α2-Macroglobulin Can Crosslink Multiple Plasmodium falciparum Erythrocyte Membrane Protein 1 (PfEMP1) Molecules and May Facilitate Adhesion of Parasitized Erythrocytes

Rosetting, the adhesion of Plasmodium falciparum-infected erythrocytes to uninfected erythrocytes, involves clonal variants of the parasite protein P. falciparum erythrocyte membrane protein 1 (PfEMP1) and soluble serum factors. While rosetting is a well-known phenotypic marker of parasites associated with severe malaria, the reason for this association remains unclear, as do the molecular details of the interaction between the infected erythrocyte (IE) and the adhering erythrocytes. Here, we identify for the first time a single serum factor, the abundant serum protease inhibitor α2-macroglobulin (α2M), which is both required and sufficient for rosetting mediated by the PfEMP1 protein HB3VAR06 and some other rosette-mediating PfEMP1 proteins. We map the α2M binding site to the C terminal end of HB3VAR06, and demonstrate that α2M can bind at least four HB3VAR06 proteins, plausibly augmenting their combined avidity for host receptors. IgM has previously been identified as a rosette-facilitating soluble factor that acts in a similar way, but it cannot induce rosetting on its own. This is in contrast to α2M and probably due to the more limited cross-linking potential of IgM. Nevertheless, we show that IgM works synergistically with α2M and markedly lowers the concentration of α2M required for rosetting. Finally, HB3VAR06+ IEs share the capacity to bind α2M with subsets of genotypically distinct P. falciparum isolates forming rosettes in vitro and of patient parasite isolates ex vivo. Together, our results are evidence that P. falciparum parasites exploit α2M (and IgM) to expand the repertoire of host receptors available for PfEMP1-mediated IE adhesion, such as the erythrocyte carbohydrate moieties that lead to formation of rosettes. It is likely that this mechanism also affects IE adhesion to receptors on vascular endothelium. The study opens opportunities for broad-ranging immunological interventions targeting the α2M—(and IgM-) binding domains of PfEMP1, which would be independent of the host receptor specificity of clinically important PfEMP1 antigens.


Introduction
About 630,000 (0.3%) of the approximately 200 million malaria cases each year are fatal [1]. The majority occur among African children below the age of five years, who die of severe Plasmodium falciparum malaria [2]. The particular virulence of P. falciparum parasites is related to the expression of adhesive proteins on the surface of the erythrocytes they infect, and the P. falciparum erythrocyte membrane protein 1 (PfEMP1) family appears to be of particular importance in this respect. Each parasite genome encodes about 60 antigenically diverse PfEMP1 proteins composed of a series of Duffy binding-like (DBL) and Cysteine-rich inter-domain region (CIDR) domains. The PfEMP1 proteins are expressed on knob-like protrusions on the infected erythrocyte (IE) surface, where they mediate adhesion of IEs to a range of host endothelial receptors [3][4][5][6]. IE sequestration can cause inflammation and organ dysfunction, and can lead to severe and life-threatening complications [7]. In addition to adhesion to receptors in various organs, some P. falciparum-IEs also have the capacity to bind receptors on uninfected erythrocytes, leading to IE/erythrocyte aggregates called rosettes [8,9]. Although the mechanisms and functional significance of rosetting remain unclear, it has repeatedly been associated with parasites causing severe P. falciparum malaria [10][11][12]. Rosetting involves soluble factors in human serum, in apparent contrast to other PfEMP1-mediated IE adhesion to clinically important endothelial protein receptors such as intercellular cell adhesion molecule 1 (ICAM-1) and endothelial protein C receptor [4,6,13]. Several candidates have been proposed (reviewed in ref. [14]), but only pentameric IgM has repeatedly been found to be necessary, although not sufficient, for rosetting to occur [15][16][17]. The molecular details of this interaction between PfEMP1 and IgM have been described [17,18]. PfEMP1 interaction with the rosetting receptors on surrounding erythrocytes is mediated by N-terminal DBL1 α1.5/6/7/8 domains [19][20][21], whereas rosette-facilitating IgM binds to the membrane-proximal, C-terminal end of PfEMP1 [17,18]. How IgM can facilitate rosetting despite binding the opposite end of the PfEMP1 molecule facing the erythrocyte receptor is not known. Based on studies of the IgMbinding and rosette-mediating PfEMP1 protein HB3VAR06, we recently proposed that the function of IgM in rosetting could be cross-linking of PfEMP1 molecules [17], thereby overcoming their low individual affinity for their erythrocyte receptors, which are suspected to be mainly sulfated carbohydrate moieties [12,[22][23][24]. However, we found that IgM alone was not sufficient to facilitate HB3VAR06-mediated rosetting, in line with earlier findings [15,16], and we therefore set out to identify the missing serum component(s) involved in rosetting and describe their involvement in the interaction.

Results
Identification of the human serum protein α 2 M binding to the PfEMP1 protein HB3VAR06 Pentameric IgM facilitates HB3VAR06-mediated rosetting, although additional unidentified serum components are required [17]. To identify these additional components, we used Histagged recombinant full-length HB3VAR06 (FV6) coupled to magnetic epoxy resin beads to identify serum proteins with affinity for FV6. The proteins isolated by this pull-down technique were separated by 2-dimensional (2D) gel electrophoresis ( Fig 1A). Two candidate spots with an apparent molecular weight of 180 kDa and a third candidate spot at 118kDa were detected, in addition to the expected 75-kDa band corresponding to the IgM heavy chain (Fig 1A, left). These spots were not observed in parallel experiments with a recombinant, Histagged DBL domain (DBLβ3_D5) of a PfEMP1 protein (PFD1235w) not mediating rosetting [25,26] (Fig 1A, right). Matrix-assisted laser desorption/ionization (MALDI) time-of-flight (TOF)/TOF mass spectrometry analysis identified all three candidate spots as monomer subunits of the human protease inhibitor α 2 -macroglobulin (α 2 M). α 2 M is a 720-kDa homotetrameric glycoprotein composed of four disulfide-linked 180-kDa subunits that form a cage-like structure [27], which circulates at physiological concentrations of 3-4 μM (2-3 mg)/mL [28]. Among its many roles in regulation and transport, it is best known as a broad-spectrum protease inhibitor (reviewed in ref. [29]). The observed affinity of α 2 M for HB3VAR06 was confirmed by enzyme-linked immunosorbent assay (ELISA) and flow cytometry, demonstrating binding of α 2 M to FV6 and to native HB3VAR06 on IEs (Fig 1B and 1C). In contrast, α 2 M did not bind to the VAR2CSA-type PfEMP1 IT4VAR04, neither recombinant full-length protein (FV2) nor native IT4VAR04 on IEs (Fig 1B and 1C). Fluorescence microscopy of α 2 Mlabeled HB3VAR06 + IEs revealed a punctate pattern of fluorescence ( Fig 1D, top panels) resembling that characteristic of antibody-labeled PfEMP1 [30,31], whereas parallel samples including the detection antibodies but omitting α 2 M did not produce this pattern (Fig 1D, lower panels).

HB3VAR06 can bind native α 2 M
Circulating native α 2 M has a serum half-life of several hours [32]. It contains four "bait" regions that are highly susceptible to proteases, and an internal thioester that is sensitive to cleavage by small nucleophiles such as methylamine (MA). Cleavage of either the "bait" regions or the thioester results in a Venus flytrap-like conformational change in α 2 M. This exposes the thioester, which then covalently "captures" the protease in the cage-like structure and exposes the receptor-binding sites in α 2 M. In vivo, the serum half-life of this "activated" α 2 M is therefore short (2-4 min), as it is rapidly cleared from circulation when the receptor-binding sites bind to the α 2 M receptor low density lipoprotein receptor-related protein 1 (CD91) [32,33]. The conformation of α 2 M activated either by protease cleavage of the bait region or by nucleophile cleavage of the thioester are structurally and functionally similar and both bind CD91 (reviewed in ref. [34]). We therefore used native α 2 M and MA-activated α 2 M (α 2 M-MA) to examine which of the two conformational states could bind HB3VAR06. ELISA analysis showed that FV6 bound approximately four times more native α 2 M than α 2 M-MA (Fig 2A).  Corresponding flow cytometry data showed concentration-dependent binding of native α 2 M binding to HB3VAR06 + IEs, whereas α 2 M-MA did not bind well to IEs, even at high concentrations ( Fig 2B). α 2 M is a promiscuous protein with about one hundred binding partners listed in the Bio-GRID database [35], and in most cases these interactions trigger activation of α 2 M. Methylpoly(ethylene glycol) maleimide (mPEG) forms a covalent bond with the thiol group liberated following α 2 M activation. This can be detected by denaturing SDS gel electrophoresis as an increase in the molecular weight of activated α 2 M. We used this system to test whether interaction with HB3VAR06 activated α 2 M. In solution and in the presence of mPEG, MA activated native α 2 M as expected (Fig 2C, lane 2  HB3VAR06 is composed of an N-terminal segment followed by eight DBL and CIDR domains ( Fig 3A). In an earlier study, we used recombinant DBL domain constructs from HB3VAR06 to show that IgM binds to the penultimate DBL domain (D8; DBLξ2) near the C-terminus [17]. Taking a similar approach here, we could show that α 2 M also bound exclusively to recombinant single-, double-, and triple-domain constructs containing DBLξ2 ( Fig 3B). False signals due to contaminating IgM in the primary or secondary antibodies used to detect α 2 M were ruled out in control experiments without α 2 M (S1 Fig).
Flow cytometry analysis of HB3VAR06 + IEs showed that FV6-and DBLξ2-specific antisera significantly reduced α 2 M binding, whereas an NTS-DBLα-CIDRδ-specific anti-serum did not ( Fig 3C). Thus, IgM and α 2 M bound to the same domain in HB3VAR06. IgM and α 2 M are high-molecular weight proteins (about 900 kD and 720 kD, respectively), but either could bind HB3VAR06 + IEs in the presence of the other (Fig 3D). In competitive binding experiments, each protein reduced the binding of the other by approximately two-fold (Fig 3E and 3F). The results indicate that IgM and α 2 M bind to separate HB3VAR06 molecules with similar affinities, although we cannot rule out that individual HB3VAR06 molecules might be able to accommodate both IgM and α 2 M simultaneously. α 2 M is required for HB3VAR06-mediated rosetting and acts synergistically with IgM Size-exclusion chromatography analysis showed that saturation of native α 2 M by FV6 occurred at a α 2 M:FV6 ratio of at least 1:4, whereas saturation of α 2 M-MA binding to FV6 was reached already at a 1:1 ratio (Fig 4A and 4B). In the absence of serum, HB3VAR06 + IEs do not form rosettes [17], but native α 2 M alone could support rosette formation in serum-free medium in a concentration-dependent manner ( Fig 4C). In contrast, α 2 M-MA had no effect on rosetting ( Fig 4C). This supports our hypothesis that HB3VAR06 rosetting requires cross-linking of multiple PfEMP1 molecules to overcome the low affinity for receptors on erythrocytes [17], although it may simply reflect the inability of α 2 M-MA to bind to IEs ( Fig 2B). Furthermore, the higher multimerization potential of α 2 M than IgM was consistent with our observation that α 2 M could induce rosetting in the absence of other serum factors (Fig 4C), in contrast to IgM [17]. IgM lowered the concentration of α 2 M required for rosetting, indicating that the two serum factors can act in synergy ( Fig 4C). This was supported by the observation that rosetting rates were higher when α 2 M and IgM were added together at equimolar concentrations than the theoretical rates calculated as the sum of rosetting rates in medium with either component alone (S2 Fig). In our earlier study [17], we found that IgM-depleted serum did not support  rosetting although α 2 M was not intentionally removed. However, addition of exogenous native α 2 M at concentrations lower than those required in serum-free medium restored the ability of IgM-depleted serum to support rosetting ( Fig 4D). Thus, endogenous α 2 M levels in IgMdepleted serum appeared to be too low to sustain rosetting, but sufficient to augment IgMdependent rosetting, when subsequently added. α 2 M-binding is not restricted to HB3VAR06 Late-stage erythrocytes infected with five of eight laboratory lines selected to express different rosette-mediating PfEMP1 variants contained sub-populations binding α 2 M, evaluated by flow cytometry (Fig 5A). α 2 M induced serum-free rosetting in one (TM284VAR1(R+)) of three α 2 M-binding lines tested, but not in the two others (PF13 and VarO) (Fig 5B). PF13 IEs express the PfEMP1 protein PF13_0003, which does not contain a DBLξ domain with potential to bind IgM, and this parasite did not form any rosettes when grown in serum-free Albumax medium. VarO + IEs have previously been reported to require at least 5% serum to form rosettes [36], but in our hands this parasite line rosetted at low levels (~10%) under serum-free conditions. Although the penultimate C-terminal DBL domain in VarO is a DBLξ domain according to the classification algorithms employed by Rask et al. [37] used here, not all such domains support Fc-mediated IgM binding (Jeppesen et al., submitted for publication). For both PF13 and VarO, it is tempting to speculate that the inefficiency of α 2 M to induce rosetting in these two parasites is related to absence of synergy between α 2 M and IgM. As expected, α 2 M had no effect on rosetting rates in two lines (HB3VAR03 and IT4VAR09) not binding α 2 M. Ex vivo analysis of IEs from 12 Ghanaian children with P. falciparum malaria showed α 2 M binding in four and IgM binding in five (Fig 5C and 5D). The two phenotypes were highly correlated (P(r -= 0.75) <0.005), and IEs from all the α 2 M-binding isolates also bound IgM (Fig 5D). Four children were clinically categorized as suffering from severe P. falciparum malaria (two had multiple convulsions, one was prostrate, one had respiratory distress). There were no statistically significant associations between clinical presentation (including parasitemia at admission) and either α 2 M-or IgM-binding phenotype (P(r) >0.4 for all). However, our study was neither designed nor powered to thoroughly investigate whether these phenotypes are preferentially found among erythrocytes infected by parasites obtained from patients with severe disease. Together, these data show that the ability to bind α 2 M is a common PfEMP1 phenotype, rather than restricted to HB3VAR06.

Discussion
Serum factors are required for PfEMP1-mediated rosetting to occur in the majority of genetically and phenotypically distinct P. falciparum isolates (reviewed in ref. [14]). Pentameric IgM has been a recurring candidate [38,39], and we recently proposed that its role in rosetting is linked to its ability to bind multiple PfEMP1 proteins [17]. However, pentameric IgM only accommodates two HB3VAR06 proteins per IgM molecule, limiting its multimerization potential, and our results showed that additional unidentified serum component(s) are indeed needed for HB3VAR06-mediated rosetting to occur. We therefore set out to close this gap in understanding, and we report here that the unknown serum factor is the abundant protease inhibitor α 2 M. This is the first report that α 2 M plays a role in the pathogenesis of P. falciparum malaria, and the first demonstration of a single, soluble serum component that is both necessary and sufficient for rosetting of P. falciparum-IEs.
Antigen-specific pull-down and mass-spectrometry showed that α 2 M bound HB3VAR06, and this finding was confirmed by ELISA and flow cytometry employing recombinant and native HB3VAR06, respectively (Fig 1). We next showed that native α 2 M bound HB3VAR06 much better than activated α 2 M, that this binding did not lead to activation of α 2 M, and that native α 2 M retained its capacity for activation after binding to PfEMP1 (Fig 2). This is similar to what has previously been observed in human-pathogenic Group G Streptococci [40,41], where it has been proposed that surface-bound native α 2 M can protect the bacteria from host protease attack [42]. It is possible that α 2 M-binding to PfEMP1 might provide IEs with a similar type of protection, but this is presently an unconfirmed hypothesis.  2 M bound to DBLξ2 near the C-terminus of HB3VAR06, which anchors it in the IE membrane (Fig 3). This is the same domain previously shown to interact with the F c part of IgM [17]. We previously proposed that IgM facilitates rosetting by aligning or "cross-linking" multiple PfEMP1 molecules to increase their combined avidity for receptors on the surrounding erythrocytes, because we found that IgM could bind up to two HB3VAR06 molecules [17]. This "cross-linking" hypothesis is markedly strengthened by the findings reported here. While native α 2 M (which binds at least four PfEMP1 molecules; Fig 4A) supports rosetting on its own, our data also show that α 2 M and IgM act synergistically in facilitating rosetting (Fig 4C  and 4D). Thus, IgM allows α 2 M-dependent rosetting to occur at much lower concentrations of α 2 M than would otherwise be required, although IgM on its own does not support rosetting [17]. Activated α 2 M (which cannot bind multiple PfEMP1 molecules; Fig 4B) appears to play no role in rosetting (Fig 4C). Finally, our results showed that α 2 M binding is not restricted to HB3VAR06, as several other genotypically distinct P. falciparum laboratory lines expressing a variety of PfEMP1 proteins on the IE surface could be labeled by α 2 M (Fig 5A), and that this in some cases led to rosetting (Fig 5B). Furthermore, α 2 M bound to the surface of IEs from Ghanaian P. falciparum malaria patients (Fig 5C), and the ability to bind α 2 M and IgM were linked phenotypes ( Fig 5D). Rosetting has been linked to parasites causing severe malaria in several studies [10,12,43], but in our small data set we did not find significant correlation between α 2 M-binding and severe malaria as defined by the WHO criteria [44]. Future studies are therefore required to determine whether the α 2 M-binding rate is higher among parasites isolated from severe malaria patients, and whether it is associated with particular clinical syndromes and/or expression of particular PfEMP1 proteins.
In this paper we have provided comprehensive evidence that α 2 M is an important component in PfEMP1-mediated rosetting. While the requirement for soluble serum factors in rosetting has long been recognized, their role has not been fully understood. An early study [38], predating the discovery of PfEMP1, suggested that soluble serum factors can form bridges between the IE and the erythrocytes bound to it in the rosette. While our evidence does not formally rule out that possibility, it suggests an alternative mechanism where the soluble serum proteins bind multiple PfEMP1 proteins at their C-terminus, further organizing the knob-specific display of PfEMP1 on the IE surface. The putative α 2 M-(and IgM-) mediated cross-linking of PfEMP1 might improve IE binding avidity by refining the overall topological organization of PfEMP1 or by facilitating intermolecular dimerization of constituent DBL domains. With respect to the former possibility, it is known that the topological and spatial organization of both mannose binding lectin (MBL 3 oligomers) and its ligand (mannose) contribute to the strength of their adhesive interaction [45]. With respect to the latter, closer packing of multiple PfEMP1 head structures might allow intermolecular dimerization of Nterminal DBL domains. DBL dimerization has been proposed to be "conserved in DBL-domain receptor engagement" and important for the adhesive interaction of DBL domains and their receptors [46,47]. Evidence that the binding sites for Blood Group A antigen and heparin are situated on opposing sides of the head structure of the PfEMP1 protein VarO is compatible with this model, as soluble heparin disrupts VarO-dependent rosetting [36]. However, experimental validation of either hypothesis will require further investigation.
Whether α 2 M binding to PfEMP1 arose to subvert a host immune response mechanism, as might be suggested by the protective function of the complement factor C3 homolog TEP1 in mosquito immunity to these parasites [48], or conversely evolved as a parasite immune-evasive strategy similar to that of Group G Streptococci [40,41] is another question that will require further investigation. In either case, it is plausible that the net result is an increased repertoire of host receptors available for IE adhesion. This is likely to enhance IE retention in tissues, either directly (by enabling IE adhesion to low-affinity vascular receptors, e.g., endothelial Blood Group A antigen [49]) or indirectly (through retention of rosetting IEs in the microvasculature). Either way, this would help the P. falciparum parasites to avoid destruction in the spleen. We propose that this at least partially explains the correlation between rosetting and severe malaria. Since the proposed α 2 M-mediated cross-linking of PfEMP1 does not depend on the host receptor specificity of the PfEMP1 ligands involved, our finding may open new avenues for PfEMP1-based immune intervention against IE adhesion that target α 2 M/IgMbinding rather than host receptor-binding epitopes. Such anti-adhesive intervention would have a broader scope than has been possible to date. Recombinant parasite proteins, animal anti-sera, human serum, and α 2 M All recombinant parasite proteins were cloned, expressed, and purified as previously described [50]. In brief, the ectodomains of HB3VAR06, IT4VAR60 (Met 1 to Ser 2,136 ) and IT4VAR09 (Met 1 to Cys 2,345 ) were codon-optimized for insect cell expression by GeneArt (Regensburg, Germany). Full-length, single-, double-, and triple-domain constructs were cloned into the pAcGP67-A vector (BD biosciences, San Jose, CA), transfected and amplified in Sf9 insect cells before being purified from the supernatant of High-Five insect cells via affinity chromatography on HisTrap HP columns (GE Healthcare, Fairland, CT). Recombinant proteins representing PFD1235w-DBLβ3_D5 and the full-length VAR2CSA-type PfEMP1 IT4VAR04 (FV2) were expressed as previously described [26,51]. The ICAM-1-binding full-length PfEMP1 protein IT4VAR13 (FV13) [52] was a kind gift from Thomas Lavstsen. PfEMP1 domain boundaries and sequences not explicitly given above can be obtained from the VarDOM server (http://genome.cbs.dtu.dk/services/VarDom/) [37].

Ethics statement
Animal anti-sera specific for recombinant HB3VAR06, IT4VAR09, and IT4VAR60 constructs were raised in rats and rabbits as previously described [17]. All the animal experiments were conducted according to Danish Law and approved by the Danish Animal Procedures Committee ("Dyreforsøgstilsynet") (permit 2012-15-2934-00567). The mouse monoclonal antibodies specific for the PfEMP1proteins VarO (D1568) and PF13 (J321) were a kind gift from Inès Vigan-Womas and Odile Mercereau-Puijalon. Anti-serum specific for the Nterminal head structure (NTS-DBLα-CIDRα1.4) of HB3VAR03 has been described previously [53] and was a kind gift from Alex Rowe. Human serum was collected and pooled from ten anonymous and healthy blood bank donors without previous exposure to P. falciparum antigens.

Mass spectrometry
Gel spots of interest were excised and cut into 1 mm 3 cubes, de-stained by washing in acetonitrile, and subjected to reduction and alkylation before in-gel digestion with trypsin at 37°C using a ProGest Investigator in-gel digestion robot (Digilab, Marlborough, MA) and standard protocols [57]. Digested peptides were extracted with 10% formic acid, and applied (0.5 μL) to the MALDI target along with α-cyano-4-hydroxycinnamic acid matrix (0.5 μL, 10 mg/mL, 50:50 acetonitrile: 0.1% trifluoroacetic acid) and allowed to dry. MALDI-mass spectrometry data were acquired, using a 4800 MALDI TOF/TOF analyzer (ABSciex, Cheshire, UK) equipped with a Nd:YAG 355 nm laser and calibrated using a mixture of peptides. The most intense peptides were selected for mass spectrometry (MS)/MS analysis using GPS Explorer (ABSciex) to interface with the Mascot 2.4 search engine (Matrix Science) and the MS/MS data using Mascot 2.4 directly. Swiss-Prot (Dec 2012) or NCBInr (Aug 2013) databases were interrogated using Homo sapiens as species restriction. The data were searched with tolerances of 100 parts-per-million for the precursor ions and 0.5 Da for the fragment ions, trypsin as the cleavage enzyme, assuming up to one missed cleavage, carbamidomethyl modification of cysteines as a fixed modification and methionine oxidation selected as a variable modification.
P. falciparum parasites, and in vitro culture and selection P. falciparum HB3 parasites [58] were maintained in vitro and selected for rosetting and expression of HB3VAR06 expression as described [17]. For rosetting laboratory isolates, we used a combination of rosette enrichment (sedimentation in gelatin as described by [59]) and PfEMP1-specific antibody selection [60]. We used rabbit anti-sera raised against the relevant recombinant full-length PfEMP1 proteins to select P. falciparum IT4 to express IT4VAR60 (such parasites have variously been described in the literature as FCR3S1.2 and PAR+; see [61]) and IT4VAR09 (also known as R29; [62]). P. falciparum HB3 was selected to express HB3VAR03 using rabbit-antisera against HB3VAR03-NTS-DBLα-CIDRα. P. falciparum 3D7 expressing PF13_0003 (PF13) and VarO (Genbank: EU9082205) were selected by mouse monoclonal antibodies J321 and D1568 respectively, as described elsewhere [21]. For the isolate TM284var1 (Genbank: JQ684046 and [63]), no PfEMP1-specific anti-sera were available to us. Instead, we used a combination of rosette enrichment and IgM selection. Rosetting in the isolate MUZ12 (Genbank: JQ684048) was maintained by gelatin selection, but no further PfEMP1-specific selection was used. For IgM selection, IEs were incubated in culture medium supplemented with 10% human serum (15 min, rotating plate) before being washed and enriched using rabbit anti-human IgM (Dako, A0426) coupled to Protein A beads (Life technologies). PfEMP1 expression was regularly monitored by flow cytometry using PfEMP1specific antisera, and only IEs where 60-95% of IEs were specifically labeled were used in the experiments described in this study. All cultures were kept synchronous by twice-weekly sorbitol treatment as described [64]. The genotypic identity and the absence of Mycoplasma infection were verified regularly as described [26]. P. falciparum IT4 was selected for IE surface expression of the PfEMP1 protein IT4VAR32b by human monoclonal antibody AB01 [65]. IEs from 12 children with P. falciparum malaria were obtained from venous blood samples collected at Hohoe Municipal Hospital, Hohoe, Ghana. The samples were incubated overnight in candle jars at 37°C, and transported to Accra for analysis by ex vivo flow cytometry (see below). These samples were collected with the permission of the Institutional Review Board at Noguchi Memorial Institute for Medical research, University of Ghana (file 026/ [13][14] and the Ethical Review Committee of the Ghana Health Services (file GHS-ERC 08/05/14).

Detection of α 2 M activation
To test whether binding of α 2 M to FV6 induced thiol-ester conversion of α 2 M to activated α 2 M, FV6 and α 2 M (10 μg) were incubated together in NaPO 4 buffer (pH 8), in the presence of 0.75 mM mPEG, MW 5000 (mPEG; Laysan Bio, Arab, AL). Incorporation of mPEG was assessed by denaturing SDS-PAGE (5 μL load), comparing activation to that obtained by MA (150 mM; Sigma) in the absence of iodoacetamide. To assess whether α 2 M could be activated after binding to FV6, the FV6: α 2 M complex was immobilized on epoxy beads as described above for the α 2 M pull-down, but using purified α 2 M (200 μg) instead of serum. After removal of unbound α 2 M by thorough washing, FV6-bound α 2 M was incubated with mPEG with or without methylamine as above in NaPO 4 buffer (pH8, 2 h, 4°C). After additional washing, the beads were incubated with SDS loading buffer and DTT (50 μL), heated (70°C, 5 min) before loading (20 μL) on denaturing SDS gels and processed as above.

Analytical size-exclusion chromatography
Size-exclusion chromatography (SEC) was performed using a 24-mL Superdex 200 10/30 HR column (GE Healthcare) equilibrated with 50 mM Tris-HCL and 150 mM NaCl (pH 7.4). α 2 M (150 μg) alone and in combination with molar ratios of FV6 were incubated on ice (15 min) and subjected to SEC analysis using a flow rate of 0.5 mL/min and absorbance detection at 280 nm.

Rosetting assessment
Late-stage IEs (5% parasitemia) were diluted in Albumax medium to 2% hematocrit to achieve a total volume of 20 μL in a 384-well plate. Rosetting agents (α 2 M, α 2 M-MA, IgM, IgA, serum) were concentrated to 10 mg/mL and dialyzed against Albumax medium before dilution to the required concentration. The effect of each reagent on IEs (20 μL culture) was tested in triplicate in 384-well plates (20 μL/well, 1 h, 37°C). IEs were labeled with ethidium bromide (2 μg/mL), and rosetting (2 erythrocytes bound per IE) assessed by fluorescence microscopy of 200 IEs. In each assay, the rosetting frequency was compared to the positive control (10% serum).
Supporting Information S1 Fig. Detection of α 2 M-binding to HB3VAR06 is not explainable by presence of IgM in antibody preparations used to label α 2 M. To test whether IgM in the primary, α 2 M-specific, antibody and/or in the HRP-conjugated secondary antibody used to detect α 2 M by ELISA would compromise the assay specificity for detection of α 2 M-binding to HB3VAR06, we coated ELISA plates with FV6 and added goat-anti-α 2 M antibody followed by HRPconjugated rabbit anti-goat antibody (A, left bar) or mouse-anti-α 2 M antibody followed by HRP-conjugated rabbit anti-mouse antibody (B, left bar). Neither produced any signal, in contrast to the strong signal in control wells coated with α 2 M instead of FV6 (right-hand bars in A and B). Graphs show means and S.D. of triplicates. (TIF) S2 Fig. Rosetting of HB3VAR06 + IEs in serum-free medium supplemented with different concentrations of α 2 M and/or IgM. To verify the synergistic effect of α 2 M and IgM on rosetting, we measured absolute rosetting rates (solid lines) of Albumax-maintained HB3VAR6 + IEs after incubation (1 h) in Albumax medium supplemented with IgM alone (black point-up triangles), α 2 M alone (black circles), or equimolar α 2 M and IgM together (black squares) at the concentrations indicated. Synergy was evident, as the observed rosetting rates when α 2 M and IgM were added in combination (black squares) were higher than theoretical rates calculated as the sum of the rates observed in medium containing either α 2 M or IgM (white squares and dashed line). Rosetting rate in medium supplemented with 10% NHS is shown for comparison (right). (TIF) pathology of fatal malaria: evidence for widespread endothelial activation and a potential role for intercellular adhesion molecule-1 in cerebral sequestration. Am J Pathol 145: 1057-1069. PMID: 7526692