Background
Sequestration of mature
Plasmodium falciparum intra-erythrocytic stages in the microvasculature is a major contributor to falciparum pathogenesis [
1,
2]. The best-characterized parasite factor implicated in cytoadherence is the family of
P.
falciparum erythrocyte membrane protein 1 (PfEMP1) variant adhesins encoded by the approximately 60-member
var gene family [
3]. PfEMP1 molecules comprise a large surface-exposed N-terminal region containing a suite of modules called Duffy-Binding Like (DBL) domains and Cysteine-rich Inter-Domain Regions (CIDR), a single transmembrane segment and a cytoplasmic C-terminal domain. DBL and CIDR domains are highly variable within the PfEMP1 family but they can be assigned to a limited number of classes according to distinct sequence signatures [
4‐
6]. The variability in sequence and domain organization in PfEMP1 variants [
6] provides the parasite with the capacity to bind to an array of host receptors and to evade host immunity [
3].
The capacity of infected red blood cells (iRBCs) to cyto-adhere to uninfected RBC, i.e., rosetting, has been associated with severe malaria in African children, with higher frequency of rosette-forming parasites and larger rosettes than in uncomplicated malaria [
7‐
11]. Rosetting is also associated with an elevated infecting parasite biomass [
10] and an increased multiplication rate in a non-human primate model [
12]. Rosetting involves specific interactions between a subset of PfEMP1 adhesins [
5,
6,
13‐
15], serum factors [
15‐
22] and a variety of RBC receptors [
20,
23‐
26]. Using vaccination or soluble inhibitors to target rosetting is thus an attractive strategy against severe malaria pathology.
To better understand critical molecular interactions and immunologic determinants implicated in rosetting, experimental models are needed. The Palo Alto VarO, a clonal rosetting line infectious for the
Saimiri sciureus monkey [
12], has been developed as a monovariant culture, in which a large majority (90–95 %) of the iRBCs express the Palo Alto varO gene [
13]. The PfEMP1-VarO extracellular region has five DBL domains (DBL
1–5) and one CIDR domain. All six domains, as well as the double DBL1-CIDR Head domain, have been produced as recombinant proteins [
13,
20,
27,
28]. RBC binding has been mapped to DBL1α and the ABO blood group determinants have been identified as the erythrocyte receptor [
20]. This model was used to explore the immune response of humans living in endemic areas showing elevated seroprevalence in Senegalese [
13] and Beninese settings [
29]. Two important features emerged from these studies, namely that the surface-reacting antibodies acquired by humans exposed to malaria were variant-specific [
30] and that there were no rosette-disrupting antibodies in children [
29].
Previous work has shown that DBL1, CIDR1, DBL2, DBL4 and the Head PfEMP1-VarO domains elicited antibodies reacting with the Palo Alto VarO iRBC surface. The work reported here aims to gain insight into the surface epitopes of PfEMP1-VarO using monoclonal antibodies (mAbs) isolated from mice immunized with these recombinant domains. The mAbs were characterized with respect to reactivity with the iRBC membrane-anchored PfEMP1-VarO by surface immunofluorescence and immunoblots of SDS-extracts of Palo Alto VarO iRBCs. Their functionality was assessed using rosette disruption and inhibition of rosette formation assays. The reactivity of surface-reacting mAbs specific for DBL1 was analysed using a panel of mutant domains, highlighting the existence of two distinct binding sites of potent rosette-disrupting mAbs. These results provide novel clues for the design of anti-rosetting strategies.
Discussion
This work investigated specificity and functionality of mAbs obtained from animals immunized with recombinant domains from PfEMP1-VarO. MAbs raised to four different PfEMP1-VarO domains were obtained; they reacted with the surface exposed PfEMP1-VarO protein and appear representative of the corresponding polyclonal sera with regard to potency of surface reactivity and inhibition of adhesion.
Potent mAbs were obtained against DBL1 that efficiently inhibited formation of rosettes down to quite low concentrations and efficiently reversed rosetting. This confirms previous data showing that polyclonal sera raised to the correctly folded, functional adhesion domain efficiently block and disrupt Palo Alto VarO rosetting [
20,
33]. MAbs BD20E4, D15-50 and E20-76 were more potent inhibitors, and mAbs BD20E4 and D15-50 were more potent disrupters, than the polyclonal rabbit IgG raised against bDBL1 (Fig.
4). This is remarkable as VarO rosettes are large (called giant rosettes by others [
38,
39]) and involve strong interactions that cannot be mechanically disrupted, in contrast to other rosetting parasites such as IT4R29 [
23,
40]. As such, these mAbs represent precious tools to dissect key molecular interactions involved in rosetting. It is worth noting that design of the recombinant domain is important to the induction of potently blocking antibodies. Indeed, antibodies elicited against a PfEMP1-VarO DBL1 domain lacking a stretch of 89 C-terminal residues compared to DBL1 expressed here (and missing 5 out of 20 Cys residues, thereby preventing formation of the Cys324–Cys442, Cys354–Cys470, Cys377–Cys467 and Cys477–Cys483 disulfide bonds) only partially inhibited Palo Alto VarO rosetting [
41].
Recently, rosetting was shown to occur in the absence of PfEMP1. In parasites in which PfEMP1 trafficking to the red cell membrane was impaired by disruption of the
Pfmahrp1 gene, variant STEVOR proteins were shown to mediate rosetting through binding to glycophorin C [
42]. A-RIFINs were identified as a mediator of rosetting involving binding to blood group A antigens, when PfEMP1 was removed by trypsin treatment or blocked by adding PfEMP1-specific antibodies [
41]. This led some authors to conclude that A-RIFINs “are conceivably the main ligand of this host-parasite interaction” and that it is possible that “PfEMP1 might work as an accessory or alternative interaction to reinforce by multiple contact points these protein-carbohydrate interactions that are usually weak” [
43]. The potent rosette blocking and disruption capacity shown here for single mAbs reacting solely with PfEMP1 in Palo Alto 89F5 VarO extracts (Fig.
3) does not support such conclusions, but rather indicates that when PfEMP1 is expressed and surface-exposed, i.e., the normal condition of
P. falciparum parasites, it is a major determinant of rosetting as PfEMP1-specific mAbs readily inhibit and disrupt rosettes.
Analysis of reactivity to a panel of constructs showed that some DBL1 mutations reduced but did not abolish binding of the mAbs, although they did completely abolish RBC binding of the recombinant domain [
20]. MAbs interfering with rosetting bound to two distinct regions on the surface of the DBL1 domain (Fig.
6) in proximity to the blood group trisaccharide binding site, which has been localized by computer docking and site-directed mutagenesis to a restricted surface area situated at the interface of subdomain 1 and subdomain 2 in the vicinity of the NTS-DBL1α
1 hinge region [
20]. This is consistent with inhibition of rosetting by impairing access to the RBC-binding site. The close proximity or partial overlap of the BD20E4 epitope with the RBC-binding site is reminiscent of findings observed for the binding site of the 24E9 Fab mAb, which overlaps with the ICAM-1 binding site on the surface of the DBLβ3-D4 domain of PfEMP1-PFD1235w [
44]. None of the anti-DBL1 VarO mAbs appear to bind to the region of subdomain 3 identified as the binding site of mAbs that potently disrupted FCR3S1.2 (alias IT4Var60) or R29 (alias IT4Var9) rosettes [
45] located in a different region of the DBL1 surface (see Additional file
4: Figure S3). Interestingly, the RBC-binding site of the DBL1α-IT4var60 domain has recently been localized adjacent to the RBC-binding site of DBL1α-VarO [
46]. Moreover, mapping of human antibody reactivity with short linear-peptides of DBL1α-FCR3S1.2 indicated the presence of epitopes recognized by anti-rosetting antibodies within subdomains 1 and 2 [
47]. Although these data are consistent with the findings reported here, it is possible that other parts of the DBL1 domain contribute to optimal display of the binding site and as such, can be targeted by rosette disrupting antibodies. It is worth noting that both titres and OD values observed with BD20E4 decreased when the constructs lacked the C-terminal residues forming the hinge with CIDR, which are located far from the RBC-binding area [
20]. The conservative conclusion from the work presented here is that mAbs displaying a potent rosette-inhibition and rosette-disruption activity bind with non-overlapping sites located close to the RBC-binding site area, but this does not exclude contribution of additional regions of the molecule or existence of additional important epitopes.
Most mAbs studied here targeted reduction-sensitive epitopes, including mAbs M21-17 and M21-30, which had an unusual profile as they recognized reduction-sensitive epitope(s) on both PfEMP1-VarO and the recombinant DBL1 domain but did not stain the VarO iRBC surface. The possibility that the binding sites are surface-exposed but masked by some serum component is ruled out by the lack of reactivity with the PfEMP1-VarO protein on immunoblots of Palo Alto 89F5 VarO SDS-extracts (Fig.
3). This indicates that the epitopes in question depend on the proper formation of disulfide bonds and the correct folding of the protein but are lost upon iRBC surface display.
Heparin and sulfated glycosaminoglycans are potent inhibitors of Palo Alto VarO rosetting [
27] as well as of multiple rosetting types [
24,
25,
46,
48‐
51]. This suggests the presence of some common motif/epitope shared by rosette-forming parasites, which constitutes an attractive intervention target. None of the mAbs isolated here proved sensitive to the presence of mutations that drastically reduced heparin binding [
28]. Interestingly, two potent FCR3S1.2 rosette-disrupting mAbs also failed to map to the heparin-binding area of the expressed PfEMP1-IT4var60 molecule [
46]. Although these properties prevent dissecting the molecular basis of VarO-heparin binding using the available inhibitory mAbs, it confirms that the heparin-binding site and the RBC-binding sites are located far from each other on the DBL1-VarO surface.
There was some relationship between ELISA titres and surface IFA titres and MFImax. The mAb with the lowest titre (G8-49) had the lowest surface reactivity, and interestingly, both ELISA titres and surface reactivity were lower than mAb N6-37, also a CIDR binder. In contrast, the mAbs with the highest ELISA titres (the anti-DBL2 mAbs B12-42 and B12-15 and the anti-DBL4 mAb D18-94) did not have the highest IFA titres and MFImax. This may reflect reduced accessibility of the individual domains on the iRBC surface compared to the N-terminal domains. Previous findings showed limited quantitative relationship between ELISA reactivity and surface staining with polyclonal sera [
30,
33]. Interestingly however, there was a good relationship between MFImax and rosette inhibition. The two most potent disrupters, BD20E4 and D15-50 had the highest IFA titres and the highest MFImax, mAbs D15-68 and E20-76 had intermediate MFImax and rosette inhibition potency, while BDEE10, the least efficient inhibitor had the lowest MFImax. In contrast, there was poor relationship of MFImax and/or IFA titres with the rosette disruption potency (Table
2). The rosette-inhibition assay was more sensitive than rosette disruption, confirming data from other rosetting types [
15]. However, there was no obvious relationship between the two assays. This is particularly striking for mAbs BD20E4, D15-50 and E20-76, which had similar rosette inhibition profiles and dissimilar rosette-disruption capacity. Whether this reflects difference in binding affinity for PfEMP1-VarO or different levels of steric hindrance within the cellular aggregates remains to be established. Affinity of the various mAbs for DBL1-VarO was not explored here, because of the unclear relevance of binding constants for the recombinant antigen with regard to binding to the iRBC surface-exposed PfEMP1-VarO. PfEMP1 is displayed as a multimodular protein (downstream modules may influence binding of inhibitory mAbs) and moreover by a specialized membrane structure, the knob, where it is presented at a high concentration. Moreover, rosette inhibition and rosette disruption involve impairment or displacement of interactions with long blood group saccharides expressed on a variety of RBC surface molecules and steric hindrance is an issue.
Conclusions
The set of anti-VarO mAbs described here will facilitate future studies to design soluble rosetting inhibitors and dissect the specificity of human responses to the VarO antigenic variant, that is commonly recognized by humans living in malaria-endemic areas [
13,
29,
30]. Importantly, all mAbs failed to react with the iRBC surface of related rosette-forming variants such as R29/IT4var9 or 3D7/PF13_0003, like polyclonal antibodies raised to the cognate DBL1 domains [
30]. This confirms the variant-specific surface reactivity observed with a panel of mAbs to the R29/IT4var9 or FCR3S1.2/IT4var60 DBL1 proteins [
45]. In line with this, none of the anti-VarO mAbs reacted with the region of subdomain 2 shown to induce variant-transcending antibodies [
52], which is distant from the binding site areas of the anti-VarO mAbs and partially masked by the surface exposed NTS-VarO domain (see Additional file
4: Figure S3). The identification of a restricted surface area as the binding site of inhibitory mAbs opens the way for fine mapping of the variant-specific interactions of the inhibitory mAbs with the DBL1 domain. This information will be essential to better understand the specificity of antibodies elicited by infection in humans [
13,
29,
30,
47] and determine which natural responses should be harnessed by vaccination and which additional specificities should be elicited by vaccination.
Authors’ contributions
OMP, MG and IVW conceived and designed the study. MG, IVW, NF, JA, HE, and MF conducted the laboratory experiments. OMP, MG and IVW analysed the data. OMP drafted the paper, with input from ALB and GAB for structural analysis.
All authors read and approved the final manuscript.