This study was part of the PregVax project (FP7-HEALTH-201588, www.pregvax.net), which studied the burden, impact, immune responses, and pathophysiology of Pv in pregnancy between 2008 and 2012 in five endemic countries: BR, Colombia (CO), Guatemala (GT), India (IN), and PNG. Approximately 2,000 women per country were enrolled at the first visit at the antenatal clinic (recruitment) and followed up until delivery. In all visits, hemoglobin (Hb) levels, Pv and Pf parasitemias by blood smear and malaria symptoms were assessed. Giemsa-stained thick and thin blood slides were read onsite following WHO standard quality-controlled procedures, and external validation of a subsample of blood slides was done at the Hospital Clinic and at the Hospital Sant Joan de Deu, in Barcelona, Spain. BW was recorded. Women with a positive smear were treated according to national guidelines, except in PNG where blood smears could not be read at the moment of the visit for logistical reasons (only symptomatic women were thus treated after confirmation of infection by rapid diagnostic test).

From the PregVax cohort, 10% of women were randomly allocated to the “immunology cohort” and were followed up with one visit at least 10 weeks after delivery (postpartum group). At recruitment, delivery and postpartum visits, a venous blood sample (5–10 mL) was collected aseptically in heparinized tubes. Submicroscopic Pv and Pf infections were also determined in a random subsample by real-time PCR, except for the Indian samples, where only Pv infection was examined. Additionally, blood samples (10 mL) were collected from 39 malaria naïve donors at the blood bank in Hospital Clinic (Barcelona, Spain) and used as negative controls.

The protocol was approved by the Hospital Clinic Ethics Review Committee (CEIC, Barcelona, Spain), the CDC IRB (USA), and the national and/or local ethics committees of each site: the Universidad del Valle de Guatemala Ethics Review Committee (CE-UVG, GT); the Comité Institucional de Revisión de Ética Humana, Facultad de Salud, Universidad del Valle de Cauca (Cali, CO); the Comissâo Nacional de Ética em Pesquisa (CONEP, BR); the Comitê de Ética em Pesquisa da Fundação de Medicina Tropical do Amazonas (FMT-AM, Manaus, BR); the Ethics Committee, Sardar Patel Medical College and A.G. Hospitals (Bikaner, Rajasthan, IN); and the Medical Research Advisory Committee in PNG (MRAC 08.02). Written informed consent was obtained from all study participants.

Plasma was separated by centrifugation and stored at −80°C. Blood cells from PNG and Spain were further fractioned in a density gradient medium (Histopaque^®-1077, Sigma-Aldrich) to obtain PBMCs and stored in liquid nitrogen. Samples from GT, CO, BR, and PNG were analyzed at ISGlobal (Barcelona, Spain) while plasmas from IN were analyzed in Delhi.

Unless otherwise specified, recombinant proteins were cloned with an Escherichia coli system using genomic DNA from different Plasmodium strains as template. Pv200L (amino acid residues 121–416 of PvMSP1) was amplified from genomic DNA obtained from a Colombian Pv-infected patient (33). PvDBP (receptor-binding domains—RII) was cloned using Pv Sal-I as template (34). PvMSP1₁₉ (19 kDa C-terminal region, residues 1639–1729) was expressed using the Belem strain as template (35) and PfMSP1₁₉ with the 3D7 strain (36). PfAMA1 (N-terminal ectodomain) was cloned using genomic DNA from the Pf 3D7 strain (37). PfEBA₁₇₅ (the receptor-binding domain—PfF2) was cloned from the Pf CAMP strain (38). The Pf A4 strain was used as template for the VAR2CSA DBL3X domain (39) and 3D7 strain for the DBL5ε (40) and DBL6ε (41) VAR2CSA domains.

PvMSP1-N (fragment 170-675), full-length PvCSP, and full-length PvMSP5 were expressed with a glutathione S-transferase (GST) tag in a cell-free wheat germ system as reported previously (11). Expressed proteins were purified on GST SpinTrap purification columns (GE Healthcare), and eluted proteins were dialyzed in phosphate buffered saline (Tube-O-DIALYZER™, GBiosciences). GST was also expressed separately for immune-reactivity control. The production of the PvCSP-N (residues 20–96), PvCSP-C (residues 301–372), and PvCSP-R (three tandem-repetitions of the residues 96–104) peptides has been described elsewhere (9). Five PvDBP peptides [VNNTDTNFH(R/S)DITFR, LYLKRKLIYDAAVEG, and LIYDAAVEGDLL(L/F)KL] containing immunogenic epitopes previously published (42) were synthesized at >80% purity by Peptide 2.0 (Chantilly, VA, USA) for cellular stimulation assays.

Measurement of plasma IgG antibodies was performed by multiplex suspension array using the Luminex™ technology, as described before (24). Briefly, 1.1–1.4 million MagPlex^® magnetic-carboxylated microspheres (Luminex Corporation, TX, USA) with different spectral signatures were covalently coated with 3 µg of each protein/peptide, following the manufacturer’s instructions. Protein-coupled beads were quantified in a Guava^® Flow Cytometer (Millipore) and mixed in equal amounts. A unique batch of microspheres was prepared for the whole study, including the samples analyzed in IN. Approximately 1,000 beads per analyte were incubated with each plasma (1:100 dilution) in duplicates, and subsequently with antihuman IgG-biotin (Sigma-Aldrich), followed by streptavidin-conjugated R-PE (Fluka, Madrid, Spain). Beads were acquired on the BioPlex100 system (Bio-Rad, Hercules, CA, USA), and results were expressed as median fluorescence intensity (MFI) of duplicates. Value against GST alone was subtracted from correspondent proteins. Raw GST data have been previously published (24). Cross-reactivity was ruled out in a pilot study analyzing a subset of plasmas in singleplex and multiplex (not shown). Samples in IN were analyzed with identical protocols and instruments.

A functional assay was carried out to study the capacity of anti-PvDBP antibodies to inhibit the binding of this protein to its receptor DARC, as previously described (43). Only plasma samples from responder women (anti-PvDBP MFI values above the mean + 3 SD of Spanish naïve controls) at recruitment were included in the functional assays. Ten microliters of plasma from each donor were preincubated with 0.01 µg/mL of PvDBP for 1 h at RT before adding to Bioplex plates containing 1,000 Luminex beads pre-coupled to DARC-Fc (N-terminal extracellular 60 amino acids of DARC fused to Fc region of human IgG). Bound PvDBP was detected with anti-PvDBP, followed by anti-human IgG-biotin and streptavidin-conjugated R-PE. A standard curve using increasing amount of PvDBP was prepared to establish concentration of PvDBP bound to DARC-Fc. Percent inhibition was calculated as % Inhibition = (1 − observed PvDBP conc./expected PvDBP conc.) × 100.

Except where indicated, all reagents were purchased from BD Biosciences, and all antibodies were monoclonal. PBMCs were thawed and rested for 10–12 h. Only samples with viability >70% (Guava ViaCount Reagent, Millipore) were used for assays. Half a million cells per well were resuspended in RPMI-1640 medium plus 10% fetal bovine serum (culture medium) and incubated in the presence of a pool containing the five PvDBP peptides (5 µg/mL) or PvMSP1₁₉ (10 µg/mL). Culture medium was used as negative control. Anti-CD3 stimulation was the positive control (24). After 12 h, an aliquot of 30 µL of culture medium supernatant was collected to measure secreted cytokines and replaced with media containing GolgiPlug™ for an additional (4 h) incubation. PBMCs were stained with LIVE/DEAD^® Fixable Violet Dead (Life Technologies), anti-CD14 Pacific Blue (clone M5E2), anti-CD19 Horizon™ V450 (clone HIB19), anti-CD4 allophycocyanin (clone RPA-T4), and anti-CD8 Peridinin Chlorophyll Protein Complex (PerCP, clone SK1). After washing, cells were fixed and permeabilized with Cytofix/Cytoperm™ and incubated with anti-CD3 phycoerythrin (PE)-Cy™7 (clone SK7), anti-interferon (IFN)-γ PE (clone 25723.11), and anti-CD69 fluorescein isothiocyanate (clone L78). Cells were acquired in an LSRFortessa flow cytometer, and data were analyzed by FlowJo (FlowJo LLC, OR, USA). The gating strategy was developed as previously (24). Supernatants were frozen at −80°C until Luminex analysis with the Cytokine Magnetic 30-Plex Panel (Invitrogen), according to the manufacturer’s instructions.

Samples from BR, CO, GT, and half samples from PNG were analyzed at the Istituto Superiore di Sanità (Rome, Italy), as described (44). Positivity for each species was established as a cycle threshold <45, according to negative controls. Pv diagnosis for IN samples was performed in Delhi following Rome’s protocol adapted for the instrument sensitivity (third step amplification 72°C for 25 s instead of 72°C for 5 s). Approximately half of PNG samples were analyzed for submicroscopic infections in Madang, following a similar protocol to Rome’s (45), except that positivity for each species was established as cycle threshold <40, according to negative controls. DNA was extracted from whole blood-spot samples on filter paper.

Any Plasmodium infection was defined as a positive smear and/or positive PCR. A positive antibody response (responder) was considered as MFI values above the mean + 3 SD of Spanish controls. To evaluate the differences in antibody levels among countries, a one-way ANOVA test was used followed by Bonferroni pairwise correction. To study the association between antibody levels and pregnancy variables, univariate (only adjusted for country of origin) and multivariate linear regression models were estimated with the following variables: country, age, gestational age, gravidity (number of previous gestations), and Pv or Pf infection during pregnancy (considering past or present infections but not future–delivery—infections). The correlation between IgG responses to different antigens was evaluated with the Spearman’s rank test. The association between IgG levels at recruitment and malaria infections at delivery was evaluated with logistic regression models. The association between antibody responses and pregnancy outcomes, i.e., Hb levels at delivery and BW, was analyzed using univariate and multivariate linear regression models, adjusted by country, Hb at recruitment, gestational age at recruitment, age, gravidity, and past or present Plasmodium infection during pregnancy. For the cellular and cytokine analyses, deviation from normality was tested using the skewness and kurtosis test. Because none of the variables except IL-13 presented a normal distribution, data were presented as medians, and comparisons between groups were done using the Mann–Whitney U test. The cytokine multiple comparisons were adjusted using the Bonferroni test. Cytokine/chemokine production in culture supernatants of unstimulated samples (medium) was not subtracted from the stimulated samples but shown side by side as it is possibly biologically relevant. Significance was defined at p < 0.05. Analyses and graphs were performed using Stata/SE 10.1 (College Station, TX, USA).

In total, 1,491 peripheral blood samples (794 at recruitment—first antenatal visit, 529 at delivery, and 168 postpartum) corresponding to 1,056 women were analyzed for IgG antibody responses against both Pv (six recombinant proteins and three peptides) and Pf (six recombinant proteins). Unfortunately many of our samples were not matched due to late attendance to antenatal clinics or low follow-up rate. The baseline study population characteristics and infection rates by country have been published previously (24).

For cellular assays, 46 samples (any gestational age including delivery, 18 Pv PCR negative, 28 Pv PCR positive) from the PNG pregnant cohort were included in the analyses. When cell counts were limited, stimulation with PvMSP1₁₉ was the last priority, resulting in a lower sample size (12 Pv PCR negative and 15 Pv PCR positive).

IgG MFI values against the four Pv preerythrocytic antigens differed among countries at all timepoints (one-way ANOVA p < 0.05). Overall, PNG had the highest levels of all anti-PvCSP antibodies at recruitment and delivery followed by CO; at postpartum CO had similar or higher levels (Figure 1). PvCSP protein was more reactive (seropositivity and/or MFI levels) than the three PvCSP peptides in PNG and GT, but they were recognized at similar levels in BR, CO, and IN. PvCSP-N was the most reactive of the three peptides in all countries and at all timepoints, except CO and BR at postpartum, where PvCSP-R was more seroprevalent (Figure 1). We also observed higher anti-PvCSP peptide levels at postpartum compared to recruitment and delivery, but this was not found for anti-PvCSP protein levels (Table S1 in Supplementary Material). In general and as expected, MFI and seroprevalence values for anti-PvCSP antibodies were low compared to merozoite antigens (Figure 2).

Similar to anti-PvCSP responses, prevalence of anti-Pv merozoite antibodies differed among countries at all timepoints (one-way ANOVA p < 0.05), and PNG presented the highest magnitudes and prevalence at recruitment and delivery followed by CO (Figures 2A,B). In contrast to anti-PvCSP responses, women from GT presented remarkable anti-merozoite levels at all timepoints in spite of lower prevalence of patent malaria infections. Reactivity of each antigen varied across countries and timepoints; for instance antibody values for Pv200L and PvMSP1-N were lower than for PvDBP in BR, CO, and IN, but similar and higher in GT and PNG, respectively. The levels of PvDBP and PvMSP1₁₉ antibodies in CO at postpartum were very high compared to the rest of countries (Figure 2C). We did not find differences in antibody levels between timepoints.

We also measured binding-inhibition percentage to Duffy antigen in the plasma of positive PvDBP responders at recruitment. Although BR had the lowest number of responders, the anti-PvDBP antibodies on those responders had the highest amount of inhibitory activity (one-way ANOVA p = 0.01, Figure 2D), with 8/18 individuals having more than 50% inhibitory activity.

Differential Pf antibody responses were observed between countries at all timepoints (one-way ANOVA p < 0.05), and PNG presented the highest magnitudes except for PfMSP1₁₉, PfDBL3x, PfDBL5ε, and PfDBL6ε at postpartum that were similar in CO (Figure 3). CO presented the second highest anti-Pf antibody levels, while antibodies were almost absent in BR and IN. Of note, GT presented remarkable anti-Pf responses, similar in some cases to CO, despite low prevalence of Pf infection in our cohort. PfAMA1 antibodies had the highest MFI values (not so for seroprevalence) in all cases, and anti-PfAMA1 IgG levels in PNG were higher than any Pv antigen-specific antibodies (Figure 3). When we compared magnitude of antibody responses between timepoints, we found lower anti-PfMSP1₁₉ at delivery and higher anti-PfDBL3x levels at postpartum compared to recruitment (Table S1 in Supplementary Material).

To study the possibility of cross-reactivity, we analyzed the correlation between antibody levels (Table 1). There was a good correlation between the levels of antibodies of the same group (preerythrocytic, merozoite, and PfEMP1 family) except for PvCSP protein and peptides. The highest intraspecies correlation was observed between anti-PfMSP1₁₉ and anti-PfAMA1 levels (Table 1). With regards to interspecies correlation, this was relatively high between Pv and Pf anti-merozoite antibody levels (Spearman’s rho ≥ 0.45 depending on the antibody, p < 0.0001 for them all, Table 1). Interestingly, we found a high correlation between levels of antibodies against the three PvCSP peptides and VAR2CSA domains, especially PfDBL3x (Spearman’s rho ≥ 0.64, p < 0.0001, Table 1; Figure 4). The percentage of PvDBP functional antibodies correlated poorly with the levels of all antibodies, except for anti-PvDBP, anti-PvMSP1-N, and anti-PvMSP5 (all p < 0.0001).

Next, we assessed how different pregnancy variables affected the levels of antibodies. As expected, country of origin was significantly associated with all antibodies, both at recruitment and at delivery (not shown). In the adjusted regression, PfMSP1₁₉ IgG responses had a positive association with age at delivery (proportional differences per 1 year increase in age = 1.04, 95% CI = 1.00–1.08, p = 0.042). PvMSP1-N antibody levels had a positive association with gravidity [proportional difference in antibody levels per belonging to category: (0 previous gestations) = 1; (1–3 previous gestations) = 2.06, 95% CI = 1.24–3.42; (4+ previous gestations) = 2.01, 95% CI = 0.87–4.63; p = 0.020] and with gestational age (proportional differences per 1 week increase in gestational age = 1.05, 95% CI = 1.02–1.08, p = 0.002) at recruitment. Most antibodies presented a positive association with current Plasmodium infection (infection status at time of sample collection), which was more prominent at recruitment than at delivery (Tables 2 and 3, data at delivery not shown). Of note, women with Pf and Pv coinfection had significantly more anti-PvMSP1₁₉, anti-Pv200L, and anti-PvDBP (borderline difference for PvDBP) antibodies than women with Pv monoinfections (Table 2). Similarly, coinfected women had more anti-Pf antibodies than women with Pf infection alone in most cases (Table 3). However, sample size for coinfections was small, and these results should be considered cautiously.

We also analyzed the association between the levels of antibody levels at recruitment in uninfected women and infection rates at delivery (future infection). Anti-PvDBP levels were positively associated with future Pv infection, and anti-PfMSP1₁₉ and anti-PfAMA1 levels were positively associated with future Pf infection (Table 4). In addition, antibody levels against PfMSP1₁₉ and PfAMA1 were associated with future Pv infection (Table 4).

Subsequently, we assessed the association between antibody levels at recruitment and delivery outcomes (Hb and BW). Anti-PvCSP-C and anti-PvCSP IgG levels had a positive and negative association, respectively, with Hb levels at delivery, but these associations were not maintained in the adjusted analysis. Anti-PvCSP-C and anti-PfDBL3x antibody levels, and the percentage of inhibitory anti-PvDBP antibodies, were positively associated with BW (Table 5), but only the last two variables maintained the association in the adjusted analysis.

We compared the cellular immune responses to PvDBP peptides and PvMSP1₁₉ between Pv infected and uninfected pregnant women in the PNG cohort. PBMCs from women with a concurrent Pv infection had less CD8⁺IFN-γ⁺ cells than uninfected women independently of whether we cultured the cells in the presence of the antigens or medium alone (Figure 5). The differences for the CD4⁺IFN-γ⁺ population were not statistically significant (medium p = 0.085, DBP p = 0.083, PvMSP1₁₉ p = 0.064, Figure 5). No differences between infected and uninfected groups were found when we stimulated with anti-CD3 (24) or when we measured CD69⁺ T cells (not shown). In the supernatants of the PBMCs cultured with either PvDBP peptides (Table 6; Figure 5) or medium alone (24), we found significantly more G-CSF and IL-4 in the infected than in the uninfected group. Specifically in response to PvDBP peptides, PBMCs from infected women secreted significantly less IL-17 than those from uninfected women (Table 6; Figure 5). However, none of these differences remained significant after adjusting for multiple comparisons, and these results should be interpreted with caution.