Measurement of ex vivo ELISpot interferon-gamma recall responses to Plasmodium falciparum AMA1 and CSP in Ghanaian adults with natural exposure to malaria

Despite the recent gains made in the battle against malaria, available control strategies may not be enough to ensure complete elimination and eradication [1]. Vaccines have been a cost-effective public health tool against many infectious diseases and the development of anti-malarial vaccines would be an important addition to existing malaria control strategies. The malaria parasite has a complex life cycle and it is believed that an effective anti-malarial sub-unit vaccine may need to target antigens in multiple stages of the parasite. Subjects in malaria-endemic areas develop partial immunity [2, 3] against severe disease and eventually against mild disease [4‐6], and this supports the feasibility of developing vaccines against malaria. The role of antibodies to malaria antigens, especially the immunodominant circumsporozoite protein (CSP) but also other liver and blood stage antigens, has been widely investigated [6]. But whether naturally acquired antibodies to CSP [7, 8], or other antigens, provide protection against malaria remains elusive [6]. Sub-unit vaccines are being widely developed and the leading candidate, RTS, S which is based on the CSP, elicits protection in up to 50 % of subjects, but begins to wane after the first year [9, 10].

However, durable sterile immunity induced in humans by bites of radiation-attenuated Plasmodium falciparum-infected mosquitoes [11] or injection of purified radiation-attenuated sporozoites [12] is by far the strongest indicator of the feasibility of a malaria vaccine, and immunity is likely mediated by antibodies and T cell responses [12]. There is as yet no clearly defined correlate of protection against clinical malaria but the immune mechanisms mediating protection, at least in animal models, likely include interferon-γ (IFN-γ)-secreting CD8 + T cells that primarily target malaria antigens expressed in liver stages [13, 14]. Therefore, a second approach has been to use heterologous prime-boost regimens such as DNA/human adenovirus-5 that target the pre-erythrocytic stages using CSP and apical membrane antigen-1 (AMA1) [15, 16]. A DNA/Ad vaccine-induced sterile immunity in 27 % of subjects and CD8 + T cell IFN-γ responses to CSP and AMA1 contributed to protection [15, 16]. Protection was associated with class 1-restricted epitopes, particularly in AMA1 [15]. These and other class 1-restricted epitopes have been identified in CSP and AMA1 in subjects immunized with gene-based vaccines [17, 18] and radiation-attenuated sporozoites [19, 20] as well as in subjects in malaria-endemic areas [21‐23].

As with antibodies, the role of naturally acquired T cell responses in providing protection against malaria infection or disease remains largely unanswered., although ELISpot IFN-gamma (IFN-γ) has been used to measure naturally acquired CD8 + T cell responses [6, 24]. Recently, in a study with subjects who were naturally exposed to P. falciparum malaria in Ghana [25], class 1-restricted T cell IFN-γ responses were successfully measured using ex vivo ELISpot. A major finding of this earlier study was that peptides containing predicted class 1-restricted epitopes recalled ELISpot responses from human leukocyte antigen (HLA)-matched subjects, although the frequencies of such responses were lower than recall responses induced by longer HLA DR-restricted peptides [25]. However, since the magnitude of responses was generally low, it was essential to use an appropriate definition of positivity that reproducibly distinguished antigen-specific ELISpot activities induced by natural exposure without vaccine intervention. On this basis, a definition of lower stringency (at least a doubling of spot-forming cells/million (sfc/m) peripheral blood mononuclear cells (PBMC) in test wells relative to control wells, and a difference of at least 10 sfc/m between test and control wells) than that used in vaccine trials conducted by the Naval Medical Research Center (NMRC) [26] was effective for assessing positivity of ex vivo ELISpot IFN-γ responses to P. falciparum CSP and AMA1 in naturally exposed subjects [25]. Of the number of positive responses identified using this stringency definition, 82 % remained positive when the positivity definition used for vaccine-induced responses at NMRC [26] was applied. This positivity definition was adapted for a study measuring naturally acquired ELISpot IFN-γ responses to the P. falciparum cell-traversal protein for ookinetes and sporozoites (CelTOS) in Ghana, namely a stimulation index of >2.0 (response obtained with PBMCs stimulated with malaria antigen peptides compared to medium alone) and a difference of 10 sfc/m between antigen-stimulated and unstimulated PBMCs [27].

These positivity criteria were applied in the current study to better determine the potency of 15mer peptides spanning the entire sequences of the vaccine candidate antigens CSP (3D7) and AMA1 (3D7) for induction of IFN-γ recall responses in PBMCs from naturally exposed subjects in Ghana using ex vivo ELISpot assays. Peptide pools containing 9–10mers that represent known or predicted class 1-restricted epitopes and grouped according to HLA supertypes were also tested against subject PBMCs. This was to assess the feasibility of using 9–10mer pools to predict the HLA restriction of responses to the 15mer overlapping peptides since this would provide further insights into the genetic restriction of naturally acquired ex vivo ELISpot IFN-γ responses to CSP and AMA1 in the same population. Since 11 of the 35 tested subjects had been HLA-typed, it was possible to determine whether such HLA-restricted pools would detect matched HLA-restricted responses of HLA-typed subjects. For subjects who were not HLA-typed, it was possible to use these HLA pools to examine the HLA-restriction of positive responses in some subjects, and the observed activities could be best explained by the promiscuity of class I-restricted epitopes [19, 28‐30]. Taken together, the results obtained from these analyses provide evidence that a malaria vaccine containing CSP and AMA1 may induce broad immune responses in a genetically diverse population. These results also demonstrate the need to include field assessment of HLA-restricted epitopes in malaria vaccine design strategies.

Forty-five healthy Ghanaian adults were screened for this study, and 35 subjects between 24 and 43 years (average age of 29 years) who met eligibility requirements and gave informed consent participated in the study. All subjects were negative for malaria by light microscopy and malaria RDTs and all female subjects were not pregnant. For each subject, the ELISpot activity (sfc/m) for the unstimulated medium control was subtracted from the activities (sfc/m) for each test peptide. All subjects in this study made positive IFN-γ responses to Con A or CEF or both, as previously reported by Anum et al. [27]. In all assays, unstimulated medium control responses ranged between 0 and 18 sfc/m except subjects v15 and v28 whose mean unstimulated medium responses were 69 and 88 sfc/m respectively.

Development of a malaria vaccine will add to currently available malaria prevention and control tools and greatly enhance ongoing elimination and eradication efforts. Identification of immunodominant T cell epitopes within malaria vaccine target antigens may be important for determining whether malaria vaccines would induce protection in a genetically diverse population. An initial human trial with a DNA-prime adenovirus-boost malaria vaccine using CSP and AMA1 achieved 27 % efficacy [16] and this strategy is being developed further to improve efficacy. Immunity to malaria develops in endemic regions, and is often related to intensity and seasonality of exposure to infected mosquitoes [34]. While naturally acquired immune responses have been better characterized in areas of high transmission [35, 36], only a few studies have shown that natural malaria transmission in areas of lower endemicity also induces significant but low ELISpot activities to antigens such as CSP, CelTOS and AMA1 at these sites [25, 27] and that it is possible to measure these reproducibly using positivity criteria of appropriate stringency that was lower than that used to define high vaccine-induced responses. The aim of this study was to assess 15mer peptides that cover the entire sequences of CSP and AMA1, as well as predicted/known HLA class I-restricted epitopes from the same antigens, for their ability to induce IFN-γ recall responses in PBMCs from naturally exposed individuals in Ghana, and to better determine optimal reagents for further site characterization.

These results show that pools of 15mer peptides spanning the full length of the CSP and AMA1 antigens recall IFN-γ responses in adults, and that these responses are approximately twice as frequent for AMA1 compared to CSP. This confirms and extends data from two previous studies in the same site (Legon) which used ex vivo ELISpot to measure activities to CSP and AMA1 [25, 27]. The first study tested the reproducibility of ex vivo ELISpot in areas of natural malaria transmission, where activities were low, and there were fewer positive responses to CSP than to AMA1 when the same Ap and Cp peptide pools described in the current study were used to stimulate subject PBMCs [25]. Similar observations of relatively lower CSP responses were made in the second study [27] where PBMCs were stimulated with single pools containing all CSP or all AMA1 15mer peptides. It has however been shown in a previous DNA/Ad vaccine trial that single pools containing overlapping peptides that span an entire antigen recall lower ex vivo ELISpot activities than summed activities of multiple peptide pools, each containing fewer peptides [15, 16]. Individual smaller peptide pools are therefore optimal for the characterization of CD8 + T Cell IFN-γ responses in potential vaccine study sites.

The greater activity of AMA1 compared to CSP results from long-term exposure of subjects to AMA1, which is also found in blood stages, during natural malaria transmission [37‐40]. In addition, AMA1 (622 amino acids) is larger than CSP (397 amino acids) and is predicted to contain relatively more epitopes and a greater number of immunodominant epitopes compared to CSP (personal observations).

The most frequent responses to CSP were to Cp9 followed by Cp1, and the most frequent responses to AMA1 were to Ap6 followed by Ap1 and Ap2. This differs from earlier observations in Ghana where the most frequent responses were to Cp1, Ap7 and Ap9 [25]. In this earlier study [25], conducted with subjects from Mampong in the Eastern Region of Ghana, smaller CSP HLA peptide pools, or individual HLA peptides, elicited positive responses in 7/19 (36 %) subjects compared to 6/35 (17 %) in this study. It is not known whether this reflects differences in HLA types of the study subjects, or other factors, such as malaria transmission intensity. Mampong, the site of the earlier study, is a semi-urban community about 35 km from Accra [25] and has higher transmission intensity than Legon. A study in a seasonal malaria transmission area in Kenya found no seasonal differences in the frequency of IFN-γ and IL5 ELISpot responses, but higher frequencies of ELISpot responses to IL10 and TNF during the high transmission season compared to the low transmission season [41]. Another study by the same group also found no differences in adult IFN-γ responses between transmission seasons [42]. The role of malaria transmission intensity on the frequency of IFN-γ responses therefore requires further investigation, and future studies may need to factor malaria transmission intensity into site selection for the possibility of identifying additional immunodominant peptides.

Studies using PBMCs from malaria-naïve subjects immunized with adenovirus-vectored AMA1 and CSP candidate vaccines found the highest responses against Cp1, Cp2, Cp6 and Cp9 for CSP [17], and Ap1, Ap3, Ap4, Ap7, Ap8, Ap10 and Ap11 for AMA1 [18]. Thus, positive responses to peptide pools Ap2, Ap6 and Cp4 were observed in the current study but not in studies with malaria-naïve subjects immunized with the gene-based vaccine [16‐18]. These differences may be attributable to the genetic (HLA) background of subjects exposed to multiple strains of P. falciparum by natural transmission or by vaccine immunization with one strain, 3D7. Subjects in malaria-endemic areas have most likely been repeatedly exposed to a diversity of parasite strains and may thus recognize a broader repertoire of immunodominant peptides. This study only investigated responses to CSP and AMA1 that are undergoing T cell-based vaccine development (16–18). However, T cell epitopes have been identified in other antigens such as LSA1 in Kenya [41, 43] and Madagascar [44], and were associated with a delay in parasitaemia after chemotherapy in Kenya [40]. Similarly, T cell responses to pre-erythrocytic antigens were identified in studies in Gambia [22], and memory responses to TRAP were associated with significantly reduced incidence of malaria [45]. These studies, along with those reported here argue for a more comprehensive investigation of the association of naturally-acquired T cell responses to malaria antigens and resistance to malaria infection and disease. This may be especially true as naturally acquired immunity may result in fewer malaria episodes based on the entomological inoculation rate [6, 46, 47]. However, a previous study in Mali suggested that repeated P. falciparum infections do not induce sterile protection [48], suggesting that further studies, especially in sites for likely malaria vaccine trials, are warranted.

It has been suggested that immunodominant T cell epitopes are mostly localized to more conserved regions of AMA1 and that the most polymorphic regions of AMA1 are poorly immunogenic [37]. The most frequent AMA1 pool responses in this study were to Ap6, Ap1 and Ap2; Ap6 contains at least nine polymorphic residues, whereas Ap1 and Ap2 are generally conserved [37]. However, the most polymorphic regions are contained in Ap4 and Ap5 that did not recall responses, in general agreement with the earlier findings [37]. These observations collectively support the need for inclusion of field assessment of antigen peptides in the search for relevant immunodominant peptides for vaccine formulation.

The current study also investigated whether epitope promiscuity was reflected in the epitope-specificity of subjects with natural exposure to P. falciparum. Predicted HLA-binding peptide pools, in contrast to the 15mer pools, recalled fewer positive responses than the Cp or Ap pools. This is not a surprise, given that there were much fewer peptides included in the HLA-specific pools (Additional file 1: Table S1 and Additional file 2: Table S2). For subjects that were not HLA-typed, by screening with a series of predicted HLA-binding peptide pools, it was possible to identify putative HLA restrictions for immunodominant epitopes based on the observed and expected ELISpot activities (Tables 3 and 4). A major factor in determining the specificity of these HLA pools was promiscuity of class I-restricted epitopes and this allowed for a more complete interpretation of the activities of the HLA peptide pools, even in the absence of HLA-typing. Similar observations of promiscuity in such epitopes have been reported previously [19, 29, 30]. Promiscuity in class I-restricted epitopes may potentially be important for broad immune coverage as promiscuous epitopes can be recognized by multiple HLA types (Tables 1 and 2). Promiscuity may however be underestimated here since only individual HLA supertypes rather than HLA alleles are included. Studies that aim to identify immunodominant epitopes require HLA-typing of subjects, but these results suggest that it may be possible to use pools of predicted HLA-binding peptides to detect potential immunodominant epitopes without the requirement for HLA-typing and strengthen the argument for inclusion of epitopes from multiple parasite antigens to ensure broad coverage of protective responses in endemic populations [24].

Of the 11 HLA-typed subjects, the HLA supertypes that the predicted epitopes bind to are amongst the most predominant HLA supertypes known globally [32]. However, a better picture of the degree of epitope promiscuity and how this affects class 1-restricted responses to CSP and AMA1 will require HLA-typing of test subjects. HLA typing and testing of a greater number of subjects to assess epitope promiscuity and association of epitope-specific responses with protection from clinical malaria is planned for further studies in Ghana. These results collectively suggest that vaccines containing CSP and AMA1 may induce effective T cell responses in a genetically diverse population.