Processing of Plasmodium falciparum Merozoite Surface Protein MSP1 Activates a Spectrin-Binding Function Enabling Parasite Egress from RBCs

1Malaria is a debilitating and often fatal infectious disease of tropical and subtropical regions. All associated pathology arises from intraerythrocytic replication of the protozoan parasite Plasmodium. For most of its erythrocytic life cycle, which lasts ∼48 hr in the most dangerous species, P. falciparum, the parasite resides within a parasitophorous vacuole (PV), sequestered from the host cell cytosol. Parasite growth leads to formation of a multinucleated schizont. Merozoites, polarized cells specialized for erythrocyte invasion, bud off from the mature schizont. Shortly thereafter, the PV membrane (PVM) ruptures, releasing the now freely mobile progeny merozoites into the residual erythrocyte cytosol. Within seconds, rupture of the host cell membrane allows egress of the merozoites to invade fresh erythrocytes (for a review of egress see ).

2At least 40 proteins localize to the merozoite surface. Many of these traffic to the parasite plasma membrane during schizont development, where they are tethered via glycosyl phosphatidylinositol (GPI) anchors or through peripheral associations with GPI-anchored proteins. The most abundant merozoite surface component, a GPI-anchored protein called MSP1, is synthesized as an ∼200 kDa protein that in P. falciparum associates with at least two other peripheral proteins belonging to the MSP3 and MSP7 families. MSP1 is conserved throughout Plasmodium and has been scrutinized as a result of its capacity to induce antibody responses that inhibit parasite replication in vitro or protect in vivo (reviewed by ). Gene targeting experiments suggest that MSP1 is essential in the haploid blood stages, but msp1 null mutants could not be established so these studies provided little insight into MSP1 function. Bioinformatic analyses have been similarly uninformative, since MSP1 has no orthologs outside Plasmodium and structural information is sparse. The merozoite surface location of MSP1 has provoked speculation that it functions in erythrocyte invasion. Supporting this are reports that MSP1 binds to erythrocyte glycophorin A, Band 3, and heparin-like molecules, while heparin and related polysaccharides block invasion by P. falciparum merozoites. However, it remains to be demonstrated that MSP1 plays a primary role in invasion, and a mechanistic understanding of MSP1 function is lacking.

3Minutes before egress, a serine protease called SUB1 is discharged from merozoite secretory organelles into the PV lumen, where it cleaves MSP1 and its partner proteins. P. falciparum MSP1 is converted in this primary processing step into four fragments, which initially remain in a non-covalent complex on the merozoite surface. Following egress, MSP1 is further cleaved at a juxtamembrane site by a second parasite protease called SUB2, shedding the bulk of the MSP1 complex. Spatiotemporal regulation of these processing steps is important for parasite viability. Discharge of SUB1—and hence the timing of primary processing—is controlled by a parasite protein kinase (PKG), and inhibition of SUB1 discharge or activity prevents egress. Despite these insights, the role of MSP1 processing is unknown and a picture of how events following SUB1 discharge lead to rupture of the bounding membranes and erythrocyte cytoskeleton has yet to be established.

4Here we show that processing by SUB1 enables MSP1 to interact with the host cell cytoskeleton to play a previously unsuspected role in egress.

5P. falciparum MSP1 is a polymorphic protein that exists in two major isoforms, typified by those of the 3D7 and FCB1 parasite isolates. N-terminal sequencing has mapped three positionally conserved primary processing sites in each of these MSP1 isoforms. The sites are referred to as 83/30, 30/38, and 38/42, after the approximate masses of the cleavage products (Figure 1A). While all are cleaved by P. falciparum SUB1 (PfSUB1), they are structurally distinct, consistent with evidence that PfSUB1 accommodates flexibility in its recognition motif. Only the 38/42 site (i.e., that closest to the C terminus of MSP1) shows significant similarity between 3D7 MSP1 (MSP1-D) and FCB1 MSP1 (MSP1-F) (Figure 1A). Cleavage at the 38/42 site is a rate-limiting processing step, implying special importance.

6Since the identification of SUB1 as the enzyme responsible for MSP1 processing, the possibility of additional processing sites has not been explored. PfSUB1 substrate recognition is dominated by a preference for an aliphatic residue at the P4 position (numbering according to, a small uncharged residue at P2, a polar residue at P1, and acidic residues at one or more of the P1′–P5′ positions. In early work examining cleavage of recombinant MSP1-D in parasite extracts, identified two additional cleavage sites adjacent to the canonical 38/42 site, suggesting redundancy. These motifs (VVQLQ↓NYDEE and PIFGE↓SEDND in MSP1-D), which are partially conserved in MSP1-F (Figure 1A), bear hallmarks of PfSUB1 sites, so we tested whether recombinant PfSUB1 (rPfSUB1) could cleave peptides based on them. All were cleaved at their central bond (Figure 1B), suggesting that both the alternative 38/42 sites in MSP1-D, and at least one of the alternative sites in MSP1-F, might be authentic processing sites. As both lie close to the canonical 38/42 site, they are referred to as the 38/42alt1 and 38/42alt2 sites (Figure 1A). In kinetic assays, the 38/42alt1 peptides from both MSP1 isoforms were cleaved ∼7-fold faster than the respective canonical 38/42 peptides (Figure 1C), an observation important for subsequent work.

7To begin to address the importance of MSP1 processing, mutations were introduced into a recombinant product called Fwt heterodimer, which comprises the two “halves” of MSP1-F refolded into a stoichiometric complex. Substitution of the P2 and P2′ positions at the 83/30 site (mutant Fmut83/30; Figure S1A) ablated cleavage by rPfSUB1 at this position (Figure S1B). Similarly, a recombinant full-length MSP1-F called Fwt with P4 and P2 substitutions at the 30/38 site (mutant Fmut30/38; Figure S1A) was refractory to cleavage at this site (Figure S1C). This showed that appropriate mutations prevent processing and indicated an absence of alternative sites at the 83/30 and 30/38 positions in MSP1-F.

8To examine the potential for preventing cleavage at the 38/42 sites, we produced further Fwt heterodimer mutants designed to block cleavage at one or more of these sites, and at a third putative alternative 38/42 site unique to MSP1-F (Figure 1A). Cleavage within the 38/42 region was abolished by simultaneous mutation of the canonical, alt1, alt2, and putative third alternative sites, but mutagenesis of only one site, or two sites together, or the canonical and alt1 sites plus the putative third alternative site, was insufficient to block cleavage (Figure S1D).

9In view of the special importance of the 38/42 site, this analysis was extended using a full-length recombinant MSP1-D called rMSP1-DCD4wt. Simultaneous mutagenesis of the canonical, alt1, and alt2 sites completely blocked cleavage within the 38/42 region (Figure S1E). Together, these results confirmed the presence of alternative 38/42 sites in both MSP1 isoforms and identified mutations that prevent all PfSUB1-mediated cleavage.

10To test whether mutations that prevent processing are tolerated by P. falciparum, we adopted two complementary strategies. First, we exploited an episomal transgene expression system that allows blasticidin-regulated control of expression levels. Constructs for expression of three forms of MSP1-F (Figure 2A; Figures S2A, and S2B) were transfected into 3D7 P. falciparum, then antibodies specific for MSP1-F used to examine transgene expression on the background of endogenous MSP1-D. Parasites harboring a wild-type msp1-f transgene (3D7pHBIMFwt), or the same gene with mutations at all putative 38/42 sites (3D7pHBIMFmut38/42), or at all primary processing sites (3D7pHBIMFmutall), correctly expressed the transgene product on developing merozoites at all blasticidin concentrations tested (Figure 2B; Figure S2C). Varying blasticidin levels from 2–15 μg ml−1 did not affect growth of the 3D7pHBIMFwt line or parasites harboring a control plasmid, pHBIRH (Figure 2C, top). However, parasites harboring mutant constructs pHBIMFmut38/42 and pHBIMFmutall showed significantly lower growth rates than the 3D7pHBIMFwt line (Figure 2C, bottom). Whereas the 3D7pHBIMFwt line responded to increases in blasticidin concentration by substantially upregulating msp1-f transcript levels (Figure 2D; Figure S2D), likely via increases in episome copy number (confirmed by copy number estimation, data not shown), much less upregulation was seen in the mutant lines, indicating an inability to respond to elevated drug concentrations. Since the episomes differed only at the msp1-f cleavage sites, these results suggested that expression of cleavage-resistant MSP1, even in the presence of endogenous MSP1, is deleterious.

11In a second approach to evaluating the importance of PfSUB1-mediated MSP1 processing, we sought to modify the endogenous msp1 locus using homologous recombination to introduce mutations that prevent processing within the 38/42 region (the importance of processing at the 83/30 and 30/38 sites was not further examined). Our approach used a previously described strategy in which we transfected 3D7 parasites with constructs containing targeting sequence fused to synthetic “recodonized” sequence encoding a chimeric MSP1 C-terminal domain (Figure 3A). Integration produces a chimeric gene, the product of which can be distinguished from unmodified MSP1-D by its reactivity with the MSP1-F-specific monoclonal antibody (mAb) 111.4. Integration thus epitope tags the gene.

12Four integration constructs were initially generated (Figure 3A). Construct pHH1MSP1chim_wt was designed to replace the 3′ region of the msp1 ORF with the chimeric sequence but leave the 38/42 processing sites unaltered. It thus acted as a control for all other genetic experiments. Constructs pHH1MSP1chim_can and pHH1MSP1chim_alt1 were identical to pHH1MSP1chim_wt except that they were designed to introduce di-leucine mutations at the P2 and P1 positions of the canonical 38/42 site or the 38/42alt1 site, respectively; these substitutions blocked processing of recombinant MSP1 (Figure S1). Construct pHH1MSP1chim_can+alt1 was designed to introduce both these sets of substitutions upon integration, thus blocking cleavage at both the canonical and 38/42alt1 sites. Parasites independently transfected with the constructs were subjected to drug cycling (growth in the absence then presence of WR99210) to select for integration. PCR analysis detected integration of all constructs by drug cycle 2 (data not shown), and clones derived from the drug-resistant lines were recognized by mAb 111.4, confirming correct integration (Figure 3B, top five rows). Sequencing of PCR products amplified from the modified msp1 locus of the clones confirmed the presence of the mutations (data not shown).

13The transgenic clones grew normally, indicating no effects of the mutations on viability (data not shown). To assess the impact on MSP1 processing, schizont extracts and culture supernatants containing shed MSP1 fragments were examined by western blot. This showed a shift in migration of the MSP142 cleavage product in the chim_can and chim_can+alt1 clones (Figure 3C), with production of progressively larger fragments termed MSP142∗ and MSP142∗∗. This was consistent with blockade of cleavage at the canonical site (chim_can) or at both modified sites (chim_can+alt1), resulting instead in cleavage at the 38/42alt1 site or the 38/42alt2 site, respectively. Examination of culture supernatants (Figure 3D) showed increases in the mass of the shed MSP133 fragment in the chim_can and chim_can+alt1 clones, again consistent with ablation of processing at the canonical 38/42 site, or both sites, respectively. To confirm the site of cleavage when processing at both the canonical and 38/42alt1 sites was prevented, we purified the MSP133∗∗ species from culture medium of a chim_can+alt1 clone (Figure S3). Edman degradation identified its N terminus as NYDEE, confirming the 38/42alt2 cleavage site and proving the presence of alternative, redundant 38/42 cleavage sites in MSP1-D. The lack of a growth defect in the chim_can, chim_alt1, and chim_can+alt1 parasite clones proved that cleavage at the canonical or 38/42alt1 sites is not essential; blocking cleavage at one or both positions simply shifted cleavage to an alternative available 38/42 site.

14To examine the effects of blocking processing at all three 38/42 positions, four additional transfection constructs were generated (Figure 3A). Constructs pHH1MSP1chim_alt2 and pHH1MSP1chim_triple were designed to introduce mutations that block cleavage at the 38/42alt2 site or all three 38/42 sites respectively (Figure S1). In addition, pHH1MSP1chim_Δ+can and pHH1MSP1chim_Δ+mut were designed to delete a 69-residue predicted unstructured (data not shown) segment of MSP1 sequence that encompasses both the 38/42alt1 and 38/42alt2 sites. Construct pHH1MSP1chim_Δ+can was designed to leave the canonical 38/42 site unaltered, whereas pHH1MSP1chim_Δ+mut would additionally render this site non-cleavable. Both pHH1MSP1chim_alt2 and pHH1MSP1chim_Δ+can rapidly integrated (data not shown), and the resulting parasite clones showed the expected reactivity with mAb 111.4 (Figure 3B, bottom 2 rows). In contrast, despite five independent transfection experiments, each with extended periods of drug cycling, integration of pHH1MSP1chim_triple and pHH1MSP1chim_Δ+mut was never detected. Since, aside from the mutations unique to these constructs, they were identical to the other six constructs that readily integrated, this result suggested that cleavage of at least one position within the 38/42 region of MSP1 is important for parasite viability.

15Size-exclusion chromatography of rPfSUB1-cleaved rMSP1-DCD4wt as well as a similar protein lacking the CD4 tag (rMSP1-Dwt; Figure S1E) showed that, like parasite MSP1, the processed products remain associated under non-denaturing conditions (Figure S4A). This encouraged us to use the recombinant proteins to examine the structural consequences of cleavage. Circular dichroism (CD) of intact and rPfSUB1-processed rMSP1-Dwt, as well as of a mutant (rMSP1-Dmut) that was refractory to cleavage in the 38/42 region (Figure S1E), showed that processing altered the secondary structure of both proteins (Figure 4A). These changes were less extensive in rMSP1-Dmut, indicating that cleavage at the 38/42 region contributed to the conformational rearrangements.

16MSP1 has been implicated in interactions with erythrocyte surface heparin-like polysaccharides, so we compared the capacity of intact and rPfSUB1-cleaved rMSP1-Dwt to bind to immobilized heparin. Cleaved rMSP1-Dwt showed ∼4-fold higher binding than intact protein, which was reduced by soluble heparin (Figure 4B). We next examined the ability of rMSP1-Dwt to bind to intact erythrocytes. No binding was detected (data not shown). However, permeabilized erythrocytes incubated with cleaved rMSP1-Dwt showed an ∼3.3-fold more intense IFA signal than cells exposed to intact rMSP1-Dwt (Figure 4C), suggesting that cleavage enhanced binding to an intraerythrocytic component. This was confirmed in pull-down assays using inside-out erythrocyte ghost vesicles (IOVs) (Figure 4D), as well as by immunoEM analysis of erythrocyte cytoskeletons (Figure 4E; Figure S4D), which in both cases showed preferential binding of cleaved rMSP1-Dwt. To determine the target(s) of binding, we probed SDS PAGE-fractionated erythrocyte ghosts in overlay assays with intact or rPfSUB1-processed rMSP1-Dwt. Cleaved rMSP1-Dwt bound exclusively to a Triton X-100-insoluble doublet migrating at the positions of α- and β-spectrin, the dominant components of the cytoskeleton (Figure 4F). This was confirmed by probing purified spectrin “spiked” with irrelevant proteins (Figure 4G). No binding was observed for intact or rPfSUB1-treated rMSP1-DCD4mut (Figure 4H), showing that cleavage within the 38/42 region—already shown to be important for parasite viability—was required for binding to spectrin.

17The erythrocyte cytoskeleton lies beneath the cell membrane so merozoites are unlikely to contact it during invasion. However, intracellular merozoites impinge on the inner face of the host cell membrane in the brief period between PVM rupture and egress (e.g., ), so we explored the possibility that direct interactions between processed merozoite surface MSP1 and host cell spectrin might play a part in egress. For this, we returned to the chim_Δ+can mutant (Figure 3) in which an MSP1 segment had been deleted to remove the 38/42alt1 and 38/42alt2 sites entirely, leaving just the canonical 38/42 cleavage site. Since this is a relatively poor substrate for PfSUB1 (Figure 1C), we predicted that cleavage within the 38/42 region in the chim_Δ+can mutant should be less efficient than in wild-type parasites. To test this, we compared the kinetics of processing in chim_Δ+can parasites with that in chim_wt parasites, which expressed the same chimeric MSP1 but retained all three 38/42 cleavage sites. For these experiments, schizonts were treated with the reversible PKG inhibitor compound 1 (C1), which prevents PfSUB1 discharge, stalling schizont development at the final stage of maturation. MSP1 processing and egress occur within minutes of washing away the inhibitor. As shown in Figure 5A, processing of MSP1 in the chim_Δ+can mutant was delayed relative to chim_wt parasites and was characterized by an unusually prominent MSP138+42 processing intermediate. Comparison of the kinetics of chim_Δ+can and chim_wt egress by time-lapse microscopy showed a reproducible delay in egress in the chim_Δ+can parasites following C1 removal (Figure 5B; Movie S1), mirroring the delay in MSP1 processing. This was confirmed in further experiments in which the clones were imaged simultaneously following fluorescent labeling of one population to identify it (Movie S2; Figure S5). Since the chim_Δ+can and chim_wt parasites differed only by the presence or absence of an MSP1 segment encompassing the 38/42alt1 and alt2 cleavage sites, these results showed that processing of MSP1 regulates the kinetics of egress.

18MSP1 is tethered to the merozoite surface via a C-terminal GPI anchor. To further test our model that direct interactions between merozoite-bound MSP1 and the erythrocyte cytoskeleton facilitates egress, we used a recently published conditional strategy to generate P. falciparum transgenics in which a 3′ segment of the msp1 gene could be deleted by rapamycin (RAP)-inducible, Cre recombinase-mediated excision (Figure 6A; Figure S6A). This was predicted to generate a truncated MSP1 that lacked a GPI anchor and so would not be bound to the merozoite surface. Analysis of RAP-treated 3D7MSP1flox42C parasites showed highly efficient excision, resulting in exclusive expression of truncated MSP1 in mature schizonts at the end of the same erythrocytic cycle (Figures 6B and 6C). No effects on merozoite development were discernible. The modified MSP1 was trafficked to the PV as expected for a non-membrane-bound merozoite surface protein (Figures 6C–6D; Figure S6B) but was not present on the surface of free merozoites (Figure 6E). Video microscopy of the RAP-treated 3D7MSP1flox42C schizonts revealed a dramatic egress defect characterized by abortive erythrocyte membrane rupture and trapping of the merozoites in the partially ruptured cell (Figure 6F; Movies S3 and S4). Consistent with this, the RAP-treated mutants displayed a substantially reduced replication rate (Figure 6G and Figure S6C). These results show that MSP1 functions at egress and that this role requires it to be tethered to the merozoite surface. Collectively, our findings support the model that processing of MSP1 facilitates host cell membrane rupture, probably through interactions between the mature merozoite surface and the erythrocyte cytoskeleton.

19We have combined genetic, structural, and functional analysis with microscopic observation of egress to produce evidence that: (1) proteolytic maturation of MSP1 by SUB1 is important for parasite viability; (2) proteolysis alters MSP1 secondary structure, conferring upon it a capacity to bind to both heparin and erythrocyte spectrin; and (3) these functional alterations regulate egress, probably as a result of interactions between MSP1 and the host cell cytoskeleton. The resistance of the erythrocyte membrane to mechanical shear stress is dependent on the structural integrity of its cytoskeleton and in particular its underlying lattice of spectrin tetramers formed by the head-to-head association of pairs of αβ spectrin heterodimers. The spectrin network is dynamic, accommodating reversible breakage and reformation of the dimer-dimer bonds in response to even moderate shear stress (e.g., ). Shear forces can also result in unfolding of the triple-helical repeat units that comprise α- and β-spectrin, providing additional flexibility. This dynamic state allows peptides and other small molecules that interfere with tetramer stability or that perturb interactions between spectrin and other cytoskeletal components such as ankyrin, protein 4.1R, and actin to destabilize the membrane. SUB1-processed MSP1 may perform an analogous role. We speculate that following PVM breakdown, the diffusive movement of intracellular merozoites impinging upon the inner face of the erythrocyte membrane—well documented by both time-lapse and diffraction phase microscopy (see also Movie S5)—enables merozoite surface-bound MSP1 to bind the spectrin lattice, producing internal shear forces that disrupt the cytoskeleton (Figure 7). This is likely aided by protease activity, perhaps involving host cell calpain-1 and/or the PfSUB1 substrate SERA6, since the cysteine protease inhibitor E64 selectively inhibits host cell membrane rupture (e.g., ). Even localized destabilization of the cytoskeleton may be sufficient to allow egress, since high-speed video microscopy has shown that erythrocyte membrane rupture initiates at a single site; subsequent elastic inversion of the membrane promotes its rapid disintegration. Interestingly, reported spectrin-binding activity for a recombinant MSP1 polypeptide, suggested by those authors as being important for intracellular parasite development. We do not favor that model, since parasites replicate within the PVM, which shields them from the host cytoskeleton. In contrast, the egress delay observed in the chim_Δ+can mutant, and the egress defect (with no effect on schizont development) when MSP1 is conditionally converted to a non merozoite-bound form, implies a role for processed merozoite-bound MSP1 in host cell rupture. Our model explains the defect associated with episomal expression of cleavage-resistant MSP1 (Figure 2), which presumably reduces egress efficiency by reducing the proportion of MSP1 at the merozoite surface able to interact with the host cell cytoskeleton. These data implicate a surface protein in the egress of an intracellular non-viral pathogen. They also provide a plausible mechanistic rationale for the timing of MSP1 processing by SUB1, which “prepares” the merozoites for partaking in their own release.

20We do not rule out additional roles for MSP1. A previous report showed that knockdown of MSP1 expression in parasite liver stages ablated merozoite formation, suggesting a role in merozoite budding. Additionally, our observation that processing enhances binding to heparin tempts speculation that SUB1 may activate MSP1 to perform a function at invasion. However, the fact that RAP-treated 3D7MSP1flox42C parasites lacking surface-bound MSP1 produce normal numbers of merozoites and replicate in vitro (albeit at a very reduced rate) shows that merozoite surface MSP1 is dispensable for merozoite development and invasion in blood stages. Compounds that inhibit MSP1 processing or that block interactions with spectrin may form the basis of antimalarial drugs that interfere with this key step in the malarial life cycle.

21P. falciparum clones FCB1, 3D7, and 1G5DC were maintained in RPMI 1640 medium with Albumax (Invitrogen) and synchronized using standard procedures. Transfection, selection with WR99210 (Jacobus Pharmaceuticals), and cloning was as described. Growth rates were determined by microscopy or fluorescence-activated cell sorting (FACS) as described. Details of transfection constructs based on the pHBIRH episome and integration plasmid pMSP1chimWT are provided in Supplemental Experimental Procedures, as are details of the construct used to flank a segment of the 1G5DC msp1 ORF with loxP sites. For conditional truncation of MSP1 in the 3D7MSP1flox42C clones, synchronous ring-stage parasites were treated for 4 hr with 100 nM RAP.

22Monoclonal antibodies 89.1, X509, and 111.4, rabbit polyclonal antibodies and their use in western blot and IFA analysis have been described. Production and purification of rPfSUB1, Fwt, and Fwt heterodimer was as described. Mutants of Fwt and Fwt heterodimer were produced using QuikChange II (Agilent) site-directed mutagenesis of parent plasmids. For rMSP1-Dwt, MSP1-D (minus its GPI anchor) was expressed in HEK293E cells; for rMSP1-DCD4wt, it was fused to domains 3 and 4 of rat CD4. Cleavage site mutants were produced by replacing segments of the expression constructs with synthetic gene fragments containing substitutions. The proteins were purified by nickel chelate and size-exclusion chromatography.

23Synthetic peptides were from Biomatik (http://www.biomatik.com). Peptide cleavage assays and product identification by RP-HPLC and mass spectrometry were as described. To purify shed MSP1 fragments, 3D7 or chim_can+alt1 schizonts were allowed to undergo egress in protein-free medium then the supernatants fractionated on a Vydac 4.6 × 150 mm 214TP C4 RP-HPLC column. The MSP133 and MSP133∗∗ species were identified by western blot then the proteins transferred to PVDF membrane for N-terminal sequencing (PNAC).

24First strand cDNA synthesis was performed using a SuperScript II First-Strand Synthesis Kit (Invitrogen) according to the manufacturer’s instructions. Quantitative real-time PCR (qRT-PCR) was performed using the ABI 7500 sequence detection system and a SensiFASTSYBR Lo-ROX kit (Bioline). Data were analyzed with SDS 1.3.1 software (Applied Biosystems). Transgene expression was displayed as a percentile of endogenous msp1-d expression (100%).

25Purified rMSP1-Dwt and rMSP1-Dmut (0.156 mg ml−1 in 500 μl 25 mM HEPES [pH 7.4], 150 mM NaCl, 15 mM CaCl2) were monitored on a Jasco J-715 spectropolarimeter for 5 hr at 37°C with or without added rPfSUB1 (6 μl at 0.84 mg ml−1). Secondary structure composition was averaged using CONTINLL, SELCON3, and CDSSTR. Secondary structure predictions were performed with JPred (http://www.compbio.dundee.ac.uk/www-jpred/).

26Heparin-agarose beads (Sigma) in assay buffer (25 mM HEPES [pH 7.4], 15 mM NaCl, 0.07% Tween 20) were incubated with intact or cleaved rMSP1-Dwt (50 μl at 0.1 μg μl−1). Control samples were additionally supplemented with heparin sodium salt (1 mg ml−1, Sigma). Following incubation for 20 min at room temperature, supernatants containing unbound proteins were recovered and the beads washed five times with assay buffer. Bound proteins were eluted into 50 μl SDS sample buffer then all samples subjected to reducing SDS-PAGE on a 4%–16% gradient gel. The gel was stained with Coomassie blue, imaged using a BioRad Chemidoc MP system and band intensities estimated using Image Lab software.

27Overlay assays to detect binding to SDS-PAGE fractionated human erythrocyte ghost proteins were as described by. IOVs were prepared using standard procedures from erythrocyte ghosts (see Supplemental Experimental Procedures) and incubated in PBS with intact or rPfSUB1-cleaved rMSP1-Dwt, rMSP1-DCD4wt, or rMSP1-DCD4mut before washing and analysis by western blot, detecting bound proteins with mAb 89.1. For immunoEM analysis, TX-100-treated cytoskeletons immobilized on grid grids were incubated with intact or cleaved rMSP1-Dwt (0.1 μg ml−1) then washed and probed with anti-MSP1 antibodies followed by 5 nm gold-conjugated anti-rabbit IgG, before staining with sodium silicotungstate.

28P. falciparum egress was imaged as described, using C1 to synchronize egress. Microscopic DIC images were routinely collected at 5 s intervals for up to 30 min. For comparison of 3D7 chim_Δ+can and 3D7 chim_wt parasites, populations were either alternately imaged or combined in the same microscopy chamber after labeling one mutant with Hoechst 33342 prior to washing away C1. An initial fluorescence image was collected prior to starting the time-lapse DIC imaging, then the fluorescence and first DIC images overlayed to identify labeled cells. Image files were exported as AVI movies using Axiovision 3.1 software. Time to individual egress events was recorded by visual examination of movie frames.

29S.D., N.H., and E.T.F. performed the experiments. A.J.P. produced recombinant MSP1. C.W.-M. and S.R.M. performed biophysical analyses. M.L.J. and C.R.C. developed conditional methodologies. H.R.S., G.J.W., M.T., C.E., and M.J.B. supervised the work. S.D., N.H., C.E., and M.J.B. wrote the manuscript.