All parasites used in this study were derived from the P. falciparum isolate, 3D7 and were cultured in O Rh+ RBCs as described previously 33. Three 3D7 sub-lines were used, namely: 3D7_SM-CD36,3D7_SM-drift and 3D7_UM.

3D7_SM-CD36 was obtained by bio-panning 3D7_SM on Chinese hamster ovary (CHO) cells expressing CD36. 3D7_SM has previously been selected from 3D7 using a pool of plasma from semi-immune Ghanaian children 2728. To select for a CD36-binding 3D7_SM population, the method described in 27 was followed. Gelatine enriched late-stage 3D7_SM was panned on a monolayer of wild-type CHO cells. Unbound iRBC and uninfected RBC (uRBC) were then panned on CHO cells transfected with human CD36. The monolayer was washed repeatedly to remove unbound iRBC and uRBC. To allow bound 3D7_SM parasites to reinvade and grow, culture medium and uRBC were added to the monolayer of CD36-transfected CHO cells and incubated overnight.

3D7_SM was allowed to drift without routine selection using the Ghanaian plasma pool to obtain the subline 3D7_SM-drift. For consistency, 3D7_SM-drift and unselected 3D7 were late-stage enriched by gelatine flotation 34 every time 3D7_SM-CD36 was selected for CD36 binding. The gelatine-enriched unselected 3D7 was termed 3D7_UM and was serologically similar to parasites causing uncomplicated malaria 27.

The genotypic identity of the 3D7 isolate and its derived sub-lines were confirmed by PCR at the polymorphic msp1, msp2, and glurp loci 35.

Immuno-staining and flow cytometry were carried out as described 2833 with minor modifications. To monitor VAR4 expression on the surface of live cells, late stage-iRBC were enriched by magnet-activated cell sorting (MACS, Miltenyl BioTec, Bergisch Gladbach, Germany)3637. MACS purified and ethidium bromide-labeled iRBC (2.5 × 10⁵) was incubated for 1 hr in 20 μL mouse sera or 40 μL rabbit sera raised against the CIDR1α or DBL5δ domains of VAR428 depleted for anti-human RBC antibodies prior to use. To detect surface bound antibodies, the iRBC were sequentially exposed to 100 μL of 1:25 goat anti-mouse or anti-rabbit Ig (DAKO), biotinylated anti-goat Ig (DAKO) and 1:1,000 fluorescein isothiocyanate (FITC)-conjugated streptavidin (BD Pharmingen) for 30 minutes each. For serological phenotyping, 20 human plasma samples from Magoda village in Tanzania with high malaria transmission were used to determine the recognition profile of the various parasite lines. Out of the 20 Magoda plasma samples, 10 came from children aged 3–5 years and 10 came from adolescents and adults aged 12–63 years. Staining of the iRBC by human IgG was done by incubating 2.5 × 10⁵ MACS purified ethidium bromide-labeled iRBC in 5 μL human plasma for 30 minutes and then sequentially exposed to 100 μL of 1:50 biotinylated rabbit anti-human IgG (DAKO) and 1:2000 FITC-conjugated streptavidin (DAKO) for 30 minutes each. All immuno-stained samples were analysed in a Coulter EPICS XL-MCL flow cytometer (Coulter Electronics, Luton, UK) or in a FACScan (Becton Dickinson). Data were analysed in WinMDI software http://facs.scripps.edu/software.html and WinList version 5.0 (Verity Software House, Inc., Maine, USA). Student's t-test was used to determine whether the samples compared were significantly different (P = ≤0.05).

RNA extraction, cDNA synthesis and quantitative real-time PCR were done as previously described 2838. Trophozoite/schizont-iRBC (36–48 h after invasion) were isolated from mixed stage in vitro cultures by exposure to a strong magnetic field as described in the previous section. To obtain ring stage parasites (30 h), the late stage enriched iRBC were cultured overnight. The ring stage has previously been shown to be optimal for studies of var gene transcription 38 and is therefore used for RNA extraction. Total RNA was purified using Trizol (Invitrogen) and treated with DNAseI (Invitrogen) for 15 min at 37°C. Superscript II was used to reverse transcribe DNA-free RNA primed with random hexamer primers (Invitrogen) at 25°C for 10 min, 42°C for 50 min followed by 70°C for 15 min. Quantitative real-time PCR was done using a Rotorgene thermal cycler system (Corbett Research, Motlake, Australia) and real-time PCR-optimized and gene-specific primers for each of the 59 full-length var genes in the P. falciparum 3D7 genome 2838. Reactions were performed in 20 μl volumes using QuantiTect SYBR Green PCR master mix (Qiagen, Merck Eurolab, Albertslund, Denmark) and 1 μM primers. Quantification was done using Rotorgene software version 4.6. The housekeeping gene seryl-tRNA synthetase, which shows a uniform transcription profile in different parasite isolates and an unchanged pattern throughout the parasite life cycle, was used as an endogenous control as previously described 38 and used for calculations of percent measured transcript out of total var transcipts by the ΔΔCT method (User Bulletin no. 2, Applied Biosystems).

The CD36 binding assay was done as described in detail elsewhere 27. To radiolabel parasites, cultures were incubated overnight in 10% non-immune Danish plasma in RPMI 1640 in the presence of ³H-hypoxanthine. Monolayers of wild type, CD36- and CD54-transfected CHO cells and transformed human bone marrow endothelial cells (TrHBMEC) were grown in 96-well microtiter plates (Nunc, Roskilde, Denmark). Late stage enriched iRBC (100 μl, 1 × 10⁷ iRBC/ml) were added to the CHO cell monolayer and incubated for 1 hour at 37°C. After removal of unbound iRBC, the number of iRBCs adhering to the CHO cells was determined by liquid scintillation spectrometry.

To determine the effect of CD36 selection on VAR4 surface expression, mouse or rabbit antibodies against the CIDR1α domain of VAR4 were used to stain three 3D7 sub-lines – 3D7_UM, 3D7_SM-CD36 and 3D7_SM-drift and the parasite protein surface expression subsequently analysed by flow cytometry (Figure 1A). In agreement with what has been previously shown 28, 3D7_SM surface expressed VAR4, whereas 3D7_UM did not show any surface staining for VAR4 at the beginning of the experiment (Figure 1i-A). Similarly 3D7_SM showed high var4 transcription, whereas 3D7_UM only transcribed var4 at a very low level (Figure 1i-B).

As a control, the VAR4 expressing 3D7_SM was left in culture for 44 generations from the day of the last antibody selection with gelatine flotation being carried out approximately every nine generations resulting in the 3D7_SM-drift sub-line. Both 3D7_SM-CD36 and 3D7_SM-drift expressed VAR4 on the surface all throughout the experiment, but with the 3D7_SM-CD36 sub-line showing a reduced level from the third panning (Figure 1ii-A and Figure 2). However, this reduction in VAR4 surface expression is not statistically significant (p = 0.19, Figure 2). Following the fifth selection, VAR4 surface expression was borderline significantly lower in 3D7_SM-CD36 compared with that in 3D7_SM (p = 0.05, Figure 2). Similarly, var4 gene transcription in the 3D7 sublines following the fifth selection was markedly lower in 3D7_SM-CD36 (10%) as compared to the level of transcription in 3D7_SM (28%) (Figure 1iii-B). By contrast the transcription level of the group B/A var gene PFF0010w (formerly known as MAL6P1.316)remained the same for both sub-lines (3D7_SM-drift and 3D7_SM-CD36) throughout the entire experiment (Figure 1ii-B and 1iii-B).

By using Western blotting for detection of VAR4 expression in 3D7_SM-CD36and 3D7_SM-drift sub-lines, a distinct 400 kDa VAR4 band was expressed following all selections, but it was not possible to accurately quantify the amount of the protein due to the insufficient sensitivity of the technique. However, the transcriptional data and flow cytometry of live immuno-stained parasites indicate that the 3D7_SM parasites which bind to CD36 have reduced levels of VAR4 PfEMP1 surface expression (Figure 1i–iii and Figure 2).

To test whether the CD36 bio-panning increased the adhesiveness of 3D7_SM to the CD36 molecule, 3D7_UM, 3D7_SM-CD36 and 3D7_SM-drift binding assays were done using CD36 and CD54 (also known as intercellular adhesion molecule or ICAM-1) expressing CHO cell lines and TrHBMEC. Due to high background adhesion of iRBC to wild type CHO cells only iRBC showing more than 8% binding to CHO-CD36 as seen following the third and fifth panning (Fig 1ii–iiiC) of 3D7_SM-CD36 were considered CD36 binders. Increased binding of 3D7_SM-CD36 to CD36 correlated well with reduced VAR4 expression and gene transcription (Figure 1A to 1C and Figure 2). The adhesion of 3D7_SM-drift to the CHO-CD36 was low in all selections (Figure 1iC–iiiC). Binding of 3D7_SM-CD36 to CHO-CD54 or TrHMBEC following bio-panning on CD36 did not differ from that of 3D7_SM-drift, indicating that bio-panning on CD36 was specifically selecting for parasites binding to the CD36 receptor.

Consistent with previous published data 2728, unselected 3D7 predominantly expressed PfEMP1_UM and were better recognized by adults (12–63 years) than by children (3–5 years) (p < 0.0001) while 3D7_SM selected using IgG from semi-immune children was equally well recognized by plasma from both groups of individuals (p = 0.8) (Figure 1i-D). Three to five rounds of gelatine flotation of 3D7_UM did not change the recognition pattern of this subline (data not shown). Following the third round of selection 3D7_SM-drift and 3D7_SM-CD36 had similar phenotypes to that of the original 3D7_SM (Figure 1i–iiD). Both sub-lines became significantly better recognized by adult plasma than by children's plasma after the fifth round (44 generations) of gelatine flotation (p = 0.02) or CD36 selection (p = 0.04, Figure 1iiiD) having serological profiles correlating slightly better with the baseline 3D7_UM compared to the 3D7_SM. Although borderline, this would indicate that both 3D7_SM-drift and 3D7_SM-CD36 have gradually shifted from expressing PfEMP1_SM-type antigens to expressing PfEMP1_UM-type antigens.

In the absence of CD36 selection VAR4 is stably expressed on the surface of 3D7_SM-drift for 44 generations – equivalent to five rounds of gelatine flotation (Figure 1iii-A). To analyse this further 3D7_UM and 3D7_SM were cultured without gelatine flotation or CD36 selection and monitored VAR4 surface expression and var4 transcription at specified time points for 94 generations. To avoid assay-to-assay variation, RBC infected with the two parasite sub-lines were frozen down at each sampling point. Frozen iRBC were thawed and cultured for another six generations and then assayed on the same day. Almost no VAR4 surface expression and var4 gene transcription was seen in the parental 3D7_UM, whereas 3D7_SM-iRBC showed high and stable surface expression of VAR4 as well as gene transcription for at least 94 generations (Figure 3). Using the percentageVAR4 positive iRBC and the ring-stage var4 transcription level at generation six (first sampling point) and 94 (last sampling point) the switch off-rate was calculated to be less than 0.005% (Figure 3). Alternatively, with a switch off-rate that is virtually 0, the minor differences observed at the transcription level may simply be due to experimental errors. Parallel to this, the two 3D7 sub-lines without prior freezing and thawing were assayed at every ~9th generation and similarly found VAR4 expression and transcription to be stable for over six months in continuous in vitro culture.