IgG Responses to the Plasmodium falciparum Antigen VAR2CSA in Colombia Are Restricted to Pregnancy and Are Not Induced by Exposure to Plasmodium vivax

Clinical immunity to malaria is associated with the acquisition of IgG specific for members of the Plasmodium falciparum erythrocyte membrane protein 1 (PfEMP1) family of clonally variant antigens on the surface of infected erythrocytes (IEs). The VAR2CSA subtype of PfEMP1 mediates IE binding in the placenta.

Outside Africa, Plasmodium vivax is responsible for about one-third of all malaria cases, including most cases in Latin America (1). Colombia, with approximately 8 million people living in areas where malaria is endemic, ranks third in the American continent in terms of malaria transmission, with half a million cases being reported between 2007 and 2013 (1,2).
In areas with stable P. falciparum transmission, children and pregnant women are at a particularly high risk of malaria. This is because substantial protective immunity is acquired during childhood and adolescence, making clinical episodes and severe cases uncommon among adults (3). Malaria protection is mainly antibody mediated, and antigens like P. falciparum erythrocyte membrane protein 1 (PfEMP1) expressed on the infected erythrocyte (IE) surface are important targets (4). When women become pregnant, particularly for the first time, they become highly susceptible to malaria, despite any clinical immunity acquired earlier in life (5). This appears related to the ability of P. falciparum parasites to express a particular PfEMP1 subtype called VAR2CSA, which is antigenically distinct from all other PfEMP1 proteins and facilitates the selective accumulation of IEs in the placenta (6,7). The expression of VAR2CSA is generally assumed to be incompatible with parasite survival in nonpregnant individuals (6,8). Therefore, the acquisition of VAR2CSA-specific IgG is normally regarded as being pregnancy restricted, despite a few reports of sporadic, low levels of VAR2CSA-specific IgG among P. falciparum-exposed men, children, and nulligravidae (9)(10)(11)(12).
The intensity of P. falciparum transmission in Colombia is low. Although at least 1 million women of reproductive age live in areas of the country where malaria is endemic, malaria in pregnancy, including placental malaria, is uncommon (13)(14)(15). It was therefore highly surprising when Gnidehou et al. reported a high prevalence of VAR2CSA-specific IgG in Colombia, not only in women with a history of pregnancy but also among nulligravidae, men, and children living in areas of the country where malaria is endemic (16). That same group recently proposed that the high VAR2CSA reactivity among Colombians might be related to cross-reactive antibodies induced by the P. vivax Duffy-binding protein (PvDBP) (17), which has low-level homology to PfEMP1.
The above-described findings from Colombia either point to a completely new and unanticipated mode of acquisition of VAR2CSA-specific IgG or suggest that the current understanding of immunity to placental malaria is incomplete. We therefore set out to shed additional light on the prevalence and specificity of IgG recognizing VAR2CSAtype PfEMP1 in Colombian populations.

RESULTS
IgG recognizing recombinant PfEMP1 proteins expressed in baculovirustransfected insect cells is prevalent in plasma from Colombian pregnant women, men, and children. Significant plasma levels of IgG specific for VAR2CSA-type PfEMP1 are usually restricted to P. falciparum-exposed women who are or have recently been pregnant, and the levels generally correlate with parity among such women (18,19). However, Gnidehou et al. recently reported that this does not appear to be the case in Colombia, as antibodies to recombinant VAR2CSA proteins did not depend on parity and were equally prevalent among malaria-exposed women, men, and children (16). To assess this surprising finding, we first measured levels of FV2 BIC -specific IgG in plasma samples from different groups of Colombian donors (Fig. 1A). All the samples listed in Table 1 were included in these enzyme-linked immunosorbent assays (ELISAs). In agreement with data from the previous Colombian study, FV2 BIC -specific IgG was detected frequently in samples from all the malaria-exposed donor groups studied here (set 1, set 2, and set 3b). With the exception of exposed men (set 1b; P Ͻ 0.001) and pregnant women (set 2c; P Ͻ 0.001), the levels were not statistically significantly different from those observed in samples from nonpregnant Ghanaian women who had been pregnant one or more times previously (set 5). Plasma levels of FV2 BIC -specific IgG were significantly lower among unexposed Colombians (set 3a, set 4a, and set 4b; enrollment samples) than among exposed Colombians (set 1 and set 2; P Ͻ 0.001) but Infection and Immunity were still significantly higher than the levels among negative-control donors from Denmark (set 6; P Ͻ 0.001) (Fig. 1A). As reported previously by Gnidehou et al. (16), levels among the pregnant Colombian women studied did not differ between primigravidae and multigravidae (Fig. 1B). Although the FV2 BIC -specific IgG levels also did not differ significantly between primigravidae and multigravidae among the Ghanaian women included here, such a difference was evident in the much larger sample set (18) from which the samples in set 5 were randomly drawn. We also measured plasma levels of IgG specific for another full-length recombinant protein (FV6 BIC ), expressed as FV2 BIC but representing a PfEMP1 antigen that is commonly recognized by plasma IgG from naturally exposed African donors. The expression of this PfEMP1 antigen is not restricted to parasites infecting pregnant women (18). The results ( Fig. 1C and D) were strikingly similar to those obtained with FV2 BIC . Furthermore, there was a significant and strong positive correlation between the FV2 BIC and FV6 BIC data among the Colombian samples (Fig. 1E) that was not observed in the corresponding data from Ghana (Fig. 1F). Taken together, these findings are in agreement with data from the previous study by Gnidehou et al. (16) and extend the surprisingly high and prevalent plasma antibody reactivity to recombinant VAR2CSAtype PfEMP1 among Colombian donors to comprise reactivity to a recombinant PfEMP1 protein (FV6 BIC ) that is not restricted to pregnancy (18,20).
IgG recognizing native VAR2CSA-type PfEMP1 expressed on the surface of P. falciparum-infected erythrocytes is rare in plasma from Colombian individuals. To substantiate the above-described findings, we next measured plasma IgG reactivity against erythrocytes infected with P. falciparum parasites selected in vitro to express the PfEMP1 antigens represented by the recombinant proteins used in the ELISAs described above. Subsets of the plasma samples in the various sets, selected to include those with the highest FV2 BIC reactivity by ELISAs, were used, as specified in Table 1. Between 7% (children and men) and 20% (pregnant women) of the samples from Colombia had IgG levels to IT4VAR04 (the native counterpart of FV2) above the negative cutoff level ( Fig.  2A). Sixty-three percent of the nonpregnant Ghanaian women had levels of IT4VAR04specific IgG above the cutoff. Between 61% (children) and 84% (pregnant women) of samples had levels of IgG recognizing IEs expressing HB3VAR06 (the native counterpart of FV6) above the cutoff (Fig. 2B). Antibody levels to the corresponding native and recombinant antigens did not correlate significantly, regardless of whether all the Colombian samples ( Fig. 2C and E) or only the samples from the Colombian pregnant women ( Fig. 2D and F) were considered. In contrast, plasma from Ghanaian women with a pregnancy history reacted strongly with both IT4VAR04 and HB3VAR06 ( Fig. 2A and B; see also Fig. S1A and S2A in the supplemental material). Furthermore, levels of IgG recognizing the corresponding native and recombinant versions of either antigen were positively and significantly correlated among the samples from the Ghanaian women ( Fig. 2G and H). These results suggest that the plasma samples from the Colombian donors recognized a component(s) that is present in both recombinant PfEMP1 antigens expressed in baculovirus-transfected insect cells but is not present in the corresponding native antigens and therefore is possibly not related to malaria exposure.
IgG recognizing recombinant VAR2CSA-type PfEMP1 expressed in CHO cells is rare in plasma from nonpregnant Colombian individuals. To assess the hypothesis that the reactivity to FV2 BIC among the Colombian donors was not P. falciparum specific, we generated another recombinant full-length PfEMP1 antigen containing the same amino acid sequence as that of FV2 BIC but expressed in a mammalian cell line   (Chinese hamster ovary [CHO]) rather than in insect cells. Levels of IgG to this protein (FV2 CHO ) were generally low among all groups of Colombian donors (Fig. 3A). The differences between the Colombian and Ghanaian donors were statistically significant for all the groups except sets 2a and 2b, despite the fact that the Ghanaian women were not pregnant and therefore were not expected to have high levels of VAR2CSAspecific IgG, due to the rapid decline of these antibodies postpartum (21,22). FV2 CHOspecific IgG levels were weakly and positively correlated with IgG levels to FV2 BIC (Fig.  3B), in contrast to the plasma samples from the Ghanaian donors, where IgG levels to the these antigens were very strongly correlated (Fig. 3C). Antibody levels to FV2 CHO and the corresponding native protein did not correlate significantly when all the samples from Colombia were considered (Fig. 3D). When the samples were further analyzed according to donor type (see Fig. S3 in the supplemental material), a weak positive correlation between FV2 CHO -and IT4VAR04-specific IgG was observed in the samples from the Colombian pregnant women (set 2c) but not in the samples from children (set 1a and set 2a) or men (set 1b and set 2b). This suggests that some of the Colombian women had been exposed to VAR2CSA-positive IEs during their pregnancy. These findings suggest that most of the apparently VAR2CSA-specific IgG in the Colombian plasma samples actually recognized so-far-unidentified epitopes present only in the recombinant proteins produced in insect cells. In contrast, levels of IgG reacting with FV2 BIC , FV2 CHO , and the corresponding native protein on IEs were positively correlated among the samples from Ghana ( Fig. 2G and H and 3C), indicating that IgG recognizing these epitopes is not induced in Ghana. IgG reactivity to FV2 BIC in Colombian plasma samples cannot be depleted in vitro by preexposure to FV2 CHO . We next sought to provide additional evidence that the reactivity to FV2 BIC among the Colombian donors was indeed directed toward non-PfEMP1-specific epitopes that are present in recombinant proteins expressed in baculovirus-transfected insect cells but absent from proteins expressed in CHO cells and absent in native PfEMP1 proteins. To this end, we sequentially measured IgG reactivity against FV2 BIC , FV2 CHO , and FV6 BIC in plasma pools transferred from one ELISA plate to the next (Fig. 4). When pools were depleted in this way of IgG reacting with FV2 BIC (Fig. 4A, D, G, and J), FV2 CHO ( Fig.  4B, E, H, and K), or FV6 BIC (Fig. 4C, F, I, and L), respectively, reactivities in the plasma pools from Colombian ( Fig. 4G to I) and Ghanaian (Fig. 4J to L) women were significantly reduced from the first ELISA to the second ELISA. In marked contrast, the reactivities to FV2 BIC in pools from Colombian unexposed donors (Fig. 4E) and Colombian pregnant women (Fig. 4H) were unaffected by two rounds of FV2 CHOspecific IgG depletion, whereas essentially no reactivity was left in the Ghanaian pool treated in this way (Fig. 4K). Conversely, reactivity to FV2 BIC was essentially absent from each of the two initially FV2 BIC -reactive Colombian pools after two FV6 BIC ELISAs ( Fig. 4F and I), whereas the FV2 BIC reactivity in the Ghanaian pool was largely unaffected (Fig. 4L). Taken together, these depletion experiments strengthen the hypothesis that the widespread Colombian IgG reactivity to recombinant PfEMP1 antigens produced in baculovirus-transfected insect cells is not P. falciparum specific.
Experimental exposure to P. vivax-infected mosquitoes does not induce IgG specific for VAR2CSA-type PfEMP1. It has been proposed that exposure to P. vivax can elicit IgG responses that cross-react with VAR2CSA-type PfEMP1 (16,23). This flow cytometry. (C and D) Correlation (r s ) between IgG levels specific for recombinant (FV2 BIC ) and native (IT4VAR04) VAR2CSA-type PfEMP1 in samples from Colombian exposed, healthy, and naturally infected individuals (set 1 and set 2) (C) or from Colombian pregnant women only (D). (E and F) Correlation (r s ) between IgG levels specific for recombinant (FV6 BIC ) and native (HB3VAR06) non-VAR2CSA-type PfEMP1 in Colombian samples (set 1 and set 2) (E) or in samples from Colombian pregnant women only (set 2c) (F). (G and H) Correlation (r s ) between IgG levels specific for recombinant (FV2 BIC ) and native (IT4VAR04) VAR2CSA-type PfEMP1 (G) and between IgG levels specific for recombinant (FV6 BIC ) and native (HB3VAR06) non-VAR2CSA-type PfEMP1 (H) in samples from Ghanaian women. The statistical significance levels of the correlations are also shown. The panel layout is the same as in Fig. 1.

VAR2CSA-Specific IgG in Colombian Individuals
Infection and Immunity hypothesis is consistent with both the pregnancy restriction of VAR2CSA reactivity in Africa (where P. vivax is rare and not an important cause of placental malaria [24]) and the apparent absence of such restriction in Colombia (where P. vivax is common). Those same authors proposed that the cross-reactivity is mediated by PvDBP (17). We therefore measured levels of FV2 BIC -, FV2 CHO -, FV6 BIC -, and PvDBP-specific IgG in plasma samples from 16 adults participating in a controlled human P. vivax malaria infection (PvCHMI) trial (set 3) (25). In all cases, the data were normalized by subtraction of the background reactivity at enrollment to reveal only the IgG reactivity induced in response to P. vivax infection. After normalization, no reactivity was observed at the time of patency (Fig. 5), meaning that antibodies did not develop after experimental , the plasma pools were transferred directly from the first to the second ELISA plates and then from the second to the third ELISA plates. In series 1, an FV2 BIC -specific ELISA was used for all three consecutive measurements. In series 2, the first two measurements were FV2 CHO specific, while the third ELISA was FV2 BIC specific. In series 3, the first two measurements were FV6 BIC specific, while the third ELISA was FV2 BIC specific. To allow comparison of data among panels, levels are expressed as how many folds higher they were than the levels in the corresponding control ELISAs without antigen (first and second ELISAs) and shown as means and SD. *, P Ͻ 0.05; ***, P Ͻ 0.001 (using Kruskal-Wallis test followed by Dunn's multiple-comparison analysis).

VAR2CSA-Specific IgG in Colombian Individuals
Infection and Immunity infection. One unexposed donor showed a weak positive response to FV6 BIC (Fig. 5C), and one exposed donor showed a strong positive response to PvDBP (Fig. 5D) (Fig. 5). Taken together, these data indicate that exposure to P. vivax parasites does not induce IgG that robustly recognizes PfEMP1 in general or VAR2CSAtype PfEMP1 in particular. PvDBP-specific antibodies do not recognize PfEMP1. To investigate further whether the Colombian plasma IgG reactivity with recombinant PfEMP1 proteins could be due to cross-reactivity with IgG induced by PvDBP, we tested the ability of PvDBP-specific antibodies raised in different species to recognize recombinant and native PfEMP1. We first tested the reactivity of the mouse monoclonal antibody 3D10 (28), which was previously reported to cross-react with VAR2CSA-type PfEMP1 (17). This monoclonal antibody in fact showed low reactivity with FV2 BIC (endpoint titer of 0.08 g/ml using mouse IgG as a reference) (Fig. 6A) but reacted only very weakly with FV2 CHO (endpoint titer of 10 g/ml) (Fig. 6C) and not with the corresponding native PfEMP1 protein IT4VAR04 (see Fig. S1B to S1D in the supplemental material). The 3D10 antibody reacted weakly with FV6 BIC (Fig. 6E), much to the same extent as it reacted with FV2 BIC (endpoint titer of 0.08 g/ml), but 3D10 did not label the native counterpart of FV6, HB3VAR06 (Fig. S2 B). It thus appears that 3D10, in addition to its specificity for PvDBP (28), also has some reactivity with recombinant proteins expressed in baculovirus-transfected insect cells. In marked contrast, we observed no reactivity of the PvDBP-specific human monoclonal antibody DBL10, generated from memory B cells of a volunteer in a recent PvDBP vaccination trial (29; Thomas Rawlinson, unpublished data), with any of the recombinant full-length PfEMP1 proteins (Fig. 6A, C, and E) or with IT4VAR04-positive (Fig.  S1E) or HB3VAR06-positive (Fig. S2C) IEs. A corresponding lack of reactivity was observed with an antiserum from PvDBP-immunized rabbits (30) (Fig. 6B, D, and F and Fig. S1F and S2D). All the following PvDBP-specific antibody reagents showed high reactivity with a recombinant PvDBP expressed in Escherichia coli (28): rabbit anti-PvDBP serum (optical density [OD] of 0.48 Ϯ 0.10 at a 1:2,560 dilution), DBL10 (OD of 2.88 Ϯ 0.16 at 0.08 g/ml), and 3D10 (OD of 1.86 Ϯ 0.09 at 0.08 g/ml). Taken as a whole, data from these experiments indicate that immunization with PvDBP does not generally induce antibodies that cross-react with PfEMP1.

DISCUSSION
VAR2CSA-type PfEMP1 proteins mediate the selective accumulation of IEs infected with P. falciparum in the placenta (19,(31)(32)(33). This can have serious consequences for a pregnant woman as well as for her unborn child (fetus), because the ensuing inflammation compromises the transfer of nutrients, antibodies, and waste products between mother and fetus (reviewed in reference 7). Placental malaria is an important component of the unacceptable infection burden imposed by P. falciparum on people living in the tropics, in particular sub-Saharan Africa. More than 50 million pregnancies each year are at risk of placental malaria (34), and Ͼ100,000 infants die because of placental malaria (5). Following exposure to VAR2CSA-type PfEMP1 during pregnancy, women acquire substantial IgG-mediated protection against placental malaria, which is therefore markedly concentrated among women of low parity, in particular primigravidae (reviewed in reference 6). This fact, and the identification of the so-called minimal binding domain in VAR2CSA-type PfEMP1 that binds to its cognate receptor, oncofetal chondroitin sulfate, has raised hopes that it may be possible to vaccinate against placental malaria (35)(36)(37)(38). Clinical phase Ia/b trials of such vaccines are currently in progress (39,40).

VAR2CSA-Specific IgG in Colombian Individuals
Infection and Immunity A few years back, Gnidehou et al. (16) reported widespread IgG reactivity to VAR2CSA-type PfEMP1 in Colombian children and adults of both sexes and in women of all parities. Those findings are in marked contrast to essentially all other studies, mostly from Africa, documenting that high levels of VAR2CSA-specific IgG are restricted to pregnant women, that levels correlate with parity, and that VAR2CSA-specific IgG declines to low levels soon after delivery (reviewed in reference 6). Those authors speculated that the reason for their unanticipated findings might be the presence of another malaria parasite, P. vivax, which is essentially absent from most of Africa. More specifically, they recently proposed that VAR2CSA-specific IgG might be induced by cross-reactive epitopes in the P. vivax antigen PvDBP (17). The studies by Gnidehou et al. (16,17) thus point to a completely new and unanticipated mode of acquisition and maintenance of VAR2CSA-specific IgG, although BLAST analysis of IT4VAR04 showed only low sequence homology (26%) with PvDBP (PVP01_0623800.1-p1). The findings from Colombia also raise questions regarding the rationale that underpins current efforts to develop VAR2CSA-based vaccines to protect young women against placental malaria (reviewed in reference 41). We therefore sought to examine in detail the unexpected results from the Colombian studies, using both recombinant and native VAR2CSA proteins. We used full-length recombinant proteins rather than the smaller constructs (DBL5, DBL3X, and ID1-ID2) used in the previous studies (16,17), as several domains are required for optimal VAR2CSA-mediated IE sequestration (42,43). We did not find evidence that VAR2CSA-specific IgG is acquired any differently in Colombia than in Africa. Furthermore, we found no indication to support the hypothesis that exposure to P. vivax plays a role in these processes. Instead, our data show that responses to recombinant antigens can be misleading if not checked carefully against data obtained with the corresponding native antigens. Specifically, many Colombians appear to possess IgG that reacts with recombinant PfEMP1 proteins expressed in baculovirus-transfected insect cells. Whether this is also the case for other proteins not related to malaria is not known. Importantly however, IgG reacting with a recombinant protein expressed in this way (e.g., FV2 BIC ) does not necessarily react with a corresponding protein expressed in another system (e.g., FV2 CHO ) or indeed with the native protein that it represents (e.g., IT4VAR04). A possible explanation for the discrepancy between FV2 BIC IgG reactivity on the one hand and FV2 CHO and IT4VAR04 reactivity on the other is the marked differences in protein glycosylation among malaria parasites (largely limited to the addition of glycosylphosphatidylinositol [GPI] anchors) (reviewed in references 44 and 45), insect cells (simple carbohydrates only) (reviewed in reference 46), and CHO cells (complex, human-like carbohydrates) (reviewed in reference 47). IgG reacting with recombinant antigens thus does not necessarily react (only) with the antigen-specific protein core but may also recognize posttranslational modifications that the antigen can share with many unrelated proteins produced in the same system. In the present case, plasma IgG reactivity to FV2 BIC among the Colombian donors seemed to be largely of the latter kind, whereas FV6 BIC reactivity (and FV2 BIC reactivity among the infected pregnant women) appeared to be a mix of both. We did not pursue the nature of the epitopes in the baculovirus proteins recognized by IgG in plasma from Colombia but not from Ghana, as this is likely to be a major undertaking that is well beyond the scope of the present work. For the same reasons, we did not investigate why these antibodies are common in Colombia while seemingly absent in Africa. It might be worth noting that the highest IgG responses to insect cell-expressed recombinant antigens among Colombian donors were found among children, both here and in the previous study. Exposure of Colombians to other pathogens or insects is another potential explanation that was not investigated here.

VAR2CSA-Specific IgG in Colombian Individuals
Infection and Immunity Apart from advising the exertion of great care when interpreting findings using recombinant proteins, we conclude from the present study that VAR2CSA-specific immunity is acquired in the same way (i.e., by exposure to VAR2CSA-positive IEs during pregnancy) in Colombia as in Africa. Furthermore, acquired immunity to VAR2CSA-type PfEMP1 seems not to be influenced by exposure to P. vivax. The rationale for the development of a VAR2CSA-based vaccine against placental malaria therefore stands (41).

MATERIALS AND METHODS
Ethical approval. The original studies from which the samples used here were obtained were approved by the Institutional Review Boards of the Malaria Vaccine and Drug Development Center (MVDC) and the Centro Médico Imbanaco, Cali, Colombia; by the Institutional Review Board of the Noguchi Memorial Institute for Medical Research, University of Ghana; and by the Regional Research Ethics Committees, Capital Region of Denmark. The Colombian clinical trials were registered at Clinical-Trials.gov (registration numbers NCT01585077 and NCT01082341). The present study was approved by the Institutional Review Board at MVDC. Study sites. Stored plasma samples from four previous studies involving five locations of Colombia were used. Cali, Buenaventura, Tumaco, and Quibdó are located in the western and Pacific regions, while Tierralta is situated in the northwestern Caribbean region (Table 1) (48,49). Cali (Department of Valle del Cauca) is a city without malaria transmission. Buenaventura had an average annual parasite incidence (API) of 3.1 in the period from 2011 to 2013. It has a population of predominantly African descent, with indigenous and mestizo minorities; most malaria cases (ϳ75%) are caused by P. vivax. Tumaco (Department of Nariño; API ϭ 10.3) has a highly predominant Afro-descendant population, and P. falciparum dominates (ϳ85%) in this region. Quibdó (Department of Chocó; API ϭ 25) is also inhabited mainly by Afro-descendants, and most malaria episodes (ϳ70%) are caused by P. falciparum. Tierralta (Department of Córdoba; API ϭ 6.7) is inhabited mostly by mestizo populations, with a small Amerindian Emberá Katío indigenous community; P. vivax dominates (ϳ85%) in this region.
Study samples. Several sets of samples were studied ( Table 1). The first set (set 1) ( Table 1) was obtained from a series of cross-sectional, active case detection studies carried out from 2011 to 2014 in Buenaventura, Tumaco, and Tierralta (50). Samples from children (set 1a) and men (set 1b) who were negative for malaria infection by quantitative PCR (qPCR) but with self-reported previous malaria exposure were included.
The second set of samples (set 2) was obtained from a passive case detection study conducted from 2011 to 2013 at malaria points of care (15) and from 2014 to 2016 in hospitals (51) of Quibdó, Tumaco, and Tierralta. Samples from children (set 2a) and men (set 2b) with acute P. vivax or P. falciparum malaria as well as a set of plasma samples from pregnant women with P. vivax or P. falciparum malaria (51, 52) (set 2c) were included.
The third set of samples (set 3) was collected from previously unexposed (set 3a) and semi-immune men and women (set 3b) participating in a controlled human P. vivax malaria infection (PvCHMI) trial carried out at MVDC in Cali. The unexposed (malaria antibody-negative) volunteers were from Cali, whereas the semi-immune volunteers were from Buenaventura. Their previous malaria exposure was confirmed by clinical history and by the presence of P. vivax-specific antibodies. Samples collected at enrollment, at patency (days 11 to 13), and 45 days after challenge (follow-up) were analyzed (25).
The fourth set of samples (set 4) was collected from previously unexposed adults during a clinical trial in Cali to assess the safety and protective efficacy of immunization with radiation-attenuated P. vivax sporozoites (PvRAS) (27). Samples from Duffy-positive (Fy ϩ ) (set 4a and set 4b) and Duffy-negative (Fy Ϫ ) individuals (set 4c) were collected at enrollment, after seven immunizations, and after PvCHMI with nonirradiated, P. vivax-infected mosquitoes.
Plasma samples from 32 nonpregnant women from Ghana with known reactivity to VAR2CSA-type PfEMP1 (18) (set 5) and from 11 Danish adults without malaria exposure (set 6) were included as positive and negative controls, respectively.
Recombinant parasite proteins. The entire ectodomains of the VAR2CSA-type PfEMP1 protein IT4VAR04 (FV2 BIC ; GenBank accession number GU249598) and of the non-VAR2CSA-type PfEMP1 protein HB3VAR06 (FV6 BIC ; GenBank accession number KP_203835) were produced in baculovirus-transfected Sf9 insect cells, as described previously (20,42). A third recombinant protein (FV2 CHO ), identical to FV2 BIC but expressed in suspension-adapted CHO cells (ExpiCHO; ThermoFisher Scientific) according to the manufacturer's instructions, was also used. In brief, a full-length construct with a C-terminal histidine tag was cloned into the vector pcDNA3.4 and used to transfect ExpiCHO-S cells. Glycosylation sites were not removed. Secreted recombinant proteins were purified from the supernatants by using Ni 2ϩ metal chelation agarose (HisTrap HP columns; GE Healthcare).
Recombinant P. vivax Duffy binding protein region II (PvDBP), produced in Escherichia coli, was a kind gift from John H. Adams (University of South Florida) (28).
Antibody reagents. For the in vitro selection of particular PfEMP1 proteins expressed on the IE surface (see below), we used human monoclonal antibody PAM1.4, which is specific for a conformational epitope in several VAR2CSA-type PfEMP1 proteins, including IT4VAR04 (53). We used a rabbit antiserum raised against FV6 BIC (20) to select for IE surface expression of the non-VAR2CSA-type PfEMP1 protein HB3VAR06.
PvDBP-specific mouse monoclonal IgG1 antibody 3D10 (28) was a kind gift from John H. Adams (University of South Florida). The human monoclonal IgG1 antibody DBL10, also specific for PvDBP (our unpublished data); the human monoclonal EBL040 IgG1 antibody, specific for an Ebola virus antigen (used here as a control reagent); as well as rabbit anti-PvDBP serum (30) were kind gifts from Thomas Rawlinson and Simon Draper (University of Oxford, UK).
Depletion of IgG reactivity to recombinant proteins. Pooled plasma samples from Colombian pregnant women (set 2c; n ϭ 56) and adults without previous malaria exposure (sets 3a, 4a, and 4b; n ϭ 22 enrollment samples), malaria-exposed nonpregnant women from Ghana (set 5; n ϭ 32), and unexposed Danish adults (set 6; n ϭ 11) were tested in three sequential ELISAs. In the first ELISA, levels of IgG specific for FV2 BIC , FV2 CHO , and FV6 BIC were measured. After incubation (1 h), the depleted pools were transferred directly to new plates coated with the same protein used in the first ELISA. After incubation (1 h), the further depleted pools were transferred to FV2 BIC -coated plates and once again incubated as described above. Further processing was performed as described above. For these experiments, the data were normalized by dividing the sample OD by the OD for the corresponding negative control (fold over control).
Malaria parasite culture and selection in vitro. P. falciparum clones IT4/FCR3 and HB3 were maintained in serum-free medium as described previously (55). Erythrocytes infected by late-stage IT4/FCR3 parasites were selected for surface expression of IT4VAR04 by immunomagnetic selection using protein A-coupled DynaBeads coated with PAM1.4, as described previously (56). A similar approach was used to select HB3-IEs for surface expression of HB3VAR06, using an HB3VAR06-specific rabbit antiserum (20). Transcription of the relevant var genes and IE surface expression of the corresponding PfEMP1 protein were monitored by quantitative real-time PCR (57) and flow cytometry (56), respectively. The genotypic identity of the parasites (58) and the absence of Mycoplasma contamination determined by using the MycoAlert mycoplasma detection kit (Lonza) were verified regularly.
Antibody reactivity with the surface of P. falciparum-infected erythrocytes. Plasma IgG reactivity against IEs expressing IT4VAR04 or HB3VAR06 was analyzed by flow cytometry, as described previously (56), using a Beckman Coulter FC500 flow cytometer for data acquisition and FlowLogic software (Inivai Technologies, Australia) for list mode data analysis. IgG binding to IEs was quantified as the relative median fluorescence intensity (rMFI) using the equation 100 ϫ [(MFI sample Ϫ MFI blank )/(MFI positive control Ϫ MFI blank )]. Negative cutoff values were calculated as the mean of rMFI values plus 2 SD obtained with the Danish control samples (set 6).
Statistical analysis. Data were analyzed by using GraphPad Prism version 6.0 (GraphPad Software, San Diego, CA, USA). All experiments reported here were repeated at least twice, with similar results. The Mann-Whitney U or Kruskal-Wallis test followed by Dunn's multiple-comparison test was used to compare two or more than two groups, respectively. Spearman's rank correlation (r s ) was used to assess the correlation between numeric variables. A chi-square test was used to compare differences in proportions. P values of Ͻ0.05 were considered statistically significant.

ACKNOWLEDGMENTS
We thank all patients, health care workers, and research staff involved in the original studies in Colombia and Ghana. We thank Maiken Visti, Lucía Buriticá, Angélica Castellanos, Carlos Castro, and Andrés Hoyos for their technical assistance. We thank John H. Adams, Simon Draper, and Thomas Rawlinson for providing the PvDBP reagents used and Peter Garred (Rigshospitalet, Denmark) for help with the production of the FV2 CHO protein.
This work was supported by the Independent Research Fund Denmark (DFF-4183-00539), the Novo Nordisk Foundation (NNF15OC0017654), and Rigshospitalet (R102-A4174). Sample collection in Colombia was supported by grants from the U.S. National Institutes of Health (NIAID/ICEMR grants U19AI089702 and 5R01HL086488), Colciencias