Background
Combined with other measures, vaccination is considered a promising approach to control and eventually eliminate malaria [
1,
2]. Extensive polymorphism in many
Plasmodium falciparum proteins may limit the efficacy of vaccines based on just one or two allelic variants that are not broadly cross-protective against diverse antigens found in natural infections. For example, a multi-antigen blood stage malaria vaccine evaluated in Papua New Guinea reduced parasite density and prevalence of infection in a strain-specific manner, suggesting selection of non-vaccine variants [
3]. In contrast, immunization with RTS,S, a vaccine directed against the pre-erythrocytic circumsporozoite protein [
4], does not appear to result in selection or allele-specific efficacy [
5,
6].
Apical membrane antigen-1 (AMA1) is a leading blood stage malaria vaccine candidate that is thought to play a critical role in erythrocyte invasion. Antibodies against AMA1 have been shown to block parasite invasion [
7‐
11], and sero-epidemiological studies have shown an association of anti-AMA1 antibodies with naturally acquired protection against malaria [
12,
13]. Animal studies [
14] have shown the ability of vaccines based on this antigen to stimulate antibody responses that were correlated with a reduction of parasite density.
AMA1 is expressed mainly during the
P. falciparum asexual or blood stage, and examination of
ama1 sequences from natural
P. falciparum infections has shown extreme diversity in this gene. There are up to 62 polymorphic amino acid sites in AMA1, representing more than 15% of the amino acid sites distributed over three domains of the protein [
15,
16]. The greatest polymorphism is seen in domain I, especially in a cluster of amino acids near a hydrophobic pocket [
17] that is thought to play a role in erythrocyte invasion, the cluster 1 loop of domain I (c1L), which includes amino acid residues 196, 197, 199, 200, 201, 204, 206 and 207 [
18]. Based on antigenic escape modeling, growth and invasion inhibition assays [
19] and molecular epidemiological studies of the impact of AMA1 polymorphism on risk of clinical malaria [
16], c1L has been identified as a key target of both strain-specific antibodies and allele-specific naturally acquired protective immune responses.
The malaria vaccine AMA1-C1 is a bivalent vaccine comprised of recombinant AMA1 based on the 3D7 and FVO reference strains of
P. falciparum expressed in
Pichia pastoris and adjuvanted with aluminum hydroxide, developed by the Malaria Vaccine Development Branch of the National Institute for Allergy and Infectious Diseases of the U.S. National Institutes of Health and tested at the Malaria Research and Training Center at the University of Bamako in Mali. Phase 1 studies of AMA1-C1 in Malian adults and children showed that it was acceptably safe and tolerable and modestly immunogenic [
20,
21]. However, as with other vaccine antigens, the high degree of polymorphism observed in this protein [
16,
22‐
24] and possible strain specificity of the immune response [
25] may limit its success as a vaccine candidate. Since failures of AMA1 vaccine in animal models have been attributed to the diversity of the antigen used in vaccine formulation [
26,
27], careful measurement of allele-specific efficacy and vaccine selection are important components in assessing the efficacy of malaria vaccine candidates in field trials conducted in endemic countries [
28].
A Phase 2 clinical trial of AMA1-C1 vaccine in Malian children showed no impact of vaccination on parasite density or clinical malaria [
29]. Even in the absence of measurable overall parasitological or clinical efficacy, it was hypothesized that children who received the AMA1 vaccine might have a decreased incidence of clinical malaria caused by parasites having AMA1 alleles similar to the vaccine alleles as a consequence of the strain specificity of the immune response. In this scenario, the low vaccine efficacy might be explained by the vaccine only providing protection against parasites with AMA1 haplotypes (based on the immunologically relevant polymorphic amino acid residues in c1L) similar to those of the strains represented in the vaccine formulation.
Data and samples collected during this trial were used to assess whether this bivalent malaria vaccine produced a response that was specific to the vaccine antigens, resulting in selection of alleles differing from the vaccine strain. It was further reasoned that even if selection could not be demonstrated directly by comparing frequencies of vaccine-type and non-vaccine type AMA1 in vaccinees and controls, these data might provide indirect evidence for allele-specific immunity elicited by the vaccine.
Discussion
The extensive genetic diversity that is maintained in malaria vaccine candidate antigens through balancing selection applied by host immunity may hamper the development of effective malaria vaccines, especially those targeting highly polymorphic blood stage antigens such as AMA1 [
35,
28]. Molecular epidemiological studies can suggest which antigen variants might best be included in vaccines based on their prevalence in natural populations [
16], and molecular epidemiological, population genetics [
30] and in vitro invasion inhibition studies [
19,
36,
37] all provide clues about which variants might offer the most cross-protection in multivalent vaccine formulations. However, only field trials of vaccine efficacy against diverse parasites can provide definitive evidence of cross-protection or the lack thereof. Here, results are reported of analyses of allele-specific efficacy of AMA1-C1, a bivalent AMA1 vaccine that was designed to overcome allelic diversity in this extremely polymorphic antigen by including two allelic variants of the target antigen. AMA1-C1 is the first AMA1 vaccine to be evaluated in a field trial measuring efficacy against malaria in a natural setting.
The AMA1 sequences included in AMA1-C1 are derived from the
P. falciparum strains 3D7 and FVO. These sequences were chosen based on the availability of these two well-characterized culture-adapted strains with divergent sequences, without knowledge of the baseline distribution of the corresponding AMA haplotypes in the natural parasite populations where the vaccine would be tested and eventually deployed. The results of this study show that, based on polymorphisms in the entire AMA1 ectodomain, fewer than 3% of AMA1 sequences examined from samples collected at the vaccine trial site had haplotypes matching 3D7 while none had the FVO haplotypes; very similar results were found at another vaccine testing site in Mali [
16]. Thus a possible explanation for the failure of AMA1-C1 to demonstrate protection in a Phase 2 trial in 300 Malian children was that allele-specific immune responses induced by the vaccine, even if highly effective against parasites carrying homologous forms of AMA1 (either with respect to the whole AMA1 ectodomain or some subset of immunologically important epitopes such as c1L on domain I), were not broadly cross-protective enough to result in measurable efficacy against parasitaemia [
29]. If an insufficiently broad immune response explained the lack of overall efficacy, allele-specific vaccine-induced immune responses should still have been directed against the fraction of parasites with partial or full homology to 3D7 and FVO AMA1, leading to a reduction in the frequency of these alleles following immunization.
The results of this study provide evidence that insufficient coverage of AMA1 diversity does not explain the lack of vaccine efficacy. Several measures of genetic diversity showed no impact of the vaccine on the diversity of AMA1 alleles in infections experienced by vaccine recipients compared to baseline or to infections in the control group. Moreover, no significant association was seen between vaccination and the risk of malaria clinical episodes caused by parasites with AMA1 similar to the vaccine strain with respect to immunologically dominant regions of the AMA1 protein. There was a non-significant ~1.6-fold lower incidence of clinical malaria caused by parasites with AMA1 c1L haplotypes corresponding to 3D7 following immunization with AMA1-C1 than with a control vaccine (Table
4). However, this hint of possible selection is at odds with the observation of increases in the frequency of 3D7-type c1L in both treatment groups during the post-immunization observation period, and the lack of any suggestion of reduced risk of clinical episodes caused by parasites with FVO-type AMA1 c1L in the vaccine group.
It is more likely that AMA-C1 vaccine failed to protect due to insufficient immunogenicity of the vaccine formulation. For this reason, a new formulation of AMA1-C1 that includes the toll-like receptor agonist CPG 7909 has been developed and tested in Phase 1 trials, which show an approximately 12-14-fold increase in post-immunization antibody levels compared to the formulation without this additional adjuvant in malaria naïve populations, and a 2.5-3-fold increase in antibody levels in malaria exposed adults [
38‐
40]. A monovalent AMA1 vaccine based on the 3D7 strain and formulated with the AS02
A adjuvant system has also shown strong and sustained antibody responses in Malian adults [
41] and children [
42]. Results of a Phase 2 safety and efficacy study of this vaccine will be available soon and may allow ascertainment of an allele-specific effect of an AMA1 based malaria vaccine.
Limitations of this study include the possibility that there was insufficient statistical power to detect a subtle degree of allele-specific efficacy. P. falciparum infections with AMA1 corresponding to the 3D7 strain with respect to the entire ectodomain represented only 3% of all the haplotypes present at the vaccine trial site at the start of the study. Furthermore, only single or predominant AMA1 sequences could be used to define haplotypes and for population genetic analyses, also contributing to reduced statistical power to detect differences between the malaria vaccine and control groups. To detect a 50% reduction in the frequency of full AMA1 haplotypes by the vaccine, 1,534 unique or predominant AMA1 sequences would have been required to have 80% power with significance at 5%. Statistical power is increased by examining only haplotypes based on the eight polymorphic amino acid positions within the putatively immunologically important region of AMA1, c1L of domain I--to detect a similar effect, 586 unique or predominant c1L haplotypes would be required, demonstrating the challenges of measuring allele-specific efficacy for vaccines based on uncommon antigenic variants. However, the conclusion that the vaccine had no effect on the distribution of AMA1 haplotypes is supported by the near-uniformity of several measures of genetic diversity and divergence from the vaccine strains.
Conclusion
In this Phase 2 trial of the bivalent blood stage malaria vaccine AMA1-C1 in Malian children, no evidence was found of allele-specific efficacy. AMA1 sequences in infections occurring in the vaccine and control groups were remarkably similar before and after immunization when examined using several complementary statistical and population genetics methods. These results strongly suggest that the AMA1-C1 vaccine failed not due to inability to overcome extensive genetic diversity in AMA1, but for another reason, most likely an insufficiently immunogenic vaccine formulation. Two second generation AMA1 vaccines with more immunogenic adjuvant systems are being evaluated in field trials in Mali. Results of these trials should provide evidence on which to base decisions about whether or not further development of AMA1-based vaccines is warranted, and to guide optimal choices of AMA1 haplotypes to include in a multivalent AMA1 malaria vaccine.
Acknowledgements
This research was supported by the National Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH), Bethesda, Maryland under Cooperative Agreement P50AI39469, and a Tropical Diseases Research/World Health Organization training grant to AO. Support for genetics analysis was provided by the Howard Hughes Medical Institute and NIAID International Center for Infectious Diseases grant U19AI065683 to CVP. We thank the research team at the University of Bamako's Malaria Research and Training Center; David Diemert from the Sabin Vaccine Institute, at George Washington University. We thank Gary Mays (Laboratory of Malaria immunology and Vaccinology, NIAID) for logistical and managerial support. We are grateful to study participants in Mali, without whose cooperation this study would not have been possible.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
AO designed the molecular typing study, performed PCR and sequencing analysis, analysed the data, wrote the paper, reviewed the manuscript and approved the final version. OKD, LHM and RDE, designed the study, coordinated study execution, wrote the report, scientifically reviewed the paper and approved the final draft. XS supervised and coordinated the molecular study, reviewed the manuscript and approved the final version. CVP supervised and coordinated the molecular study, analysed the data, reviewed and revised the manuscript and approved the final version. JM and ST-H reviewed the design of the molecular study, analysed the data, reviewed the manuscript and approved the final version. JD performed part of the PCR analysis and sequencing, reviewed the manuscript and approved the final version. RS, AN, AD and IS conducted field study, and approved the final version.