Background
The development of effective vaccines is a major global goal towards achieving malaria control and elimination. However, a major challenge in the development of highly efficacious vaccines is antigen polymorphism, which is an important issue for many leading vaccine candidates. This includes the RTS,S vaccine, which recently completed phase III trials, and vaccines based on apical membrane antigen 1 (AMA1) and merozoite surface protein 2 (MSP2); for all these candidates, vaccine efficacy against malaria or
Plasmodium falciparum infection was higher for episodes caused by vaccine-like strains compared to vaccine-dissimilar strains [
1‐
3]. Currently, knowledge on the distribution and prevalence in populations of functional immune responses to different alleles or strains for polymorphic vaccine candidates is very limited, but would be highly valuable for guiding vaccine design. A further constraint to vaccine development is a paucity of data on the targets of functional immune responses that may mediate protective immunity. Antibodies form an important component of acquired human immunity [
4‐
6]. Merozoite antigens are important targets of antibodies that inhibit erythrocyte invasion, limit parasite replication and control disease associated with blood stage replication [
5,
7]. However, the major targets of acquired invasion-inhibitory antibodies are unclear.
The merozoite protein apical membrane antigen 1 (AMA1) is a leading polymorphic vaccine candidate that plays an essential role in host cell invasion and is a prominent target of naturally acquired antibodies [
7‐
9]. AMA1 binds the rhoptry neck protein, RON2, a key interaction that is required for formation of the tight junction during invasion [
10,
11], and antibodies to AMA1 inhibit invasion in vitro [
12‐
18]. In malaria-exposed individuals, antibodies to AMA1 are highly prevalent, increasing with age and exposure [
19‐
22], and some studies have found antibodies to AMA1, measured by standard enzyme-linked immunosorbent assay (ELISA), are associated with protection from malaria in longitudinal studies [
7,
8,
20,
23‐
27]. Affinity-purified human antibodies to AMA1 can inhibit invasion [
12], and some people acquire antibodies to inhibitory epitopes of AMA1 [
28]. However, the significance of AMA1 as a target of acquired invasion-inhibitory antibodies and the strain specificity and prevalence of these antibodies remain unknown [
29].
AMA1 is highly polymorphic with more than 200 haplotypes, and reflects the challenges faced in vaccine development of overcoming antigenic diversity to enable highly efficacious vaccines [
12‐
15,
30‐
32]. Humans generate both allele-specific and cross-reactive antibodies to AMA1 [
20,
21,
32], but how these antibodies are acquired and their relative contribution to protection remain uncertain, particularly for functional antibodies. A phase II trial in Malian children of an AMA1 vaccine containing a single allele demonstrated significant strain-specific efficacy, reducing the risk of malaria caused by vaccine-like strains (defined by genotype) [
2]. These results provide an important proof of concept for AMA1-based vaccines, but highlight the need to understand AMA1 antigenic diversity and address this diversity in vaccine design. Although there are more than 200 AMA1 haplotypes, antigenic diversity appears more limited than suggested by sequence diversity [
31,
32]. Population genetics suggest that the distribution of AMA1 haplotypes, or major haplotype groupings, is similar across different geographic regions and that there is a similar proportion of major haplotype clusters within a population [
33,
34]. However, data on the acquisition of functional antibodies in populations is absent due to a lack of tools to measure these antibodies. Understanding these issues is crucial for vaccine design with respect to selecting alleles for inclusion in future vaccines and determining whether vaccine formulations may need to differ between regions. Similar needs exist for many other vaccine candidates for which antigenic diversity is a key issue [
1,
35].
To date, there has been no way of quantifying naturally acquired AMA1-specific invasion-inhibitory antibodies in human populations or quantifying the prevalence of allele-specific functional antibodies. Moreover, the prevalence of allele-specific functional antibodies has not been reported for any malaria antigen to date, because of a lack of tools and approaches. In this study we used AMA1 as a model to investigate the patterns, prevalence and nature of acquired functional immunity to a polymorphic vaccine candidate. We developed a novel approach using genetically engineered
P. falciparum lines expressing one of six antigenically distinct AMA1 alleles that broadly represent global antigenic diversity [
15,
31]. The development of these novel tools enabled us to evaluate the importance of AMA1 as a target of human invasion-inhibitory antibodies, determine the extent to which inhibitory antibodies are cross-reactive and allele-specific and determine whether the prevalence of inhibitory antibodies varies for different allelic serotypes. This knowledge will be highly valuable in selecting alleles for potential vaccine inclusion and advancing our understanding of strain-specific protection, which is a feature of immunity to malaria and relevant to many malaria vaccine candidates.
Discussion
The polymorphic nature of many malaria vaccine candidates presents major challenges to achieving highly efficacious vaccines. To date, there has been very little knowledge on the prevalence and patterns of functional immune responses to any polymorphic vaccine candidate in populations to guide vaccine design. Here we studied AMA1 as an important vaccine candidate and as a model for understanding functional immunity to polymorphic vaccine antigens in populations. We successfully developed a novel approach using P. falciparum engineered to express different polymorphic variants of AMA1 to quantify AMA1 as a target of naturally acquired inhibitory antibodies, thereby determining their prevalence, strain specificity and functional importance in immune sera. Our results reveal that AMA1 is a major target of invasion-inhibitory human antibodies, and these antibodies have a strong strain-specific component to their activity. A highly significant finding is that the prevalence of inhibitory antibodies varied substantially for different alleles, and this prevalence differed between geographic regions. This has major implications for the selection of alleles for inclusion in a multi-allele vaccine. Our studies showed that inhibitory antibodies to the FVO allele were highly prevalent in both geographic regions, and inhibitory antibodies were also prevalent to W2Mef and XIE alleles/serotypes, identifying them as strong candidates for inclusion in future AMA1 vaccines. These findings have major implications for vaccine design, and it has not previously been possible to detect such serotypic differences in functional immune responses by standard immunoassays. Moreover, for the first time, our approach enables measurement of inhibitory antibodies at a population level, which can be used to guide the selection of AMA1 alleles for inclusion in a future multi-allele vaccine. To our knowledge, the allele/serotype-specific prevalence of functional antibodies in populations has not been reported for any malaria antigen or vaccine candidate to date. These functional immunoassays will also be highly valuable for evaluating immune responses and monitoring vaccine escape in vaccine trials.
The selection of specific alleles, strains or serotypes for inclusion in multi-allele or multi-strain vaccines typically needs to consider the predominant alleles or strains circulating in target populations. This approach has been used extensively with vaccines for other pathogens, including influenza, pneumococcus, meningococcus and human papilloma virus. We propose that the prevalence of allele-specific inhibitory antibodies in target populations should be considered in vaccine design, and this study develops new tools to measure this. Our studies establish an approach that could be used to evaluate functional antibodies for other vaccine candidates and quantify the importance of allele specificity and population prevalence of allele-specific antibodies. Given the enormous investment cost in taking vaccine candidates through pre-clinical studies, GMP production and into clinical trials, ensuring vaccine design addresses issues of antigenic diversity and vaccine escape is essential.
Data from multi-allele competition ELISAs amongst human populations and studies of vaccine-induced antibodies in rabbits suggest that the diversity in AMA1 may be covered by only a small number of different AMA1 alleles [
31,
32,
42,
43]. Tailoring vaccine design to include the most prevalent alleles/serotypes, based on assays measuring functional antibodies, would further facilitate allele selection to maximize population coverage by AMA1 vaccines. In this study, the higher prevalence of inhibitory antibodies to FVO and W2mef alleles in the PNG population and FVO and XIE in the Kenyan population almost certainly reflects greater exposure to these alleles in the respective study populations. Our findings support the potential inclusion of these three alleles (or closely related alleles) in a multi-allele AMA1 vaccine, particularly FVO, which was a prominent target of inhibitory antibodies in both populations. However, data are needed from other populations to further inform these selections and better define the extent of regional differences. As there are more than 200 recorded unique alleles of AMA1 [
12‐
15,
30‐
32] and no evidence of geographic clustering [
33,
34], it is unlikely that identical forms of any one of the six AMA1 alleles tested in this study are circulating at a high prevalence in either population examined. Instead, it is likely that there are a number of antigenically similar variants circulating, leading to the acquisition of functional antibodies. As such, each of the AMA1 alleles included in this study appears to represent a member of a distinct serotype or serogroup of AMA1 alleles that each share a collective set of epitopes targeted by invasion-inhibitory antibodies.
Analysis of AMA1 sequences suggested that the ability of AMA1 alleles to recruit T cell help is not a differentiating factor in the ability of each AMA1 allele to elicit functional antibody responses. Whilst the number of human major histocompatibility complex (MHC) class II binding peptides predicted to be present in AMA1 varies greatly depending on the human leukocyte antigen (HLA) allele examined, there is no significant difference in the number of MHC class II binding peptides for a given HLA allele across each of the six AMA1 alleles examined in this study (data not shown; NetMHCII 2.2 server:
http://www.cbs.dtu.dk/services/NetMHCII/). There are also no data to suggest that the FVO, W2mef or XIE alleles are inherently more susceptible to inhibitory antibodies, and studies have shown comparable inhibition of different alleles by vaccine-induced antibodies [
42,
43].
This is the first human population study to quantify AMA1 as a target of naturally acquired invasion-inhibitory antibodies and estimate the prevalence of these inhibitory antibodies. Both children and adults in PNG showed similar prevalence of AMA1-specific invasion inhibition, and adults had a tendency for a higher prevalence of inhibitory antibodies. In contrast to AMA1, inhibitory antibodies to merozoite surface protein 1 (C-terminal region) were lower in both populations [
39,
44].
Acquired human inhibitory AMA1 antibodies have substantial allele-specific activity, rather than cross-reactive inhibition. The strain-specific nature of human inhibitory antibodies we observed is consistent with the strain-specific efficacy reported in a phase II AMA1 vaccine trial [
2]. There was limited overlap in the inhibitory activity of antibodies to different alleles. Whilst a high prevalence of antibodies to all AMA1 alleles was found by ELISA in PNG and Kenyan samples, invasion-inhibitory antibodies to the FVO and W2mef AMA1 alleles in PNG and FVO and XIE AMA1 alleles in Kenya were much more prevalent than invasion-inhibitory antibodies to other AMA1 alleles. We explored the relationship between antibodies to AMA1 measured by standard ELISA, or allele-specific antibodies measured by competition ELISA, and AMA1-specific inhibitory antibodies. We did find some significant associations between antibodies measured by ELISA and AMA1-specific inhibitory antibodies for some AMA1 alleles. However, these associations were not consistent for all alleles, and overall, ELISA values were not highly predictive of allele-specific inhibitory antibodies. This highlights the potential limitations of relying on standard immunoassays that do not assess functional activity, which is influenced by antibody fine specificity, avidity and other factors that are not measured by ELISA. For example, a previous study reported that the growth-inhibitory activity of naturally acquired antibodies did not significantly correlate with AMA1 antibodies measured by standard ELISA; however, measuring antibodies to an inhibitory epitope of AMA1 showed a better correlation with growth inhibition [
28]. Furthermore, AMA1 vaccine trials in malaria-exposed populations have reported that the induction of AMA1 antibodies measured by ELISA does not necessarily predict the induction of functional inhibitory antibodies [
45,
46]. It has also been reported that some acquired antibodies may interfere with the activity of AMA1-specific invasion-inhibitory antibodies [
29]. We did find significant correlations between AMA1 antibody reactivity and total growth-inhibitory activity of serum antibodies. However, total growth-inhibitory activity is mediated by antibodies to multiple antigens that are co-acquired with increasing exposure to malaria [
47], not just by antibodies to AMA1. These observations and considerations further emphasise the need for functional assays to evaluate AMA1 in populations and clinical trials.
Our data are consistent with malaria-exposed individuals acquiring a repertoire of allele-specific invasion-inhibitory antibodies of multiple specificities, rather than cross-inhibitory antibodies. Although ELISA data indicate that humans acquire some cross-reactive antibodies [
12,
20,
21,
32], it might be that polymorphisms have concentrated around functional epitopes in order to facilitate evasion of protective antibodies [
21]. The most polymorphic residues in AMA1 occur adjacent to the hydrophobic trough, which acts as a binding site for RON2 [
10,
48,
49]. Malaria exposure leads to the acquisition of antibodies to a highly polymorphic inhibitory epitope around this region [
28]. A longitudinal study showed a strong association between polymorphisms in AMA1 and the development of symptomatic malaria and infection episodes [
50]. Previous studies have shown that rabbit antibodies raised against different AMA1 alleles inhibit invasion of a panel of
P. falciparum isolates to varying degrees, but may also show some cross-inhibitory activity depending on the alleles being tested [
31]. The present data suggest that human antibodies are more strongly allele-specific than antibodies from immunised animals; this highlights the importance of studying human antibody responses to inform vaccine development and the caveats in extrapolating results from vaccine studies performed in small laboratory animals to human populations. Although our ability to detect any samples with broadly cross-inhibitory antibodies to AMA1 could potentially be reduced using our approach, this appears unlikely to be a major issue because our data, and those of others [
32], suggest that broadly cross-reactive antibodies to AMA1 do not occur, or are rare, amongst humans; furthermore, they are not generated by standard immunisation approaches [
31,
42]. It is not possible to delete the AMA1 gene and use AMA1-knockout parasites in functional assays, because AMA1 is essential in
P. falciparum; therefore, allele replacement was the only approach possible to quantify antibodies in functional assays. The predominantly allele-specific nature of AMA1 inhibitory antibodies demonstrated by these data indicates that a multi-allele vaccine approach, or alternate strain-covering approach, will be required for AMA1.
Acknowledgements
We thank all study participants and the Papua New Guinea Institute of Medical Research staff involved in the study. We also thank the Australian Red Cross Blood service for providing red blood cells for culture and Colleen Woods and Carter Diggs for constructive feedback.