Pathogenicity Determinants of the Human Malaria Parasite Plasmodium falciparum Have Ancient Origins

Cytoadhesion of P. falciparum-infected erythrocytes in the microcirculation is a major virulence determinant. P. falciparum is descended from a subgenus of parasites that also infect chimpanzees and gorillas and exhibits strict host species specificity. Despite their high genetic similarity to P. falciparum, it is unknown whether ape parasites encode adhesion properties similar to those of P. falciparum or are as virulent in their natural hosts. Consequently, it has been unclear when virulent adhesion traits arose in P. falciparum and how long they have been present in the parasite population. It is also unknown whether cytoadhesive interactions pose a barrier to cross-species transmission. We show that parasite domains from the chimpanzee malaria parasite P. reichenowi bind human receptors with specificity similar to that of P. falciparum. Our findings suggest that parasite adhesion traits associated with both mild and severe malaria have much earlier origins than previously appreciated and have important implications for virulence evolution in a major human pathogen.

Cytoadhesion of falciparum malaria involves significant remodeling of the erythrocyte membrane cytoskeleton to form distinctive, knob-like protrusions (12)(13)(14)(15). These modifications result in reductions in the deformability of IEs (16) and render them vulnerable to splenic clearance (17). Parasite binding to vascular endothelium is mediated by a clonally variant gene family, termed the var gene or P. falciparum erythrocyte membrane protein 1 (PfEMP1) family (18)(19)(20). PfEMP1 proteins contain multiple Duffy binding-like (DBL) and cysteine-rich interdomain region (CIDR) domains that interact with endothelial receptors (21). Surface exposure places PfEMP1 proteins under strong selection for immune evasion and binding properties, resulting in high intra-and interstrain sequence variability (22). Nevertheless, the var gene family is organized similarly between parasite genotypes into three major types (A, B, and C) and a placenta-specific E variant, as defined by the chromosomal location and 5= upstream sequence region (23)(24)(25). Furthermore, var groups have diverged into CD36 binding (groups B and C) and endothelial protein C receptor (EPCR) binding (group A) subsets, with both traits mapping to the CIDR domain in the PfEMP1 head structure (26)(27)(28)(29). Infections dominated by CD36 binding parasites are associated with mild malaria, while parasites transcribing var genes that are predicted to encode EPCR binding properties are preferentially expressed in malaria-naive hosts and in subjects with severe malaria (29)(30)(31)(32)(33)(34)(35)(36). EPCR plays an important role in regulating coagulation, vascular inflammation, and endothelial permeability (37), and it is thought that parasite blockade of EPCR function may contribute to malaria disease mechanisms (30,(38)(39)(40). Both the conservation of var gene genomic organization and protein functional specialization suggest that adhesion selection has strongly shaped the PfEMP1 repertoire, although some adhesion traits appear to be more dangerous than others.
Despite the importance of IE sequestration in virulence, the vast majority of P. falciparum infections do not lead to severe disease, suggesting that cytoadhesion is relatively well adapted. An approach to investigating the evolutionary history of pathogenicity determinants is to study other Laverania parasites. Although limited genetic data exist for most Laverania parasite species, the chimpanzee malaria parasite P. reichenowi has been fully sequenced and presents extensive gene synteny with P. falciparum, including multigene families involved in erythrocyte remodeling and a repertoire of var genes with similar gene copy numbers and multidomain architectures (41)(42)(43)(44). By comparison, P. gaboni, a Laverania parasite more distantly related to P. falciparum, contains divergent var-like genes (41). Notably, var genes are absent in non-Laverania malaria species (45,46), indicating that the var-mediated cytoadhesion phenotype originated within the Laverania subgenus.
The var gene/PfEMP1 family has a capacity for rapid evolution through high rates of recombination and mutation (47). These features endow the members of the protein family with a far greater versatility than ordinary malaria proteins with respect to escaping immunity and potentially acquiring new adhesion traits. Notably, minor sequence variation in the P. reichenowi and P. falciparum reticulocyte binding protein homologue 5 (RH5) invasion ligand has a major role in determining host tropism for red blood cells (48). However, it is unknown whether cytoadhesion interactions impose a similar host restriction barrier for cross-species transmission of ape Laverania parasites to humans. It is also not known when virulent adhesion traits arose in P. falciparum. Here, we performed the first functional characterization of domains from P. reichenowi erythrocyte membrane protein 1 (PrEMP1). We provide evidence that CD36 and EPCR head structure binding properties have ancient origins that predate P. reichenowi and P. falciparum speciation into chimpanzee and human hosts, thereby revealing deep evolutionary roots of parasite adhesion traits that have been linked to both mild and severe infection outcomes.
severe malaria (29,30), additional subtyping of P. reichenowi CIDR␣1 domains was performed. Of the eight CIDR␣1 subtypes in P. falciparum (25) (26). In the CIDR␣1 subtype tree (Fig. 1C), all of the P. reichenowi CIDR␣1 sequences clustered with CIDR␣1.4 sequences from P. falciparum. Viewing the data as a whole, this analysis supports the conclusion that the divergence of CIDR sequence types occurred prior to speciation of P. reichenowi and P. falciparum and that the same major sequence types have been maintained during parasite adaptation to their chimpanzee and human hosts.
Similarly, in cell binding assays, the P. reichenowi CIDR domains bound with predicted specificity to CHO-EPCR cells ( For the CD36 interaction, both the P. falciparum and P. reichenowi domains were inhibited at levels greater than 90% by the anti-CD36 MAb FA6-152 and there was limited or no inhibition by an isotype control antibody ( Fig. 3D and Fig. S2). A recent analysis of a CD36:CIDR␣2.8 co-crystal structure identified 14 residues in CD36 that interact with Malayan Camp var1 CIDR␣2.8 (50). Human and chimpanzee CD36 sequences are identical at 13 of 14 contact residues (conservative substitution I157V), and human and gorilla CD36 sequences are identical at 8 of 14 contact residues (conservative substitutions at M156V and I157V) (Fig. S3). Taken together, the results of this analysis indicate that CIDR domains from P. reichenowi and P. falciparum interact with similar regions on CD36 and EPCR and further suggest that these binding properties originated in a common ancestor of P. reichenowi and P. falciparum.
CIDR domains from P. reichenowi interfere with the endothelial barrier protective response of the EPCR pathway. EPCR is a receptor for protein C/activated protein C (APC) and plays a critical role in coagulation, inflammation, and endothelial barrier properties (37). It has been postulated that EPCR binding P. falciparum parasites contribute to cerebral malaria brain swelling (51) by inhibiting the APC-EPCR interaction (29,(38)(39)(40). To explore the origins of this presumptive virulence phenotype in the Laverania subgenus, we investigated whether P. reichenowi CIDR domains inhibit the APC-EPCR interaction to the same extent as P. falciparum domains.
To study whether P. reichenowi CIDR domains interfere with APC binding, competition assays were performed with CHO-EPCR cells. For these assays, CIDR domains were used at 50 g/ml or 250 g/ml to achieve approximately 70% and 100% binding levels on CHO745-EPCR cells at the lower and higher concentrations (Fig. S4). As expected, the CD36 binding, P. reichenowi var85 CIDR␣5 domain did not inhibit APC binding ( Fig. 4A and B), whereas the positive-control P. falciparum var07 CIDR␣1.4 domain almost completely abolished APC binding at 50 g/ml (85% reduction) or 250 g/ml (98% reduction). Notably, the three CIDR␣1.4 domains from P. reichenowi inhibited APC binding by 62% to 85% at the higher concentration ( Fig. 4A and B).
To study if P. reichenowi CIDR domains interfere with APC barrier protective properties, thrombin-induced barrier dysfunction assays were performed with primary human brain microvascular endothelial cells. Thrombin induced a rapid drop in electrical impedance across the brain endothelial monolayer that peaked at approximately 30 min and returned to baseline by 2 h (Fig. 4C). APC diminished thrombin-induced barrier disruption by 50%. As expected, pretreatment with the negative-control CD36 binding P. reichenowi var85 CIDR␣5 domain led to a minimal reduction in APC barrier protection (25% reduction at the higher concentration) (Fig. 4D). Conversely, the positive-control P. falciparum var07 CIDR␣1.4 domain and the three CIDR␣1.4 domains on January 7, 2021 by guest http://msphere.asm.org/ from P. reichenowi caused a 60% to 70% reduction in APC protective function at the Consistent with the functional analysis, comparison of CIDR␣1.4 sequences of P. falciparum and P. reichenowi shows that they are relatively conserved at nine contact residues from the solved IT4var07 CIDR␣1.4-EPCR co-crystal structures (26) (Fig. 5A). This includes the highly conserved dual phenylalanine residues at positions F655 and F656 in IT4var07 CIDR␣1.4; the F656 inserts into the EPCR lipid binding groove in a location similar to that of a phenylalanine residue from APC (26) (Fig. 5A). Moreover, chimpanzees and gorillas differ at only 6 amino acid positions from human EPCR (Fig. S5) and all of them are distant from the P. falciparum var07 CIDR␣1.4-EPCR contact interface (Fig. 5B). Taking the results together, this analysis suggests that P. reichenowi and P. falciparum CIDR␣1.4 domains engage similar surfaces on EPCR and that both compete for binding with its native ligand protein C/APC.

DISCUSSION
The uniquely virulent character of the P. falciparum parasite, along with its capacity to cytoadhere in the host's microvasculature, has led to debate regarding the origins of deadly cytoadhesion traits. Several lines of evidence support the hypothesis that widespread human infection with P. falciparum is a relatively recent phenomenon and may have emerged in only the past 5,000 to 10,000 years, associated with changes in human populations from hunter-gatherer societies to agriculture-based communities (52). P. falciparum's closest relative is a gorilla parasite, P. praefalciparum, and its next closest relative is a chimpanzee parasite, P. reichenowi (8). Despite their high genetic similarity to P. falciparum, it is unknown whether ape parasites are as virulent in their natural hosts (53). While only limited DNA sequence data exist for P. praefalciparum, the genome of P. reichenowi has been fully sequenced (42). Here, we investigated the evolutionary origins of P. falciparum pathogenicity determinants by studying the binding properties of P. reichenowi domains.
For most of the 20th century, the nature of the evolutionary relationship between P. falciparum and P. reichenowi was largely limited to inferences based upon the morphological similarities between the two parasites (54). These have been extended by genomic studies that show similar catalogs of var genes of P. falciparum and P. reichenowi, as well as of multigene families involved in erythrocyte remodeling (41)(42)(43)(44). Although the members of the var gene family are among the fastest-evolving genes in P. falciparum, there appears to be significant conservation of var organization and coding features between the two species (25,42). In contrast, P. gaboni, a more distant Laverania relative, contains var-like genes that have DBL domains that are highly Sequence logos comparing variations between P. falciparum (top logo) and P. reichenowi (bottom logo) with respect to CIDR␣1 residue positions corresponding to those that have direct contact with EPCR in two solved co-crystal structures from P. falciparum (26). The residues of the three P. reichenowi CIDR␣1.4 sequences tested in this study are shown. (B) Structure of the IT4var07 CIDR␣1.4-EPCR interaction (26) showing the locations of EPCR residues that differ in gorilla or chimpanzee sequences from human sequences (G124 and I194, gorilla only ϭ blue; S138, chimpanzee only ϭ yellow; A127, gorilla plus chimpanzee ϭ red).  (41,42). Taken together, the genomic data suggest an evolutionary history in which DBL domains containing var-like genes originated in an early precursor of the Laverania subgenus. However, the ancestral var gene diverged between Laverania species and the characteristic DBL-CIDR head structure of PfEMP1, a key determinant of P. falciparum cytoadhesion, is present in only a subset of the members of the Laverania clade. While the origin of the var gene family is becoming clearer from comparative genomics, nothing was known about the adhesion traits encoded by var genes present in other members of the clade. Extensive work on P. falciparum has revealed that the binding properties of CD36 and EPCR are predicted by sequence classification of CIDR domains (26)(27)(28)(29). Nevertheless, while phylogenetic classification of PfEMP1 domains is highly predictive of binding (55), exceptions are known. For instance, sequence variation between CIDR␣1 domains can determine the ability to bind EPCR (26) or influence the extent of APC blockade activity (30,38,40). Here we demonstrated that CIDR domains in P. reichenowi can be categorized into sequence types similar to those of P. falciparum and that domains bind in a predictable manner to human CD36 and EPCR. More significantly, the CIDR␣1 domains from P. reichenowi share the capacity of CIDR␣1.4 domains from P. falciparum to disrupt APC-EPCR binding in a manner that exacerbates the permeability of human brain endothelial cell monolayers in the presence of thrombin (30,(38)(39)(40), a feature that may contribute to brain swelling and cerebral malaria pathophysiology (51,56,57).
A limitation of this study was that recombinant proteins were analyzed. This was necessary because chimpanzees are an endangered species and because it is technically challenging to work with P. reichenowi parasites. However, previous work has shown that recombinant CIDR domains predict P. falciparum parasite binding to both CD36 (58) and EPCR (29,40,59). These results show that binding to CD36 and EPCR is conserved between P. falciparum and the chimpanzee malaria parasite P. reichenowi.
The remarkable persistence and specificity of the CD36 and EPCR binding traits in a rapidly evolving variant antigen gene family have several implications. First, they provide evidence that CD36 and EPCR binding have early origins and have been maintained in the populations of both P. reichenowi and P. falciparum parasites, despite the association of EPCR binding with severe malaria in human infections (29). Consistent with this concept, CD36 and EPCR sequences are highly conserved between humans, chimpanzees, and gorillas. The stability of adhesion receptor specificity suggests a strategy in which the parasite co-opts a key functional molecular interface which the host cannot easily modify without compromising protein function. Indeed, P. falciparum CIDR␣1 domains interact with the APC binding site in EPCR (26,29,30,39,40) and CIDR␣2-6 domains bind to the oxidized low-density lipoprotein (OxLDL) binding site in CD36 (50). The corresponding EPCR contact residues are completely conserved between human, chimpanzee, and gorilla sequences, and the corresponding CD36 contact residues are identical at 13 of 14 positions in humans and chimpanzees. While gorilla sequences differ at six of the human CD36 contact residues, two of the differences are conservative amino acid substitutions.
An additional implication is that cytoadhesion interactions may have posed less of a barrier than red blood cell invasion (48) for the crossing of the progenitor parasite from gorillas to humans. The ability of the ancestral parasite to cytoadhere to CD36 and EPCR may have been an important factor in the P. praefalciparum zoonotic event of transfer to humans, since poor sequestration would otherwise be expected to impose a high parasite fitness cost due to splenic entrapment and clearance of more-rigid IEs (16,17). Studies of fecal samples have revealed six different Laverania species in African chimpanzee and gorilla populations (7)(8)(9)(10)(11). It is possible that cytoadhesion interactions may pose a stronger barrier to human transmission for more-distant Laverania parasites that lack CIDR domains, but this remains to be determined. Our findings suggest that the adhesion traits encoded in var genes may have at least partly preequipped the ancestral P. falciparum parasite with traits necessary to survive in the human host. Furthermore, the parasite that crossed into humans was already endowed with the dangerous EPCR binding trait. The long evolutionary persistence of the CIDR-EPCR interaction raises the possibility that it may also possess adaptive properties that have led to its retention in human and chimpanzee parasites, especially since the vast majority of P. falciparum infections are not deadly. While CD36 binders are predicted to be more common than EPCR binders, both parasite species have invested considerable genomic resources in retaining both adhesion traits. In addition to OxLDL and APC, a diverse array of ligands bind to CD36 and EPCR (60,61). This suggests additional possible roles of CD36-and EPCR-based parasite adhesion in influencing a range of physiological and pathological processes which will be of interest to explore.

MATERIALS AND METHODS
Sequence analysis. To identify CIDR domains in P. reichenowi, we conducted searches of public databases (http://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGEϭProteins) using amino acid sequences from representatives of the four CIDR sequence types (␣, ␤, ␥, and ␦) found in the P. falciparum genome. This approach identified 93 unique var sequences, which were submitted to the VarDom server (http:// www.cbs.dtu.dk/services/VarDom/) for the identification of DBL and CIDR domain boundaries. All CIDR domain sequences were extracted, combined with a data set comprising all of the CIDR sequences present in the 3D7 genome, and assembled into a neighbor joining tree using the Geneious Tree Builder tool. For subtyping of P. reichenowi CIDR␣1 domains, additional representative CIDR␣1-8 type domains, as defined by Rask et al. (25), were included in the tree.
Protein expression. Recombinant P. reichenowi CIDR domains were synthesized as Gblocks gene fragments (Integrated DNA Technologies, Inc.), and P. falciparum CIDR sequences were amplified from parasite genomic DNA (gDNA) (see Table S2 in the supplemental material). Proteins were expressed as His6-maltose binding protein (MBP)-tobacco etch virus (TEV)-CIDR-StrepII constructs in pSHuffle expression hosts (New England Biolabs, Inc.) and purified using a two-step process, as described previously (62). Purified proteins were analyzed by SDS/PAGE according to standard procedures. BLI analysis. CIDR binding kinetics data were determined using an Octet Qke instrument. The protocol for analyzing the CIDR-EPCR interaction was conducted as reported previously (40). The CIDR-CD36 interaction was evaluated by immobilizing hisCD36 to nickel-nitrilotriacetic acid (Ni-NTA) biosensors (ForteBio). For the association phase, binding was measured by immersion of the sensors into wells containing CIDR domains diluted in kinetics buffer (phosphate-buffered saline [PBS], 0.02% Tween-20, 100 g/ml bovine serum albumin, 0.005% sodium azide) for 600 or 900 s. For the dissociation phase, sensors were then immersed in kinetics buffer for 300 to 600 s. Mean association rate constant (K on ), dissociation rate constant (K off ), and equilibrium dissociation constant (K d ) values were calculated using subtracted double-reference data fitted to a 1:1 binding mode using the data analysis software furnished with the Octet instrument (ForteBio).
Permeability assays. Endothelial barrier permeability assays were measured in real time using an xCELLigence system from ACEA Biosciences as described previously (30). In brief, primary human brain microvascular endothelial cells (ACBRI376; Cell Systems) were grown for several days to reach confluence in 96-well plates on integrated electronic sensors. For thrombin-induced barrier dysfunction assays, cell monolayers were treated with recombinant CIDR domains (50 or 250 g/ml) for 30 min, followed by 100 nM APC (Haematologic Technologies, Inc.) for 2 h, followed by 5 nM thrombin. Control wells received thrombin plus thrombin inhibitor hirudin (Hyphen Biomed, France) (500 nM), recombinant CIDR domains alone, APC alone, thrombin alone, or APC plus thrombin treatment. Measurements of transendothelial resistance were initiated before the first treatment and continued until impedance measurements returned to baseline (~2 h after thrombin treatment). CIDR blockade of APC function was measured at the peak of thrombin barrier disruption. Blockade activity was calculated by determining the percentage of APC protection in the presence or absence of CIDR domains. The funders had no role in the study design, data collection and interpretation, or the decision to submit the work for publication.