Background
Cytoadhesion of infected erythrocytes plays a key role in malaria pathogenesis and contributes to disease severity [
1‐
5]. During the intra erythrocytic part of their life cycle
Plasmodium spp. invade erythrocytes and remodel the erythrocytic surface both in terms of exposed proteins, nanoprotrusions (‘knobs’) and rigidity [
6]. These changes render the infected erythrocytes susceptible to splenic removal and thus cytoadhesion to endothelial cells in the microcirculation is essential for parasite survival.
The cytoadhesion is mediated by variant surface antigens (VSA) that the parasites export to the erythrocyte surface [
7]. The binding is a strong selective force in vivo and parasites have multiple VSAs binding to multiple ligands [
8‐
10] including CD36, a well-known glycoprotein receptor [
11]. Studies of cytoadhesion and its role in malaria pathogenesis have mostly been performed by various in vitro assays using recombinant proteins, glycans or immobilized cells as ligands [
7,
10‐
13]. However, the cytoadhesion assays have so far ignored the endothelial glycocalyx, which is a thick, negatively-charged carbohydrate-rich matrix anchored to the cell membrane by proteins and lipids [
14]. Although the glycocalyx has been studied extensively on endothelial cells it is commonly overlooked in malaria research despite its relevance for endothelial homeostasis [
14,
15]. Previous studies indicate that malaria affects the endothelial glycocalyx thickness and structure [
16]. The present study examined the effect that the glycocalyx may have on parasite cytoadhesion. It is well known that the endothelial glycocalyx shields leukocytes and platelets from undesired binding to the endothelium [
17,
18]. This led to the proposal that cytoadhesion of parasite-infected erythrocytes may similarly be affected by the glycocalyx [
19]. The glycocalyx grows continuously during in vitro culture [
20] and in order to assess how this affected cytoadhesion a simple culture system was used to quantify changes in parasite binding to CD36 as a consequence of glycocalyx growth on Chinese hamster ovary (CHO) cells.
Methods
Cultivation of Chinese hamster ovary cells (CHO), endothelial cells and Plasmodium falciparum parasites
In short cultivation was performed essentially as previously described [
12]. The following CHO cell lines were used: CHO K1 [CHO WT, Cat No CCL-61™, American Tissue Culture Collection (ATCC)] and CHO CD36 (stably express human CD36, Cat No CRL-2092™, ATCC). CHO cells were cultured in HEPES-buffered RPMI 1640 (Cat No 01-106-1A, Biological Industries) supplemented with fetal bovine serum (FBS, final concentration 10%, Cat No 10500064, Gibco, Thermo Fischer Scientific) and gentamicin (final concentration 50 µg/ml, Cat No 15710064, Gibco). Cells were grown at 37 °C at 5% CO
2.
Immortalized, human cerebral microvascular endothelial cells (hCMEC/D3 [
21]) were kindly provided by Pierre-Olivier Couraud (Institut Cochin, Paris, France). hCMEC/D3 cells were grown in ECM2 medium (Cat No CC-3156, Lonza) supplemented with growth factor bullet (Cat No CC-3202, Lonza). Cells were grown at 37 °C at 5% CO
2. Passage 27–29 was used for the described studies.
Plasmodium falciparum strain IT/FCR3 was cultured in culture flasks at 37 °C, at 4% haematocrit in an atmosphere of 2% oxygen, 5.5% CO
2 and 92.5% N
2 [
12]. They were grown in HEPES-buffered RPMI Cat No 01-106-1A, Biological Industries) supplemented with Albumax (final concentration 5 mg/ml, Cat No 11021029, Gibco), hypoxanthine (0.02 mg/ml, Cat No H9636, Sigma-Aldrich),
l-glutamine (0.18 mg/ml, Cat No G5792, Sigma-Aldrich) and gentamicin (final concentration 50 µg/ml, Cat No 15710064, Gibco). Subculture with the addition of blood group O erythrocytes was done throughout the study. Human blood was obtained with verbal informed consent from healthy volunteers, a procedure that is permitted without ethical approval from the Ethics Committee in the Capital Region of Denmark.
Seeding cells at different densities
Several seeding densities were tested in order to obtain a confluent monolayer at the time of the experiment. For CHO cells the following densities were used: confluent day 1: 8 × 104 cells/ml, confluent day 2: 2.5 × 104 cells/ml, and confluent day 4: 6 × 103 cells/ml. For endothelial hCMEC/D3 cells the following densities were used: confluent day 1: 2 × 105 cells/ml, confluent day 2: 105 cells/ml, and confluent day 4: 5 × 104 cells/ml. These densities were seeded in 24- and 96-well plates and in transwell inserts for the experiments described below.
CHO and hCMEC/D3 cells were seeded in chamber slides (Ibidi, Germany) and extracellular carbohydrates detected by adding N-azidoacetylgalactosamine-tetraacylated (GalNaz, 50 µM, Cat No 88905, Thermo Fischer Scientific) to the culture medium. After a various number of days in culture the medium was removed and cells washed with 2% FBS in phosphate buffer (0.15 M Sørensen’s Buffer). The azido sugars were detected by incubating the cells with 100 mM DyLight488-labeled phosphine for 60 min at 37 °C (Cat No 88907, Thermo Fischer Scientific) diluted in phosphate buffer. Negative controls were cells not fed GalNaz but only the normal cell culture medium. Cells were co-stained with Hoechst 33342 (5 µg/ml, Cat No H3570, Thermo Fischer Scientific) and cell mask orange (1000× dilution, Cat No C10045, Thermo Fischer Scientific). After 60 min incubation cells were washed with FBS supplemented phosphate buffer. Labelling was assessed in live cells with a confocal microscope (Zeiss LSM780) at 37 °C using a 20× objective (NA 0.8).
Analyses of azido sugar binding
All cells were imaged with identical microscopy settings using Carl Zeiss software (Zen). Image analyses were performed using Image J version 1.47 [
22]. The green channel was separated from the original image and a transect was made from the upper left corner to the lower right corner. The raw data are shown as a histogram of intensity. Imaging depth was 16 bit and the maximal intensity is 65,355.
Transmission electron microscopy (TEM)
Cells were grown on hanging polyethylene terephthalate filters with pore size 0.4 µm (Cat No PIHT12R48, Milipore, Merck, Germany) and seeded in different densities allowing them to become a confluent monolayer at days 1, 2 and 4. When confluent, cells were washed in 5% bovine serum albumin (BSA, Cat No A8022, Sigma Aldrich) and incubated with 0.5 mg/ml cationized ferritin (Cat No F7879, Sigma Aldrich) in 5% BSA for 30 min at room temperature. Cells were initially washed in 5% BSA then in 0.2 M cacodylate buffer [Cat No 11653, electron microscopy sciences (EMS)]. Cells were fixed ON in 2.5% glutaraldehyde (Cat No 16210, EMS) in 0.05 M cacodylate buffer and dehydrated according to standard methods [
23]. Cells were osmificated at 1% for 60 min. (Cat No AGR1017, Agar Scientific) and infiltrated with an epoxy resin (Cat No T031, Embed 812 Medium, TAAB) by using propylene oxide as an intermediate (Cat No 20401, EMS). Cells were initially cut as semi thin sections at 1 µm and stained with toluidine blue. Then cut for TEM at 50 nm thickness and stained with uranyl acetate and lead citrate (EMS) as previously described [
23]. All TEM was performed by using a Philips CM100 equipped with an Olympus Veleta camera connected to a workstation with SIS Analysis software (iTEM).
Selection procedures
Parasites were selected for binding to CHO CD36 cells as described in detail previously [
12]. The selection procedure was repeated at least four times. All CHO CD36 cells were seeded and used for selection the following day.
Binding experiments
In essence binding experiments were performed as described in detail previously [
12]. All binding experiments were performed using 1% haematocrit and 20% parasitaemia. Binding to CHO CD36 was compared with binding to CHO WT cells and with unselected IT/FCR3 parasites and uninfected, healthy erythrocytes. Inhibition of binding to CD36 was performed as described previously [
12] where CD36 antibody at different concentration was incubated with CHO CD36 cells and allowed to bind prior to the addition of parasites.
CD36 expression
The expression of CD36 in CHO cells was compared with an on-cell ELISA design. After growth a different number of days, the cells were washed in PBS and fixed with formaldehyde 1% for 5 min. After washing, cells were blocked with 5% BSA for 30 min at room temperature and incubated for 60 min at room temperature with primary antibodies at 1 µg/ml (5% BSA in PBS as diluent). The primary antibody was CD36 (clone FA6.152, Cat No IM0765, Beckman Coulter). Cells were washed with PBS and bound antibodies detected by horse-radish peroxidase-conjugated anti-mouse antibodies (Cat No P0447, Dako). Labelling was quantified with a luminescent substrate (Cat No 34094, West femto, Thermo Fisher Scientific) in a luminescence reader (500 ms exposure. Victor2, Perkin Elmer, USA). After detection, substrate was removed and cells lysed with 50 µl triton x-100 (0.1% in PBS, Sigma-Aldrich). Protein content in the lysate was determined by a modified Lowry method (DC kit, Bio Rad, USA) and read at 650 nm (Multiscan, Thermo Fischer).
To test for total CD36 expression CHO cells grown for 1 and 4 days were lysed in ice-cold RIPA buffer (Cat No R0278, Sigma-Aldrich) with protease inhibitors (Cat No, 04693124001, Complete Mini, Roche). Lysate was spun and the supernatant was collected. Protein content was determined with Lowry method (Cat No 5000112, DC protein assay kit, Bio-Rad). 2 µl (5 µg protein/ml) was spotted on nitrocellulose membrane (Cat No 1620150, Bio-Rad), left to dry and blocked with 5% skim milk powder in tris-buffered saline (Sigma-Aldrich) for 60 min. The membranes were incubated over night with primary antibodies as described above and detected with matching Alexa647-conjugated secondary antibodies. An antibody against β-tubulin was used as a loading control (1000× diluted, Cat No Ab6046, Abcam). Content was quantified by using a LAS4000 scanner (CY5 filter, GE Healthcare) and analysed by densitometry using Image J 22.
Var gene profiling
The expression profile of
var genes was performed by quantitative PCR analyses of mRNA. Ring-stage parasites were enriched with sorbitol as previously described [
24]. The washed pellet (100 µl) was thoroughly mixed with 900 µl Trizol (Cat No 15596026, Thermo Fischer Scientific) and stored at −80 °C use. RNA was reverse transcribed from random hexamers, using Superscript II (Cat No 18064014, Thermo Fischer Scientific), according to the manufacturer’s instructions. Quantitative primers for each
var gene of the
P. falciparum clone IT/FCR3 and quantitative PCR was performed on a Rotorgene RG-3000 thermal cycler (Corbett Research) was as previously described [
25], where gene-specific standard curves were produced by determining the amplification efficiency relative to the single copy housekeeping gene,
seryl-
tRNA synthetase, based on quantitative measurements of 10-fold dilutions of genomic DNA and used to calculate the transcript copy number of each gene in tested cDNA.
Statistical analyses
Data from binding experiments were assessed for equal variance and Gaussian distribution by using R [
26]. Data followed the criteria and were analysed with one-way ANOVA followed by post hoc tests (Holm–Sidak correction). Data from on-cell ELISA failed equal variance tests and were analysed with non-parametric tests (Kruskal–Wallis followed by Mann–Whitney U test). If data followed criteria for parametric testing they are shown as mean with standard deviation, otherwise as medians with interquartile ranges. All experiments were run in quadruplicate and repeated at least twice or in triplicate and repeated at least thrice. Graphs are designed in Graphpad Prism (version 7).
Discussion
Malaria parasite cytoadhesion has been widely studied since it is believed to be involved in pathogenesis and malaria severity [
11]. So far, in vivo studies of the cytoadhesion of
P. falciparum-infected erythrocytes have been difficult due to the lack of a good animal model [
33], and studies of malaria-induced changes to the microcirculation in vivo rely mostly on ophthalmoscopy [
34]. Thus, in vitro models have been used for most in-depth studies of cytoadhesion [
10,
12,
35,
36].
The present data show that in vitro cytoadhesion to CD36 could be blocked by glycocalyx growth in CHO cells. CD36 selected parasites recognized CD36 on CHO cells but only when the cells were not covered by a glycocalyx. In this case specific recognition was reduced significantly.
Two independent markers of glycocalyx growth were used: azido sugars visualizing total O-linked proteins and cationized ferritin binding, which mainly reacts with negatively charged glycosaminoglycan chains on proteoglycans [
15]. This allowed for confirmation of the previously described spatio-temporal in vitro growth of the glycocalyx in cell cultures [
20]. This suggests that the thickness of the glycocalyx also has functional importance in terms of cytoadhesion in malaria and that it should be taken into account.
There is a significant difference in the two described labelling approaches for the glycocalyx. By using the azido sugars one can quantify the processing of glycocalyx by the cells. Thus, if the cells do not export O-linked glycoproteins to the cell surface no labelling will be observed. The method does not demonstrate O-linked glycoproteins that had been formed prior to adding GalNaz. In contrast, ferritin labelling will detect all negative surface charge present when added. Demonstration of ferritin shortly after trypsinization gives indications of either a robust negatively charged surface coat that is resistant to trypsin, or a coat that is very quickly rebuilt. Together the two approaches give a picture of the dynamics taking place at the cell surface confirming the cell surface changes over time as previously reported [
20]. Thus, after trypsinization the majority of the negatively charged glycans are located
intracellularly and they require 2–4 days in culture to reach an extent, where they are able to cover surface receptors such as CD36.
An alternative explanation of these findings could be a reduced CD36 expression on the CHO cells with time in culture. However, dot blot analysis showed that CD36 expression was stable during 4 days of incubation. In contrast, antibody accessibility and binding to CD36 was reduced at the same times that parasite binding was inhibited. This further supports a role of the increased glycocalyx thickness in inhibiting binding.
Although the parasite cell line used was selected for CD36 binding a complete inhibition of binding with anti-CD36 was not possible. In concordance with this some degree of unspecific binding by unselected parasites was noticed. Interestingly, anti-CD36 reduced the binding to levels comparable to background binding to WT CHO cells.
The profile of the CHO CD36 selected parasite isolate matches what has previously been shown [
30].
Var genes encode more than 60 types of the protein called
P. falciparum erythrocyte membrane protein-1 (PfEMP1), which is composed by multiple CIDR and Duffy binding-like (DBL) domains [
37].
P. falciparum exports multiple VSAs to the cells surface and in this study it was not assessed whether one or more of these smaller VSA contribute to CD36 adhesion.
Various PfEMP1 variants have been shown to exhibit binding to both glycans and proteins [
11,
13,
30,
36]. It would be obvious to take into account how the glycocalyx [
14] affects binding to ligands involved in malaria pathogenesis. A loss of glycocalyx has been demonstrated in murine models of malaria [
16] and the inflammatory conditions may prime for both loss of glycocalyx as well as upregulation of adhesion molecules including CD54 enabling improved cytoadhesion [
19].
The glycocalyx is a permeability barrier and plays a considerable role in preventing loss of albumin via the kidney [
38]. In vivo imaging has also revealed that different sizes of dextran have different permeability through the glycocalyx. A 70 kDa fluorescein dextran only enters the glycocalyx after enzymatic removal of glycans [
39] and these molecules are half the molecular weight of an IgG molecule. The glycocalyx could be expected to block binding to large molecules such as immunoglobulins as well as infected erythrocytes in vivo.
From a glycocalyx perspective, the use of static binding assays may be questioned due to the tendency of erythrocytes to sink into the glycocalyx if they are placed on top of it [
40]. Reduced erythrocyte binding may not only be caused by steric hindrance but also by a different surface charge of the immobilized cells. The use of endothelial cell lines for in vitro studies of parasite adhesion could also be reconsidered since such cell lines appeared to develop a less pronounced glycocalyx layer in vitro compared with what has been found in vivo [
41,
42]. In accordance with previous studies less intense staining of glycocalyx on brain endothelial cells compared with CHO cells was also noticed in this study. Thus, the functional importance of the glycocalyx in endothelial cells was not tested.
The present data suggest that the loss of endothelial glycocalyx during malaria may be an important factor allowing the sequestration of infected erythrocytes to endothelial receptors. Although the study only shows data for CD36, further studies should address the possibility of a similar effect on receptors implicated in cerebral malaria such as CD54 and endothelial protein C receptor. It is clear that the inflammatory conditions during clinical malaria may be a direct cause of the shedding of the glycocalyx [
43]. In line with this, glycocalyx loss in murine malaria models has been shown; the loss being more pronounced in severe than in uncomplicated disease [
16]. However, the glycocalyx loss was only significant several days after parasites were detected in the bloodstream, and there is a need to study the initial steps allowing contact between infected erythrocytes and surface receptors. Conversely, it was recently speculated that the glycocalyx is essential for the propagation of malaria since the parasites rely heavily on cytoadhesion in order to avoid splenic destruction [
19]. Glycan-mediated cytoadhesion has been known for several years [
13,
44,
45] but an understanding of the spatial and temporal interplay between parasite VSA, glycocalyx and endothelial protein receptors is lacking.
Authors’ contributions
CH developed the hypotheses, performed experiments, interpreted data, drafted the manuscript. CW performed experiments, contributed to writing of the manuscript. JALK contributed to interpreting the data and writing of the manuscript. TS refined the hypotheses, performed experiments, contributed to writing of the manuscript. All authors read and approved the final manuscript.