Background
Malaria during pregnancy is a major public health problem, associated with severe maternal anaemia, low birthweight (LBW) delivery, preterm delivery (PTD), pregnancy loss and infant mortality [
1]. During pregnancy,
Plasmodium falciparum parasites sequester in the placenta and adhere to chondroitin sulfate A (CSA) expressed on the surface of syncytiotrophoblast [
2]. Infected erythrocyte (IE) adhesion to CSA is mediated by VAR2CSA, a member of the
P. falciparum erythrocyte membrane protein family (PfEMP1) that is preferentially expressed by placental parasites and parasites selected to bind CSA [
3‐
5].
In areas of stable malaria transmission, primigravidae are at highest risk of
P. falciparum pregnancy malaria and women develop resistance over 1–2 pregnancies that controls infection and prevents severe sequelae [
6,
7]. This unique gravidity-dependent epidemiology correlates with the acquisition of immunity to placental parasites, and efforts to develop a placental malaria vaccine (PMV) that mimics naturally acquired immunity are on-going (reviewed in [
8]). VAR2CSA is currently the leading candidate for a PMV. VAR2CSA is a large protein of about 350kD composed of 6 Duffy binding-like (DBL) domains together with inter-domain regions. Its large size and sequence variation pose unique challenges in designing a vaccine based on VAR2CSA [
8].
Despite a central role in placental malaria pathogenesis, the evidence that VAR2CSA is targeted by protective antibodies has been inconsistent. Several studies compared primigravid and multigravid women for their seroreactivity to various VAR2CSA domains. Most studies reported that multigravidae have higher antibody levels to the relatively immunogenic DBL5 domain, while parity-dependent reactivity to other domains was less consistent [
8]. A common finding across studies is that multigravidae have higher antibody levels to at least one VAR2CSA domain or to the full-length protein [
3,
9‐
13]. Different results from studies of recombinant VAR2CSA seroreactivity patterns could be due to differences in allelic forms, domain boundaries, infection status, gestational age at the time of antibody measurement, or malaria prevalence in the study area.
Two studies in Cameroon reported that the breadth of reactivity to multiple domains, and levels of high avidity antibodies to full length recombinant VAR2CSA, were associated with a reduction in PM at delivery [
14‐
16]. In Benin, high antibody levels at enrollment variously associated with improved outcomes: antibodies to full length recombinant VAR2CSA and to DBL3 were associated with a reduction in the number of infections during the follow up period; antibody to DBL3 with a reduction in placental malaria; and antibody to DBL1-DBL2 with a reduction in low birthweight [
12]. Among Mozambican women who were infected at least once during pregnancy, above-the-median antibody levels to DBL3 and DBL6, as well as to malaria proteins MSP and AMA1 but not to DBL2, were associated with increased birthweight and gestational age [
11].
Malaria infections boost antibody levels to the IE surface [
17], and this could confound efforts to relate specific antibody titers to protection. In the current study of a longitudinal cohort of pregnant women, gravidity-dependent changes in antibody levels during and after malaria infection were evaluated. Further, relationships between VAR2CSA antibodies and preterm delivery or early pregnancy loss have not been studied. The present study examined whether VAR2CSA antibody levels predict pregnancy loss or preterm delivery risks, newborn gestational age and birthweight, and maternal anaemia risk.
Methods
Human subjects and clinical procedures
Pregnant women enrolled between November 2010 and October 2013 into a longitudinal cohort study of mother-infant pairs conducted in Ouelessebougou, Mali. The study site is located 80 km south of Bamako, an area of intense seasonal malaria transmission during the rainy season from July to December. Pregnant women aged 15–45 years without clinical evidence of chronic or debilitating illness were asked to participate in the study and gave signed informed consent after receiving a study explanation form and oral explanation from a study clinician in their native language. The protocol and study procedures were approved by the institutional review board of the National Institute of Allergy and Infectious Diseases at the US National Institutes of Health (ClinicalTrials.gov ID NCT01168271), and the Ethics Committee of the Faculty of Medicine, Pharmacy and Dentistry at the University of Bamako, Mali.
After enrollment, women underwent a clinical evaluation including a thorough medical history review of the current and previous pregnancies. The follow-up included scheduled monthly clinical examinations until 1 month post-delivery and at any time women felt sick. Blood smear samples for malaria infection and venous blood samples for the evaluation of the antibodies to VAR2CSA were collected at enrolment, week 30–32 and at delivery. Blood smear samples for malaria infection were also taken at scheduled and sick visits. Malaria infections were treated with quinine or with artemether–lumefantrine as clinically indicated. Women received 0–3 doses of intermittent preventive treatment with sulfadoxine–pyrimethamine (IPTp-SP), depending on recruitment trimester, number of doses received prior to enrollment, and on incidental parasitaemia which mandated treatment with other anti-malarial drugs. Gestational age was determined by ultrasound examination with Siui CTS-7700+ ultrasound scanner. Malaria infection was determined by thick blood smear. Infection was defined as the presence of any
P. falciparum parasite in a peripheral or placental blood smear. At least 100 high power fields in the thick smear were examined before concluding that the result was negative. Submicroscopic infections were defined by nested PCR as previously described [
18]. Miscarriage was defined as pregnancy ending at gestational week ≤ 28 weeks, stillbirth as a delivery of a non-viable baby at a gestational age of > 28 weeks, and early neonatal death as death occurring within 7 days of birth. Pregnancy loss included cases of miscarriage, stillbirth and early neonatal death. Preterm delivery was defined as birth prior to gestational age of 37 weeks. LBW was defined as a birth weight of < 2500 g. Small for gestational age (SGA) was defined as a birth weight below the tenth percentile for the gestational age and gender, based on a weight chart developed in Tanzania [
19].
Multiplex antibody assays
Recombinant VAR2CSA DBL domains were prepared based on alleles from laboratory parasite lines FCR3 and 3D7, and from the dominant VAR2CSA alleles identified by whole genome sequencing of maternal isolates M1010, M711, and M466 (Table
1). Eleven VAR2CSA fragments were cloned and expressed in
Escherichia coli or baculovirus as previously described [
20‐
22]. Recombinant glutamic acid rich protein (GARP, PlasmoDB ID PF3D7_0113000), a protein that is not unique to placental parasites, was included as a control. Antibody levels in plasma samples diluted 1:80 were measured using a multiplex bead-based assay (Bioplex: Bio-Rad, Irvine, CA), as previously described [
23]. Antibody levels were expressed as median fluorescence intensity (MFI) after subtracting background reactivity with beads coated with BSA. Seropositivity was defined for each of the domains as the level above the mean + 2 SD obtained from 22 control plasma samples collected in the US.
Table 1
Recombinant DBL domains
DBL2 | FCR3 | 543–858 |
E. coli
|
ID1-ID2a | Maternal isolate 1010 (M1010) | 428–1024a | Baculovirus |
DBL3 | FCR3 | 1220–1541 |
E. coli
|
DBL3-4 | FCR3 | 1445–1989 |
E. coli
|
DBL4 | FCR3 | 1583–1989 |
E. coli
|
DBL4 | 3D7 | 1570–1926 |
E. coli
|
DBL4 | Maternal isolate 1010 (M1010) | 1583–1989a |
E. coli
|
DBL4 | Maternal isolate 711 (M711) | 1583–1989a |
E. coli
|
DBL5 | FCR3 | 2003–2281 |
E. coli
|
DBL5 | Maternal isolate 1010 (M1010) | 2003–2281a |
E. coli
|
DBL5 | Maternal isolate 466 (M466) | 2003–2281a |
E. coli
|
Statistical analysis
The analyses included singleton deliveries only. Differences in antibody levels between groups were analysed by nonparametric methods (Mann–Whitney or Kruskal–Wallis tests). p values were corrected for multiple comparisons by Holm method using R function p adjust. Adjusted p values < 0.05 were considered significant.
To estimate if antibody levels at enrollment and at gestational week 30–32 predict reduced risk of pregnancy loss and PTD, the Lunn & McNeil Competing Risks Model (LMCRM) [
24] were fitted. Three distinct modelling approaches were utilized for antibody levels as predictor: (A) the mean log
10 antibody levels, (B) all 11 antibodies in 1 model, and (C) each antibody in 11 different models. A likelihood ratio test confirmed that approach (A) was more parsimonious than approach (B) without information loss and approach (C) gave similar results to approach (A), and results of model (A) are presented. Because of the longitudinal nature of the study, gestational age was used as the time variable. All models were adjusted for the same covariates. Time dependent variables were used for covariates cumulative IPTp and malaria infection during follow up visits, and malaria at the time of antibody measurement. In addition, the model was adjusted for gravidity and insecticide treated net (ITN) use. The analysis was conducted in R with packages survival and data table.
Multivariate logistic regression models were used for each cross-sectional analysis (enrollment, gestational week 30–32 and delivery) to examine the relationships between antibody levels to VAR2CSA and LBW or SGA. Normal birthweight or normal weight for gestational age served as the reference group. Multivariate linear regression models were used for each cross-sectional analysis (enrollment, gestational week 30–32 and delivery) to examine the relationships between antibody levels to VAR2CSA and birthweight or gestational age. Models were adjusted for gravidity, malaria infection, ITN use and number of IPTp doses. Models fitted for the analysis of birthweight and LBW were also adjusted for gestational age.
Multinomial logistic regression was used to examine the relationships between antibody levels and maternal anaemia at enrollment, and multivariate linear regression model was used to examine the relationships between antibody levels and hemoglobin. The 3 levels of the categorical outcome were defined as no anaemia, mild anaemia (haemoglobin levels 10–10.9 g/dl), and moderate/severe anaemia (haemoglobin levels < 10 g/dl). No anaemia served as the reference group. The predictors include mean log10 antibody levels with adjustment for malaria infection at enrollment, insecticide-treated bed net (ITN) use, and gestational age.
Discussion
Over successive pregnancies, women develop antibodies against IE surface proteins expressed by placental parasites that block parasite adhesion to the placental receptor CSA [
7]. The presence of anti-adhesion antibodies has been associated with increased birth weight and gestational age [
12,
28]. Anti-adhesion antibodies broadly react with placental IE regardless of their geographical origin, indicating that they target antigenically conserved epitopes [
7]. IE adhesion to CSA is mediated by VAR2CSA, the leading candidate for developing a vaccine to prevent PM. In the current study, the effects of gravidity and infection on VAR2CSA antibody levels during pregnancy, and the associations between VAR2CSA antibody levels and pregnancy outcomes were examined.
Similar to previous reports, antibody levels to VAR2CSA domains DBL2, DBL3, DBL4 and DBL5 but not to the control protein GARP increased with gravidity [
11,
12]. Antibody levels between VAR2CSA domains correlated, but not between antibody levels to VAR2CSA and the control protein GARP (Additional file
3). Patent infection at enrollment was associated with an increase in VAR2CSA antibody levels (Fig.
2), a difference that was not observed at gestational week 30–32 and at delivery. However, among women who were BS− at week 30–32 or at delivery, those who had been BS+ at any earlier timepoint often had higher antibody levels, but this was primarily seen in the paucigravid groups: for primigravidae, antibody levels were higher to 11/11 and 8/11 domains at the two respective timepoints; for secundigravidae, 0/11 and 8/11; for multigravidae, 0/11 and 1/11 (Additional file
2). These results are consistent with a previous study in Mozambique reporting that at delivery, antibody levels to domains DBL2, DBL3 and DBL5 were higher in women with histological evidence of acute or past infection than women without evidence of infection [
11].
Submicroscopic infections at enrollment were associated with a significant increase in antibody levels to all VAR2CSA domains in multigravidae, but to only one domain in primigravidae. In mice, subpatent infections with
Plasmodium chabaudi chabaudi induce antibody responses to conserved antigens, but not to variant surface antigen, and this has been ascribed to several possible causes including relatively low variants surface antigen dose [
29]. Here, submicroscopic infections probably represent a low antigen dose sufficient to boost existing immune responses to the variant antigen in multigravidae but not to elicit new antibodies in primigravidae.
An earlier study conducted in northwestern Thailand, a low malaria transmission area described that antibody levels to malaria antigens including VAR2CSA declined during pregnancy in both infected and uninfected women. However, the half-life of antibodies to VAR2CSA was much longer than the half-life of antibodies to merozoite antigens (36–157 years) [
30]. Another study of non-pregnant women residing in an area with stable malaria transmission, described a progressive increase during pregnancy in VAR2CSA-specific IgG levels and B cell numbers in both primigravidae and multigravidae, and then a decline postpartum [
26]. Despite the rapid decline in antibody levels post-delivery, gravidity-dependent differences in antibody levels to VAR2CSA were observed in non-pregnant women suggesting an increased longevity of plasma cells secreting IgG over successive pregnancies [
31]. Consistent with this observation, in the current study, antibody levels rapidly declined and were significantly lower at delivery in primigravidae who were infected at enrollment but not thereafter, indicating that short-lived antibody is induced in primigravidae, who are relatively naïve.
At enrollment and at gestational week 30–32, secundigravidae had antibody levels that were similar to those of primigravidae, but were lower than those of multigravidae to four domains. By delivery, secundigravidae who experienced at least one infection had similar antibody levels as multigravidae. These results suggest that at least one infection during the 2nd pregnancy is required to achieve antibody levels similar to those of clinically resistant multigravidae.
Two vaccine candidates based on the N-terminal region of VAR2CSA are currently being evaluated in clinical trials. One product, PRIMALVAC, is based on the DBL1-DBL2 domain combination of 3D7 allele, and the other, PAMVAC, is based on the ID1-ID2a region of FCR3 allele [
32]. These fragments were selected for placental malaria vaccine (PMV) development based on their binding affinity to CSA [
33,
34]. In addition, these fragments elicited heterologous functional antibodies that inhibited IE binding to CSA [
35,
36]. Conflicting results have been reported on naturally acquired antibodies to the N-terminal VAR2CSA fragments. In Cameroon, antibody levels to ID1-ID2a were similar between men and pregnant women [
37]. In Benin and in the current study, antibody levels to DBL1-DBL2 were significantly higher in multigravidae than primigravidae [
12]. Variation in allelic forms, domain boundaries, timing of antibody measurement and assay among others complicates direct comparisons between studies, that may be addressed by harmonizing these variables.
Antibody levels to VAR2CSA domains were not associated with reduced risk of pregnancy loss and preterm delivery, nor with increased gestational age and birthweight or decreased maternal anaemia. The negative association between birthweight and antibody levels in primigravidae could be explained by the increase in antibody levels during malaria infection. The results here confirm that the recombinant VAR2CSA domains including those containing the minimal CSA-binding domain used in this study are reactive to malaria-induced antibodies, but may fail to re-create the epitopes targeted by antibodies associated with protection. In a recent study, multigravidae plasma depleted of antibodies directed to the domains evaluated here, did not reduce the level of functional anti-adhesion antibodies, but reduced the level of reactivity with IE surface measured by flow cytometry [
22]. Induction of functional anti-adhesion antibodies is sensitive to changes in domain boundaries and allelic type [
36], which may explain the lack of an association between antibodies to domains evaluated here including those containing the minimal CSA-binding region and pregnancy outcomes.
Authors’ contributions
MF, AD, and PED designed the study. MF and PED with contributions of JDK and AD wrote the main text. SPT, SK, AM, NA OA, ABD, KBC, BSD, and MBK performed experiments. BS and RM performed statistical analysis. AB and YS collected clinical information. DLN Provided recombinant proteins. All authors read and approved the final manuscript.