Plasmodium vivax ligand-receptor interaction: PvAMA-1 domain I contains the minimal regions for specific interaction with CD71+ reticulocytes

The malarial parasite’s invasion is complex, active and coordinated, involving many low and high affinity interactions with receptors on target cell membrane. Proteomics analysis has described around 40 proteins in P. vivax which could be involved in reticulocyte invasion; few have been studied with the aim of elucidating how many of them establish specific interactions with their respective host cells. Given the importance of knowing which of the parasite’s protein regions are functionally important for invasion, minimum regions mediating specific interaction between Plasmodium vivax apical membrane antigen 1 (PvAMA-1) and its host cell were here elucidated. The region covering PvAMA-1 domains I and II (PvAMA-DI-II) specifically bound to the CD71+ red blood cell subpopulation. A 20 residue-long region (81EVENAKYRIPAGRCPVFGKG100) located in domain I was capable of inhibiting PvAMA-DI-II recombinant protein binding to young reticulocytes (CD71+CD45−) and rosette formation. This conserved peptide specifically interacted with high affinity with reticulocytes (CD71+) through a neuraminidase- and chymotrypsin-treatment sensitive receptor. Such results showed that, despite AMA-1 having universal functions during late Plasmodium invasion stages, PvAMA-1 had reticulocyte-preferring binding regions, suggesting that P. vivax target cell selection is not just restricted to initial interactions but maintained throughout the erythrocyte invasion cycle, having important implications for designing a specific anti-P. vivax vaccine.


Figure 1. COS7 cells expressing PvAMA-DI-DII bound to UCB RBC. (a) Schematic representation of
PvAMA-1 primary structure. Signal peptide (SP) is indicated in red, pro-sequence (Pro) in light grey, ectodomain (consisting of domain I (DI), domain II (DII) and domain III (DIII)) in dark grey, transmembrane region (TR) in black and cytoplasmic tail in white. The three fragments were expressed both on COS-7 cell membrane and as soluble recombinant proteins in E. coli. Black shows the complete PvAMA-1 ectodomain (rPvAMA-1), dark grey represents domains I-II (rPvAMA-DI-II) and light grey domains II-III (rPvAMA-DII-III). The first and last residues in each region are indicated. (b) Immunofluorescence assay. Each of the three PvAMA-1 fragments was cloned in the pRE4 vector and expressed on COS-7 cell membrane. The presence of HSV gD protein signal peptide and transmembrane domain in pRE4 vector enabled the expression of each fragment on cells surface, as shown by green fluorescence. Cell nuclei were stained with DAPI. (c) Rosette formation assays. The binding assay was performed with UCB RBC or reticulocyte-depleted normocytes. Only PvAMA-DI-DII and DBPRII transfected cells could form rosettes (black arrows).The amount of rosettes for each protein expressed on COS-7 cells is shown. DBPRII was used as positive control while non-transfected cells and DBPRII binding to chymotrypsin-treated UCB were used as negative controls. (d) Trypsin-, chymotrypsin-and neuraminidase-treated UCB RBC binding assays. It can be seen that the amount of rosettes became reduced with PvDI-DII protein when RBC were treated with chymotrypsin and neuraminidase. Only intact erythrocyte and chymotrypsin-treated DBPRII binding was evaluated. At least three independent experiments (with replicates) were performed for all experiments. The error bars indicate the standard deviation.
the domain I and II region is responsible for PvAMA-1 binding. PvAMA-DI-II binding was affected regarding neuraminidase-or chymotrypsin-treated RBC, being reduced by 71% or 73%, respectively, while limited impact on binding was observed concerning trypsin-treated RBC (Fig. 1d). Enzyme treatment of RBC led to no significant change in the amount of rosettes for either of the other two PvAMA-1 fragments, different from what has been reported for PfAMA-1 domain III which only binds to its receptor (known as Kx) on trypsin-treated erythrocytes 38 .
Taking into account that P. vivax preferentially infects reticulocytes, an evaluation was made of whether PvAMA-DI-II binding concerns interaction with all RBC populations or only a minor subpopulation (reticulocytes). As purified intact reticulocytes (>99%) could not be obtained, peripheral blood (containing only 1% reticulocytes) depleted in white blood cells and reticulocytes was used to obtain only normocytes and determine binding properties regarding the latter cells. The amount of rosettes decreased with normocytes regarding UCB (p = 0.01) for COS-7 expressing PvAMA-DI-II (Fig. 1d), suggesting that these domains bound to a specific non-normocyte RBC sub-population while only a slight reduction in the el amount of rosettes was observed for DBP region II thereby agreeing with that published previously for this region 23 . CD71 + cells were the main PvAMA-DI-II binding targets. Previous studies have shown that immature reticulocytes (CD71 + , I-II-III stages) are the major P. vivax target cells, whereas older CD71 − (IV stage) reticulocytes were rarely invaded 69 . A cytometry-based erythrocyte-binding assay was thus used to evaluate whether PvAMA-DI-II binding was dependent on a specific RBC subpopulation (i.e. reticulocytes (CD71 + ). Three soluble recombinant proteins (rPvAMA-1, rPvAMA-DI-II and rPvAMA-DII-III) (Fig. 1a) were obtained in E. coli and purified by affinity chromatography (Fig. 2a). These proteins displayed a molecular mass of ~54 kDa (rPvAMA-1), ~40 kDa (rPvAMA-DI-II) and ~27 kDa (rPvAMA-DII-III) in SDS-PAGE, and were recognized by WB with a monoclonal anti-histidine antibody (Fig. 2a). rPvAMA-1 and rPvAMA-DII-III were recognized by mAb F8. 12.19 in dot blot and ELISA assays, confirming correct recombinant protein folding (Fig. 2a). This mAb recognized a conformational epitope on full-PvAMA-1, specifically in domain III 65 .
Each recombinant protein was incubated with UCB and reticulocytes were stained with transferrin receptor (CD71) antibodies; this receptor is known to be expressed during the earliest erythroid precursor stages and becomes progressively depleted during reticulocyte maturation into normocytes 70 . It has also been found in activated lymphocytes 71 ; an antibody against the CD45 marker (found in all leukocytes) having greater expression in lymphocytes was thus included to rule out this population from the binding study ( Supplementary Fig. S1). The study population thus consisted of CD71 + CD45 − cells (mean 2.07%: 3.58% maximum and 1.54% minimum) (Fig. 2b) corresponding to the population of preference in P. vivax invasion. Each recombinant protein's binding to CD71 + CD45 − cells was detected with a phycoerythrin-conjugated anti-histidine monoclonal antibody. The histograms show that soluble recombinant DBPRII bound to CD71 + CD45 − cells (16.1% ± 2.5) as did rPvAMA-DI-II (9.3% ± 1.2) whereas only 2.5% ± 0.6 binding was found for rPvAMA-1 and 2.9 ± 0.5 for rPvAMA-DII-III (p < 0.05) ( Fig. 2b and c). Only 0.8% binding of DBPRIII-IV was found; earlier studies have shown that this fragment did not bind to either normocytes or reticulocytes, so it was considered as a negative control 17,72 . rPvAMA-DI-II binding activity was shown to increase in a concentration-dependent manner and was saturated at 50 µg/mL, similar to DBPRII (Fig. 2d). Only 0.03% to 0.3% binding was found when each PvAMA-1 recombinants' binding to the CD71 − CD45 − population (normocytes and stage IV reticulocytes) was evaluated ( Supplementary Fig. S2). 3.4% ± 0.5 binding was found for PvDBP RII thereby agreeing with previous work showing that DBPRII-DARC interaction is more abundant with reticulocytes than total RBC 23 .
Conserved peptide 21270 had binding properties which were CD71 + reticulocyte-specific for the PvAMA-DI-II región. Seventeen peptides covering the whole PvAMA-1 domains I and II (residues 41 to 380) were chemically synthesized and included in a CD71 + cell binding inhibition experiment where each peptide competed with rPvAMA-DI-II to delineate the specific region within PvAMA-DI-II involved in CD71 + reticulocyte interaction. Figure 3a shows that adding peptide 21270 (located in DI) produced a 52% reduction regarding rPvAMA-DI-II binding to CD71 + CD45 − cells and 14% concerning 21289 (located in DII).
We have previously reported a sensitive and specific binding test for detecting receptor-ligand interactions between peptides and receptors on RBC surface 73 . Each synthetic peptide derived from malarial Spz and Mrz antigens was radiolabeled and incubated with human hepatocytes or RBC in the presence or absence of the same unlabeled peptide. The specific binding plot was obtained from the difference between total binding (binding in the absence of non-radiolabeled peptide) and unspecific binding (binding in the presence of non-radiolabeled peptide) plots 74 . Peptides having ≥0.02 slope specific binding plot were considered high activity binding peptides (HABPs) and used as templates for designing an anti-malarial vaccine 75 . Some HABPs were conserved (no amino acid sequence variation) and others had genetic variability and elicited a strain-specific immune response allowing malarial parasites to escape vaccine-induced immunity, meaning that they were discarded from further immunological studies 74 .
Taking into account that AMA-1 is significantly polymorphic in all Plasmodium species 60, 76 , the specific binding activity of PvAMA-1 peptides which did not have amino acid changes was measured ( Supplementary Fig. S2) to describe this protein's conserved functional regions involved in interaction with specific target cells. Five conserved peptides were selected, three of them in DI and two in DII, peptides 21268 and 21270 being found to be HABPs having ≥2% specific binding activity (Fig. 3b). Only 0.7% to 1.2% specific binding (not surpassing the ≥2% threshold) was found when these peptides were assayed for normocyte binding showing that the specific binding activity measured for these two peptides was directed towards their interaction with reticulocytes, as already shown by cytometry (Fig. 2). Due to HABP 21270 being able to inhibit recombinant protein interaction by more than 50% (1:20 molar ratio) (Fig. 3a), it was evaluated whether it could inhibit rosette formation. Adding UCB RBCs pre-incubated with peptide 21270 to COS-7 cells expressing PvAMA-DI-II inhibited the amount of rosettes by 87% compared to control (Fig. 3c). Measuring HABP 21270 binding to UCB RBC at different radiolabeled peptide concentrations in the presence or absence of unlabeled peptide yielded a 1.9 ± 0.34 µM K d , Hill coefficient = 3.0 ± 0.47 indicating positive cooperativity (Fig. 3d).
HABP 21270-CD71 + reticulocyte interaction was sensitive to treatment with neuraminidase and chymotrypsin (Fig. 3e), similar to the pattern observed for the whole of domains I-II, suggesting that this peptide located in domain I had PvAMA-1-specific binding properties. PvAMA-1 HABPs were located near the hydrophobic binding Groove. The PvAMA-1 structure was solved by Pizarro et al. 31 , who described three domains comprising the antigen's ectodomain. When HABPs 21268 and 21270 were located on PvAMA-1 structure it was found that both peptides were located on the same face ( Fig. 4), opposite to the face where most polymorphic sites which contribute towards PvAMA-1's overall diversity 61 were found.

Discussion
Determining P. vivax antigens' functional regions could be one of the best strategies for blocking this parasite's biological functions during host cell entry. Comparative studies involving P. falciparum have led to a significant amount of new antigens being identified in P. vivax 77 ; in spite of this, few studies have emphasized the identification of P. vivax proteins' functional regions during invasion, mainly due to this species preferentially invading reticulocytes (low concentration in peripheral blood) together with the inherent characteristics in each P. vivax strain impeding the development of a continuous culture in vitro 78 . Even though short-term cultures have been developed, it has been observed that parasite density becomes reduced as intra-erythrocytic cycles advance, even when adding fresh reticulocytes 79 . The foregoing, plus variability concerning parasitemia percentages depending on the isolate 79,80 , could lead to a poor interpretation of the role played by a particular antigen during P. vivax Mrz invasion.
Evaluating some P. vivax proteins' binding to human reticulocytes represents an approach to begin describing the possible antigens involved in host cell invasion. Some regions having specific binding to and high affinity for receptors on reticulocyte membrane suggested the participation of the proteins evaluated here in some steps related to host cell invasion. Using recombinant proteins and reticulocytes in host-pathogen interaction has provided one of the best alternatives to date 21,22,81 .
Bearing in mind that obtaining pure reticulocytes in high amounts for binding assays is extremely difficult due to their low concentration in blood and little stability, UCB RBC or reticulocyte-depleted normocytes were here used for characterizing specific PvAMA-1 reticulocyte binding properties using three different receptor-ligand interaction techniques. Rosette formation led to recombinant expression in a mammalian system containing all the machinery for the correct P. vivax protein folding. Flow cytometry was used for determining which specific RBC population each recombinant protein assayed here bound to and which short regions might be mediating Peptides having ≥2 binding percentage (red line) were considered HABPs, according to previously described criteria 74 . Each assay was carried out in triplicate. A tyrosine residue was added at the C-terminus of peptides 21268 and 21287 to enable radiolabeling. such binding. Finally, a robust methodology was used for confirming peptide-cell interaction specificity and determining physicochemical parameters.
Binding assays using the aforementioned methodologies (including soluble recombinant proteins and proteins expressed on COS-7 cell membrane) showed that PvAMA-DI-II recognized a receptor just on CD71 + reticulocytes (i.e. a P. vivax target cell) ( Fig. 1 and 2). Receptor-ligand interaction was affected by treating UCB RBC with chymotrypsin and neuraminidase. The same enzymatic profile observed for PvAMA-DI-II has been described for glycophorin B which is a receptor for erythrocyte binding ligand-1 (EBL-1) 12 and the receptor for Rh1 binding; however, future experiments should explore the identity of the PvAMA-DI-II receptor. "Why do only AMA-1 domains I and II bind to CD71 + reticulocytes if the full PvAMA-1 ectodomain contains domains I-II-III?" represents an interesting question which emerged from the binding assays. Miss-folding could explain this pattern; however, these fragments were expressed in COS-7 ensuring correct protein folding (Fig. 1). mAb F8.12.19 which recognized a conformational epitope in DIII 65 recognized both complete rPvAMA-1 and rPvAMA-DII-III (Fig. 2a). Some studies have shown that different AMA-1 recombinant domain combinations (i.e. DI-DII or DIII only) maintain the same structure as the complete AMA-1 protein 32 .
A previous study found that PfAMA-1 domain III was the only fragment which could specifically interact with a receptor on RBC membrane, only becoming exposed when RBC were treated with trypsin, while the complete ectodomain did not interact. The authors suggested that this might have been due to PfAMA-1 cleavage and that DIII acts as a receptor for invasion 38 . Even though the presence of fragments has not been determined for PvAMA-1, it cannot be ruled out that the protein undergoes proteolytic processing similar to that for PfAMA-1 and that the resulting fragments play a role in binding. A similar pattern has been described for PfMSP-1 and its cleavage products 8,82 .
It has been suggested that AMA-1 participates in the invasion stage which is conserved for most apicomplexa, as well as having similar tertiary structure in Plasmodium spp. 60 ; however, different binding properties have been reported for its homologues. It has been reported through rosetting assays that different PfAMA-1 domains did not bind to intact human RBC 38 ; however, DIII have bound to human erythrocytes pre-treated with trypsin 38 while PyAMA-1 DI-DII has only interacted with intact mouse and rat erythrocytes but not with trypsin-or chymotrypsin-treated erythrocytes 37 . Complementing other authors' reports 32 , the results presented here suggested that AMA-1 recognizes specific receptors on different Plasmodium host cells in close relationship with target cell preference.
Comparative analysis of the proteome of human RBC and reticulocytes has shown that membrane or cytosol protein detected thereby was not unique to just erythrocytes or reticulocytes 83 . The difference between these two cells thus lying in the abundance of each protein, having around 500 membrane proteins, being lower in erythrocytes than in reticulocytes 83 . It seems that such differences could be critical for P. vivax to have tropism regarding reticulocytes compared to normocytes; however, it is still not clear whether P. vivax selectivity for its target cells could be just mediated during initial invasion stages where low affinity interactions occur, or whether this happens during late stages where high affinity interactions are established or simply that selectivity is maintained throughout reticulocyte invasion.
When defining domain I and II regions responsible for binding to CD71 + reticulocytes it was found that peptide 21270 located between residues 81 E to 100 G in domain I significantly inhibited the soluble recombinant protein (rPvAMA-DI-II) binding and rosette formation, strongly suggesting that this peptide has PvAMA-1 binding properties. Furthermore, to reinforce this inhibition experiment, it was found that peptide 21270 is a HABP which binds specifically and with high affinity to reticulocytes, having an enzyme binding profile similar to that reported for the soluble recombinant protein (rPvAMA-DI-II). Interestingly, this peptide's amino acid sequence is conserved among strains having various geographical origins. Several studies have shown that AMA-1's polymorphic nature prevents it being considered a vaccine candidate 61 ; in fact, the results of different PfAMA-1 formulations assayed at clinical level have shown that a protective response has been strain specific. Thus, finding conserved regions of PvAMA-1 having a binding function could be used for resolving one of the parasite's main evasion mechanisms: antigenic polymorphism 84 . No binding peptides were found in PvAMA-1 domain II; it has been found in other species, such as P. falciparum, that AMA-1 domain II does not participate in binding to RBC 38 but does play a role during PfRON2 interaction-related invasion 16,85 . Structural studies have suggested that DII protects a significant portion of the binding site in AMA-1 against a host's immune response and can be readily displaced to extend the hydrophobic groove for effective binding to PfRON2 16 . Such structural changes regarding AMA-1-RON2 interaction have been used for explaining invasion inhibiting antibodies (i.e. 4G2) which have recognised the base of a loop in PfAMA-1 ectodomain DII 43 . MAb 4G2 binding has prevented DII loop displacement for effective PfRON2 binding, thereby affecting tight junction formation during parasite invasion and thus inhibiting invasion 16 . Even though such interactions have not been studied in P. vivax, PvAMA-1 DII participation in invasion cannot be ruled out.
A methodology based on identifying conserved sequences having high specific host cell binding capability has been used with P. falciparum and other microorganisms such as Mycobacterium tuberculosis during the last 25 years for characterizing conserved HABPs (cHABP) directly involved in target cell binding 73,86 .
Regarding P. falciparum it has been described that many cHABPs form niches or clefts where they or the receptors can bind 87 87 .
Previous studies have highlighted AMA-1's role as adhesin during invasion 36 . It has been described recently that P. falciparum and T. gondii AMA-1 DI-II domains interact with a RON2-derived protein which might be mediating TJ formation 16,88 . Structural and inhibition studies regarding this interaction have supported the argument concerning TJ/complex formation being crucial for parasite entry 16,40 . However, AMA-1 conditional knockout studies have shown that modified parasites can form the characteristic TJ ring 89 , suggesting varied roles for AMA-1 during invasion. The present work thus found that PvAMA-1 domain I and II were interacting with reticulocytes, i.e. acting as an adhesin.
Two P. vivax cHABPs were identified here (both located in PvAMA-1 domain I) having structural differences regarding their localization; cHABP 21270 formed part of the trough formed between PAN domains and (as in PfAMA-1) could be the critical pocket where the receptor binds. Out of the 8 HABPs identified previously in the whole PfAMA-1 ectodomain 36 , cHABP 4313 was a homolog of cHABP 21270, having 95% similarity. Interestingly, cHABP 4313 inhibited Mrz invasion of erythrocytes by 75% 36 suggesting that this region´s critical role during this stage of the cycle. However, this region does not participate as adhesin during Spz invasion of hepatocytes; rather, the parasite uses different PfAMA-1 regions to interact 90 . P. vivax cHABP 21268 was located on the same face as HABP 21270 but opposite the trough; even though this cHABP did not inhibit rPvAMA-DI-II binding by flow cytometry in 1:20 molar ratio, it displayed a low inhibition of rosette formation (by 25%) (Supplementary Fig. S4).
Bearing cHABP 21268 location in mind and its low inhibitory capacity, it can be suggested that, even though it might specifically interact with a receptor on reticulocyte membrane, it is not the determinant region in protein binding. When PvAMA-1 polymorphisms were located on the 3D structure, it was found that most were located on the opposite face of the identified cHABPs, suggesting that one face of PvAMA-1 is more exposed to the immune system than others 60 . It has been reported that different microorganisms hide relevant regions from the immune system while exposing immunodominant (but non-functional) regions under immune pressure, to protect antigen function during invasion 91 .
To date, PvMSP-1 19 , PvRBP-1a 92 and PvDBP 72 HABPs have been identified which bind specifically to reticulocytes; this, together with characterizing PvAMA-1 regions, has increased current knowledge about P. vivax Mrz specific binding properties used for invading reticulocytes. Future work should address evaluating naturally-acquired antibodies capability for blocking PvAMA-1 binding to reticulocytes and correlate the results with some degree of protection. Future work should also be aimed at identifying critical binding residues which could be modified (mHABP) to increase cHABP immunogenicity and protection-inducing capability, following the methodology proposed for developing a vaccine against P. falciparum 93 .

Constructing recombinant plasmids.
Three PvAMA-1 fragments corresponding to the ectodomain (residues 43-487), domain I and II (residues 43-386) and domain II and III (249-487) were amplified from Plasmodium vivax VCG-I strain genomic DNA for PvAMA-1 expression as a recombinant protein and expression on COS-7 cells (American Type Culture Collection CRL-1651). Gene specific primers were designed using the PvAMA-1 genomic sequence (PVX_092275) available in PlasmoDB 93 as template. Each fragment was amplified using a KAPA HiFi HotStart ReadyMix PCR kit at 25 μL final reaction volume, containing 12.5 μL 2x KAPA HiFi Ready Mix, 1.5 μL of each primer (Supplementary Table S1) at 5 μM concentration and 7.5 μL of nuclease-free water. The amplification conditions for the three products consisted of one 5-min cycle at 95 °C followed by 35 cycles lasting 20 sec at 98 °C, 45 sec at 56 °C and 2 min at 72 °C with a final extension cycle lasting 5 min at 72 °C. Purified products were digested with DraI or PvuII and ApaI restriction enzymes and cloned in pRE4 vector 64 in frame with the HSVgD signal sequence and transmembrane segment 17 to express these fragments on COS-7 cells membrane or digested with BamHI and SalI and cloned in pQE30 vector in frame with histidine tag at the N-terminal to be expressed in E. coli. Each construct was transformed in JM109 cells and selected on Luria Bertani (LB) plates containing ampicillin and confirmed by sequencing. The resulting plasmids were labelled pRE4-PvAMA-1, pRE4-PvAMA-DI-II, pRE4-PvAMA-DII-III, pQE30-PvAMA-1, pQE30-PvAMA-DI-II and pQE30-PvAMA-DII-III.

COS-7 cell culture and transfection. COS-7 cells were cultured in Dulbecco-Modified Eagle Medium
(DMEM) with 10% FCS in a humidified 5% CO 2 incubator at 37 °C. For all transfection experiments, 3 × 10 4 cells were seeded in 3.5-cm diameter wells (40-60% confluent) and then transfected with 300 ng of each recombinant plasmid using FuGENE HD transfection reagent (Promega), following the manufacturer's protocol, in the presence of OptiMEM medium for 48 h in a humidified 5% CO 2 incubator at 37 °C. The cells were grown on coverslips for use in immunofluorescence to verify the correct protein expression on COS-7 cell surface.
Immunofluorescence assay. Transfected cells grown on coverslips were washed in phosphate buffered saline (PBS), fixed with 2% formaldehyde (in PBS) for 15 min at room temperature (RT) and blocked with 10% FCS in PBS for 1 hr to verify correct protein expression on COS-7 cell surface. The cells were stained overnight with anti-DL6 antibodies (Santa Cruz) in 1:1,000 dilution at 4 °C which recognized a proline-rich region adjacent to the mature HSV gD protein transmembrane segment or F8.12.19 monoclonal antibody which recognized a conformational epitope in PvAMA-1 domain III 65 . Fluorescein-conjugated goat anti-mouse antibodies were used as secondary antibodies. Cell nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) and fluorescence was visualized by fluorescence microscope (Olympus BX51) using an Olympus DP2 camera and Fiji software. Transfection efficiency (%) was calculated as total amount of fluorescent COS-7 cells × 100/total amount of COS-7 cells counted in 30 fields, as described previously 67 .
COS-7 cells were tested 48 h after transfection with pRE4-PvAMA-1, pRE4-PvAMA-DI-II or pRE4-PvAMA-DII-III plasmids regarding their binding to normal, neuraminidase-treated, chymotrypsin-treated and trypsin-treated human RBC from UCB (10%) or reticulocyte-depleted erythrocytes (10% haematocrit) from peripheral blood without any treatment for 2 hr at 37 °C. Regarding competition assays, UCB (10%) was pre-incubated with 10 µL peptide 21270 (1 mg/mL) for 1 hr at 37 °C followed by two washes with incomplete DMEM and then added to transfected COS-7 cells. After incubation, the cells were washed twice with PBS. The COS-7 cells with adherent erythrocytes or normocytes (i.e. rosettes) were scored in 20 randomly-chosen fields at 200× magnification. Rosettes were considered positive when adherent erythrocytes covered more than 50% of COS-7 cell surface. COS-7 cells transfected with plasmid pHVDR22 (which expressed DBP RII) were used as positive control 17 while non-transfected cells were used as negative control. At least three independent experiments were performed with replicates.

Soluble recombinant expression and purification. E. coli (JM109) recombinant proteins transformed
with pQE30-PvAMA-1, pQE30-PvAMA-DI-II or pQE30-PvAMA-DII-III were grown in 25 mL LB broth supplemented with 0.1 mg/mL ampicillin and 0.1% D-glucose for 12 hr at 37 °C with constant shaking. This starting culture was used to inoculate 475 mL LB broth containing 0.1 mg/mL ampicillin and 0.1% D-glucose maintained at 37 °C with constant shaking until reaching 0.6-0.8 optical density (OD) at 620 nm. Recombinant protein expression was induced with isopropyl β-D1-thioglactopyranoside (IPTG) at 1 mM final concentration. The cultures were grown for another 16 hr at 30 °C and E. coli cells were harvested by centrifugation. The cell pellet was suspended in lysis buffer (50 mM Tris pH 8.0, 300 mM NaCl, 25 mM imidazole) supplemented with 15 mM 2-β-mercaptoethanol, 0.1 mM ethylene glycol-bis (β-aminoethyl ether)-N,N,N' ,N'-tetraacetic acid (EGTA), 0.1 mg/mL lysozyme, 0.5% Tween 20 and a protease inhibitor cocktail. Bacterial cells were lysed by sonication (Torebeo Ultrasonic Processor 36800). The lysate was spun at 15,000 × g for 30 min at 4 °C and the supernatant was incubated with nickel-nitrylotriacetic acid (Ni 2+ -NTA) agarose resin for purifying recombinant proteins by affinity chromatography. Bound protein was eluted with increasing imidazole gradients (50-500 mM). The eluates were analyzed on SDS-PAGE and by Western Blot and the fractions containing the recombinant protein having a clear single band were pooled. Protein concentration was determined using the BioRad protein assay system (BCA method) and a standard bovine serum albumin (BSA) curve.
ELISA and dot blot assays. The ELISA assay described elsewhere 40 had some modifications. Briefly, 1 µg recombinant protein (rPvAMA-1 or rPvAMA-DII-III) was pre-absorbed onto 96-well plates in coating buffer (15 mM sodium carbonate and 35 mM sodium bicarbonate) and then blocked with 5% BSA in PBS and 1% Tween 20 at 37 °C. The plates were incubated for 1 hr with F8.12.19 monoclonal antibody in 1:16,000 dilution at 37 °C, followed by three washes with 1% PBS-Tween. Peroxidase-conjugated anti mouse IgG was used as secondary antibody, diluted 1:5,000 for 1 hr. Antigen-antibody reaction was detected using a TMB Microwell peroxidase substrate system kit (KPL Laboratories). The reaction was stopped after 10 min incubation with 1 M phosphoric acid. Absorbance was measured at 405 nm using a microtiter reader.
The dot blot assay involved seeding 10 µL rPvAMA-1 or rPvAMA-DII-III on nitrocellulose membrane and then incubating for 15 min in a damp chamber at 37 °C. The membrane was blocked with 5% milk in 0.5% PBS-Tween for 30 min; the membrane was then incubated with F8.12.19 monoclonal antibody (1:16,000) for 1 hr. Alkaline phosphatase conjugated anti-mouse antibody was used as secondary antibody. A BCIP/NBT (Promega) kit was used for revealing the reaction, according to the manufacturer's instructions.
Flow cytometry assays. The flow cytometry-based RBC-binding assay was performed as previously described, with the following modifications 21 . Twenty-five micrograms of each recombinant protein were incubated with 500 µL UCB suspension overnight at 4 °C with constant shaking. Competition assays involved adding each PvAMA-1 DI-and DII-derived synthetic peptide (>99% purity) in a 1:20 (protein:peptide) molar ratio. The erythrocytes were washed with 1% BSA/PBS and then incubated with anti-human CD45 allophycocyanin (APC) (Becton Dickinson) and anti-human CD71-APC-H7 monoclonal antibodies (Becton Dickinson) for 20 min at RT following incubation with mouse anti-histidine phycoerythrin conjugated monoclonal antibody (Miltenyi Biotec). RBC were washed and suspended in 2 mL PBS. One million events were read on a FACSCanto II (BD Bioscience) and the results were analyzed using FlowJo software (TreeStar). DBPRII recombinant soluble protein was used as positive binding control and DBPRIII-IV as negative control. At least three independent experiments were performed with different UCB samples.
Radiolabeled peptide binding assays. 2 × 10 7 RBC from UCB were incubated at RT for 90 min with different concentrations of each radiolabeled peptide (0-800 nM) in the absence (total binding) or presence (non-specific binding) of the same unlabeled peptide dissolved in 4-(2-hydroxyethyl)-1-pi perazineethanesulphonic acid buffered saline (HBS; 200 μL final volume). Following incubation, cells were washed twice with HBS and the amount of cell-bound radiolabeled peptide was quantified on an automatic gamma counter.
Peptides' specific binding activity was evaluated in a binding assay with enzyme-treated RBC from UCB. The RBC were treated as described in the COS-7 cell RBC binding assay. 2 × 10 7 untreated RBC and trypsin-, chymotrypsin-or neuraminidase-treated RBC were incubated at RT for 90 min with 400 nM of each radiolabeled peptide in the absence or presence of the same unlabelled peptide, as described previously 97 .
Determining dissociation constants. Modified binding assays were used for determining HABP 21270 dissociation constants (K d ), and Hill coefficients (nH). Briefly, 1.5 × 10 7 RBC were incubated with increasing concentrations of radiolabeled peptide (0-4,000 nM) in the absence or presence of unlabelled peptide (20 μM), using a 255 μL final volume. Cell-bound radioactivity was quantified as in binding assays 97 . GraphPad Prism software v5 was used for calculating K d and nH, following an interaction model, one by one. Statistical analysis. SPSS v20.0 was used for statistical analysis of RBC-binding assays. Comparative statistics were reported from non-parametric univariate analysis. Differences between means were compared by Kruskal-Wallis tests when comparing multiple groups or Mann-Whitney U-test for comparing two groups.
Ethics approval and consent to participate. All parents from the donors of umbilical cord samples here used signed an informed consent form after receiving detailed information regarding the study's goals. The umbilical cord samples were collected by Bogotá's District Blood Centre (Colombia), following the protocols established by this institution, in accordance with Colombian laws and regulations. All procedures were approved by FIDIC's ethics committee.