Adhesion between P. falciparum infected erythrocytes and human endothelial receptors follows alternative binding dynamics under flow and febrile conditions

Characterizing the adhesive dynamics of Plasmodium falciparum infected erythrocytes (IEs) to different endothelial cell receptors (ECRs) in flow is a big challenge considering available methods. This study investigated the adhesive dynamics of IEs to five ECRs (CD36, ICAM-1, P-selectin, CD9, CSA) using simulations of in vivo-like flow and febrile conditions. To characterize the interactions between ECRs and knobby and knobless IEs of two laboratory-adapted P. falciplarum isolates, cytoadhesion analysis over time was performed using a new tracking bioinformatics method. The results revealed that IEs performed rolling adhesion exclusively over CD36, but exhibited stationary binding to the other four ECRs. The absence of knobs affected rolling adhesion both with respect to the distance travelled by IEs and their velocity. Knobs played a critical role at febrile temperatures by stabilizing the binding interaction. Our results clearly underline the complexity of the IE-receptor interaction and the importance of knobs for the survival of the parasite at fever temperatures, and lead us to propose a new hypothesis that could open up new strategies for the treatment of malaria.


Results
Experimental setup. Endothelial cytoadhesion and the binding capacities of various laboratory strains and patient isolates of P. falciparum have been mainly investigated under static conditions 6,31,32 . This is classically achieved by directly incubating IEs with cells or recombinant receptors of interest 8,16,33,34 . Finally, adherent IEs are counted manually. These experiments demonstrated that laboratory-adapted trophozoite-stage IT4 and 3D7 isolates show different binding capacities to various ECRs under static conditions 31,33 . The main objective of this study was to characterize the binding behavior of erythrocytes infected with P. falciparum to the ECRs CD36, ICAM-1, P-selectin, CD9, and CSA under controlled flow and temperature conditions. For this purpose, we used transfected CHO cells expressing GFP fusions of the receptors of interest on the cell surface 33 . P. falciparum populations with increased binding for ICAM-1, P-selectin and CD9 receptors were obtained by performing at least seven rounds of enrichment, as described previously 33 . Whereas the IEs populations used for characterizing the binding over CD36 were obtained from the starting cultures without any enrichment, since over 80% of PfEMP1 proteins contain CIDRα2-6 domains, which are involved in CD36 binding 35 . All the IEs populations were tested for the presence or absence of knobs on their cell surfaces by scanning electron microscopy (SEM) (Fig. 1A). The HBEC-5i cell line was used as a model of endothelial cell cytoadhesion to CSA. Transcriptome profiling showed that IEs populations enriched for binding to HBEC-5i showed exclusive expression of var2csa and the HBEC-5i enriched population bind exclusively to CSA on the surface of HBEC-5i cells (Fig. S6A) 36 . A laminar flow system was used to imitate the hydrodynamic forces exerted on IEs in blood microvessels (Fig. 1B). In all experiments, different wall shear stresses (6, 4, 3, 2, 1.5, and 0.9 dyn/cm 2 ) were applied, which resemble the range of shear stress exerted on ECs lining blood microvessels 26,37 . Each magnitude of shear stress was applied for 10 min, except for 0.9 dyn/cm 2 , which was applied for 30 min. Bioinformatics image analysis was performed using the trackdem R package 38 . The IEs number was adjusted to 1 × 10 7 before the start of each experiment. Individual IEs were tracked through the entire sequence of images, and then trajectories were constructed and plotted over the static background. The accumulated distance traveled by IEs was defined as the length of each trajectory during the entire 10 min tracking time and was obtained from the output of trackdem analysis (Fig. 1C).
Effects of the presence of knobs and the parasite stage on the distance traveled by and velocity of 3D7-IEs rolling over CHO-745 CD36 cells. At 3 dyn/cm 2 , 3D7-IEs troph−K traveled a minimum distance  38,58 . The static cell background was subtracted from all images and then moving particles (i.e., IEs) were detected. The Shiny R package 38,58 was used with thresholding of 0.1. The PixelRange of moving particles was set to 30-500. Individual trajectories were subsequently reconstructed based on a minimum present (incThres) equal to six frames (i.e., 30 sec). Finally, trajectories were plotted over the still background. Examples of trajectories are shown in S1 and S2 Movies. of 13.43 µm and a maximum distance of 201 µm, with a geometric mean of 48 µm (Fig. 3A). At 2 dyn/cm 2 , 3D7-IEs troph−K traveled a minimum distance of 10 µm and a maximum distance of 212 µm, with a geometric mean of 34 µm. At 1.5 and 0.9 dyn/cm 2 , the minimum distance traveled remained the same (10 µm), while the maximum distance traveled decreased to 174 and 146 µm, respectively. The presence of knobs on the surface of 3D7-IEs affected the distance traveled. Specifically, the maximum distance traveled by 3D7-IEs troph+K was 330-400 µm at 4, 3, 2, and 1.5 dyn/cm 2 , while the minimum distance traveled was 11.5 µm at 4 dyn/cm 2 and 10 µm at all other shear stresses (Fig. 3B). On the other hand, at the lowest shear stress (0.9 dyn/cm 2 ), the maximum distance traveled was only 122.9 µm and the minimum distance traveled remained 10 µm (Fig. 3B). The maximum distance traveled by 3D7-IEs schi+K was longer (430-490 µm) at 4, 3, and 2 dyn/cm 2 but decreased to 338 and 198 µm at 1.5 and 0.9 dyn/cm 2 , respectively. The minimum distance traveled by 3D7-IEs schi+K remained the same (10 µm) at all shear stresses (Fig. 3C).
As a control, static binding assays were performed for IT4 and 3D7 IEs troph+K over CHO-745 GFP , where non-specific binding was not observed (Fig. S6B). Additionally, flow experiments were performed with IEs over CHO-745 GFP cells. After 30 min at 0.9 dyn/cm 2 , very few bound IEs were detected (3-6 IEs/0.2 mm 2 ). Another control was performed by allowing IT4-IEs ICAM-1 and 3D7-IEs ICAM-1 to bind to CHO-745 CD36 cells under flow conditions using the same experimental set-up. Interestingly, IT4-IEs ICAM-1 showed rolling binding behavior on CD36, as observed with the long-term cultured parasites.
We also observed that, on performing the flow assays at 22 and 37 °C at lower shear stresses (1.5-0.9 dyn/ cm 2 ) some of the rolling IEs stopped rolling and adhere to CD36 (about 60%). Whereas, this was not observed as we performed the assays at 40 °C, where a lower number of IT4-IEs (3-9 IEs/0.2 mm 2 < 1%) adhered at all shear stresses.

Discussion
We developed a new approach to study the binding behavior of IEs. This method integrates a flow system, which mimics vascular mechanobiology under physiological and febrile temperatures, and bioinformatics image tracking procedures to accurately characterize the cytoadhesive behavior of IEs. Wall shear stress in vivo varies along the venular tree as vessel diameter and flow rate change. Therefore, we performed experiments under six magnitudes of shear stress. We demonstrated that IEs interacting with CD36 performed rolling adhesion under different shear stresses and at different temperatures. Additionally, the rolling behavior of disc-shaped trophozoite-stage IEs (flipping) differed from that of round-shaped schizont-stage IEs (smooth continuous rolling). Studies by Fedosov and colleagues using multiscale IE modeling are consistent with this finding 20,23 . However, binding behavior to CD36 is controversial. In agreement with the current study, Herricks and colleagues observed rolling adhesion of isogeneic IEs over a microfluidic slide coated with CD36 40 . Rolling adhesion was also observed over human dermal microvascular ECs, which express numerous CD36 molecules on their surface (86 ± 14 × 10 6 molecules/mm 2 ) 41 . Lansche and colleagues reported that even deformed trophozoite-stage IEs with sickle cell traits perform rolling adhesion over human dermal microvascular ECs 42 . A chimeric SCID mouse model was used to mimic the human microvasculature in vivo. The authors also observed rolling adhesion performed of patients' IEs on human skin grafts with CD36 on their surface 19 . On the other hand, other studies suggested that IEs only bind statically to CD36 16,18 . CD36 is the main interacting partner of IEs because more than 40 of the (B) Experiments were performed using IT4-IEs enriched over HBEC-5i cells for at least seven rounds of enrichment. HBEC-5i cells were cultivated in a µ-slide I 0.8 Luer ibiTreat and the fluidic unit was connected with 1 × 10 7 IEs. Different shear stresses were applied using the ibidi pump system. Temperature was controlled using a hood over the microscope. An example is shown in S7 and S8 Movies. The number of bound IEs was counted using the trackdem R package 38,58 in the total image field (0.2 mm 2 ) and is represented as the mean ± SEM (**p < 0.001).
CD36 is a class B scavenger transmembrane glycoprotein found on the microvascular endothelium, macrophages, microglia, adipocytes, and platelets 45 . This receptor binds to a diverse range of ligands including pathogen-associated molecular patterns and modified self-molecules 46 . CD36 forms a 'hairpin-like' structure with a large ectodomain and two transmembrane domains 45 . The ectodomain of CD36 is capped by a three-α-helix bundle and an apex region that contains an accumulation of cationic residues, which are proposed to form a binding site for polyanionic ligands. Co-crystallization studies of CIDRα2-6 domains and the CD36 ectodomain identified a small conserved hydrophobic pocket in CIDRα peptides that forms a non-covalent hydrogen bond with a complementary hydrophilic phenylalanine residue (F153) in CD36. This combination generates a high affinity and stable binding site that allows the domains to interact with their ligands with a slow off rate, which stabilizes cytoadhesion of IEs against the strong forces generated by blood flow 47 . In the current study, cytoadhesive dynamics to CD36 differed according to the conditions. The absence of knobs on the surface of IEs shortened the rolling distance, whereas the presence of knobs led to maintenance of the rolling pattern for longer distances in the case of both trophozoite-and schizont-stage IEs. We also observed that the distance travelled by the IEs at higher shear stresses in both isolates have two peaks. This might be due to the control of the shear stresses to the (A) Cells were cultivated in a µ-slide I 0.8 Luer ibiTreat and the fluidic unit was connected with 1 × 10 7 IEs. Different shear stresses were applied using the ibidi pump system. Binding to cells was analyzed with longterm cultures. The number of adherent IEs was determined after flow for 30 min at 0.9 dyn/cm 2 . The number of rolling IEs was counted using the trackdem R package 38,58 in the total image field (0.2 mm 2 ) and is represented as mean ± SEM (**p < 0.001). (B) Average velocity of IT4 troph+K@37 °C and IT4 troph+K@40 °C , respectively. The distance travelled by each IE was divided by the individual contact time with cells to calculate velocity (n = 20).

Scientific RepoRtS |
(2020) 10:4548 | https://doi.org/10.1038/s41598-020-61388-2 www.nature.com/scientificreports www.nature.com/scientificreports/ rolling movement of the IEs. Some PfEMP1 molecules on the surface of IEs might perform a stronger bond with CD36, so that the IEs resist flow while others with weaker bond might not be able to roll longer distance and will be washed with the flow.
The mean rolling velocity of IEs over CD36 was different between different stages. We observed that schizonts roll more stable at different shear stresses than trophozoites. The flipping of trophozoites was quicker at higher shear stresses than the lower ones. In this study, all the recorded velocities at different stages and conditions were less than 1 µm/sec which was not recorded by Antia and colleagues. In their study, they observed higher velocities few to tens µm/sec. This might be due to the fact that they performed the assays over recombinant proteins (CD36 and ICAM-1) which might change the binding behavior 16 .
Comparing the mean rolling velocities of the knobby IT4-IEs at 37 °C versus 40 °C showed no statistical significance, in both cases they roll with a mean velocity of 30-50 µm/min which is higher than the mean velocity at room temperature (30-20 µm/min). This difference highlights the effect of temperature on binding dynamics between CD36 and PfEMP-1.
Shear stress significantly affected the velocity of IEs without knobs on their surface. It seems that flow influenced IEs lacking knobs. This might be because the density of PfEMP1 is supposed to be lower on the surface of IEs without knobs than on the surface of IEs with knobs 48 . On the other hand, tethering and receptor attachment of IEs with knobs seems to be controlled independently of shear stress. This was observed from the stable mean of velocity observed during the experiments performed with IEs +k (Figs. 3 and 4). Thus, the presence of knobs benefits parasites by stabilizing the ligand-receptor interaction due to the concentrated amount of PfEMP1 on the knob's surface. Slip binding occurs between CD36 and PfEMP1, and its life time is decreased by external forces (e.g., shear stress) 34 , which might explains the rolling behavior on higher shear stresses and static binding on the lower ones. Interestingly, rolling of schizont-stage IEs was more stable and continued for longer distances than that of trophozoite-stage IEs. The surface of schizont-stage IEs is homogenously covered by knobs, which is assumed to slow down rolling movement 41 . In addition, the adhesion force is smaller at the schizont stage than at the trophozoite stage, despite the high concentration of adhesion proteins in the former stage 49 .
We made the following findings regarding binding of IEs to ICAM-1, P-selectin, and CD9: (i) binding occurs mostly at lower shear stresses (starting from 2 dyn/cm 2 ), (ii) IEs mostly bind in a conglomerate pattern and resist flow, and (iii) no rolling pattern is observed. This suggests that catch bonds mediated binding. These are defined as non-covalent bonds that have increased lifetimes only when external force is applied 27,28 . On the other hand, IT4-IEs ICAM-1 showed rolling binding behavior on CD36, this IEs population is known to express mainly IT4-var01, this encodes for a protein which have both head structure for CD36 binding and also an ICAM-1 binding domain 33 . In accordance with the current study, such static binding to ICAM-1 was previously reported 19 . The low binding capacity at higher shear stress might be due to the lack of CD36 on the surface of CHO-745 cells. We believe that CD36 is the main receptor responsible for tethering of IEs and allows them to roll slowly before being immobilized by other ECRs. In addition, the presence of knobs did not affect the binding capacity or pattern to ICAM-1, P-selectin, and CD9, suggesting that these receptors play a secondary role in sequestration of IEs and only when ECs are stimulated with inflammatory cytokines. Previous studies provided more details about the binding dynamics of ECs to different receptors (including selectins and ICAM-1). When an activation signal is detected, ECRs switch from a low to a high affinity state due to changes in bond formation rates and bond dissociation properties 50 .
We investigated the thermodynamics of IE-ECR interactions. Most of these interactions occur at body temperature (37 °C) and during fever (up to 40 °C). We used HBEC-5i cells, which express CSA on their surface. Our flow experiments demonstrated the importance of knobs for parasite survival. Only IEs with knobs could withstand the high temperature and bind under shear stress of 0.9 dyn/cm 2 . In addition, rolling adhesion to CD36 was preserved and the number of rolling IEs was increased at 40 °C and under higher shear stress (starting from 4 dyn/cm 2 ). Normally, fever is a reaction to a pathogen in order to protect the host 27,50 . Fever induces expression of adhesive receptors (i.e., ICAM-1) on the surface of ECs and leukocyte trafficking 27 , but parasites are well prepared for these changes. Carvalho and colleagues measured the cytoadhesive force of IEs to CSA via force spectroscopy. They found that the binding force is notably decreased at febrile temperature, but the number of bound IEs increases 51 . They speculated that this increase in binding despite the decrease in force is due to non-specific binding. Another study demonstrated that, the binding affinities to both CD36 and ICAM-1 were decreased at febrile temperature (41 C°) 34 . The difference between our experiments and those performed by Lim and colleagues that 34 , they used recombinant proteins (ICAM-1 and CD36) in their binding studies and we used transfected CHO-cells that express the receptor of interest. We do believe that these two methods lead to different binding phenotypes between the receptor and ligand. On the other hand, it might be also that other proteins on the surface of the cells might help in increasing the binding capacity (for example heat shock proteins), but it will be difficult in the current study to confirm this explanation.
In conclusion, our data completely describe the cytoadhesive dynamics between IEs and ECRs in the presence and absence of knobs. IEs are first captured from the circulation via CD36, and this is followed by tethering and slow rolling. Other ECRs are expressed as the parasite load increases, which enhances immobilization of IEs to the surface of ECs. We also highlighted the importance of knobs for the thermodynamics of IE-ECR interactions. Higher temperatures also affect the binding dynamics between IEs and ECRs: (i) increased binding of knobby IEs to CSA, (ii) increased number of knobby IEs rolling over CD36, (iii) decreased number of immobilized knobby IEs over CD36.

Transfection and culture of CHO-745 cells. CHO-745 cells defective in glycosaminoglycan biosyn-
thesis (American Type Culture Collection (ATCC); no. CRL-2242) were transfected and cultured as previously described 8,31,33 . CHO-745 cells transfected with GFP, CD36, ICAM-1, P-selectin, and CD9 were used. These five cell lines were routinely sorted for surface expression of the receptors via fluorescence-activated cell sorting.
Selection and enrichment of IEs that bind to ECRs of interest. As previously described 33 , IEs (10% parasitaemia) were pre-absorbed over CHO-745 GFP cells for 60 min at 37 °C in binding medium (RPMI containing 2% glucose, pH 7.2). Thereafter, non-binding IEs were incubated with transfected CHO-745 cells for 60 min at 37 °C. Washing was then performed using binding medium for 7 times and then controlled under inverted microscope for any unbound IEs. Bound IEs were cultivated up to a parasitaemia of 10%. The entire enrichment procedure was repeated weekly for 7 weeks.
Selection and enrichment of IEs that bind to HBEC-5i cells. For enrichment of IEs for binding to HBEC-5i, the same procedure as described above was performed. At least seven rounds of selection were performed at 37 °C and 40 °C 36 . The HBEC-5i cell line was cultivated according to ATCC guidelines.
Enrichment of knob-containing IEs using gelatin flotation. IEs were resuspended in two volumes of 1% pre-warmed gelatin (175 g Bloom, Sigma-Aldrich, Germany) and incubated for 45 min at 37 °C 54 . The supernatant, which harbored knob-containing IEs, was washed with RPMI and cultivated as usual. This procedure was repeated each week. The presence of knobs was confirmed by electron microscopy.
qPCR. As a control, expression levels of var genes and kahrp in each parasite population were determined by qPCR. qPCR was performed using sense and antisense primers designed to amplify 100-120 bp fragments of the respective genes (S8 Text). After RNA isolation, cDNA was synthesized using SuperScript II Reverse Transcriptase (Invitrogen, Germany) and random hexamers (Invitrogen) at 50 °C for 1 h. The cDNA template was mixed with SYBR Green PCR Master Mix (QuantiTect SYBR Green PCR Systems, Qiagen, Germany) and 0.5 µM forward and reverse primers in a final volume of 10 µl. Samples were incubated at 95 °C for 15 min, and then subjected to 40 cycles of 95 °C for 15 sec and 60 °C for 1 min, followed by a melting step (60-95 °C). The specificity of each primer pair was confirmed after each qPCR run by performing dissociation curve analysis. var gene expression was normalized against expression of the conserved ring-stage gene Skeleton-binding protein 1 (sbp1) and the housekeeping genes Fructose-bisphosphate aldolase and arginyl-tRNA. Expression of each var gene was calculated relative to the geometric mean expression of these three normalizers 55,56 . Electron microscopy. Trophozoite-stage P. falciparum cultures were separated by Percoll density gradient centrifugation and fixed with 2.5% glutaraldehyde and 2.5% paraformaldehyde (Electron Microscopy Science, Hatfield, PA, USA). Samples were washed with 50 mM sodium cacodylate trihydrate buffer, pH 7.4 (Electron Microscopy Science), and post-fixed with 2% OsO 4 prepared in H 2 O for 45-60 min on ice in the dark. After heavy metal staining using 0.5-1% uranyl acetate (Electron Microscopy Science) for 45-60 min at room temperature, samples were dehydrated using an increasing ethanol series prepared in H 2 O. Embedding and polymerization were performed using increasing concentrations of epoxy resin (EPoN 812; Carl Roth, Karlsruhe, Germany) at 60 °C. Polymerized samples were cut into 55-60 nm sections (Ultracut UC7, Leica, Germany), placed on 300 mesh copper grids, and analyzed with a transmission electron microscope (Tecnai Sprit TEM, FEI, Netherland) at an acceleration voltage of 80 kV. By counting 150 IEs from each sample, the percentage of knobby to knobless IEs was determined.
Binding experiments under flow conditions. 1. Seeding of CHO cells. A total of 1.5 × 10 5 CHO cells were seeded into a µ-slide I 0.8 Luer ibiTreat in 200 µl Ham's F12 medium containing G418 as a selection marker 2 days before the assay. Cells were 90-100% confluent on the day of the assay. 2. Enrichment of IEs containing parasites at the trophozoite or schizont stage by magnetic-activated cell sorting. On the day of the assay, IEs containing trophozoites (28-32 hpi) or schizonts (36-44 hpi) were enriched using various MACS magnetic separation columns as described previously 57 . Briefly, cultures were centrifuged, resuspended in binding medium such that hematocrit was 10%, and loaded onto the top of the column. The outflow, which contained non-IEs and ring-stage and young trophozoites, was discarded. Finally, trophozoite-or schizont-stage IEs were eluted in binding medium and incubated over CHO-745 GFP cells for 60 min at 37 °C 33 . Finally, unbound cells were washed with binding medium and the cell count was adjusted to 1 × 10 7 using a Neubauer chamber. 3. Flow assay set-up: a unidirectional laminar flow system (ibidi pump system) was used. According to the manufacturer's instructions, IEs obtained in step 2 were resuspended in binding medium and added to the fluidic unit of the flow system. The fluidic unit was then connected to the CHO-745 cell slide and pump system. A range of wall shear stresses similar to those detected in venular microcapillaries 26 were applied using software to control the pump. Specifically, 6, 4, 3, 2, and 1.5 dyn/cm 2 was applied for 10 min, and 0.9 dyn/cm² was applied for 30 min. Imaging was performed using an EVOS FL Auto inverted microscope (2020) 10:4548 | https://doi.org/10.1038/s41598-020-61388-2 www.nature.com/scientificreports www.nature.com/scientificreports/ (Thermo Fisher Scientific, Waltham, USA). An image was acquired every 5 sec using a control program. The experiment was started by simultaneously switching on the pump system control program and the microscope control program.
Bioinformatics to track IEs and analyze trajectories. The trackdem (0.4.2) R package was used to characterize the behavior of IEs 38 . In each experiment, the image sequence recorded under each shear stress was loaded in the R-environment (3.5.1) 58 . A background image containing all motionless objects was created. Moving particles were detected by subtracting this background from all images. Identification was optimized using machine learning. The Shiny R package was used with thresholding of 0.1. In the PixelRange argument, a range of 30-500 was used to filter any unwanted noise. Particle tracking was then performed to generate tracked segments connecting the moving IEs in each frame. Individual trajectories were subsequently reconstructed based on a minimum present (incThres) equal to six frames (i.e., 30 sec). The reconstructed trajectories were plotted as images and presented in AVI animations. The accumulated distance covered by each IE was calculated by multiplying the total displacement in pixels by 0.4572. The detailed R-script used is provided in S10 Text. To calculate the average velocity of each individual IE, the accumulated distance was divided by the calculated contact time with cells. IEs that traveled less than 50 µm were excluded from the analysis.
Robustness checks. Manual tracking was performed and compared with automated tracking. ImageJ (Version 1.48 V, National Institutes of Health, Maryland, USA) was used to evaluate each experiment. The accumulated distance traveled by individual IEs was tracked using the Manual Tracking and Chemotaxis plugins. Statistical analysis. Geometric mean was calculated using GraphPad Prism 8. Statistical significance was assessed using the Kolmogorov-Smirnov test in flow experiments and the Kruskal Wallis test in qPCR experiments (GraphPad Prism 8).