Background
Antibody immunity to
Plasmodium falciparum malaria is central to the prevention and clearance of symptomatic disease [
1], but is acquired only after years of exposure [
2]. Immunity to malaria is non-sterile, and older children and adults in malaria-endemic areas are frequently parasitaemic, but asymptomatic. Children in sub-Saharan Africa carry the burden of morbidity and mortality associated with
P. falciparum infection [
3]. It is still unclear why some parasitaemic children develop severe malarial syndromes, including cerebral malaria (CM).
Although not all studies agree [
4], prior studies have reported ‘protective’ antibody responses that differentiate asymptomatic cases from symptomatic malaria [
5‐
7]. Most studies have found few differences in antibody profiles along the symptomatic disease spectrum [
8‐
10], but tested a limited number of antigens. Newer technologies, including protein microarrays, provide an alternative platform for large-scale evaluation of seroreactivity to thousands of antigens [
11]. An initial large
P. falciparum proteome-wide array encompassed approximately a quarter of the inferred proteome (2320 proteins encoded by 1204 genes) and identified 491 immunoreactive proteins recognized by sera from a cohort in Mali [
12]. A subsequent 824 feature array representing 699 Pf genes was down-selected on the basis of initial array data [
5].
In CM,
P. falciparum erythrocyte membrane protein 1 (PfEMP1), a parasite variant surface antigen (VSA) encoded by ~ 60
var genes, is critical for cerebral sequestration [
13,
14]. Since antibodies to VSA correlate to protection from symptomatic malaria [
10], children who progress to CM may lack antibodies targeting antigens essential for microvasculature sequestration. Each PfEMP1 protein is comprised of multiple extracellular Duffy-binding like (DBL) and cysteine-rich interdomain region (CIDR) domains that are classified into sequence types (e.g. α, β, γ) and subtypes (e.g. CIDRα1) [
15,
16]. The N-terminal head structure (DBL-CIDR) of each PfEMP1 protein is a determinant of PfEMP1-host binding specificity. CIDRα2–6 domains encode CD36 binding properties, and CIDRα1 domains encode endothelial protein C receptor (EPCR) binding properties. CIDRβ/γ/δ domains have unknown binding properties but have been associated with rosetting, which occurs when parasitized red blood cells (pRBCs) bind and aggregate uninfected RBCs [
17]. Studies of paediatric malaria implicate EPCR-binding parasites in severe disease [
18‐
25]. Breadth and magnitude of PfEMP1 seroreactivity correlates with age and exposure [
26,
27], and antibodies to EPCR-binding CIDRα domains are more abundant than antibodies to other CIDR domains and are likely acquired early in life [
28].
Here, partial proteome microarrays were used to identify and characterize differences in global antibody level, breadth and magnitude of
P. falciparum antibody responses, and magnitude of PfEMP1 specific antibody responses between children with uncomplicated malaria (UM) and children with stringently defined CM (Ret + CM; WHO definition + malarial retinopathy) in acute infection and at 30 days of convalescence. A new ~ 1000 feature partial proteome array based upon the 3D7
P. falciparum reference genome that prioritized antigens with relevance to diagnostics and human anti-malarial immunity was used. The antigens represented on this array include vaccine candidates as well as proteins that have been associated with either exposure to
P. falciparum or protection from clinical disease. The study results indicate that children with UM and CM have broad seroreactivity to the panel of ~ 1000
P. falciparum antigens and have lowest seroreactivity during acute disease to EPCR-binding “virulent” PfEMP1 antigens represented on the array [
17,
25]. Despite exposure to “virulent” PfEMP1 during acute disease, antibody responses to the corresponding antigen in convalescence were not detected using this platform that was generated based upon the reference 3D7 genome.
Discussion
Differences in antibody repertoire between UM and CM cases could explain why some children with
P. falciparum infection progress to severe disease. Despite evidence of general suppression of antibody response in acute Ret + CM (Fig.
1), similar robust reactivity to a broad range of
P. falciparum antigens was seen in UM and CM cases alike (Fig.
2). Reactivity to unfractionated malaria antigen and specific antigens of interest (e.g. MSP-1, CSP) by EIA or ELISA in children from the Gambia showed that evidence of prior exposure did not prevent progression to severe disease [
45]. Here, a large proportion of the
P. falciparum proteome was assayed and no differences in the breadth or magnitude of
P. falciparum reactivity by case severity were found (UM vs. Ret + CM; Fig.
2). The antigens on the array were selected based upon their interest as vaccine or diagnostic candidates as well as seroreactivity in both natural and experimental infection and enabled measurement of antibody responses during acute disease and in convalescence. While it cannot be definitively ascertained whether antibody present during acute disease reflects current infection or prior infection, a similar pattern in IgM reactivity that is classically indicative of acute infection was observed in the UM and CM groups across both time points, suggesting that the level of prior exposure and generation of antibody to
P. falciparum antigens in Ret + CM cases parallels that of children with UM.
In a study employing an 824 feature
P. falciparum protein microarray, antibody to a subset of antigens (including MSP1, MSP2, LSA1, LSA3, Pf70, and PfEMP1; Table
2) was associated with protection from symptomatic disease in sub-Saharan African children [
5]. Many of these antigens are current or previous vaccine candidates [
42,
46‐
48] and have inconsistently been reported as protective in the literature [
4,
7,
49,
50]. Independent of disease severity, high seroreactivity to this group of antigens was identified in children with symptomatic malaria (Table
2). Thus, high seroreactivity is common across children living in endemic regions, such as Malawi, and may be a marker of prior (or repeated)
P. falciparum exposure. Dent et al. followed a cohort longitudinally, monitoring seroreactivity pre- and post- infection [
5], as did Crompton et al. [
12]. Longitudinal studies suggest that the antibody response of younger children to malaria antigens including PfEMP1 is short-lived [
12,
26]. The antibody response to
P. falciparum appears to increase with age, and it is not yet clear what factors promote antibody stability and which antibodies are important for the prevention of symptomatic disease versus for prevention of progression to severe manifestations such as CM. Taken together with the results presented here, the data suggests that antibody against some of these targets may be more important for the prevention of symptomatic disease than for the progression to severe manifestations, such as CM.
With regard to PfEMP1 specific antigens, and in agreement with previous studies [
5] [
27], magnitude of reactivity to the highly conserved ATS proteins was greatest. As the ATS is an intracellular domain of PfEMP1, reactivity to this region should be regarded as a surrogate of exposure rather than protection. The CIDR domains are the most hypervariant of all PfEMP1 domains/regions, and seroreactivity to CIDR domains in the patient population described here was highly variable across sequence type (i.e. α/β/γ) and subtype (e.g. CIDRα1) (Fig.
3 and Additional file
7). Within the subset of CIDR domains, reactivity to PfEMP1 domains associated with EPCR-binding was lowest and reactivity to those predicted to bind CD36 was highest [
43]. The breadth and magnitude of response to all EPCR-binding CIDR domains on the microarray was lower in acute infection than that to CD36-binding or rosetting CIDR domains (Fig.
3d). It is important to note that while the strongest seroreactivity signal observed corresponded to CD36-binding CIDR domains, there are more probes that encode this binding property on the array (45 vs. 5 EPCR-binding and 2 rosetting) yielding a higher likelihood of measuring a maximum response as a central measure of IgG reactivity. Abundance of features does not fully explain our results, as there are over twice as many EPCR-binding antigens on the array as rosetting antigens, yet the seroreactivity to EPCR-binding antigens remains lowest. The PfEMP1 antigens on the Pf1000 array that were annotated for this study contain two adjacent domains: DBL-CIDR head structure (N-terminus) domains or in the case of ICAM-1 binding PfEMP1 antigens, DBL–DBL domains (see Additional file
3). Further studies of the specificity of seroreactivity observed for individual binding domains or to variants from other sequenced strains can be pursued readily using the same platform, and the consistent annotation schema provide an initial view of antibody responses to different classes of PfEMP1.
Seroreactivity to the newly identified
var A ICAM-1 binding motif (2 spots on the array) [
32] was found to parallel seroreactivity to EPCR-binding head structures (Fig.
4b). EPCR-binding PfEMP1 often also encode ICAM-binding motifs (Table
3 and Additional file
7). While the magnitude of seroreactivity to this ICAM-1 binding motif was lower than that to CD36-binding and rosetting PfEMP1, a greater fraction of children had detectable antibodies to this motif compared with PfEMP1 extracellular domains associated with EPCR-binding. Exposed children and adults in malaria endemic regions have reactive, inhibitory IgG to this conserved ICAM-1 binding motif, possibly reflecting the greater conservation of the ICAM-1 binding motif than the EPCR-binding CIDR [
32]. Here, the data demonstrating greater seroprevalence in children to this motif than to linked CIDR domains is consistent with those results. In adult hyperimmune sera, seroreactivity to ICAM-1 domains on the array was also greater than seroreactivity to EPCR domains (see Additional file
6).
Neither a consistent response to the corresponding antigen in children exposed to a specific PfEMP1 variant in acute infection or a general significant increase in PfEMP1 IgG antibody at 30 days convalescence were detected. Prior studies employing in vitro functional assays demonstrated that children seroconvert against the VSA of the isolate seen in acute infection following disease resolution [
51] and subsequent studies suggest that the major VSA recognized by immune sera is PfEMP1 [
52]. An array study performed in Papua New Guinea showed that the anti-PfEMP1 antibody response was more pronounced in children with detectable parasitaemia [
26]. It is possible that the study subjects made PfEMP1-specific antibody in response to acute infection, but this antibody was not maintained in convalescence.
The array was developed using the 3D7 proteome, which lacks some EPCR-binding proteins. 3D7 PfEMP1 proteins may lack sufficient homology to Malawian PfEMP1 to be recognized by specific antibody. The PfEMP1 family is hypervariable with less than 50% sequence identity found in homologous domains of homologous genes [
15,
16]. Strain-specific polymorphism of MSP1 and AMA1, vaccine antigens that are more conserved than PfEMP1, is proposed to be a major limiting factor in development of protective antibodies [
53]. Additionally, incomplete or inaccurate protein folding for certain PfEMP1 antigens on the array may also contribute to the observed differences across studies, especially since many
P. falciparum-specific antibodies recognize conformational epitopes. The cell free in vitro expression system and direct printing protocol used in production of the microarray employed here involves minimal manipulation of expressed proteins and no denaturation steps, but correct conformation of recombinant protein is not readily verifiable.
Seroreactivity in both acute infection and convalescence to antigens from all PfEMP1 binding phenotype groups was identified in UM and CM cases. Seroreactivity to EPCR domains is acquired relatively early in life [
28] and may explain why older children are less likely to develop CM. Children with symptomatic malaria had a lower level of seroreactivity to EPCR-binding
P. falciparum antigens than to other PfEMP1 antigens. The median age of our subjects was 3–4 years, whereas peak EPCR-binding domain antibody prevalence was detected in children aged 8–10 years [
28]. Although the rate of antibody acquisition is dependent upon a number of factors including exposure to
Plasmodium, studies that have examined acquisition of “protective” antibody responses to VSA or
Plasmodium antigens, significant protection begins to appear in children 8–10 years of age [
12,
52]. Children appear to be at risk for symptomatic malaria until they are older than the children described here. Pooled immune IgG from Malawian adults [
34] (see Additional file
6) was seroreactive across all PfEMP1 binding groups. Multiple exposures are likely required to develop significant cross-reactive antibody to hypervariable extracellular PfEMP1 domains. An array that encompasses Malawian PfEMP1 sequences, as reported for the Papua New Guinea study [
26], may be useful to determine when children with malaria develop antibodies to local PfEMP1 variants.
Antibody reactivity to PfEMP1 binding groups should be compared with caution and the findings would be strengthened by supplemental functional assays and comparison with other platforms [
25]. The high throughput microarray provides a cost-effective platform to evaluate reactivity to a large number of
P. falciparum antigens using small volumes of sample and is readily amenable to simultaneous querying of reactivity of sera to proteins from strains of disparate genetic backgrounds, as recently demonstrated for AMA1 and MSP1 [
53]. Due to recent advances in HIV vaccine efforts, there has been renewed interest in vaccine development of a number of infectious diseases associated with chronic antigen exposure. For many of these diseases, including malaria, there is often a robust host antibody response but very slow acquisition of disease-modifying immunity. Given progress in HIV vaccine research, there has been renewed interest in identifying strain-transcendent protective epitopes.
Authors’ contributions
AK, JC, and KK conceptualized and designed the study. AK and VH collected and processed patient samples. AK performed the experiments. AK and JC performed the formal data analysis, and WBM supervised the statistical analysis. WLM, SR, and KS oversaw the malaria cohorts. AK, JC, and KK wrote the manuscript with input from all authors. KS and KK supervised all work. All authors read and approved the final manuscript.