Parasites used in this study were either well-characterized laboratory lines or were isolated from blood samples obtained at the KEMRI-Wellcome Trust, Kilifi, Kenya. Parasite lines and their origin are listed (Table 1). The African parasites were collected at Kilifi district hospital, situated 60 km north of Mombasa on the Kenyan coast. More than 10% of the children under five years of age, who reside within the Kilifi district, are admitted into hospital each year. In this area, P. falciparum transmitted by mosquitoes of the Anopheles gambiae complex 28, is seasonal and usually occurs after the long and short rainy seasons.

This study was approved by the Institutional Ethical Review Board of Nanyang Technological University, Singapore, and by the Kenyan Medical Research Institute National Ethics Committee

Laboratory parasite lines were cultured in fresh human RBC (supplied by National UK Blood Donation Service) in RPMI 1640 complete medium supplemented with Albumax and 2 mM glutamine (Gibco) as previously described 29. Parasitaemia was monitored daily by microscopy examination of Giemsa stained thin blood smears.

Plasmodium falciparum parasite isolates from Kilifi were recovered from liquid nitrogen frozen stock and cultured in RPMI 1640 medium supplemented with 10% AB human serum (Kilifi General Hospital, Kenya), 2 mM L-glutamine (Gibco), 37.5 mM HEPES (Gibco), 20 mM Glucose, and 2 μg/ml Gentamycin (Gibco). No fresh red blood cells (RBCs) were added and the parasites were maintained in culture for approximately 30 hours, or until they reached late trophozoite/schizont-stages. Parasite samples were separated into two fractions for later processing of parasite genomic DNA (gDNA) and RNA isolation.

Plasmodium falciparum iRBCs were pelleted at 560× g for 5 minutes. Resulting pellets were stored at -80°C in screw top cryovials. A phenol/chloroform extraction method was used for larger (greater than 500 μl) iRBC pellets as previously described 30. The DNeasy^® kit (Qiagen) protocols 1 and 2 were used for small (less than 500 μl) blood samples according to manufacturer's recommendations. Genomic DNA was resuspended in nuclease-free water.

For RNA isolation, iRBC pellets (approximately 100 μl) were lysed in 1 ml prewarmed (37°C) Trizol LS reagent (Gibco Invitrogen) and then stored at -80°C in screw top cryovials. RNA was extracted as described previously 31, using DNase I digestion (Invitrogen) to remove contaminating DNA. RNA was resuspended in nuclease-free water.

The primers were assessed using multiple sequence alignments to ensure that the chosen sequences were in regions conserved within the stevor gene family, but not in other genes (for example the rif gene family). A combination of two sets of primers in a nested PCR was used RepF1, RepF2 and RepR internal-PCR primers were designed as previously described 16. The RepF1/2 primers were redesigned to remove an opal stop codon (TGA) replacing it with (TGC). The smf1/smr1 external primers were previously described 32. All primers were synthesized by Eurogentec/Oswel (Southampton, UK) or Research Biolabs (Singapore). The use of a nested-PCR protocol for RT-PCR was problematic due to amplification of contaminating gDNA, therefore a stevor-specific single step amplification was designed, facilitated by the complete genome sequence 2, using the primers JBSTEVORF1 and JBSTEVORR1. Primers sequences are summarized in Table 2.

DNA and RNA products were separated by electrophoresis in 1% multi purpose agarose (Roche) in 1× Tris buffer contained 0.25 μg/ml ethidium bromide to enable UV visualization. RNA gels contained 5 mM guanidine thiocyanate.

Plasmid DNA was isolated using the QIAprep spin SNAP Miniprep^® Kit (Invitrogen). The QIAgen Maxiprep^® kit was used according to manufacturer's protocol to isolate larger quantities of pET-24a (+) vector.

Following DNA expansion, purification, and confirmation that a stevor insert was present; the DNA was then quantified by spectrophotometry, pelleted and dried before sequencing.

Three systems have been used for sequencing reactions:

a) The BigDye^® Terminator Cycle Sequencing Ready Reaction kit (Perkin-Elmer, Applied Biosystems, Warrington, UK) was used according to manufacturer's instructions. The reaction was carried out using approximately 600 ng of pCR^®2.1 or pCR^® II DNA (vector plus insert) per sample with either M13 forward or reverse primers in the Eppendorf Mastercycler. Unincorporated dye terminators were removed by ethanol precipitation. Samples were resuspended in 3 μl loading dye 16.5% v/v Blue Dextran/EDTA (Applied Biosystems) in formamide (Amresco, Solon USA), followed by electrophoresis and data collection on the ABI PRISM™ 377 DNA sequencer (Perkin -Elmer).

b) The BigDye^® reactions were also used with T7 primers and pET-24a (+) DNA, or TrxFus forward or T7 reverse primers with pET102/D-TOPO DNA. Sequencing was performed at the National Neuroscience Institute, Singapore.

c) The MegaBACE system (Amersham Biosciences) was used. The ABI PRISM^® BigDye^® Terminator v1.1 Cycle Sequencing kit (Applied Biosystems) cycle sequencing mix was used according to the manufacturer's protocols. Samples were ethanol precipitated before addition of MegaBACE™ loading solution (Amersham Biosciences) and loading onto a MegaBACE 96-capillary machine.

A fragment of approximately 280 base pairs (bp) was successfully isolated through PCR and RT-PCR amplification. However, due to the variant nature of the HVR, sequences varied in length, coding for a maximum of 82 amino acids. Clones have been obtained from multiple PCR reactions to reduce PCR bias, and were sequenced in forward and reverse directions. P. falciparum 3D7, IT and Ghanaian isolates sequences were obtained from BLAST and Tools-oligo searches within PlasmoDB 33.

Editing of nucleotide and amino acid (translated) sequences was carried out using the Autoassembler^®, Chromas^®, DNAstar (2001) freeware programmes. Sequences from different parasite sources and different amplification reactions were aligned by using ClustalX^® 34, and Bioedit7.0.1^® softwares. Multiple alignments and their corresponding chromatograms were edited by eye for discrepancies and mistaken nucleotide assignment by the automated sequencing method. BLAST analysis, version 4.4 from PlasmoDB 33 was used in order to confirm that DNA sequences were the result of the amplification of stevor exon-2 hyper-variable loop locus.

Phylogenetic trees showing genetic distance were constructed using the following programmes: Molecular Evolutionary Genetics Analysis (MEGA) 2.1 or 3.0^®, PAUP version 4.0 3536, and Tree view^® 37. These programmes enable the construction of trees using three-building models: Neighbor-Joining, Maximum Parsimony and Maximum likelihood, taking into account the observed nucleotide frequency, substitution rate and transition/transversion ratio, based on Nei genetic distance estimates 35. The bootstrap analysis 38 was used in order to statistically assign support to hypotheses of phylogenetic relationship among sequences estimated by the above-mentioned methods.

The P. falciparum genome sequencing project 2 as well as subsequent sequencing efforts has given some insight into the diversity of stevor in different P. falciparum lines. Still these data are limited and, therefore, additional sequences from both established laboratory lines and freshly isolated parasite samples would provide valuable additional information. For this purpose, genomic DNA was extracted from five different laboratory lines of various geographic origins: A4; FcB1; C10; T9/96 and 3D7 (Table 1) and also from 10 representative blood samples taken from patients in the Kilifi region of Kenya.

Genomic DNA was amplified using the previously described stevor-specific nested primers flanking the predicted hyper-variable loop region 32 (Table 2). PCR from all laboratory lines (Figure 1A) as well as Kilifi isolates (Figure 1B) gave amplification products of the expected size (600 bp and 300 bp) indicating that these primers were able to amplify stevor sequences from both the laboratory lines and Kilifi field isolates. Similar to what was found with the P. falciparum 3D7, more than one PCR product was detected in the expected size range, amplified by both the external (550–700 bp), and internal primers (250–350 bp) (Figure 1A). This is indicative of the amplification of multiple stevor sequences from both laboratory lines and field isolates. The 300 bp PCR products from the different samples were cloned and a number of clones were sequenced in both forward and reverse directions. The sequence data obtained here from genomic DNA was further expanded by the addition of the transcribed stevor sequences which had previously been obtained by RT-PCR from RNA extracted from the same original blood samples 39.

Despite the low throughput approach, a total of 213 genomic DNA and RNA products were successfully sequenced, 133 were unique DNA sequences [see Additional file 1] coding for 108 unique amino acid sequences. Comparison with known STEVOR sequences confirmed that the PCR and RT-PCR products indeed represented STEVOR. The number of times each sequence was obtained per isolate is shown (Figure 2A). Overall, six different stevor sequences were identified from genomic DNA in the reference 3D7 clone isolated in The Netherlands but of unknown geographic origin 40 that had been included as a positive control. Interestingly, two (3D7_2_41202 and 3D7_1_181202) of the sequences obtained from the 3D7 clone maintained at the NIMR in London were absent from the published 3D7 genome stevor repertoire (Figure 2A). All other stevor sequences present in the 3D7 genome are shown with their gene identification number for the purpose of comparison. Of the 133 unique stevor DNA sequences, twenty-one were found to be present in more than one parasite line or isolate (Figure 2A, asterisk), five (5) of these paralogues were identified in A4, C10, FcB1 and D10 (blue asterisk), six (6) were found in A4, C10 and FcB1 (green asterisk), three (3) were found in T9/96, C10 and A4 (orange asterisk), while two (2) were found in 3D7 and T9/96 (red asterisk), 3D7 and A4 (brown asterisk) as well as K11 and K14 (black asterisk). Finally a single sequence was found as part of the 3D7 genome and K1640 (Figure 2A, pink asterisk). The highly represented stevor sequences most likely represent "common" stevor that can be found in a variety of parasite isolates from around the world. For a number of these more common stevor transcription was also detected by RT-PCR indicating that these genes are of potential functional relevance.

From this initial analysis it is clear that there are a number of stevor sequences that are common to a number of different laboratory lines; in contrast only a single sequence (K1640_1_BD141105) from a Kilifi isolate was found to be identical to a gene (PFL2610w) from the 3D7 genome. In contrast despite the large number of stevor sequences identified in the Kilifi isolates, only a single stevor sequence was found to be present in more than one isolate. This particular sequence was transcribed in both K11 (K11+_111104) and K14 (K14+_1_111104) parasites (Figure 2B).

The RT-PCR data indicate that stevor is transcribed in all the P. falciparum lines examined with the exception of the parasite clone A4, supporting previous observations that this parasite clone does not transcribe nor express STEVOR 439. This lack of STEVOR expression in the A4 parasite is not due to the absence of the genetic information as this clone contains at least 20 different stevor genes (Figure 2A). In line with published observations 21 that field isolates transcribe multiple var at the same time, numerous unique stevor transcripts were also detected in both laboratory and patient samples (Figure 2B).

The observation that identical HVR sequences could be found in different parasite lines raised the possibility that stevor could be grouped into distinct subgroups with different biological functions, similar to the rif and pir multigene family 254142. For this purpose the sequences identified here, together with the corresponding regions from the 3D7 stevor repertoire were clustered into similarity groups by constructing a Neighbor Joining as well as a Minimum Evolution Tree (Figure 3).

The Neighbor Joining tree (Figure 3A) clustered the 152 HVR DNA sequences into two groups with one large sub-group containing approximately one-third of the stevor sequences (Figure 3A, orange branch). This subgroup contained sequences from all parasites except the D10 laboratory line and the K14, K11 and K6 Kilifi field isolates. At this stage it is not clear whether the absence of this subgroup of stevor in these parasites is a reflection of the limited sequence data available or truly represents a lack of this subgroup. It is also interesting to note that there is further sub-clustering of sequences within this branch, where clusters either contain sequences from the same line or alternatively from a number of different parasites. The second large group containing the remaining two-thirds of stevor sequences appears to contain numerous branch points of related sequences rather than containing a single defined cluster. These data would indicate that overall the sequence divergence is much higher in this group. Sequences obtained from a single parasite line also appear to form distinct clusters, with each cluster containing a number of different related sequences. This can be clearly seen for T9/96 stevor (Figure 3A, blue circles) that contains at least three distinct clusters. Similar clustering is also seen for sequences from A4, C10, FcB1 and D10 with the majority of sequences falling within just four sub-groups (Figure 3A, green circles) while the majority of Kilifi isolates stevor sequences were contained within seven clusters (Figure 3A, yellow circles). Importantly, these specific sequence clusters often also contain sequences from other parasite lines or isolates indicating that they do not result from a specific phylogenetic branch. Rather it appears that while there is a trend for some sequences from the same parasite to cluster together, most are distributed across the tree and cluster with sequences from other strains. To further validate these findings, the stevor sequences were clustered using the Minimum Evolution tree construction method at 95% significance bootstrap values (Figure 3B). This approach produced the same clusters as the Neighbor Joining method (compare Figure 3A with 3B). The 3D7 stevor repertoire is distributed throughout the trees (Figure 3A, B; pink points) with clusters containing a maximum of two 3D7 sequences.

To assess whether or not this clustering of stevor DNA sequences is maintained at the amino acid level, 108 unique STEVOR amino acid sequences were used to construct a Neighbor Joining as well as a Minimum Evolution Tree with a bootstrap value of 85% or higher (Figure 4A, B). The two approaches resulted in nearly identical trees with a number of broad clusters containing sequences from different lines and isolates. Using this analysis a single major sub-group was observed (highlighted by an orange circle) reflecting an apparent trend for sequences of the same geographic region to be more closely related (Figure 4A, B). More precisely, one obvious cluster was observed containing majority of Kilifi protein sequences (Figure 4A, orange circle, yellow points). However, as already identified in the DNA sequence trees, clusters are not completely exclusive and a number of sequences from laboratory lines are found within this cluster (Figure 4A, orange circle, green, pink, blue and light blue points). Conversely a number of Kilifi isolate sequences are found distributed across the other branches of the tree where they cluster both with other Kilifi sequences as well as sequences from a range of laboratory lines (Figure 4, yellow points).

Similar to the outcome of the DNA sequence analysis, the T9/96 amino acid sequences clustered within three small sub-groups (Figure 4, blue circles) while sequences from laboratory lines A4 (green), C10 (green), FcB1 (light blue) could be found dispersed in different clusters on the tree (Figure 4). Interestingly, the significant clustering seen within the trees produced from the nucleotide sequences (Figure 3) was also maintained at the amino acid level. This clustering is consistent using either of two methods for generating the phylogenetic trees.

This analysis was further extended to include all STEVOR sequences including those from published databases (IT and Ghanaian isolates) and 75 South American region sequences obtained from Albrecht et al. (2006)18. A total of 226 HVR amino acid sequences were used for the analysis (Figure 5). However, no obvious cluster was observed, though a number of non-significant groups were present when the tree was condensed at 50% bootstrap values. The high divergence within the HVR encoding the predicted erythrocyte surface-exposed region of the STEVOR family is likely in response to host immunity.