Differential Recognition of P. falciparum VAR2CSA Domains by Naturally Acquired Antibodies in Pregnant Women from a Malaria Endemic Area

Background Plasmodium falciparum infected red blood cells (iRBC) express variant surface antigens (VSA) of which VAR2CSA is involved in placental sequestration and causes pregnancy-associated malaria (PAM). Primigravidae are most susceptible to PAM whereas antibodies associated with protection are often present at higher levels in multigravid women. However, HIV co-infection with malaria has been shown to alter this parity-dependent acquisition of immunity, with more severe symptoms as well as more malaria episodes in HIV positive women versus HIV negative women of a similar parity. Methods Using VAR2CSA DBL-domains expressed on the surface of CHO-745 cells we quantified levels of DBL-domain specific IgG in sera from pregnant Malawian women by flow cytometry. Dissociations constants of DBL5ε specific antibodies were determined using a surface plasmon resonance technique, as an indication of antibody affinities. Results VAR2CSA DBL5ε was recognized in a gender and parity-dependent manner with anti-DBL5ε IgG correlating significantly with IgG levels to VSA-PAM on the iRBC surface. HIV positive women had lower levels of anti-DBL5ε IgG than HIV negative women of similar parity. In primigravidae, antibodies in HIV positive women also showed significantly lower affinity to VAR2CSA DBL5ε. Conclusions Pregnant women from a malaria-endemic area had increased levels of anti-DBL5ε IgG by parity, indicating this domain of VAR2CSA to be a promising vaccine candidate against PAM. However, it is important to consider co-infection with HIV, as this seems to change the properties of antibody response against malaria. Understanding the characteristics of antibody response against VAR2CSA is undoubtedly imperative in order to design a functional and efficient vaccine against PAM.


Introduction
Pregnancy-associated malaria (PAM) has a major impact on the mother and child [1] and is often associated with the sequestration of Plasmodium falciparum infected red blood cells (iRBC) in the placenta [2,3]. IRBC expressing the adhesin VAR2CSA adhere to the receptor CSA expressed by syncytiotrophoblast [4,5,6,7], and potentially enable the parasites to evade the immune system and successfully proliferate. Hence, despite previously acquired immunity to non-VAR2CSA expressing P. falciparum malaria, pregnant women are highly susceptible to PAM [8], leading to maternal anemia, low birth weight, miscarriage and stillbirths [9,10,11].
VAR2CSA is a member of the PfEMP1 family but displays a different domain architecture than typical PfEMP1s and has an unusual high sequence conservation between isolates [12,13]. VAR2CSA contains six Duffy-binding-like (DBL) domains of which four (DBL2x, DBL3x, DBL5e and DBL6e) domains have been shown to bind CSA in vitro [14,15,16]. Studies have shown that placental parasites and CSA-selected parasites bind normal non-immune IgG and IgM [17,18,19] and also that VAR2CSA DBL domains harbour non-immune IgG and IgM binding regions [19]. It has also been shown that immune antibodies are targeting various epitopes within the DBL-domains [20]. Antibody immunity against PAM is acquired in a parity dependent manner, and important functions of these antibodies are the ability to block adhesion to placental receptors [21] and to facilitate opsonic uptake by phagocytes [22,23,24]. VAR2CSA is the current vaccine candidate for PAM, but in order to design a rational and protective vaccine based on this large protein, increased knowledge of anti-VAR2CSA antibody response characteristics is needed. Since malaria and HIV co-exist to an extremely high extent in sub-Saharan Africa, it is also important to consider the fact that HIV positive women infected with malaria have more febrile illnesses, more anemia, and more adverse birth outcomes than HIV negative women with malaria [25]. Several studies have suggested HIV to affect the immune memory mechanism, which is responsible for the parity dependent acquisition of immunity to PAM, thus rendering women of all parities highly susceptible to PAM [26,27]. HIV infection in multigravid women seems to impair the ability to control malaria parasitemia, resulting in more frequent and higher parasite density than in HIV negative women of the same parity [25]. HIV positive women in their first pregnancy do not experience a significantly increased risk of malaria prevalence, but do retain significantly higher parasite density [28]. In several studies, it has been shown that HIV positive women receiving intermittent preventive treatment in pregnancy (IPTp) need additional doses of drugs in order to be protected [28,29,30], further reflecting the need for increased knowledge of antibody dynamics.
In vitro studies often focus on investigating the mere presence of antibodies against the variant surface antigens present on the surface of iRBC causing pregnancy associated malaria (VSA-PAM), with less knowledge on the potentially important affinity to their target. Being a neglected factor in PAM immune response studies, we here investigate not only the levels of VAR2CSA DBLdomain specific antibodies in sera from pregnant Malawian women but also further scrutinize the elicited antibody responses by exploring affinity of antibodies targeting VAR2CSA DBL5e.

Patient Characteristics
Pregnant women attending the Queen Elizabeth Central Hospital, Blantyre, Malawi in late third trimester of pregnancy were enrolled into a study of interactions between HIV and malaria in pregnancy, as described elsewhere [26,31]. A convenience selection of serum samples collected on enrollment was used in the various assays of the present study, including a total of 189 serum samples from primigravidae, 21 from secundigravidae and 72 from multigravidae. HIV infection rates were lower than expected in the primigravidae group compared with the secundigravidae and multigravidae (x 2 test, p = 0.02316), which is probably a function of age and repeated exposures. Parasitemia was more common in primigravid women and much less common in multigravid women (x 2 test, p = 0.01594), consistent with the hypothesis that previous malaria infection during pregnancy produces VAR2CSA (and thus protective) antibodies. Maternal anemia rates and infant birth weights were not different among the groups (x 2 test, p.0.8). Patient characteristics are displayed in Table 1.

VAR2CSA DBL-Domains Are Recognized by Immune Sera from Pregnant Adults
We used flow cytometry to measure antibody levels in a panel of sera from Malawian pregnant women to the variant surface antigens (VSA) of P. falciparum VAR2CSA expressing lab isolate CS2 [32] ( figure 1A and figure S1). This analysis showed multigravid (MG) women to have significantly higher levels of VSA-PAM antibodies than primigravid (PG) women from the same endemic area (Kruskal-Wallis ANOVA, p = 0.0001 followed by Dunn's Multiple Comparison test, p,0.001). We then used CHO-745 cells transfected with 3D7 VAR2CSA DBLdomains, described elsewhere [14], to investigate the domain specificity of these acquired antibodies (figure S2 and S3). Previous studies showed the various domains to be expressed on the surface of the transfected CHO-745 cells at similar magnitudes [19], and the level of surface expression was also monitored for each experiment in this study (figure S3). Using the panel of sera from Malawian pregnant women described above, we assayed levels of DBL domain specific antibodies. Flow cytometry assays using these VAR2CSA DBL domain transfected cells showed various patterns. After initial screening of all six DBL domains using serum samples from Malawian women with different parities (data not shown), additional serum samples were tested using domains DBL3x, DBL5e and DBL6e ( Figure 1B). The reason for choosing these domains is the distinctly higher recognition of these domains by pooled hyperimmune multigravid sera than pooled male immune sera from individuals living in the same area, demonstrating a gender specific antibody recognition pattern of these three domains. The pattern of DBL3x recognition indicated that antibodies in sera from primigravidae show similar recognition of this domain to those in sera from multigravidae (Kruskal-Wallis ANOVA, p = 0.0589, figure 1C). Neither did antibody levels to DBL3x correlate with antibody levels to VSA-PAM antibodies (Spearman r = 20.08980, p = 0.3972, figure 1D). Notably, A4 DBL3x transfected CHO-745 cells were used as a complement to 3D7 DBL3x transfected CHO-745 cells since the 3D7 variant was shown to contain a 12-mer residue deletion and did not bind CSA [33]. From this further analysis, it is apparent that DBL5e was the only domain to induce a clear parity dependent recognition pattern, with multigravid women boasting notably higher levels of anti-DBL5e antibodies than primigravidae (Kruskall-Wallis AN-OVA, p = 0.0141; Dunn's Multiple Comparison test, p,0.05 (PG versus MG), figure 1E). Levels of VSA-PAM antibodies and DBL5e specific antibodies showed a significant positive moderate correlation (Spearman r = 0.5632, p,0.0001), suggesting that DBL5e may be important in the acquired immune response against PAM ( figure 1F). The pattern of VAR2CSA DBL6e antibody recognition indicates this domain to also be recognized by multigravid women to a similar extent as primigravidae (Kruskall-Wallis ANOVA, p = 0.5118, figure 1G). Levels of antibodies against DBL6e did show a significant positive weak correlation with levels of antibodies against VSA-PAM (Spearman r = 0.2905, p = 0.0013, figure 1H). Analyzing the effect of HIV infection on antibody levels to DBL5e revealed HIV negative women to have higher antibody levels than HIV positive women  in all parity groups, but the difference was only significant in multigravidae (Kruskall-Wallis ANOVA, p = 0.023, Dunn's Multiple Comparison test, p,0.05 (PG versus MG), figure 2). We also investigated the possible correlation between parasitemia (parasites per ml) at the time of serum collection and antibody levels against VSA-PAM as well as against DBL5e ( figure S4). We found a weak significant correlation (Spearman r = 0.2134, p = 0.0005) between antibodies against VSA-PAM and parasitemia, something that holds true also for primigravidae (Spearman r = 0.2819, p = 0.0002) and multigravidae (Spearman r = 0.2833, p = 0.0175), but not for secundigravide. We also found a weak significant correlation (Spearman r = 0.2000, p = 0.0123) between antibodies against VAR2CSA DBL5e and parasitemia in primigraviade but nor for secundi-or multigravidae nor for these three groups together.

Antibody Affinity against VAR2CSA DBL5e
Due to the interesting pattern of VAR2CSA DBL5e recognition by IgG in immune sera, we chose to continue working with this domain, investigating affinity properties of DBL5e specific antibodies. Using recombinant FCR3 VAR2CSA DBL5e and a surface plasmon resonance technique, we investigated the dissociation rate constant (k d ) of immune antibodies as an indicator of their affinity. This technique allows comparison of different sera even though the concentration of the specific antibodies is not known, since k d is concentration independent. Even though multigravid women have higher levels of antibodies against DBL5e than primigravidae, these antibodies did not show higher affinity against this VAR2CSA DBL5e domain (figure 3A). However, sera from primigravid HIV positive women showed significantly higher k d , hence lower affinity, than sera from primigravid HIV negative women, indicating HIV infection to impair the affinity of VAR2CSA DBL5e domain specific antibodies (t-test, p = 0.0230, figure 3B). Among multigravid women, even though HIV negative women had higher levels of antibodies than HIV positive women, there was no significant difference in the affinity of these antibodies against DBL5e (figure 3C). Malaria naïve sera showed binding only at background levels and as an extra control, we also used non-immune IgG and  IgM as well as IgG depleted non-immune sera, neither exhibiting binding using the surface plasmon resonance technique (data not shown). Both antibody levels to total VSA-PAM and to DBL5e correlated significantly with the dissociation rate constant (Spearman r = 20.3571, p = 0.0002 and r = 20.3414, p = 0.0024 respectively, figure 4A and B).

Discussion
In this study investigating VAR2CSA DBL-domain specific immune response, we show the VAR2CSA domains DBL5e and DBL6e to be recognized in a gender specific manner ( figure 1B). DBL5e also displays a parity dependent recognition, as previously described for VSA-PAM antibodies [34,35] suggesting this domain to be important in a protective response against PAM ( figure 1E). All DBL-domains contain both conserved and polymorphic regions targeted by surface reactive antibodies, but the conserved regions are most prominent in DBL3x and DBL5e [36]. Further, DBL5e is highly conserved between both laboratory and clinical isolates, and most sequence differences localize to flexible loops in the protein [37,38]. A recent study indicates that antibodies raised against DBL3x and DBL5e domains are highly cross-reactive with several placental isolates [39]. We demonstrate a correlation between antibodies against a VAR2CSA expressing parasite line and DBL5e domain specific antibodies, suggesting this domain to be a prominent target of antibodies developed in malaria infection during pregnancy. A study by Oleinikov et al showed that most antibodies elicited following immunization with DBL5e and DBL6e recognize polymorphic epitopes in the native protein [40]. Levels of adhesion-inhibitory antibody correlated with levels of anti-DBL5e antibodies [41], and a recent study illustrated how antibodies raised against DBL3x and DBL5e were highly crossreactive with heterologous parasites [42]. Furthermore, in this study we used DBL5e from three different parasite isolates (CS2, to measure total VSA-PAM antibody levels; 3D7 domains expressed in CHO cells; and FCR3 in affinity studies) with results correlating significantly ( figure 4A and B). Also, the rMFI for antibodies against VSA-PAM and antibodies against DBL5e have a very strong resemblance to each other across the population ( figure 4A and B). Levels of anti-DBL5e IgG have also been positively associated with birth weight, again indicating their protective effect [6]. A VAR2CSA vaccine will need to elicit a cross-reactive antibody response in order to be widely effective, and our data in conjunction with earlier studies [36,41,43] indicate DBL5e to be a promising vaccine candidate.
In order to succeed in blocking iRBC adhesion to CSA in the placenta of pregnant women, it is possible that antibodies need not only to bind, but also to do so at strong affinity. Using the surface plasmon resonance technique, we found that even though multigravid women have slightly higher levels of antibodies against DBL5e than primigravidae, there was no significant difference in the dissociation rate constant of antibodies against this VAR2CSA domain between the groups. Diving deeper into the different parities and the effect of HIV co-infection, antibodies against DBL5e in primigravid HIV positive women tested had lower affinity than sera from HIV negative primigravidae from the same region (figure 3B), suggesting HIV infection to impair the strength of antibody binding. Among multigravid HIV, levels of DBL5e antibodies were lower in HIV positive women (figure 2) but affinity was similar to that of antibodies present in HIV negative women ( figure 3C). However, levels of total VSA-PAM antibodies as well as anti-DBL5e antibodies correlate significantly with the dissociation rate constant used as an indicator of affinity of these antibodies ( figure 4A and B). Since HIV in multigravid women seems to impair a pregnant woman's ability to control parasitemia, levels of antibodies are an important determinant of favorable birth outcomes and as a sign for less disease in these women [28].
Our data indicate that HIV infection impairs antibody affinity in primigravidae more than in multigravidae. One hypothesis would argue that antibody affinity is important in primigravidae when antibody levels are low whereas antibody levels and functionalities such as facilitation of opsonic activity may be more important determinants of outcome in multigravid women. Phagocytic activity of cytophilic antibodies is known to be impaired in multigravidae infected with HIV whereas no difference is seen in HIV infected and HIV uninfected primigravid women [23,24].
Many studies have investigated presence, levels and function of antibodies targeting VSA-PAM, but less is known concerning the dynamics of binding interactions. It could be so that antibodies, to be protective, need to bind with high affinity. Some studies found no association between high levels of VSA-PAM antibodies and adhesion inhibition [44], whereas some clearly show this correlation [34]. We hypothesize that high affinity antibodies are protective whereas low affinity antibodies are less so and thereby rely more on increased levels and various antibody functions independent of the antibody binding equilibrium such as phagocytic activity and ability to block placental adhesion.
When designing a vaccine against pregnancy-associated malaria, various properties of antibody response against VSA-PAM are important to consider, especially in the presence of HIV coinfection. Conserved surface epitopes are interesting as possible vaccine components, if immune sera from several parts of the world recognize these epitopes. Considering that the DBL5e specific response correlates well with acquired antibodies to VSA-PAM expressed on iRBC surface selected for VAR2CSA expression, this is an interesting domain for vaccine development. This study presents important information on altered mechanisms of antibody response toward VSA-PAM in presence of HIV infection, something that is of great interest in the continued quest for an efficient vaccine against pregnancy-associated malaria.

Ethics Statement
The study was approved by the College of Medicine Research and Ethics Committee, University of Malawi, the Royal Melbourne Hospital Clinical Research Ethics Committee, and the Institutional Review Boards of the Universities of Michigan and North Carolina. Written informed consent was sought from all eligible women involved in the study for blood collection and HIV testing. Adult males gave written informed consent for collection of blood as part of studies approved by the same committee.

VAR2CSA Transfectants
The six domains of 3D7 PFL0030c var2csa (Genbank accession no. NP_701371) and DBL3x of A4 PFL0030c var2csa (Genbank accession no. AAQ73926) were amplified from genomic DNA by PCR and cloned into the pSRa5-12CA5 vector (Affymax Research Institute, Palo Alto, CA) as described [14]. Chinese hamster ovary PgsA-745 (CHO-745) cells deficient in glycosaminoglycans (American Type Culture Collection, Manassas, VA) were transfected with these plasmids, and stable transfectants expressing the various DBL-domains were used for flow cytometry assays.

Measurements of VSA-PAM Antibodies by Flow Cytometry
IRBCs of P. falciparum laboratory strain CS2 at 5-10% parasitemia of trophozoites were washed three times in PBS containing 1% fetal calf serum (PBS-1% FCS). Cells at 0.1% hematocrit (100 ml) were incubated with test sera at a 1:20 dilution in PBS-1% FCS for 30 min in 96-well v-bottom plates. This was followed by a 30-min incubation with 50 ml rabbit anti-human IgG (DAKO A0424) at a 1:100 dilution in PBS-1% FCS and a final incubation with 50 ml Alexa Fluor 488-conjugated donkey anti-rabbit IgG (Molecular Probes A-21206) at a 1:500 dilution and 10 mg/ml ethidium bromide at a 1:100 dilution in the dark for 30 min. All incubations were performed at room temperature and after each incubation, the cells were washed three times in PBS-1% FCS. After the final wash, cells were resuspended in 200 ml PBS and analyzed with a FACSCalibur flow cytometer with Cell Quest software (BD Biosciences). Included in all runs were a positive control made up of a pool of hyperimmune sera from pregnant women resident in Blantyre, Malawi, and two negative controls made up of pooled sera from malaria non-exposed adults in Melbourne, Australia. The geometric mean fluorescence intensity (MFI) was calculated as a measure of IgG binding to iRBC. IRBCs were gated and 1000 positive cells were collected on the basis of ethidium bromide fluorescence. Samples were determined to have antibodies if the MFI was greater than the mean of the negative controls plus 2 standard deviations (SD). Sample readings were assigned relative values by using the formula ((sample MFI) -(negative control MFI))/((positive control MFI) -(negative control MFI))*100.

Biacore Surface Plasmon Resonance
Antibody affinity to VAR2CSA DBL5e measurements was performed on a BIAcore 3000 instrument (Biacore AB, Uppsala Sweden). Recombinant FCR3 VAR2CSA DBL5e and the control 3D7 DBL6c (kind gift from Ali Salanti) was coupled to a CM5 sensor chip (Biacore AB) using an amine coupling kit, according to the instructions of the manufacturer. FCR3 VAR2CSA DBL5e in sodium acetate pH 4.5 and 3D7 DBL6c in sodium acetate pH 4.8 were injected to flow cell 2 and 3 (FC2 and 3) respectively to be immobilized on the sensor surface reaching a total of 1200 resonance units (RU). No protein was injected into FC1, in order to serve as control for background binding to the dextran matrix. A continuous flow of HBS-EP buffer (0.01 M Hepes, pH 7.4, 0.15 M NaCl, 3 mM EDTA and 0.005% (v/v) Surfactant P20, Biacore AB) passing over the sensor surface at 30 ml/min was maintained and experiments were performed at 25uC. Each serum sample to be tested was diluted 1:7.5 and 1:15 in HBS-EP buffer. Pooled serum samples from Malawian hyperimmune multigravid women were used as a positive control. Sera from malaria-naïve Melbourne donors, non-immune IgG and IgM as well as sera depleted from IgG were used as controls. The sensor chip was regenerated with a pulse of 10 mM glycine pH 1.5. The dissociation rate constant, k d , was evaluated for each sample showing specific DBL5e binding, using the software BIAevaluation 3.0 (Biacore AB).

Statistical Analysis
Statistical analyses were performed using Excel and GraphPad Prism and InStat3. Appropriate tests were performed (Kruskal-Wallis ANOVA followed by Dunn's Multiple Comparison Test ( figure 1 and 2), Spearman rank test ( figure 4, figure S4) for the non-normal distributed data) and t-test for the normal distributed data (figure 3)) depending on properties of respective data. respectively, and non-transfected CHO cells (black). All cells were incubated with an anti-hemagglutinin (HA) antibody (targeting the transfection construct) followed by a secondary antibody labelled with Alexa-488 (FL1). Transfection levels of these three stable transfectants were consistently great and highly similar to one another. Found at: doi:10.1371/journal.pone.0009230.s003 (0.91 MB TIF) Figure S4 Correlation between antibody levels and parasitemia. Correlation between parasitemia (parasites per ml) at the time of serum collection and antibody levels against VSA-PAM as well as against DBL5e. We found a weak significant correlation (Spearman r = 0.2134, p = 0.0005) between antibodies against VSA-PAM and parasitemia (A), something that holds true also for primigravidae (Spearman r = 0.2819, p = 0.0002) (C) and multigravidae (Spearman r = 0.2833, p = 0.0175) (G), but not for secundigravide (E). We also found a weak significant correlation (Spearman r = 0.2000, p = 0.0123) between antibodies against VAR2CSA DBL5e and parasitemia in primigravidae (D) but nor for secundi-or multigravidae (F and H) nor for these three groups together (B). Found at: doi:10.1371/journal.pone.0009230.s004 (0.94 MB TIF)