Evidence That the Erythrocyte Invasion Ligand PfRh2 is a Target of Protective Immunity against Plasmodium falciparum Malaria

Plasmodium falciparum malaria remains a major global health problem, with ∼500 million cases and 1 million deaths annually (1, 2), and is a leading cause of death among children younger than 5 years of age (3). Protective immunity develops after repeated exposure and prevents severe disease and symptomatic episodes by control of blood-stage parasitemia (4).

During blood-stage replication, P. falciparum merozoites invade erythrocytes, and Abs to merozoite Ags are believed to be important in mediating acquired immunity and immunity generated by candidate blood-stage vaccines (5). However, the targets of protective human Abs are largely undefined. More than 40 merozoite proteins have been identified that may play a role in invasion and/or represent potentially important targets of acquired immunity (6). However, very few of these have been studied in detail as targets of protective immunity against P. falciparum (7, 8). Abs to merozoite Ags are thought to act by inhibition of invasion, Ab-dependent cellular inhibition, and opsonization for phagocytosis or neutrophil-mediated killing (9–15).

The P. falciparum reticulocyte-binding homologs (PfRh; also known as P. falciparum normocyte-binding proteins) are located in the rhoptries of merozoites and are thought to be released to bind erythrocyte receptors for invasion after initial interactions of the merozoite with the erythrocyte surface. The members of this protein family have a region of 500 aa that has identity with the P. vivax reticulocyte-binding proteins 1 and 2; these are thought to form a heterocomplex involved in the direct binding of reticulocytes (16). In P. falciparum, PfRh1, PfRh2a, PfRh2b, PfRh4, and PfRh5 are expressed members of this family, whereas PfRh3 seems to be a pseudogene (17–28). PfRh2a and PfRh2b are identical for the first 2776 aa (88%) of the protein. After a breakpoint, the sequence diverges into PfRh2a- and PfRh2b-specific sequences (19). Substantial evidence suggests the PfRh proteins play important roles in erythrocyte invasion, although their precise functions are yet to be elucidated. PfRh1, PfRh4, and PfRh5 were shown to bind erythrocytes, and Abs generated in experimental animals to PfRh1, PfRh2, PfRh4, and PfRh5 can inhibit invasion (22, 24–27). Furthermore, genetic disruption of specific members alters the erythrocyte receptor preference of merozoites during invasion (6, 13).

Merozoites can use different pathways for erythrocyte invasion mediated by variation in the expression and/or use of erythrocyte binding Ags (EBAs) and PfRh proteins (6, 13). Invasion phenotypes can be broadly classified into two main groups: sialic acid (SA)-dependent invasion, demonstrated by poor invasion of neuraminidase-treated erythrocytes (neuraminidase cleaves SA on the erythrocyte surface) and SA-independent invasion, demonstrated by efficient invasion of neuraminidase-treated erythrocytes. SA-dependent (neuraminidase-sensitive) invasion involves the EBAs and PfRh1. PfRh2 and PfRh4 are important in SA-independent invasion (27–29); however, receptors for these ligands are unknown. Recent studies showed that the use of different invasion pathways through variation in EBA and PfRh utilization mediates evasion of human invasion inhibitory Abs (13).

The important role of the PfRh proteins in invasion, as well as the ability of Abs raised against PfRh ligands to inhibit invasion, suggests that they could be important targets of protective immunity and may be suitable for vaccine development. However, very little is known about immune responses to PfRh proteins. An initial study in Kenya reported that Abs to PfRh2 and PfRh4 were acquired in an age-dependent manner, reflecting the acquisition of immunity in the population (13). Differential inhibition by human Abs of P. falciparum isolates that varied in their use of PfRh proteins pointed to this ligand family as a potentially important target of inhibitory Abs (13). Although the expression of PfRh proteins may vary, PfRh2 seems to be expressed in most isolates, although the expression of gene transcripts was reported to vary (27, 30–33).

In this study, we examined the acquisition of Abs to PfRh2 and their association with protective immunity. We expressed multiple recombinant proteins covering the entire ectodomain of PfRh2 and measured Abs in a longitudinal cohort study of children exposed to P. falciparum. The study design enabled prospective examination of Abs in relation to the risk for reinfection, symptomatic malaria, and high-density parasitemia. IgG subclasses were assessed as one measure of potential Ab function. To further investigate whether PfRh2 may be an important target of protective immune responses, we performed sequence analysis to determine whether there is evidence of diversifying selection in the pfrh2 gene.

Eight recombinant proteins of PfRh2a/b were expressed in Escherichia coli. Proteins were named according to the starting amino acid in the full-length protein sequence. The PfRh2-297 fragment (i.e., commencing at amino acid 297) was amplified by PCR from a codon-optimized 3D7 PfRh2 gene segment. All other PfRh2a and PfRh2b gene fragments were amplified by PCR from 3D7 genomic DNA (European Molecular Biology Laboratory nucleotide sequence database [http://www.ebi.ac.uk/embl/] accession numbers: AY138496 [P. falciparum normocyte-binding protein 2a gene (3D7)] and AY138500 [P. falciparum normocyte-binding protein 2b gene (3D7)]). The following oligonucleotide primers were used (restriction sites are denoted by lower case letters): PfRh2-34: 5′-CAATCAAGTggatccCATGGAGCATCTTCAG-3′, 5′-CTAATAGCTCTTTctcgagttaTAAGTATAAATCAATAGGTGT-3′; PfRh2-297: 5′-agctggatcccGAAAGCTATGTGATGAAC-3′, 5′-agctctcgagttaGCTGGTGTTCAGAATGG-3′; PfRh2-673: 5′-aggaacgatcatcatttgaatggatccaaaatacatg-3′, 5′-CATCTATTATATTTTGTTGTTCTGActcgagttaATATAAGTT-3′; PfRh2-1288: 5′-ATAAATGATTTTggatccGAAAAGAATATATCACAAG-3′, 5′-CTTCATAAAAAAAAGActcgagttaTTTATCTCCATT-3′; PfRh2a: 5′-agctggatcccCACATAAAAAGTAAACTAGAATC-3′, 5′-AgctctcgagttaTGATCGAGAAAAATTTCTATC-3′; and PfRh2b: 5′-AGAAATATCCAAGAAggatccGAGCAAAAAAAG-3′, 5′-CTCCAGCATTATATACctcgagttaCATTTTGTTATG-3′.

Gene fragments for PfRh2a/b-34, -673, and -1288 and PfRh2b were subcloned into pPROEX HTb (Life Technologies, Sydney, Australia). Proteins were expressed as N-terminal (His)₆ fusion proteins in E. coli and affinity purified on NiNTA columns (Invitrogen, Melbourne, Australia), following the manufacturer’s instructions. Insoluble protein (PfRh2-34 and PfRh2-673) was refolded by rapid dilution (0.1 M Tris [pH 8], 0.4 M l-Arginine, 0.2 mM PMSF, 0.5 mM glutathione disulfide, 5 mM reduced glutathione). Residual insoluble protein was precipitated by centrifugation (10,000 × g, 20 min 4°C). All proteins were dialyzed against PBS. Gene fragments for PfRh2-297 and PfRh2a were subcloned into pET45b and were expressed as (His)₆ fusion proteins. Purified PfRh2-297 was refolded by dilution in 2 M urea/20 mM Tris (pH 8)/100 mM NaCl. n-Octyl glucoside (Sigma-Aldrich, St. Louis, MO) detergent (20 mM) was added to dissociate the fusion protein from a contaminating protein. The resulting precipitate contained relatively pure fusion protein and was resuspended in PBS/2 M urea for use in ELISA assays. PfRh2a was expressed as a soluble fusion protein, purified on a NiNTA column, and further purified by size fractionation. PfRh2-2030 and PfRh2-2530 [previously published as PfRh2A9 and PfRh2A11 (19)], were expressed as GST fusion proteins in E. coli and purified on glutathione agarose (Sigma-Aldrich, Sydney, Australia), following the manufacturer’s instructions, and then dialyzed overnight against PBS. All recombinant proteins were run in reducing and nonreducing sample buffer in SDS-PAGE gels (Supplemental Fig. 1) to check for purity and integrity. Coomassie-stained protein bands were cut out and evaluated by mass spectrometry to verify their identity. PfRh2-1288 occurred as two clear bands on reducing and nonreducing gels; both bands were labeled by anti-His tag Abs in Western blots, and analysis by mass spectrometry confirmed that both bands were PfRh2 recombinant proteins. Based on their M_r, the lower molecular mass species seems to have resulted from truncation at the C terminus.

Details of the cohort were described elsewhere (34). Briefly, plasma samples were obtained from a prospective treatment reinfection study of 206 children from Madang Province on the north coast of Papua New Guinea (PNG). The median age at enrollment was 9.3 y (range, 5–14 y; interquartile range, 8.1–10.3 y). Baseline clinical observations and venous sampling were performed. The prevalence of P. falciparum and P. vivax, detected by PCR, at enrollment was 67.5 and 33.9%, respectively. All children received artesunate orally for 7 d to clear parasitemia. Over the subsequent 6 mo, the cohort was followed by active surveillance every 2 wk, including finger-prick blood collection for parasitemia and passive case detection. Merozoite surface protein (MSP)2-based genotyping was used to distinguish between reinfection and treatment failure. A clinical episode of P. falciparum malaria was defined as the presence of fever and parasitemia >5000/μl, but all children with fever and any parasitemia by light microscopy were treated with antimalarials. Plasma samples used for this study were those taken at enrollment, prior to treatment with antimalarials. Sera for positive controls were obtained from adults within the Madang area. Sera for negative controls were obtained from anonymous Australian blood donors who were malaria naive.

Ethics approval was obtained from the Medical Research Advisory Committee, PNG, and the Walter and Eliza Hall Institute of Medical Research Ethics Committee. Written informed consent was obtained from all study participants or their guardians.

Abs to recombinant proteins were measured by ELISAs using established methods (13, 35). Briefly, Ags were coated onto Maxisorp microtiter plates (Nunc, Roskilde, Denmark) at 1–2 μg/ml (schizont lysate at 4.6 μg/ml). Sera samples were tested at 1/500 for total IgG. Polyclonal goat anti-human IgG HRP-conjugated Ab (Chemicon, Melbourne, Australia) was used at 1/5000. For measurement of IgG subclasses, secondary Abs were added at a dilution of 1/1000 using monoclonal anti-human IgG1, IgG2, IgG3, and IgG4 (Zymed, Melbourne, Australia). We used an HRP-conjugated sheep anti-mouse Ab (Chemicon, Melbourne, Australia) at 1/2500 as a tertiary Ab for IgG subclass detection. ABTS liquid substrate (Sigma, Castle Hill, Australia) was used to measure enzymatic reactivity, and the reaction was stopped with 1% SDS. OD was determined at 405 nm. Nine nonimmune serum donors were used as negative controls, and sera from three malaria-exposed adults were used as positive controls for validation and standardization purposes. All sera were tested in duplicate, which was repeated if there was a discrepancy >25% between duplicates. PfRh2-2030 and PfRh2-2530 were expressed as GST-fusion proteins. Therefore, reactivity to GST alone was assessed and deducted from all values for these Ags; however, reactivity to GST was very low or negligible among samples. In addition, we tested for IgG to the NANP repeats of circumsporozoite protein. A sample was classified as Ab⁺ if the OD was greater than the mean plus three SDs of the reactivity observed with nonexposed samples. Data on IgG to MSP1–19, MSP2, and apical membrane Ag 1 (AMA1) were determined previously (35) and included in this study for comparison with PfRh2.

Statistical analysis was performed using STATA 9.2 (STATACorp, College Station, TX). Differences in seroprevalence and IgG levels between categorical variables were assessed using χ² tests or Kruskal–Wallis tests, respectively. Correlations between Ab responses were examined using Spearman’s rho. To determine the association between Ab levels and P. falciparum infection and symptomatic malaria, OD values for each Ag were stratified into three equal groups (tertiles) reflecting low, medium, and high responders. Kaplan–Meier curves were generated for time-to-event (infection or clinical episode) data and compared using the Wilcoxon test. Assumptions of proportional hazards were assessed, and Cox proportional hazards models were used to calculate hazard ratios and evaluate the effects of possible confounders. For Ab variables that showed nonproportional hazards, results from Cox regression including an interaction term between Ab variable and time are reported. Hazard ratios for Ab variables were adjusted by multivariate Cox regression using predefined covariates of age (binary variable: <9 y, ≥9 y) and location of residence, which were reported previously (34, 35). To understand the extent of differences in relative IgG levels to PfRh2 among protected versus malaria-susceptible children, they were stratified into three groups according to their history of symptomatic malaria episodes in the 6-mo period after drug treatment, and median Ab levels between groups were compared. For this analysis, we only included children who had recorded exposure to P. falciparum during follow-up, as determined by a detectable parasitemia by PCR or light microscopy of any density; this included 95% of children. Children who had no detected parasitemia during follow-up were regarded as not exposed and were excluded from this analysis. Children were classified as “protected” if they had no recorded symptomatic P. falciparum malaria and low-grade parasitemia only (parasitemia < 5000 parasites/μl), “susceptible 1” if they had one symptomatic episode (and a parasitemia >5000 parasites/μl), and “susceptible 2” if they two or more symptomatic episodes.

P. falciparum genomic DNA was extracted from the cell pellet of blood samples from children in the cohort using the QIAamp 96 DNA Blood kit (34). A 1662-bp fragment was amplified in a PCR reaction with the following oligonucleotides: Rh2_1aF: 5′-AATGAAGAACAATGTTTAGTTGGTGGGAAAACAG-3′ and Rh2_1R: 5′-TTTCTTTAAGACTTAATAGATGACTTAATTCAG-3′.

PCR fragments were cloned into pCRII-TOPO (TOPO TA Cloning, Invitrogen) and used to transform E. coli (TOP10, Invitrogen). Forward- and reverse-sequencing reactions for several clones were done by the sequencing service of the Applied Genetic Diagnostics Unit, Department of Pathology, University of Melbourne. Single nucleotide polymorphisms were identified by visual inspection of high-quality raw chromatogram data. SNPs were considered valid if they were observed in multiple clones from the same isolate and found in more than one isolate or verified by sequencing a second independent PCR product from the same isolate.

Sequences were submitted to GenBank (http://www.ncbi.nlm.nih.gov/) under new accession numbers HM802474–HM802488.

For comparison with the PNG sequences, we downloaded the 3D7, 7G8, and Dd2 reference sequences (36) from the public database, as well as those from an additional 12 isolates from worldwide origins being sequenced in genome projects at the Broad (U.S.) and Sanger (U.K.) Institutes. Basic Local Alignment Search Tool homology searching identified PfRh2 sequences for Senegal, RO33 and Ghana clinical (Ghana), RAJ116 and ICH-CR14 (India), HB3 (Honduras), IT (Brazil), SL (El Salvador), K1 (Thailand), FCC-2 (China), and FVO and VS/1 (both from Vietnam). Only partial sequences covering the 1600-bp region sequenced were available for isolates K1 and FCC-2. Sequence data for the IT clone and the Ghana field isolate were produced by the pathogen genomics group at the Wellcome Trust Sanger Institute and are available at http://www.sanger.ac.uk/Projects/P_falciparum/. Sequences for all other P. falciparum laboratory clones were generated by the P. falciparum sequencing project at the Broad Institute of Harvard and Massachusetts Institute of Technology (http://www.broad.mit.edu) and are available at http://www.broadinstitute.org/annotation/genome/plasmodium_falciparum_spp/Blast.html. Sequences of the reference lines 3D7, Dd2, FVO, and 7G8 were downloaded from the European Molecular Biological Laboratory nucleotide database (http://www.ebi.ac.uk/embl/; accession numbers AY138500, AY138502, AY138503, and AY138501, respectively). We also included sequences that we generated from recent laboratory-adapted isolates from Thailand [HCS3 (37); http://www.ncbi.nlm.nih.gov/; new GenBank accession number is HM802473] and West Africa [Pf2004 and Pf2006 (38); http://www.ncbi.nlm.nih.gov/; new GenBank accession numbers are HM802472 and HM802471, respectively], giving a total of 18 sequences from diverse geographic origins.

Tajima’s D test was used to test for departure from selective neutrality by comparing S (number of segregating sites) and π (the average pairwise nucleotide diversity). Balancing selection increases π compared with S; thus, the value for Tajima’s D becomes positive. Additionally, Fu and Li’s test was used to test for selective neutrality. It compares η_s, the number of singletons, with η, the total number of mutations, expressed as D*, or η_s and π (the average number of nucleotide differences between pairs of sequences), expressed as F*. The average evolutionary divergence over all sequence pairs was estimated using the Tamura three-parameter model, which corrects for multiple hits and takes into account differences in transitional and transversional rates and guanine and cytosine content bias. Codon-based tests for neutrality or positive selection were performed using the Nei–Gojobori method (39). All tests were performed using DNASp5.00.04 (40) or MEGA v4 (41). Cluster analysis of amino acid haplotypes was done using Structure v. 2.2 software (42, 43). Structure 2.2 assigns individual multilocus haplotypes probabilistically to a user-defined number of clusters (K) (43). For PfRh2 amino acid haplotypes, Structure was run 20 times for K = 1–10 for 10,000 Monte Carlo Markov Chain iterations after a burn-in period of 10,000 (44) using the admixture model and correlated allele frequencies. The estimated value of K was defined by calculating ΔK, according to the method of Evanno et al. (44). Ancestry coefficients were then plotted for each PfRh2 sequence.

To determine the acquisition of Abs to PfRh2, we expressed consecutive regions of nearly the entire PfRh2a and PfRh2b ectodomain (Fig. 1). In total, we expressed eight proteins: five proteins covered the common region shared by the full-length (fl-) PfRh2a and PfRh2b (referred to as PfRh2-34, PfRh2-297, PfRh2-673, PfRh2-1288, and PfRh2-2030; numbers indicate starting amino acid); one protein covered the last 246 aa of the common region and the first 256 aa of the PfRh2a-specific region (PfRh2a-2530); and proteins representing each of the PfRh2a-specific and PfRh2b-specific sequences (referred to as PfRh2a and PfRh2b throughout this paper). Analysis of total IgG among the cohort samples showed that all proteins were recognized by the majority of serum samples (Table I), whereas there was little reactivity among nonexposed Melbourne residents. Seroprevalence ranged from 66.5% for PfRh2-673 to 94.2% for PfRh2-2030 and PfRh2-2530. Median OD was highest for PfRh2-2030, PfRh2-2530, and PfRh2a (Table I). Among the entire cohort, 123 individuals (60.6%) were seropositive for all PhRh2 proteins tested, and only 2 individuals (0.99%) were seronegative to all PfRh2 proteins.

IgG levels were significantly higher in children who had PCR-detectable parasitemia at the time of sample collection for all PfRh2 proteins tested (Table I; p ≤ 0.001). Despite a restricted age range in the cohort (median age, 9.3 y; range: 5–14 y), higher IgG levels or higher prevalence of Abs were observed among older children (>9 y of age) compared with younger children (≤9 y of age) for all proteins, which was significant for PfRh2-34, PfRh2-1288, PfRh2-2030, PfRh2-2530, and PfRh2b (Table I). IgG levels to all rPfRh2 proteins were positively correlated with IgG to P. falciparum schizont protein extract, used as a marker of exposure to blood-stage infection (p ≤ 0.001; Table II). This observation, together with the acquisition of Abs with age, higher IgG levels in parasite-positive children, and lack of reactivity among samples from nonexposed donors, suggest that the responses measured are malaria specific.

There was a modest to strong correlation between Abs against a particular protein and Abs against the adjacent proteins (Table II). The highest correlation was found between Abs against PfRh2-2030 and PfRh2-2530 (Spearman’s rho = 0.844; p < 0.001). Correlations between Abs against N-terminal proteins PfRh2-34 and PfRh-297 were less strong but were still significant (Spearman’s rho = 0.470; p < 0.001).

The design and longitudinal nature of the cohort study, incorporating initial antimalarial treatment and active follow-up with the application of sensitive molecular methods to detect P. falciparum infections, enabled us to investigate the relationship between levels of PfRh2-specific Abs (high, medium, and low) and reinfection, parasitemia of different densities, and symptomatic malaria in the same cohort of children.

Comparing children with the highest versus lowest Ab levels showed that IgG against PfRh2 was associated with a significantly decreased risk for symptomatic malaria (defined as parasitemia >5000 parasites/μl plus fever) for all eight PfRh2 proteins (Table III). Even after adjustments for age and spatial confounders, high IgG responses to all PfRh2 proteins showed a >50% reduction in risk for symptomatic malaria (Table III). Protective associations were strongest for PfRh2-673, PfRh2-2030, and PfRh2a (adjusted hazard ratio [aHR], 0.25, p < 0.001; aHR, 0.2, p = 0.012; and aHR, 0.27, p = 0.01, respectively). Of note, aHRs for all PfRh2 proteins were lower than or comparable to those for 3D7-AMA1 (aHR, 0.39; 95% confidence interval [CI]: 0.143–1.089; p = 0.073), 3D7-MSP1-19 (aHR, 0.60; 95% CI: 0.240–1.505; p = 0.276), MSP2-3D7 (aHR, 0.56; 95% CI: 0.211–1.489; p = 0.246), and MSP2-FC27 (aHR, 0.72; 95% CI: 0.263–1.994; p = 0.532), which are leading blood-stage vaccine candidates. Furthermore, Abs to the pre-erythrocytic Ag circumsporozoite protein or the control protein GST were not associated with risk for malaria (data not shown).

We also examined associations between Abs and reinfection with a parasitemia of any density (as detected by PCR or light microscopy) or episodes of moderate (>500 parasites/μl) or high-density (>5000 parasites/μl) parasitemia. High Ab levels to all PfRh2 proteins were significantly associated (p < 0.05) with protection against high-density parasitemia (>5000 parasites/μl), except for PfRh2-297, which was of borderline significance (hazard ratio [HR], 0.51; p = 0.065). Abs to PfRh2-673 (HR, 0.39; p = 0.002), PfRh2-2030 (HR, 0.44; p = 0.023), and PfRh2-34 (HR, 0.45; p = 0.005) showed the strongest association with protection from high-density parasitemia. After adjusting for covariates, a high level of IgG to PfRh2-673 (aHR, 0.42; p = 0.005) and PfRh2-2030 (aHR, 0.46; p = 0.034) remained strongly predictive of protection against high-density parasitemia, whereas associations for PfRh2-2530 (aHR, 0.53; p = 0.051) and PfRh2a (aHR 0.52; p = 0.056) were of borderline significance, and the aHRs for PfRh2-34, PfRh2-297, PfRh2-1288, and PfRh2b were no longer significant. No significant associations were observed between high levels of Abs against any PfRh2 protein and protection from parasitemia detected by PCR or light microscopy (p > 0.2) or against moderate-density parasitemia (>500/μl; p > 0.19). Fig. 2 shows representative examples of Kaplan–Meier survival analyses for two of the PfRh2 proteins (PfRh2-673 and PfRh2-2030) that showed the strongest association with protection against clinical malaria. These demonstrate the protective effect of Abs against clinical disease (Fig. 2E, 2F) and high-density parasitemia (Fig. 2C, 2D) but not against reinfection per se (Fig. 2A, 2B).

To further understand the extent of differences in relative IgG levels to PfRh2 among protected versus malaria-susceptible children, we stratified children into three groups according to their history of symptomatic malaria episodes in the follow-up period. Children were classified as “protected,” “susceptible 1” (one symptomatic episode), and “susceptible 2” (at least two symptomatic episodes). IgG levels were highest among the protected group of children compared with susceptible children (Table IV). The difference in median Ab values was 1.3–2.5-fold higher for PfRh2a and PfRh2-673, respectively, in protected children compared with susceptible group 2.

Prior studies suggested that IgG1 and IgG3 are the predominant subclasses that respond to merozoite Ags (35, 45, 46, 47); however, to the best of our knowledge, no study has examined IgG subclass responses to any PfRh protein. IgG1 and IgG3 are cytophilic Abs with a high potential to activate complement and with high affinity for FcγRs, thus having the ability to activate T cell-mediated cytotoxicity and Ab-dependent phagocytosis. To gain insights into the potential function of Abs to PfRh2, we evaluated IgG subclasses to two representative proteins: PfRh2-2030 and PfRh2-2530. We observed predominantly IgG1 and IgG3 subclass responses to PfRh2, with very little IgG2 and IgG4 detected (Supplemental Table I). The seroprevalence of IgG1 against PfRh2-2030 and PfRh2-2530 was higher compared with IgG3 (85 versus 63.1% for PfRh2-2030; 85 versus 78.6% for PfRh2-2530). IgG subclass responses also showed increased prevalence and levels in association with age, concurrent parasitemia, and high schizont IgG levels at enrollment, as were observed for total IgG (Supplemental Table I).

Examining correlations between IgG subclass responses suggested that the same subclass response to the two different PfRh2 proteins was more strongly correlated than different subclass responses to the same Ag (Supplemental Table II). For example, IgG1 to PfRh-2030 versus PfRh2-2530 was more strongly correlated (Spearman’s rho = 0.846; p < 0.001) than the correlation between IgG1 and IgG3 levels to the same Ag (Spearman’s rho = 0.460; p < 0.001 and Spearman’s rho = 0.570; p < 0.001 for PfRh2-2030 and PfRh2-2530, respectively). The same was observed for the correlation between IgG3 responses to the different PfRh2 Ags (Spearman’s rho = 0.725; p < 0.001). IgG2 and IgG4 against each of the proteins were less strongly correlated (Spearman’s rho = 0.429; p < 0.001 and Spearman’s rho = 0.487; p < 0.001, respectively). There was a moderate correlation between IgG2 and IgG4 against PfRh2-2030 (Spearman’s rho = 0.353; p < 0.001) as well as between IgG2 and IgG4 against PfRh2-2530 (Spearman’s rho = 0.552; p < 0.001).

IgG1 and IgG3 against PfRh2-2030 and PfRh2-2530 were associated with a reduced risk for symptomatic malaria episodes when comparing children with high versus low Ab levels (Table V). The strongest association was for IgG3 to PfRh2-2530 (HR = 0.21; p = 0.005). Although the reduction in risk was greater for IgG3 (79% reduction) versus IgG1 (66% reduction) to PfRh2-2530, it was similar for IgG1 and IgG3 to PfRh2-2030 (70 and 64% reduction, respectively). After adjustment for age and location, HRs for IgG1 and IgG3 against PfRh2-2530 remained significantly reduced, whereas only IgG1 remained significant for PfRh2-2030. There was some evidence of a protective association for IgG2 against PfRh2-2530 (aHR, 0.38; p = 0.064). However, because levels of PfRh2-2530–specific IgG2 were very low (median OD, 0.022; interquartile range: 0.006–0.067, Supplemental Table I), it is unclear whether this association is biologically significant. No significant associations were observed for IgG4. There is evidence that high levels of IgG1 against PfRh2-2530 are also protective against high-density parasitemia (p = 0.008). High levels of IgG3 showed an association with protection against high-density parasitemia with borderline significance (p = 0.06). There were no significant associations between Abs of any IgG subclass and risk for low-grade parasitemia or light microscopy- or PCR-detectable parasitemia (i.e., reinfection per se; data not shown).

Ab levels among children who were classified as protected were significantly higher for IgG1 and IgG3 against PfRh2-2530 as well as for IgG3 against PfRh2-2030 compared with children who had malaria episodes; levels were increased 1.63–3.75-fold in protected children compared with those who had two or more episodes (Supplemental Table III). No significant differences were found for IgG2 or IgG4. This further suggests that IgG1 and IgG3 are the main contributors to the protective effect of PfRh2-specific Abs.

To obtain further evidence that PfRh2 may be a target of protective immunity, we generated PfRh2 gene sequences from the P. falciparum isolates derived from the participants of the study cohort. Previous sequence analysis of four reference isolates (3D7, 7G8, Dd2, FVO) that were available in the public database suggested that polymorphisms are concentrated in the N-terminal region of PfRh2 and that this region might be under diversifying selective pressure (36). To determine the number of SNPs and the rate and ratio of synonymous versus nonsynonymous changes within this region, we amplified, cloned, and sequenced an N-terminal fragment of PfRh2 that corresponded to the most polymorphic region based on analysis of P. falciparum reference isolates. This sequence is contained within the PfRh2-34, PfRh2-297, and PfRh2-673 recombinant proteins (Fig. 1).

We obtained 15 PfRh2 sequences from genomic DNA samples obtained from 12 infected individuals in the cohort, including four different sequences from one individual. Among these gene sequences, we identified a total of 20 polymorphic sites; 19 were nonsynonymous. For comparison, we also obtained sequences from 18 isolates from different worldwide locations. Among these, we identified 28 nonsynonymous polymorphisms and 1 synonymous polymorphism. Two of the nonsynonymous polymorphisms were contained within the same codon, which resulted in coding for three different amino acids. The synonymous site resulted from a third nucleotide variant at one position found only in two isolates (HCS3 and RAJ116), in addition to the more common nonsynonymous variant at the same site. Combining the PNG and worldwide datasets, there were 35 nonsynonymous (located within 34 codons) and 2 synonymous polymorphisms; only 12 polymorphisms were shared between the two sets of isolates, and all of these were nonsynonymous. Therefore, there were many polymorphisms unique to each dataset, suggesting that there may be more polymorphisms still to be sampled for this gene. A multiple alignment of the DNA sequences is shown in Supplemental Fig. 2. By contrast, the 1497-bp gene fragment coding for PfRh2-2030 further downstream within the PfRh2 gene showed markedly less diversity compared with the N-terminal region with only one nonsynonymous and one synonymous polymorphism among 10 isolates (RAJ116, K1, IGH-CR14, HB3, FVO, Dd2, 7G8, 3D7, IT, Ghana; data not shown).

To investigate whether the pattern of polymorphisms in PfRh2 N-terminal coding region were consistent with immune selection, we determined the number of nonsynonymous changes per nonsynonymous site (d_n) and the number of synonymous changes per synonymous site (d_s) among the 15 sequences from PNG. We found that ω, the ratio between d_n and d_s, was positive (4.95), pointing toward diversifying selection (Table VI). Neither Fu and Li’s D* and F* statistics (0.89 and 1.021, respectively) nor Tajima’s D (1.021) reached significance, most likely as a result of the limited sample size. Therefore, a more sensitive sliding window analysis was performed, which revealed highly positive values for Tajima’s D and Fu and Li’s statistics at various window midpoints, suggesting a departure from neutrality in certain regions of the analyzed sequence (Fig. 3). We also compared d_n and d_s using a codon-based test (Z-test of selection, using the Nei–Gojobori method, p-distance model), and found a significant departure from neutrality (p = 0.014) in favor of positive selection (p = 0.008). Together, these findings point toward diversifying selective pressure acting on the N-terminal region of PfRh2 proteins in clinical isolates among the cohort, consistent with it being a target of protective immune responses.

If the PfRh2 N-terminal domain is a target of immune selection, we expect high levels of diversity to be found within the PNG parasite population, with important implications for immunity and its development as a vaccine candidate. Among the 15 PNG PfRh2 sequences, there were 13 haplotypes, with only two pairs of PNG isolates sharing the same haplotype, showing an extremely high degree of haplotype diversity for PfRh2 in this parasite population. Because there are a variable number of polymorphisms among different haplotypes (Supplemental Fig. 2), and this is likely to be associated with the degree of antigenic diversity, we also performed a Bayesian cluster analysis on the combined dataset of 33 PfRh2 sequences (28 haplotypes) to identify groups of similar haplotypes. The distribution of polymorphisms among sequences was most consistent with two major clusters, each with a broad geographic distribution (Fig. 4). Both clusters were found in PNG, suggesting that the full spectrum of diversity is found in this local sample. Importantly, several PNG haplotypes clustered with the 3D7 haplotype, showing that 3D7-like sequences are present within the PNG parasite population. Thus, the measurement of Abs using 3D7-derived PfRh2 protein seems to be appropriate for this cohort.

An important and major finding of this study was that Abs to PfRh2 were strongly associated with a reduced risk for symptomatic malaria using a prospective study design. Remarkably, this association was observed for all eight PfRh2 protein fragments, and associations remained significant after adjusting for age and location for all proteins, except PfRh2-297, which was of borderline significance. Furthermore, the strength of association between PfRh2 total IgG and protection from malaria was similar to or greater than that for the leading vaccine candidates MSP1-19, MSP2, and AMA1. We extended our analysis by comparing Ab levels among children who did or did not have episodes of malaria during the 6-mo follow-up period. Consistent with the survival analysis, Ab levels were significantly higher among protected children than among those who had malaria episodes.

An important question in understanding human immunity to malaria is whether Abs to merozoite Ags contribute to protection against parasitization per se or only act to prevent symptomatic malaria. Our results demonstrate that Abs to PfRh2 are significantly associated with protection from high-density parasitemia and symptomatic malaria but not against parasitization per se. Very few studies have examined associations of immune responses with risk for reinfection and clinical malaria in the same study (7). To address these important questions in our study, we cleared parasitemia among participants at enrollment with a highly effective treatment (>90% cure rate), actively screened for reparasitization during follow-up using sensitive molecular methods, and genotyped infecting parasites to distinguish reinfection from treatment failure (34). No PfRh2 IgG or subclass response was associated with a reduced risk for reinfection as detected by PCR or light microscopy. Instead, our results suggest that Abs to PfRh2 mediate their protective effect by control of blood-stage parasitemia, thereby preventing high-density infections. High-density parasitemia and overall parasite biomass are linked with the pathogenesis of severe malaria (48). Therefore, these Abs may contribute to protection from severe malaria. Abs to PfRh2 would be expected to function, in part, by inhibiting erythrocyte invasion and blood-stage parasite replication because Abs raised against PfRh2 in laboratory animals have growth-inhibitory activity in vitro (27). Our findings also demonstrated antigenicity across the entire PfRh2a and PfRh2b proteins. Abs to all PfRh2 proteins were higher among those with active parasitemia at the time of testing and were generally higher among older children, and Abs correlated with Abs to schizont protein extract, used as a marker of exposure to blood-stage infection. In contrast, there was little or no reactivity among samples from nonexposed donors, and control proteins, such as GST, showed little reactivity among cohort samples. A single prior study showed age-associated acquisition of Abs to PfRh2-2030 among a Kenyan population (13), but other regions of PfRh2 were not studied.

Analysis of IgG subclass responses revealed a predominance of IgG1 and IgG3, as found for other Ags in this cohort and more broadly (35, 46, 47, 49, 50). A significant age-associated increase in IgG1 and IgG3 levels and associations with parasitemia and exposure (reactivity to schizont extract) point toward malaria-specific responses. IgG1 and IgG3 against PfRh2-2030 and PfRh2-2530 were significantly associated with protection against clinical malaria. Although IgG2 to PfRh2-2530 was associated with protection, the low levels of IgG2 detected suggest that the contribution of these Abs to immunity is small. Examining correlations between IgG subclass responses revealed that the same subclass response to the two different PfRh2 proteins was more strongly correlated than were different subclass responses to the same Ag. This suggests that subclass responses among individuals may be related and that there is a tendency for individuals to produce a similar IgG subclass response to different Ags (35). The cytophilic nature of IgG1 and IgG3 suggests that Abs to PfRh2 may interact with macrophages, granulocytes, and/or NK cells or activate complement to mediate parasite clearance.

Analysis of PfRh2 sequence diversity provided additional evidence that PfRh2 is an important target of protective immunity. Prior analysis of four reference lines indicated that polymorphisms are predominantly located in an N-terminal region of the protein (36). Our sequence analysis demonstrated significant polymorphism in this region and suggested it is under diversifying selection, presumably resulting from immune pressure. Many polymorphic residues were common among isolates from different geographic regions, although some polymorphisms were restricted to specific isolates and, thus, geographic locations. Codon-based tests for neutrality and positive selection were in favor of diversifying selection. Although values for Tajima’s D, Fu and Li’s D*, and Fu and Li’s F* were not significantly associated with departure from neutrality, the sliding-window analysis showed areas with elevated values. This is further supported by the excess of nonsynonymous changes compared with synonymous changes and a resulting high d_n/d_s ratio observed within all P. falciparum sequences that we analyzed in this study. The average pairwise nucleotide diversity (π = 0.00477) was greater than that reported for EBA175 region II [π = 0.003 (51); π = 0.004 (52)] but less than AMAI domains I, II, or III or the entire sequence [π = 0.025, 0.006, 0.014, and 0.014, respectively (53)]. By contrast, region II of EBA140 and EBA181 reportedly have limited diversity (π = 0.001 and 0.0003, respectively) and no evidence for selection (51, 52). Consistent with these findings, when we investigated how the sequence diversity was distributed, we found that the majority of parasite isolates had unique PfRh2 haplotypes. The identification of only two main sequence clusters shows that many of these PfRh2 haplotypes may be antigenically similar, and both of these were found among the PNG PfRh2 sequences that we obtained. This pattern of high within-population diversity for PfRh2 in this study was also found in several other merozoite Ags known to be important targets of host immunity (38, 54). Merozoite Ags seem to have little regional diversification, but diversification has been observed for other Ags (38, 54). Our analysis did not clarify whether the observed changes occurred in PfRh2a or PfRh2b because the polymorphic region is contained within the common region of PfRh2. The 14 nonsynonymous changes described previously in the reference lines were shown to be identical in PfRh2a and PfRh2b, and it was suggested that this might be due to continual gene conversion (36). Therefore, we consider it safe to assume that the changes that we observed are the same in PfRh2a and PfRh2b.

Our findings suggest that PfRh2 is an important target of protective immune responses and support PfRh2 as a potential vaccine candidate. We found that the protein is naturally immunogenic, and Abs are associated with protection from malaria and high-density parasitemia. Sequence diversity also supports PfRh2 as an important target of immunity, and there is some evidence that the PfRh ligands may be targets of human inhibitory Abs (13). Importantly, Abs raised against the PfRh2-2030 protein in rabbits were shown to inhibit invasion (27), and Abs to this region were strongly associated with a reduced risk for malaria in our study. As is the case for many P. falciparum merozoite Ags, there are no proteins that are similar to PfRh ligands in structure and function in rodent malarias. Therefore, responses to PfRh ligands cannot be easily studied in animal models, emphasizing the importance of human studies to evaluate these Ags as targets of protective immunity and to aid in the identification and prioritization of candidates for vaccine development.

We thank all study participants and Papua New Guinea Institute of Medical Research staff involved in the study and Danny Wilson and Jack Taraika for assistance with sample processing. We thank Jake Baum and John Donelson for helpful discussions regarding sequence analysis. We also thank the Broad Institute and the Sanger Institute for permission to use the PfRh2 sequence data.

Disclosures The authors have no financial conflicts of interest.