ABSTRACT
Plasmodium falciparum erythrocyte membrane protein 1 (PfEMP1) is a variant antigen expressed on the surface of malaria-infected erythrocytes. PfEMP1 attaches to the vascular lining and allows infected erythrocytes to avoid filtration through the spleen. Each parasite genome encodes about 60 different PfEMP1 variants, each PfEMP1 comprises several domains in its extracellular region, and the PfEMP1 repertoire in different parasites contains domain types that are serologically cross-reactive. In this longitudinal study, we followed 672 children living in an area of high malaria transmission during the first years of life and compared the acquisitions of antibodies to 32 Duffy-binding ligand-like (DBL) domains representing different types. For each child, we determined whether an antibody response to each domain was acquired before, after, or at the same time as responses to each of the other domains. We next used this information to calculate population-level odds ratios to measure the odds that antibodies to a given domain were acquired before antibodies to other domains. Odds ratios for 269 of the 496 possible domain combinations were statistically significant. Thus, the sequence in which individuals acquire antibodies to different PfEMP1 domains is ordered, and children in areas of endemicity first acquire antibodies to particular PfEMP1 domains encoded by the so-called group A and B/A var genes. The results imply that anti-PfEMP1 antibodies effectively structure PfEMP1 expression and play a major role in limiting parasite multiplication in the blood.
The parasite-encoded Plasmodium falciparum erythrocyte membrane protein 1 (PfEMP1) family is expressed on the surface of infected erythrocytes (IEs) (25). PfEMP1s are large, 200- to 350-kDa multidomain proteins encoded by a repertoire of about 60 var genes per parasite genome (13, 38) and contain a large extracellular polymorphic part consisting of up to 10 Duffy-binding ligand-like (DBL) and cysteine-rich interdomain region (CIDR) domains, a transmembrane domain, and a conserved intracellular part anchoring the protein in the IE. DBL and CIDR domains can be subclassified by phylogeny into DBLα0, -α1, -β, -γ, -δ, -ε, and -ζ and CIDRα0, -α1, -β, and -γ. PfEMP1s can be grouped into three main groups (A, B, and C) and two intermediate groups (B/A and B/C) relating to the DBLα type and number of domains they carry (22, 24). Thus, group A and B/A PfEMP1 proteins most often contain DBLα1 and more than four domains, whereas group B, B/C, and C proteins contain DBLα0, most often in a four-domain structure.
The large, extracellular part of PfEMP1 mediates adhesion of IEs to a number of different host vascular receptors (1, 37), leading to tissue-specific sequestration of the parasites, which thereby avoid passage through the spleen, where they would otherwise be destroyed (17). Sequestration is undoubtedly an important element in malaria pathogenesis (6, 11, 28, 31), and acquisition of PfEMP1-specific antibodies therefore appears to be a critical element in naturally acquired protective immunity to the disease (5, 26, 27, 36). Interestingly, acquisition of clinical immunity to severe complications precedes protection from uncomplicated disease and asymptomatic parasitemia (2, 14). Parasites isolated from severe malaria cases and patients with limited immunity tend to express PfEMP1 variants encoded by group A and B/A var genes (18, 34, 41). These findings suggest that the order of acquisition of antibodies with specificity for different PfEMP1 domains is nonrandom (20) and that identification of domains to which immunity is acquired early could provide important leads in the development of PfEMP1-based vaccines against severe P. falciparum malaria.
In a recent paper, we analyzed samples collected from children of different age groups during cross-sectional surveys of villages located in areas of different malaria endemicities. Using a high-throughput assay, we showed that the acquisition of antibodies to PfEMP1 was structured so that antibodies to some group A and B/A DBL domains consistently were developed first (8). In the present study, we analyzed the IgG reactivities to 32 DBL domains in 1,274 samples collected during a longitudinal study following infants and toddlers exposed to conditions of high-level malaria transmission in Muheza, Tanzania. This study design allowed us to investigate whether antibodies to a particular DBL domain was more likely to be developed before or after antibodies to the other domains. The results showed that the acquisition of anti-PfEMP1 antibodies was remarkably structured, and for 269 of 496 domain comparisons, there was a statistically significant likelihood of a response to one of the domains occurring before a response to the other. The results indicate that preexisting anti-PfEMP1 antibodies shape the expression of PfEMP1 during subsequent infections and that these antibodies play an important role in controlling parasite growth.
MATERIALS AND METHODS
Study population.The study population included mother-infant pairs enrolled in the Mother-Offspring Malaria Studies (MOMS) project conducted in the Muheza district, northeastern Tanzania, an area of intense malaria transmission. The study population has been described previously (15, 29). Briefly, parturient women between the ages of 18 and 45 years and with no evidence of chronic or debilitating illness, such as recent significant weight loss or chronic diarrhea, were invited to enroll themselves and their newborns in the study. Written informed consent was obtained from the mothers prior to enrollment.
Children were seen at birth, at 2-week intervals during infancy, and at 4-week intervals postinfancy, as well as at the time of any illness, for full clinical evaluation by the clinician of the research team and blood smear analysis. Venous blood samples were collected when the children were 3 and 6 months of age and every 6 months thereafter, as well as at the time of illness, and processed to recover plasma.
Protocols for procedures used in this study were approved by the International Clinical Studies Review Committee of the Division of Microbiology and Infectious Diseases at the U.S. National Institutes of Health, and ethical clearance was obtained from the institutional review boards of the Seattle Biomedical Research Institute and the National Institute for Medical Research in Tanzania.
Protein expression.Thirty-two recombinant proteins representing DBL domains present in different PfEMP1s from P. falciparum 3D7 were expressed. The domains were chosen to represent PfEMP1s belonging to different groupings (groups A, B, B/A, C, and B/C), DBL domains of different types (α, β, γ, δ, ε, or ζ), and domains located in different parts of the proteins (see Fig. 1). Protein expression was performed essentially as described previously (7, 16). Primer pairs were designed to contain restriction enzyme sites (Table 1) and used to amplify DBL domain-encoding fragments from 3D7 genomic DNA. The digested PCR products were cloned into the baculovirus vector pAcGP67-A (BD Biosciences), designed to contain a V5 epitope upstream of a histidine tag in the C-terminal end of the construct. The identity of the cloned fragments was verified by sequencing. Linearized Bakpak6 baculovirus DNA (BD Biosciences-Clontech) was cotransfected with pAcGP67-A into Sf9 insect cells for generation of recombinant virus particles. Histidine-tagged proteins secreted into the supernatant of infected High-Five insect cells were purified on Co2+ metal-chelate agarose columns. Eluted products were dialyzed overnight against phosphate-buffered saline (PBS) and verified by SDS-PAGE and Western blotting using anti-V5 antibodies. Based on these assays, the purity of the recombinant proteins varied between 79 and 90%.
Description of recombinant proteins used in the study and the overall serorecognition of these domains in the study population. The DBL domains used for the analysis are highlighted, and the PfEMP1s to which they belong are annotated and grouped according to reference 24. Also shown is the percentage (and 95% CI [Conf. Interval]) of children with a measurable IgG response to each of the 32 DBL domains. Plasma samples from 72 children at 24 and 100 weeks of age were tested. McNemar's test was used to test if a significant number of individuals had acquired an antibody response to a given domain between the two sampling points. Resp., responders.
Primers used for cloning of var gene domainsa
Covalent coupling of recombinant PfEMP1 proteins to microspheres.Carboxylated Luminex microspheres were covalently coupled with the different PfEMP1 domains according to the manufacturer's protocol. Each protein was coupled to a particular fluorescence-coded microsphere type. In each case, the microspheres (1.25 × 107/ml) were washed repeatedly in distilled water, activated in NaH2PO4 buffer, reacted with 1-ethyl-3-[3-dimethylaminopropyl]carbodiimide hydrochloride and N-hydroxysulfosuccinimide (Pierce Biotechnology, Rockford, IL), washed twice in 2-[N-morpholino]ethanesulfonic acid, and covalently linked to the target protein (100 μg/ml microsphere suspension). Equal volumes of microspheres coupled with individual domains were pooled, mixed, split into single-use aliquots, lyophilized in polypropylene vials, sealed under nitrogen gas, and stored at −80°C (7).
Luminex analysis of plasma samples.The analysis of plasma DBL domain-specific IgG levels was as described previously (7). In brief, lyophilized microspheres were reconstituted with distilled water immediately prior to use and diluted 1:333 in assay buffer E (ABE; 0.1% bovine serum albumin [BSA], 0.05% Tween 20, and 0.05% sodium azide in PBS, pH 7.4). Aliquots (50 μl) were dispensed into the wells of 1.2-μm filter-bottom 96-well microtiter plates (MSBVS 1210; Millipore) prewetted with ABE and washed three times with ABE by using a vacuum manifold (Millipore). Frozen plasma samples were thawed at room temperature, mixed by vortexing, and spun (16,000 × g for 5 min) to remove particulates. Plasma samples were diluted 1:80 in ABE and 50-μl aliquots added to the microsphere wells. After incubation in the dark on a shaking platform (30 s at 1,100 rpm followed by 30 min at 300 rpm), the plates were washed three times in ABE to remove unbound antibody. Biotinylated anti-human IgG (Sigma) antibody (25 μl/well at 1:500 dilution) was added to the microspheres, which were incubated and washed as described above. This was followed by use of streptavidin-phycoerythrin (Sigma) (50 μl/well at 1:500), incubation in the dark with shaking (30 s at 1,100 rpm followed by 10 min at 300 rpm), and washing as described above. Finally, the microspheres were resuspended in 125 μl of ABE and analyzed on a Luminex 100 IS instrument set to read a minimum of 100 microspheres per microsphere region. Antibody levels for each domain were expressed as median fluorescence intensities (MFI). We used a negative cutoff value of 249 MFI units, based on the reactivity measured in plasma from unexposed control donors.
Statistical analysis.The purpose of this study was to test the hypothesis that the acquisition of antibodies to PfEMP1 domains is structured and that individuals are likely to develop antibodies to certain domains before others. This was tested in a pairwise comparison of the 32 domains. The results from the longitudinal set of samples from each individual were inspected to establish whether the response to domain x occurred before or after the response to domain y. It was then scored (1, 0) if the response to x occurred before the response to y or (0, 1) if the response to x occurred later than the response to y. If the responses to the two domains occurred at the same point in time or there was no response to any of the domains in any of the individuals samples, the score was (1, 1) or (0, 0), respectively. The scores for all individuals were then compiled, and the number of individuals scoring (1, 0) was compared to the number of individuals scoring (0, 1) with McNemar's test for matched pairs. The results were computed as odds ratios and 95% confidence intervals (CI) for the likelihood that the response to domain x occurs before the response to domain y and a two-tailed P value for the likelihood that the odds ratio is equal to 1. If in all individuals the response to domain x occurred before the response to domain y or vice versa, an odds ratio could not be computed and the odds ratio was scored as high or low, respectively. Sorting of data and calculations were performed with Excel. To reduce the risk of obtaining spurious significant results, we first tested the hypothesis that antibodies to the highest-ranking domain (domain 20) were acquired before antibodies to the other domains were acquired and then proceeded to test whether antibodies to the second-highest-ranking domain (domain 69) were more likely to be acquired before antibodies to the lower-ranking domains, and so on. With this strategy, P values lower than 0.05 were considered significant until a test had shown a P value higher than 0.05 (which occurred when domain 49 was compared to domain 11), and then P values lower than 0.01 were considered significant until the second nonsignificant association occurred (comparison of domains 11 and 211). Hereafter, only associations associated with two-tailed P values lower than 1 × 10−4 were considered to be statistically significant.
From 72 children, samples were available from weeks 24 and 100. McNemar's test was used to test if a significant number of individuals had acquired an antibody response to a given domain between the two sampling points.
RESULTS
Verification of cohort appropriateness.Plasma samples for this study were obtained from children 5 to 142 weeks of age. To ascertain that this period adequately covered the period of acquisition of antibodies to our PfEMP1 domain constructs, we first compared the levels of antibody against our 32 PfEMP1 domain constructs in the 72 children, in cases in which samples obtained when the children were 24 weeks and 100 weeks of age were available. The percentage of children who had seroconverted by week 100 was higher than the percentage who had seroconverted by week 24 for most constructs, although the difference was statistically significant for four (Fig. 1). The domains in Fig. 1 are largely ordered according to how often they were recognized in a previous study (8).
Similar to previous observations (8), there was a statistically significant association between the age of a child and the number of PfEMP1 domains recognized by plasma antibodies (P = 0.01; linear regression for longitudinal data by using Stata).
Analysis of temporal order of acquisition of PfEMP1 domain-specific antibodies.Having verified the suitability of the antigen constructs and the relevance of the study period covered (Fig. 1), we proceeded to analyze the order of seroconversion in more detail, employing all 1,274 samples from all 672 individuals. The number of samples available for each infant/child ranged from 1 to 9. Five hundred ninety-five children seroconverted to at least one of the domains during the study, whereas 77 (245 samples) did not seroconvert to any of the constructs during the period of study. The results of our analysis are summarized in Fig. 2. Construct 20 (DBLγ of the group A PfEMP1 variant PF11_0008) topped the list, as seroconversion to this construct was very likely to occur before seroconversion to any of the other constructs, with odds ratios ranging from 2.5 to >254. Antibodies to construct 69 (DBLε of MAL6P1.4) were more likely to be acquired before antibodies to the third-highest-ranking domain, domain 22 (DBLδ of PF11_0008) (odds ratio, 1.9 [95% CI, 1.4 to 2.7]; P = 0.00028). Construct 22 (DBLδ of PF11_0008), in turn, was more likely to be recognized before the fourth-highest-ranking domain, domain 49 (DBLβ of PFL0020w) (odds ratio, 1.5 [95% CI, 1.1 to 2.2]; P = 0.0269). Nine of the 10 constructs most likely to show early seroconversion belonged to group A or group B/A, compared to 4/10 constructs least likely to show early seroconversion.
Odds ratios for the pairwise comparison of which children were most likely to seroconvert to one DBL domain before another. A total of 1,274 samples from 672 children were tested. For each donor and pair of constructs (x, y), the numbers of children who seroconverted to x before y [NBEFORE(x,y)] and after [NAFTER(x,y)] was counted. Odds ratios and levels of statistical significance were calculated using McNemar's test on paired proportions. Data with two-tailed P values lower than 1 × 10−4 are shown in red, those with P values between 1 × 10−4 and 1 × 10−2 are shown in orange, and those with P values between 0.05 and 0.01 are shown in green. In cases where either NBEFORE(x,y) or NAFTER(x,y) was zero and odds ratios therefore could not be calculated, results were simply recorded as high (H) or low (L), respectively.
As odds ratios were statistically significant for 269/496 comparisons, the order of acquisition of antibodies to the different domains was highly structured. For many of the remaining comparisons, the odds ratio was high but did not reach statistical significance (see supplemental material), because very few children acquired a response to either of the two domains being compared. Early seropositivity could reflect maternal transfer of antibodies rather than early acquisition. To verify that our results reflect a difference in acquisition of antibodies rather than a difference in maternal transfer of antibodies, we repeated the statistical analyses for domain 20 but included only results of samples collected at week 48 or later. Domain 20 was used for this analysis because it was the first recognized and results from this domain thus were most likely to be compounded by maternal transfer. For all comparisons and excluding samples collected before week 48, domain 20 was more likely to be recognized before any of the other 31 domains, with an odds ratio of ≥2.5 and those scored as high (see above) (all P values, <0.0001).
DISCUSSION
This study was performed in an area where malaria transmission is high and where infants and children are expected to be exposed to several infectious mosquito bites every week (10). Under such circumstances, infants experience only a few malaria episodes during the first 6 months of life, probably because of passive transfer of immunoglobulin from their mothers, but can then expect to suffer from repeated clinical attacks during the following period (2). These first clinical attacks can be severe, and the number of children in these areas who die from malaria before the age of 3 years is very high. After the first years of life, the children become far less vulnerable to severe noncerebral malaria, and in the following years, their risk of developing uncomplicated malaria diminishes progressively as they acquire immunity (3). Acquired malaria immunity is mediated at least in part by IgG (9, 35), and several studies indicate that PfEMP1 variants are important targets (5, 26, 27). This is best demonstrated by the high susceptibility to infection in first-time pregnant women, in which parasites expressing a particular pregnancy-specific PfEMP1 variant, VAR2CSA, can sequester in the placenta (11). Women are less susceptible in successive pregnancies, due to acquisition of VAR2CSA-specific IgG as a result of the placental infection (12, 36).
Several studies have indicated that parasites causing severe malaria in young children express a limited set of PfEMP1s on the surface of the infected erythrocytes and that children acquire protection from severe noncerebral malaria early in life (14) because antibodies against these parasite forms are acquired after the first infections (4, 30, 39). These results imply that the host environment shapes PfEMP1 expression by parasites. The question is how. P. falciparum parasites carry about 60 PfEMP1-encoding var genes per haploid genome (13, 38). Evidence from experimental infections indicates that parasites released from the liver are not imprinted to express particular var genes, and many PfEMP1 variants can be detected during early blood-stage infection (23, 40). In contrast, blood-stage parasites are genetically imprinted, and most daughter parasites therefore express the same PfEMP1 as their parent (33). The global repertoire of PfEMP1 variants is probably very large but is organized into structurally related groups that are present in comparable proportions in all P. falciparum parasites (21, 32). This is perhaps not surprising, given the important function of these proteins, the relatively long generation time of the parasite, and the complexity of the parasite life cycle and cell biology. Serological cross-reactivity among PfEMP1 variants encoded by the same genome is limited (19), whereas interclonal cross-reactivity is considerable (8).
With this background, we produced a large set of recombinant PfEMP1 domain constructs representing VAR1 and the major PfEMP1 types of P. falciparum 3D7 and used them to follow seroconversion in a cohort of infants and children during their first years of life. As expected, the children acquired a broadening anti-PfEMP1 IgG repertoire over time as they were exposed to repeated P. falciparum infections. Despite the fact that only one or a few plasma samples from some children were available, the longitudinal cohort design of the study and the employed strategy for statistical analyses allowed us to evaluate the order of acquisition of domain-specific IgG in individual children, and this gave us considerable statistical power to evaluate the order of acquisition of domain-specific IgG in the cohort as a whole. We found that acquisition of these antibodies was remarkably structured. Since antibody responses reflect exposure, this strongly suggests that the children in the present study were exposed to parasites expressing particular PfEMP1 variants in an ordered sequence. This notion is supported by the recent study of Warimwe et al. (41) linking expression of PfEMP1 containing DBLα domains typical for group A and B/A genes to young host age, severe disease manifestations, and low levels of anti-PfEMP1 antibodies. We find the alternative explanation for our results, namely, differences in antibody detection efficiencies, unlikely, as the constructs employed induced comparable IgG responses in immunized rats and rabbits (data not shown). Moreover, results of assays run with different batches of beads coated with different protein production batches were consistent (data not shown).
Interestingly, the best-recognized domains in this study and our previous cross-sectional study (8) (DBLγ of PF11_0008, DBLε of MAL6P1.4, and DBLδ of PF11_0008) are all components of so-called PfEMP1 protein “cassettes” composed of 2 to 4 domains, which can be identified in most P. falciparum clones annotated so far (32). This suggests that these “cassettes” form functional units of importance in pathogenesis and protective immunity, a hypothesis that we are now investigating. Constructs representing domains belonging to the same cassettes seemed to be recognized almost similarly (e.g., constructs 127, 129, 131, and 135 belonging to a single cassette and constructs 20 and 22 belonging to a single cassette [cassettes 1 and 5, respectively]) (32), whereas constructs representing domains located on the same PfEMP1 in 3D7 but not belonging to the same cassette (e.g., constructs 69 and 63) showed a marked difference in serorecognition. This could indicate that individuals are exposed to domains represented in PfEMP1 cassettes simultaneously, whereas protein domains not belonging to a cassette (e.g., constructs 69 and 63) often are not inherited as part of the same PfEMP1. The identification of cassette types expressed by parasites causing severe malaria and establishing the endothelial ligands for the PfEMP1 proteins containing these cassettes is important for our understanding of the pathogenesis of severe malaria. It could also be the first step in defining vaccine constructs that should prevent severe malaria by inducing antibodies that prevent the sequestration of parasites causing severe malaria.
We agree that we can draw conclusions only from the domains tested and that other domains that are not well represented in the plex could be as well recognized and antibodies to these could be acquired earlier than are antibodies to the domains tested.
In conclusion, our results support the hypothesis that exposure to malaria and acquisition of anti-PfEMP1 antibodies are highly structured and that antibodies acquired during the first infections shape PfEMP1 expression during subsequent infections. It appears likely that infections in nonimmune individuals are dominated by parasites expressing the PfEMP1 variant(s) that mediates the most-effective sequestration of infected erythrocytes, as this will ensure maximum protection from clearance in the spleen. Such parasites will therefore have a higher effective growth rate than parasites expressing less-effective binding phenotypes. As a result, antibodies to this dominant variant will be acquired first, removing its selective advantage relative to parasites expressing variants conferring less-efficient adhesion phenotypes. The net result is that acquisition of anti-PfEMP1 antibodies in individual children is highly structured. The fact that this structuring extends beyond the acquisition of the first couple of anti-PfEMP1 antibodies indicates that these antibodies play a role not only in protecting against severe malaria infections but also in controlling blood-stage parasite multiplication in the later stages of acquisition of immunity.
ACKNOWLEDGMENTS
This work was supported by a grant from the Foundation for the National Institute of Health through the Grand Challenges in Global Health initiative, the U.S. National Institutes of Health, the Danish International Development Agency (DANIDA), and the Danish Research Council. G.K.K.C. is supported by the Gates Malaria Partnership, and A.T.R.J. is supported by the Howard Hughes Medical Institute.
We thank Sunthorn Pond-Tor and Susanne Pedersen for excellent technical assistance, the MOMS Project clinicians, nurses, and technicians for managing the clinical cohort and performing blood smear analyses, and Wonjong Moon for clinical data management. Thomas Scheike, University of Copenhagen, is thanked for providing statistical advice.
FOOTNOTES
- Received 3 June 2010.
- Returned for modification 27 June 2010.
- Accepted 23 August 2010.
- Copyright © 2010 American Society for Microbiology
