Duffy Antigen is Expressed During Erythropoiesis in Duffy Negative Individuals

Summary

The erythrocyte silent Duffy blood group phenotype in Africans is thought to confer resistance to Plasmodium vivax blood-stage infection. However, recent studies report P. vivax infections across Africa in Fy-negative individuals. This suggests that the GATA-1 SNP underlying Fy-negativity does not entirely abolish Fy expression or that P. vivax has developed a Fy-independent red blood cell (RBC) invasion pathway. We show that RBCs and erythroid progenitors from in vitro differentiated CD34 cells and from bone marrow aspirates from Fy-negative samples express a functional Fy on their surface. This suggests that the GATA-1 SNP does not entirely abolish Fy expression. Given these results, we developed an in vitro culture system for P. vivax and show P. vivax can invade erythrocytes from Duffy-negative individuals. This study provides evidence that Fy is expressed in Fy-negative individuals and explains their susceptibility to P. vivax with major implications and challenges for P. vivax malaria eradication.

Graphical Abstract

eTOC blurb

The Duffy gene (Fy) polymorphism has historically been linked to Plasmodium vivax natural susceptibility/resistance, but clinical infections have recently been detected in Fy-negative individuals. Dechavanne and colleagues observe Fy expression during early erythropoietic stages in Fy-negative individuals, which enables P. vivax invasion and helps explain infections seen in Fy-negative individuals.

Introduction

Duffy gene (FY) polymorphism has been prominent historically. In 1968, Donahue et al. used Fya/Fyb serological inheritance patterns and human chromosome 1 centromeric staining features to make FY the first gene mapped to a human autosome¹. Functionally, the Duffy blood group protein (Fy)² also known as Duffy antigen chemokine receptor (DARC) or CD234, was the first identified member of the diverse seven-transmembrane chemokine receptor family, albeit with several interesting structural and functional differences that make its biology unique². Fy expression on both erythroid and non-erythroid cells² modulates homeostatic levels and gradients of chemokines between blood circulation^3–7 and tissues^8–12.

The decades-long observations that Africans and African Americans, who are for the most part Fy-negative, were highly resistant to Plasmodium vivax blood-stage infection^13–18 suggested that interaction with the Fy protein was required for parasite invasion of human red blood cells (RBCs)¹⁹. P. vivax, selectively invades reticulocytes in a process that requires region II of Plasmodium vivax Duffy-binding protein (rPvDBPII) with DARC receptor^20–26. With gene sequencing, a single nucleotide polymorphism (SNP) in the FY gene GATA-1 transcription factor binding site has been shown to inhibit FY erythroid expression^27,28 and homozygosity underlies the “Fy erythrocyte silent” (herein, Fy-negative) phenotype characterized by standard serological methods²⁷. In vitro studies using Fy-dependent P. knowlesi^20,29 show that while the merozoite can orient apically to the RBC surface, the parasite failed to invade Fy-negative erythrocytes. Therefore, the onward formation of a tight mobile junction needed for infection of the RBC failed³⁰. This background has inextricably linked Duffy blood group genetics and malaria. Recent studies now report P. vivax infections in Fy-negative individuals from many malarious African countries^31,32. These observations lead to at least two hypotheses. First, constitutive low-level or periodic Fy expression may occur in Fy-negative people and facilitate occasional blood-stage infection. Second, P. vivax may not interact with Fy exclusively and has developed a Fy-independent RBC invasion pathway. Even if other P. vivax proteins appear to play a role in parasite invasion^33,34 (e.g., reticulocyte-binding proteins) no alternative pathway has been so far identified.

Previous studies have demonstrated a gene dosage effect of the FY GATA-1 SNP that reduces erythrocyte surface expression of Fy^a and Fy^b by 50% in heterozygous individuals^28,35, and others have identified additional coding region SNPs associated with reduced expression of both Fy^a (Fy^a-weak)³⁶ and Fy^b (Fy^b-weak)^37–39; very rare nonsense SNPs and deletions in the FY gene coding region have also been reported (see text with Supplemental Information (SI) Table S1)^40,41. Further, Fy is more highly expressed on reticulocytes versus older RBCs³⁵. With this range of Fy expression phenotypes, we tested the hypothesis that the GATA-1 SNP (FY^ES mutation) may not entirely block Fy protein expression and here we describe how this may influence susceptibility to P. vivax blood stage infection.

Results

Fy expression on Fy-positive and Fy-negative pbRBC

To assess the presence of Fy on peripheral blood red blood cells (pbRBC), two anti-Fy6 antibodies were used: the conventional mouse monoclonal antibody 2C3^42,43 and the camelid single-domain antibody CA111⁴⁴, which overlaps the binding site of 2C3 and possess a higher affinity⁴⁵. We tested whether the anti-Fy6 antibodies recognized Fy on the surface of Fy-negative (FY*B^ES/*B^ES) and Fy-positive RBC (Figures S1, S2 and S3). Working on freshly collected pbRBC, we observed CA111 binding to pbRBC of FY*A/*A, FY*B/*B donors and FY*B^ES/*B^ES pbRBC at different time points (Figure 1A). Collection of different time points from the same donors indicated variations in CA111 binding from 74.0% to 89.4% for FY*B/*B pbRBC, 77.2% to 90.8% for FY*A/*A pbRBC and 11.5% to 47.6% for FY*B^ES/*B^ES. This latter result was markedly higher than previously observed on FY*B^ES/*B^ES donors⁴⁴. Another set of experiments was performed using the conventional anti-Fy6 2C3 and the single-domain anti-Fy6 CA111 (Figure 1B) and anti-Fya and anti-Fyb on 10 other Fy-negative donors (Figure 1C). CA111 detected Fy on the pbRBC from FY*B/*B (N=7) and FY*B^ES/*B^ES donors (N=10, Figure 1B) confirming the observation of Figure 1A. Using the 2C3, Fy protein was detected on the pbRBC of FY*B/*B donors but not on the FY*B^ES/*B^ES donors. As expected, the anti-Fya antibody showed reactivity on pbRBC from FY*A/*B donor only and the anti-Fyb antibody on FY*A/*B and FY*B/*B pbRBC. Taken together, these results suggest that the Fy protein is consistently detected on the pbRBC surface of FY*B^ES/*B^ES by CA111 compared to 2C3 and the anti-Fya and anti-Fyb antibodies used for clinical serology and detection of Fy is in a reduced range in Fy-negative compared to Fy-positive individuals.

Figure 1. Binding of Fy-specific antibodies on Fy-positive and Fy-negative pbRBC.

A - Direct binding of CA111 anti-Fy6 to pbRBC from FY*B/*B, FY*A/*A and FY*B^ES/*B^ES donors. Experiments were performed in triplicate (histograms in red, orange and blue in overlay) ; each donor was tested at three independent time points. Individual donor samples minus CA111 (plus anti-HA mouse antibody and goat anti-mouse PE-conjugated antibody) served as negative controls represented in each histogram in dark. The percentage of cells with CA111 binding (PE: phycoerythrin) over the background is indicated above the gate. The x-folds over negative control (geometric mean of fluorescence intensity (gmfi)) were higher than 48 for FY*B/*B #3 or FY*A/*A #5 and were equal to 6, 5 and 4 for FY*B^ES/*B^ES #1 at the three time points respectively.

B - Binding of 2C3 and CA111 anti-Fy6 to pbRBC from 1 FY*B/*B donor (orange) and 10 FY*B^ES/*B^ES donors (red). Each individual served as its own negative control: no primary antibody, only secondary antibodies (anti-HA (CA111), anti-mouse (2C3)). The x-fold is calculated using the gmfi of the binding of antibodies targeting the Fy6 epitope divided by the gmfi of the binding of the secondary antibody only.

C - Binding of anti-Fya and anti-Fyb to pbRBC from 1 FY*A/*B (green), 1 FY*B/*B donor (orange) and 10 FY*B^ES/*B^ES donors (red). Each individual served as its own negative control: no primary antibody, only secondary antibodies (anti-human-Fya or -Fyb). The x-fold is calculated using the gmfi of the binding of antibodies targeting the Fy epitope divided by the gmfi of the binding of the secondary antibody only.

rPvDBPII binding on FY*B^ES/*B^ES pbRBC

Given the unexpected interaction of the CA111 Fy6-specific antibody with FY*B^ES/*B^ES pbRBC, we tested in additional experiments the host-parasite functionality of the Fy protein by evaluating rPvDBPII binding to pbRBC. As expected, a high binding of rPvDBPII on FY*A/*B pbRBC was observed, while a much lower binding of rPvDBPII was also observed on FY*B^ES/*B^ES pbRBC (Figure 2A). We then exposed both Fy-positive and Fy-negative pbRBC to rPvDBPII, washed and eluted pbRBC bound proteins and then probed them with a polyclonal antibody to PvDBP as previously described^46,47. This specific detection revealed a single protein band of approximately 35kDa - consistent with the detection of the rPvDBPII protein – from both FY*B^ES/*B^ES and FY*A/*B pbRBC on a Western blot (Figure 2B).

Figure 2. rPvDBPII binding on FY*B^ES /*B^ES pbRBC.

A - rPvDBPII binding on FY*A /*B and FY*B^ES /*B^ES pbRBC. The pink cross provides a point of reference for cell populations in negative controls (no rPvDBPII and secondary antibody only). The percentage of rPvDBPII binding cells over background is indicated above the gate.

B - rPvDBPII binding and dissociation to Fy-positive and Fy-negative pbRBC - pbRBC were exposed to rPvDBPII (20 μg/mL, 2 hr.; room temperature) to allow protein binding. The western blot of dissociated proteins from FY*B^ES/*B^ES (left) and FY*A/*B (right) pbRBC was probed with a polyclonal anti-PvDBP. The western blot included here is from an original intact gel and blot. This experiment was repeated 3 more times on fresh RBC from 2 FY*B^ES/*B^ES donors (data not shown). Those data were complemented with flow cytometry data.

C - Chymotrypsin binding assessment - Without chymotrypsin treatment (light gray boxes); with chymotrypsin treatment (dark gray boxes); individual data points (filled black circles); FY*A/*B - without chymotrypsin (N=12); with chymotrypsin (N=5); for FY*B^ES/*B^ES - without chymotrypsin (N=9); with chymotrypsin (N=5); evaluated by non-parametric paired Wilcoxon rank-sum test. FY*A/*B (P-value = 0.0016); FY*B^ES/*B^ES (P-value = 0.038).

D - rPvDBPII dissociation assessment of non-specific binding - No significant reduction in rPvDBPII binding was observed across increasing NaCl concentrations from 0.15M to 1.0M; addition of 1.5M NaCl caused significant reduction in rPvDBPII binding (Pearson Chi² test, p=0.014: Pearson Chi²=6.0). Each curve represents a different donor: FY*B/*B #3 (black circle), FY*A/*B #4 (white square), FY*B^ES/*B^ES #1 (black triangle).

E - Comparative binding between folded (gray boxes) and unfolded rPvDBPII (dark boxes; same amino acid sequence) to pbRBC - Non-parametric Wilcoxon matched pairs sign rank test evaluated the median differences between rPvDBPII folded vs unfolded; n=8 for each Fy-positive and n=7 for each Fy-negative donors. FY*A/*A: p=0.012; FY*B/*B: p=0.012; FY*B^ES/*B^ES #1: p=0.018; FY*B^ES/*B^ES #2: p=0.042. Unfolded rPvDBPII interaction with pbRBC from four study participants showed no significant difference (non-parametric median test Pearson Chi²: p=0.853).

Given historically consistent demonstration of the chymotrypsin-sensitive nature of the Fy protein^20,48–51, we then tested the reproducibility of rPvDBPII binding to both FY*A/*B and FY*B^ES/*B^ES pbRBC after chymotrypsin treatment. We examined the chymotrypsin sensitivity of rPvDBPII binding for one FY*A/*B and one FY*B^ES/*B^ES (Donor #1) at multiple independent times. Results consistently showed that the binding of rPvDBPII to both FY*A/*B and FY*B^ES/*B^ES pbRBC were significantly higher before than after chymotrypsin treatment (Figure 2C).

To further evaluate the relative affinity of rPvDBPII interaction to Fy-positive and Fy-negative pbRBC, we tested the dissociation of PvDBPII from pbRBC with increasing NaCl concentrations (Figure 2D). While we observed that the relative binding of rPvDBPII to FY*B^ES/*B^ES pbRBC was markedly lower than that observed for FY*B/*B and FY*A/*B, dissociation of rPvDBPII from FY*B/*B and FY*A/*B versus FY*B^ES/*B^ES pbRBC had a similar pattern across increasing NaCl concentrations (0.15M, 0.3M, 0.6M and 1M NaCl). At 1.5M NaCl the pbRBC for all FY genotypes were lysing and therefore rPvDBPII signal decreased across all samples.

In a more specific test of rPvDBPII binding to Fy-positive and Fy-negative pbRBC, we compared the binding of properly folded rPvDBPII to partially denature rPvDBPII (same amino acid construct) on pbRBC (Figure 2E). Results showed that irrespective of the Duffy genotype, denatured rPvDBPII (dark boxes) compared to properly folded (light boxes) had significantly reduced binding to Fy-positive and Fy-negative pbRBC. Taken together these results show that rPvDBPII bind specifically to Fy-positive and Fy-negative pbRBC and are consistent with CA111 anti-Fy6 binding experiments.

Variable rPvDBPII binding on pbRBC over time

To observe the persistence of the interactions between rPvDBPII and pbRBC over time, we examined multiple samples from both Fy-positive and Fy-negative study participants in a time-course study (Figure 3A). The pbRBC samples were collected and examined over 282 days for one FY*B/*B donor (orange), one FY*A/*A donor (green) and two FY*B^ES/*B^ES donors (red and blue). Among the FY*B^ES/*B^ES pbRBC sampling time points (7 samples for both FY*B^ES/*B^ES #1 (red) and FY*B^ES/*B^ES #2 (blue)), five samples showed rPvDBPII binding at least 2-fold above the negative control; both FY*B^ES/*B^ES donors showed evidence of rPvDBPII binding 5-fold above the negative control. The variability observed in rPvDBPII binding among the Fy-negative donor samples was also observed for both Fy-positive donors (regardless of their genotype). Overall, the rPvDBPII binding (mean fold-increase over negative control and range) was 32.4 for FY*B/*B (1.8 –110.9), 17.2 for FY*A/*A (4.9–47.7), 1.9 for FY*B^ES/*B^ES #1 (0.7 – 5.9), 2.4 for FY*B^ES/*B^ES #2 (0.5 – 6.7). Interestingly, across this series of samples from the four pbRBC donors featured, the higher levels of Fy expression on FY*B^ES/*B^ES pbRBC were comparable to the lower level of rPvDBPII binding for the FY*A/*A pbRBC (Figure 3A). As further evidence of Fy protein expression on the FY*B^ES/*B^ES pbRBC surface, we were able to immunoprecipitate, after solubilizing the pbRBC membranes, the Fy protein using CA111 anti-Fy6 and detected it using the commercial anti-Fy [ACKR1] antibody (Figure 3B).

Figure 3. Specific rPvDBPII binding on Fy-positive and Fy-negative pbRBC over time.

A - rPvDBPII binding on Fy-positive and Fy-negative pbRBC over time – Binding of rPvDBPII to pbRBC was followed for up to 282 days (10 months): one FY*B/*B (donor #3, orange), one FY*A/*A (donor #5, green), two FY*B^ES/*B^ES (donor #1, red; donor #2, blue); binding tested in triplicate (minimum, maximum and mean value) and represented as the x-fold increase over negative control. Those data complete the Fy-expression in the cross sectional study represented in Fig1B. Each donor served as its negative control for each time point. The same antibodies and flow cytometer (Attune NxT) were used in each experiment.

B - Specific capture and detection of the Fy protein from the surface of pbRBC.

This figure presents direct evidence of the specific capture, immunoprecipitation and detection of the Fy protein from the surface of the pbRBC of FY*B/*B and FY*B^ES/*B^ES donors. In the left panel, nDARCIg panel – the monomeric and dimeric nDARCIg species were distinguishable. Experimental panels – Fy6-specific CA111-based capture of the Fy protein from pbRBC of FY*B/*B (Donor #3; day 112), FY*B^ES/*B^ES (Donor #1; day 112) and FY*B^ES/*B^ES (Donor #2; day 189). Western blot detection of CA111-captured Fy protein was performed using polyclonal anti-ACKR1 (LS Bioscience). See SI Methods Section 7 for western blot raw data. This experiment was repeated twice on fresh RBC from 2 FY*B^ES/*B^ES donors (data not shown). Those data were complemented with flow cytometry data.

C - Blocking of the rPvDBPII binding to Fy-positive and Fy-negative pbRBC – Upper Panel – Inhibition of rPvDBPII binding with addition of the anti-Fy6-specific CA111. Lower Panel – Inhibition of rPvDBPII binding with addition of the nDARCIg. Only the experiments showing a decrease of at least two times were represented on the graphics. Experiments included in the statistical analysis: n=9 for FY*B/*B#3, n=7 for FY*A/*A#5, n=7 for FY*B^ES/*B^ES#1, n=7 for FY*B^ES/*B^ES#2 - Non-parametric Wilcoxon matched pairs sign rank test evaluated the differences of rPvDBPII binding without or with blocking for each individual. The dashed line represents equivalent binding compared to the negative controls. Each dot is the mean of 2 to 3 technical replicates. Data of rPvDBPII binding in Fig3C are also represented in Fig3A with error bars. Two connected dots refer to one experiment.

To further confirm the specificity of rPvDBPII binding to pbRBC in this time-course study, we used both CA111 and nDARCIg (a chimeric protein including Fy and IgG1-Fc region⁵²) in competition assays to block rPvDBPII access to the RBC. The Fy-specific CA111 nanobody (Figure 3C left panels) decreased rPvDBPII binding from 25.2-fold (range, 5.8;110.9) over negative controls to 2.1-fold (0.2;12.1) for both Fy-positive donors (P-value =0.001 for FY*B/*B; P-value =0.001 for FY*A/*A). Under the same conditions, rPvDBPII binding decreased from 2.0-fold (0.5;6.4) over negative controls to 0.8-fold (0.2;1.9) for both Fy-negative donors (P-value =0.002 for FY*B^ES/*B^ES #1; P-value =0.003 for FY*B^ES/*B^ES #2). In the partner experiments (Figure 3C right panels), nDARCIg was observed to decrease rPvDBPII binding to Fy-positive pbRBC from 25.2-fold (5.8; 110.9) over negative controls to 2.81-fold (0.5;7.7) (P-value=0.005 for FY*B/*B; P-value=0.005 for FY*A/*A). For the FY*B^ES/*B^ES pbRBC, nDARC decreased rPvDBPII binding to pbRBC from 2.0-fold (0.5;6.4) over negative controls to 1.0-fold (0.3;2.8) (statistically significant P-values all <0.05 for both FY*B^ES/*B^ES samples). Thus, both CA111 and nDARCIg blocked the binding of rPvDBPII to FY*B^ES/*B^ES pbRBCs, providing evidence of Fy-specific binding of rPvDBPII to FY*B^ES/*B^ES pbRBC.

Fy protein expression on erythroid progenitor cells in vitro

Given a wide range of results pointing to P. vivax preference for infection of reticulocytes^53–61 and the apparent low, variable expression of Fy protein on Fy*B^ES/*B^ES pbRBC (Figures 1A, 1B and 3A), we postulated that this expression pattern reflects the declining Fy expression on older pbRBC³⁵. Indeed, recent evidence from murine studies showing that the Fy protein is expressed at higher levels on pro-erythroblasts and normoblasts, compared to mature RBC⁶². We initiated these in vitro studies by expanding CD34^pos (expressed on megakaryocyte erythrocyte progenitors^63–67) from the peripheral blood of one FY*B/*B donor and FY*B^ES/*B^ES Donors #1 and #2⁶³. At Days 11, 15 and 18 of differentiation, cells were collected to monitor erythroid lineage maturation using the surface markers CD36, CD71, Glycophorin A and Band 3 and query Fy protein expression (Figure 4A and Figure 4B). As expected, during the differentiation process, CD36 and CD71 were decreasing, whereas Glycophorin A and Band 3 increased among the most mature RBCs. No difference in the maturation stages from the different donors was observed over time (Figure 4A). Expression of the Fy protein during these differentiation time points was monitored by either flow cytometry (CA111 binding, Figure 4B) and/or Western blot (CA111 immunoprecipitation; anti-ACKR1 detection, Figure 4C and FigureS4). For the FY*B/*B donor, at Day11, 48.9% of the cell population expressed the Fy protein at the cell surface whereas from Day15 to Day18 >96% of the population expressed the Fy protein. From the two different FY*B^ES/*B^ES donors, the expression of the Fy protein was observed in vitro as early as Day11 of maturation. Consistent with our previous results on pbRBC, the Fy protein expression was lower on FY*B^ES/*B^ES cells (11.1 and 35.5%) than on FY*B/*B cells (48.9%). However, we observed that significantly higher proportions of the erythroid precursor cells from FY*B^ES/*B^ES donors tested positive for the Fy protein by Day18 (49.7% for FY*B^ES/*B^ES #1 and 83.8% for FY*B^ES/*B^ES #2) (Figure 4B) than was observed previously for pbRBC (highest proportion of Fy-positive pbRBC from Figure 3A - 6% for FY*B^ES/*B^ES #1 and 7% for FY*B^ES/*B^ES #2). Expression of the Fy protein on erythroid precursors differentiated in vitro at Day18 was further confirmed by immunoprecipitation and Western blot (Figure 4C).

Figure 4. Fy protein expression during in vitro CD34^pos erythroid precursors differentiation.

The CD34^pos erythroid precursor cells were expanded and differentiated for over 21 days.

A - Monitoring of CD34 positive cells differentiation in vitro. CD34^pos erythroid precursor cells were collected from the peripheral blood of 3 donors (1 Fy-positive and 2 Fy-negatives). In vitro differentiation was monitored by the presence/absence of the surface markers CD36, CD71, Glycophorin A and Band 3 on CD34^pos erythroid precursor cells. As expected, during the differentiation process, the cells were losing the CD36 and CD71 signals whereas Glycophorin A and Band 3 signals were increasing. Overall, the cells from all the donors were observed to be at comparable stages over time.

B - Fy expression on differentiated CD34^pos from FY*B/*B and FY*B^ES/*B^ES donors. At Day11, Day15 and Day18, the expression of Fy protein was detected by flow cytometry using CA111 anti-Fy6. For each donor, the percentage of positive cells for the binding of CA111 is annotated in the black box (population in red within the CA111-PE positive gate) in comparison to its negative control (population in black, e.g. secondary antibody only). mfi: mean of fluorescence intensity.

C, D - Expression of Fy protein in erythroid precursors isolated from peripheral blood of FY*B/*B and FY*B^ES/*B^ES donors (C) and from bone marrow of FY*A/*B^ES and FY*B^ES/*B^ES donors (D). Differentiated cells were harvested on Day18, membrane proteins were extracted and immunoprecipitated with CA111 anti-Fy6. The protein captured by Fy-specific CA111 were detected with a commercial polyclonal anti-ACKR1 antibody by western blot. All original western blots and re-organization of lanes to present salient results are described SI Methods Section 7. The experiments of the Fig4C were repeated at Day11 and Day18 on differentiated CD34^pos from two FY*B^ES/*B^ES donors (data not shown). The experiments of the Fig4D were repeated at Day11 and Day15 on differentiated CD34^pos from two FY*B^ES/*B^ES donors and Day15 on differentiated CD34^pos from one other FY*B^ES/*B^ES donor (data not shown). Those data were complemented with flow cytometry data.

The same in vitro differentiation was performed on human bone marrow aspirates (HBMAs)⁶³. From expanded CD34^pos bone marrow cell populations of both Fy-positive and Fy-negative samples, we were also able to immunoprecipitate and detect the Fy protein (Figure 4D).

These studies confirm the ability of individuals with FY*B^ES/*B^ES genotype to express the Fy protein in erythroid progenitor cells cultured in vitro. Nevertheless, these results required confirmation of Fy expression in erythroid progenitor cells ex vivo.

Ex vivo Fy protein expression on bone marrow erythroid precursors.

Fy expression was evaluated ex vivo with CA111 anti-Fy6 (n=5 Fy-positive, n=2 Fy-negative) and rPvDBPII (n=2 Fy-positive, n=2 Fy-negative) binding on erythroid precursor cells from bone marrow samples (Figure 5). On the same samples, we monitored the expression of the following cell surface markers as accepted indicators of erythroid precursors: CD34^63–67, CD45 (marker to differentiate erythroid from lymphoid development⁶⁴), CD105 (marker of human erythrocyte progenitor (hEP)^67–69), and CD71 (transferrin receptor preferentially expressed on young reticulocytes⁷⁰) (Figure 5A). The cell size and expression of these markers on bone marrow subpopulations were similarly observed among all Fy phenotypes. We further assessed Fy protein expression in bone marrow erythroid precursors using CA111 and rPvDBPII recognition (Figure 5B). The highest levels of anti-Fy6 binding was observed within the CD34^neg/CD45^neg/CD105^mid/CD71^pos subpopulation (skewed toward reticulocytes; CD105 declining) for Fy-negative donors and CD34^neg/CD45^neg/CD105^neg/CD71^pos subpopulation for Fy-positive donors. Further detailed comparative analyses of bone marrow samples from individuals representing specific FY genotypes are provided in Figures S5 and S6.

Figure 5. Ex vivo expression of Fy protein among erythroid precursor sub-populations from bone marrow aspirates.

A, B - In the bone marrow, all stages of erythroid precursors are present except mature erythrocyte (that are circulating in the peripheral blood). The gating strategy is detailed in the SI and summarized in the table up left (pos: positive, mid: middle, neg: no expression). The density (mean of fluorescence intensity divided by the size of the cells) of each CD markers as well as CA111 anti-Fy6 or rPvDBPII binding to Fy protein was represented in different graphics. The Fy-negative donors were represented in dashed lines in the graphics. The profile of CD marker expression was similar between all Fy genotypes (A) whereas Fy expression appeared to be earlier during erythropoiesis in Fy negative compare to Fy positive individuals (B).

C - Expression of Fy protein in Fy*B^ES/*B^ES CD45 negative erythroid precursors from ex vivo bone marrow. CD45 negative cells (erythroid precursors in bone marrow) of three bone marrow samples were sorted by flow cytometry. The membrane proteins of the sorted cells were then immunoprecipitated with the CA111 anti-Fy6 and revealed by Western blot with a commercial polyclonal anti-ACKR1 antibody (arrow). Fy-positive and Fy-negative showed a signal for the Fy protein in erythroid precursors. The western blot included here is from an original intact gel and blot. See SI Methods Section 7 for western blot raw data. This experiment was complemented with flow cytometry data.

Finally, we confirmed expression of Fy protein by CA111-capture and commercial anti-ACKR1-detection from both Fy-negative and Fy-positive donors in the erythroid precursor bone marrow population (Figure 5C; Figure S7).

P. vivax in vitro invasion into Fy-positive and Fy-negative erythroid cells.

With the Fy protein detected by Fy-specific antibodies and rPvDBPII on genotypically Fy-negative erythroid cells, we determined whether cryopreserved P. vivax field isolates (asynchronous) would interact with and/or invade host cells. Short-term invasion assays used peripheral blood of North American Fy-positive and Fy-negative participants as target cells seeded with stock P. vivax cells. In order to distinguish between P. vivax re-infection of stock culture cells and new invasion into target cells, we labeled the target cell populations with CFSE (carboxyfluorescein succinimidyl ester). Imagestream analyses were performed after 72 hours of culture. Overall assessment of the RBC invasion experiments by imaging flow cytometry provides an overview for selecting populations of cells down to assessments of individual parasite-target cell interactions (Figure 6, Figure S8, S9, S10 and S11, Table S2). Figure 6A shows separation of single cells (black data points) versus multiple (gray data points) cells by aspect ratio and area in the brightfield image. Results in Figure 6B show that strategic CSFE labeling of target RBCs in a control cell population without exposure to parasites serving as a background control. Results from the invasion assays show newly infected Fy-positive and Fy-negative target cells (CFSE+ and Hoechst+) in Figures 6C and 6D, respectively. To confirm P. vivax infected target cells, we examined individual cells of interest (circled cells) with bright field and fluorescent images (Figure 6C and 6D), with Fy-positive target cells (Figures 6E1–6 correspond to circled/numbered cells in Figure 6C) and Fy-negative target cells (Figures 6F1–6 correspond to circled/numbered cells in Figure 6D). These results show P. vivax merozoites interacting with and invading Fy-positive (Figures 6E1 to 6E4) and Fy-negative (Figures 6F1 to 6F4) target cells. Figures 6E5 & 6F5 are uninfected donor cells (stock culture) and Figures 6E6 & 6F6 are infected donor cells (stock culture). Our experiments also show that cells from FY*B/*B compared to FY*B^ES/*B^ES donors are more readily invaded by P. vivax by a factor of 2.2 (Table S2).

Figure 6. In vitro invasion of pbRBCs from Fy-positive and Fy-negative North Americans by Malagasy P. vivax isolate.

6A. single cells (black data points) versus multiple (gray data points) cells by aspect ratio and area in the brightfield image; 6B; Control image using uninfected cells previously stained with CFSE. Y-axis CFSE; X-axis Hoechst; 6C. AMP2016.14 (CFSE[-]) mixed with Fy-positive, FY*B/*B target cells (CFSE[+]); 6D. AMP2016.14 (CFSE[-]) mixed with Fy-negative, FY*B^ES/*B^ES target cells (CFSE[+]); 8E1- E6. Individual cells identified in Panel 6C from AMP2016.14 added to FY*B/*B (Fy-positive). 8F1- F6. Individual cells identified in Panel 6D from AMP2016.14 added to FY*B^ES/*B^ES (Fy-negative). Additional data and plots are provided in the SI (Figure S8, Table S2, Figure S9, Figure S10, Figure S11).

Discussion

Given our findings on P. vivax infection of Fy-negative people from Madagascar^71,72 and reports from other malarious sites across Africa^31,32,73, we wanted to investigate how parasite and host invasion proteins interacted with Fy-negative RBCs. We therefore assessed whether or not two different anti-Fy6 antibodies and rPvDBPII bind to the surface of different erythroid target cells (pbRBC; expanded CD34 cell populations; bone marrow). Our data show that Fy is expressed at the surface of pbRBC and this expression is variable in both Fy-negative and Fy-positive individuals. Using competitive inhibition studies, we demonstrated that adding of CA111 or nDARCIg significantly reduced rPvDBPII binding to Fy-positive pbRBC and essentially blocked rPvDBPII binding to Fy-negative pbRBC. These results suggest that rPvDBPII binding is focused on the anti-Fy6 epitope and thus limited to the Fy protein.

Our initial studies on pbRBC, showed that Fy expression on Fy-negative pbRBC was positive at some sample collection time points and that these pbRBC could interact with rPvDBPII at levels similar to RBC from FY*A/*A people. Because of P. vivax preference for infection of reticulocytes^33,53–60, we studied Fy expression across a broader time frame of erythroid development. Human and murine studies show higher levels of Fy protein expression on pro-erythroblasts, normoblasts and reticulocytes compared to mature RBC^35,58,62. Additionally, Malleret et al. have demonstrated that immature reticulocytes (CD71^pos), mostly found in the bone marrow and matured in the spleen⁵⁹ are preferentially invaded by P. vivax compared to older reticulocytes and erythrocytes (CD71^neg), than in the peripheral blood. Recent studies emphasize that the bone marrow and especially the spleen, the latest accounting for more than 98% of the estimated total-body P. vivax biomass, are important reservoirs of P. vivax infection in humans and non-human primate model systems^74–76. Since the spleen harbor more mature reticulocytes than in the bone marrow, the distribution of the P. vivax biomass is inconsistent with the relative paucity of invasion-receptive reticulocytes in the spleen. One possibility could be that upon P. vivax infection, the erythropoiesis within the bone marrow is impaired, resulting in extramedullary hematopoiesis (EMH)⁷⁷ induction within the spleen and then allowing the expansion of parasite biomass beyond the bone marrow. With this focus on the importance of reticulocytes as P. vivax invasion-targets, interactions between reticulocyte binding proteins and CD71 and CD98 heavy chain suggest additional molecular components of the P. vivax erythrocyte invasion mechanism⁷⁸.

Our studies on earlier erythroid stages focused on enriching CD34^pos megakaryocyte erythrocyte progenitors that circulate in the peripheral blood, allowing them to mature in vitro under conditions favoring erythroid development⁶³. We purposefully performed a time-course study on the same Fy-negative donors to evaluate the variability of erythroid precursors released in the peripheral circulation over time. Interestingly, results showed that significantly higher proportions of the erythroid precursor cells from FY*B^ES/*B^ES donors were positive for Fy protein detection than their pbRBC.

With the observation of Fy protein expression from the CD34 cell maturation experiments, we repeated our studies on bone marrow aspirate samples as the site for natural erythroid cell development and also an important reservoir of P. vivax. We monitored the expression of CD34, CD45, CD105 and CD71 expression as markers distinguishing erythroid from lymphoid maturation. Results from these bone marrow studies showed Fy protein expression on early-stage cells committing to erythroid development in both Fy-positive and Fy-negative people. Our observations suggest that Fy expression initiates in an early erythroid subpopulation (CD34^pos/CD45^pos/CD105^mid/CD71^neg) in both Fy-positive and Fy-negative people. Because Fy expression drops for Fy-negative individuals whereas the expression increased for Fy-positive individuals, we speculate that, before proerythroblast differentiation and transition to GATA-1^68,79, Fy expression may be driven by another transcription factor, such as GATA-2 and this may contribute to the results we have observed. Consistent with Tournamille et al. findings (i.e., compromised FY gene expression associated with the GATA-1 FY^ES SNP)²⁷, the amount of Fy protein begins to decay earlier in Fy-negative than in Fy-positive erythroid cells. The observed decline of Fy signal in CD34^neg/CD45^neg/CD105^mid/CD71^pos cells, appears to be consistent with RBC extrusion of its nucleus, after which RBC precursors are no longer able to express their genes and the cell can only be equipped with the gene products expressed prior this important developmental event.

The potential that bone marrow could be the source of reticulocytes for in vitro cell culture is becoming an increasingly important area of P. vivax research. Fernandez-Becerra et al. have expanded CD34^pos hematopoietic stem cells to generate CD71^pos reticulocytes for in vitro P. vivax culture⁸⁰. We followed in this direction by using human bone marrow aspirates from US donors, to provide target cells for in vitro propagation of P. vivax field isolates from the Madagascar communities where we have previously observed significant frequency of Fy-negative P. vivax infections^71,72. Human bone marrow aspirates have enabled in vitro propagation of two field isolates as source parasite material for the short-term invasion assays presented here. Using these in vitro propagated P. vivax, we have shown that both Fy-positive and Fy-negative RBCs can be invaded by P. vivax.

Limitations

Although our work provides a clear demonstration that the dogma about Fy-negative individuals no longer holds and represents significant advances for understanding P. vivax malaria, we recognize limitations that future studies should address. While we have shown that P. vivax can invade Fy-negative cells in vitro, it was not possible to demonstrate that RBC invasion was Fy protein dependent and completely rule out the possibility that P. vivax can use an alternative invasion pathway. With the development of monoclonal antibodies and peptide-based inhibitors, interrupting interactions between PvDBP, PvEBP2, PvRBP and other P. vivax merozoite proteins, more specific invasion inhibition studies are warranted. For this to occur, it will be necessary to refine P. vivax in vitro invasion studies to allow more extensive studies. The demonstration of P. vivax invasion of Fy-negative cells in vitro opens potential that laboratory-adapted P. vivax strains could be used to study a range of candidate protein-protein interactions necessary for blood stage infection. Additionally, while the focus of our experiments provided assessment of Fy expression from multiple different sample types (pbRBC; expanded CD34 cell populations; bone marrow aspirates), we were not able to examine Fy-expression from splenic reticulocytes of Fy-negative and Fy-positive individuals. Emerging studies that have probed P. vivax infections outside of the peripheral blood are critically important to advance current perspectives on vivax malaria pathogenesis and evaluate Fy genotype variation in future studies of this nature.

Conclusions

Our results provide a mechanism to explain why Fy-negative people are not completely resistant to P. vivax blood-stage malaria^3,4 by showing that Duffy antigen can be variably expressed in subjects with the GATA1 polymorphism associated with Fy negativity. With the peak of Fy expression during early reticulocyte stages and declining as RBCs enter peripheral circulation, our findings support observations that the bone marrow and sites of extramedullary erythropoiesis are important for P. vivax blood stage propagation. Our results are consistent with observations that P. vivax infections in Fy-negative individuals are usually detected by PCR, not blood smear microscopy. P. vivax infection in Duffy negative individuals expands human populations potentially capable of P. vivax transmission. Our results also raise questions regarding P. vivax malaria pathogenesis, especially given a wide range of clinical observations showing bone marrow involvement, from in vitro studies^78,81, individual case histories^82–84 and cohort studies^85–87. With potential significant involvement of the bone marrow, it becomes easier to understand how considerable anemia and other markers of hematological disruption are out of proportion with low levels of observed P. vivax peripheral blood parasitemia (summarized in Baird³³ and Silva-Filho et al.⁸⁸). Given evidence that peripheral blood infections significantly underestimate the total body biomass of a P. vivax infection and therefore global prevalence, closer examination of P. vivax infection and disease biomarkers (lactate dehydrogenase and other circulating parasite antigens)^88,89 is required to more accurately estimate the burden of vivax malaria in Fy-negative Africa and globally.

STAR Methods

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Peter A. Zimmerman (paz@case.edu).

Materials availability

Proteins generated in this study will be shared by the lead contact upon request if necessary with a Material Transfer Agreement.

Data availability

Any additional information required to reanalyze the data reported in this paper will be available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Study Participants and Sample Collection

Study protocols were approved by the University Hospitals of Cleveland Institutional Review Board (#08-03-33 and #09-90-195). All methods were carried out in accordance with guidelines and regulations included within our protocols. Informed consent was obtained from all subjects, and for subjects who were under 18, from a parent and/or legal guardian. Eleven Fy-positive (three females and eight males) and sixteen Fy-negative persons (nine females and seven males) were included (Table S1); all Fy-negative persons were homozygous for the negative GATA-1 mutation (rs2814778; FY*B^ES/*B^ES). All blood and bone marrow samples were processed within 2 hours of collection.

METHOD DETAILS

DNA Extraction and PCR-based Genotyping

All study participants were genotyped by previously described methods⁷¹. Genomic DNA was extracted from the whole blood samples using the QIAGEN QIAmp DNA Blood Kit following recommended protocols with a starting blood volume of 200 μL and an elution volume of 200 μL of Buffer AE (Valencia, CA).

PCR amplifications of Duffy blood group gene sequences were performed in reaction mixtures (28μL) with 3 μL of genomic DNA, 180 μM each dNTP, 67mM Tris-HCl (pH 8.8), 6.7 mMMgSO4, 16.6mM(NH₄)₂SO4, 10mM 2-mercaptoethanol, 0.1 μM each primer (Primers inclusive of promoter (rs2814778) and FY*A/FY*B (rs12075) snps - forward primer, Duffy-200up 5′-CAGGCAGTGGGCGTGGG-3′; reverse primer, Duffy +730dn 5′-CTGCTAGCTAGGATACCCAG-3′; Primers inclusive of FY*B/FY*X (rs34599082) snp - forward primer, FYBXup 5′-AGCACTGTCCTCTTCATGCTTT-3′; reverse primer, FYBXdn 5′-GCAGAGCTGCGAGTGCTAC-3′) and 2.5 units of thermostable DNA polymerase under the following conditions: 95 °C for 2min, followed by 40 cycles of 95 °C (30 s), 60 °C (30 s), and 72°C (90 s) and a final extension at 72 °C (5 min) (PCR products 912 and 1,033 bp). Following PCR amplification, products were further processed by a ligation detection reaction (LDR). The LDR was performed in a reaction mixture (15 μL) containing 20mMTris-HCl buffer (pH 7.6), 25 mM potassium acetate, 10mM magnesium acetate, 1mM NAD+, 10mM DTT, 0.1% Triton X- 100, 13nM each LDR probe,1 μL of PCR product, and 2 units of Taq DNA ligase (New England BioLabs). LDR probes consisted of six allele-specific oligonucleotides and three fluorescently labeled conserved-sequence oligonucleotides. The allele-specific probes contained an MTAG sequence for further hybridization with complementary sequence oligonucleotides bound to Luminex FlexMAP fluorescent microspheres; conserved-sequence probes were 5′ phosphorylated and 3′ biotinylated.

Sequences of the LDR probes used were as follows:

Duffy null promoter snp (rs2814778) at nucleotide −67 (t, wild-type; c, erythrocyte silent): PRO ntT MTAG_A018: 5′-ACACTTATCTTTCAATTCAATTACcattagtccttggctcttat-3′ PRO ntC MTAG_A020: 5′- CTTTCTCATACTTTCAACTAATTTtcattagtccttggctcttac-3′ PRO common: 5′-phosphate-cttggaagcacaggcgctg-biotin-3′.

Codon 42 snp (rs12075) associated with Fya and Fyb phenotypes (g, Fya; a, Fyb): FY*A ntG MTAG-A022: 5′- CAAACAAACATTCAAATATCAATCttcccagatggagactatgg-3′ FY*B ntA MTAG-A026: 5′- TACATTCAACACTCTTAAATCAAActtcccagatggagactatga-3′ Codon42 common: 5′-phosphate-tgccaacctggaagca-biotin-3′.

Codon 89 snp (rs34599082) associated with Fyb or the Fyb^weak phenotypes (c, Fyb; t, Fyb^weak): FY*B ntC MTAG-A028: 5′- CACTTAATTCATTCTAAATCTATCtgcttttcagacctctctcc-3′ FY*X ntT MTAG-A034: 5′- ACTTATTTCTTCACTACTATATCAtgcttttcagacctctctct-3′ Codon89 common: 5′-phosphate-gctggcagctctgccctggct-biotin-3′.

Protein expression

CA111 anti-Fy nanobody

Variable domain of a heavy chain only (VHH), nanobody of camelids. This nanobody includes the CA52 VHH sequence⁴⁴ and an HA (human influenza virus hemagglutinin)-tag. This nanobody was kindly provided by Olivier Bertrand (National Institute of Blood Transfusion (INTS), Paris, France)⁴⁴. CA111 was obtained by subcloning the CA52 VHH sequence into a Novagen pET 28-b derived plasmid allowing expression in the cytoplasm of SHuffle (C3029H) cells. This plasmid encodes (from 5’ to 3’) a polyhistidine tail, a thrombin cleavage site, the VHH, and an HA-tag; it is used for routine subcloning of other VHHs using the PstI and Eco91I sites⁹⁰.

The resulting nanobody exhibits the following characteristics.

• Interferes with the interleukin-8 binding to Fy on RBCs.

• Interferes with P. vivax infection of red blood cells (short term culture).

• Upon linkage to a solid substrate, this nanobody acts as a powerful adsorbent to purify native Fy in a single step from a detergent extract of cells.

• Pepscan analysis identified 22FEDVW26 as the peptide sequence of the linear epitope.

A single dromedary was immunized with the N-terminal extracellular domain constructs of Fy expressed in E. coli (referred to as ECD1; Amino acids 6 [H] to 63 [P] – HRAELSPSTENSS QLDFEDVWNSSYGVNDSFPDGDY[G/D]ANLEAAAPCHSCNLLDDSALP)⁴⁴. This domain covers the region to which Fy6 epitope, chemokines and the P. vivax Duffy binding protein (PvDBP) bind (amino acids 19 to 25 – 19QLDFEDV25) and carries the polymorphic G/D site at amino acid 42, responsible for the Fya/Fyb allotypes. From the immunized animal, a VHH library of dromedary lymphocytes was exposed to ECD1 to yield several Fy-specific VHHs. Smolarek et al. focused on one VHH, referred to as CA52. The linear epitope of CA52 exhibits capacity to recognize the glycosylated Fy protein present on human cells, although the immunogen was a non-glycosylated polypeptide.

To further validate the specificity of the CA111 antibody, we assessed its reactivity on K562 cells expressing or not the Fy antigen and Saimiri RBC⁹¹. Blood samples (2 mL) from Bolivian squirrel monkeys (S. boliviensis), collected in EDTA vacutainers, was obtained from the Biologics Production Program, Michael E. Keeling Center for Comparative Medicine and Research, The University of Texas MD Anderson Cancer Center (Protocol number: 00000451-RN01-AR001). The blood was shipped on ice overnight and used in our studies on the day of arrival.

rPvDBPII recombinant protein

Region II of Salvador I strain PvDBP (rPvDBPII; 37.6 KDa) was expressed as inclusion bodies in Escherichia coli and then refolded and purified as previously described^21–23. The protein purified before refolding served as a negative control.

Duffy Antigen Receptor for Chemokine chimeric protein construct (nDARCIg Fyb)

Briefly, as previously described⁵², HEK293H cells were co-transfected with pCDM8-DARC-Fc (first N-terminal 60 amino acids of Fy) and pRc/CMV-TPST (human sulfotransferase).

Fy detection by flow cytometry

Erythrocyte binding assays were performed based on previously developed strategies and reagents^46,47. Peripheral blood RBC (1×10⁶) from human or Saimiri) or K562 cells⁹² were first blocked for 30 min with PBS 1% BSA. Anti-Fy antibodies were then incubated for 20 min at 37°C at the following concentrations: CA111 (30 μg/mL), murine monoclonal antibody, 2C3 (10 μg/mL)^42,43, anti-FyA (ref: 808 186 BioRad, 1:10 dilution) or anti-FyB (ref: 808 191 BioRad, 1:10 dilution) antibodies. A mouse anti-HA antibody (Biolegend; 30 μL of a 1:1000 dilution in PBS 0.2% BSA; 20 min at 37°C) followed by a goat-anti-mouse PE-conjugated antibody (eBioscience; 40 μl of a 1:80 dilution in PBS 0.2% BSA; 20 min at 37°C in the dark) for CA111, an anti-mouse for 2C3 or an anti-human for anti-Fy antibodies were used to detect the binding. Following 2 washes in PBS 0.2% BSA, a total of 100,000 cells (all samples in triplicate) were analyzed by flow cytometry. Between each staining step, pbRBC were washed two times with PBS 0.2% BSA. For ex vivo fresh bone marrow samples, a direct anti-HA tag phycoerythrin (PE)- tagged antibody was used to reduce the background (Miltenyi, dilution 1:11). Similar procedures were followed for other Fy-specific antibody reagents. For the experiments on the maturation of CD34+ hematopoietic stem cells, the same protocol was performed and at least 50,000 cells were analyzed by flow cytometry. In evaluation of bone marrow precursors, the BSA concentration change from 1% to 2% and at least 500,000 cells were evaluated to ensure sufficient data capture for erythroid precursor subpopulations of interest.

rPvDBPII binding monitored by flow cytometry

Peripheral blood RBC (1×10⁶) were first blocked for 30 min with PBS 1% BSA and then further incubated with rPvDBPII Sal1 (20μg/mL final concentration; 60 μL of a 1:20 dilution in PBS 0.2% BSA; 30 min at 37°C). After 2X washes with PBS 0.2%BSA, these cells were incubated with a rabbit polyclonal anti-PvDBP serum (30 μL of a 1:50 dilution in PBS 0.2% BSA; 20 min at 37°C). Cells were washed again 2X with PBS 0.2%BSA and then incubated with rabbit phycoerythrin (PE)-tagged antibody (Sigma; 40 μL of a 1:32 dilution in PBS 0.2% BSA; before 20 min at 37°C in the dark). After 2 washes in PBS 0.2% BSA and a final wash in PBS, cells were subjected to analytical flow cytometry^46,47. Erythrocyte binding assays were further optimized by adjusting the blocking concentrations for BSA.

Unfolded rPvDBPII was used as the same concentration as folded rPvDBPII and the protocol is identical.

Erythrocytes were pre-treated with chymotrypsin (1 mg/mL) 30 min at 37 °C. The enzyme treatment was stopped by washing five times with PBS 0.2% BSA. A blocking step (1% BSA for an hour at 37°C) were performed before rPvDBPII binding.

For the rPvDBPII dissociation assessment, the rPvDBPII binding protocol as described above was performed and increasing NaCl concentrations from 0.15M to 1.0M were added 750 μL of each gradually. The measures were acquired on Attune NTX.

Competition assays

Blocking of rPvDBPII binding:

the competition assay was performed using pre-incubated erythrocytes with CA111 (30 μg/mL) or nDARCIg (60 μg/mL) for an hour at 37°C before rPvDBII binding following the same steps as described above.

Blocking of CA111 binding:

the competition between CA111 and 2C3 was performed on fresh FY*B/*B pbRBC, FY*B^ES/*B^ES pbRBC or enriched reticulocytes from FY*B^ES/*B^ES donor. The enrichment protocol was described elsewhere⁹³. Briefly, after leucocyte depletion using NWF syringe filter, cells were resuspended in 3 ml high-KCl buffer (pH 7.4) and then incubated 20 minutes at room temperature on a rocker. After centrifugation, cells were resuspended in 500 μL of high-KCl buffer and layered over 300 ul 19% Nycodenz-KCl solution in a 1.1 mL microtube. Tube was centrifuged at 3,000g for 30 min without braking. Reticulocytes at the interface were harvested and washed 2 times in PBS. After a blocking step, cells were pre-incubated with increasing concentrations of CA111 (molar ratio from 0 to 25) an hour at 37°C. 2C3 was then directly added (no wash) at 5 μg/mL. The detection of the 2C3 binding was following the same steps as described above.

rPvDBPII Erythrocyte Capture Assay

This assay was previously described⁸⁵. Briefly, ~10⁶ cells were collected from packed RBC and 10 μg of recombinant PvDBPII was added. After 2 hours of incubation at room temperature, the preparation was layered on top of 500 μL of dibutylphtalate and centrifuged at 11,000 rpm for 30 sec. The pellet was collected, and the proteins bound at the RBC surface were eluted adding 20 μL of 1.5 M NaCl drop wise and shaking. After incubation (5 min), this procedure was repeated with 1.0 M of NaCl and then 0.3 M of NaCl. The preparation was centrifuged at 14,000 rpm for 2 min. The proteins present in the supernatant were separated on a Tris-Glycine SDS 4–20% gel (Biorad) in reducing conditions and then transferred onto a PVDF membrane. The membrane was blocked for one hour with TBS 0.1% tween, 5% milk, incubated with rabbit anti-PvDBPII (1: 2,000), washed, then followed by the secondary antibody anti-rabbit HRP (Thermo Scientific). Chemiluminescent signal was detected using SuperSignal (Pierce) on X-ray film.

Fy immunoprecipitation by CA111 and Western Blot detection

Solubilization and protein extraction –

Packed RBC (~5 mL); cells matured during the in vitro erythroid differentiation; and CD45 negative sorted-cells from the bone marrow were lysed by three cycles of freeze/thaw followed by sonication. After adding 5 mM Tris·HCl, pH 8.0 with protease inhibitor (Sigma Aldrich), the protein preparation was centrifuged 20 min at 16,000 × g at 4°C. The pellet was resuspended twice in 0.1 M Na₂CO₃ pH 11.5 with protease inhibitor and centrifuged 20 min at 16,000 × g at 4°C. To solubilize the membrane proteins from the pellet, 1mM triton X100 + protease inhibitors were added and incubated at room temperature for 30 min.

Immunoprecipitation of Fy antigen -

CA111 was covalently coupled to magnetic beads (Dynabeads M-270 Carboxylic Acid, Invitrogen) for 30 min at room temperature with slow rotation. EDC at 100mg/ml (1-ethyl-3-[3-dimethylaminopropyl] carbodiimide) was dissolved 100mM MES (2-(N-morpholino) ethanesulfonic acid)) buffer (pH 5.0) for the activation reaction and incubated for 2 hr at 4°C with slow rotation. The solubilized extract was incubated with the activated beads for 1h. Beads were resuspended in Laemmli buffer for SDS gel and western blot.

Western blot -

The beads + extract solutions were separated on a Tris-Glycine SDS 4–20% gel (Biorad) in reducing conditions and were then transferred onto a PVDF membrane. The membrane was blocked for one hour with TBS 0.1% Tween 5% milk. Polyclonal rabbit anti-ACKR1 (2ug/ml; LS Bioscience) and HRP anti-rabbit antibody (diluted 1/10,000 into 10 ml TBS 0.1% Tween 5% milk; Thermo Scientific) were used to probe Fy from the immunoprecipitated elution. The signal was detected on X-ray films using SuperSignal (Pierce).

Specificity of CA111 -

Tests of specificity were performed on recombinant proteins by direct Western blotting, or immunoprecipitating cells from patient bone marrow and subsequent Western Blot. For the direct probe of the Western Blot, 0.1 μg of nDARCIg and rCXCR2 were loaded on a gel; 0.2 μg/mL of CA111 were used to probe the Western blot and reveal the signal. For the immunoprecipitation experiment, 0.5 μg of rCXCR2 were used. The anti-CXCR2 antibody (R&D system, 0.5 μg/mL) was used to probe recCXCR2 on a Western Blot as a positive control.

Erythroid differentiation of CD34 positive cells

Isolation of CD34 positive cells was performed from peripheral blood donors (FY*B/*B and FY*B^ES/*B^ES donors; #1 and #2) and from bone marrow samples. Mononuclear cells were isolated by Ficoll Plaque Plus (GE Healthcare) and frozen in 10% DMSO. The thawed cells were positively selected for CD34 by magnetic beads (human CD34 MicroBead Kit; Miltenyi Biotech). Between 10,000 and 100,000 cells were obtained after selection and differentiated in culture over 21 days. A detailed protocol is published by Jingping Hu et al.⁶³. The in vitro differentiation was monitored by flow cytometry with 400,000 cells per day (Day 11, Day 15, Day 18) monitoring surface markers CD36 (BD Biosciences), CD71 (BD Biosciences), Band 3 (American Research Product) and Glycophorin A (Life Technology).

Monitoring maturation and expansion of CD34+ Hematopoietic Stem Cells (HSCs).

Mononuclear cells were isolated by Ficoll Paque Plus (GE Healthcare) and frozen in 10%DMSO. The thawed cells were positively selected for CD34 by magnetic beads (human CD34 MicroBead Kit; Miltenyi Biotech). Between 10,000 and 100,000 cells were obtained after selection and differentiated in culture over 21 days. A detailed protocol is published by Jingping Hu et al⁶³. For the FY*B/*B donor and 2 FY*B^ES/*B^ES donors, the cells were counted by light microscopy during the in vitro differentiation at different time points (Day 0, 11, 15 and 18).

The in vitro differentiation was monitored by flow cytometry with 400,000 cells per day (Day 11, Day 15, Day 18) monitoring the presence/absence of the surface markers CD36 (BD Biosciences), CD71 (BD Biosciences), Band 3 (American Research Product) and Glycophorin A (Life Technology). As expected, during the differentiation process, the cells are losing the CD36 and CD71 expressions whereas Glycophorin A and Band 3 are increasing. Overall, the cells from all the donors were observed to be at comparable stages over time.

Bone marrow cell preparation and staining

Fresh bone marrow was collected in heparin. Erythroid precursors from fresh bone marrow samples were isolated by Ficoll Paque Plus (GE Healthcare) separation. Cells were first blocked for 30 min with FcR blocking reagent (Miltenyi – 20 μL per sample) following manufacturer’s recommendations. rPvDBPII and CA111 binding experiments were performed as described above. Sub-populations of erythroid precursors were defined using directly-labeled antibodies - anti-CD45 PerCP Cy5.5 (Becton Dickinson), anti- CD34 FITC, anti-CD105 APC, anti-CD71 APC Cy7 (all from Biolegend) and an anti-Band3 PE (American Research Product). Data were acquired with a BD Biosciences LSR II (San Jose, CA) or a Thermo Fisher Attune Nxt (Waltham, MA) operated within manufacturer’s specifications, which was tested with performance beads. To ensure sufficient relevant data, we aimed to acquire 500,000 cell events. Data were analyzed with FlowJo 10.0 (Becton Dickison & Company (BD), Ashland, OR) or WinList 9.0 (Verity Software House, Topsham, ME). Two types of analysis were performed.

Flow cytometric analysis of erythroid progenitor subpopulations in bone marrow

Forward scatter (FSC-A) and side scatter (SSC-A) plots were used to select erythroid populations then standard gating was done to select positive and negative population from orthogonal quadrants on bivariate plots or univariate histograms. For each of the samples analyzed, unstained cells were used to evaluate cell auto-fluorescence. Unstained and single antibody stained cells, or unstained or single stained antibody-binding beads, were used to set fluorescence compensation. Negative controls for CD234 (Fy) were “secondary only” with either CA111 or rPvDBPII omitted in otherwise complete assays. Secondary and tertiary reagents were PE labeled anti HA-tag for CA111and PE labeled goat anti-rabbit, respectively for rPvDBPII (with a polyclonal rabbit anti- PvDBP as secondary antibody). Results collected following comparisons with these negative controls allowed for calculating fold-increase of the geometric mean of fluorescence intensity (CA111 or rPvDBPII) over the negative control or for calculating the fraction of positive cells.

Second, for data collected on bone marrow samples, we analyzed a subset of bone marrow donors to show the level of expression of CD34, CD45, CD105, CD71, and Fy. The antibodies and staining protocol was as above. However, the sub-populations were gated multi-dimensionally using a series of bivariate plots to create a sequential series of erythroid differentiation states. The fluorescence value of each parameter was divided by a correlated measure of cell size (forward scatter) providing a measure of antigen density. The expression level of Fy was made more accurate by subtracting the fluorescence of the negative control on a subpopulation basis (median fluorescence – median background fluorescence).

Erythroid cell protein extraction and Western blot analysis

Extracting the Fy protein from the surface of erythroid cells was conducted to specifically capture evidence of FY gene expression. These experiments were performed on pbRBCs from Fy-positive and Fy-negative donors. Target cell populations included pbRBC, expanded CD34 cell populations and bone marrow. Details of protein extraction, immunoprecipitation and Western blot analyses are provided earlier in the Star Method.

Competition assay of CA111 or 2C3 binding to Fy-negative RBCs in the presence of nDARC (Fc conjugated DARC)

Blood derived from three Fy negative individuals was incubated with soluble nDARC to compete with the binding to RBC bound DARC. Samples were taken on multiple days and each time tested in quadruplicates for each nDARC concentration (range: 1000 nM to 0.48 nM, twofold serial dilutions) as well as N=12 for 1000 nM nDARC, 0 nM nDARC and 2nd antibody staining alone for higher statistical power. Samples from the three individuals were tested on three different days. Samples were collected on an Attune Next using the appropriate lasers for HO, PE and FITC as well as Forward Scatter and Side Scatter. For each sample, 200,000 events were collected.

Imagestream Analysis

Samples from invasion experiments were collected on an Imagestream IS100 instrument. Target cells were labeled with CFSE, while the donor cells were CFSE negative. Each experiment was seeded using peripheral blood of North-American Fy-positive or Fy-negative participants as target cells and stock P. vivax isolates (stock culture: AMP2016.14 and AMP2016.36) at a ratio of 80% to 20% (or 90% to 10%) respectively for a total RBCs 7.2×10⁷. After 72h incubation the cells were stained with Höchst 33342 and then fixed with 4 % PFA in 1 × PBS. For each data file 200K events were collected, for most samples we performed four technical replicate (resulting to ~1000K events). All datasets were collected in Extended Depth of Field (EDF) mode, resulting in an extended depth of field when focusing the cells. Only focused cells were further analyzed by first separating them by Area and Aspect Ratio in the brightfield channel, allowing the distinction of single versus multiple cells in an image. Single cells were then separated by intensity in CFSE and HO staining. Five areas are distinguished in the CFSE/HO plot. Unlabeled uninfected donor cells (CFSE-/HO−), infected donor cells (CFSE-/HO+), uninfected labeled target cells (CFSE+/HO−), infected target cells (CFSE+/HO+) and white blood cells (CFSE+/HO>10⁵). The HO positive cells detected in the donor cell population versus the uninfected donor cells reflect the parasitemia of that individual sample at the time of collection. In addition, a cutoff value was determined based on a CI95 level of the mean fluorescent intensity of uninfected cells (HO channel). Cells that were above that threshold were deemed positive.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses were performed using STATA version 13.0 and are described on the legend of each concerned figure. Additionally, Graphpad Prism 10 was used for statistical analysis of the SPR data as well as the competition assay. Imagestream data was analyzed by the IDEAS Software version 6.2.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Recombinant proteins
rPvDBPII	Gift from Dr. Niraj Tolia	https://doi.org/10.1038/nsmb.2088
unfolded rPcDBPII	Material produced at CWRU	https://doi.org/10.1038/nsmb.2089
nDARCIg	Construct is a gift from Chetin Chitnis, material produced at CWRU	https://doi.org/10.1371/journal.pmed.0040337
rCXCR2	NovusBio	H00003579-P01
Antibodies
conventional mouse 2C3 anti-Fy6	Gift from Dr. Yves Colin	https://doi.org/10.1046/j.1365-2141.2003.04533.x
single domain camelid CA111 (HA tag) anti-Fy6	Gift from Dr. Olivier Bertrand and Dr. Yves Colin	https://doi.org/10.1007/s00018-010-0387-6
Anti-ACKR1 / DARC Antibody (aa1-63)	LS Biosciences	LS-C371451
HRP anti rabbit antibody	Milipore	AP187P
Mouse Anti-HA (IgG1)	Bioloegend	901502
Goat anti-mouse PE	eBioscience	12-4010-82
Goat anti-mouse APC	BD Biosciences	550826
Polyclonal rabbit anti-PvDBP (Rabbit 2)		https://doi.org/10.1371/journal.pmed.0040337
Goat anti-rabbit APC	Invitrogen	31984
Goat anti-rabbit PE	Sigma	P9537
CD45 MicroBeads, human	Miltenyi	130-045-801
anti-CXCR2	R&D system	MAB331
Anti-Fy a (FY1)	Biorad	808-186
Anti-Fyb (FY2)	BioRad	808-191
anti-GlycoA PE-Cy7	BD Biosciences	563666
anti-CD36 APC	BD Biosciences	550956
anti-Band3-PE	American Research Product	IBGRL-BRIC6-PE
anti-CD71 APC	BD Biosciences	551374
anti-CD105 PE-Cy7	BD Biosciences	568753
anti-CD71 FITC	BD Biosciences	555536
anti-CD34 BV421	BD Biosciences	562577
anti-CD45 APC-Cy7	BD Biosciences	557833
anti-CD71 BV786	BD Biosciences	563768
mouse anti-HA tag	Miltenyi	130-098-404
mouse anti-His tag	Qiagen	34650
mouse anti-PE	Invitrogen	12-4010-82
mouse anti-PE	Jackson ImmunoResearch	115-115-164
Immunoprecipitation
Protease inhibitor	Sigma Aldrich	4693159001 ; P8849
Dynabeads M-270 Carboxylic Acid	Invitrogen	14305D ; 10003D
EDC 1-(3-Dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride	ACROS organics	171440010
Trypsin	Sigma Aldrich	T6567
IgG Elution Buffer	Thermo	21004
Neutralizing buffer 1 M tris Base pH 9 10X	Sigma Aldrich	1083820100
Western Blot
Amersham standard autoradiography cassette, 24 × 30 cm	Dutscher	RPN11643
Amersham Hyperfilm™ ECL	Dustcher	28-9068-37
Mini-P Tetra Cell,2precast gel	Biorad	1658005
Dithiothreitol (DTT), 5 g	Biorad	1610611
Mini-PROTEAN TGX,4–20%,30μl 10 well,10	Biorad	4561093
Precision Plus Protein All Blue Standards, 500microliter, 50 applications	Biorad	1610373
Immun-Blot PVDF Membranes, 7 × 8.4 cm, 10	Biorad	1620174
Clarity Max ECL substrate,100ml	Biorad	1705062
Surebeads Protein G Magnetic Beads 3 ml	Biorad	1614023
Surebeads Magnetic Rack, 16 tube Holder	Biorad	1614916
Tris, 500 g	Biorad	1610716
Glycine, 1kg	Biorad	1610718
SDS, 100 g	Biorad	1610301
Bio-Safe Coomassie Stain, 1 L	Biorad	1610786
CD34+ Maturation
CD34 MicroBead Kit	Miltenyi	130-046-702
holotransferrine	Sigma Aldrich	T0665 100mg
Human insulin	Sigma Aldrich	I9278 5ml
heparin	Sigma Aldrich	2106-10VL
AB serum	Sigma Aldrich	H4522-100ML
Hydrocortisone (dexamethasone)	Sigma Aldrich	D4902-25MG
Stem Cell Factor (SCF)	Peprotech	300-07
IL3	Peprotech	100-64
EPO	Peprotech	200-03
IMDM + glutamax	Gibco Thermo Fisher Scientific	31980030
Others
Chymotrypsin	Roche Diagnostic	11418467001
Bovine Serum Albumin Fraction V	Sigma Aldrich	810531
Ficoll paque plus	Sigma Aldrich	GE17-1440-02
Nycodenz	MP Biomedicals	157750
Primers
Duffy forward	Sigma Aldrich	5′-CAGGCAGTGGGCGTGGG-3′
Duffy reverse	Sigma Aldrich	5′-CTGCTAGCTAGGATACCCAG-3′
FY*B/FY*X forward	Sigma Aldrich	5′-AGCACTGTCCTCTTCATGCTTT-3′
FY*B/FY*X reverse	Sigma Aldrich	5′-GCAGAGCTGCGAGTGCTAC-3′
LDR probes
PRO ntT MTAG	Sigma Aldrich	5′-ACACTTATCTTTCAATTCAATTACcattagt ccttggctcttat-3′
PRO ntC MTAG	Sigma Aldrich	5′-CTTTCTCATACTTTCAACTAATTTtcattagt ccttggctcttac-3′
PRO common	Sigma Aldrich	5′-cttggaagcacaggcgctg-biotin-3′
FY*A ntG MTAG	Sigma Aldrich	5′-CAAACAAACATTCAAATATCAATCttccca gatggagactatgg-3′
FY*B ntA MTAG	Sigma Aldrich	5′-TACATTCAACACTCTTAAATCAAActtccca gatggagactatga-3′
Codon42 common	Sigma Aldrich	5′-tgccaacctggaagca-biotin-3′
FY*B ntC MTAG	Sigma Aldrich	5′-CACTTAATTCATTCTAAATCTATCtgcttttc agacctctctcc-3′
FY*X ntT MTAG	Sigma Aldrich	5′-ACTTATTTCTTCACTACTATATCAtgcttttc agacctctctct-3′
Codon89 common	Sigma Aldrich	5′-gctggcagctctgccctggct-biotin-3′
Surface Plasmon Resonance
CM5 Sensor chip	Cytivia	29104988
Anti-DARC [2C3]	Absolute Antibody	https://absoluteantibody.com/product/anti-darc-2c3/Ab00893-10.0_human_igg1/standard/
ImageStream
CellTrace CFSE Cell Proliferation Kit	Thermo Fisher	https://www.thermofisher.com/order/catalog/product/C34554
Höchst 33342	Life Technologies	H3570
Software
FlowJo version 10.0	Becton Dickison & Company (BD)	https://www.bdbiosciences.com/en-us/products/software/flowjo-v10-software
BD FACSDiva™ Software version 8.0	Becton Dickison & Company (BD)	https://www.bdbiosciences.com/en-us/products/software/instrument-software/bd-facsdiva-software
WinList 9.0	Verity Software House	https://www.vsh.com/products/winlist/index.asp
STATA version 13.0	StataCorp LLC	https://www.stata.com/products/
Graphpad Prism version 10	GraphPad software, Inc.	https://www.graphpad.com/features
IDEAS Application Version 6.2.187	Amnis part of EMD Millipore	https://ideas.com/

Highlights.

• Duffy blood group protein (Fy) is naturally expressed on Fy-negative red blood cells.

• Fy expression is higher on erythroid precursors than on red blood cells.

• Fy protein is expressed in early erythropoietic stages in Fy-negative individuals.

• P. vivax invades Fy-positive and Fy-negative erythroid cells in vitro.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.