The Structure of Plasmodium falciparum Blood-Stage 6-Cys Protein Pf41 Reveals an Unexpected Intra-Domain Insertion Required for Pf12 Coordination

Plasmodium falciparum is an apicomplexan parasite and the etiological agent of severe human malaria. The complex P. falciparum life cycle is supported by a diverse repertoire of surface proteins including the family of 6-Cys s48/45 antigens. Of these, Pf41 is localized to the surface of the blood-stage merozoite through its interaction with the glycophosphatidylinositol-anchored Pf12. Our recent structural characterization of Pf12 revealed two juxtaposed 6-Cys domains (D1 and D2). Pf41, however, contains an additional segment of 120 residues predicted to form a large spacer separating its two 6-Cys domains. To gain insight into the assembly mechanism and overall architecture of the Pf12-Pf41 complex, we first determined the 2.45 Å resolution crystal structure of Pf41 using zinc single-wavelength anomalous dispersion. Structural analysis revealed an unexpected domain organization where the Pf41 6-Cys domains are, in fact, intimately associated and the additional residues instead map predominately to an inserted domain-like region (ID) located between two β-strands in D1. Notably, the ID is largely proteolyzed in the final structure suggesting inherent flexibility. To assess the contribution of the ID to complex formation, we engineered a form of Pf41 where the ID was replaced by a short glycine-serine linker and showed by isothermal titration calorimetry that binding to Pf12 was abrogated. Finally, protease protection assays showed that the proteolytic susceptibility of the ID was significantly reduced in the complex, consistent with the Pf41 ID directly engaging Pf12. Collectively, these data establish the architectural organization of Pf41 and define an essential role for the Pf41 ID in promoting assembly of the Pf12-Pf41 heterodimeric complex.


Introduction
Apicomplexan parasites of the Plasmodium genus are the etiological agents of malaria, one of the most severe and widespread infectious diseases in the developing world. While several species of Plasmodium can cause malaria in humans, P. falciparum is the most lethal [1]. In 2013, there were an estimated 198 million cases of malaria that led to 584,000 deaths, 90% of which occurred in Africa and the majority of which were caused by P. falciparum infection [1]. The success of Plasmodium spp. is due, in part, to a complex life cycle that relies on sophisticated molecular strategies to both survive transmission in mosquito vectors and access the immunoprotective environment of host cells [2,3]; to accomplish this, Plasmodium parasites encode a diverse arsenal of surface displayed proteins capable of interfacing with biomolecular partners on vector and host cells. The 6-Cys s48/45 family of surface antigens has garnered particular attention as they are differentially expressed at every stage of the P. falciparum life cycle and have been shown to play important biological roles [4].
Of the 14 members that comprise the 6-Cys s48/45 protein family in Plasmodium falciparum [5,6], Pf36, Pf52, sequestrin and B9 are expressed on the sporozoite or liver-stage parasite surface and play critical roles in invasion of or growth within hepatocytes [6][7][8][9]. Pfs230, Pfs48/ 45 and Pfs47, designated with the "s" for sexual stage specific antigens, are presented on the surface of gametocytes and are involved in fertilization of male and female gametes in the human blood stream as well as escape from the mosquito immune system [10][11][12]. Pf12, Pf38, Pf41 and Pf92 are localized on the surface of the blood-stage merozoite, which is the form of the parasite that invades erythrocytes, proliferates, and causes the symptoms of malaria [2]. On the merozoite surface, Pf12 and Pf41 are displayed as a heterodimer with Pf12 tethered to the outer membrane via a glycophosphatidylinositol (GPI) anchor and soluble Pf41 appropriately localized through its interaction with Pf12 [13][14][15][16]. Notably, Pf12 is the fifth most prevalent GPI-anchored protein on the merozoite surface [13], both Pf12 and Pf41 are strongly recognized by antibodies from naturally infected patients [16][17][18][19], and Pf41 was recently identified as one of five top-ranked potential malaria vaccine candidates [20]. Strikingly, however, no phenotype was observed in Pf12 or Pf41 knockout parasites, although it was noted that this observation may be due to the ability of the parasites to adapt in culture through activating compensatory mechanisms [14].
Recent structural analysis of Pf12 provided the first insight into the architecture of the 6-Cys domain and the organization of the tandem repeats with the two disulfide pinned βsandwich domains connected via a short linker [5,15]. Intriguingly, Pf41 incorporates an additional 120 residues of unknown structure that appear to form a spacer linking the two predicted 6-Cys domains (D1 and D2) ( Fig 1A). Based on the expected domain organization of Pf41 [17], the structure of Pf12 [15], and chemical cross-linking data [15], we previously proposed a preliminary model where Pf12 and Pf41 adopt an antiparallel organization [15]. A weakness of this model, however, is the absence of structural information describing Pf41 and, in particular, the large sequence insertion. To address this limitation, we report here the crystal structure of Pf41 and complement the structural data with solution binding studies to reveal the molecular determinants of Pf12-Pf41 complex formation. Based on these data, we re-evaluate and refine the model of the assembly mechanism and overall architecture of the Pf12-Pf41 heterodimer.

Materials and Methods
Cloning, protein production and purification Two forms of Pf41 and one form of Pf12 were recombinantly produced in this study. The mature full length Pf41 construct extends from the predicted signal peptide cleavage site to the C-terminus (Lys21 to Ser378; PlasmoDB: PF3D7_0404900), while the Pf12 construct extends from the signal peptide cleavage site to the GPI anchor site (His26 to Ser321; PlasmoDB: PF3D7_0612700) as described previously [15]. For both Pf12 and Pf41, N-linked glycosylation sites were mutated and the genes codon optimized for insect cells, synthesized and subcloned into a modified pAcSecG2T vector (Pharmingen) with a TEV protease cleavable N-terminal hexahistidine/maltose binding protein (MBP) tag. The Pf41 construct lacking the inserted domain-like region (Pf41ΔID) was cloned out of Pf41 by overlap extension PCR enabling the replacement of the ID (Thr117 to Ile225) with a GSGGSG linker. Protein production and Anomalous difference electron density map of Pf41 calculated at 8 sigma (orange mesh) around one of two high confidence zinc sites (grey sphere) used for phasing. Two Glu residues (Glu 132 and 134) were found to coordinate the zinc ion (along with His23 and Glu369 of two neighboring chains, not shown) and are shown in green ball-and-stick colored by element. (C) Left/middle-Orthogonal views of the Pf41 structure, shown as a cartoon backbone colored as in (A). Disulfides are shown in the left panel as ball-and-stick colored by element. A four residue un-modeled region in D1 is indicated by a dotted green connecting loop. Grey dotted lines extending out of α1" (inserted domain-like region, ID) and into β9 (D1) indicate the un-modeled region of the ID (~85 residues). Right-green surface of Pf41 in same orientation as (C, middle) colored as in (A), with the previously undefined region shown with a semi-transparent grey surface with underlying cartoon. (D) Left-Lys349 shown as green sticks colored by element with the positive Fo-Fc map shown as an orange mesh contoured at 2.5σ. Right-final σ-A weighted 2Fo-Fc electron density map shown as a purple mesh contoured at 1σ around the dimethylated lysines (MLY347 and 349). The symmetry mate against which the methylated lysines pack is shown as a semi-transparent grey surface. (E) Sequence of the ID and flanking secondary structure elements colored as in (A). Regions observed in the structure have a solid underline; un-modelled sequence and predicted secondary structure elements are indicated by dashed lines. doi:10.1371/journal.pone.0139407.g001 Impact of Pf41 Domain Insertion on Pf12-Pf41 Assembly purification was performed using established protocols [15]. Briefly, expression viruses were generated and amplified in Spodoptera frugiperda 9 cells and protein production was performed in Hi5 cells. Growth media was harvested after a 65 hour infection and secreted proteins purified by nickel affinity chromatography. The hexahistidine/MBP tag was cleaved with TEV protease and removed by cation exchange chromatography. Proteins were further purified by size exclusion chromatography (SEC) in HEPES-buffered saline (HBS: HEPES pH 7.5, 150mM NaCl) with 2% glycerol.

Protein methylation
Prior to crystallization, Pf41-MBP fusion protein was concentrated to 1.4 mg/mL and dialyzed into 50 mM HEPES pH 7.5, 250 mM NaCl and 2% glycerol for lysine methylation based on a previously published protocol [21]. Briefly, borane-dimethylamine complex (Acros Organics) and methanol-free formaldehyde (Thermo Scientific) were sequentially added to final concentrations of 50 mM and 80 mM, respectively, and incubated with the protein overnight at 4°C. The methylation reaction was stopped by addition of 100 mM glycine. Methylated Pf41-MBP was cleaved with TEV protease and purified by cation exchange and SEC as described above. Methylated protein was used only for crystallization experiments.

Crystallization and data collection
Crystallization trials were set using a Crystal Gryphon (Art Robbins Instruments). Crystals of methylated Pf41 were identified in the PACT Premier Screen (Molecular Dimensions) using sitting drops at 18°C. The final drops consisted of 0.3 μL Pf41 at 15 mg/mL with 0.2 μL of reservoir solution (10 mM zinc chloride, 0.1 M MES buffer pH 6.0, 20% PEG6000) and were equilibrated against 55 μL of reservoir solution. Crystals were cryoprotected in mother liquor with 15% glycerol and flash cooled in liquid nitrogen. Diffraction data were collected on beamline 12-2 at the Stanford Synchrotron Radiation Lightsource.
Data processing, structure determination and refinement Diffraction data were initially processed to 2.61 Å resolution using Imosflm [22] and Aimless [23]. The structure of Pf41 was determined by zinc single-wavelength anomalous dispersion. Two high confidence Zn sites were identified and refined using Phenix.autosol [24,25], which enabled building of approximately 90% of the backbone using Phenix.autobuild [26]. The nearly complete model was improved following refinement in Phenix.refine [27] against the higher resolution (2.45 Å resolution) data set. Final model building and solvent atom selection was performed in COOT [28]. Structural validation was performed with MolProbity [29]. Overall, 5% of reflections were set aside for calculation of R free . Data collection and refinement statistics are presented in Table 1. The atomic coordinates and structure factors for Pf41 have been deposited in the Protein Data Bank under the following PDB ID: 4YS4.

Isothermal titration calorimetry
ITC measurements were carried out at 25°C on a MicroCal iTC200 instrument (Malvern). The sample cell contained Pf12 at a concentration of 10 μM, with full length Pf41 or Pf41ΔID at a concentration of 100 μM added in 16 injections of 2.4 μL each separated by 180 seconds. Data were analyzed with Origin software (MicroCal) and the dissociation constant (K d ) was determined using a one-site model. Values were derived from a single experiment, but are representative of two independent experiments.

Trypsin protection assay
Pf41 was incubated with an equal molar ratio of Pf12 at 4°C for 10 min to allow for complex assembly. Trypsin was added to the protein samples (Pf12, Pf41, or Pf12-Pf41 mixture) at a trypsin:protein ratio of 1:100 (w/w). Reactions were incubated at 4°C. Aliquots of each reaction were taken at time intervals between 0 min and 2 hours and inactivated by the addition of protease inhibitor in SDS loading buffer. Aliquots were heated at 95°C, separated by SDS-PAGE, and visualized with Coomassie Brilliant Blue. Select protein bands were cut out of the polyacrylamide gel, trypsin digested, and analyzed by MALDI-TOF mass spectrometry as described previously [15].

Organization of the Pf41 6-Cys domains reveals an unexpected juxtaposition
To investigate the domain organization of Pf41, including the architecture of the predicted spacer region (Fig 1A), we adopted an insect cell based expression system to produce sufficient quantities of correctly folded recombinant protein to support structural and biophysical studies [15]. Pf41 initially proved recalcitrant to crystallization potentially due to the high surface entropy resulting from 40 lysine residues present in the mature protein, representing approximately 11% of the total sequence. To compensate for this innate property of Pf41, surface lysines were chemically methylated during purification, and crystals of methylated Pf41 were observed after eight months. Intriguingly, attempts to solve the structure of Pf41 by molecular replacement were unsuccessful, likely indicating architectural divergence from the only other structurally characterized 6-Cys protein, Pf12 [5,15]. Ultimately, the structure of Pf41 was determined by zinc single-wavelength anomalous dispersion to a resolution of 2.45 Å (Fig 1B). One Pf41 molecule was present in the asymmetric unit, and PISA analysis [30] indicated that Pf41 crystallized as a monomer consistent with a previous analysis showing no homo-multimerization capabilities [14]. The refined model of Pf41 begins at Lys21 and extends through Glu369, with a four residue disordered region (Lys59 to Ser62) and an extended, un-modelled region between Ser137 and Ile224 (Fig 1C, left). During refinement, two clear positive difference density peaks were observed near the epsilon amino groups of each of three lysine residues consistent with dimethyl modifications (Fig 1D, left). Two of the dimethylated lysines were located at a crystal packing interface (Fig 1D, right), underscoring the importance of lysine methylation in obtaining diffraction quality crystals of Pf41. Structural analysis confirms that both Pf41 6-Cys domains (D1 and D2) adopt a β-sandwich fold with expected disulfide connectivity of C1-C2, C3-C6 and C4-C5; the first two disulfides pin together the two sheets of the β-sandwich and the latter disulfide anchors an ancillary loop to the domain core (Figs 1C and 2). As evident from structural overlays, the architecture of the two Pf41 6-Cys s48/45 domains are similar to those of Pf12 (Fig 2) [5,15]. Specifically, the D1 domains overlay with a root mean square deviation (rmsd) of 1.1 Å over just 44 Cα atoms indicating conservation of the small domain core but divergent conformations of the inter-strand loops (Fig 2A), which is consistent with our inability to solve the structure of Pf41 by molecular replacement. The D2 domains were more similar as they overlay with an rmsd of 0.99 Å over 86 Cα atoms, indicating conservation of both the domain core and most inter-strand loops ( Fig 2B). Intriguingly, despite a rotation of D1 relative to D2, the relative spatial positioning of the two 6-Cys domains of Pf41 is analogous to that of Pf12 (Fig 2C), with overall dimensions for these two domains of approximately 65 by 40 Å, despite the prediction that the 6-Cys domains of Pf41 are separated by a large spacer (Fig 1A). The unexpected topological similarity is mediated by a relatively conserved, at least with respect to sequence length, inter-domain linker in Pf41 comprising residues SNNNV (Fig 2C) (Pf12 linker sequence: SLENK [15]).
As only the second 6-Cys protein to be structurally characterized, Pf41 confirms that the overall architecture of the Plasmodium 6-Cys domains is reminiscent of the SRS fold from related Apicomplexa such as Toxoplasma gondii [4,[31][32][33]. However, one key difference is the placement of the sixth cysteine residue, which has important implications for the bioinformatic assignment of C-terminal domain boundaries: C6 of the SRS fold is located in the final β-strand resulting in a clear delinieation of the domain boundaries, whereas both C5 and C6 of the 6-Cys fold are located in the penultimate β-strand. Furthermore, there is no sequence motif defining the final inter-strand loop or β-strand of the 6-Cys fold [6], which has led to ambiguity in the definition of the C-terminal domain boundaries for members of the 6-Cys family.
Structural analysis reveals that the large sequence insertion in Pf41 is not an inter-domain spacer, but is inserted within D1 To further investigate the nature of the 120 residue sequence insertion in Pf41 (Thr117 to Val241) relative to Pf12, we first sought to establish whether the lack of electron density between Ser137 and Ile224 was due to flexibility within the crystal or due to the region being proteolytically processed. Analysis of the crystal packing revealed insufficient room in the crystal lattice to accommodate the un-modelled residues leading us to conclude that the region was proteolyzed during the extended time frame in which it took the crystals to form. While a significant portion of the Pf41 sequence insertion residues are un-modelled in the structure, it is clear that the labile region is inserted between the last two strands of the D1 domain: β8 and β9. Altogether, 108 residues of the Pf41 sequence insertion lie between D1 β8 and β9 (the remaining inserted residues complete β9 and extend through the short inter-domain linker), with the modelled residues extending from β8 into an alpha helix (α1": Thr117 to Phe129) followed by a region of coil (Leu130 to Ser137) ordered by crystal packing (Fig 1C and 1E). Based on the region observed in the crystal structure combined with secondary structure and disorder predictions [34,35], the region between β8 and β9 is expected to contain approximately 40% α-helical character, with the presence of the longest putative helix partially supported by previous circular dichroism analysis of a component peptide (Ala162 to Asp181, Fig 1E) showing predominant helical content [36]. The remaining 60% of this region, predominately mapping to the C-terminal portion, is predicted to be present as disordered, random coil (Fig 1E). Since much of the sequence comprising the D1 β8-β9 loop is predicted to be disordered, this region could be classified as intrinsically disordered [37,38]. However, such a designation will require validation by other methods such as nuclear magnetic resonance spectroscopy. Thus, we have termed the Pf41 β8-β9 loop (Thr117 to Ile225) an inserted domain-like region (ID).
Multiple sequence alignments of the Pf41 ID with other 6-Cys proteins anchored by the last Cys of D1 (C6 in β8) and the first Cys of D2 (C1' in β1') combined with secondary structure predictions clearly showed that the α-helical/coil composition of the ID is unique to Pf41 (data not shown). It is worth noting, however, that the related sexual stage 6-Cys protein Pfs47 has a predicted inter-domain region of approximately 100 residues with two cysteines and predominant β-structure that has the potential to be a well-ordered inserted domain in the analogous loop of D1. Altogether, our data reveal that the majority of the inserted sequence in Pf41 maps to an ID embedded in D1, and since this region appears to be unique to Pf41, we reasoned that its corresponding function is similarly unique. This led us to postulate a potential role for the Pf41 ID in coordinating Pf12.

ITC data reveal that the Pf41 ID is necessary to coordinate Pf12
To assess the contribution of the Pf41 ID to coordinating Pf12, we first used the ID boundaries defined from our structure to engineer a form of Pf41 where the ID was strategically replaced with a short glycine-serine linker (Pf41ΔID; Fig 3A). Following proteolytic removal of the purification hexahistidine/MBP tags, both recombinant full length Pf41 and Pf41ΔID eluted from the SEC column consistent with the expected molecular weights (Fig 3B) indicating that truncation of the ID did not alter protein folding. It is important to note that the exceptionally stable disulfide-pinned core of the 6-Cys/SRS domain is commonly accessorized with loops of different lengths and compositions [4,15,[31][32][33], supporting our observation that interchanging the ID sequence with a shortened loop does not disrupt the core 6-Cys structure. Moreover, we suspect that truncation of the ID does not appreciably increase inter-domain mobility, as we were able to obtain crystals of the tandem Pf41 6-Cys domains despite proteolytic removal of a large portion of the ID. Having established the quality of our full length and ID truncated Pf41, we next investigated the solution binding characteristics of these constructs with Pf12. Titration of Pf41 into the ITC cell containing Pf12 produced a dissociation constant (K d ) of 27.7 ± 3.7 nM (Fig 3C), which reflects approximately one order of magnitude tighter binding than previously obtained by surface plasmon resonance using rat CD4d3/4 fused Pf12 and Pf41 constructs [14]; the difference in measured affinity likely arises from a combination of the different techniques used and the presence/absence of a fusion partner. ITC data were fitted using a single-site binding model and showed a stoichiometry of approximately 1:1 (0.85 ± 0.01 Pf41: Pf12), consistent with a Pf12-Pf41 heterodimer shown previously by SEC co-elution volume and cross-linking of both recombinant and parasite-surface proteins [14,15]. The strongly negative binding enthalpy (ΔH of -33.9 ± 0.2 kcal/mol) indicates formation of numerous favorable interactions between full length Pf41 and Pf12, which compensate for a highly unfavorable entropy change (-TΔS of 23.6 kcal/mol). The thermodynamic parameters for the Pf12-Pf41 interaction are more extreme than typical protein-protein interactions and approach the values for interactions that require significant conformational changes [39], which may indicate that the interaction requires organization of the predicted disordered regions of the Pf41 ID and/or the shorter flexible loops on Pf12. In contrast to the tight interaction measured for the Pf12-Pf41 complex, no binding was observed between the ID truncated form of Pf41 and Pf12 indicating that the Pf41 ID is necessary for coordinating Pf12 (Fig 3C, bottom).
Having established that the absence of the Pf41 ID in the context of the tandem 6-Cys domains essentially eliminated binding to Pf12, we next sought to determine if the Pf41 ID was sufficient for Pf12 recognition. In support of this approach we engineered constructs where the ID was fused to either MBP or protein G B1 domain. We also engineered a form of the ID with cysteines at each end to better mimic the structural constraints observed in the crystal structure, where the N-and C-termini of the ID are separated by only 6.5 Å. Unfortunately, production of these fusion constructs in insect cells or E. coli did not yield a stable protein. To assess whether sufficient stability of the ID could be imparted by the inclusion of Pf41 D1, we engineered an MBP fusion construct extending from Lys21 through Arg236. While the incorporation of D1 markedly improved solubility and stability of the ID, removal of the MBP tag led to significant protein precipitation. Thus, it appears that both Pf41 D1 and D2 are necessary to stabilize the ID.
The Pf41 ID becomes protected from proteolysis in the Pf12-Pf41 heterodimer While it is evident that the Pf41 ID is necessary for binding Pf12, the underlying mechanism by which it promotes assembly is unclear. For example, the Pf41 ID could directly engage Pf12, it could serve an indirect role by optimizing the orientation of the Pf41 6-Cys domains, or it could promote heterodimer assembly through a combination of indirect and direct mechanisms. We reasoned that a direct role in binding would likely provide the labile ID with protection from proteolysis while a purely indirect function would maintain its proteolytic susceptibility. To test this, we carried out a series of trypsin protease protection assays (Fig 4).
Pf41 was approximately 50% degraded by trypsin after 20 minutes of incubation at 4°C; degradation bands were confirmed by mass spectrometry to be D2 and two forms of D1, the smaller of which lacks β9 and a portion of the ID (Fig 4, top). The observed degradation pattern fits with Pf41 having an exposed trypsin cleavage site immediately preceding the D1-D2 linker as well as several potential cleavage sites within the second half of the ID that is predicted to be largely void of stabilizing secondary structure elements (Fig 1E). Accessibility of the Pf41 linker is also consistent with the conformational flexibility of D1 and D2 enabled by similar linkers in the T. gondii SRS proteins [31,32]. Pf12 was marginally more stable than Pf41 with approximately 50% degraded after 40 minutes (Fig 4, middle). The two predominant Pf12 degradation products likely correspond to cleavage in the β4-β6 loop of D1 that was disordered in the Pf12 crystal structure [15], resulting in 10 kDa and 24 kDa products. The doublet for Pf12 that appears shortly after trypsin addition likely represents cleavage of the exposed and disordered C-terminal tail. In stark contrast, proteolysis of the Pf12-Pf41 heterodimer with trypsin was substantially reduced even after 120 minutes (Fig 4, bottom). These data support a model where the Pf41 ID becomes more tightly organized upon complex formation, likely through directly mediating complex formation with Pf12. Notably, a previous study using western blots of reduced protein samples extracted from schizont and merozoite stage parasites showed that only one Pf41 protein band was observed [14]. Thus, complex formation with Pf12 appears to protect the Pf41 ID from degradation both in an in vitro and a biological context.

Discussion
The 6-Cys s48/45 surface antigens are differentially presented on all life cycle stages of Plasmodium species suggesting important roles in enabling these parasites to interact with their environment. While a recent study provided intriguing insight into the role of Pfs47 in modulating the immune response to P. falciparum in the mosquito [12], detailed functional profiles of the human blood-stage merozoite expressed family members have been elusive. This is especially true for Pf12 and Pf41 where recent knockout studies showed no clear phenotype, although a complicating factor in interpreting these experiments may lie in the ability of parasites to adapt during the extensive time required for in vitro culturing [14]. Furthermore, a general lack of structural information describing both the individual 6-Cys proteins and the overall assembled Pf12-Pf41 heterodimeric complex has limited mechanistic insight.
The structural characterization of Pf12 reported in 2013 provided the first detailed architectural insight into tandem 6-Cys domains [15]. With respect to Pf41, however, an additional 120 residues predicted to be inserted as a spacer between the two 6-Cys domains have complicated Pf41 modeling efforts and led to ambiguity in defining its assembly with Pf12. Towards resolving this uncertainty, we first determined the structure of Pf41. Strikingly, structural Impact of Pf41 Domain Insertion on Pf12-Pf41 Assembly analysis revealed that the additional residues in Pf41 map to three distinct regions: a large inserted domain-like region (ID) of 108 residues (Thr117 to Ile225) between the β8 and β9 strands of the D1 core that is largely proteolyzed in the final structure, the D1 β9 strand, and the short inter-domain linker (Fig 1). The observation that the newly identified ID within D1 appears to be unique to Pf41 suggests a Pf41-specific function that we reasoned involved coordination of Pf12. To test this, we engineered a form of Pf41 where the ID was replaced by a short glycine-serine linker and showed by ITC that binding to Pf12 was abrogated (Fig 3). While these data showed the Pf41 ID is necessary to coordinate Pf12 with high affinity, we were unable to show sufficiency due to poor stability of the ID in the absence of both 6-Cys domains. We then used a protease protection assay to investigate the mechanism by which the ID promotes Pf12-Pf41 complex formation. These data showed that the proteolytic susceptibility of the ID was nearly eliminated in the context of Pf12 consistent with a direct role in coordinating Pf12 (Fig 4).
The structural and solution binding data reported here provided an opportunity to re-evaluate our previous Pf12-Pf41 cross-linking data [15]. Originally, cross-linked peptides were assigned to Pf41 D1, Linker, or D2 based on the previously predicted architecture of Pf41 with the spacer/linker region separating the two 6-Cys domains ( Fig 1A). However, with the unambiguous domain assignments enabled by our Pf41 structure (Fig 3A), we can now confidently reassign the ambiguous spacer peptides to the ID, β9 of D1, or the short inter-domain linker. Thus, with our reassigned cross-link designations we obtain a more accurate model of the Pf12-Pf41 interaction: Pf12 is in close proximity to the Pf41 ID as evidenced by ten crosslinks to this region, while cross-links between Pf12 and Pf41 D1 (three), D2 (five), or the C-terminal tail (three) were much fewer (Fig 5A), which is consistent with the ITC and protease protection data (Figs 3 and 4). While it is possible that the Pf41 ID solely contributes to the binding interface between Pf41 and Pf12, the cross-linking data combined with polymorphism analyses of Pf12 and Pf41 homologs are consistent with at least weak interactions between the 6-Cys domains of Pf41 and Pf12 to form an antiparallel interface [40,41]. Moreover, our trypsin protease protection assay indicates that the Pf41 inter-domain linker (RSNNNVI) is stabilized in the presence of Pf12, indicating a tightening of the Pf41 6-Cys domains upon complex formation with Pf12 (Fig 4). Based on these data, we propose a refined model of the Pf12-Pf41 heterodimer, where the Pf41 ID serves as the key structural bridge to anchor the complex (Fig 5B).
While the Pf41 ID is clearly a critical component of the Pf12-Pf41 complex assembly, the ID may endow Pf41 with multi-functional capabilities as its extended size appears to be substantially over-engineered for simply mediating the interaction with Pf12. This is particularly evident when analyzed in the context of the SRS29B and SRS16C proteins from T. gondii where the analogous SRS domains are sufficient for homodimerization [31,33]. Notably, however, these SRS proteins are tethered to the membrane via a GPI anchor thereby minimizing the entropic penalty that must be overcome compared to soluble Pf41 binding a membraneanchored protein. One possibility for an additional role for the Pf41 ID comes from a previous study that found that in parasites with Pf12 genetically deleted, a small amount of Pf41 was observed on the parasite surface suggesting that Pf41 is able to bind at least one other merozoite membrane-anchored molecule [14]. It is thus tempting to speculate that the proteolytic susceptibility and lack of predicted secondary structure elements within the ID, particularly in the second half, may indicate a level of flexibility that enables the ID to bind other molecules with variable induced fit depending on the partner [37]. Moreover, the binding of Pf41 to Pf12, or other parasite membrane proteins or even host derived partners, could be enhanced and/or enabled by post-translational modifications, such as the phosphorylation of Pf41 ID Ser137 [43] (Fig 5B, yellow arrow). Whether or not the Pf41 ID has functions additional to its role in Pf12 coordination, the importance of the Pf41 ID is supported by a previous study showing that antibodies in human immune sera specifically recognize this region [17].

Conclusions
Pf12 and Pf41 form a stable heterodimer on the surface of the infective Plasmodium merozoite, but the function of this complex remains elusive [14]. In the absence of a definitive biological role, however, the biophysical studies of Pf12 [15] and Pf41 (here) reveal important insight into the unique architectural features of the individual proteins and help unravel key mechanistic details underpinning assembly of the heterodimer. In particular, the identification and characterization of the ID in Pf41 as the crucial region enabling coordination to Pf12 has allowed us to refine the model of the Pf12-Pf41 complex on the parasite surface. As genetic tools continue to improve it will be important to revisit the function of the Pf12-Pf41 complex and assess what role, if any, the Pf41 ID plays in directly interfacing with the host.  [15]; Pf41 ID specific interactions are bolded and shaded light grey. (B) Refined model of Pf12-Pf41 complex generated by manual docking and guided by cross-linking data. Pf12 is shown as a magenta surface, and Pf41 as a green surface except a model of the ID generated in iTASSER [42] displayed as a grey semitransparent surface. The Pf41 ID model represents the relative size of this region compared to D1 and D2 and approximates how the ID could contact both Pf12 D1 and D2, but further studies probing the detailed structure and flexibility of this region are still needed. A previously identified Pf41 ID phosphorylation site [43] is shown in yellow and indicated by a yellow arrow. The black dotted line indicates uncertainty in the exact interface between the tandem 6-Cys domains of Pf12 and Pf41. doi:10.1371/journal.pone.0139407.g005