Structurally conserved erythrocyte-binding domain in Plasmodium provides a versatile scaffold for alternate receptor engagement

Significance

Plasmodium vivax is responsible for the most widely distributed recurring human malaria infections whereas Plasmodium falciparum inflicts the most mortality and morbidity in human populations. Malaria parasites enter our blood cells by making proteins that recognize and bind to their cognate receptors on the red blood cell surface. Our research describes, to our knowledge, the first crystal structure of PvRBP2a, an erythrocyte-binding protein from P. vivax, which revealed a structural scaffold similar to that of PfRh5, the essential erythrocyte-binding protein in P. falciparum. Structural comparisons between PvRBP2a and PfRh5 provide an important foundation toward understanding how P. vivax and P. falciparum parasites use a homologous erythrocyte-binding protein family to engage alternate erythrocyte receptors and ultimately govern host cell specificity.

Abstract

Understanding how malaria parasites gain entry into human red blood cells is essential for developing strategies to stop blood stage infection. Plasmodium vivax preferentially invades reticulocytes, which are immature red blood cells. The organism has two erythrocyte-binding protein families: namely, the Duffy-binding protein (PvDBP) and the reticulocyte-binding protein (PvRBP) families. Several members of the PvRBP family bind reticulocytes, specifically suggesting a role in mediating host cell selectivity of P. vivax. Here, we present, to our knowledge, the first high-resolution crystal structure of an erythrocyte-binding domain from PvRBP2a, solved at 2.12 Å resolution. The monomeric molecule consists of 10 α-helices and one short β-hairpin, and, although the structural fold is similar to that of PfRh5—the essential invasion ligand in Plasmodium falciparum—its surface properties are distinct and provide a possible mechanism for recognition of alternate receptors. Sequence alignments of the crystallized fragment of PvRBP2a with other PvRBPs highlight the conserved placement of disulfide bonds. PvRBP2a binds mature red blood cells through recognition of an erythrocyte receptor that is neuraminidase- and chymotrypsin-resistant but trypsin-sensitive. By examining the patterns of sequence diversity within field isolates, we have identified and mapped polymorphic residues to the PvRBP2a structure. Using mutagenesis, we have also defined the critical residues required for erythrocyte binding. Characterization of the structural features that govern functional erythrocyte binding for the PvRBP family provides a framework for generating new tools that block P. vivax blood stage infection.

The most widely distributed recurring malaria infections globally are caused by Plasmodium vivax, which accounts for 80–100 million malaria infections per year (1). The majority of clinical symptoms associated with malaria are due to blood stage infection (2). The merozoite forms of malaria parasites invade human erythrocytes through a multistep process that involves initial contact with the red blood cell, apical reorientation of the merozoite, and the formation of a tight junction that moves progressively toward the posterior end of the parasite until host cell membrane fusion is completed. These steps in invasion are dependent on specific interactions between parasite adhesins and their cognate erythrocyte receptors (reviewed in ref. 3).

P. vivax preferentially invades reticulocytes: i.e., immature red blood cells (4). The basis of host cell selectivity by merozoites from Plasmodium spp. seems to be mediated primarily by families of adhesin proteins. The two erythrocyte-binding protein families of P. vivax are called the Duffy-binding protein (PvDBP) and reticulocyte-binding protein (PvRBP) families (5). In laboratory-adapted P. vivax strains, there is only one PvDBP protein in P. vivax that binds to Duffy antigen receptor for chemokines (DARC) (6, 7). De novo assembly of a Cambodian field isolate genome identified a potential new member called P. vivax erythrocyte-binding protein (PvEBP), which shares a Duffy binding-like domain and the C terminus cysteine-rich domain present in PvDBP (8). DARC is strongly implicated in malaria invasion because Duffy-negative individuals are resistant to P. vivax infection and antibodies that interfere with PvDBP–DARC interaction inhibit P. vivax invasion (9). For many decades, the leading paradigm was that P. vivax invaded only Duffy-positive erythrocytes, leading to the hypothesis of the existence of only one functional P. vivax invasion pathway, in contrast to the multiple pathways involved in Plasmodium knowlesi and Plasmodium falciparum invasion. This concept has been overturned by mounting evidence that P. vivax infection is observed in individuals that were genetically and phenotypically Duffy-negative in diverse geographical regions (reviewed in ref. 10).

Upon sequencing of the P. vivax genome from the Salvador 1 strain in 2008, 10 proteins were identified as belonging to the PvRBP family of proteins, comprising three partial genes and seven full-length genes, of which two are predicted pseudogenes (5). The partial genes are PvRBP1-P, PvRBP2-P1, and PvRBP2-P2; the five full-length coding genes are PvRBP1a, PvRBP1b, PvRBP2a, PvRBP2b, and PvRBP2c; and the two pseudogenes are PvRBP2d and PvRBP3. Sequencing of a field isolate genome also identified PvRBP2e, which seems to be absent from Salvador 1 but present in most P. vivax strains (8). Because DARC is present on both reticulocytes and normocytes, the binding of PvDBP to DARC is not sufficient for reticulocyte specificity. Restricted host cell selectivity of P. vivax is thought to be governed by the PvRBP protein family, of which it has been previously shown that PvRBP1a and PvRBP2c bind specifically to reticulocytes (11). The molecular function of the rest of the PvRBP family in P. vivax invasion is currently unknown. Furthermore, no known erythrocyte receptor has been identified that interacts with any PvRBP family members.

Through gene structure and sequence similarity, the homologs of PvRBPs were identified in P. falciparum and Plasmodium yoelii as the PfRh family and Py235 family, respectively (12, 13). PfRh5, a member of the PfRh family of P. falciparum, has been a focus of intense research as a leading blood stage vaccine candidate due to its essential function in parasite invasion and low levels of sequence polymorphisms (reviewed in ref. 14). PfRh5 binds erythrocytes through the recognition of the Ok blood group antigen, basigin (15). This receptor-adhesin pair mediates an essential molecular event during parasite entry, and sequence polymorphisms within PfRh5 and/or basigin have been shown to modulate host tropism (16, 17). The addition of antibodies against PfRh5, basigin, or PfRh5-interacting partners (PfRipr and PfCyRPA) in P. falciparum growth assays results in the strong inhibition of parasite growth across multiple strains (15, 18–21). Further promising results show that Aotus nancymaae monkeys immunized with the anti-PfRh5 vaccine are protected against severe infection (22). The crystal structure of PfRh5 alone or with its receptor basigin provides, to our knowledge, the first structural description of how PfRh proteins engage with their receptors (23, 24). PfRh5 adopts a fold comprising an α-helical scaffold that provides binding sites at the tips of helices for basigin and some inhibitory monoclonal antibodies (24).

To date, no experimentally determined molecular structure for any PvRBPs has been reported. In this paper, we present, to our knowledge, the first crystal structure of an erythrocyte-binding domain within PvRBP2a that shows structural similarity to the erythrocyte-binding domain of the essential P. falciparum invasion ligand, PfRh5. Our analyses of its erythrocyte-binding characteristics and sequence diversity patterns from field isolates provide, to our knowledge, the first structure-function analyses of this important family of P. vivax erythrocyte-binding proteins.

Results

Gene Structure and Expression of PvRBP2a.

Bioinformatic analyses suggest that full-length pvrbp genes contain two exons that are separated by an intron located immediately downstream of the signal peptide-coding sequence (11). To verify this putative gene structure for PvRBP2a, we prepared cDNA from a Thai patient infected with P. vivax and used it as the PCR template. Splicing was detected when primers flanking the putative intron were used (Fig. 1A). Sequencing of the gene fragments indicated splicing between nucleotides position 55–235 of pvrbp2a (Fig. 1B). No additional intron was found within the first 4,000 base pairs of the gene by cDNA amplification and sequencing.

Fig. 1.

Gene structure and expression of PvRBP2a. (A) Amplification of pvrbp2a confirms splicing near the start codon. Primers flanking the putative intron were used to amplify gDNA and cDNA, yielding amplicons consistent with the 356- and 177-base pair product lengths predicted by the gene structure in B. The no-reverse-transcriptase (−RT) control detected no gDNA contamination in cDNA. The molecular weight for the DNA standard is highlighted in base pair (bp). (B) Schematic of the gene structure for pvrbp2a. The pvrbp2a gene consists of two exons and one intron, with the coding sequence and the noncoding intron shown in capital and small italicized letters, respectively. ATG highlights the start methionine in the 5′ region whereas TAA represents the stop codon on the 3′ end. (C) Western blots of P. falciparum and P. vivax protein extracts probed with anti-PvRBP2a and anti-EBA175 antibodies. Molecular mass marker is shown on the left hand side in kDa.

Full-length PvRBP2a is predicted to consist of 2,487 amino acid residues with a molecular weight of 286 kDa. To determine whether PvRBP2a is expressed in P. vivax, we generated specific anti-PvRBP2a polyclonal antibodies in rabbits. Using Western blotting, we probed both P. falciparum and P. vivax protein extracts with anti-PvRBP2a antibodies. We detected a doublet running above 250 kDa only in the P. vivax extracts (Fig. 1C). Anti-EBA175 antibodies were used as a positive loading control for P. falciparum lysate.

The Crystal Structure of PvRBP2a.

Full-length PvRBP2a includes a signal peptide at the N terminus and a single transmembrane domain at the C terminus, with no other predicted structural domains. Most secondary structure predictions highlighted a disordered region within the first 160 amino acids of PvRBPs. Therefore, for our initial structural studies, we used a construct of PvRBP2a encompassing amino acid residues 160–1135 (PvRBP2a_160–1135). The protein was expressed in Escherichia coli and purified using Ni²⁺-affinity chromatography, followed by tag removal using tobacco etch virus (TEV) protease and size exclusion chromatography. Protein crystals were obtained by in situ proteolysis with porcine elastase. The structure of PvRBP2a was determined through the single isomorphous replacement with anomalous scattering (SIRAS) method using iodine-derivatized crystals, with subsequent refinement against the native diffraction data to a resolution of 2.12 Å. Crystallographic data collection and refinement statistics are presented in Table 1.

Table 1.

Crystallographic data collection and refinement statistics

Protein	PvRBP2a
Data collection
Space group	P212121
Cell parameters
a, b, c, Å	58.79, 93.45, 126.72
Resolution, Å	49.76–2.12 (2.19–2.12)
Rmerge	0.134 (1.916)
〈I/σI〉	13.3 (1.8)
Completeness, %	100.00 (100.00)
Multiplicity	14.3 (14.1)
CC1/2*	0.999 (0.538)
Refinement statistics
Resolution, Å	43.84–2.12 (2.24–2.12)
Observed reflections	1,249,569
Unique reflections	40,139
Rwork	0.203 (0.286)
Rfree	0.229 (0.347)
No. of atoms	4,945
Protein	4,739
Ligand	11
Solvent	195
Average B-factors, Å2	56.8
Protein	57.0
Ligand	77.1
Solvent	51.1
rmsd
Bond lengths, Å	0.004
Bond angles, °	0.709
Ramachandran plot, %
Most favored	97.6
Allowed	2.4
Outlier	0

Statistics for the highest-resolution shell are shown in parentheses. Data were collected on a single crystal.

^*CC_1/2, Pearson correlation coefficient between independently merged halves of the dataset.

The crystal structure of PvRBP2a reveals two molecules in the asymmetric unit (Fig. S1A). Both molecules are almost identical, with rmsd of atomic positions between two molecules 0.6 Å over 281 aligned atoms Cα (Fig. S1B). As such, the following discussion will focus solely on molecule A, which we will refer to as PvRBP2a_160–455. The structure of PvRBP2a_160–455 consists of 10 α-helices surrounding one short β-hairpin (Fig. 2A). We denote these helices as α1, α2a, α2b, α2c, α3a, α3b, α4, α5, α6, and α7, where numerical order indicates progression of the polypeptide chain from N to C terminus. Helices α2 and α3 are divided into smaller fragments due to the presence of the breaks into the canonical (i, i + 4) hydrogen bonding pattern. Within α2, helices α2a and α2b are divided, with a bend introduced by residue P229, whereas a short helix α2c is separated from the rest at the C-terminal extremity. Helix α3 is divided into two equal parts by a β-turn formed by amino acids Q273 to M276.

Fig. S1.

Asymmetric unit content analysis. (A) An overall view of the asymmetric unit. Molecule A includes 298 residues spanning 160–455 amino acids of PvRBP2a and a di-peptide GS fragment that was introduced as a cloning artifact. Molecule B spans residues 160–450 but is less well-defined, in which residues 392–398 could not be convincingly modeled, yielding a structure with a total of 284 residues. The buried interface area between two molecules present in the ASU is around 1,340 Ų as calculated using the PISA server. Molecules A and B are colored in blue and green, respectively, and are shown as ribbon and surface representation. (B) Superimposition of the two molecules present in the asymmetric unit. Molecules are shown as ribbons colored with the colors corresponding to the rmsd between two molecules. Dark blue shows good alignment, and higher deviations are in orange/yellow/red. Residues not used for alignment are colored white.

Fig. 2.

Crystal structure of the PvRBP2a erythrocyte-binding domain. (A) Three orthogonal views of the molecule. The overall fold of the protein is formed by 10 antiparallel α-helices and one short β-hairpin. Molecule is shown in ribbon representation colored in rainbow from blue at the N terminus until red at the C terminus. The cysteine residues forming disulphide bridges are shown as yellow sticks. (B) Sample of the electron density around the residue F181 forming a γ-turn. The 2F_obs − F_calc map is contoured at 2σ and shown as blue mesh. Protein is shown in ball and stick representation with carbon shown in green, nitrogen in blue, oxygen in red, and sulfur in yellow.

The molecule adopts a flat, ellipsoidal shape, with the helices arranged parallel to the long axis. It consists of two subdomains of similar length, the N- and C-terminal, that are related to each other by a pseudo-twofold rotation symmetry (Fig. S2 A and B). The N-terminal subdomain includes helices α1, α2a, α2b, α2c, α3a, and α3b whereas the C-terminal subdomain is composed of helices α4, α5, α6, and α7. The N-terminal subdomain begins with a very short helix 3₁₀ forming one full turn, followed by a small β-hairpin, helix α1, and two long helices α2 and α3. The β-hairpin localized in the center of the molecule consists of two short anti-parallel β-strands (β1 and β2) that are connected by a beta bulge (Fig. 2A). The main axis of the β-hairpin is oriented perpendicularly to the plane formed by the long helices and protrudes slightly out on both sides. It may contribute to the overall shape of the molecule either by keeping helices α2 and α3 apart from the C-terminal subdomain due to the stacking interactions or by forming an extensive network of hydrogen bonds, with the rest of the molecule thus holding two subdomains together (Fig. S2C). The connection between strand β2 and helix α1 includes a γ-turn, with F181 in its center (Fig. 2B). The C-terminal subdomain exhibits simpler organization and consists of four relatively long α-helices forming a coiled-coil domain (Fig. 2A).

Fig. S2.

Detailed analysis of the PvRBP2a structure. (A) The molecule is schematically divided into two subdomains of similar size that are related to each other through pseudo-twofold rotation symmetry. The alpha-helices are represented as cylinders and labeled. The N-terminal domain is colored in blue, and the C-terminal in red. (B) B-factors analysis. Molecule A is drawn in schematic putty representation. The tip of the molecule, including a disulphide bond formed between C299 and C303, is relatively flexible compared with the rest of the molecule. The protein regions with high temperature factors are shown as wide orange/red tubes. (C) Enlargement of the N-terminal part of the molecule. This fragment is in contact with both subdomains, forming an extensive network of hydrogen bond interactions. Several amino acid residues are also involved in the stacking interactions: like F181 with I224 and F167 with K269, for example. Interacting amino acid residues are shown in sticks and labeled. Residues belonging to the N-terminal part of the protein are underlined.

There are two disulfide bridges present within the PvRBP2a_160–455 structure (Fig. 2A). Cysteine residues 227 and 271 form a covalent bond, connecting helices α2 and α3, that is localized in the middle of the N-terminal subdomain in the close proximity of the kinks present in the long helices. The second disulfide bridge formed between cysteine residues 299 and 303 and is located on the tip of the molecule in the loop connecting helices α3b and α4 at the border of the N- and C-terminal subdomains. By holding different helices or subdomains together, the covalent bonds between those residues may play an important structural role in stabilizing an overall fold of the protein.

The asymmetric unit of the crystal lattice contains two molecules of PvRBP2a in the extensive contact with the buried surface area between them of 1,340 Ų (Fig. S1A). To address whether this dimeric form was representative of the biological unit, we analyzed the oligomeric state of PvRBP2a using analytical ultracentrifugation (AUC). We expressed a construct that encompassed residues 160–1000 (PvRBP2a_160–1000). Sedimentation velocity experiments indicated that the protein is exclusively monomeric, with the sedimentation coefficient of 3.5 S corresponding to an apparent molecular mass of 93.6 kDa (Fig. 3A), which is consistent with the theoretical molecular mass of 98.4 kDa. The frictional coefficient obtained from AUC data analyses for all three protein concentrations was 2.1, which is characteristic for protein molecules with an elongated shape.

Fig. 3.

AUC and SAXS data analysis for PvRBP2a. (A) Sedimentation velocity analytical ultracentrifugation for PvRBP2a_160–1000 at 0.4, 1.0, and 1.7 mg/mL (green, red, and black, respectively) is consistent with a monomeric protein. The measured sedimentation coefficient 3.5 S corresponds to an apparent molecular mass of 93.6 kDa with the theoretical calculated molecular mass as 98.4 kDa. No formation of higher order oligomers was detected. (B) Experimental scattering profile (black squares) overlaid with the calculated scattering pattern (red line) of a representative ab initio model generated using the program DAMMIF. The data are presented as the natural logarithm of the intensity vs. q. (Inset) The Guinier plot was linear, which is consistent with an absence of detectable aggregates in the sample. (C) The interatomic distance distribution function, P(r), of PvRBP2a_160–1000. The curve was calculated by indirect Fourier transform using the program GNOM. (D) Kratky plot analysis of the SAXS data, suggesting that the protein molecule is fairly rigid. (E) Two orthogonal views of the crystal structure of PvRBP2a_160–455 docked into an averaged and filtered ab initio SAXS envelope of PvRBP2a_160–1000 derived from 20 independent DAMMIF calculations. The PvRBP2a_160–1000 construct adopts an elongated boomerang-like shape. The crystal structure was fitted into one of its extremities by rigid-body docking. The fragment of PvRBP2a not present in the crystal structure forms a long tail after the C terminus.

Although we were able to crystallize the N-terminal domain of PvRBP2a, the overall topology of the larger fragment remained of broad interest, especially considering that the C-terminal domains in the 160–1000 fragment have a presently unknown structure and are nonhomologous to known domain structures. To this end, we collected small angle scattering (SAXS) data for PvRBP2a_160–1000 at the Australian Synchrotron (Fig. 3 B–E and Table 2). Analysis of the Guinier region indicated that the protein was monodisperse and that data were not affected by aggregation or interparticle repulsion (Fig. 3 B, Inset). The real space interatomic distance distribution [P(r)] plot was consistent with PvRBP2a existing in a highly elongated form (Fig. 3C) (D_max = 300 Å). The analysis of the Kratky plot confirmed that the protein molecule is fairly rigid (Fig. 3D). We performed ab initio bead modeling with DAMMIF, revealing PvRBP2a to exist in an elongated, boomerang-like shape. The crystal structure of PvRBP2a_160–455 was docked into the obtained SAXS envelope by rigid body modeling to illustrate the relative size and position of the erythrocyte-binding domain in the context of the longer fragment of PvRBP2a (Fig. 3E).

Table 2.

SAXS data collection and analysis statistics

Protein	PvRBP2a 160–1000
Data collection parameters
Instrument	Australian Synchrotron SAXS/WAXS beamline
Beam geometry	120-μm point source
Wavelength, Å	1.033
q range, Å-1*	0.00369–0.22821
Exposure time	2-s exposures
Protein concentration	8.0 mg/mL protein via inline gel filtration chromatography
Temperature, K	289
Structural parameters
I(0), cm−1 [from P(r)]	0.089 ± 0.001
Rg, Å [from P(r)]	91.06 ± 0.91
Dmax, Å	300
I(0), cm−1 (from Guinier)	0.087 ± 0.001
Rg, Å (from Guinier)	83.70 ± 1.53
Software used
Primary data reduction	Scatterbrain (Australian Synchrotron)
Data processing	PRIMUS, GNOM
Rigid-body modeling	COLORES
Computation of model intensities	CRYSOL
Porod–Debye analysis	ScÅtter
3D graphics representations	Sculptor

^*q is the magnitude of the scattering vector, which is related to the scattering angle (2θ) and the wavelength (λ) as follows: q = (4π/λ)sinθ.

Structure Comparison Between PvRBP2a and PfRh5.

The crystal structure of PfRh5, a member of the homologous PfRh family, has been recently solved alone as well as in complex with its cognate receptor basigin or with inhibitory monoclonal antibodies (23, 24). Crystal structures of PfRh5 and PvRBP2a_160–455 can be superimposed with rmsd of atomic positions 3.4 Å over 217 aligned atoms Cα (Fig. 4A). However, the general architecture of the PvRBP2a_160–455 molecule seemed similar to that of PfRh5. Structural alignment of both proteins shows that the two disulfide bridges superimpose very well (Fig. 4A). The bonded pair of cysteine residues C227 and C271 in PvRBP2a_160–455 corresponds to the bond present between C224 and C317 in PfRh5 whereas the second connection formed between C299 and C303 overlaps with a bond between C345 and C351.

Fig. 4.

Structural comparison between PvRBP2a and PfRh5. (A) Superimposition of PvRBP2a (magenta) and PfRh5 (blue) structures in ribbon representation. The molecules can be aligned with the calculated rmsd of atomic positions 3.4 Å. The cysteine residues forming disulfide bridges are shown as yellow sticks. The PDB ID code for PfRh5 is 4WAT. (B) Comparison of the electrostatic potential surfaces of two orthogonal views of PfRh5 (Upper) and PvRBP2a (Lower). The negatively charged area is clearly visible in the upper part of the PvRBP2a molecule. Electrostatic surface potentials were calculated using the program APBS with the nonlinear Poisson–Boltzmann equation and contoured at ±5 kT/e. Negatively and positively charged surface areas are colored red and blue, respectively. (C) Basigin-binding site in PfRh5 is superimposed with the corresponding region in PvRBP2a. Residues forming the negatively charged area in PvRBP2a are localized in helix α4 and include E304, D306, E309, E313, E317, and E321. This glutamate-rich region overlaps with the basigin-binding site in PfRh5 (shown as a white surface representation, with the residues interacting with basigin highlighted in teal). The PvRBP2a molecule is shown as ribbons colored in magenta, and amino acid residues bearing negative charge are shown as yellow/red sticks.

A striking difference between PfRh5 and PvRBP2a_160–455 concerns the distribution of the charged amino acid residues exposed on the surface (Fig. 4B), particularly within the region in PfRh5 involved in binding basigin (Fig. 4C). The corresponding area within PvRBP2a has a predominantly negative electrostatic potential due to several acidic amino acid residues localized in helix α4 and the preceding loop: e.g., E304, D306, E309, E313, E317, and E321 (Fig. 4 B and C). This negatively charged area forms a distinctive patch on the surface of the protein that may be a putative receptor-binding site.

Detailed comparisons of PfRh5 and PvRBP2a_160–455 revealed several differences in the placement of the secondary structure elements. The orientation of the β-hairpin within the N terminus is rotated by 180 degrees between both proteins (Fig. S3A). The different arrangement of this fragment of the PvRBP2a molecule is related to the presence of an additional short loop, including a γ-turn that connects strand β2 with helix α1. Moreover, helices α2a and α3b in PvRBP2a_160–455 are shifted by one helix breadth compared with equivalent helices in PfRh5 (Fig. S3B). Another structural difference concerns the presence of a long loop in PfRh5, including residues S257 to D294. This region of the molecule is disordered and not visible in the electron density (23). The corresponding sequence in PvRBP2a forms only a very short helix α2c and consists of residues P244 to H247 (Fig. S3C).

Fig. S3.

Structural differences between PvRBP2a (magenta) and PfRh5 (blue). The PDB accession code for PfRh5 is 4WAT. (A) Close view of the N-terminal part of the molecule. The β-hairpin is rotated by 180 degrees between two structures. The N termini of both molecules are labeled. The loop connecting strand β2 and helix α1 in PvRBP2a includes a classic γ-turn with F181 localized in its center. The loop in PvRBP2a is colored in yellow, with F181 shown as sticks. (B) Helix α3 in PvRBP2a is divided into two parts by a β-turn formed by amino acids Q273 to M276. Additionally, the helix α2a is shifted compared with the corresponding helix in PfRh5. The β-turn formed by amino acids Q273 to M276 is colored in yellow. The cysteine residues forming an overlapping disulphide bridge are shown as yellow sticks. (C) The disordered loop in PfRh5, including residues S257 to D294 and represented schematically as a dashed line. A corresponding place in PvRBP2a forms only a very short helix α2c and consists of residues P244 to H247.

Sequence Alignment of the PvRBPs Family.

Sequence alignment of PvRBP2a_160–455 with the six other PvRBP members highlighted a common domain that features conserved placement of one or two disulfide bridges (Fig. 5, Left) and is localized close to the first 500 amino acids of the N terminus of each molecule. The majority of the conserved amino acid residues within this aligned region are hydrophobic in nature and may play a structural role in maintaining the proper architecture of the domain (Fig. 5, Right, pink residues). On the other hand, the residues on the solvent-exposed surface are highly variable. The region of highest divergence between PvRBP subgroup 1 (PvRBP1a and 1b) and subgroup 2 (PvRBP2a, -2b, and -2c, PvRBP-P1, and PvRBP-P2) is localized at the N terminus surrounding the β-hairpin, where the sequence alignment contained several gaps.

Fig. 5.

Sequence alignment of PvRBP family. (Left) Alignment of the sequence corresponding to the crystallized fragment of PvRBP2a with sequences of six other members of the PvRBP family highlights a conserved domain. The secondary structure elements of PvRBP2a are shown above the alignment. The bonded pairs of cysteine residues are highlighted in yellow and marked with green numbers. (Right) Two orthogonal views of the PvRBP2a structure, with conserved amino acid residues highlighted as magenta sticks. The conserved residues have a mostly hydrophobic character and are involved in packing of the helical core.

The position of cysteine residues is conserved for almost all members of subgroup 2, with the exception of PvRBP2c, which is missing the second pair. In contrast, for subgroup 1, the first pair of disulfide bonds is not present. The most variable region between all PvRBP proteins includes the loop connecting helices α3b and α4, which, in PvRBP2a, includes one of the disulfide bridges formed between C299 and C303. This localization of the loop in PvRBP2a overlaps with the basigin-binding site in PfRh5 (Fig. 4C).

Sequence Diversity and Evolution of the Genes Encoding PvRBP2a.

High quality consensus sequences for the entire 7,464-bp pvrbp2a coding locus were obtained from 22 P. vivax clinical isolates from Papua New Guinea (Dataset S1). These sequences were aligned together with published data for five reference strains (25) and four field isolates from Thailand (26). A total of 105 single nucleotide polymorphisms were identified that encoded 81 polymorphic amino acids. All 31 P. vivax isolates surveyed had unique amino acid haplotypes (Fig. S4), demonstrating very high diversity of PvRBP2a. The highest nucleotide diversity was concentrated in the first 2.2 kb (Fig. 6A). This region also contained evidence of significant balancing selection as measured by the Tajima’s D statistic, consistent with this region being an important target of host immune responses (Fig. 6B). Of the 31 polymorphic sites with minor allele frequencies more than 10%, 10 mapped to the corresponding region of the crystal structure at residues N186S, E277K, K285N/I, K289E, E304K, D306V, E351Q, P399S, K421M, and G438E (Fig. 6C). Of these polymorphisms, three polymorphic residues (E304K, D306V, and P399S) are within the negatively charged region that may function as a putative erythrocyte-binding region. Comparing these sites to polymorphism data publically available in PlasmoDB, 7 of the 10 residues were also found commonly among isolates from other malaria endemic areas (N186S, E277K, K285N/I, K289E, E304K, D306V, and P399S).

Fig. S4.

Plasmodium vivax RBP2a amino acid haplotypes and allele frequencies. High quality consensus sequences for the entire 7,464-bp pvrbp2a coding locus were obtained from 22 P. vivax clinical isolates from Papua New Guinea, and these sequences were aligned together with published data for five reference strains and four field isolates from Thailand. A total of 105 single nucleotide polymorphisms were identified, which encoded 81 polymorphic amino acids. All 31 P. vivax isolates surveyed had unique amino acid haplotypes demonstrating very high diversity of PvRBP2a. Of the 31 polymorphic sites with minor allele frequencies of more than 10%, 10 mapped to the corresponding region of the crystal structure at residues N186S, E277K, K285N/I, K289E, E304K, D306V, E351Q, P399S, K421M, and G438E.

Fig. 6.

Polymorphism and natural selection on the genes encoding PvRBP2a. (A) Sliding window analysis showing nucleotide diversity (π) values in pvrbp2a for the 31 sequences analyzed. A window size of 100 bp and a step size of 3 bp were used. (B) Sliding window calculation of Tajima’s D statistic was performed for the PNG population (n = 22). A window size of 100 and a step size of 3 were used. The gray box in A and B refers to a highly polymorphic region within pvrbp2a that is under balancing selection. (C) Three views of the surface representation of the crystallized fragment of PvRBP2a, with polymorphic residues highlighted in green. The polymorphic residues cluster around three main sites. The first one is localized on the tip of the molecule in the region overlapping with the basigin-binding region in PfRh5 and includes residues E304K, D306V, and P399S. The second region is localized on the side of the molecule within helix α3b and includes E277K, K285N/I, and K289E. The third region is localized in the cleft between helices α1 and α7 and includes N186S and K421M. There are two isolated single polymorphic sites, including E351Q and G438E.

Erythrocyte-Binding Characteristics of PvRBP2a.

We used a flow cytometry-based erythrocyte-binding assay to determine the binding characteristics of PvRBP2a (27). This flow cytometry method entailed incubating recombinant PvRBP2a_160–1000 (or its variants) with red blood cells, and binding was detected using an anti-PvRBP2a rabbit polyclonal antibody, followed by an anti-rabbit secondary antibody conjugated to a fluorophore. We initially performed the binding studies using an enriched population of reticulocytes that were stained by Thiazole Orange (TO⁺). CD71, the transferrin receptor, is present on immature reticulocytes and is progressively depleted as reticulocytes mature into normocytes. It was used as an additional surface marker for different stages of reticulocytes (Fig. 7 A and D). We observed that PvRBP2a_160–1000 bound equally to the mature erythrocytes (TO⁻) and reticulocyte populations (TO⁺) (Fig. 7 B and D). This binding property was similar to PfRh4 binding profile which cognate receptor complement receptor 1 is present on both mature erythrocytes and reticulocytes (Fig. 7 C and D). Because PvRBP2a_160–1000 binding was not reticulocyte-specific, we performed all subsequent binding assays using mature erythrocytes. This protein fragment bound to erythrocytes treated with neuraminidase and chymotrypsin, but not when treated with trypsin (Fig. 7E). This binding property represents an enzyme profile different from PfRh4 binding, which is neuraminidase-resistant but trypsin- and chymotrypsin-sensitive (28).

Fig. 7.

PvRBP2a erythrocyte-binding characteristics. (A) Dot plots showing the enriched reticulocyte purification stained with both TO and CD71-PECy5 (Upper) or CD71-PECy5 alone (Lower). (B) Dot plots showing the binding of PvRBP2a to red blood cells (Upper). Binding was detected using an anti-PvRBP2a rabbit IgG antibody, followed by a secondary anti-rabbit Alexa 647 antibody. (Lower) A binding control where no protein was added before incubation with primary and secondary antibodies. (C) Dot plots showing the binding of PfRh4 to red blood cells (Upper). Binding was detected using an anti-PfRh4 rabbit IgG antibody, followed by a secondary anti-rabbit Alexa 647 antibody. (Lower) A binding control where no protein was added before incubation with primary and secondary antibodies. (D) Bar charts showing the percentage of binding of CD71-PECy5, PvRBP2a, and PfRh4 to mature erythrocyte (TO⁻) vs. reticulocyte (TO⁺) populations. Error bars represent SEM of three independent repeats. (E) PvRBP2a binding to erythrocytes is neuraminidase- and chymotrypsin-resistant but trypsin-sensitive whereas PfRh4 binding is unperturbed by neuraminidase treatment. A flow cytometry-based red blood cell-binding assay was performed using untreated erythrocytes (Un), neuraminidase (Nm), low trypsin (LT), high trypsin (HT), and chymotrypsin-treated erythrocytes (CHY). Low trypsin and high trypsin refer to treatments with 0.1 and 1.5 mg/mL enzyme, respectively.

Due to the striking structural similarity with PfRh5, we hypothesized that the putative erythrocyte-binding domain within PvRBP2a resides within PvRBP2a_160–455. Unfortunately, we were unsuccessful in purifying PvRBP2a_160–455 as a recombinant fragment in E. coli using both native and refolding methods. Alternatively, we expressed a construct that contained a deletion of amino acids 160–460 within PvRBP2a_160–1000, which we will refer to as PvRBP2a_Δ160–460 (Fig. 8 A–C)_. PvRBP2a_Δ160–460 did not bind red blood cells (Fig. 8D), showing that amino acid 160–460 is the erythrocyte-binding domain of PvRBP2a.

Fig. 8.

Identification of amino acid residues in PvRBP2a important for erythrocyte binding. (A) Far UV CD spectrum for the WT PvRBP2a_160–1000 (WT) compared with the spectra of various structural mutants, including PvRBP2a_E/DmutK, PvRBP2a_{C299S, C303S}, PvRBP2a_Δ160–460, and PvRBP2a_{P180A, F181A, Y182A}, as labeled. Only the last triple-alanine mutant exhibits a different CD spectrum compared with the WT protein. (B) Far UV CD spectrum for the WT PvRBP2a_160–1000 (WT) and various SNP mutants, including PvRBP2a_{E277K, K285N, K289E}, PvRBP2a_{E304K, D306V}, PvRBP2a_{E304K, D306V, P399S}, PvRBP2a_G438E, PvRBP2a_E351Q, PvRBP2a_{N186S, K421M}, and PvRBP2a_{E277K, K285I, K289E}, as labeled. All presented spectra superimpose very well, suggesting that the introduced mutations did not alter the proper folding of the protein. (C) SDS/PAGE gel of purified PvRBP2a constructs loaded, as labeled. Two micrograms of each construct were loaded onto a NuPAGE gradient gel (4–12%) under reducing conditions and stained with Coomassie Brilliant Blue. Molecular mass marker indicated in kDa. (D) Deletion and mutational analyses of the erythrocyte-binding domain within PvRBP2a. Recombinant PvRBP2a proteins were tested for their capability to bind erythrocytes in a flow cytometry-based assay. % RBC binding, the percentage of erythrocytes with bound PvRBP2a protein determined by normalizing the number of erythrocytes exhibiting a positive Alexa Fluor 488 signal that is above the background (which is the Alexa Fluor 488 signal of erythrocytes without protein added) on the total number of erythrocytes.

In this study, we sought to identify amino acid residues within PvRBP2a that are important for erythrocyte binding. We designed and expressed a variety of individual or clusters of mutations within PvRBP2a_160–1000 based on the structural comparisons with PfRh5 and the identified field polymorphisms (Fig. 8 A–C). In most cases, these mutations were named after their respective positions within PvRBP2a, preceded by the original amino acid residue followed by the mutant amino acid residue: e.g., in PvRBP2a_E351Q, PvRBP2a was mutated in residue 351 from glutamic acid to glutamine. To ensure that these protein variants were well-folded, we performed circular dichroism spectroscopy and observed that only PvRBP2a_{P180A, F181A, Y182A} (which has three mutations within the γ-turn) showed different secondary structure features from the WT protein PvRBP2a_160–1000 (Fig. 8A). As expected, PvRBP2a_{P180A, F181A, Y182A} did not bind erythrocytes (Fig. 8D).

Comparison of the PvRBP2a and PfRh5 structures highlighted differences in surface properties within the PfRh5-basigin–binding region (24). The corresponding area within PvRBP2a has predominantly a negative electrostatic potential due to the presence of several acidic amino acid residues localized in helix α4 and the preceding loop: e.g., E304, D306, E309, E313, E317, and E321 (Fig. 4C). We mutated the six residues mentioned above to lysine (positive charge) and expressed the protein PvRBP2a_E/DmutK (Fig. 8 A–C). This protein was not able to bind erythrocytes, suggesting that the negative charge patch on the PvRBP2a surface is important for modulating its ability to bind erythrocytes (Fig. 8D).

Structural alignment of PfRh5 and PvRBP2a proteins shows that both disulfide bridges superimpose very well. Reduced and alkylated PfRh5 results in a threefold reduction in binding to basigin (23). In PvRBP2a_160–455, there are two disulfide bonds at C227 and C271 and between C299 and C303. We proceeded to mutate C277 and C271 or C299 and C303 as pairs into serine residues to prevent disulfide bond formation. Mutation of C277S and C271S resulted in an expressed protein that eluted in the void volume of size exclusion chromatography, indicative of misfolded proteins. We were successful, however, in generating a double mutant of C299S and C303S, which we will refer to as PvRBP2a_{C299S, C303S} (Fig. 8 A and C). This mutant did not bind erythrocytes, highlighting the importance of the disulfide loop connecting helices α3b and α4 in PvRBP2a function (Fig. 8D).

We also expressed variant forms of PvRBP2a_160–1000 with mutated clusters of existing polymorphisms identified in field isolates (Fig. 6C). Although mutation of residues within the first cluster of polymorphisms PvRBP2a_{N186S, K421M} seemed to have a detrimental effect on PvRBP2a ability to bind erythrocytes, we did not see a reduction in erythrocyte binding for the other variants: PvRBP2a_{E277K, K285N, K289E,} PvRBP2a_{E277K, K285I, K289E,} PvRBP2a_{E304K, D306V, P399S,} PvRBP2a_{E304K, D306V,} PvRBP2a_E351Q, and PvRBP2a_G438E (Fig. 8D). We observed a slight increase in erythrocyte binding in proteins with mutations within the negative patch (PvRBP2a_{E304K, D306V, P399S} and PvRBP2a_{E304K, D306V}), but this increase was not statistically significant.

Discussion

Here we show, to our knowledge, the first high-resolution crystal structure of the erythrocyte-binding domain of a member of the PvRBP family of proteins. This erythrocyte-binding domain from PvRBP2a shares a structural fold similar to PfRh5, an essential invasion ligand in P. falciparum and one of the current leading vaccine candidates against blood stage infection (22–24). Alignment of PvRBP2a erythrocyte-binding domain and other coding PvRBPs shows a conserved domain that is characterized by the placement of disulfide bonds. Sequence diversity analyses using field and reference isolates show that the PvRBP2a erythrocyte-binding domain lies within a region of significant polymorphism and balancing selection. We propose that, whereas the PfRhs and PvRBPs family of proteins share a similar fold, the surface properties for each member are distinct and reflect the specific nature of individual ligand–receptor interaction.

Our binding assays show that PvRBP2a recognizes an erythrocyte receptor that is present on both reticulocytes and mature erythrocytes. Its interaction with the receptor is completely abolished upon trypsin treatment but is resistant to neuraminidase and chymotrypsin treatment. Unfortunately, none of the known erythrocyte receptors in P. vivax and P. falciparum invasion share a similar enzyme profile. We propose that, whereas PvRBP2a does not govern entry into reticulocytes specifically, it may function in the subsequent steps in parasite invasion that are universal between P. vivax and P. falciparum. These molecular events may include modulation of the erythrocyte cytoskeleton, the assembly of the tight junction, or signaling within the parasite for rhoptry secretion. However, the elucidation of PvRBP2a function remains challenging due to the lack of a long-term in vitro culture and of genetic manipulation techniques for P. vivax.

The crystal structure of PfRh5 with basigin identified critical residues within the binding interface (24). Overlay of PfRh5′s binding interface upon PvRBP2a_160–455 shows that the surface properties within this region are quite different. PvRBP2a_160–455 shows a distinct patch of negative charge residues mostly contributed by E304, D306, E309, E313, E317, and E321 within helix α4. Mutation of these residues to lysine resulted in a marked reduction in the ability of PvRBP2a to bind erythrocytes. Sequence analyses of field isolates show E304 and D306 as lysines in their respective positions, and we also observed a slight increase in erythrocyte-binding in proteins with mutations within the negative patch (PvRBP2a_{E304K, D306V, P399S} and PvRBP2a_{E304K, D306V}). Collectively, these results highlight that the negative patch on PvRBP2a is an important surface for the interaction with erythrocyte proteins. The structural mechanism of how these polymorphisms will affect interaction with the erythrocyte receptor will be important questions for the future.

It remains a matter of outstanding interest how PvRBP1a and PvRBP2c mediate specific reticulocyte binding (11). It is notable that the site implicated in erythrocyte binding in PvRBP2a aligns with the basigin-binding region in PfRh5 (24). However, the preponderance of negatively charged residues in this region in PvRBP2a is not conserved among other PvRBP family members, suggesting that divergent interfaces for receptor binding are present within the family, as observed for DBL and EBA families (29, 30).

Cocrystallization studies between PfRh5 and inhibitory monoclonal antibodies have identified regions important for receptor–ligand interaction . Monoclonal antibody QA1 raised against PfRh5 blocks PfRh5–basigin interaction and inhibits parasite growth in vitro (18). QA1 binds to the loop at the tip of PfRh5 that is held by a disulfide bond between C345 and C351 (24). Residues within this loop are important for receptor specificity because SNP N347 has been linked with the ability of P. falciparum to invade Aotus monkey erythrocytes (16). In PvRBP2a, the cysteine bond between C299 and C303 overlaps with the bond between C345 and C351 in PfRh5. Mutation of C299 and C303 to serines resulted in a loss of PvRBP2a’s capability to bind erythrocytes, showing that the presence of the bond and the loop is required for PvRBP2a function. From our alignment of the PvRBP family, we observed that the most variable region between all PvRBP proteins includes the loop connecting helices α3b and α4, which includes the disulfide bridge formed between C299 and C303 on PvRBP2a. Future experiments should explore the intriguing possibility of targeting both the PfRh and PvRBP family of proteins by generating monoclonal antibodies that recognize specific conformational epitopes within the different loops.

The two-stranded β-sheet within the N terminus of PfRh5 and PvRBP2a is rotated by 180° between both proteins, with the orientation of the β-hairpin in PvRBP2a a result of a γ-turn with F181 in its center. Residues within the β-hairpin may contribute to stacking interactions or extensive hydrogen-bonding interactions that assist in forming the overall shape of PvRBP2a by keeping helices α2 and α3 apart. In particular, α2a and α3b in PvRBP2a are shifted by a whole diameter compared with equivalent helices in PfRh5, and the difference in positioning may confer receptor specificity. Helix α3b within PfRh5 is an important interacting surface for inhibitory monoclonal antibody 9AD4, which shows the strongest effect on the reduction of P. falciparum invasion (18). We hypothesize that the orientation of the β-hairpin within the erythrocyte-binding domain in the PfRh and PvRBP family of proteins will confer some level of receptor specificity by modulating the 3D architecture of helix α2 and α3.

Sequence diversity analyses using field and reference isolates show that the erythrocyte-binding domain of PvRBP2a is localized within a region of significant balancing selection, suggesting functional importance. From our erythrocyte-binding assays, it was observed that PvRBP2a_{N186S, K421M} did not bind very well to red blood cells from Australian donors but may reflect an adaptation to its cognate receptor in malaria endemic populations. Sequence polymorphisms in members of the P. falciparum erythrocyte-binding antigen (PfEBA) protein family result in changes either in receptor affinity or alternate receptor recognition, thus allowing functional flexibility in the use of erythrocyte receptors that are under selection in malaria-endemic regions. The identification of the erythrocyte receptor for PvRBP2a will be important to help elucidate the functional relationship between its sequence polymorphisms and receptor interaction.

PvRBP2a is highly polymorphic, with 81 polymorphic amino acids among the 31 isolates analyzed. In contrast, for PfRh5, 50 nonsynonymous polymorphisms were found among 204 isolates (Plasmodium falciparum Sequencing Project, Broad Institute of Harvard and MIT) (31, 32). The higher diversity of PvRBP2a is consistent with higher levels of diversity genome-wide for P. vivax compared with P. falciparum and is at least partly the result of a more stable and ancient parasite population (25). However, the high density of polymorphisms found in the elucidated binding region of PvRBP2a and evidence of balancing selection suggest antigenic diversity that allows parasites to escape host immune responses. However, some polymorphisms may have roles other than antigenic diversity, such as adaptation to a polymorphic host receptor. The differential binding of mutants to red blood cells is also consistent with this scenario.

SAXS data show that the longer fragment of PvRBP, including residues 160–1000, forms an elongated shape in solution with the erythrocyte-binding domain localized at one end of the molecule. Furthermore, ultracentrifugation analyses of PvRBP2a_160–1000 show that this elongated form exists solely as a monomer in solution. The erythrocyte-binding domain of PvDBP is monomeric in the absence of its receptor DARC but dimerizes upon receptor engagement to promote receptor specificity, potentially allowing invasion-activated signaling processes to occur. Whether or not the PvRBP family undergoes receptor-induced dimerization will be an important structural aspect to examine in the future. Some attempts have been also undertaken to structurally and functionally characterize different fragments of Py01365, a member of Py235 family (reviewed in ref. 33). This protein is thought to function as an ATP/ADP sensor with its nucleotide-binding properties residing within a 94-kDa domain, called NBD94 (34). The N-terminal portion of NBD94, so-called EBD_1–194, has been proposed to be involved in erythrocyte binding (35). However, SAXS data indicate that NBD94_444–547 and NBD94_674–793 do not have a similar architectural arrangement to PvRBP2a (36).

An outstanding question in P. vivax biology is the function of the PvRBP family in parasite invasion of red blood cells. PvRBP1a and PvRBP2c bind specifically to reticulocytes and thus allow P. vivax to enter this particular cell type (11), although the reticulocyte-specific receptors are yet to be identified. Mounting evidence shows that P. vivax is capable of invading Duffy-negative red blood cells, and one possibility is that PvRBPs may provide additional parasite adhesins for engaging with other erythrocyte receptors as entry portals. Expression of several PvRBPs at the apical tip may form a local concentration of adhesins to facilitate binding and affinity of the merozoite to erythrocyte. Future examination of the red blood cell-binding profiles of each individual PvRBP and identification of their cognate receptors will be important to further our understanding of P. vivax invasion and host cell selection.

Materials and Methods

cDNA Sequencing of PvRBP2a.

Fifty microliters of P. vivax patient blood was collected in 250 μL of TRIzol (Life Technologies) for RNA extraction. A TURBO DNA-free Kit (Life Technologies) was used to remove the trace amount of genomic DNA from purified RNA. Preparation of cDNA from total RNA with oligo dT followed the standard protocol of the SuperScript III First Strand Synthesis Kit (Life Technologies). cDNA was used as the template for PCR amplification of several overlapping gene fragments covering the first 4,000 base pairs of the gene. Each amplicon was sent to Macrogen for dye terminator sequencing. An intron was identified by sequencing and subsequently confirmed by PCR with primers 5′-AGG TGC TCT GGG CAG TTT T-3′ and 5′-TTG GGG GAT TTT CCT TCC-3′ (Fig. 1A).

Antibody Production.

Antibody production was performed at the Walter and Eliza Hall Institute Monoclonal Antibody Facility and approved by the Walter and Eliza Hall Institute Animal Ethics Committee (2014.009). Rabbits were immunized five times with 150 μg of PvRBP2a_160–1135. The first immunization was administered with complete Freund’s adjuvant, and the rest with incomplete Freund’s adjuvant. Rabbit IgG were purified from serum using Protein G Sepharose.

Cloning of pvrbp2a and Variants into Expression Vectors.

E. coli codon-optimized DNA of PvRBP2a (www.plasmodb.org/plasmo/) (PVX_121920) was purchased from Life Technologies. The fragment encompassing residues 160–1135 was cloned into the pPROEX HTb vector (Life Technologies). To generate mutations within PvRBP2a, we used a construct encompassing residues 160–1000 in the same vector. Restriction-free cloning was used to introduce the specific amino acid substitutions following a protocol described before (37). The list of all primer sequences is presented in Table S1. All positive clones were sequenced to verify the presence of each mutation.

Table S1.

Oligonucleotides used for the construction of plasmids

Oligonucleotide	Sequence
FOR160	CGCGGATCCGATATTCTGCGTTATCTGGATTTTAGC
REV1135	CGCTGCCTCGAGTCAATTGTCGCTCACGGTAAAAC
L1001STOP FOR	TATAACATTATCGCCAAAGATGACAAGTGAATCTTCGAAAAACGTCTGAACGAAGAAAAA
L1001STOP REV	TTTTTCTTCGTTCAGACGTTTTTCGAAGATTCACTTGTCATCTTTGGCGATAATGTTATA
pPROEXHTb FOR	ACGGTTCTGGCAAATATTCTG
E277K K285N K289E	TGCTGTAAATCTCCTCGTTGAAGTTGCTCATACGATCGGTATAGCTTTTCATGTTTGCCT
E304K D306V	TTCCAGCATAATTTCATAGGTAACGGTTTTACATGCATCGGTACATTTAACGCTGTT
E304K D306K E309K E313K E317K E321K	CACTTTTTTCACGCGTTTGACATAGATTTTCAGCATAATTTTATAGGTTTTGGTTTTACA
C227S	TTCCAGTGCTGCTTTTTTCGGAACGCTGCTATTAATCTGGTTCTGAATGCT
C271S	GCTTTCCATGTTTGCCTGAACGCTGTTTTTATAGGCATCCAGCTG
C299S C303S	AATTTCATAGGTATCGGTTTCGCTTGCATCGGTGCTTTTAACGCTGTTCAGAATGCT
P399S	TTTTTTCAGGGTTGACAGCGGAACCTGTGACGGAATGTTATTGTTCTGCAGTTTCAG
G438E	TTTGTTGTACAGGTTTGCCAGCAGTTTTTCTTTTTCTTTCAGTGCTGCGGTGGTGGT
E351Q	GGCGCTAATGGTAACGTTGTTATCGATTTGTTGTTTAACTTTCAGCATCAGGGTAAC
N186S	ACCGTTTATCCGTTTTATGTGCAGATGAGCTATTTTGCCGAGATTAAATACTATATTACC
K421M	GGTATCGGCACGTTTCAGGCTGAACATATAGGTGGCGTAGAAGTTGGCGCT
E277K K285I K289E	TGCTGTAAATCTCCTCGTTGAAAATGCTCATACGATCGGTATAGCTTTTCATGTTTGCCT
P180A F181A Y182A	AGCGGTCAGATTATTAGCACCGTTTATGCCGCCGCCGTGCAGATGAACTATTTTGCCGAG

Expression and Purification of Recombinant PvRBP2a.

Protein expression was performed using the SHuffle T7 E. coli strain (New England Biolabs). Soluble protein was captured using a HisTrap column (GE Healthcare), and eluted protein was dialyzed overnight in the presence of TEV protease. The protein sample was passed once again through a HisTrap column, and the collected flow-through was subjected to size exclusion chromatography. Expression and purification of PvRBP2a_160–1000 and variants were performed in a similar fashion.

Crystallization of PvRBP2a, X-Ray Data Collection, and Structure Determination.

A PvRBP2a_160–1135 construct was used for crystallization trials. Crystal screenings were performed at the CSIRO Collaborative Crystallization Center (Parkville, Australia). The initial crystal hit was optimized manually in the presence of porcine elastase (MP Biomedicals). Diffraction data were collected at the MX2 microfocus beamline at the Australian Synchrotron Facility in Clayton. Molecular replacement was attempted with the PfRh5 structure as a search model (PDB ID code 4WAT) but was unsuccessful. The single isomorphous replacement with anomalous scattering (SIRAS) approach was attempted, combining native and derivative datasets and using AutoRickshaw (38). The obtained initial model was rebuilt automatically using the program AutoBuild (39), followed by manual improvement using the program Coot (40). The structure was refined using the program Phenix Refine (39), and the refinement statistics are given in Table 1. The atomic coordinates and structure factors have been deposited in the Protein Data Bank with accession number 4Z8N. The PDB Structure Validation Report is provided in Dataset S2.

Small-Angle X-Ray Scattering Data Collection, Processing, and Analysis.

SAXS data were collected on PvRBP2a_160–1000 (8 mg/mL) eluted in Dulbecco's Phosphate-Buffered Saline (DBP) by inline size exclusion chromatography (SEC) via a capillary in the path of the X-ray beam. Data reduction was performed using Scatterbrain, and processing is detailed in Table 2, as before (41).

Analytical Ultracentrifugation Data Collection, Processing, and Analysis.

Samples were analyzed using an XL-I analytical ultracentrifuge (Beckman Coulter) equipped with an AnTi-60 rotor. The data were processed using the program SEDFIT (42).

Circular Dichroism Data Collection, Processing, and Analysis.

Circular dichroism (CD) data were collected using CD spectrometer Model 410 (Aviv Biomedical). The spectra were processed using the DichroWeb server (43, 44). The content of secondary structures was predicted using the CDSSTR algorithm (45) and SMP180 reference set (46) for 190–240 nm.

Sequence Alignment of PvRBP Family.

Sequence alignment was performed using ClustalW2 (47), corrected manually and visualized using program ESPript3.0 (48). All protein sequences were obtained from the PlasmoDB Database under the following accession numbers: PvRBP1a PVX_098585, PvRBP1b PVX_098582, PvRBP2a PVX_121920, PvRBP2b PVX_094255, PvRBP2c PVX_090325, PvRBP2-P1 PVX_090330, and PvRBP2-P2 PVX_101590.

Population Genetics Analyses.

To investigate naturally occurring diversity among PvRBP2a gene sequences, the full-length sequence was analyzed (nucleotides 1–7464), and nine previously published PvRBP2a gene sequences were available for analysis, including four sequences from Thailand (nucleotides 1303–7464 only) (26) and five P. vivax reference isolates under the following PlasmoDB database accession numbers: Sal-1 (PVX_121920), Brazil I (PVBG_04239), India VII (PVIIG_00453), Mauritania (PVMG_03227), and North Korea (PVNG_00022). In addition, 22 high quality PvRBP2a gene sequences were extracted from available genome sequence data from field isolates collected in Madang (n = 18), East Sepik (n = 3), and Milne Bay (n = 1) Provinces, in Papua New Guinea where P. vivax malaria is hyperendemic. Data were generated by the MalariaGEN Plasmodium vivax Community Project (www.malariagen.net/projects/parasite/pv) and by the Plasmodium vivax Sequencing Project at the Broad Institute of Harvard and MIT (www.broadinstitute.org/) using different Illumina short read sequencing platforms. Briefly, Illumina short reads were aligned to the reference genome Salvador 1 using bwa-short (49). Additional read trimming was performed if the base quality score was below 15, and reads were aligned in either single-end or paired-end mode depending on their read-group. PCR and optical duplicates were marked using PicardTools (broadinstitute.github.io/picard/). Variant calls in each isolate were obtained using samtools mpileup (50) for reads mapping to the PvRBP2a region, and then consensus sequences were extracted using vcf-consensus from the vcftools package (51). All 31 consensus pvrbp2a sequences are available in Dataset S1.

Multiple alignments of the consensus sequences were performed using the MUSCLE algorithm implemented in MEGA version 5.0 software (52). Genetic polymorphism and diversity were estimated using DnaSP version 5.0 (53) by calculating the total number of polymorphic sites, synonymous and nonsynonymous polymorphisms, and number of haplotypes (i.e., different combinations of polymorphisms). Average pairwise nucleotide diversity (π) was calculated across the length of the gene using a sliding window approach, with a window size of 100 and a step size of 3.

To identify regions under selection, the Tajima’s D test statistic was calculated using DnaSP version 5.0 (53), using a sliding window approach with a window size of 100 and a step size of 3. Positive values indicate an excess of alleles at intermediate frequencies, indicative of a recent population bottleneck or balancing (immune) selection (54). Negative values of the Tajima’s D test statistic indicate an excess of rare alleles, consistent with directional selection or recent population expansion.

Flow Cytometry-Based Erythrocyte-Binding Assay.

The flow cytometry-based erythrocyte-binding assay was performed as described with the following modifications (27). Five micrograms of recombinant PvRBP2a was incubated with 100 μL of erythrocyte suspension for 1 h at room temperature. After binding, erythrocytes were washed, and PvRBP2a binding was detected using anti-PvRBP2a rabbit IgG (1 mg/mL, 1:100, followed by Alexa Fluor 647 chicken anti-rabbit secondary antibody; Life Technologies). Erythrocytes were washed twice and resuspended in 600 μL of PBS. In the experiments where purified reticulocytes were used, 100 μL of thiazole orange (BD Bioscience) was added and incubated at room temperature for 30 min before final resuspension in PBS. A total of 50,000 erythrocytes were read on the FACSCalibur flow cytometer (BD Bioscience), and the results were analyzed using FlowJo software (Three Star). The percentage of erythrocytes with bound recombinant PvRBP2a was determined by normalizing the number of erythrocytes exhibiting a positive Alexa Fluor 647 (or 488) signal that is above the background [which is the Alexa Fluor 647 (or 488) signal of erythrocytes without recombinant protein added] on the total number of erythrocytes.

SI Materials and Methods

cDNA Sequencing of PvRBP2a.

Fifty microliters of P. vivax patient blood was collected in 250 μL of TRIzol (Life Technologies) for RNA extraction. A TURBO DNA-free Kit (Life Technologies) was used to remove the trace amount of gDNA from purified RNA. Preparation of cDNA from total RNA with oligo dT followed the standard protocol of the SuperScript III First Strand Synthesis Kit (Life Technologies). cDNA was used as the template for PCR amplification of several overlapping gene fragments covering the first 4,000 base pairs of the gene. Each amplicon was sent to Macrogen for dye terminator sequencing. An intron was identified by sequencing and was subsequently confirmed by PCR with primers 5′-AGG TGC TCT GGG CAG TTT T-3′ and 5′-TTG GGG GAT TTT CCT TCC-3′ (Fig. 1A).

Antibody Production.

Antibody production was performed at the Walter and Eliza Hall Institute Monoclonal Antibody Facility. Rabbits were immunized five times with 150 μg of PvRBP2a_160–1135. The first immunization was administered with complete Freund’s adjuvant and the rest with incomplete Freund’s Adjuvant. Rabbit IgG were purified from serum using Protein G Sepharose.

Cloning of pvrbp2a and Variants into Expression Vectors.

The PvRBP2a sequence of P. vivax strain Salvador I was obtained from the PlasmoDB Database (www.plasmodb.org/plasmo/) (PVX_121920). An E. coli codon-optimized version of PvRBP2a (Life Technologies) containing the DNA sequence for amino acids 24–1323 was cloned into the pMA-RQ (ampR) vector. The fragment encompassing residues 160–1135 was amplified from this vector, which also introduced BamHI and XhoI restriction sites at the ends of the DNA sequence. The PCR product was digested with BamHI and XhoI (New England Biolabs) and cloned into compatible sites in the pPROEX HTb vector (Life Technologies). Upon transformation, positive clones were selected and sequenced at the Melbourne Translational Genomics Platform. The final construct PvRBP2a_160–1135 enables expression of the protein fused to a tobacco etch virus (TEV) protease cleavage site cleavable N-terminal hexa-histidine tag. To generate mutations within PvRBP2a, we used a construct encompassing residues 160–1000 in the pPROEX HTb vector as a template. Restriction-free cloning was used to introduce the specific amino acid substitutions into the original template using a universal forward primer and the reverse primer bearing particular mutations and following a protocol described before (37). The list of all primer sequences is presented in Table S1. All positive clones were sequenced to verify the presence of each mutation.

Expression and Purification of Recombinant PvRBP2a.

Protein expression was performed using the SHuffle T7 E. coli strain (New England Biolabs) grown in Terrific Broth medium supplemented with 100 μg/mL carbenicillin. Expression was induced using 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG) (Astral) for 20 h at 16 °C. Cells were harvested by centrifugation at 6,000 × g and stored at −80 °C until further processing. For protein purification, frozen pellets were thawed, resuspended, and supplemented with 0.5 mg/mL DNase and 1.0 mg/mL lysozyme (Sigma). Cells were sonicated and centrifuged at 30,000 × g for 30 min at 4 °C. The supernatant was filtered through a 0.22-μm syringe filter and applied on the 5-mL HisTrap column (GE Healthcare). Unbound material was washed away using at least 10 column volumes of washing buffer: 50 mM Tris⋅HCl, pH 7.5, 500 mM NaCl, 10% (vol/vol) glycerol, and 20 mM imidazole. Bound protein was eluted from the column using the same buffer with 300 mM imidazole. Eluted fractions were analyzed on SDS PAGE, and appropriate fractions were pooled and dialyzed overnight in the presence of TEV protease. The protein sample was passed once again through a 5-mL HisTrap column to bind the unprocessed protein, cleaved hexa-histidine tag, and other impurities. The collected flow-through was concentrated and injected onto an S200 Superdex 16/600 size exclusion column (GE Healthcare) preequilibrated with the buffer containing 20 mM Tris⋅HCl, pH 7.5 and 150 mM NaCl. The monodisperse peak fractions containing pure protein were pooled, concentrated, and used immediately for the crystallization trials. A typical yield was 10 mg of the protein from 2 L of bacterial culture. Expression and purification of PvRBP2a_160–1000, as well as all of the mutant proteins, were performed analogically to that described for the PvRBP2a_160–1135 protein, except that DPBS was used as a final buffer for size exclusion chromatography. Protein used for SAXS and AUC experiments was flash-frozen in liquid nitrogen upon addition of 10% (vol/vol) of glycerol and stored at −80 °C. The protein sample ran on SDS/PAGE as a single band around 115 kDa, and purity was assessed to be around 98%. We confirmed the identity of the protein using mass spectrometry.

Crystallization of PvRBP2a.

The PvRBP2a construct used for crystallization trials included residues from 160 to 1135. Crystal screens using six different sparse matrix screens were set up at the CSIRO Collaborative Crystallization Center (Parkville, Australia). Then, 150 nL of protein sample at two different concentrations, 7.5 and 15 mg/mL, were mixed with the equal volume of precipitation solution. Plates were incubated at 20 °C, and the drops were monitored regularly over a period of 3 mo. Crystals appeared in the condition containing 25% (wt/vol) PEG 3350 and 100 mM Hepes, pH 7.5. The initial crystal hit was optimized manually using the hanging drop method in 24-well plates with varying pH and PEG concentration in the presence of porcine elastase (MP Biomedicals).

X-Ray Data Collection and Structure Determination.

Diffraction data were collected at the MX2 microfocus beamline at the Australian Synchrotron Facility in Clayton at 0.9537 Å wavelength using an ADSC Quantum 315r detector. The best crystal diffracted till 2.12 Å resolution. The crystal belonged to the orthorhombic space group P2₁2₁2₁. The cell dimensions were as follows: a = 58.79 Å, b = 93.45 Å, and c = 126.72 Å. The collected diffraction data were processed using iMosflm (55). Scaling and merging were performed using the program Aimless from the CCP4 package (56). The cell content analysis was performed using the program Matthews (57) and the suggested presence of two molecules in the asymmetric unit. To obtain the phases from the experiment, crystals were soaked for about 1 min in the crystallization solution supplemented with 500 mM sodium iodide (NaI). Iodine-derivatized crystals diffracted to 2.7 Å and were isomorphous with the native crystals. The single-wavelength anomalous dispersion (SAD) data were collected at the same beamline at 1.4586 Å wavelength. The single isomorphous replacement with anomalous scattering (SIRAS) approach was attempted, combining native and derivative datasets and using AutoRickshaw. The obtained initial model was rebuilt automatically using the program AutoBuild, followed by manual improvement using the program Coot. The structure was refined using the program Phenix Refine (39), including TLS (Translation/Libration/Screw) vibrational motion that was generated using the TLS Motion Determination web server. Non-crystallographic symmetry restrains were used only in the initial cycles of model building and refinement but were removed during the final steps. The structure was eventually refined against native data to 2.12 Å resolution with the final R_work = 20.3% and R_free = 22.9%. The refinement statistics are given in Table 1.The atomic coordinates and structure factors have been deposited in the Protein Data Bank with accession number 4Z8N. Interactions between molecules in the structure were analyzed using the program PISA (58). A geometry check of the protein model was performed using PROCHECK, PDBsum, and PROMOTIF. Figures were generated using the program PyMOL.

Small-Angle X-Ray Scattering Data Collection, Processing, and Analysis.

SAXS data were collected on PvRBP2a_160–1000 (8 mg/mL) eluted in DPBS by inline SEC via a capillary in the path of the X-ray beam. Data reduction was performed using Scatterbrain and was processed as detailed in Table 2, as before (41).

Analytical Ultracentrifugation Data Collection, Processing, and Analysis.

Samples were analyzed using an XL-I analytical ultracentrifuge (Beckman Coulter) equipped with an AnTi-60 rotor. PvRBP2a_160–1000 samples at three different protein concentrations (0.4, 1.0, and 1.7 mg/mL) were loaded in the sample compartment of double-sector Epon centerpieces, with DPBS buffer in the reference compartment. Radial absorbance data were acquired at 20 °C using a rotor speed of 50,000 rpm and a wavelength of 280 nm, with radial increments of 0.003 cm in continuous scanning mode. The sedimenting boundaries were fitted to a model that describes the sedimentation of a distribution of sedimentation coefficients with no assumption of heterogeneity [c(s)] using the program SEDFIT (42). Data were fitted using a regularization parameter of P = 0.95, floating frictional ratios, and 150 sedimentation coefficient increments in the range of 0.1–15 S.

Circular Dichroism Data Collection, Processing, and Analysis.

CD data were collected using circular dichroism spectrometer Model 410 (Aviv Biomedical). Before the experiment, protein samples were dialyzed extensively overnight into DPBS buffer. For the measurements, the protein concentration was set to 3 μM. The CD spectra were recorded between 260 and 190 nm with 1-nm wavelength step and 4-s averaging time at 25 °C. Three hundred microliters of the protein sample was placed into a quartz cell (Hellma) with the optical path of 0.1 cm. After subtracting the background from the buffer, the spectra were processed using the DichroWeb server. The experimental data were converted to mean residue molar ellipticity (θ). The content of secondary structures was predicted using the CDSSTR algorithm and the SMP180 reference set for 190–240 nm.

Sequence Alignment of the PvRBP Family.

Sequence alignment was performed using ClustalW2 (47), corrected manually and visualized using the program ESPript3.0. All protein sequences were obtained from the PlasmoDB Database under the following accession numbers: PvRBP1a PVX_098585, PvRBP1b PVX_098582, PvRBP2a PVX_121920, PvRBP2b PVX_094255, PvRBP2c PVX_090325, PvRBP2-P1 PVX_090330, and PvRBP2-P2 PVX_101590.

Flow Cytometry-Based Erythrocyte-Binding Assay.

The flow cytometry-based erythrocyte-binding assay was performed as described with the following modifications (27). Forty microliters of packed erythrocytes were washed twice with 1 mL of 1% BSA/PBS and resuspended to a final volume of 1 × 10⁷ cells per mL. Five micrograms of recombinant PvRBP2a was incubated with 100 μL of the resuspended erythrocytes for 1 h at room temperature. After binding, erythrocytes were washed three times with 1% BSA/PBS. To detect PvRBP2a binding, anti-PvRBP2a rabbit IgG (1 mg/mL, 1:100 dilution) was added and incubated for 1 h, followed by three washes with 1% BSA/PBS. Alexa Fluor 647 chicken anti-rabbit secondary antibody (Life Technologies) was added and incubated at room temperature for 1 h. Erythrocytes were washed twice with 1% BSA/PBS, followed by one wash in PBS, and resuspended in 600 μL of PBS. In the experiments in which purified reticulocytes were used, 100 μL of thiazole orange (BD Bioscience) was added and incubated at room temperature for 30 min before final resuspension in PBS. A total of 50,000 erythrocytes were read on the FACSCalibur flow cytometer (BD Bioscience), and the results were analyzed using FlowJo software (Three Star). The percentage of erythrocytes with bound recombinant PvRBP2a was determined by normalizing the number of erythrocytes exhibiting a positive Alexa Fluor 647 signal that is above the background (which is the Alexa Fluor 647 signal of erythrocytes without recombinant protein added) on the total number of erythrocytes.