Genomic DNA from field isolates of the parasite blood stages was obtained by extraction with the Qiamp DNA blood minikit (Qiagen, USA) on blood samples received from Dielmo in Senegal (n = 7), Brazil (n = 5), Comoro islands (n = 6) and Thailand (n = 2). Nucleotide sequence data reported in this paper are available in the GenBank™, EMBL and DDBL databases under the accession numbers GQ222688-GQ22707.

The complete sequence of lsa-3 was obtained by a set of six PCR amplifications (Figure 1) with the corresponding PCR primer pairs (Table 1) designed after the lsa-3 sequence found in the strain K1 of P. falciparum (GenBank accession number: AJ007010) 7. The sequences of the PCR primers originally designed to amplify lsa-3 from the strain K1 matched the sequence in the strain 3D7 (GenBank accession number: AE001424) 18. The PCR reactions were performed on genomic DNA with 2,5 units of AmpliTaq DNA polymerase (Roche, USA) in a final volume of 50 μL of a buffer solution containing 1,5 mM of MgCL₂, 800 μM of dNTPs and 500 nM of primers. The cycling programme was; 94°C/2 min, [(94°C/15 sec, 57°C/30 sec, 72°C/1 min) × 35], 72°C/2 min, on a PTC-200 thermal cycler (MJ Research). Two sets of PCR amplifications were serially completed on the same samples to prevent ambiguities introduced by the Taq polymerase and thereby confirm any observed mutation. The PCR fragments were purified by electrophoresis in an agarose gel and extracted with the Qiaquick gel extraction kit (Qiagen) before to be introduced in the sequencing plasmid vector pCR4-TOPO. Chemically competent Escherichia coli Top 10 bacteria were transformed with an aliquot of the reaction according to the manufacturer instructions (Invitrogen, USA). Plasmids from recombinant bacteria were prepared with the QIAprep miniprep kit (Qiagen, USA) and analysed by restriction enzymatic digestion. The fragment-containing plasmids were sent to a genomics company for sequencing (Cogenics, Meylan, France).

The mutations observed after the two sets of PCR amplifications were considered as true polymorphisms whereas mutations observed after a single set of PCR were considered as errors introduced by the Taq polymerase in the amplification step. The sequencing traces were aligned with the SeqMan software (Lasergene, Germany). The derived nucleic and protein sequences were aligned by the MegAlign software (Lasergene, Germany) using the clustalW algorithm and compared to the lsa-3 sequence of the P. falciparum laboratory strain 3D7 believed to have originated from Africa 18. Sequences coding for repeats in the molecule were further aligned manually (Additional file 1). The sequences corresponding to the short intron (Figure 1) were ignored because of the difficulty to reliably amplify this region composed by long strings of A and T nucleotides. Numerations of nucleic and protein sequences were done according to the sequence of lsa-3 mRNA and LSA-3 protein of the P. falciparum strain 3D7.

The amplification of the gene was achieved by using PCR primers based on the sequence of the previously cloned and fully sequenced lsa-3 gene from the strain K1 of P. falciparum 7. A scheme of the DNA sequence of lsa-3 in the generic strain 3D7 is shown in the Figure 1 with the location and size of the six fragments amplified. The primer pairs (Table 1) were fully operative on the 20 clinical isolates analysed, even with a stringent annealing temperature of 57°C in the PCR programme.

The sequences of the full length of the lsa-3 gene from the 20 clinical isolates were compared with the published sequence of the laboratory strain 3D7 18. The number of punctual mutations in the sequence coding for the large non-repeated regions of the molecule among the 20 isolates was remarkably low with only 15 single nucleotide polymorphisms (SNPs) out of 3444 base pairs (Figure 2 and Additional file 2). Five of these were synonymous mutations (33%) whereas 10 were non-synonymous resulting in amino acid changes. This was not unexpected because the full sequencing of lsa-3 in the laboratory strain K1 of Thai origin reported six SNPs along the same sequence with two synonymous mutations (33%) 7. Moreover, the complete conservation at the genetic level of a sequence coding for an immunodominant region of LSA-3 (lsa-3_410-775) that was previously observed in 27 samples from diverse geographical areas 10 was confirmed in the samples analysed in this study. These frequencies are much lower than in other P. falciparum vaccine candidates, such as for example PfAMA1 with 130 non-synonymous mutations out of 1,800 base pairs in 13 Indian samples 19. The silent mutation a3882c identified on 12 out of 18 Gambian isolates in a sequence coding for la90 20, a peptide recognized by the molecule of the histocompatiblity complex HLA-B53 that is associated with resistance to severe malaria in West Africa was not observed in this study.

The singleton variations could be classified in three categories (Figure 2). Firstly, singletons observed in a single isolate as a157g in Com 119, t247c in Sen 5522, a1778g in Sen 5514, g2629a in Sen 5505, and c3556t in Sen 5510. Secondly, singletons or group of singletons that were associated with a group of isolates; a1957g, c1998t and g3108a in Bra 1905 and Bra 1884, g2136a in Sen 5510, Sen 5514, Tha 28, Tha 52 and the laboratory strain K1, c4340t in Brazilian isolates 1915, 1882 and 1853. Thirdly, singletons seen in most of the isolates compared to the generic sequence of the strain 3D7, a237t, g2144a, g3108a, a3786g and t4656a. A deletion of the codon gaa (lsa-3_4121-4122) was detected in isolates Bra 1905, Bra 1884 and Sen 5522.

The very low number of SNPs in lsa-3 compared to most of the actual malaria candidate vaccines, mainly expressed during the asexual blood stages (ABS), might relate to the PE expression of the protein. The PE antigens are likely less exposed to immunological pressure; Firstly, because of the tremendously lower numbers of PE schizonts as compared to ABS schizonts 21. Secondly, once injected in the host blood by an infected mosquito, the sporozoites infect liver cells within a few minutes were they expand and mature inside hepatocytes hidden from antibodies. Finally, the liver maintains a tolerogenic response towards incoming harmless antigens 22 that might favour the development of the parasite.

However, the sequence of msp-3 coding for the C-terminal region of the merozoite surface protein-3, which is expressed on the surface of merozoites in ABS that are more exposed to immune surveillance, was also remarkably conserved in the same clinical isolates 23. These results suggest that, independently of the stage of expression, mechanisms such as structural constraints may drive the genetic stability observed in these antigens and act against the occurrence of mutations.

The amplification of the sequence coding for the repeated contiguous regions I and II of LSA-3 showed considerable size variation ranging from 264 nucleotides in the isolate Bra 1853 to 1800 in the K1 strain, respectively (Figure 3). The figure 4 illustrates the corresponding number of tetrapeptide repeats ranging from 22 in the sample Bra 1853 to 150 in the laboratory strain K1.

The region I is composed of 14 tetrapeptides in the strain 3D7. Deletions of six tetrapeptides in isolates Bra 1915, Bra 1882 and Bra 1853 and insertions of two tetrapeptides in Com 524 were detected (Additional file 1). The polymorphism E253G was observed in all strains but not in the strain 3D7. This suggests that the glutamic acid residue at the position 253 of the LSA-3 protein in the strain 3D7 could result from an artificially introduced mutation during the sequencing process of the gene.

The region II contributes the most to the repeat length heterogeneity between isolates (Figures 3 and 4, Additional file 1). Unique tetrapeptides were identified in isolates Sen 5522 (DDGS, VASS, IASS, VDES, IDSS), Sen 5514 (VVEN), Com 119 (VAED, VAEK) and Com 524 (VVPS).

The region III show little variations compared to the sequence of 3D7 composed of eleven tetrapeptides with only a tetrapeptide deletion in all strains but Tha 28 (no deletion) and Com 119 (two deletions) (Additional file 1). The repeats composition were identical in all strains apart from two polymorphisms in Sen 5533 changing a tetrapeptide IDED to a unique IEEN.

The results show that apart from some unique repeats in region II, the composition of most of the tetrapeptide repeats was identical in all strains but solely their number and organisation in the repeated regions varied (Figures 3 and 4, Additional file 1).

Overall, twenty-two SNPs scattered along the lsa-3 sequence (Figure 2 and Additional files 1 and 2) were identified. The polymorphisms in the repeated regions of lsa-3 consisted mainly in insertions, deletions and/or reorganizations of sequences coding for the same tetrapeptide units in LSA-3. Although the number of clinical isolates analysed did not allow to draw definitive conclusions, the samples from Brazil seemed to contain the shorter repeated regions whereas samples originating from Thailand with the laboratory strain K1 contained the largest (Figure 4). At a time when the polymorphism of LSA-3 had not yet been investigated, it is this particular criterion, of longer length, that led us to choose K1 for molecular characterization of the gene. Hence, the particular length of LSA-3 in K1 is not a surprise, and results now indicate that this size is only due to repeats duplications.

The strong disparity in the number of repeats in region I and II suggest that the length of this area of the molecule is dispensable for the fitness of the parasite. The repetitive organization of these sequences usually produces B cell epitopes that are immunodominant in other genes, as compared to non-repeated regions. It has been proposed that repeated sequences might act on the intensity and quality of the immune response 24 and thus contribute to the immune escape of the parasite. A putative escape mechanism could be to divert the immune response from protective epitopes towards these repeated regions of the parasite antigens. However, in the case of LSA-3, detailed immunological studies in hyperendemic areas revealed that B-cell epitopes defined in non-repeated regions were as much the target of antibodies as the repeat blocks 13. Hence, in contrast to other genes encoding repeats and particularly Glu-rich repeats, the Glu-rich block of LSA-3 does not seem to be immunodominant. The presence of such repeats in several malarial antigens and the network of cross-reactivity they generate across those molecules have been frequently stressed, however their functions remain poorly understood.