Comparative proteomic analysis of metabolically labelled proteins from Plasmodium falciparum isolates with different adhesion properties

The virulence of the human malaria parasite Plasmodium falciparum is believed to relate in part to adhesion of parasitized erythrocytes to postcapillary venular endothelium and other uninfected blood cells. At the molecular level, surface antigens from infected erythrocytes interact with several host receptors such as CD36 [1‐3], Intercellular adhesion molecule-1 (ICAM-1) (CD54) [4], thrombospondin [5], E-selectin (CD62E), vascular cell adhesion molecule VCAM-1 (CD106) [6], P-selectin (CD62P) [7], α_vβ₃-integrin [8], platelet endothelial cell adhesion molecule-1 [PECAM-1 (CD31)] [9], Chondroitin sulfate A (CSA) [10], Complement receptor-1 (CR-1)),(CD35) [11], Heparan sulphate (HS)-like Glycosaminoglycan (GAG) [12], ABO blood group antigens [13] resulting in the rosetting, agglutination [14] and cytoadherence of iRBC. A parasite-derived adhesion molecule that mediates these interactions has been identified as P. falciparum erythrocyte membrane protein 1 (PfEMP1), which is encoded by the var multigene family [15‐17]. There are about 60 var genes per parasite genome [18] that switch on expression in a mutually exclusive manner and contain copies of motifs that bind to diverse host receptors. Direct association of specific adhesive phenotype with disease has not been easy to demonstrate, except for pregnancy-associated malaria, in which the ability of iRBC to bind to placental receptors (such as CSA) is closely related to poor maternal and neonatal outcome [19]. A number of other parasite proteins associated with the erythrocyte membrane and exported on the surface of parasitized erythrocytes have been reported, including, Rifin, Stevor, ring surface protein (RSP) 1/2, (for a review see [20]) and, more recently, Surfins [21]. These proteins form part of a group of P. falciparum proteins that contain signals for export from the parasitophorous vacuole [22, 23]. Further bioinformatic analysis has proposed over 200 proteins that contain these signals and this behaviour appears to have been adopted much more extensively in P. falciparum than in other Plasmodium species [24], perhaps in part due to its requirement to undergo antigenic variation and cytoadherence.

There is a correlation between parasite clones of different antigenic type and cytoadherence [25, 26], rosette formation and pathogenesis, which has led to the studies on the diversity of these properties in P. falciparum isolates by testing binding to purified host proteins or cell lines. The parental P. falciparum line IT4/25/5 was originally selected by binding to ICAM-1, C32 amelanotic melanoma cells and Human umbilical vein endothelial cells (HUVEC), followed by sub-cloning to produce ItG [27], A4 and C24 [25]. These parasite clones have distinctive cytoadherence properties [28, 29]. ItG is a high avidity ICAM-1 and CD36 binder, A4 has high avidity for CD36 and lower avidity to ICAM-1, and C24 binds strongly to CD36 with high avidity but not to ICAM-1. An unrelated parasite line, 3D7, is a low binding line, which has low avidity for CD36 and does not bind ICAM-1. However, the molecular basis underpinning these distinctive binding phenotypes and the molecules involved, other than PfEMP1, are poorly understood. Therefore, the very high resolving power of 2D autoradiography of metabolically-labelled proteins has been used to investigate these genotypically related clones that exhibit distinct adhesion phenotypes to identify the repertoire of molecules involved.

P. falciparum infection is a sophisticated system including adaptation and interaction of the human erythrocyte and the parasite. At the early ring stage, very few parasite proteins have been synthesized and so the contribution of the ring-stage parasite to the overall protein biomass of the infected erythrocyte is much less than the human proteins, making it hard to distinguish differences between parasite strain protein profiles at this stage. However, the mature human erythrocyte has no nucleus, no protein synthesis capability and no protein trafficking machinery. Metabolic labelling of proteins of P. falciparum can easily indicate parasite protein changes. Moreover, metabolic labelling is a more sensitive method for protein detection; the detection limit for silver stained protein spot is 200 ng while the limit for autoradiography of radioisotope labelled protein is 1 ng. In this study, metabolic incorporation was used to label cultured isolates of P. falciparum with [³⁵S]methionine combined with biochemical enrichment of the infected erythrocyte membrane compartment. The objectives of this study were to use proteomic approaches to obtain a snapshot of P. falciparum infection at the protein level in the iRBC membrane and free parasite fractions, compare specific parasite strains to find molecules differentially expressed in parasite isolates with defined adhesion properties and to identify the proteins of interest using MALDI-TOF.

Red blood cells (RBCs) are relatively simple cells that lack internal organelles and are enclosed by a plasma membrane. For iRBCs, the main parasite structure resides within a vacuolar membrane but progressively interacts with host proteins such as ankyrin and spectrin [36, 37] in the RBC membrane by inserting parasite-derived proteins, forming knob-like structures and changing the rigidity of the infected cell [38]. In order to analyze parasite protein expression and distinguish free parasite protein from RBC membrane proteins, iRBCs were treated with hypotonic ionic buffer, to rupture and release their internal haemoglobin and free the parasite, leaving membrane ghosts that can be collected from the surface of the free-parasite pellet after centrifugation. This ghost fraction contains mainly erythrocyte membrane, as well as proteins from parasite source that are still associated with the erythrocyte membrane even after extensive washing. The free-parasite fraction more precisely reflects the expression of parasite proteins in general. When the 2D gel of RBC ghost fraction was stained with Coomassie Blue (Figure 1), A4 iRBC looked almost identical to 3D7 iRBC, with the only visible change being a 200 kilo Dalton (kDa) protein with different iso-electric point (PI), which was more acidic in A4 than in 3D7. This protein was identified as ankyrin by MALDI-TOF. As neither Coomassie Blue nor silver stain methods were able to highlight differential protein patterns between 3D7 and A4 isolates (Figure 1), [³⁵S] methionine labelling in vivo was used so that only parasite proteins would be labelled and detected, as mature erythrocytes have no protein synthesis. As expected, the uninfected erythrocyte gel (data not shown) did not show any protein spots, while the iRBC ghost fraction contained a number of parasite proteins (Figures 2 and 3), with more parasite proteins seen in trophozoite than ring stage iRBC, suggesting that a number of parasite proteins have a tight association with the RBC membrane. The free-parasite fractions were also analysed after [³⁵S] methionine labelling; the autoradiographs of the free parasite preparations of ring and trophozoite stages from three different parasite lines are shown in Figures 4 and 5. Consistent with the genotype identity of A4, C24 and ItG, their 2D protein maps show very similar protein profiles, with the small number of variable proteins indicated by circles on the gels. Nevertheless, since A4, C24 and ItG were selected in vitro for different adhesive phenotypes and are known to express variant PfEMP1 proteins, clone-specific protein expression profiles would be expected, particularly in proteins associated with the membrane or surface fraction. However, the autoradiographs of both iRBC erythrocyte membrane ghost and free-parasite fractions showed only very subtle changes. Using the 2D Evolution software, a number of changes were identified with over three-fold difference in signal intensity (see marked spots in Figures 2 to 5). Interestingly, most changes were between C24 and either A4 or ItG profiles, suggesting that A4 and ItG were more similar, consistent with their ability to bind ICAM-1. Another P. falciparum strain, 3D7, is a poorly cytoadherent line but the complete sequence of its genome makes it the best candidate for functional genomics and proteomics study. Therefore, 2D autoradiographs of parasites were compared from the IT4/25/5 lineage with 3D7. As shown in Figure 1, comparison of 3D7 with A4 did not reveal many changes, reflecting the predominance of host proteins. However, autoradiography revealed several distinctions between the protein profiles of the 3D7 trophozoite stage iRBC and either A4 or ItG (Figure 6A), while iRBC of A4 and ItG, as before, displayed almost identical maps. In the enlarged image of the boxed area of Figure 6A (Figure 6B), examples of protein position changes have been listed. In a limited area around pH4-5, molecular weight 45 to 60 kDa, three spots show significant changes from 3D7 and A4 or ItG, in all three cases resulting from more basic proteins in the 3D7 sample. This could be due to lower phosphorylation of these proteins in 3D7 than A4 or ItG, however, the role of phosphorylation in P. falciparum and its regulation are not known. The higher degree of difference in protein profiles between 3D7 and IT4/25/5-derived strains was also reflected in the comparison of profiles from ring stage iRBC, which also showed differences (data not shown). These changes were not seen on autoradiographs of the free-parasite fraction, suggesting that they could be membrane-associated events.

Autoradiography does not only describe the changes in parasite proteins but also can be used as an indicator to excise proteins for identification. As seen in Figure 7, radio-labelled iRBC samples showed over 100 proteins of parasite origin in the iRBC membrane fraction. When this type of gel is stained with Coomassie Blue, the most abundant proteins identified by MALDI-TOF were from human source. Therefore, the positions of spots on the autoradiograph were used to trace parasite proteins. Since [³⁵S] methionine is a low energy isotope, the X-ray film needs to be in direct contact with the 2D gels, so the gel has to be dried, necessitating protein identification on dry spots. Despite these technical difficulties, identified a number of parasite proteins were identified (Table 1). However, there remain difficulties in identifying all the proteins, particularly low abundant ones. Indeed a number of well-characterized parasite proteins known to be involved in membrane association or surface localization (e.g. PfEMP1) could not be seen from the 2D map, most likely because of low abundance and hydrophobic nature, setting them out of the range of the protein purification and separation methodologies used here.

Parasite proteins only account for a small amount of total infected erythrocyte proteins, with over 70% of the protein spots identified as human proteins (data not shown). To further extract parasite proteins for identification, parasites were synchronized by Plasmagel flottation to enrich trophozoite stage parasites up to 80%, treated infected erythrocytes with saponin for less than 7 min and then washed the free-parasite fraction extensively. The brief saponin treatment produces large pieces of erythrocyte membrane and free parasite, repeated washing with ice cold saline solution removes most of the contaminating host erythrocyte lysate and, therefore, enriches for parasite proteins. This sample was separated by 2D gel electrophoresis (Figure 8) and using MALDI-TOF, 58 proteins were identified from this gel, of which 28 came from parasite source.

A number of techniques can be used to increase the rate of protein identification and try to find the surface protein specifically. For example, Multidimensional Protein Identification Technology (MudPIT), a two-dimensional liquid chromatograph coupled with tandem mass spectrometry [39]; molecules exposed on the iRBC surface can be digested by enzyme treatment such as trypsin or chymotrypsin [40, 41]; insertion molecules can be purified from membrane fractions [42]. Yimin Wu's group used biotin-streptavidin labeling to purify all proteins on the iRBC surface and identified proteins by shotgun proteomic analysis [43]. Such techniques certainly provide high throughput scale and improve the identification sensitivity. However, a problem still remained as some cytosol proteins were also present in the pool of surface proteins because of internal biotin labelling and high abundant protein non-specifically binding to streptavidin. Recently, Lingelbach's group has used furosemide, a non-specific anion transport inhibitor, to block parasite-induced biotin permeability, making biotin-streptavidin technique more specific [44]. Another way to trace protein trafficking or protein remodelling or protein targeting is by in silico prediction of host-targeting signals required for the export of proteins from the parasite vacuole to the erythrocyte membrane [22, 23].

The predominant modification seen on 2D gels was horizontal position changes rather than presence or absence of a protein, indicating protein post-translational modification such as phosphorylation, which increases the negative charge of the protein and shifts the iso-electric point (PI) to the more acidic side. In yeast, protein phosphorylation is estimated to affect 30% of the proteome. Using proteome chip technology, scientists have determined the in vitro substrates recognized by yeast protein kinases. 4,192 phosphorylation events involving 1,325 different proteins were identified and kinase-substrate relationships distinguished [45]. The results suggest that host and parasite proteins may experience phosphorylation changes during parasite development inside the erythrocytes.

Using a sensitive in vivo labelling technique allied to 2D electrophoresis, highly reproducible, limited changes in P. falciparum protein profiles from a number of parasite lines have been shown, differing in their adhesive phenotypes, with even fewer differences seen in antigenically distinct clones sharing the ability to bind to ICAM-1. Of the changes detected, most were alterations in PI, indicative of differential phosphorylation. Only a subset of the protein spots could be identified using mass spectrometry. Other studies have used non-gel methodologies, such as nano-LC/MS/MS to overcome this problem, but these also have limitations for low abundance, membrane proteins and several proteins known to be at the surface of the infected erythrocyte were not identified by this approach. Further development of biochemical techniques, detection systems and sequence databases, particularly for variant surface antigen families, will be required to identify the full complement of parasite proteins in the infected erythrocyte membrane.