Background
There is an unmet need for highly efficacious malaria vaccines, which would substantially reduce worldwide morbidity and mortality and accelerate malaria elimination. To date, only one candidate has been tested in phase III clinical trials, the RTS,S subunit vaccine administered with AS01
B adjuvant (liposome-based adjuvant). Thousands of children and infants were enrolled from multiple study sites across sub-Saharan Africa, and RTS,S was moderately efficacious against clinical malaria at 38–48 months follow-up (36.3% and 25.9% after the fourth booster at 20 months in children and infants, respectively) [
1]. RTS,S has received a positive scientific opinion by the European Medicines Agency [
2] and will undergo pilot implementation in areas of Ghana, Kenya, and Malawi to further evaluate vaccine safety, reduction in childhood mortality, and feasibility of the four-dose vaccine regimen [
3,
4]. Despite this achievement, vaccine efficacy against clinical malaria was well below the target of 75% as set by the World Health Organization [
5], and longitudinal studies show that protection rapidly declines after vaccination [
1].
RTS,S is based on the major surface expressed antigen on
Plasmodium falciparum sporozoites, the circumsporozoite protein (CSP). The vaccine construct is a fusion protein between a truncated form of CSP (containing the central repeat and C-terminal regions only) and hepatitis B surface antigen that is co-expressed with unfused hepatitis B surface antigen, which self-assemble as virus-like particles. The central repeat region is a tandem repeat of amino acid sequence NANP that is a known B cell epitope, and the C-terminal region contains B and T cell epitopes [
6]. High levels of immunoglobulin G (IgG) to the central repeat region have been broadly associated with protection against clinical malaria in RTS,S vaccine trials [
7]. A recent study was the first to report the specific IgG subclass responses induced by RTS,S/AS01
B in infants and young children (age groups, 6–12 weeks and 5–17 months) and identified that IgG3 to the central repeat region and IgG1 and IgG3 to the C-terminal region were associated with protection [
8]. However, the mechanisms of such antibodies and how they confer protection are undefined. Further research is needed to understand the mechanisms of RTS,S-induced immunity and why efficacy is short-lived so that strategies to enhance vaccine efficacy and longevity in target populations can be developed.
Healthy malaria-naïve adults vaccinated with RTS,S develop antibodies that can block sporozoite infectivity and mediate opsonic phagocytosis using the THP-1 monocyte cell line in vitro. How this relates to protection is somewhat unclear, particularly for the latter mechanism, which has had conflicting findings [
9‐
11]
. Complement fixation and activation by antibodies is an important mechanism of humoral immunity to some viral and bacterial pathogens [
12‐
14] and has been proposed to play a role in immunity against blood-stage
P. falciparum infection [
15]. We have recently demonstrated that naturally acquired human antibodies, or antibodies generated by repeated experimental inoculation of healthy volunteers with sporozoites, can promote fixation of complement factor C1q and activate the classical complement pathway against
P. falciparum sporozoites [
16,
17]. This antibody-complement interaction inhibited the traversal mode of sporozoite motility and led to cell death. Antibodies targeting CSP could also promote complement fixation [
17]. Additionally, high levels of naturally acquired complement-fixing antibodies were associated with a reduced risk of clinical malaria in a longitudinal study of children [
17]. This indicated that antibody-complement activity is an important mechanism of human immunity against sporozoites and raised the question of whether it may also be a functional mechanism induced by RTS,S.
Here, we examined antibody responses from a phase IIb clinical trial of RTS,S administered with AS02
A adjuvant (formerly used adjuvant that is an oil-in-water emulsion) in children resident in Manhiça, Mozambique, that was conducted at two study sites of different malaria transmission intensity [
18]. We aimed to characterize RTS,S-induced antibodies by isotype, IgG subclass, and reactivity to different CSP regions, and to determine whether antibodies induced by RTS,S function by fixing complement. Furthermore, we aimed to understand the influence of antibody types and levels of malaria exposure on complement-fixing activity. We then developed statistical models of the decay of complement-fixing antibodies, IgG subclass responses, and IgM over 5 years of follow-up and explored which responses influenced the decay of complement-fixing antibodies.
Methods
Study participants
Serum samples were obtained from a previously conducted randomized control phase IIb clinical trial of the RTS,S/AS02
A vaccine (
ClinicalTrials.gov registry number NCT00197041). The study was conducted in Mozambique at two study sites to evaluate protective efficacy against clinical disease (Manhiça, cohort 1) and new infection (Ilha Josina, cohort 2). Notably, malaria transmission intensity was low to moderate at the Manhiça study site and high at the Ilha Josina study site. Children aged 1 to 4 years were enrolled and administered three doses of RTS,S/AS02
A at months 0, 1, and 2 of the study or comparator vaccines as previously described [
18]. We tested sera from a random selection of participants collected at baseline (month 0, M0) and 30 days after receiving the third and final RTS,S vaccination (month 3, M3) from Manhiça (RTS,S
N = 50, comparator
N = 25) and Ilha Josina (RTS,S
N = 49, comparator
N = 24) cohorts. We additionally tested sera from the Manhiça cohort collected during a 5-year follow-up to assess long-lived immune responses (months 8.5, 21, 33, 45, and 63,
n = 30). Sample selection was at random and summary statistics closely matched that of the entire cohort (median age, 2.8 years and 2.5 years; male-to-female ratio, 0.93 and 0.83; Manhiça malaria negative-to-positive ratio, 0.17 and 0.15; Ilha Josina malaria negative-to-positive ratio, 2.32 and 1.41, respectively). Individuals tested at later time points were randomly selected from those with sufficient volumes at all time points (out of
n = 371). Note that all serum samples were heat treated for 45 min at 56 °C to inactivate any host complement proteins.
Experimenter blinding
Experimenters were blinded from the vaccine group, age group, and study site until all participants had been tested for IgG, immunoglobulin M (IgM), and C1q-fixation to CSP.
Antigens
The following antigens used in this study were all based on
P. falciparum 3D7: recombinant CSP beginning at amino acid residue 50 that contained (NANP)
22 repeats and (NVPD)
4 repeats expressed in
Pichia pastoris (Sanaria, Rockville, USA), synthetic (NANP)
15 peptide (NANP) representing the central repeat region of CSP (Life Tein, Hillsborough, USA), recombinant C-terminal region (CT) of CSP expressed in HEK293 cells at Burnet Institute (as described in Additional file
1: Figure S1), recombinant merozoite surface protein 2 (MSP2), and apical membrane antigen 1 (AMA1) both expressed in
Escherichia coli as previously described [
19,
20].
Antibody detection assay
Sera were tested by standard enzyme-linked immunosorbent assay (ELISA) to detect antibody reactivity as follows. Ninety-six-well flat bottom MaxiSorp plates (Thermo Fisher Scientific, Waltham, USA) were coated in 0.5 μg/ml antigen in phosphate-buffered saline (PBS) overnight at 4 °C. Plates were washed (thrice with PBS-Tween20 0.05%) using the ELx405 automated plate washer (BioTeck, Winooski, USA) and blocked with 1% casein in PBS (w/v) for 2 h at 37 °C. Plates were washed and then incubated with human sera (tested in duplicate) in buffer (0.1% casein in PBS, v/v) for 2 h at room temperature (RT). Sera were tested at 1/4000 dilution for IgG and IgG subclass detection and 1/500 for IgM detection (note that for the decay analysis, samples collected between M3 and M63 were tested at 1/500 for IgG subclass detection). Plates were washed, and antibody isotypes were detected using goat anti-human IgG and IgM conjugated to horseradish peroxidase (HRP, Millipore, Burlington, USA) at 1/2500 dilution in buffer for 1 h at RT. Plates were washed for a final time and incubated with 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) substrate (ABTS, Thermo Fisher Scientific) shielded from light at RT. IgG and IgM reactivities were measured after 20 and 60 min, respectively, at an optical density (OD) of 405 nm using the Multiskan Go plate reader (Thermo Fisher Scientific).
To detect IgG subclasses, plates were instead incubated with mouse anti-human IgG1, IgG2, IgG3, and IgG4 (Thermo Fisher Scientific), washed, and then incubated with goat anti-mouse IgG HRP (Millipore) all at 1/1000 dilution in buffer for 1 h at RT. Plates were washed for a final time and incubated with tetramethylbenzidine substrate (TMB, Thermo Fisher Scientific) shielded from light at RT. IgG subclass reactivity was stopped after 30 min of incubation using 1 M sulfuric acid, and OD was measured at 450 nm. We have previously confirmed the specificity and sensitivity of the IgG subclass reagents used [
21]. A prior report of IgG subclass responses induced by RTS,S in the phase III trial [
8] used a suspension bead array method rather than ELISA. While there may be differences between these methods, prior evaluation of the two approaches for detection of malaria antibodies using a range of antigens suggested the two methods were comparable [
22].
Complement fixation assay
Sera were tested for the ability to fix complement using the plate-based complement fixation assay as previously described [
17]. Briefly, coating and blocking were conducted as described above for antibody detection. Plates were incubated with human sera (tested in duplicate) at 1/250 dilution in buffer for 2 h at RT (note that for the decay analysis, samples collected M3-M63 were re-tested at 1/110 dilution for C1q-fixation). Plates were washed and incubated with purified human C1q (Millipore) at 10 μg/ml in buffer for 30 min, followed by washing. To detect C1q-fixation, plates were incubated with rabbit anti-C1q IgG (in-house), washed, and then incubated with goat anti-rabbit IgG HRP (Millipore) both at 1/2000 dilution in buffer for 1 h at RT. Detection antibodies were validated as described in Additional file
1: Figure S2. Plates were washed for a final time and incubated with TMB shielded from light at RT. C1q-fixation reactivity was stopped after 30 min of incubation using 1 M sulfuric acid, and OD was measured at 450 nm.
We additionally tested a random selection of n = 20 individuals (from the RTS,S vaccine group collected at 3 M) for C5b-C9-fixation to CSP. This was conducted using the same protocol, but fresh human serum pooled from malaria-naïve donors was used as a source of complement (1/10 dilution), and rabbit anti-C5b-C9 detection antibodies were used (1/1000 dilution).
Experimental controls and standardization
To ensure accurate results, human serum samples were tested in duplicate, and those with high variability (> 25%) were re-tested or excluded (unless duplicates differed OD < 0.1). Raw data were corrected for background reactivity using no-serum/blank negative controls and adjusted for plate-to-plate variation using malaria-exposed positive controls that were included on each plate. We also included malaria-naïve negative controls from Melbourne donors (N = 24), and test samples with an OD greater than the mean + 3 standard deviations of the Melbourne controls were considered positive.
We also examined the epitope specificity of individuals and considered the ratio of NANP-to-CT IgG as equal if the variability between NANP and CT responses was < 25%, which was the acceptable range for duplicate variability. Individuals who exceeded this were considered as having a NANP (ratio > 1.25) or CT-skewed (ratio < 0.75) IgG response.
Statistical analysis
For descriptive and comparative analysis, the median and interquartile range (IQR) were reported, and the following two-tailed non-parametric tests were performed where appropriate (GraphPad Prism 7; p < 0.05 was considered significant): Mann-Whitney U test, Wilcoxon matched-pairs signed-rank test, and Spearman’s correlation coefficient (Rho).
Statistical models were developed to investigate (i) decay rates of complement-fixing antibodies, antibody isotypes, and IgG subclasses and (ii) the relationship between complement-fixing antibodies and different antibody types using repeated measures longitudinal data (Stata version 14, StataCorp). Given the dependency on the data from repeated measurement, multilevel modelling was used to estimate latent growth curve models exploring the subject-specific nature of the association between time and C1q-fixation, IgG subclasses, and IgM. Given the non-linear functional form of the association between time and C1q-fixation, IgG subclasses, and IgM, latent growth curve models regressed the log of C1q-fixation and IgG subclass and IgM on log time, and post-estimation non-linear equations using exponentiated model coefficients were used to provide proportional rates of change at specific time points. Latent growth curve models comprised two levels, individuals at level 2 (i.e., random intercept and coefficient for time) and their C1q-fixation, IgG subclass, and IgM response across time at level 1 (see Additional file
1: Supplementary equation 1). Nested model-based likelihood ratio statistics were used to provide statistical inference for model fit when relaxing model constraints (random effects and the functional form of fixed effects for time) and Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) fit statistics used to inform functional form modeling across non-nested models. To further assess the fit of the estimated latent growth curve models, diagnostic plots comparing participants observed marker levels with Bayesian model-based (best linear unbiased predictions) predicted levels over time were produced and examined.
Contemporaneous (i.e., both outcome and factor variable responses from the same time period were regressed) unadjusted and adjusted longitudinal associations between C1q-fixation, IgG subclass, and IgM levels were also estimated using multilevel modelling. Level-1 error variance estimates from nested regression modeling based on these analyses were used to derive Cohen’s
f2 measures of effect for each IgG subclass and IgM [
23]. In these multilevel models, IgG subclass and IgM were estimated as log time-varying fixed factors and their effects unconstrained by time. In addition to these terms, both fixed and random terms for log time were also estimated and a random effect (intercept (level 2)) to account for the dependency C1q-fixation level measurement over-time was also estimated. Statistical inference was assessed at the 5% level.
Discussion
Knowledge of the induction and maintenance of functional immune responses by RTS,S is valuable for strategically improving vaccine design and delivery to generate a highly efficacious and longer-lasting malaria vaccine. Here, we investigated RTS,S-induced antibody responses in children and present new evidence of a functional mechanism that may contribute to vaccine-induced immunity. We show that RTS,S-induced anti-CSP antibodies were predominantly IgG1, and there was also a substantial induction of IgG3, IgG2, and IgM. Antibodies among vaccinated children could strongly fix complement on the whole, but functional responses did vary among individuals. Our data suggest that antibodies targeting both the central repeat and C-terminal regions of CSP are important for complement fixation. Furthermore, functional antibodies were poorly induced in older children with greater malaria exposure, and functional antibodies were poorly maintained after vaccination. This decline in complement-fixing antibodies was largely due to a decline in IgG1 and to a lesser extent IgG3. These findings highlight the diversity of RTS,S-induced immune responses in malaria-exposed populations, demonstrate a new functional mechanism of vaccine-induced antibodies, and shed some light on the maintenance and decay of vaccine immunity. Our findings suggest that generating higher levels of complement-fixing antibodies among a greater proportion of children, and inducing responses that are durable over time, may contribute to greater vaccine efficacy and durability.
IgG1 and IgG3 subclasses have the highest complement-fixing activity, whereas IgG2 and IgG4 have little or no activity [
25]. RTS,S/AS02
A strongly induced anti-CSP IgG1 followed by IgG3 and IgG2 in children aged 1–4 years. Similar findings showing a predominant IgG1 and IgG3 response were recently reported for a subset (
n = 195) of subjects from the phase III RTS,S/AS01
B vaccine trial of infants and young children (6–12 weeks and 5–17 months). A higher ratio of cytophilic antibodies (IgG1 and IgG3) to non-cytophilic antibodies (IgG2 and IgG4) was associated with vaccine-induced protection [
8]. Therefore, we propose that one way in which RTS,S-induced cytophilic antibodies may confer protection is by interacting with human complement, a mechanism that has been previously associated with protection in studies of naturally acquired immunity to sporozoites in children [
17]. The AS01
B and AS02
A adjuvants both contain TLR4 agonist MPL and the saponin QS-21, but AS01
B is a liposome-based adjuvant whereas AS02
A is an oil-in-water emulsion [
26]; RTS,S administered with either adjuvant appears to induce similar IgG subclass profiles in malaria-exposed children. In the phase III study, post-vaccination levels of IgG2 and IgG4, but not IgG1 and IgG3, to CSP were significantly higher in the high malaria transmission site compared to the low malaria transmission site [
8]. It is noteworthy that studies in malaria-naïve adults found RTS,S to predominately induce IgG1 and IgG2 only [
27‐
29], which demonstrates that immunogenicity can substantially differ in malaria-exposed populations compared to malaria-naïve individuals in non-endemic countries. Such differences are poorly understood but require a greater understanding, particularly because IgG3 is a potent mediator of complement fixation and activation. The use of adjuvants or vaccine regimens that increase IgG3 and reduce IgG2 induction is likely to lead to responses with greater complement-fixing activity and potentially greater efficacy.
Complement activity can also be mediated by IgM antibodies. Interestingly, anti-CSP IgM seropositivity was moderate at baseline, although functional C1q-fixation responses were only apparent after RTS,S vaccination when IgM (and IgG) had significantly increased. Modeling the decay of antibodies over time suggests IgM is not a major mediator of complement-fixing activity, and this was instead largely mediated by IgG1 and IgG3. The role of IgM in immunological memory is not well established, but a recent study demonstrated that IgM memory B cells were long-lived and involved in secondary responses using a murine model of malaria [
30], and IgM responses remain present even in those with extensive malaria exposure [
31].
Complement activity is also influenced by epitope-specificity [
32‐
34]. We found that RTS,S-induced antibodies were most potent at fixing complement to the central repeat region of CSP, but antibodies to the C-terminal region could also fix complement. Interestingly, children with higher complement-fixing antibodies tended to have equal levels of IgG to both CSP regions, whereas those with low functional antibodies were more frequently epitope-skewed. Therefore, antibodies to both regions may better promote immune complex formation that leads to complement fixation and activation. Overall, there were highly diverse IgG epitope profiles among children. In field evaluations of RTS,S, only IgG to the central repeat region have been typically measured and reported. This is likely because RTS,S-induced antibodies to the central repeat region, but not the C-terminal region, show some association with protection in studies of healthy malaria-naïve adults [
10,
29,
35‐
37]. However, the relationship between protection and repeat-specific antibodies has been inconsistently reported in field evaluations of malaria-exposed children and infants [
18,
38,
39]. A recent study found that antibodies to both regions of CSP, that were of a specific IgG subclass, were associated with protection in the phase III clinical trial of RTS,S/AS01
B [
8]. Additionally, the central repeat region is considered immunodominant as it is a major target of antibodies induced by whole irradiated attenuated sporozoite vaccines and natural malaria exposure [
40,
41], and repeat-specific antibodies can block sporozoite infectivity in vitro [
42]. However, antibodies to non-repeat regions are also naturally acquired and have demonstrated inhibitory activity in vitro [
43‐
47]. Taken together, our data support the importance of antibodies to the central repeat region of CSP, but also encourage further investigation of antibody responses to the C-terminal region, particularly because repeat-specific antibodies alone are a poor correlate of protection.
RTS,S-induced antibodies strongly fixed complement to CSP, but overall functional activity among vaccinated children markedly declined by M8.5. Although total IgG remained moderate at this time, reactivity may have dropped below the threshold required to detect functional activity using our methods, particularly because antibody density is known to influence complement activity, as multiple C1q monomers (six in total) bind neighboring IgG molecules to initiate the classical complement pathway [
32]. Our statistical modeling indicated that the rapid decay of complement-fixing antibodies was mostly explained by the decline in IgG1 and to a lesser extent IgG3, likely because both subclasses can potently fix complement and because IgG1 was the predominant subclass induced by vaccination. Perhaps, if RTS,S induced greater levels of IgG3 that were above the antibody threshold needed for functional activity and were more durable, complement-fixing antibody responses may have better maintained after vaccination. Including measures over time enabled a better understanding of the relationships between specific antibody types and complement fixation than analysis performed at a single cross-sectional time point. While the decay post-immunization of IgG to the central repeat region has been reported [
48], there are no data on the decay of functional antibodies, IgG subclasses, or IgM. Additionally, no malaria vaccine studies have reported the relationship between IgG or IgM and functional antibodies over time. Therefore, our new statistical methods developed for this study may be valuable for understanding these responses in other RTS,S trials and studies of other malaria vaccines. It is possible that functional antibodies, IgG subclasses, and IgM have different responses and kinetics after booster immunizations. The decay of complement-fixing antibodies was broadly consistent with the decline in RTS,S vaccine efficacy for this trial. Statistical analyses estimated that initial efficacy against clinical malaria was ~ 30% in the first 4–5 months but quickly waned to limited efficacy 18–30 months after vaccination. In our studies, some children retained significant complement-fixing antibodies at M8.5, which may contribute to some continuing vaccine efficacy. Our findings suggest these questions warrant investigation in the phase III trials to better understand the efficacy and durability of RTS,S.
The present study was conducted in a malaria-endemic setting, and therefore, children will have varying degrees of malaria exposure and naturally acquired immunity. Older children in the higher malaria transmission site, Ilha Josina, had significantly lower levels of complement-fixing antibodies. This was also observed for total IgG to CSP, but the effect was not as pronounced. Older children also had higher levels of antibodies to blood-stage antigens, and this inversely correlated with vaccine-induced anti-CSP antibodies, which has been previously reported [
49,
50]. Interestingly, these trends were only clear in the Ilha Josina study site, where transmission intensity was approximately tenfold higher than the Manhiça site. These findings suggest that age and higher levels of malaria exposure negatively impact on the induction of functional antibodies by the RTS,S vaccine. This is a significant finding that has implications for generating immunogenic and efficacious malaria vaccines in high transmission zones. The biological basis for this effect is not known but could be due to T or B cell exhaustion from repeated exposure [
51], or naturally acquired immunity could be interfering with the induction of vaccine immunity. Understanding this issue is a priority for further research.
Acknowledgements
We thank all children and parents/guardians who participated in this study, all CISM and ISGlobal staff who facilitated the Mozambican RTS,S/AS02A Phase IIb clinical trial, and GSK Biologicals for the conduct of the clinical trial. We thank Pedro Alonso who was involved in all phases of the clinical trial. We thank Joe Campo for managing the samples and databases used for all immunogenicity studies, particularly the M63 cross-sectional follow-up, as well as Caterina Guinovart, Eusebio Macete, Pedro Aide and Augusto Nhabomba who were involved in the clinical trials. We also thank Nana Aba Williams and Núria Díez for coordination and management, Ruth Aguilar, Gemma Moncunill and Clarissa Valim for their inputs in the design of this study, Marta Vidal for preparing samples for shipment and Laura Puyol for arranging shipment.