Interferon-γ responses to Plasmodium falciparum vaccine candidate antigens decrease in the absence of malaria transmission

Study population and malaria surveillance

The study was conducted in the adjacent highland locations of Kipsamoite and Kapsisiywa in North Nandi District in western Kenya, which have historically had unstable malaria transmission (Ernst et al., 2006; John et al., 2005). Intensified annual indoor residual spraying of long-lasting insecticides and introduction of artemether-lumefantrine as first-line therapy for symptomatic malaria produced a period of apparent interrupted malaria transmission from April 2007 to March 2008 in which no clinical cases of microscopy-confirmed malaria were recorded on site (John et al., 2009).

Written informed consent was obtained from the participants or their guardians (in case of children). The study was approved by the ethical and scientific review committees at the Kenya Medical Research Institute and the Institutional Review Board at the University of Minnesota (SSC Protocol No. 939).

Three hundred and five individuals, randomly selected (from a total population ∼8,000 in 2008) over a one-year period, consented to participate in the longitudinal study. Healthy individuals over the age of two years who had lived at the site for more than six months were eligible for inclusion. Samples were collected in April 2008, October 2008, and April 2009. Two hundred and forty-three of the 305 individuals were present at all three time points and had sufficient peripheral blood mononuclear cells (PBMC) for testing of IFN-γ responses to all six antigens to be studied (mean age 25.9, median age 22.3, age range 2.4 years to 82 years), though IFN-γ responses to individual antigens are missing at some time points.

Clinical malaria was assessed by weekly active surveillance of study participants by trained village health workers for malaria symptoms (fever, headache, chills, severe malaise). Participants with symptoms of malaria were referred to the local dispensary for further evaluation and treatment.

Blood samples obtained in April 2008, October 2008, and April 2009 were also tested for asymptomatic P. falciparum parasitemia by microscopy (John et al., 2005) and nested PCR (Menge et al., 2008) as previously described (Noland et al., 2012). Further assessment for malaria transmission was by measuring prevalence of antibodies to P. falciparum antigens CSP (Esposito et al., 1988) and schizont extract using ELISA previously described (Wipasa et al., 2010).

Blood sample collection and processing

Venous blood (3–5 mls from children and 10–15 ml from adults) was collected into heparinized vacutainers (BD, Plymouth, UK) separated by histopaque density gradient centrifugation. PBMC were cultured in 96 well plates and stimulated with P. falciparum peptides. Phytohemagglutinin (PHA) and PBS were used as positive and negative controls respectively.

Selection of peptides of P. falciparum pre-erythrocytic and blood stage antigens

P. falciparum pre-erythrocytic and blood stage antigens were selected for testing based on prior studies demonstrating immunogenicity or association with protection from P. falciparum infection and disease in malaria endemic areas. All antigens tested except MB2 have been assessed as vaccine candidate antigens. Peptides selected for testing are detailed in Table 1. They included peptides from the pre-erythrocytic antigens: circumsporozoite protein (CSP (Reece et al., 2004)), liver stage antigen-1 (LSA-1), thrombospodin-related adhesive protein (TRAP) (Flanagan et al., 1999), and the blood-stage antigens: apical membrane antigen-1 (AMA-1) (Lal et al., 1996; Udhayakumar et al., 2001), MB2 (pool of peptides MB1 (Nguyen & James, 2001) and MB2, (Nguyen & James, 2001)), and merozoite surface protein-1 (MSP-1, pool of peptides M1 (Parra et al., 2000) and M2, (Udhayakumar et al., 1995)). The peptides were synthesized and purified by high-performance liquid chromatography (HPLC) to >95% purity (Sigma Genosys, St. Louis, MO, USA).

Antibody ELISA testing

To detect any evidence of recent malaria exposure in the population, plasma samples were tested for antibodies to circumsporozoite protein (CSP) and schizont extract by ELISA. The (NANP)₅ repeat peptide was used for the circumsporozoite protein (CSP) while the schizont extract was obtained from cultures of 3D7 P. falciparum strain. Both CSP peptide and schizont extract were were dissolved in 0.01 M PBS to a concentration of 10 µg/ml. Briefly, 50 µl of antigen solution was added to Immulon-4HBX plates (Dynex Technologies, Chantilly, VA, USA). Following overnight incubation at 4 °C, and blocking in 5% nonfat powdered milk in PBS, 50 µl samples of serum diluted at 1:100 in 5% powdered milk were added to wells and incubated for 2 h at room temperature. After washing with PBS −0.05% Tween 20, 50 µl of alkaline phosphatase conjugated goat anti-human IgG (Jackson ImmunoResearch, West Grove, PA, USA) diluted 1:1,000 in 5% powdered milk was added and incubated for 1 h. After extensive washing with PBS-0.05% Tween 20, p-nitrophenylphosphate was added in accordance with the manufacturer’s instructions (S0942; Sigma, St. Louis, MO, USA). The optical density (OD) was measured at 405 nm (Molecular Devices, Sunnyvale, CA, USA). Antibody levels were expressed in arbitrary units (AUs), calculated by dividing sample OD for test plasma sample by the mean plus 3 SD from plasma from malaria naïve North American controls. An AU >1.0 was considered a positive response.

Cell culture and ELISA for cytokine testing

PBMC was diluted in RPMI media (containing 10% human AB serum, 10 mM L-glutamine, 10 µg/mL of gentamicin) at a concentration of 10⁶ cell/ml then plated in duplicate at 2 × 10⁵ PBMC/well in round-bottomed plates (Corning^® Costar^® microtiter plates). P. falciparum antigens were added at a concentration of 10 µg/ml while the PHA positive control was added at 5 µg/ml. Cells were incubated at 37°C, 5% CO₂ and humidity for ∼120 h. After which, supernatants were harvested and stored −20°C for later batch analysis. IFN-γ responses were measured by ELISA assay as previously described (John et al., 2004). To obtain the cut-off, PBMCs from 13 malaria-naïve North American controls were cultured in similar conditions as test samples above. The culture supernatants were collected after 120 h incubation and tested for IFN-γ. A positive IFN-γ response to an antigen (cut-off) was defined as a response >mean + 2 standard deviations of the IFN-γ responses of 13 North American individuals never exposed to malaria. Cut-off IFN-γ levels (pg/ml) for individual antigens were: CSP, 43; LSA-1, 101; TRAP, 27; AMA-1, 49; MB2, 67; MSP-1, 31.

Statistical analysis

Prevalence of IFN-γ responses was compared between age groups using Pearson’s χ² test; change between two time points was tested using the exact two-tailed McNemar test.

In analyzing levels of IFN-γ responses, an antigen’s dependent variable was the common (base 10) logarithm of its IFN-γ response. To include zero responses, a number was added to the levels before taking the log; for each antigen, this number was the 2.5th percentile of positive IFN-γ levels. To test change over time in log IFN-γ response for an antigen, a mixed linear model was used. Site and site-by-time interaction were included to allow differences between sites, but these were not significant and are not discussed further. Preliminary analyses also found that age and sex had negligible effect on the results, so they were excluded from the analyses reported here. Pair wise comparisons of times used Tukey’s post-hoc test separately for each antigen. The mixed linear model analysis was checked using a parametric bootstrap (sampling persons), which gave nearly identical P-values.

The half-life of IFN-γ response was estimated using the method of Wipasa et al. (2011). The analysis used a mixed linear model with a random intercept for each individual and fixed effects for site (Kipsamoite or Kapsisiywa) and for time, treated as a continuous measure. Half-life was estimated as follows. The mixed linear model used to analyze the data was ${\displaystyle log\left(\text{IFN}-\gamma \text{level at time}t\right)=\text{intercept}+\alpha \times t+\left[\text{other items}\right]+\text{error},}$ where time t measures time from time 0—for each person, time 0 was the collection date in the April 2008 wave—with units ”years” and the “other items” are constant for a given person. The coefficient α estimates the decline per year in log(IFN-γ level). The half-life is the time t such that 0.5 = IFN-γ level at time t/IFN-γ level at time 0; taking the common logarithm of both sides of the latter equation gives ${\displaystyle -log\left(2\right)=log\left(\text{IFN}-\gamma \text{level at time}t\right)-log\left(\text{IFN}-\gamma \text{level at time}0\right)=\alpha \times t.}$

Solving for t gives the half-life as −log(2)∕α, and half-life was estimated by replacing α with its estimate. The 95% confidence interval for half life was computed by transforming the upper and lower limits of α’s confidence interval; bootstrap confidence intervals were nearly identical.

Analyses used Stata 12 (Stata Corp, College Station, TX, USA) and the R system (R Development Core Team, 2012), specifically the lme function in the lme4 package (Bates, Maechler & Bolker, 2011).

Study population characteristics and incidence of P. falciparum infection and disease

The age distribution of the 243 study participants was <5 years, n = 22 (9.1%); 5 to 14 years, n = 73 (30.0%); 15 to 40 years, n = 94 (38.7%); and >40 years, n = 54 (22.2%). There were 118 females (48.5%) and 125 males (51.4%). None tested positive for P. falciparum infection by microscopy during the sample collections in April 2008, October 2008 or April 2009 (when all were asymptomatic). While none was PCR positive for P. falciparum in April 2008, one individual (0.4%) tested positive in October 2008 and another two (0.9%) in April 2009. All were part of the area-wide passive surveillance for clinical malaria from March 2007 to April 2008, and none developed clinical malaria during the one-year period of study as determined by weekly active surveillance by village health workers.

Intensity of malaria transmission was, further, assessed by serology (Drakeley et al., 2005; Esposito et al., 1988). Between April 2008 and April 2009, prevalence of seropositivity of IgG antibodies to CSP remained unchanged (April 2008, 8.3%; April 2009, 4.5%; P = 0.11) while prevalence of seropositivity of antibodies to schizont extract decreased significantly (April 2008, 7%; April 2009, 0%; P < 0.0001). Antibody levels to both CSP and schizont extract decreased significantly over the period (Table S1). The implication was that no recent malaria transmission had occurred at the two sites over the study period.

Prevalence of IFN-γ responders to P. falciparum antigens decreases in the absence of malaria transmission

The prevalence of IFN-γ responders to P. falciparum antigens decreased significantly for all six antigens between April 2008 and October 2008 (Fig. 1). Between October 2008 and April 2009, prevalence further declined or remained the same for TRAP, AMA-1 and MB2, while non-significant increases occurred for CSP, LSA-1 and MSP-1. Compared to April 2008, prevalence of IFN-γ responders in April 2009 remained significantly lower for all antigens except CSP, which was of border-line significance (P = 0.06).

IFN-γ levels to P. falciparum antigens decrease significantly in the absence of transmission

IFN-γ levels to all P. falciparum antigens decreased significantly between April 2008 and April 2009 (all P ≤ 0.001, Fig. 2). However the pattern of decrease depended on antigen, with IFN-γ levels to CSP, MB2 and MSP-1 decreasing significantly between April and October 2008 but not between October 2008 and April 2009, IFN-γ levels to LSA-1 similar between April and October 2008, then decreasing significantly between October 2008 and April 2009, and IFN-γ levels to AMA-1 and TRAP decreasing at both time points (so the overall decrease from April 2008 to April 2009 was statistically significant, though neither individual decrease was). Using the data from the three time points, the estimated half-lives in years for the level of IFN-γ responses for the different antigens ranged from 0.42 to 0.71 years (Table 2).

Given that sample testing for each time point was done separately from the other time points, the study also investigated whether differences in experimental condition could have led to the observed differences IFN-γ levels. Thirty randomly selected samples were tested such that all the three time-points were in the same plate. A decrease in IFN-γ responses was observed for CSP, LSA-1, TRAP, MB2 and MSP-1 between April 2008 and October 2008 but between October 2008 and April 2009 the levels remained similar. In contrast, levels of IFN-γ responses to AMA-1 did not vary between the time points (Table  S2). The decrease in levels of IFN-γ to CSP, LSA-1, TRAP, MB2 and MSP-1 over time could, therefore, be attributed to decrease in malaria transmission rather than batch effect. However, the trend of IFN-γ responses to AMA-1 may partly have been due to batch effect.

IFN-γ prevalence and levels to P. falciparum antigens differ by antigen but not by age

The proportion of IFN-γ responders to P. falciparum antigens in April 2008, the period of highest IFN-γ responses, was similar for CSP (15.6%), LSA-1 (11.5%), AMA-1 (12.4%) and MB-2 (14.0%), but lower than the proportions of responses to TRAP (23.1%) and MSP-1 (33.8%) (P = 0.0028 for TRAP and P <  0.001 for MSP-1 compared to AMA-1). Frequency of IFN-γ responses to P. falciparum antigens did not differ by age, except for responses to MSP-1, which similar in frequency across ages in October 2008, but higher in frequency in younger individuals in April 2009 and in older individuals in October 2009 (Fig. 3). IFN-γ levels were not significantly associated with age at any measurement time, except LSA-1 in October 2008 (Pearson’s r = 0.19; P = 0.0034) and MSP-1 in April 2009 (Pearson’s r = 0.16; P = 0.0131) (Fig. S1).