Evidence needed to establish a role for a specific host receptor in rosetting

In order to prove that a particular molecule acts as a host receptor for P. falciparum rosetting, a variety of different types of evidence have been provided. Essential data include proof that the molecule in question is found on normal human erythrocytes and that erythrocytes lacking the molecule show reduced/absent rosetting. Direct binding between IEs and/or recombinant parasite adhesion proteins and the receptor molecule should be demonstrated. Ideally, a crystal structure of the parasite adhesion molecule–host receptor complex should show the precise binding interaction site. Supportive evidence includes the ability of antibodies against the receptor or soluble receptor proteins to inhibit rosetting, and biochemical approaches to remove or alter the receptor on erythrocytes. Human genetic evidence can also provide indirect supportive evidence that particular molecules are important in life-threatening malaria. Several putative rosetting receptors have high-frequency polymorphisms in populations from malaria endemic regions that reduce rosetting and are associated with protection against severe malaria and death [reviewed in Rowe et al. (Reference Rowe, Claessens, Corrigan and Arman2009a, Reference Rowe, Opi and Williams2009b)]. These various lines of evidence are summarized below for each potential host rosetting receptor.

Blood group A and B trisaccharides

The most well-validated rosetting receptors are the blood group A and B trisaccharides (Fig. 3). In vitro experiments have shown that rosetting parasites have a ‘preference’ for blood groups A, B or AB rather than O (Carlson and Wahlgren, Reference Carlson and Wahlgren1992; Udomsangpetch et al., Reference Udomsangpetch, Todd, Carlson and Greenwood1993; Barragan et al., Reference Barragan, Kremsner, Wahlgren and Carlson2000b; Pipitaporn et al., Reference Pipitaporn, Sueblinvong, Dharmkrong-at and Udomsangpetch2000; Vigan-Womas et al., Reference Vigan-Womas, Guillotte, Juillerat, Hessel, Raynal, England, Cohen, Bertrand, Peyrard, Bentley, Lewit-Bentley and Mercereau-Puijalon2012; Moll et al., Reference Moll, Palmkvist, Ch'ng, Kiwuwa and Wahlgren2015). This varies by parasite genotype, with A-preference being the commonest. Clinical isolates from non-O (i.e. A, B or AB) patients show higher levels of rosetting than isolates from group O patients in studies from sub-Saharan Africa (Rowe et al., Reference Rowe, Obeiro, Newbold and Marsh1995, Reference Rowe, Handel, Thera, Deans, Lyke, Koné, Diallo, Raza, Kai, Marsh, Plowe, Doumbo and Moulds2007) and India (Rout et al., Reference Rout, Dhangadamajhi, Ghadei, Mohapatra, Kar and Ranjit2012), although the same result was not seen in one Thai study (Lee et al., Reference Lee, Malleret, Lau, Mauduit, Fong, Cho, Suwanarusk, Zhang, Albrecht, Costa, Preiser, McGready, Renia, Nosten and Russell2014). When parasites are cultured in their ‘preferred’ blood group, they form larger, stronger rosettes that are more resistant to disruption by antibodies or chemical agents than in group O cells (Carlson and Wahlgren, Reference Carlson and Wahlgren1992; Barragan et al., Reference Barragan, Kremsner, Wahlgren and Carlson2000b; Ch'ng et al., Reference Ch'ng, Moll, Quintana, Chan, Masters, Moles, Liu, Eriksson and Wahlgren2016). Enzymatic removal of the terminal sugars (N-acetyl-D-galactosamine for A and D-galactose for B) results in smaller, weaker rosettes, equivalent to those seen in group O erythrocytes (Barragan et al., Reference Barragan, Kremsner, Wahlgren and Carlson2000b). Rosettes do, however, still occur with blood group O erythrocytes (that express the H antigen), and also in Bombay phenotype red cells that lack the ABO blood group core fucose residue (Fig. 3) (Carlson and Wahlgren, Reference Carlson and Wahlgren1992; Rowe et al., Reference Rowe, Moulds, Newbold and Miller1997). This indicates that other red cell surface molecules in addition to the A and B antigens can act as host receptors for rosette formation.

Fig. 3.

Diagram of the ABO blood group sugars. Schematic representation of the terminal structure of the A (blue square), B (purple) H (green; H is the antigen carried on blood group O erythrocytes) and Bombay (orange) antigens. Yellow circle: D-Galactose (Gal), yellow square: N-acetyl-D-galactosamine (GalNac), red triangle: L-Fucose (Fuc). The symbols α and β indicate the position of the hydroxyl group and the numbers indicate the specific carbon atoms that are linked between the sugars. The H, A and B antigens are synthesized by a series of glycosyltransferase enzymes that add monosaccharides to create oligosaccharide chains attached to lipids and proteins in the erythrocyte membrane.

For the blood group A-preferring parasite line, Palo Alto 89F5, direct binding between the VarO PfEMP1 adhesion molecule and the blood group A trisaccharide was shown by Surface Plasmon Resonance (Vigan-Womas et al., Reference Vigan-Womas, Guillotte, Juillerat, Hessel, Raynal, England, Cohen, Bertrand, Peyrard, Bentley, Lewit-Bentley and Mercereau-Puijalon2012). The VarO PfEMP1 variant also binds to the B trisaccharide, but with lower affinity (Vigan-Womas et al., Reference Vigan-Womas, Guillotte, Juillerat, Hessel, Raynal, England, Cohen, Bertrand, Peyrard, Bentley, Lewit-Bentley and Mercereau-Puijalon2012). A crystal structure of the PfEMP1 N-terminal region was obtained and the A-trisaccharide binding site mapped (Vigan-Womas et al., Reference Vigan-Womas, Guillotte, Juillerat, Hessel, Raynal, England, Cohen, Bertrand, Peyrard, Bentley, Lewit-Bentley and Mercereau-Puijalon2012). A recent study suggests that P. falciparum RIFIN molecules may also be able to interact with blood group A sugars to contribute to rosette formation (Goel et al., Reference Goel, Palmkvist, Moll, Joannin, Lara, Akhouri, Moradi, Öjemalm, Westman, Angeletti, Kjellin, Lehtiö, Blixt, Ideström, Gahmberg, Storry, Hult, Olsson, von Heijne, Nilsson and Wahlgren2015), although direct RIFIN-A trisaccharide interaction was not shown.

The importance of the A and B antigens in rosetting is emphasized by the fact that the non-O blood groups are associated with increased risk of severe malaria and death compared to O (Rowe et al., Reference Rowe, Handel, Thera, Deans, Lyke, Koné, Diallo, Raza, Kai, Marsh, Plowe, Doumbo and Moulds2007; Fry et al., Reference Fry, Griffiths, Auburn, Diakite, Forton, Green, Richardson, Wilson, Jallow, Sisay-Joof, Pinder, Peshu, Williams, Marsh, Molyneux, Taylor, Rockett and Kwiatkowski2008; Tekeste and Petros, Reference Tekeste and Petros2010; Rout et al., Reference Rout, Dhangadamajhi, Ghadei, Mohapatra, Kar and Ranjit2012; Malaria Genomic Epidemiology Network, 2014; Ndila et al., Reference Ndila, Uyoga, Macharia, Nyutu, Peshu, Ojal, Shebe, Awuondo, Mturi, Tsofa, Sepulveda, Clark, Band, Clarke, Rowlands, Hubbart, Jeffreys, Kariuki, Marsh, Mackinnon, Maitland, Kwiatkowski, Rockett, Williams and Malaria2018; Degarege et al., Reference Degarege, Gebrezgi, Ibanez, Wahlgren and Madhivanan2019). Reduced rosetting in blood group O, and therefore reduced microvascular obstruction and reduced downstream pathological effects, is the proposed mechanism for the protective association with group O (Udomsangpetch et al., Reference Udomsangpetch, Todd, Carlson and Greenwood1993; Rowe et al., Reference Rowe, Handel, Thera, Deans, Lyke, Koné, Diallo, Raza, Kai, Marsh, Plowe, Doumbo and Moulds2007). ABO blood group does not influence parasite burden (Rowe et al., Reference Rowe, Handel, Thera, Deans, Lyke, Koné, Diallo, Raza, Kai, Marsh, Plowe, Doumbo and Moulds2007; Degarege et al., Reference Degarege, Gebrezgi, Ibanez, Wahlgren and Madhivanan2019), and evidence for an effect of ABO on P. falciparum invasion or other host–parasite interactions is conflicting and requires further study (Chung et al., Reference Chung, Gardiner, Hyland, Gatton, Kemp and Trenholme2005; Wolofsky et al., Reference Wolofsky, Ayi, Branch, Hult, Olsson, Liles, Cserti-Gazdewich and Kain2012; Pathak et al., Reference Pathak, Colah and Ghosh2016; Theron et al., Reference Theron, Cross, Cawkill, Bustamante and Rayner2018). The ABH antigens are known to be present on endothelial cells (Ito et al., Reference Ito, Nishi, Kawahara, Okamura, Hirota, Rand, Fechner and Brinkmann1990) and it is likely, but has not been shown experimentally, that cytoadhesion and overall levels of sequestration of rosetting parasites are enhanced in group A/B/AB patients compared to O.

Despite the progress in identifying the A and B trisaccharides as rosetting receptors and key genetic determinants of host susceptibility to severe malaria, there have been no attempts to develop specific therapies to block P. falciparum interaction with A/B antigens. The PfEMP1-blood group A trisaccharide binding pair described above (Vigan-Womas et al., Reference Vigan-Womas, Guillotte, Juillerat, Hessel, Raynal, England, Cohen, Bertrand, Peyrard, Bentley, Lewit-Bentley and Mercereau-Puijalon2012) remains the most clearly defined molecular interaction between parasite ligand and host receptor in rosetting, and could be used as a starting point to develop rosette-blocking therapeutics. Vigan-Womas et al. did report that the interaction between PfEMP1 and the A and B trisaccharides is indirectly inhibited by heparin (Vigan-Womas et al., Reference Vigan-Womas, Guillotte, Juillerat, Hessel, Raynal, England, Cohen, Bertrand, Peyrard, Bentley, Lewit-Bentley and Mercereau-Puijalon2012), and the development of a heparin-derivative as a potential adjunctive therapy for severe malaria is described below (Leitgeb et al., Reference Leitgeb, Charunwatthana, Rueangveerayut, Uthaisin, Silamut, Chotivanich, Sila, Moll, Lee, Lindgren, Holmer, Färnert, Kiwuwa, Kristensen, Herder, Tarning, Wahlgren and Dondorp2017).

Complement receptor one (CR1, CD35)

CR1 is a red cell membrane glycoprotein that regulates complement activation on cell surfaces (Thielen et al., Reference Thielen, Zeerleder and Wouters2018) and carries the Knops Blood Group antigens (Moulds, Reference Moulds2010). In malaria, CR1 plays a role in both rosetting and parasite invasion of erythrocytes (Schmidt et al., Reference Schmidt, Kennedy and Tham2015). CR1 was first identified as a rosetting receptor from a screen of 23 naturally occurring erythrocyte null mutants, each missing a particular blood group molecule or membrane glycoprotein (Rowe et al., Reference Rowe, Moulds, Newbold and Miller1997). The only variant to show substantially reduced rosetting with five P. falciparum parasite lines was Knops null cells, which are deficient in CR1. Normally, erythrocytes have between 100 and 1000 molecules of CR1 per cell (Wilson et al., Reference Wilson, Murphy, Wong, Klickstein, Weis and Fearon1986), whereas Knops null cells have fewer than 100 molecules per cell (Moulds et al., Reference Moulds, Moulds, Brown and Atkinson1992). Erythrocytes with fewer than 50 CR1 molecules per cell form rosettes poorly (Rowe et al., Reference Rowe, Moulds, Newbold and Miller1997), with normal rosetting occurring above a threshold of around 100 molecules per cell (JA Rowe, unpublished data).

Soluble CR1 and CR1 antibodies were shown to inhibit rosetting in some but, not all P. falciparum rosetting laboratory strains and clinical isolates, with only monoclonal antibodies (mAbs) that map to the C3b binding site on CR1 being effective inhibitors (Rowe et al., Reference Rowe, Moulds, Newbold and Miller1997, Reference Rowe, Rogerson, Raza, Moulds, Kazatchkine, Marsh, Newbold, Atkinson and Miller2000; Vigan-Womas et al., Reference Vigan-Womas, Guillotte, Juillerat, Hessel, Raynal, England, Cohen, Bertrand, Peyrard, Bentley, Lewit-Bentley and Mercereau-Puijalon2012). A recent paper suggested that the commercially available CR1 mAb E11 that recognizes epitopes outside the C3b binding site (Nickells et al., Reference Nickells, Hauhart, Krych, Subramanian, Geoghegan-Barek, Marsh and Atkinson1998) may inhibit P. falciparum rosetting (Lee et al., Reference Lee, Malleret, Lau, Mauduit, Fong, Cho, Suwanarusk, Zhang, Albrecht, Costa, Preiser, McGready, Renia, Nosten and Russell2014), but this was not seen in our hands (Rowe et al., Reference Rowe, Rogerson, Raza, Moulds, Kazatchkine, Marsh, Newbold, Atkinson and Miller2000). Further evidence of a role for CR1 in rosetting came from the expression of recombinant PfEMP1 domains in COS-7 cells, which bound to normal erythrocytes but not to CR1-deficient cells (Rowe et al., Reference Rowe, Moulds, Newbold and Miller1997).

Despite these supportive data, direct binding of IEs to CR1 protein has not been demonstrated, and recombinant rosette-mediating PfEMP1 proteins produced in E. coli (Ghumra et al., Reference Ghumra, Semblat, Ataide, Kifude, Adams, Claessens, Anong, Bull, Fennell, Arman, Amambua-Ngwa, Walther, Conway, Kassambara, Doumbo, Raza and Rowe2012) do not bind to CR1 in Surface Plasmon Resonance experiments (Tetteh-Quarcoo et al., Reference Tetteh-Quarcoo, Schmidt, Tham, Hauhart, Mertens, Rowe, Atkinson, Cowman, Rowe and Barlow2012). This could reflect a genuine lack of interaction between the two molecules, or could be due to technical reasons (e.g. the recombinant CR1 used in experiments was produced in mouse rather than human cells, whereas CR1 glycosylation, which may affect function, is cell-type specific) (Lublin et al., Reference Lublin, Griffith and Atkinson1986). It is also possible that a serum protein mediates the interaction between PfEMP1 on IEs and CR1 on uninfected erythrocytes, as the original experiments were all performed in the presence of serum (Rowe et al., Reference Rowe, Moulds, Newbold and Miller1997).

Human genetic studies provide additional support for the importance of CR1 in malaria host–parasite interactions. Erythrocyte CR1 deficiency is common in some malaria-endemic countries such as Papua New Guinea (Cockburn et al., Reference Cockburn, Mackinnon, O'Donnell, Allen, Moulds, Baisor, Bockarie, Reeder and Rowe2004) and India (Sinha et al., Reference Sinha, Jha, Anand, Qidwai, Pati, Mohanty, Mishra, Tyagi, Sharma, Venkatesh and Habib2009), and is associated with protection against severe malaria in medium to high transmission areas (Cockburn et al., Reference Cockburn, Mackinnon, O'Donnell, Allen, Moulds, Baisor, Bockarie, Reeder and Rowe2004; Sinha et al., Reference Sinha, Jha, Anand, Qidwai, Pati, Mohanty, Mishra, Tyagi, Sharma, Venkatesh and Habib2009; Rout et al., Reference Rout, Dhangadamajhi, Mohapatra, Kar and Ranjit2011; Panda et al., Reference Panda, Panda, Tripathy, Pattanaik, Ravindran and Das2012). However, erythrocyte CR1 deficiency may be detrimental in areas such as Thailand, where malaria transmission is low (Nagayasu et al., Reference Nagayasu, Ito, Akaki, Nakano, Kimura, Looareesuwan and Aikawa2001; Teeranaipong et al., Reference Teeranaipong, Ohashi, Patarapotikul, Kimura, Nuchnoi, Hananantachai, Naka, Putaporntip, Jongwutiwes and Tokunaga2008). There is also evidence that the CR1 Swain Langley 2 (Sl2) Knops blood group polymorphism that is common in African populations (Moulds, Reference Moulds2010) is associated with protection against severe malaria (Thathy et al., Reference Thathy, Moulds, Guyah, Otieno and Stoute2005; Opi et al., Reference Opi, Swann, Macharia, Uyoga, Band, Ndila, Harrison, Thera, Kone, Diallo, Doumbo, Lyke, Plowe, Moulds, Shebbe, Mturi, Peshu, Maitland, Raza, Kwiatkowski, Rockett, Williams and Rowe2018). Red cells carrying the Sl2 antigen on CR1 show reduced rosetting (Rowe et al., Reference Rowe, Moulds, Newbold and Miller1997; Opi et al., Reference Opi, Swann, Macharia, Uyoga, Band, Ndila, Harrison, Thera, Kone, Diallo, Doumbo, Lyke, Plowe, Moulds, Shebbe, Mturi, Peshu, Maitland, Raza, Kwiatkowski, Rockett, Williams and Rowe2018), and Sl2 may have additional effects on complement activation and regulation (Opi et al., Reference Opi, Swann, Macharia, Uyoga, Band, Ndila, Harrison, Thera, Kone, Diallo, Doumbo, Lyke, Plowe, Moulds, Shebbe, Mturi, Peshu, Maitland, Raza, Kwiatkowski, Rockett, Williams and Rowe2018).

Overall, the ability of CR1 mAbs and soluble protein to reverse rosettes suggests that CR1 plays a role in rosetting for some P. falciparum isolates. However, further work is needed to fully investigate the molecular interactions between parasite adhesion molecules and CR1, and to explore the potential for CR1 reagents (Li et al., Reference Li, Jaggers and Anderson2006; Reddy et al., Reference Reddy, Siedlecki and Francis2017) as therapeutic disruptors of rosetting.

Heparan sulphate and chondroitin sulphate

The glycosaminoglycans HS and chondroitin sulphate (CS) are found on cell surfaces and in the extracellular matrix of many tissues, and have a role in multiple aspects of the P. falciparum life cycle including hepatocyte invasion (Frevert et al., Reference Frevert, Sinnis, Cerami, Shreffler, Takacs and Nussenzweig1993), endothelial cell cytoadherence (Vogt et al., Reference Vogt, Barragan, Chen, Kironde, Spillmann and Wahlgren2003; Adams et al., Reference Adams, Kuhnrae, Higgins, Ghumra and Rowe2014) and, for CS, placental sequestration (Fried and Duffy, Reference Fried and Duffy1996). A number of papers have showed that heparin (which is a highly-sulphated form of HS found only in mast cells) can partially disrupt rosettes in about one-third to one-half of P. falciparum clinical isolates in vitro (Udomsangpetch et al., Reference Udomsangpetch, Wåhlin, Carlson, Berzins, Torii, Aikawa, Perlmann and Wahlgren1989; Carlson et al., Reference Carlson, Ekre, Helmby, Gysin, Greenwood and Wahlgren1992; Rogerson et al., Reference Rogerson, Reeder, al-Yaman and Brown1994; Rowe et al., Reference Rowe, Berendt, Marsh and Newbold1994; Barragan et al., Reference Barragan, Spillmann, Kremsner, Wahlgren and Carlson1999). It was shown that treating erythrocytes with heparinase III, which selectively cleaves HS chains, reduces rosetting in two P. falciparum lines (Barragan et al., Reference Barragan, Spillmann, Kremsner, Wahlgren and Carlson1999), and therefore suggested that ‘HS-like’ molecules on red cells are receptors for rosetting (Chen et al., Reference Chen, Heddini, Barragan, Fernandez, Pearce and Wahlgren2000). However, there has been only one paper reporting the existence of HS on normal human erythrocytes (Vogt et al., Reference Vogt, Winter, Wahlgren and Spillmann2004) and we have been unable to confirm this, and unable to detect any rosette-reducing effect of heparinase III in a range of parasite lines (McQuaid and Rowe, unpublished data).

Fluorescently-labelled heparin does bind to the surface of erythrocytes infected with rosetting parasites more than non-rosetting lines (Barragan et al., Reference Barragan, Fernandez, Chen, von Euler, Wahlgren and Spillmann2000a; Heddini et al., Reference Heddini, Pettersson, Kai, Shafi, Obiero, Chen, Barragan, Wahlgren and Marsh2001), and some rosette-mediating PfEMP1 variants bind directly to heparin (Barragan et al., Reference Barragan, Fernandez, Chen, von Euler, Wahlgren and Spillmann2000a; Vogt et al., Reference Vogt, Barragan, Chen, Kironde, Spillmann and Wahlgren2003; Juillerat et al., Reference Juillerat, Igonet, Vigan-Womas, Guillotte, Gangnard, Faure, Baron, Raynal, Mercereau-Puijalon and Bentley2010, Reference Juillerat, Lewit-Bentley, Guillotte, Gangnard, Hessel, Baron, Vigan-Womas, England, Mercereau-Puijalon and Bentley2011; Adams et al., Reference Adams, Kuhnrae, Higgins, Ghumra and Rowe2014). The heparin binding site in the N-terminal region of the varO PfEMP1 variant was mapped onto a crystal structure (Juillerat et al., Reference Juillerat, Lewit-Bentley, Guillotte, Gangnard, Hessel, Baron, Vigan-Womas, England, Mercereau-Puijalon and Bentley2011), and shown to be on the opposite side of the molecule from the erythrocyte binding site (Vigan-Womas et al., Reference Vigan-Womas, Guillotte, Juillerat, Hessel, Raynal, England, Cohen, Bertrand, Peyrard, Bentley, Lewit-Bentley and Mercereau-Puijalon2012). Hence, the rosette-disrupting effect of heparin is not due to direct blocking of receptor binding, but may result from aggregating PfEMP1 monomers and preventing their interaction with erythrocyte receptors (Vigan-Womas et al., Reference Vigan-Womas, Guillotte, Juillerat, Hessel, Raynal, England, Cohen, Bertrand, Peyrard, Bentley, Lewit-Bentley and Mercereau-Puijalon2012). Similarly, for another rosette-mediating PfEMP1 variant IT4var60, site-directed mutagenesis studies of recombinant proteins showed that mutations that disrupt heparin binding are distinct from mutations that disrupt erythrocyte binding, indicating that heparin-like molecules are not the main host rosetting receptor in this case (Angeletti et al., Reference Angeletti, Sandalova, Wahlgren and Achour2015).

Overall, whether HS is present on normal erythrocytes and is a host receptor for rosetting requires further confirmation. HS in present on the luminal surface of microvascular endothelial cells (albeit at a much lower density than on basolateral surfaces) (de Agostini et al., Reference de Agostini, Watkins, Slayter, Youssoufian and Rosenberg1990; Stoler-Barak et al., Reference Stoler-Barak, Moussion, Shezen, Hatzav, Sixt and Alon2014), therefore interactions between IE and endothelial HS (Vogt et al., Reference Vogt, Barragan, Chen, Kironde, Spillmann and Wahlgren2003; Adams et al., Reference Adams, Kuhnrae, Higgins, Ghumra and Rowe2014) are physiologically relevant and are likely to contribute to cytoadherence and sequestration in vivo.

Despite the uncertainty on the precise role of HS as an erythrocyte rosetting receptor, heparin and other sulphated glycoconjugate compounds have clear potential as adjunctive therapies for severe malaria due to their rosette-disrupting effects (Udomsangpetch et al., Reference Udomsangpetch, Wåhlin, Carlson, Berzins, Torii, Aikawa, Perlmann and Wahlgren1989; Carlson et al., Reference Carlson, Ekre, Helmby, Gysin, Greenwood and Wahlgren1992; Rogerson et al., Reference Rogerson, Reeder, al-Yaman and Brown1994; Rowe et al., Reference Rowe, Berendt, Marsh and Newbold1994; Kyriacou et al., Reference Kyriacou, Steen, Raza, Arman, Warimwe, Bull, Havlik and Rowe2007). There are reports of successful heparin treatment in severe malaria (Rampengan, Reference Rampengan1991) but its use is not recommended due to a high incidence of bleeding complications (World Health Organisation, 1986). As an alternative, Wahlgren and coworkers have developed a low anti-coagulant heparin derivative, Sevuparin, that reverses rosetting and cytoadherence in some P. falciparum isolates (Leitgeb et al., Reference Leitgeb, Blomqvist, Cho-Ngwa, Samje, Nde, Titanji and Wahlgren2011; Saiwaew et al., Reference Saiwaew, Sritabal, Piaraksa, Keayarsa, Ruengweerayut, Utaisin, Sila, Niramis, Udomsangpetch, Charunwatthana, Pongponratn, Pukrittayakamee, Leitgeb, Wahlgren, Lee, Day, White, Dondorp and Chotivanich2017) and also blocks merozoite invasion (Leitgeb et al., Reference Leitgeb, Charunwatthana, Rueangveerayut, Uthaisin, Silamut, Chotivanich, Sila, Moll, Lee, Lindgren, Holmer, Färnert, Kiwuwa, Kristensen, Herder, Tarning, Wahlgren and Dondorp2017). Sevuparin has been shown to be safe in adults with uncomplicated malaria (Leitgeb et al., Reference Leitgeb, Charunwatthana, Rueangveerayut, Uthaisin, Silamut, Chotivanich, Sila, Moll, Lee, Lindgren, Holmer, Färnert, Kiwuwa, Kristensen, Herder, Tarning, Wahlgren and Dondorp2017), but has not yet been tested in severe malaria patients.

The evidence for CS as a rosetting receptor is minimal. One report shows that rosetting in the P. falciparum line TM284 was partially inhibited by soluble CS and by chondroitinase enzyme treatment of erythrocytes, and that several clinical isolates showed reduced rosetting in the presence of CS (Barragan et al., Reference Barragan, Spillmann, Kremsner, Wahlgren and Carlson1999). However, other studies have found no effect of CS on rosetting in a variety of culture-adapted lines and clinical isolates (Rogerson et al., Reference Rogerson, Reeder, al-Yaman and Brown1994; Rowe et al., Reference Rowe, Berendt, Marsh and Newbold1994). There is also no convincing evidence that CS is found on the surface of normal human erythrocytes. Overall, current data do not support a role for CS in rosetting.

CD36

The membrane glycoprotein CD36 is a scavenger receptor for oxidized lipoproteins and a fatty acid translocase (Silverstein and Febbraio, Reference Silverstein and Febbraio2009). It is expressed on a variety of cell types including monocytes, macrophages, platelets, microvascular endothelial cells and adipocytes (Silverstein and Febbraio, Reference Silverstein and Febbraio2009), and at low levels on erythrocytes (van Schravendijk et al., Reference van Schravendijk, Handunnetti, Barnwell and Howard1992). The binding of PfEMP1 (group B and C variants) to CD36 on microvascular endothelial cells plays a major role in P. falciparum sequestration (Baruch et al., Reference Baruch, Gormely, Ma, Howard and Pasloske1996; Robinson et al., Reference Robinson, Welch and Smith2003). Almost all P. falciparum isolates bind to CD36, and increased CD36 binding (Newbold et al., Reference Newbold, Warn, Black, Berendt, Craig, Snow, Msobo, Peshu and Marsh1997; Ochola et al., Reference Ochola, Siddondo, Ocholla, Nkya, Kimani, Williams, Makale, Liljander, Urban, Bull, Szestak, Marsh and Craig2011) and predominant expression of group B and C PfEMP1 (Kraemer and Smith, Reference Kraemer and Smith2006; Kyriacou et al., Reference Kyriacou, Stone, Challis, Raza, Lyke, Thera, Koné, Doumbo, Plowe and Rowe2006) are associated with uncomplicated malaria.

The role of CD36 in rosetting is less clear. Anti-CD36 mAbs are capable of disrupting rosettes in a single culture-adapted parasite line, Malayan Camp (Handunnetti et al., Reference Handunnetti, van Schravendijk, Hasler, Barnwell, Greenwalt and Howard1992), but not in a wide range of other laboratory lines or clinical isolates (Udomsangpetch et al., Reference Udomsangpetch, Wåhlin, Carlson, Berzins, Torii, Aikawa, Perlmann and Wahlgren1989; Wahlgren et al., Reference Treutiger, Hedlund, Helmby, Carlson, Jepson, Twumasi, Kwiatkowski, Greenwood and Wahlgren1992; Rowe et al., Reference Rowe, Rogerson, Raza, Moulds, Kazatchkine, Marsh, Newbold, Atkinson and Miller2000; Niang et al., Reference Niang, Bei, Madnani, Pelly, Dankwa, Kanjee, Gunalan, Amaladoss, Yeo, Bob, Malleret, Duraisingh and Preiser2014). The PfEMP1 variants identified as parasite rosetting ligands (Rowe et al., Reference Rowe, Moulds, Newbold and Miller1997; Vigan-Womas et al., Reference Vigan-Womas, Guillotte, Juillerat, Vallieres, Lewit-Bentley, Tall, Baril, Bentley and Mercereau-Puijalon2011; Ghumra et al., Reference Ghumra, Semblat, Ataide, Kifude, Adams, Claessens, Anong, Bull, Fennell, Arman, Amambua-Ngwa, Walther, Conway, Kassambara, Doumbo, Raza and Rowe2012) are mostly of the group A type, which do not bind to CD36 (Robinson et al., Reference Robinson, Welch and Smith2003).

Intriguingly, while CD36 deficiency is fairly common in African populations, large-scale genetic studies have shown that CD36 polymorphisms do not influence severe malaria risk (Fry et al., Reference Fry, Ghansa, Small, Palma, Auburn, Diakite, Green, Campino, Teo, Clark, Jeffreys, Wilson, Jallow, Sisay-Joof, Pinder, Griffiths, Peshu, Williams, Newton, Marsh, Molyneux, Taylor, Koram, Oduro, Rogers, Rockett, Sabeti and Kwiatkowski2009). There is some evidence that interaction between IEs and CD36 may benefit the host, as CD36 may contribute to innate immune clearance of IEs and platelet-mediated parasite death (McGilvray et al., Reference McGilvray, Serghides, Kapus, Rotstein and Kain2000; McMorran et al., Reference McMorran, Wieczorski, Drysdale, Chan, Huang, Smith, Mitiku, Beeson, Burgio and Foote2012; Cabrera et al., Reference Cabrera, Neculai and Kain2014). Overall, it is unlikely that CD36 is a clinically significant rosetting receptor or a useful therapeutic target in severe malaria (Cabrera et al., Reference Cabrera, Neculai and Kain2014).

Glycophorin C (GYPC; GPC; CD236)

GYPC is a red cell membrane glycoprotein that carries the Gerbich blood group antigens (Jaskiewicz et al., Reference Jaskiewicz, Peyrard, Kaczmarek, Zerka, Jodlowska and Czerwinski2018). It is a P. falciparum invasion receptor bound by the merozoite protein EBA-140/BAEBL (Maier et al., Reference Maier, Duraisingh, Reeder, Patel, Kazura, Zimmerman and Cowman2003; Mayer et al., Reference Mayer, Jiang, Achur, Kakizaki, Gowda and Miller2006). Recently, two studies have suggested that GYPC is a rosetting receptor for both P. falciparum (Niang et al., Reference Niang, Bei, Madnani, Pelly, Dankwa, Kanjee, Gunalan, Amaladoss, Yeo, Bob, Malleret, Duraisingh and Preiser2014) and P. vivax (Lee et al., Reference Lee, Malleret, Lau, Mauduit, Fong, Cho, Suwanarusk, Zhang, Albrecht, Costa, Preiser, McGready, Renia, Nosten and Russell2014; Niang et al., Reference Niang, Bei, Madnani, Pelly, Dankwa, Kanjee, Gunalan, Amaladoss, Yeo, Bob, Malleret, Duraisingh and Preiser2014). Niang et al. showed that the rosetting of a 3D7-derived P. falciparum laboratory strain (5A-R+) was partially inhibited by a GYPC mAb (clone Ret40f) and by soluble recombinant GYPC (Niang et al., Reference Niang, Bei, Madnani, Pelly, Dankwa, Kanjee, Gunalan, Amaladoss, Yeo, Bob, Malleret, Duraisingh and Preiser2014). Furthermore, cultured GYPC knockdown erythrocytes failed to rosette, providing strong evidence that GYPC is an essential rosetting receptor for 5A-R+ parasites (Niang et al., Reference Niang, Bei, Madnani, Pelly, Dankwa, Kanjee, Gunalan, Amaladoss, Yeo, Bob, Malleret, Duraisingh and Preiser2014). Other parasite lines or clinical isolates were not tested, therefore the wider role of GYPC in P. falciparum rosetting was not determined.

Lee et al. (Reference Lee, Malleret, Lau, Mauduit, Fong, Cho, Suwanarusk, Zhang, Albrecht, Costa, Preiser, McGready, Renia, Nosten and Russell2014) focussed mainly on P. vivax, but also assessed the ability of GYPC mAb fragments to inhibit rosette formation in ten P. falciparum clinical isolates from Thailand. A significant decrease in rosetting was reported with GYPC mAb BRIC 4, although the reduction in the median rosette frequency was small (from 11.5 to 5.5%), and was based on a single count for each isolate with no replication (Lee et al., Reference Lee, Malleret, Lau, Mauduit, Fong, Cho, Suwanarusk, Zhang, Albrecht, Costa, Preiser, McGready, Renia, Nosten and Russell2014). The biological significance of these results is difficult to assess, given the low starting rosette frequencies and inherent variation in the rosetting assay. Lee et al. also used a different definition of rosetting to all previous studies, defining a rosette as an IE binding one or more uninfected erythrocytes. The usual definition requires the binding of two or more uninfected erythrocytes, which helps to identify genuine cell–cell interactions and avoid spurious identification of rosettes due to close packing of cells under the coverslip during microscopy.

For P. vivax, Lee et al. showed that the GYPC mAb reduced the median rosette frequency from 30 to 22% when tested on 11 Thai isolates, and that GYPC knockdown cultured erythrocytes formed rosettes poorly compared to GYPC-positive control cells (median rosette frequency 6.2 vs. 35.4% in controls, tested on three isolates). Plasmodium falciparum isolates were not tested with the GYPC knockdown erythrocytes.

If GYPC is a rosetting receptor, it is possible that the ‘Gerbich-negative’ blood group type, which is common in Melanesian populations (Patel et al., Reference Patel, Mehlotra, Kastens, Mgone, Kazura and Zimmerman2001), might influence rosetting. As part of a screen of null blood group erythrocytes with five high-rosetting P. falciparum culture-adapted parasite lines, Rowe et al. (Reference Rowe, Moulds, Newbold and Miller1997) tested two donors with the Gerbich-negative blood group (formed by deletion of exon 3 of the GYPC gene on chromosome 2, giving a truncated protein with altered glycosylation). Gerbich-negative erythrocytes formed rosettes normally with the five parasite lines tested. Goel et al. also report normal rosetting of Gerbich-negative erythrocytes from two donors (Goel et al., Reference Goel, Palmkvist, Moll, Joannin, Lara, Akhouri, Moradi, Öjemalm, Westman, Angeletti, Kjellin, Lehtiö, Blixt, Ideström, Gahmberg, Storry, Hult, Olsson, von Heijne, Nilsson and Wahlgren2015). The true null phenotype for GYPC, called the Leach phenotype (which arises due to the deletion of exon 3 and exon 4, encoding the transmembrane and cytoplasmic domains, respectively) is rare and has not been tested in rosetting assays to our knowledge.

Taking into account all existing evidence, further investigation of a wider range of parasite lines is needed to determine whether GYPC is an important host receptor for both P. falciparum and P. vivax rosetting.

Glycophorin A (GYPA, GPA, CD235a)

GYPA is a highly-expressed erythrocyte surface glycoprotein that carries the MNS blood group antigens. It is known to be a receptor for P. falciparum erythrocyte invasion (Sim et al., Reference Sim, Chitnis, Wasniowska, Hadley and Miller1994), and polymorphisms in GYPA are associated with resistance to severe malaria (Band et al., Reference Band, Rockett, Spencer, Kwiatkowski and Network2015; Leffler et al., Reference Leffler, Band, Busby, Kivinen, Le, Clarke, Bojang, Conway, Jallow, Sisay-Joof, Bougouma, Mangano, Modiano, Sirima, Achidi, Apinjoh, Marsh, Ndila, Peshu, Williams, Drakeley, Manjurano, Reyburn, Riley, Kachala, Molyneux, Nyirongo, Taylor, Thornton, Tilley, Grimsley, Drury, Stalker, Cornelius, Hubbart, Jeffreys, Rowlands, Rockett, Spencer and Kwiatkowski2017).

There is some limited evidence to suggest that GYPA may have a role in rosetting. Parasites of the strain FCR3S1.2 transfected with a specific RIFIN gene formed rosettes that were largely dependent on blood group A (Goel et al., Reference Goel, Palmkvist, Moll, Joannin, Lara, Akhouri, Moradi, Öjemalm, Westman, Angeletti, Kjellin, Lehtiö, Blixt, Ideström, Gahmberg, Storry, Hult, Olsson, von Heijne, Nilsson and Wahlgren2015). However, rosetting of the RIFIN-transfected parasites was significantly reduced with GYPA null cells from blood group O and B donors, whereas blood group A GYPA null erythrocytes formed rosettes normally. These data suggest that GYPA may have an accessory role for RIFIN-mediated rosetting in the absence of the A antigen (Goel et al., Reference Goel, Palmkvist, Moll, Joannin, Lara, Akhouri, Moradi, Öjemalm, Westman, Angeletti, Kjellin, Lehtiö, Blixt, Ideström, Gahmberg, Storry, Hult, Olsson, von Heijne, Nilsson and Wahlgren2015), although whether this applies to rosetting in non-genetically manipulated parasites in unknown.

Despite the above positive evidence, there are no other data supporting a role for GYPA in rosetting. GYPA mAb fragments had no inhibitory effect on rosetting in ten P. falciparum and 11 P. vivax clinical isolates (Lee et al., Reference Lee, Malleret, Lau, Mauduit, Fong, Cho, Suwanarusk, Zhang, Albrecht, Costa, Preiser, McGready, Renia, Nosten and Russell2014), and a GYPA mAb did not inhibit 3D7 5A-R+ rosettes (Niang et al., Reference Niang, Bei, Madnani, Pelly, Dankwa, Kanjee, Gunalan, Amaladoss, Yeo, Bob, Malleret, Duraisingh and Preiser2014). Furthermore, GYPA null erythrocytes (MkMk cells, lacking both GYPA and glycophorin B) formed rosettes with five culture-adapted P. falciparum lines (Rowe et al., Reference Rowe, Moulds, Newbold and Miller1997). Overall, existing evidence does not support a major role for GYPA in rosetting, but as with GYPC, further investigation is needed.

New receptors and new approaches

None of the receptors described above fully account for the adhesion interactions between infected and uninfected erythrocytes, and it is likely that other host rosetting receptors remain to be identified. There is evidence to suggest that these unknown host receptors are carbohydrates or protease-resistant proteins, because uninfected group O erythrocytes treated with trypsin and other proteases are still able to form rosettes (Udomsangpetch et al., Reference Udomsangpetch, Wåhlin, Carlson, Berzins, Torii, Aikawa, Perlmann and Wahlgren1989; Rowe et al., Reference Rowe, Berendt, Marsh and Newbold1994).

In order to progress rosetting research, alternative methods are needed. Rosetting experiments with GYPC and CR1 knockdown cultured human red cells derived from CD34+ haematopoetic stem cells have been performed (Lee et al., Reference Lee, Malleret, Lau, Mauduit, Fong, Cho, Suwanarusk, Zhang, Albrecht, Costa, Preiser, McGready, Renia, Nosten and Russell2014; Niang et al., Reference Niang, Bei, Madnani, Pelly, Dankwa, Kanjee, Gunalan, Amaladoss, Yeo, Bob, Malleret, Duraisingh and Preiser2014), using lentiviral transduction of short hairpin RNA (Bei et al., Reference Bei, Brugnara and Duraisingh2010). However, these cultured erythrocytes have a short life-span, limiting their usefulness. The development of immortalized erythroid lines (Kurita et al., Reference Kurita, Suda, Sudo, Miharada, Hiroyama, Miyoshi, Tani and Nakamura2013; Kanjee et al., Reference Kanjee, Grüring, Chaand, Lin, Egan, Manzo, Jones, Yu, Barker, Weekes and Duraisingh2017; Trakarnsanga et al., Reference Trakarnsanga, Griffiths, Wilson, Blair, Satchwell, Meinders, Cogan, Kupzig, Kurita, Nakamura, Toye, Anstee and Frayne2017; Scully et al., Reference Scully, Shabani, Rangel, Gruring, Kanjee, Clark, Chaand, Kurita, Nakamura, Ferreira and Duraisingh2019) may overcome this limitation. Nevertheless, attention must be paid to the subtle but real differences between mature erythrocytes and these, still relatively immature, immortalized CD34+ derived cells (Wilson et al., Reference Wilson, Trakarnsanga, Heesom, Cogan, Green, Toye, Parsons, Anstee and Frayne2016; Dankwa et al., Reference Dankwa, Chaand, Kanjee, Jiang, Nobre, Goldberg, Bei, Moechtar, Grüring, Ahouidi, Ndiaye, Dieye, Mboup, Weekes and Duraisingh2017; Trakarnsanga et al., Reference Trakarnsanga, Griffiths, Wilson, Blair, Satchwell, Meinders, Cogan, Kupzig, Kurita, Nakamura, Toye, Anstee and Frayne2017). CRISPR-Cas9 technology (Doudna and Charpentier, Reference Doudna and Charpentier2014) has led to an explosion in the ability to genetically manipulate multiple cell types, including erythrocyte precursors and immortalized haematopoietic lines (Song et al., Reference Song, Fan, He, Zhu, Niu, Wang, Ou, Luo and Sun2015; Kanjee et al., Reference Kanjee, Grüring, Chaand, Lin, Egan, Manzo, Jones, Yu, Barker, Weekes and Duraisingh2017; Hawksworth et al., Reference Hawksworth, Satchwell, Meinders, Daniels, Regan, Thornton, Wilson, Dobbe, Streekstra, Trakarnsanga, Heesom, Anstee, Frayne and Toye2018; Chung et al., Reference Chung, Magis, Vu, Heo, Wartiovaara, Walters, Kurita, Nakamura, Boffelli, Martin, Corn and DeWitt2019; Scully et al., Reference Scully, Shabani, Rangel, Gruring, Kanjee, Clark, Chaand, Kurita, Nakamura, Ferreira and Duraisingh2019), potentially giving the opportunity to generate multiple knockout lines for rosetting research. A consistent supply of knockout erythrocytes would allow large-scale screens for new rosetting receptors using cells as close to their normal physiological form as possible, raising exciting prospects for future work.