The low linkage disequilibrium (LD) and high recombination parameter across the PvDBP
II gene suggest that meiotic recombination partially accounts for the observed haplotype diversity. This result corroborates the findings of Martinez
et al. [
29], who also found high recombination rates in PvDBP
II among isolates from Papua New Guinea, Colombia and Korea. Using omegaMap with PNG sequences, similar results were obtained (see additional file
4: Spatial variation in recombination (ρ) and omega parameter (ω) calculated using omegaMap software among Papuan New Guinea PvDBP
II sequences). However, Cole-Tobian and King [
30] found a low recombination rate in PvDBP
II, but the test used (Sawyer's permutation test) is not appropriate for estimating the recombination rate in relatively short regions like this [
57]. The results presented here are supported by three different analyses: Haploview; PHASE and omegaMap. These results reinforce the importance of recombination as well as natural selection in the generation and maintenance of the genetic diversity of PvDBP
II. Multiple recombination events have also been described in other
P. vivax antigens, such as the Apical Membrane Antigen 1 (PvAMA1), Merozoite Surface Protein-3α (MSP) and MSP-1 [
48,
58,
59].
Previous attempts to infer the action of natural selection in PvDBP
II with different neutrality tests yielded inconclusive results. Martinez
et al. [
29] used Tajima's [
60] and Fu-Li's [
61] neutrality tests and found an excess of rare alleles for PvDBP
II, which is consistent with either the action of a selective sweep or the purifying natural selection (as suggested by the authors). A limitation of these tests is that the allelic spectrum is influenced both by natural selection and by the demographic history of the population, and it is not trivial to discriminate between the influence of these two variables without significant knowledge of the demographic history of
P. vivax populations, which is not the case. A second type of neutrality test applied to the PvDBP
II data compares the pattern of synonymous (assumed to be neutral) and non-synonymous polymorphisms (for which natural selection is possible). Because the demographic history of a population affects synonymous and non-synonymous sites equally, these tests are less sensitive to the potential confounding effects of the demographic history of populations. Applications of these analyses to PvDBP
II data include the McDonald-Kreitman test [
62] and comparisons between
d
N
and
d
S
estimated using the Nei-Gojobori method [
63], and these suggest the action of positive natural selection acting on PvDBP
II[
29,
30]
. However, the new method of Wilson-McVean [
33] has enabled us to map how the action of natural selection varies across PvDBP
II and to integrate this evolutionary information with an epitope analysis and structural modelling. Indeed, the Wilson-McVean method has four features that improve its performance in correctly identifying the effect of diversifying selection. First, it assumes a more realistic mutational model [
34] that allows for different rates of transitions and transversions and for differences in codon frequencies. Second, its fully Bayesian approach allows ω parameters and their associated HPD intervals to be estimated, providing a measurement of uncertainty. Third (and more importantly), the Wilson-McVean method uses a reversible-jump MCMC to explore the posterior distribution of ω, allowing it to vary along the length of the sequence. Although incorporating these more realistic assumptions implies a computational cost, the analyses have demonstrated their value. The Wilson-McVean method allowed us to map how diversifying selection has acted across PvDBP
II and to test the association between the presence of epitopes and the action of natural selection. Last (but not least), recombination, which has been important in shaping PvDBP
II diversity, does not affect the performance of Wilson-McVean estimations. These features, the fact that a highly polymorphic and functionally important region of PvDBP was examined for which results of functional studies are available, and the complementary structural and epitopes analyses partially overcome one of the limitations of this study: the small size of the studied region. A thorough examination of the entire
PvDBP sequence in different populations will provide a better understanding of how natural selection has shaped the different regions of the gene across different populations, including Brazil. However, on the basis of the results presented here, it is anticipated that, if the observed level of recombination affects the entire gene, signatures of natural selection might not propagate across large genomic regions. Therefore, population genetics methods that are able to examine signatures of selection at localized regions may be more appropriate for an evolutionary analysis of PvDBP.