (Figure 1A). Sera of SMA cases (N = 21) reacted less intensely to two of 268 AMA1 antigen variants than matched controls and did not react more intensely for any of the AMA1 variants (Figure 1B). In contrast, sera of concurrent CM and SMA cases (N = 9) did not react less intensely to any AMA1 antigen variants than matched controls and reacted more intensely to 50 of 268 AMA1 variants (Figure 1C).

Figure 1.

Seroreactivity to apical membrane antigen-1 (AMA1) variants in children with severe malaria as compared with matched controls with uncomplicated disease. Median immunofluorescence (blue bars) to AMA1 variants in (A) cerebral malaria (CM), (B) severe malarial anemia (SMA), and (C) concurrent CM and SMA. Uncomplicated control median immunofluorescence intensity was sorted from largest to smallest for each severe malaria subtype (solid orange line). Log of Wilcoxon signed-rank test P value on right-hand axis, with P value denoted with black bars if < 0.1. This figure appears in color at www.ajtmh.org.

Seroreactivity of concurrent CM and SMA cases was significantly higher than that of CM alone for 80.5% (216/268) of AMA1 variants, with no significant difference for the remaining 52 variants. Seroreactivity of concurrent CM and SMA cases was significantly higher than that of SMA alone for 40.3% (108 of 268) of AMA1 variants and was not significantly different for the remaining 160 variants. Seroreactivity of SMA cases was significantly higher than that of CM cases for 28.4% (76/268) of AMA1 variants and was not significantly different for the remaining 192 variants.

Cytokine data in severe versus uncomplicated disease comparisons.

Children with CM had significantly higher levels of IL-6 (P = 8.13E-7) and IL-10 (P = 2.05E-5) when compared with matched children with uncomplicated malaria ( Supplemental Figure 1). Children with SMA had significantly higher levels of IL-10 (P = 3.60E-3) compared with matched children with uncomplicated malaria. Children with concurrent CM and SMA had higher levels of IL-6 (P = 0.025), IL-10 (P = 0.019), and IL-8 (P = 0.014) compared with matched children with uncomplicated malaria ( Supplemental Figure 2).

Cytokine data in severe malaria comparisons.

Children with concurrent CM and SMA had significantly higher levels of IL-6 (P = 0.018) and IL-10 (P = 0.037) than children with severe malarial anemia alone (Figure 2A). Children with concurrent CM and SMA had significantly lower levels of IL-1β (P = 0.042) and IL-12p70 (P = 2.70E-3) than children with CM alone (Figure 2B, Supplemental Table 4).

Figure 2.

Serum levels of cytokines in severe Plasmodium falciparum malaria cases. Geometric mean of serum levels of interleukin (IL)-6 and IL-10 (A) and IL-1β, IL-8, IL-12, and tumor necrosis factor (TNF)-α in panel (B) in concurrent cerebral malaria and severe malarial anemia, cerebral malaria alone, and severe malarial anemia alone. CM = cerebral malaria; SMA = severe malarial anemia. * P < 0.05 in Kolmogorov-Smirnov test. This figure appears in color at www.ajtmh.org.

DISCUSSION

This study found that children with concurrent cerebral malaria and severe malarial anemia had a distinct immunological profile compared with children with severe malarial anemia or cerebral malaria alone. Children with concurrent CM and SMA had a significantly higher seroreactivity to subsets of parasite AMA1 variants than matched controls with uncomplicated disease, suggesting more prior exposure to P. falciparum.¹⁰ The seroreactivity differences to particular AMA1 variants observed may reflect subtle differences in exposure to P. falciparum variants. The concurrent CM and SMA population had a higher mortality than CM alone and had lower average parasitemia than each severe phenotype alone, although the parasitemia differences were not statistically significant.

Children with CM or SMA alone had weaker seroreactivity to AMA1 variants than controls with uncomplicated disease, suggesting that they had fewer P. falciparum exposures. The pathogenesis of these severe phenotypes appears to be related to a lack of antigen recognition, which could be related to a lower amount of exposure and thus more vulnerability to severe disease. This finding is congruent with the hypothesized development of natural immunity over years of exposure through the gradual acquisition of allele-specific immune responses against diverse variants of parasite antigens.¹⁴ Previous work by our group has demonstrated that adults from Mali had significantly higher magnitude and breadth of seroreactivity to AMA1 antigens than children.¹⁰

Direct comparisons between severe malaria subtypes found that the concurrent CM and SMA phenotype had significantly higher seroreactivity than SMA alone and CM alone for 40% and 81% of the tested AMA1 probes, respectively. None of the tested probes were more reactive in SMA alone and CM alone compared with the concurrent phenotype. This likely means that the cases with the concurrent phenotype have been previously exposed to more P. falciparum variants than CM alone and SMA alone and that the pathophysiology of concurrent CM and SMA is not due to low allele-specific immune response to various parasite antigens.

There was no significant difference in AMA1 serorecognition between cases of CM alone, SMA alone, concurrent CM and SMA, and their respective matched controls. This likely reflects that there was sufficient exposure to P. falciparum in each severe malaria subtype and matched uncomplicated malaria group such that AMA1 variant serorecognition thresholds could not differentiate between cases and controls.

To further investigate the immunological response associated with the various clinical presentations of severe malaria, we studied serum levels of cytokines previously associated with the clinical trajectory of P. falciparum infection.⁸ In our cohort, children with concurrent cerebral malaria and severe malarial anemia had higher levels of IL-8 than uncomplicated cases, an association that was not present when comparing either syndrome alone with uncomplicated controls. We did not observe a significant difference in IL-8 levels when directly comparing unmatched concurrent CM and SMA to CM alone or SMA alone. IL-8 is a proinflammatory cytokine that plays a role in neutrophil recruitment and stimulation of phagocytosis, and it has been proposed as a key mediator of excessive inflammatory responses in P. falciparum infections.¹⁵ Increased IL-8 levels have also previously been associated with noncerebral severe malaria in a study from Thailand.¹⁶ When directly comparing immune response in concurrent CM and SMA cases to SMA alone, we noted that children with concurrent CM and SMA had higher levels of IL-6 and IL-10. When comparing concurrent CM and SMA to CM alone, the concurrent phenotype had lower IL-1β and IL-12(p70) levels. This suggests that the concurrent phenotype is not the result of a generalized inflammatory response but is instead associated with a more targeted/specific inflammatory pathway. It is also possible that the inflammatory response in the concurrent CM and SMA group is driven by cytokines that were not assayed in our analysis. Further investigation of host genetics may also shed light on what underlies these differences in immune response and seroreactivity.

One of the limitations of our study is the small sample size of the concurrent CM and SMA phenotype sample. The proportions of severe malaria subtypes were largely representative of their relative incidence in sub-Saharan Africa, and early intervention at disease onset has reduced the number of severe cases that present with both phenotypes.² Our future goals include performing a larger case–control study focusing on this patient population to validate our findings and further characterize their immunological response. We will also aim to include an assessment of P. falciparum expression to assess whether seroreactivity to AMA1 variation is associated with other parasite specific factors. We did not examine P. falciparum genetic variation in our samples, so it is possible that the clinical phenotype of concurrent CM and SMA was caused by a unique set of parasite variants.

These findings suggest that the concurrent CM and SMA phenotype is immunologically distinct from the response in CM or SMA alone. Our results could have therapeutic implications because severe malaria has been perceived as being a result of inadequate immune recognition, whereas the increased seroreactivity in the concurrent CM and SMA patients implies a unique inflammatory pathway that warrants further study to determine whether a tailored approach to therapy and vaccine targets is necessary.

Note: Supplemental materials appear at www.ajtmh.org.