Acquisition of IgG to ICAM-1-Binding DBLβ Domains in the Plasmodium falciparum Erythrocyte Membrane Protein 1 Antigen Family Varies between Groups A, B, and C

Plasmodium falciparum erythrocyte membrane protein 1 (PfEMP1) is an important malaria virulence factor. The protein family can be divided into clinically relevant subfamilies. ICAM-1-binding group A PfEMP1 proteins also bind endothelial protein C receptor and have been associated with cerebral malaria in children. IgG to these PfEMP1 proteins is acquired later in life than that to group A PfEMP1 not binding ICAM-1.

cysteine-rich interdomain region (CIDR) domains (2)(3)(4)14). Particular subtypes of DBL␤ and CIDR␣ domains have been associated with binding to endothelial receptors such as intercellular adhesion molecule 1 (ICAM-1), endothelial protein C receptor (EPCR), and CD36 (15)(16)(17). More recently, we identified particular group A PfEMP1 proteins that can bind both ICAM-1 and EPCR (18). The ICAM-1-binding DBL␤ domains of such group A PfEMP1 proteins are characterized by a specific sequence motif, and IgG specific to them is acquired later in life than IgG specific for group A DBL␤ domains that do not bind ICAM-1 (18,19).
The acquisition pattern of ICAM-1-binding group B and C DBL␤-specific IgG is unknown. Therefore, the current study was designed to provide such data and to compare IgG reactivity to that of different subtypes of DBL␤ domains in Ghanaian children with or without P. falciparum malaria. The aim was to provide increased understanding of how antibody-mediated immunity to PfEMP1 is acquired following natural exposure to P. falciparum.

RESULTS
Identification of ICAM-1-binding DBL␤ domains. We have previously identified a novel family of group A ICAM-1-binding DBL␤ domains associated with cerebral malaria (18). Here, we used BLASTP searches and amino acid sequences encoding ICAM-1binding group B and group C DBL␤ domains from P. falciparum IT4 (20) to search for additional DBL␤ domains predicted to bind ICAM-1. Seven new sequences were identified by this approach. The encoded domains were a DBL␤3-type domain (GenBank accession no. KOB58843/HB3VAR34) and a DBL␤5-type domain (KOB63129/HB3VAR21) from HB3, two DBL␤5-type domains from Dd2 (AAA75396/Dd2VAR01A and KOB84711/ Dd2VAR21), one DBL␤5-type domain from 3D7 (PFL0020w), and one DBL␤5-type domain from each of two field isolates (ERS009963 and ERS010653). Dd2VAR21/ KOB84711 was identical to the previously published IT4VAR13, except for one residue (E instead of V) in DBL␣ and one residue (C instead of R) in the ATS region. All seven new domains bound ICAM-1 as predicted (Fig. 1A) and clustered together with other ICAM-1-binding DBL␤ domains from groups B and C (Fig. 1B). The average sequence similarity of the new group B and C ICAM-1-binding DBL␤ domains was 50%, which is comparable to that of previously identified ICAM-1-binding group A domains (58%) (18). Domains downstream of the ICAM-1-binding DBL␤ domains belonged to groups and subgroups similar to those in the previously identified ICAM-1-binding group B and C PfEMP1 proteins (Fig. 1C). To validate these findings further, we immunized rats with one of the domains (HB3VAR21-DBL␤5_D4) and used the antiserum to select P. falciparum HB3 to express VAR21 on the surface of IEs (Fig. 1D). HB3VAR21 ϩ IEs bound ICAM-1 at high levels ( Fig. 1E), confirming the ability to predict the IE adhesion phenotype from var gene sequences.
IgG specific for ICAM-1-binding group A DBL␤ domains dominates in healthy children. Group A PfEMP1-specific IgG is acquired earlier in life than antibodies targeting group B and C PfEMP1 antigens, and IgG to group A PfEMP1 therefore tends to dominate among healthy individuals living in areas with natural transmission of P. falciparum parasites (12,21,22). This was also the case here, when we compared IgG reactivity to ICAM-1-binding DBL␤ domains in group A and group B PfEMP1 proteins, employing plasma from a cohort of healthy Ghanaian children. Because only small sample volumes were available for this testing, we selected five of the ICAM-1-binding group B DBL␤ proteins identified above and five previously identified corresponding domains from group A (19). The IgG reactivity to each of the ICAM-1-binding DBL␤ domains from group A was higher than the reactivity to any of the domains from group B. This was consistently the case with samples from the same donors but collected at six different time points over a 1-year period ( Fig. 2; also see the data set in the supplemental material). This finding extends the earlier reports by demonstrating that the dominance of group A PfEMP1-specific IgG among healthy individuals remains when the comparison is restricted to ICAM-1-binding DBL␤ domains only. However, when we used plasma from children with acute P. falciparum malaria instead of plasma from healthy children, the pattern was the opposite, as levels of IgG specific for group B DBL␤ domains were higher than those for group A domains (Fig.  2). This suggests that acute malaria episodes markedly perturb the steady-state ("healthy") hierarchy of IgG reactivity to group A and group B PfEMP1 proteins, leading to a transient inversion of the group-specific IgG ratio. To examine this possibility further, we proceeded with a more detailed analysis of DBL␤-specific antibody responses, including kinetics and a larger panel of domains.
The PfEMP1 group hierarchy of DBL␤-specific IgG is influenced by malaria episodes. Plasma levels of IgG to P. falciparum antigens, including PfEMP1, tend to increase in relation to malaria episodes among individuals with natural exposure to these parasites but decline again shortly after resolution of the infection (23,24). The IgG responses to a large panel of DBL␤ domains in Ghanaian children monitored over 6 weeks after acute P. falciparum malaria episodes showed a similar pattern. Responses were highly variable, and marked but mostly transient IgG responses to all the different types of DBL␤ domains were observed in some but not all children ( Fig. 3; also see the data set in the supplemental material). Nevertheless, the most prominent overall increase in DBL␤-specific IgG reactivity associated with malaria episodes was to group B and C antigens (Fig. 3C). This finding was underpinned by analysis of IgG responses to the individual DBL␤ domains 2 weeks after admission, where responses generally peaked compared to the levels on admission and at week six ( Fig. 3 and 4; also see the data set in the supplemental material). At that time, IgG reactivity to all but one of the ICAM-1-binding DBL␤ antigens from group B and C PfEMP1 was higher than that to ICAM-1-binding group A domains (Fig. 4B). The difference between the two groups of PfEMP1 antigens was due to low IgG reactivity against group A DBL␤ domains in children younger than 7 years (Fig. 4C). Furthermore, the IgG reactivity to group B and C domains did not differ between age groups (P ϭ 0.5) (Fig. 4C), and significantly more  of the children were seropositive for group B DBL␤-specific IgG (62 to 85%) than for corresponding domains from group A (30 to 60%) (P Ͻ 0.05) (Table S3). Overall, it appears that most of the clinical episodes involved parasite populations expressing a mixture of PfEMP1 variants, including proteins containing different types of DBL␤ domains (Table S3), although parasites expressing group A ICAM-1-binding DBL␤ domains seemed underrepresented among the younger children. Furthermore, our data suggest that prominent responses to group B and C DBL␤ domains can cause a transient inversion of the ratio of IgG specific for ICAM-1-binding group A as well as group B and C DBL␤.
IgG reactivity to ICAM-1-binding DBL␤ domains in group A and groups B and C is similar in children with uncomplicated and those with severe, noncerebral malaria. Previous studies have shown that IEs obtained from young children with severe malaria primarily express PfEMP1 encoded by group A var genes, while expression of PfEMP1 encoded by group B and group C var genes appears associated with uncomplicated disease in slightly older children (8,(25)(26)(27). We therefore proceeded to compare the DBL␤-specific IgG responses in children with severe malaria to those in children with uncomplicated malaria. IgG reactivity to ICAM-1-binding group B and C DBL␤ domains was higher than that to corresponding group A domains when each group of patients was considered separately ( Fig. 5; also see the data set in the supplemental material). However, no statistically significant differences were noted when IgG reactivity to ICAM-1-binding DBL␤ domains from either group A or groups B and C was compared between children with severe or uncomplicated malaria (Fig. 5). While this may seem at variance with the well-documented relationship between expression of group A PfEMP1 and severe malaria, it should be noted that the subset of group A dual-receptor-binding PfEMP1 containing ICAM-1-binding DBL␤ domains has been associated specifically with cerebral malaria and not severe malaria in general (18), and that only 2 of the 124 children studied fulfilled the criteria for such a diagnosis.

DISCUSSION
P. falciparum causes the most severe form of malaria and is responsible for the vast majority of malaria-related deaths (28). This is not least due to the presence in this species of the PfEMP1 family of adhesive proteins, which are expressed on the surface of IEs in a mutually exclusive manner (only one variant expressed at a time) (4). Different members of the PfEMP1 family enable the adhesion of IEs to a range of vascular host receptors, which facilitate IE evasion of splenic clearance (5). It furthermore promotes tissue inflammation and organ dysfunction, while parasite switching among different  PfEMP1 family members (variants) frustrates the development of protective PfEMP1specific immunity (reviewed in reference 1).
Severe P. falciparum malaria has repeatedly been linked to IE adhesion to particular host receptors, mediated by groups and subgroups of structurally related group A PfEMP1 (reviewed in reference 1). Group A PfEMP1 proteins mediating adhesion to EPCR appear to be of particular relevance to the pathogenesis of severe malaria in patients with or without cerebral symptoms (8,17,25,26,29,30). Cerebral malaria, one of the most severe complications of P. falciparum infection (reviewed in reference 31), is associated with expression of a subgroup of group A PfEMP1 variants with a dualreceptor adhesion phenotype (18,29). These proteins carry an ICAM-1-binding DBL␤ domain next to an EPCR-binding CIDR␣1 domain. Although ICAM-1-binding DBL␤ domains also occur in group B and C PfEMP1, these domains are structurally distinct, do not have a neighboring EPCR-binding but a CD36-binding CIDR␣2-6 domain, and do not appear to play a role in cerebral malaria pathogenesis (18). Of the two domain types, CIDR␣ domains are more commonly recognized than DBL domains (12,32), and some studies have shown acquisition of group A CIDR␣1-specific IgG to precede immunity to group B and C CIDR␣2-6 domains (33), while others found no such link (12,34). Two recent studies found similar antibody reactivity against group A CIDR␣1 in uncomplicated and severe malaria during acute disease (34,35), while at convalescence older children with severe (likely noncerebral) malaria had higher antibody levels against such EPCR binding CIDR␣1 than those with uncomplicated malaria (34). The PfEMP1 reactivity between convalescent groups did not differ in the study by Kessler et al. (35), although higher seroprevalence to the conserved group A-associated ICAM-1binding DBL␤ domain (18) was observed relative to that of CIDR␣1.
In a longitudinal study assessing antibody acquisition against 32 non-ICAM-1binding DBL domains (three ␣, eight ␤, five ␥, nine ␦, six , and one ), asymptomatic Tanzanian children were shown to acquire antibodies to group A prior to group B and C domains (13). In addition, it has recently been shown that the breadth of antibodies that inhibit adhesion of IEs to ICAM-1 increases with age in Malian children (36), although the study did not investigate whether these IEs expressed group A, B, or C PfEMP1 antigens. We find that among group A PfEMP1 proteins, acquisition of IgG to DBL␤ domains that do not bind ICAM-1 appears to precede acquisition of IgG to those that do (19).
As the acquisition pattern of IgG specific for ICAM-1-binding DBL␤ domains from groups B and C is not currently known, we first used a BLASTP search to extract new potential ICAM-1-binding DBL␤ domains from these PfEMP1 groups. Seven such domains were identified and shown to bind ICAM-1 and to be structurally related to known ICAM-1-binding DBL␤ domains from group B and C PfEMP1 (Fig. 1).
In healthy children, we found that IgG reactivity to ICAM-1-binding DBL␤ domains IgG to ICAM-1-Binding DBL␤ Domains Infection and Immunity from group A was higher than reactivity to corresponding domains from group B and C PfEMP1 (Fig. 2). Overall, these findings are in agreement with the previously reported dominance of responses to group A PfEMP1 over other PfEMP1 groups (12,13,37). However, when the assays were repeated with plasma obtained from children with acute malaria, we found higher overall IgG reactivity to ICAM-1-binding DBL␤ domains from groups B and C rather than group A (Fig. 2). Furthermore, the most prominent change after acute malaria was a transient increase in IgG reactivity to ICAM-1-binding DBL␤ domains from groups B and C (Fig. 3), and 2 weeks after the acute attack, reactivity to each of the group B and C ICAM-1-binding DBL␤ domains was higher than that to the corresponding group A domains (Fig. 4). The most parsimonious explanation for the disease-related inversion of the ratio of IgG reactivity to ICAM-1-binding DBL␤ domains from group A versus groups B and C is that the disease episodes in our study children were caused mainly by parasites expressing group B and C PfEMP1 or group A PfEMP1 without ICAM-1-binding DBL␤ domains. This interpretation is consistent with the fact that only two of the study children were diagnosed with cerebral malaria (associated with parasites expressing group A DBL␤ domains [18]). This in turn may explain why we did not observe differences in IgG reactivities between children of similar age with severe and uncomplicated malaria (Fig. 5). In agreement with this, a recent study found that children with uncomplicated and cerebral malaria had similar breadth and magnitude of responses to different P. falciparum antigens, including DBL␤ domains (35). Finally, our data suggest that IgG specific for ICAM-1-binding DBL␤ domains from group A and associated specifically with cerebral malaria (18) is acquired later in life than IgG specific for ICAM-1-binding DBL␤ domains from group B and C PfEMP1 (Fig. 4C). We previously made a similar observation when IgG reactivity to ICAM-1-binding DBL␤ domains in group A was compared to reactivity to DBL␤ domains from the same group that do not bind ICAM-1 (19). Thus, the age where IgG specific for group A ICAM-1-binding DBL␤ domains is acquired coincides with the age where cerebral malaria incidence peaks (38,39). This is in striking contrast to the case for group A DBL␤ domains that do not bind ICAM-1 (19) and for group B and C DBL␤ domains that do (this study). However, whether a causal relationship exists remains to be investigated in a study area where the incidence of cerebral malaria is higher.
Opsonizing antibodies against PfEMP1 have been suggested to play a role in immunity to P. falciparum malaria (40), while other mechanisms, such as recruitment of complement (41), interaction with immune cells (42), and inhibition of vascular adhesion, might also play a role. Sequestration of P. falciparum IEs to the microvascular endothelium contributes to the pathogenesis of severe malaria in children, and broadly, cross-reactive antibodies inhibiting the interaction between ICAM-1 and DBL␤ domains are detectable in immune plasma (18,19,29). We were unable to test these potential effector functions due to the limited plasma volumes available, but these are aspects that should be investigated in future studies.
In conclusion, our study demonstrates significant differences in the acquisition of IgG to ICAM-1-binding DBL␤ domains from group A as well as group B and C PfEMP1. These differences are likely to be of significance in the development of PfEMP1-based vaccines to prevent severe P. falciparum malaria in general and cerebral malaria in particular (43).

MATERIALS AND METHODS
Study site and participants. The study was conducted from 2014 to 2015 at Hohoe Municipal Hospital in the Volta Region of Ghana. Plasma samples were collected from children aged 1 to 12 years (n ϭ 124) who reported with P. falciparum malaria (Table 1) (18,19,44). The inclusion criteria were a positive rapid diagnostic test for malaria, a positive blood smear of asexual P. falciparum parasites (Ͼ2,500/l), and fever or a history of fever (Ͼ37.5°C) in the preceding 24 h. Patients were categorized as having severe malaria if they had unarousable coma (Blantyre coma score of Յ2) without other known causes, severe malarial anemia (hemoglobin of Ͻ5 g/dl), hyperparasitemia (Ն250,000/l), or respiratory distress (i.e., rapid, deep, and labored breathing) (45). Among the severe malaria patients, 15 children were diagnosed with respiratory distress, eight with severe anemia, and two with cerebral malaria. Children with uncomplicated malaria were cases without any of the severe disease symptoms; all children in this group were treated as outpatients. Blood samples were collected on the day of hospital  Values are medians (25th; 75th percentile). Two study participants were diagnosed with cerebral malaria.
IgG to ICAM-1-Binding DBL␤ Domains Infection and Immunity admission and 2 and 6 weeks later. Patients receiving blood transfusion prior to follow-up were excluded from the study. Plasma collected from healthy children (n ϭ 91) ( Table 2) was also included in the study (46). A minority of those children had occasional, asymptomatic parasitemia. The study was approved by the Ethical Review Committee of the Ghana Health Services (file GHS-ERC 08/05/2014).
Production of recombinant proteins. His-tagged DBL␤ domains (see Table S1 in the supplemental material) were expressed in Escherichia coli Shuffle C3030 cells (New England Biolabs) from synthetic genes (https://eurofins.dk) or from DNA constructs generated by PCR from genomic DNA using specific primers (Table S2). The domains were purified by immobilized metal affinity chromatography (18,19,48). Recombinant ICAM-1-Fc (D1-D5 with a human Fc tag) was expressed in HEK293 cells and purified on a HiTrap protein G HP column (GE Healthcare) as described previously (48).
The plates were coated with 22 group A and nine group B and C recombinant DBL␤ proteins (2 g/ml; overnight at 4°C). Sixteen domains were already known to bind ICAM-1 (18,20), eight were already known not to bind ICAM-1 (19), and the remaining seven DBL␤ proteins (groups B and C) were predicted to bind ICAM-1 based on the sequence analysis done in this study.
Binding of ICAM-1-Fc and human IgG antibodies to the immobilized DBL␤ domains was detected using horseradish peroxidase-conjugated rabbit anti-human IgG (1:3,000; Dako) (18). Plates were developed using TMB PLUS2 (Kem-En-Tec) according to the manufacturer's instructions. Optical density (OD) values were read at 450 nm using a VersaMax microplate reader (Molecular Devices). Plasma antibody reactivity was expressed in arbitrary ELISA units (EU) calculated by the equation (OD sample Ϫ OD background )/ (OD positive control Ϫ OD background ) ϫ 100 (49) and translated into IgG level scores 0 (0 to 25 EU), 1 (26 to 50 EU), 2 (51 to 75 EU), 3 (76 to 100 EU), and 4 (Ͼ100 EU). A pool of P. falciparum-exposed Tanzanian individuals and 26 nonexposed Danish individuals were used as positive and negative controls, respectively. The mean (plus two standard deviation [SD]) value among the negative-control plasma samples was used as a cutoff value to define seropositivity.
Plasmodium falciparum parasite culture. P. falciparum HB3 was maintained in vitro and selected with rat anti-HB3VAR21-DBL␤_D4 as described previously (48,50). IE surface expression of PfEMP1 was regularly monitored by flow cytometry, and only cultures with more than 60% HB3VAR21 ϩ IEs were used. The identity and clonality of the parasites used were routinely verified by genotyping as described previously (51). Mycoplasma infection was excluded regularly using the MycoAlert mycoplasma detection kit (Lonza) according to the manufacturer's instructions.
Parasite adhesion spot assay. Falcon 1007 petri dishes were coated overnight (4°C) with recombinant CD36 (0.05 g/spot), ICAM-1-Fc (0.5 g/spot), or EPCR (0.1 g/spot) in triplicates (48). Dishes were blocked (1 h) in phosphate-buffered saline (PBS) with 3% bovine serum albumin (BSA). Mature IEs were adjusted to 3% parasitemia and 1% hematocrit in RPMI 1640 supplemented with 2% normal human serum, added to the dishes, and incubated (37°C, 30 min) as described previously (52). After removal of nonadherent IEs by sequential washing using prewarmed RPMI wash buffer, the remaining cells were fixed in 1.5% glutaraldehyde (15 min) and stained with Giemsa. After rinsing with water, the dishes were air dried overnight before the number of adherent IEs per square millimeter was quantified using ImageJ software (http://rsb.info.nih.gov/ij/). A minimum of three independent experiments in triplicate were done. All assays were blinded. PfEMP1 sequence similarity and phylogenetic trees. The Praline multiple-sequence alignment tool (http://www.ibi.vu.nl/programs/pralinewww/) was used to calculate the average amino acid similarity of DBL␤ domains. Multiple alignments of DBL␤ domains were made using Muscle (https://www .ebi.ac.uk/Tools/msa/muscle/) and analyzed using Mega 5.0 software (53) to create cladograms.
Statistics. We used Kruskal-Wallis one-way analysis of variance on ranks and Mann-Whitney test to assess intergroup differences. Data analysis was done using SigmaPlot 13.0 (Systat Software Inc., United Kingdom).

ACKNOWLEDGMENTS
We acknowledge the technical support of the field team in Ghana and thank all study participants for their generous contribution of blood samples. Mette U. Madsen is thanked for excellent technical assistance.
This work was supported by Novo Nordisk Fonden (NNF16OC0022298), Lundbeckfonden (R180-2014-3098), the Danish Council for Independent Research (4004-00032), the Consultative Committee for Development Research (12-081RH), and the Division of Microbiology and Immunology Diseases, NIAID, NIH (contract number HHSN266200400016C). The funders had no role in the study design, data collection and interpretation, or the decision to submit the work for publication.