Red blood cell tension protects against severe malaria in the Dantu blood group

Abstract

Malaria has had a major effect on the human genome, with many protective polymorphisms-such as the sickle-cell trait-having been selected to high frequencies in malaria-endemic regions^1,2. The blood group variant Dantu provides 74% protection against all forms of severe malaria in homozygous individuals^3-5, a similar degree of protection to that afforded by the sickle-cell trait and considerably greater than that offered by the best malaria vaccine. Until now, however, the protective mechanism has been unknown. Here we demonstrate the effect of Dantu on the ability of the merozoite form of the malaria parasite Plasmodium falciparum to invade red blood cells (RBCs). We find that Dantu is associated with extensive changes to the repertoire of proteins found on the RBC surface, but, unexpectedly, inhibition of invasion does not correlate with specific RBC-parasite receptor-ligand interactions. By following invasion using video microscopy, we find a strong link between RBC tension and merozoite invasion, and identify a tension threshold above which invasion rarely occurs, even in non-Dantu RBCs. Dantu RBCs have higher average tension than non-Dantu RBCs, meaning that a greater proportion resist invasion. These findings provide both an explanation for the protective effect of Dantu, and fresh insight into why the efficiency of P. falciparum invasion might vary across the heterogenous populations of RBCs found both within and between individuals.

Extended Data Figure 1. Erythrocytic cycle of malaria parasites.

Illustration of the erythrocytic stage of the malaria parasite. (a) The merozoites undergo repeated rounds of asexual multiplication, progressing through ring, trophozoite and schizont stages. (1) This cycle starts when merozoites contact, attach and successfully invade RBCs in the circulation. The invasion period lasts for less than a minute and we are only able to follow the dynamics of such a fast event with real-time live microscopy^,. The merozoite has a slightly ovoidal shape of 1 micron in diameter and is adapted for invasion of erythrocytes. The apical end of the parasite contains secretory organelles called rhoptries and micronemes that release proteins for helping merozoite’s internalisation. In the host red cell, the parasite develops and multiplies, digesting haemoglobin, protected from immune attack. In the case of Plasmodium falciparum, the erythrocytic cycle lasts about 48 hours, and infected cells progress from (2) the ring stage (first 16 hours) throughout (3) trophozoite stage (around 16-36 hours) and, finally to (4) the schizont phase (lasts a few hours). The infected schizont then ruptures (5) releasing 15-30 daughter merozoites ready to infect new nearby red cells. In each cycle, less than 10% of parasites develop into the sexual form of the parasite called gametocytes (6). (b) Detailed illustration of the parasite invasion process into the red blood cell (1) described further in the Materials and Methods, involving the “pre-invasion” phase (contact, merozoite reorientation which triggers RBC membrane deformation, and tight attachment of the merozoite to the RBC membrane), the “invasion” phase (initiation of invasion, penetration, complete internalization of the merozoite, and resealing of the RBC membrane), and the echinocytosis phase (formation of echinocyte). Icons adapted from ©biorender.com

Extended Data Figure 2. Invasion process across Dantu genotype groups studied by timelapse video microscopy.

(a) RBC detachment force from a merozoite was measured using optical tweezers across genotype groups. RBCs attached to merozoites were pulled using optical traps, and the adhesive forces at the merozoite-RBC interface were quantified by evaluating the elastic morphological response of the RBC as it resisted merozoite detachment. Mean and standard deviation (SD) in Supplementary Table 3. Six biologically independent samples per genotype group were tested obtaining 21 events for non-Dantu, 19 for Dantu heterozygote, and 24 for Dantu homozygote. Importantly, the experimentalist was blinded to the RBC genotype group. The median is indicated by the middle red line in the boxplots, with the 25^th and 75^th percentiles indicated by the tops and bottoms of each plot, while whiskers denote total data range. If the median is not centered in the box, it shows sample skewness. Pairwise comparisons between genotype groups were performed using the two-sided Mann-Whitney U test. (b) The degree to which merozoites deformed RBCs during invasion was given by a simplified four-point deformation scale (0, 1, 2, and 3), based on the most extreme degree of deformation achieved (Weiss et al., PLoS Pathog 2015). The degree of deformation was compared across genotype groups with no significant difference noticed, and between successful and failed invasions. The percentage of RBCs undergoing strong deformations (score 2/3) is significantly higher in case of successful invasions, while RBCs that were contacted but not invaded experience weak deformations (score 0/1). Data from 155 cells for non-Dantu, 191 for Dantu heterozygote, and 233 for Dantu homozygote. Number of successfully invaded cells: 53 for non-Dantu, 43 for Dantu heterozygote, and 41 for Dantu-homozygote. The visual assessment was done with blinded data by two different experimentalists. (c) Successful invasions are usually followed by a reversible echinocyte phase that lasts between 5 and 11 minutes (Gilson and Crabb, Int J Parasitol, 2009), until the recovery of newly infected RBC biconcave shape. The time of echinocytosis was in agreement with literature and not significantly different across genotype groups. Data from 53 cells non-Dantu, 43 Dantu heterozygote, and 41 Dantu-homozygote.

Extended Data Figure 3. Distribution of reticulocytes and RNA concentrations across Dantu genotypes.

Reticulocyte counts, and concentrations of RNA extracted from reticulocytes, were compared across Dantu genotypes. Reticulocyte count data was tested for 8 non-Dantu, 7 Dantu heterozygotes and 7 Dantu homozygote individuals, while RNA concentrations were measured in 9 non-Dantu, 7 Dantu heterozygotes and 7 Dantu homozygote individuals. The median is indicated by the middle red line in the boxplots, with the 25^th and 75^th percentiles indicated by the tops and bottoms of each plot, while whiskers denote total data range. Statistical comparison across the three genotype groups was performed using a one-way ANOVA test, while all pairwise comparisons between genotype groups were performed using the Tukey HSD test correcting for multiple pairwise comparison tests using Benjamini-Hochberg FDR. Significant differences were observed in reticulocyte count (non-Dantu vs. Dantu homozygote p=0.0023) and RNA concentrations (non-Dantu vs. Dantu homozygote p=0.0015; non-Dantu vs. Dantu heterozygote p=0.0088). ** p < 0.01; * p < 0.05.

Extended Data Figure 4. Plasma membrane profiling by tandem mass tag (TMT)-based MS3 mass spectrometry.

The impact of the Dantu polymorphism on RBC membrane protein expression levels was quantified using mass spectrometry. (a) Hierarchical cluster analysis of all proteins quantified and annotated as described in the Methods. Fold change was calculated for each donor by (signal:noise (donor) / average signal:noise (non-Dantu). (b) Proteomic quantification of markers shown in Figure 2a (3 biologically independent samples per genotype group). All markers were quantified by proteomics apart from GYPB. Statistical comparisons of quantitative protein expression across Dantu genotype groups were performed using two-tailed t-test with Benjamini-Hochberg multiple hypothesis correction: * p< 0.05, **p< 0.01. All p-values are listed in Supplementary Table 4.

Extended Data Figure 5. Representative membrane fluctuation spectra for non-Dantu, Dantu heterozygous, and Dantu homozygous red blood cells.

Example of contour detection and flickering spectra across genotype groups. (a) Contour of the RBC (dashed blue line), with inner and outer bounds used in image analysis (green lines). (b) Contour of the RBC. (c) Mean square fluctuation amplitudes for non-Dantu (green line), Dantu heterozygote (orange line), and Dantu homozygote (purple line) RBCs. Fitted modes 8-20. The error bars (not shown for clarity) were calculated as $SD/\sqrt{(n\times dt)}$, where SD is the standard deviation, n total number of frames, and dt the time gap between frames.

Extended Data Figure 6. Relationship between biophysical properties in non-Dantu and Dantu homozygote RBCs.

(a) Scatter plot showing correlation between tension and radius in non-Dantu and Dantu homozygote RBCs. The shaded points in the background are all the data considered for non-Dantu (249) and Dantu (247) RBCs from six different biological replicates. The big marks in the foreground represent the mean and standard deviation in tension and radius of the six samples for non-Dantu and Dantu RBCs. There is a linear inverse relation between radius and tension, where we observe that RBCs with higher tension have lower radii. Radius change is very small (0.3 μm) and, we believe, equatorial radius decrease is due to a shape change caused by the increased tension, and the two biophysical parameters have no different fluctuation modes. (b) The impact of tension on RBC deformation during preinvasion, induced by merozoites contacting RBCs, was compared across Dantu genotype groups. RBCs having tension above the tension threshold tended to be weakly deformed (scores 0 and 1), whereas RBCs with tensions below the threshold were more strongly deformed (scores 2 and 3). Deformation scores as defined in Weiss et al. PLoS Pathog. 2015 (Reference 10 of the manuscript).

Extended Data Figure 7. Reduction of membrane tension both in non-Dantu and Dantu homozygous RBCs with phloretin.

Biophysical properties in non-Dantu (a,b) and Dantu homozygous (c,d) RBCs after phloretin treatment **p < 0.01. Phloretin treatment causes a decrease in tension without affecting bending modulus at 150 (p = 0.0015) and 200 (p = 1.72 ×10⁻⁴) μM for both non-Dantu and Dantu samples. Above 200 μM concentration of phloretin, most RBCs become crenated and cannot be used for flickering spectroscopy. Phloretin has an effect on RBC tension only when it is present in the medium, i.e. RBCs recover their normal tension when washed. (a-b) Data from about 30 cells from 3 biologically independent non-Dantu samples. (c-d) Data from 60 cells from 4 biologically independent Dantu samples. Between untreated and phloretin 150 μM (p = 0.01) and between untreated and phloretin 200 μM (p = 0.0022) using the two-sided Mann-Whitney U test.

Extended Data Figure 8. Comparing parasite invasion and biomechanical properties of frozen and fresh RBCs.

Comparison of parasite invasion efficiency and biomechanical properties in frozen vs fresh RBCs. (a) The invasion efficiency of P. falciparum laboratory strain, 3D7, was compared across frozen and fresh RBCs (n=6 frozen and n=14 non-Dantu, 12 Dantu heterozygote and 12 Dantu homozygote fresh biologically independent RBC samples per genotype group were tested). The percentage of parasitised RBCs that successfully invaded each genotype group was measured using a flow cytometry-based invasion assay. Boxplots indicate the median (middle line) and interquartile ranges (top and bottom of boxes) of the data, while whiskers denote the total data range. Statistical comparison across the three genotype groups was performed using a one-way ANOVA test, while pairwise comparisons between genotype groups were performed using the Tukey HSD test, with significant differences in 3D7 invasion observed in frozen RBCS (non-Dantu vs. Dantu homozygote p=0.001) and in fresh RBCs (non-Dantu vs. Dantu homozygote p=0.001). **p < 0.01; *p < 0.05. (b) Membrane flickering spectrometry enabled measurement of RBC biomechanical properties (bending modulus, tension, radius, and viscosity) of fresh (n = 53) and frozen (n = 51) RBCs from the same donor. No statistically significant differences were detected between the two conditions for all the measured biophysical properties. Pairwise comparisons were performed using the two-sided Mann-Whitney U test; p Bending modulus = 0.1, p Tension = 0.6, p Radius = 0.7, p Viscosity = 0.6.

Extended Data Figure 9. Decoupling tension and bending modulus with flickering analysis.

To test our ability to decouple tension and bending modulus from our data through the flickering analysis, we have taken the 20 highest tension and the 20 lowest tension cells from our database and shown that on analysing the fluctuation power spectra of these, which cover a wide enough range of q-values that both tension and bending moduli can be robustly extracted. (a-b) Boxplots for the tensions and bending moduli of the 20 cells with extreme high and extreme low tensions. While there is an obvious significant difference in tension (p=4.0302 × 10⁻¹³, two-sided Mann-Whitney U test), bending modulus is similar. (c) This is also evident from the overlapping of the two spectra for the high modes where a bending-dominated regime prevails, whereas the divergence of the fluctuation amplitudes between the two spectra becomes noticeable when tension predominates. Each mean-square fluctuation spectrum is obtained averaging all 20 fluctuation spectra for both low (blue) and high (yellow) tension cells. Since tension dominates low modes (q⁻¹ behavior) and bending modulus dominates high modes of the spectra (q⁻³ trend), the decoupling between tension and bending modulus becomes evident from these two spectra (Extended Data Section S2).

Extended Data Figure 10. Membrane flickering spectroscopy amplitude analysis.

(a) To justify our choice of modes for fitting Eq. S4, we calculated the residuals of mean square fluctuation amplitudes at different ranges of modes for the same RBC. The figure shows that the residues derived from fitting modes above 20 increase steadily, suggesting a systematic error in fitting modes above 20. Our range of modes (8-20) seems the most convincing range, as well as range 5-20, with no systematic deviations. By studying the dynamics of modes it is possible to extract the viscosity of the cell interior, and this analysis can be used as further proof of the method. From the timescale of decorrelation of mode amplitudes, it is possible to obtain the viscosity of the RBC interior, using the values of tension and bending modulus obtained from the static spectrum of the same cell. This is achieved by fitting the relaxation time with Eq. S7. The viscosity is statistically the same, across the non-Dantu and Dantu groups which have statistically different tension values. This is thus a further independent check confirming the static study is measuring tension values reliably. (b) The viscosities of RBCs with extreme low and high tension are not significantly different (p value=0.14, two-sided Mann-Whitney U test). The fit in the inset shows data from one of the RBCs in the sample. (c) The relaxation times, plotted vs q_x, modes 5-11, are represented for both the low- and high-tension RBCs; the trend is 1/q consistent with the limiting behaviour of Eq. S7 for $\left(\sigma \gg \kappa q_x^2\right)$. The range of modes that can be studied dynamically is limited by the camera acquisition rate, as well as by the other factors that limit also the static analysis.

Figure 1. Reduced invasion of Dantu variant RBCs by multiple P. falciparum strains.

(a) The relative ability of P. falciparum strains from multiple geographic locations (3D7 and GB4 West Africa; Dd2 Southeast Asia; SA075 East Africa; 7G8 South America) to invade RBCs was measured using a flow cytometry-based preference invasion assay. The percentage of parasitised RBCs in each genotype group is indicated on the y-axis. Statistical comparison across groups was performed by one-way ANOVA, while pairwise comparisons between groups used the Tukey HSD test. Significant differences in invasion were observed between non-Dantu and Dantu homozygotes in 3D7 (p=0.001), Dd2 (p=0.015) and SAO75 (p=0.028). Statistical data listed in Supplementary Table 2. (b) The invasion process was also followed by live video microscopy, where the invasion rate of 3D7 merozoites was measured as the proportion of merozoites that contacted and successfully invaded RBCs, relative to all merozoites that contacted RBCs. Pre-invasion time - from first merozoite contact through RBC membrane deformation and resting; invasion time - from beginning of merozoite internalization to beginning of echinocytosis. 6 RBCs per genotype group were tested in both flow and video microscopy assays. In the video microscopy assays, the number of contacted and successfully invaded RBCs counted were as follows: non-Dantu: 144/53, Dantu heterozygote: 191/43, Dantu homozygote: 233/41. Boxes indicate the median and interquartile ranges, while whiskers denote the total data range, with the dots outside the whiskers indicating the outliers. Bars show the mean and standard deviation of the video microscopy invasion data. Pairwise comparisons between genotypes were performed using the two-sided Mann-Whitney U test. **p < 0.01; *p < 0.05.

Figure 2. RBC membrane protein characteristics vary across Dantu genotypes but do not directly correlate with invasion efficiency.

(a) The relative expression of essential RBC membrane proteins was assessed using fluorescent monoclonal antibodies in flow cytometry assays. 13 non-Dantu, 12 Dantu heterozygotes and 11 Dantu homozygotes were tested. Statistical comparison across groups was performed by one-way ANOVA, while pairwise comparisons between groups used the Tukey HSD test. Significant differences were observed in GYPA (non-Dantu vs. Dantu homozygote p=6.25×10⁻¹¹; non-Dantu vs. Dantu heterozygote p= 4.62×10⁻⁶; Dantu heterozygote vs. Dantu homozygote p= 6.86×10⁻⁴), GYPC (non-Dantu vs. Dantu homozygote p=0.03), Band3 (non-Dantu vs. Dantu homozygote p=6.25×10⁻¹¹; non-Dantu vs. Dantu heterozygote p=0.0136), CD71 (non-Dantu vs. Dantu homozygote p=0.006), and CR1 (non-Dantu vs. Dantu heterozygote p=0.003; Dantu heterozygote vs Dantu homozygote p=0.045). (b) Scatter plot of all proteins quantified by mass spectrometry (n=3 RBCs per genotype). Fold change was calculated by average signal:noise (Dantu homozygote/non-Dantu). GYPA was split into two parts: identified by peptides unique to GYPA (‘GYPA unique’, originating from extracellular region) or shared with the Dantu protein (‘GYPA shared’, originating from intracellular region). Mass spectra were processed with the quantitative proteomics platform “MassPike” and the method of significance A with Benjamini-Hochberg multiple testing correction was used to estimate the p-value that each protein ratio was significantly different to 1. (c) Graph of the relative abundance of ‘unique’ and ‘shared’ GYPA peptides across all donors. Signal:noise values were normalised to a maximum of 1 for each protein. Statistical data for (b) and (c) listed in Supplementary Table 4. (d) Comparison of invasion efficiency of a genetically modified parasite strain, ΔPfEBA175, across genotypes (n=13 non-Dantu, 12 Dantu heterozygotes and 12 Dantu homozygotes) using the flow-cytometry-based preference invasion assay. The percentage of parasitised RBCs in each genotype is indicated on the y-axis. Statistical comparison across groups was by one-way ANOVA, while pairwise comparisons between groups used the Tukey HSD test (non-Dantu vs. Dantu homozygote p=0.04). ** p < 0.01; * p < 0.05.

Figure 3

Biomechanical properties of the RBC membrane differ across Dantu genotypes and correlate with invasion. (a) Membrane flickering spectrometry enabled measurement and comparison of RBC bending modulus, tension, radius, and viscosity across genotypes (n=6 RBCs per genotype). Mean and standard deviation were obtained from the averages of cell tensions for each sample: non-Dantu RBCs - (6.0 ±1.9) *10⁻⁷ N/m; Dantu heterozygotes - (7.9±2.8) *10⁻⁷ N/m; Dantu homozygotes - (8.8 ± 0.7) *10⁻⁷ N/m. The impact of tension on parasite invasion was evaluated by simultaneously measuring tension from flickering analysis and live video imaging of the invasion process from rupturing schizonts (“egress”, “deformation”, then either “invasion” and “echinocytosis”, or a failed invasion) (b), in non-Dantu and Dantu homozygote RBCs (c). The threshold range for tension, marked in (a) and (c), was obtained by comparing distributions of tension across Dantu genotypes with their invasion efficiency. (d) The contact region between merozoites and RBCs, represented in the snapshots, was measured during pre-invasion at the point of RBC maximum deformation for 2 sets of very high (n = 15) and low tension (n = 23) cells (p=1.30 × 10⁻³²). Merozoite-RBC contact section was significantly smaller in high tension RBCs meaning that parasites were much more wrapped around RBCs with a lower membrane tension. (e) Parasite invasion efficiency and RBC tension for six increasing concentrations of glutaraldehyde (0.00001 - 0.01%). Parasite invasion was significantly decreased for RBC tensions around 8.8 × 10⁻⁷ N/m (22% decrease) and 12.2 × 10⁻⁷ N/m (43% decrease). Median values are reported from 2 technical replicates of 2 biologically independent samples. Pairwise comparisons between genotypes used the two-sided Mann-Whitney U test. ** p < 0.01. Number of cells and tension reported in Supplementary Tables 5, and 6.