PfEMP1 A-Type ICAM-1-Binding Domains Are Not Associated with Cerebral Malaria in Beninese Children

Cerebral malaria pathophysiology remains unknown despite extensive research. PfEMP1 proteins have been identified as the main Plasmodium antigen involved in cerebrovascular endothelium sequestration, but it is unclear which var gene domain is involved in Plasmodium cytoadhesion. EPCR binding is a major determinant of cerebral malaria whereas the ICAM-1-binding role is still questioned. Our study confirmed the EPCR-binding role in CM pathophysiology with a major overexpression of EPCR-binding domains in CM isolates. In contrast, ICAM-1-binding involvement appears less obvious with A-type ICAM-1-binding and dual receptor-binding domain expression in both CM and UM isolates. We did not find any variations in ICAM-1-binding motif sequences in CM compared to UM isolates. UM and CM patients infected with isolates expressing the ICAM-1-binding motif displayed similar IgG levels against DBLβ3 recombinant protein. Our study raises interrogations about the role of these domains in CM physiopathology and questions their use in vaccine strategies against cerebral malaria.

about the role of these domains in CM physiopathology and questions their use in vaccine strategies against cerebral malaria. KEYWORDS cerebral malaria, var genes, cytoadherence, dual receptor binding, ICAM-1-binding motif C erebral malaria (CM) is the most severe form of Plasmodium falciparum (P. falciparum) infection, mainly affecting children under 5 years in areas of endemicity (1). One of the most-described pathophysiologic mechanisms of CM is the ability of P. falciparum-infected erythrocytes (iE) to adhere to the cerebrovascular endothelium through variant surface antigen (VSA) proteins. These proteins are expressed on the iE cell membrane and bind to multiple endothelial receptors (2). In CM, this phenomenon causes brain swelling, coma, and even death (3). Currently, there is no specific marker for CM, but histopathology of malarial retinopathy (MR) (4,5) correlates with cerebral pathology.
The major VSA involved in iE sequestration is Plasmodium falciparum erythrocyte membrane protein 1 (PfEMP1). This parasitic protein is encoded by var genes (6), a hypervariable multigenic family classified in four groups, A, B, C, and E, and two intermediate groups, B/A and B/C (7). PfEMP1 proteins from A (8,9) and B/A (10) groups are associated with severe malaria (SM). PfEMP1 sequences are divided into two types of hypervariable extracellular domains: cysteine-rich interdomain regions (CIDR) and Duffy-binding-like (DBL). They have been classified into domain cassettes (DCs), composed of at least two conserved domains, DBL/CIDR (11).
Intercellular adhesion molecule 1 (ICAM-1) receptor has previously been proposed to be implicated in iE adhesion, and ICAM-1 overexpression in cerebrovascular tissue has been shown (21). Moreover, it has been demonstrated that EPCR and ICAM-1 are coinvolved in iE binding to brain cells (22). In this context, an ICAM-1-binding amino acid sequence, found in A-type DBL␤1/3 domains and mainly associated with CM, was recently identified (23)(24)(25). Interestingly, each DBL␤1/3 domain predicted to bind ICAM-1 is preceded by an EPCR-binding CIDR␣1 domain which defines a dual receptorbinding PfEMP1 sequence, associated with CM (23). Jensen et al. proposed a pathogenic cascade where the iE first binds to EPCR, leading to a proinflammatory cytokine release and increased expression of ICAM-1 (26).
Despite extensive research, few formal associations between the expression of specific domains of PfEMP1 and the adhesion properties of clinical isolates of P. falciparum have been demonstrated (18,19,27). In order to fill this gap, we conducted a comparative study on CM and uncomplicated malaria (UM) in Beninese children, in which we attempted to correlate the cytoadherence phenotype of P. falciparum isolates with clinical presentation and the expression of specific domains of PfEMP1. We also focused on the ICAM-1-binding motif and its implication in CM pathogenesis. and splenomegaly (P ϭ 0.0046). Sixty-seven percent of diagnosed CM cases presented retinopathy. CM cases had lower hemoglobin and blood glucose concentrations, lower platelet counts, higher creatinine and bilirubin concentrations, and higher leukocyte counts (P Ͻ 0.0001). Notably, 35% of children with CM died in our study in spite of adequate management including parenteral treatment with artesunate. There was no difference in multiplicity of infection (MOI) between CM and UM isolates (4 [3 to 9] versus 4 [3 to 7]). Values in parentheses and brackets are median [10-90th percentile].
Associations between var genes expression were also evaluated. Expected connections (such as CIDR␣1.7 and CIDR␣1.7-DBL␤1/3) were observed. Interestingly, a significant correlation was shown between DBL3 expression and double domains in CM isolates ( Fig. 2 and Table S3).  A-Type ICAM-1-Binding Domains in Cerebral Malaria ® Partial correlation between cytoadherence and var gene expression. Since expression of no more than one var gene was associated with Hbec-5i cytoadherence (Table S3), partial correlation only was studied for CHO-ICAM-1. Expression levels of CIDR␣1.4 and DBL␤1/3 in CM isolates and CIDR␣1.4-DBL␤1/3 and CIDR␣1.6 in UM isolates were associated independently with CHO-ICAM-1 cytoadherence. Partialcorrelation results are summarized in Table S4.
Recombinant protein sequences and binding property validations. PfEMP1 recombinant domains purity was evaluated by SDS-PAGE and Coomassie blue staining (Fig. S1a) and the domains' sizes were as expected (35 kDa for CIDR␣1.4, 65 kDa for DBL␤3, and 100 kDa for the CIDR␣1.4-DBL␤3 double domain). To ensure the amino acid sequences of our recombinant proteins, a mass spectrometry analysis was carried out (Text S1 and Fig. S1b). All identified peptides matched and covered 70% of the sequence. Furthermore, for the CIDR␣1.4-DBL␤3 domain, a single peptide overlapping both CIDR␣1.4 and DBL␤3 domains was detected (blue highlighting). We validated the interactions between the recombinant domains and the receptors by ligand receptor assay (29) and far-Western blotting (30) (see Text S1 and Fig. S2 (Fig. 3). Among isolates that displayed ICAM-1-binding DBL␤1/3 expression in RT-qPCR, no significant difference in IgG levels was found in CM compared to UM (Fig. 4). Moreover, no difference was observed for UM isolates that expressed or did not express ICAM-1-binding DBL␤1/3 in RT-qPCR. In the same way, UM children carrying a high CIDR␣1.4 level of expression did not have a higher IgG level against CIDR␣1.4 PF3D7_1150400 than did other children (data not shown).
IgG levels were lower for children with an MR compared to normal fundus (NF) children for both CIDR␣1.4 PF3D7_1150400 (38 [  21 days later (D21) for CM children, IgG levels were significantly higher in D21 plasma (Fig. 5).

DISCUSSION
We conducted a field study in Benin and included children presenting CM and UM. We evaluated the association between var gene expression and in vitro cytoadherence to Hbec-5i and CHO-ICAM-1 cell lines as well as the prevalence of ICAM-1-binding DBL␤1/3 domain expression.
CM isolates showed higher cytoadherence levels to both cell lines than did UM isolates. Hbec-5i is a well-described but complex cellular model of CM (12, 13) express-  ing different membrane receptors such as EPCR, ICAM-1, and PECAM, which could explain a more pronounced variation of cytoadherence than CHO-ICAM-1. These results might be the consequence of the ability of CM isolates to bind to multiple receptors. Higher cytoadherence on CHO-ICAM-1 may result from the nature of transfected cell lines, which express more cell membrane receptors than native ones (31). Our results are consistent with a previous study from Storm et al. (18). However, Hbec-5i and CHO-ICAM-1 are not physiological models of endothelial cells. In addition, although we have validated the cytoadherence assay protocol with a selected HB3 strain, we emphasize that the experiments were performed at room temperature, which could have influenced proteins interaction at the interface between iE and the cellular model.
To investigate the higher cytoadherence capacity of CM isolates, we quantified the expression levels of var gene domains. As expected, we did not observe increased CD36-binding domains (CIDR␣2.3/5/6/7/9/10) expression in CM isolates compared to UM isolates. Indeed, CD36-binding properties are a common feature shared by CM and UM isolates and have been rarely described as an explanatory factor in severe malaria pathophysiology (28). Then, we focused on var gene domains previously described as overexpressed in CM isolates analyzed by RT-qPCR (18) and transcriptome sequencing (RNA-seq) (32) techniques. CM isolates overexpressed EPCR-binding domains (CIDR␣1.1, CIDR␣1.4-8) and DBL␣ domains upstream (DBL␣2/1.1/2/4/7) in accordance with the literature, confirming the major role of EPCR in CM pathophysiology (15,17,18). Interestingly, DBL2, DBL3, and DBL␥1 domains were also overexpressed in CM isolates while no binding properties to endothelial receptors have been described for these domains. However, we assume that they could have an impact on the folding or maintenance of PfEMP1 three-dimensional structure as well as in IgM and ␣2macroglobulin binding (33,34). We might not focus only on domains with known endothelial receptor-binding properties to fully understand CM pathophysiology.
Surprisingly, we did not find any difference in ICAM-1-binding DBL␤1/3 domain expression in RT-qPCR between CM and UM, contrary to previous publications (18,23). Respectively, 50% and 41% of CM and UM isolates had ICAM-1-binding DBL␤1/3 domain expression in RT-qPCR. As a confirmation, we sequenced the ICAM-1-binding motif in isolates with ICAM-1-binding DBL␤1/3 domain expression and found it in both CM and UM isolates (see Table S5 in the supplemental material). We successfully sequenced the CIDR upstream ICAM-1-binding DBL␤1/3 domain in 11 isolates (9 CM and 2 UM), all predicted to bind EPCR. These results, combined with those of RT-qPCR, pointed out that the ICAM-1-binding DBL␤1/3 domain and dual receptor-binding domains are expressed in both CM and UM isolates, questioning their role in CM physiopathology. Further studies are needed to investigate it more precisely.
To investigate the role of immune response against var gene domains in cerebral malaria protection, we measured IgG levels against three recombinant proteins. We hypothesized that UM children infected with parasites expressing ICAM-1-binding DBL␤1/3 domain in RT-qPCR were immunized against that domain and against dual receptor-binding domains. Surprisingly, they did not have a higher IgG level than CM children infected with parasites expressing ICAM-1-binding DBL␤1/3 domain in RT-qPCR, meaning that those UM patients did not seem to be more protected against those domains than CM patients (Fig. 4). However, we tested a single recombinant protein (PF3D7_1150400, reference sequence) for each domain, and these results must be confirmed with several recombinant proteins with various amino acid sequences.
We carried out fundus examination for each cerebral malaria patient within 1 day of inclusion (35). Of note, 67% of cerebral malaria children had MR. No difference in var gene expression and cytoadherence level was evidenced between children with MR and NF (Table S2), as previously described (28). Interestingly, MR children had lower IgG levels against CIDR␣1.4 PF3D7_1150400 and CIDR␣1.4-DBL␤3 PF3D7_1150400 than did NF children (Fig. 5). All together, these results suggest the role of immune response against EPCR-binding domains, such as CIDR␣1.4, in MR physiopathology. Finally, we focused on the 10% of CM patients with the highest IgG level against each recombinant protein.
We did not find any difference in var gene expression compared to the remaining CM patients.
This work showed that CM is associated with parasites presenting a higher level of cytoadherence and EPCR-binding domain expression. Contrary to previous studies, the ICAM-1-binding DBL␤1/3 domain was not overexpressed in CM isolates in comparison to UM isolates. Those isolates expressed dual receptor-binding domains as well as CM. Besides, no difference was found in IgG against DBL␤3 PF3D7_1150400 between CM and UM isolates that express ICAM-1-binding DBL␤1/3 in RT-qPCR. To conclude, this study raises questions about the role of the A-type ICAM-1-binding domain in CM pathophysiology and finds a potential interest in CIDR␣1-DBL␤1/3 double domains without the ICAM-1-binding motif. Patients were enrolled at Cotonou in southern Benin from December 2017 to November 2018. We included children between 2 and 6 years displaying either CM or UM. Briefly, we defined UM by fever at inclusion or within 24 h before and positive thick or thin blood smear, without clinical or biological sign of severe malaria. We defined CM by a Blantyre score at diagnosis of Յ2 and a confirmed presence of P. falciparum infection with exclusion of other causes for coma, particularly meningitis (35). Peripheral venous blood samples have been collected from all study individuals in a Vacutainer tube containing EDTA. Giemsa-stained thick blood film confirmed P. falciparum infection, and parasitemia was quantified by counting against 1,000 leukocytes. We separated plasma from total blood by centrifugation and stored it at Ϫ20°C. Ring-stage parasites were conserved in TRIzol LS reagent (Life Technologies). Finally, parasites were cultured for 18 h to 24 h in vitro to obtain mature forms (Fig. 6) for cytoadherence assays.

Recruitment. Ethical clearance has been obtained from Comité
DNA extraction. To confirm P. falciparum infection by PCR, DNA extraction of isolates was required. Blood samples were extracted with the QIAamp DNA blood minikit (Qiagen) following the manufacturer's instructions. Briefly, 200 l of whole blood cells from each patient was incubated for 10 min at 56°C with 20 l of Qiagen protease and 200 l of lysis buffer. After incubation, 200 l of ethanol was added and the mix was transferred on silica columns. After several washes with Qiagen buffers, DNA was eluted with 100 l of elution buffer.
Confirmation of Plasmodium falciparum infection and assessment of the MOI. P. falciparum identification was confirmed by quantitative PCR (qPCR)-TaqMan (Fast-Track; Launch Diagnostic) following manufacturer's instructions. The MOI was estimated with a fragment analysis method using the P. falciparum msp2 polymorphic gene (36).
RNA collection. Plasmodium falciparum RNA preserved in TRIzol LS reagent (Life Technologies) was extracted using the phenol-chloroform method as previously described (37). Briefly, 0.2 ml chloroform was added to 1 ml blood-TRIzol sample, mixed, and then centrifuged at 4°C for 30 min at 12,000 ϫ g. Supernatants were precipitated with ice-cold isopropanol and centrifuged as previously described. Pellets containing RNA were recovered with 20 l RNase-free water. To clean our samples of any potential DNA contamination, RNA samples were digested and purified using the RNeasy minikit (Qiagen) following the manufacturer's guidelines. qPCR on the Rotor-Gene system (Qiagen) was carried out on seryl-tRNA synthetase gene (P90 primers [38]) to assess the absence of remaining parasite genomic DNA. RNA samples were reverse transcribed to cDNA using SuperScript II (Invitrogen; Thermo Fisher Scientific Inc.).
Cytoadherence of isolates on Hbec-5i and CHO-ICAM-1. Static binding assays were carried out on Hbec-5i (cerebral microvascular endothelium from Homo sapiens brain, ATCC CRL-3245, USA) and Chinese hamster ovary (CHO) cell lines transfected with ICAM-1 receptor (ATCC CRL-2093, USA). Blood was centrifuged at 800 ϫ g for 20 min for peripheral blood monocellular cell removal. Two hundred to four hundred microliters of ring-stage iE was matured in vitro during 24 to 36 h until they reached the mature stages in the first developmental cycle to avoid var switching (39). Briefly, isolates were cultured at 10% hematocrit in complete RPMI medium (RPMI 1640 with 0.2 mM hypoxanthine, 0.5% AlbuMAX II We selected the HB3 strain on Hbec-5i and CHO-ICAM-1 cell lines as previously described (13,40) and measured their cytoadherence value to validate this assay. We also measured the HB3-ICAM-1 cytoadherence value on CHO parental cell lines to ensure ICAM-1-binding specificity.
var gene expression quantification. Nineteen var gene domains were chosen based on previous publications (10,17,18,32), and 122 degenerated primers were designed based on var gene sequences assembled (41) from whole-genome sequencing (WGS) of field isolates previously collected in 2014 and 2016 in the same geographical region (1131-Pf-BJ-Bertin, CIVIC study) (32). We evaluated our primer sensitivity on African countries from the MalariaGEN Pf3K data set (Table S1) (42). RT-qPCRs were performed using the SensiFAST Sybr No-ROX kit (Bioline) on the Rotor-Gene qPCR system (Qiagen) following these PCR conditions: 95°C for 2 s; 40 cycles of 95°C for 2 s, 50 to 60°C for 10 s, and 72°C for 10 s; and a dissociation phase from 60°C to 90°C. var gene transcript levels were calculated using the relative standard curve method with several plasmid dilutions (GenScript). Quantified domains were normalized with the quantification of the seryl-tRNA synthetase gene expression (38). Human and P. falciparum genomic DNAs were used as negative and positive controls, respectively, at each qPCR run.
PfEMP1 domain recombinant protein production. Codon-optimized sequences coding for CIDR␣1.4, DBL␤3, and CIDR␣1.4-DBL␤3 domains from the Pf3D7_1150400 gene were synthesized for Escherichia coli (E. coli) expression (GenScript). Following the manufacturer's guidelines, these three domains were subcloned in pET100/d-TOPO vector (Invitrogen; Thermo Fisher Scientific) and used to transform TOP10 E. coli bacteria. Purified plasmid constructions (PureLink HQ mini-plasmid purification kit; Invitrogen) were confirmed in Sanger sequencing. These plasmids were used to transform Schuffle T7 competent E. coli (New England Biolabs). Protein expression was induced by addition of 1 mM IPTG when OD 600 reached 0.7 (in LB plus 100 g/ml ampicillin). Then, cultures were incubated for 4 h at 30°C. Recombinant proteins were purified as previously described (43). Briefly, bacteria were incubated for 2 h at room temperature in lysis buffer (250 mM NaCl, 20 mM Tris, 1 mM DTT, 1ϫ protease inhibitor cocktail [Roche] and 2% Sarkosyl). Samples were then sonicated and centrifuged at 3,000 ϫ g for 30 min to pellet cellular fragments. Supernatants were purified on nickel-agarose resin, which specifically binds the histidine tag. Finally, proteins were eluted three times with 2 ml of buffer containing 300 mM imidazole. The purity and the specificity of all our constructions were assessed by mass spectrometry with an Orbitrap Fusion Tribrid (3P5 Cochin Facility) and SDS-PAGE/Coomassie blue staining. We tested the interaction between recombinant proteins and endothelial receptors (ICAM-1 and EPCR) in a ligand receptor assay (29) and a far-Western blotting assay (30) (Text S1).
Mass spectrometry analysis. Digestion was carried out using an S-trap micro spin column. Briefly, samples were heated, reduced, and alkylated at the same time for 5 min at 95°C in a buffer containing 100 mM tetraethylammonium bicarbonate, 2% SDS, 10 mM Tris(2-carboxyethyl)phosphine hydrochloride (TCEP), and 55 mM chloroacetamide. Then, samples were loaded on S-trap columns and incubated with sequencing-grade-modified trypsin (Promega) overnight at 37°C, and after digestion, peptides were eluted. Mass spectrometry analyses were performed on a Dionex U3000 RSLC nano-LC system coupled to an Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific). After drying, peptides were solubilized in 10 l, and 1 L was loaded, concentrated, and washed for 3 min on a C 18 reverse-phase precolumn. Peptides were separated on a C 18 reverse-phase resin with a 1-h gradient ending in 90% of solvent B containing 80% acetonitrile (ACN), 0.085% formic acid (FA) in H 2 O. The mass spectrometry data were analyzed using Mascot v2.5 (Matrix Science) jointly on a homemade data bank and on bacteria (333,999 sequences) from the Swiss-Prot data bank containing 560,537 sequences (July 2019). The enzyme specificity was trypsin, and up to 1 missed cleavage was tolerated. Carbamidomethylation of cysteines was set as variable modifications, and oxidation of methionines was set as fixed modifications.
ELISA. Enzyme-linked immunosorbent assays (ELISAs) were performed to measure the total IgG responses against the different PfEMP1 recombinant domains. Briefly, 1 g/ml of CIDR␣1.4 PF3D7_1150400 , DBL␤3 PF3D7_1150400 , or CIDR␣1.4-DBL␤3 PF3D7_1150400 recombinant proteins was applied as a coating to MaxiSorp plates (Nunc, Thermo Fisher Scientific) overnight at 4°C. After each step, wells were washed 3 times with PBS-0.1% Tween using an ELISA plate washer. Plates were blocked for 2 h at room temperature with PBS-4% BSA. Then, PBS-diluted plasma samples (1:100) were incubated for 1 h at room temperature and bound IgG was detected using a horseradish peroxidase (HRP)-conjugated mouse anti-human IgG antibody (1:5,000) [IgG(HϩL) F(ab=) 2 -goat anti-human-HRP; Invitrogen]. The HRP substrate 3,3=,5,5=-tetramethylbenzidine (TMB Plus2; Eco-Tek) was added for exactly 3 min in darkness, and the reaction was stopped by adding 0.2 M H 2 SO 4 . Optical densities were measured at 450 nm using the Tecan system. ELISA arbitrary units (AU) were calculated using the following equation: AU ϭ (OD of sample Ϫ OD of negative control)/(OD of positive control Ϫ OD of negative control) (44). The positive control was a pool of pregnant Beninese women, and the negative control was a pool of plasma from pregnant French women.
Statistical analysis. We compared UM and CM in terms of clinical and biological characteristics, cytoadherence, var gene expression, and immune response levels. Nonparametric variables were presented by median (10th to 90th percentile).
Mann-Whitney U-test was used to compare medians of clinical and biological characteristics, cytoadherence, var gene expression, and IgG levels and Chi-2 test was used to compare proportions of selected clinical characteristics. Correlations between cytoadherence and var gene expression level and between var gene expression levels were assessed calculating Spearman's rank correlation coefficient rho, and P values were adjusted (p-adj) for multiple testing using the Benjamini-Hochberg correction (45). For a given cytoadherence, when several var gene expressions were associated (correlations with a p-adj of Ͻ0.35), partial correlations were calculated to control for complementary var gene expression (46).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.02 MB.