Plasmodium vivax contributes to a significant amount of morbidity in the Peruvian Amazon, accounting for 90% of all reported malaria cases in 2011 [1]. P. vivax is a predominant Plasmodium species in most areas of South America. The presence of the dormant hypnozoite stage along with reports of resistance to the most commonly used drugs to treat the blood stage, complicating treatment options for P. vivax complicate treatment options. Accordingly, the treatment and control of P. vivax remains an important issue to public health programmes.

P. vivax is characterized by the presence of a hypnozoite stage, which can persist in the liver for months to years, causing recurrent disease even though treatment and clearance of the initial blood stage infection were achieved. When a patient presents with recurrent P. vivax parasitaemia following treatment, there are several possible causes: 1) recrudescence of blood stage parasites; 2) relapse from latent hypnozoites; or, 3) a new infection (if the patient resides in an endemic area). Previous studies have shown that the risk and timing of relapse depends on the geographical origin of the initial infection. It is known that isolates of P. vivax from the tropics relapse sooner than those from more temperate areas, with 70% of relapses occurring within five months of initial infection [2]. The probability of relapse is higher than 20% for P. vivax patients when anti-hypnozoite therapy is not prescribed, and, thus, the ability to distinguish between relapse, recrudescence and re-infection is important in clinical efficacy studies [3].

Molecular genotype profiles of the parasites in an infection are critical to the assessment of clinical efficacy studies. These profiles may also be used to estimate the rate of new infections in populations that are continuously exposed. Increasingly, molecular genotyping has been used to differentiate treatment recrudescence from re-infection by comparing the profiles or haplotypes in a paired patient samples [4,5]. The probability that a given haplotype is identical in paired samples taken from two different time points greatly decreases with the use of a greater number of genetic loci and, also, with loci that have a high amount of polymorphism within a parasite population [6]. Using antigen-encoding genes is problematic as they are affected by immune selection. These genes alone or in concert are not able to distinguish recrudescence from re-infection with high levels of confidence [2].

Highly variable neutral microsatellite loci are ideal tools to assess population-level studies of Plasmodium falciparum and, more recently, P. vivax[7-16]. Complicating factors include the existence of multiple infections or clones within a single host and, in the case of P. vivax, the presence of dormant hypnozoites that persist for some time after the initial infection. However, studies that employ population level sampling and estimates of variation at each of the loci, in addition to paired patient samples taken at different points in time during an infection, can yield estimates of recurrence and drug failures [9,15]. An analysis of P. vivax populations and recurrent parasitemia was conducted. The haplotypes and the genetic variation allow for further assessment of paired patient samples as recrudescence/relapse or new infection.

The high levels of genetic diversity seen in these samples from the Peruvian Amazon are consistent with neutral levels of microsatellite and tandem repeat variation measured in P. vivax from multiple sites in Brazil, Colombia, and Peru [7,13,15,29-32]. This pattern is curious because it is entirely different from South American P. falciparum populations, as they show much lower levels of variation [33]. These differences could be explained by a number of factors, including differences in demography, natural selection due to drug pressure or differences in microsatellite dynamics between the two Plasmodium species. A low number of multiple clone infections in individual patients were observed in this study. This is an area where there is relatively lower transmission of parasites, and it is expected that these patients would experience less inoculations than patients in a hyperemic area of disease, and, thus a reduced chance for multiple clones existing within a single patient. Further, the specific composition of loci examined (i.e. the microsatellite motifs) and the effects of natural selection (i.e. strong immune selection on PALs) on these loci, may affect the detection of multiple infections [34].

There was evidence of population differentiation between Padrecocha and San Juan; this differentiation was not observed between Santa Clara and either of the two sites. This result could be due to a sample size bias, as only 54 samples were available from Santa Clara. Extensive LD among microsatellite markers has been noted in Colombian and some (but not all) Brazilian P. vivax populations. The existence of LD in South American populations is not surprising, given the low rate of parasite transmission and high levels of genetic inbreeding documented in many areas of the continent [13,29,35]. The Santa Clara population had less observed LD, but this population was represented by a smaller number of samples, which could affect the calculation of LD as it is dependent on the frequency of allele or loci pairs in a population. An alternative explanation could be that the Santa Clara population has experienced a larger amount of genetic recombination or mutation. Additional studies would be needed to better understand this observation.

There have been several studies that have utilized population-based estimates of genetic variation to calculate the probability that a pair of samples from the same patient would have the same genetic profile or haplotype by chance. P (identical), the probability that a random pair of haplotypes is identical, is affected by low frequency haplotypes in a population [36]. There are many low frequency haplotypes in each of the three Peruvian populations evaluated here. Another estimate, p (maximum), the maximum probability of two clones carrying the same haplotype by chance, requires independence of loci [9]. There was extensive LD between loci in the populations from Padrecocha and San Juan; thus, non-independence of loci was found. The third method, P(match), utilizes the population haplotype frequencies to assess whether the paired haplotype frequency occurs above or below a defined frequency value (for example, 0.05) [28]. Given the before mentioned limitations of the populations, P(match) was chosen as a method to assess the maximum probability of two clones or parasites carrying the same haplotype by chance in a population. P(match) has the limitation of requiring a population-level assessment of parasite diversity prior to assigning probability values. In areas with low genetic diversity, as is observed in South America, sample sizes must be relatively large in order to appropriately define the extent of diversity in the parasite populations [28].

Eighty percent of the paired samples examined by both microsatellite and antigen loci methods obtained the same overall result. Five out of six non-concordant paired samples (83%) were found to have different haplotypes at the two sampling times by microsatellite loci, but not by using antigen loci. This result is not unexpected, as microsatellite loci are putatively neutral with respect to selection, and, thus, should exhibit a greater amount of variation. Loci under strong selection, such as antigen loci, exhibit lower levels of variation in the population. These results emphasize the utility of using neutral loci to determine the P. vivax strain characteristics, and this can be illustrated by considering the application of these tests to assess a treatment regimen. If, for example, this study assessed drug failures amongst these 30 paired samples, results from antigen-encoding loci would lead one to believe that 12 (40%) patients potentially had a relapse of liver-stage parasites, whereas microsatellite loci would assess this rate at eight (26.7%) patients. It should be emphasized, though, that this type of simple interpretation has many caveats. Of note, the microsatellite methods used here employed a population-based level of sampling and analysis - such an approach is critical when assessing and interpreting genetic variation. Although use of PAL and microsatellite loci can be used in conjunction, researchers should use caution in the interpretation of the data and consider the effect of selection or neutral mechanisms on each of type of loci.

Interpretation of recurrent parasitaemia and drug treatment regimens in P. vivax is complicated by the presence of hypnozoites. Although the transmission of P. vivax in the Iquitos area of Peru is relatively low, the possibility of re-infection for the subjects in this study cannot be eliminated. Paired samples with different initial and recurrent haplotypes are the result of relapse with a heterologous hypnozoite or a new infection; here, 50% of the paired samples had different parasite lineages. Paired samples with the same lineage (here, 50%) are the result of recrudescence of the blood stage parasite (potentially a chloroquine treatment failure), relapse with a homologous hypnozoite (potentially a primaquine treatment failure), or a new infection with the same parasite strain. Furthermore, the population-level analysis here indicates that 90.2% of the paired samples that had the same haplotype were unlikely due to a new infection. These results indicate a potential large frequency of recurrent infections resulting from hypnozoite activation, either heterologous or homologous. This could arise from either an initial infection or multiple infections over time with different P. vivax strains, all resulting in dormant hypnozoites. The activation of heterologous hypnozoites has been noted before in patients from Asia [9].

There are limitations to microsatellite, tandem repeats, and PCR-based techniques to discern treatment outcomes. PCR bias, potential artifacts, and the inability to discriminate multiple clones are significant hurdles with these methods [14]. Indeed, microsatellite diversity is associated with the underlying repeat motif structure of each loci, and, thus, there may be an inherent bias in the results based on the loci that are chosen [32,34,37]. Careful consideration of the underlying motif structure of these loci is an important factor for researchers to consider. Further, the ability to rapidly screen many infections and assess genetic variation of the larger parasite population is key to understanding and interpreting these assays, irrespective of the locus. Population-based approaches have an advantage over smaller sampling strategies in both analyses and assessments [15].

In conclusion, the P. vivax parasite population in the Peruvian Amazon has considerable diversity and population differentiation. Antigen and microsatellite-based P. vivax genotyping methods can be useful tools to evaluate drug treatment regimens and characterize genetic profiles of parasite populations; however, researchers should be careful when interpreting data. Antigen-encoding loci do incorporate a bias due to immune selection. The description of the extent of hypnozoite activation in this population emphasizes the importance of treatments that target this stage. The P. vivax dormant hypnozoite stage poses challenges in the assessment of paired patient samples; nevertheless, population-based studies with appropriate methods can provide insights into the biology and epidemiology of P. vivax populations. The further development and assessment of molecular tools that allow a broad analysis of an individual parasite’s genetic construction will be immensely useful to drug treatment and resistance studies.

The authors declare that they have no competing interests.

AMM, VU, AAE, and DJB designed the study. AMM, VS, CS, and MS conducted the molecular genetics studies and data analysis. AMM, PCFG and DJB drafted the manuscript. SD, CC and DJB participated in the field study design and coordination of sample collection. All authors read and approved the final manuscript.