Genetic diversity and population structure of Plasmodium falciparum in Thailand, a low transmission country

The parasitized blood samples were collected from patients who attended for malarial diagnosis at clinics in seven provinces of Thailand from 2002 to 2007 (Figure 1). All patients had uncomplicated malaria. Five of these provinces (Tak, Yala, Maehongson, Kanchanaburi and Ranong) were ranked in the top ten provinces for the occurrence of P. falciparum based malaria incidences in Thailand [29]. In 2006, the malaria incidence in these provinces ranged from the highest in Tak, at 8,648 cases; through 3,544, 2,411, 1,250, 1,072 and 629 cases in Yala, Maehongson, Kanchanaburi, Ranong and Trat provinces, respectively. In addition, although there have been no cases reported in Ubonratchathani since 2005, in 2004 there were 979 cases reported in this province [29]. Four of these seven provinces, Maehongson, Tak, Kanchanaburi and Ranong, are located along the Thailand-Myanmar border in the west of Thailand, whilst Ubonratchathani and Trat locate along the Thailand-Laos and Thailand-Cambodia borders, respectively, in Eastern Thailand. Lastly, Yala is located in the southern part of Thailand, close to Malaysia (Figure 1). There is a relatively high exchange rate of people who move between international borders, especially between Thailand and Myanmar [4]. The samples were collected from Mueang and Mae Sot districts in the Maehongson and Tak provinces, respectively, where the highest malaria incidence was reported, whilst in Kanchanaburi and Ranong provinces, the blood sample collections were conducted in the Sai Yok and Mueang districts, respectively. Finally, in the Yala province, the sample collection was performed in the malaria clinic in the Bannang Sata district, where the malaria incidence has increased in frequency in relation to recent political violence in the three southernmost provinces (Yala, Narathiwat and Songkhla) since 2004. The malaria incidence in Yala increased annually between 2004 and 2006 at 1,903 to 3,544 cases. Because of the difficulty of access to this, only a single year of samples were obtained from Yala. In the eastern part, the blood samples from Na Chaluai and Buntharik districts in Ubonratchathani province and Bo Rai district in Trat province were obtained. These two provinces had a lower malaria incidence than that of the other five.

The P. falciparum-infected human blood samples from two sources in Thailand were used in the study: 1) samples from in vitro culture, which had been preserved in liquid nitrogen at the Malaria Research Programme, College of Public Health Sciences, Chulalongkorn University, and 2) blood samples from P. falciparum malaria patients preserved on filter papers. Blood samples collected before 2004 were in vitro cultivated after blood collection from the malaria patient, and the cultivated samples were then cryopreserved in liquid nitrogen until being used in this study. The DNA was extracted from cultured parasites by standard phenol/chloroform extraction of whole blood. Thereafter, between 2004 and 2007, parasitized blood samples were collected from thick film-confirmed falciparum malaria patients, before drug administration. These patients were informed of the objectives of this study, agreed to participate in it, and signed a standard consent form. Blood samples were taken from finger-pricks of P. falciparum- infected (malaria symptomatic) patients, and were absorbed onto 3 MM Whatman filter paper and air-dried. DNA from the dried blood sample was subsequently extracted using a DNeasy Tissue kit (Qiagen) according to the manufacturer's instructions. The blood collection protocol was approved by the ethical committee of the Institute of Health Research, Chulalongkorn University. Socio-demographic information of patients was collected by interview, especially the most likely place of infection, to help identify the parasite's origin. The provincial origins and year of collection of the infected blood samples, and the likely country of infection source where not Thailand, are summarized in Table 1.

For genotyping P. falciparum from each infected blood sample, the DNA was extracted and used as template for the PCR based amplification of 12 established polymorphic and apparently neutral microsatellite loci distributed on eight of the 14 different chromosomes [30], as listed in Table 2. Both a one-step PCR and two-step semi-nested PCR strategy with fluorescent end-labeled primers were used for the microsatellite amplification, as described in Razakandrainibe and colleagues [22], following the methodology developed by Anderson et al [30]. For comparison of the international samples and with the published literature, the first seven microsatellite markers listed in Table 2 were used, as well as for screening if the samples were single or multiple isolate infected (see below). All Thai samples were further screened with the other five loci. The size (length) of the amplified microsatellite repeats were resolved and sized relative to an internal size standard on an ABI Prism 310 Genetic Analyzer using Genescan software. Thus, and importantly for the analysis, size variation can be derived from indels in the flanking sequences as well as in changes in the microsatellite core repeat unit numbers, whilst substitution mutations in any part of the amplicon are overlooked.

Only isolates that successfully amplified at all 12 loci within the Thai samples or at the selected seven loci for international samples were included in the respective analyses. Moreover, for population genetics, only those samples that showed a single P. falciparum clone infection were analysed, because of the difficulty in defining multi-locus genotypes in samples with a polyclonal infection. To this end, given that these 12 loci are all single genome copy, not in linkage, and that P. falciparum from the blood (schizonts and early trophozoites) has a haploid genome, then isolates with more than one allele at more than one locus were deemed to be from multiple infections. Genetic diversity was measured by the allelic richness per locus and sample (R_S) and Nei's unbiased expected heterozygosity (H_S), after adaption to haploid data, using FSTAT, v2.9.4 [31]. Differences in the allelic richness and expected heterozygosities were tested by an exact Wilcoxon rank-sum test using the SYSTAT software.

Linkage disequilibrium analyses between pairs of loci were performed with FSTAT v2.9.4. Genetic variation and differentiation were analyzed using FSTAT. Canonical correspondence analysis (CCA) was carried out to illustrate the measures of population structure, using CANOCO^® software [32, 33]. Only isolates scored at each locus were considered for CCA. Analysis was performed with only the single infected samples. The significance of the canonical axes was tested with a Monte Carlo permutation test [32], which also allowed estimation of the 95% confidence intervals of the centroid of each population.

Because the haploid stage of P. falciparum, i.e. schizonts and early trophozoites, present in human blood samples were analysed, the presence of multiple infections was assessed in a given isolate by the presence of more than one allele at any of the microsatellite locus surveyed. The minimal multiplicity of infection for each sample (i.e. the number of genetically distinct parasite genotypes in each sample) was analysed locus by locus, and was estimated from the locus that exhibited the highest number of alleles in a given isolate. Therefore, a single P. falciparum isolate infection is deemed to correspond to an isolate exhibiting only one allele at each of the 12 loci investigated. Of course, multiple infections that are genetically distinct overall but identical at the 12 microsatellite loci, for example recombinants or new mutations at other sites, or whose differences lead to null alleles and so would not be detected, would not be distinguished. Although the former would still be likely to be closely related isolates all the same, the later maybe quite different and thus the data represents the minimal rather than the actual level of multiple infections. Accordingly, the minimal proportion of multiple infections in each sample group was assessed for each locus and between loci in a sample group (Table 3). In the case of multiple infections, it is impossible to match the different alleles of each distinct locus to reconstruct a valid, rather than a chimeric (random) genotype, and so the analyses of the population structure must thus be conducted on single infected isolates only. In total, 523 blood samples from seven provinces of Thailand were genotyped at 12 microsatellite loci. Amongst these, 42 samples were unsuccessfully genotyped due to missing data at any loci, leaving 481 samples that were successfully genotyped at 12 polymorphic loci. Of these, 135 samples were found to be multiple P. falciparum isolate infected (28%), leaving 346 single isolate infected (72%) samples suitable for genotyping and population genetic analysis. The multiple infected samples were distributed across six of the seven provinces assayed, with frequencies of multiple infections as a ratio of infected individuals to the total number of samples ranging from 44% in Maehongson to 22% in the Ubonratchathani population (Table 3). An important result was that, in the Yala population, all 31 isolates analysed displayed an identical haplotype (i.e. genotype) across all 12 loci. However, since they were only sampled in one year this may reflect a transient expansion of a single clonal focal point, rather than a stable but isolated (inbred) population.

For the population genetic analysis, based on 12 polymorphic nuclear microsatellite loci, only the 346 single P. falciparum isolate infected samples were considered. The number of alleles per locus varied from 4 to 20 alleles (Tables 2 and Additional file 1), and the allelic richness varied from 1.0 (Yala for all loci) to 10.1 (Tak for locus C1M8) (Table 4). When the average allelic richness per locus was compared amongst the seven Thai populations, Trat and Yala populations had a significantly lower allelic richness (Table 4). Gene diversity per microsatellite locus varied from 0.391 (locus TA109) to 0.841 (locus C1M8), whilst the gene diversity (H_S) in each population was 0.68 ± 0.23, 0.63 ± 0.25, 0.60 ± 0.18, 0.70 ± 0.21, 0.56 ± 0.25, 0.62 ± 0.20 and 0 in the Maehongson, Tak, Kanchanaburi, Ubonratchathani, Trat, Ranong and Yala populations, respectively.

Linkage disequilibrium (LD) was tested at three levels, (i) for the whole population (ii) for the three regional areas; and (iii) amongst sites.

Genetic differentiation (F_ST) was estimated using the FSTAT software. A permutation test was applied (n = 10,000) permuting alleles amongst all samples for all loci to test whether F_ST significantly differed from zero. A significant differentiation (p-value = 0.0024) of P. falciparum populations between sites was observed, with a very high differentiation between the Yala population and all the other populations (Table 5). The analysis of the genetic differentiation in P. falciparum populations between the three regions revealed a significant level of differentiation (i.e. p-value = 0.017) between the western, eastern and southern parts of Thailand with F_ST values equal to (i) 0.0478 for the western versus the eastern regions, (ii) 0.1321 for the western versus the southern regions and (iii) 0.1199 for the eastern versus the southern regions.

Three hundred different genotypes were obtained from the 346 single P. falciparum isolate infected individuals, with 14 identical haplotypes (Table 6). Thus, essentially identical genotypes were found from patients:

- in the same year but infected in two different areas: In Kanchanaburi, an identical genotype was found in seven individuals infected in 2007 where four of them were infected in Myanmar and three of them infected in Thailand,

- in the same site but collected in two successive years: In Ranong, an identical genotype was found in two individuals infected in 2005 and 2006, and another identical genotype was found in 2006 and 2007,

- in Yala, only one haplotype was found in 31 parasite samples.

These genotypes which are found in more than one individual would have an impact on the epidemiology of the disease because, (i) they can pass through several generations, and (ii) these genotypes could carry important traits, such as drug-resistance genes.

In order to compare the population genetics and structure of the Thai P. falciparum populations with the already published reports for other populations, the observed patterns at seven common microsatellite loci (Poly α, TA60, ARA2, Pfg377, PfPK2, TA87 and TA109), were compared with those observed from blood samples of two African countries, Congo (n = 15) and the Cameroon (n = 35), and one South American country, French Guyana (n = 137) (unpublished data). The gene diversity in the Thai populations (H_S = 0.65 ± 0.17) was lower than those found in the two African countries (H_S = 0.81 ± 0.07, and 0.78 ± 0.18 for the Congo and Cameroon, respectively), but higher than that of French Guyana (H_S = 0.54 ± 0.08). Significant LD was observed in the two low transmission countries (12 and 6 out of 21 combinations for Thailand and French Guyana populations, respectively). The high LD value observed in Thailand is due to the Yala sample. The canonical correspondence analysis (CCA), searches for multivariate relationships between two data sets (e.g., genetic data of P. falciparum and countries environmental data) [33], was also used to determine the relative contribution of the countries (i.e. sites) to the global genetic structure of P. falciparum populations. A graphic representation of the results showed overlapping distributions (i.e. centroids and ellipses of the 95% CI – Figure 2) of the genetic variability of P. falciparum observed amongst sites. A Monte Carlo permutation test on the first four canonical correspondence analysis axes that combines parasite genotypes and sites is statistically significant (p-value = 0.001). This suggests some degree of geographic differentiation. Indeed, all populations in different countries displayed significant genetic differentiation except between the two African countries, Congo and Cameroon (Figure 2).

At the level of continents, the genetic structure of P. falciparum populations in different continents revealed that African countries with high transmission intensities had the highest population genetic diversity. The genetic diversity decreased with low transmission intensities, as in Thailand (Asia) and French Guyana (South America). The average genetic diversity of P. falciparum populations in Thailand (H_S = 0.65 ± 0.17) was, however, close to that of the high transmission intensity level of Papua New Guinea [5], and slightly higher than that reported for populations in Malaysia [19] and Brazil [20], also based on assaying genetic diversity at these seven microsatellite loci.

The deduced mating patterns of P. falciparum in Thailand showed a deviation from random mating, in congruence with the French Guyana population, whilst, in contrast, the two African populations showed likely panmixia. The proportion of multiple isolate infections, and the number of clones in an individual host, which are related to the transmission intensity of malaria parasites in any endemic area, are important parameters that affect P. falciparum mating, because fertilization may occur between gametes of the same or different genotypes in mosquitoes. Therefore, an inverse correlation between the proportion of multiple clone infections and the degree of linkage disequilibrium is expected. Even though the genetic diversity of Thai populations was observed at an intermediate level, and multiple isolate infections occurred frequently, a significant linkage disequilibrium was found which could be caused by, (i) a higher chance of inbreeding (selfing) in this parasite population, or (ii) structure formation within populations into sub-populations of each foci in Thailand (i.e. the Wahlund effect). The latter reason would be more likely to explain this finding because samples from all the foci from across the country (the seven different provinces) were pooled in the analysis.

In Thailand, the malaria incidence rate shows a high spatial heterogeneity across the country, with high incidence regions found nearby the borders with Myanmar and Cambodia [34]. This study reports the genetic structure of P. falciparum analysis with 12 polymorphic nuclear microsatellite loci in isolates from multiple populations in seven provinces of Thailand.

The P. falciparum populations in the areas bordering with Myanmar revealed a similar level of genetic diversity (0.60 <H_S <0.68), whilst the populations bordering with Laos and Cambodia had slightly higher or lower levels of genetic diversity at 0.70 and 0.56 in Ubonratchathani and Trat, respectively. Although the multiple isolate infection rates in those areas bordering with Myanmar (mean, 34%) were higher than that of Ubonratchanthani (22%) and comparable to that in Trat (36%), this could be explained by the notion that (i) the parasite mating in those areas is preferentially inbreeding, or (ii) the relatively small amount of sampled individuals in Trat and Ubonratchathani leading to a stochastic masking of the true population structure. The Yala population, however, showed an extreme difference in the gene diversity compared the other samples since it showed only a unique haplotype in all sampled individuals, but again may be a transient clonal expansion of an immigrant infection.

In contrast to the Yala population, a low but significant population structure was observed amongst the provinces that border with Myanmar, Laos and Cambodia. In the case of the Myanmar bounding provinces, the significant genetic structure detected between populations (KAN, TAK, MAE and RAN) was likely to be due to the fact that the pattern of human movement from neighbouring countries into each of these four provinces was different. In the Tak and Ranong provinces, the numerous movements in to Thailand from Myanmar from immigrant laborers, and from patients resident in Myanmar who cross the border to seek free malaria treatment, account for the high proportion (approximately 70%) of sampled individuals being likely to have been infected in Myanmar. Indeed, the same genotype was found in seven samples collected in 2007 in Kanchanaburi, from patients infected in Thailand and Myanmar. These, imported cases may play an important role on the P. falciparum gene pool in each endemic area and affect the epidemiology of the P. falciparum caused malaria in these areas. Besides this migration rate of humans in each foci, there is additionally the factor of the geographic barrier between Tak and Kanchanaburi provinces which may be another cause of genetic differentiation by partial isolation by distance and breaking down of panmixis. Thus, the genetic differentiation found between populations within each province of Thailand may reflect both a variation of the transmission rate in each location and the migration rate of people from one location to another. Certainly seasonal migration has been suspected as a leading cause of malaria transmission in these areas [35]. However, not only does the transmission rate and immigration of inhabitants play a role, but the differential dispersal of the Anopheles vectors in local areas may also have an effect on the parasite structure. In Thailand, Anopheles aconitus sensu lato, Anopheles baimaii, Anopheles dirus, Anopheles maculatus, Anopheles minimus and Anopheles pseudowillmori have been incriminated as important vectors of P. falciparum in humans, and they are differentially distributed throughout the country [36‐38], and may each have different vectoring capabilities for different isolates of P. falciparum. Thus, although differences in the population structure of P. falciparum between two vectors (Anopheles gambiae and Anopheles funestus) have been preliminarily investigated [39], no such information currently exists for the different vector species in each region of Thailand.

Importantly, the same multilocus genotypes of P. falciparum infected blood samples collected one year apart (2005 – 2006 and 2006 – 2007) was observed only in Ranong, where one case was found in a patient infected from Myanmar and the following year was found from a patient infected in Thailand. If a generation time is assumed to be two months, this parasite genotype has been transmitted through six generations without detectable change across the 12 loci (eight/14 chromosomes) due to recombination. This may imply that if this P. falciparum strain has an advantage (fitness) in that particular environment, it will propagate and expand in this population.

In this study, both a spatial and a temporal distribution of P. falciparum haplotypes collected in the same year and over two years were detected. A unique spatial haplotype, distinct from the other populations, was found in all patients in the Yala population, which were obtained from individual residents in four different sub-districts in Bannang Sata and other three nearby districts during May – July 2007. Three possible explanations are available to explain this trend. Firstly, that a P. falciparum isolate with this genotype had a high infectivity in the mosquito's midgut or carried some genetic characteristics, such as drug resistance, which gave it a significant benefit over other haplotypes. Secondly, that there is no inhabitant movement into this area due to the political violence, and so as a result no new haplotype parasite was introduced into this area (no gene flow). Thirdly, there maybe a rare new gene influx by immigration but which was subsequently lost due to selection or stochastic failure to progress to a stable infection threshold. Finally, is the possibility that the decrease in sanitary cover in this area allowed this particular genotype to spread into the population rapidly. Thus, a longitudinal study in the population dynamics in this area is required. Nevertheless, this is the first study to reveal the potential clonal expansion of a single haplotype in a malaria endemic area.