This study indicates that Peruvian P. falciparum populations expanded from bottlenecked populations or migrants in the post-eradication era. During the peak malaria incidence (1998–2000), Peruvian parasite populations consisted of at least five clonal lineages with varying drug resistant genetic backgrounds. Since the 1990s the increase in the transmission intensity has favored sexual recombination, especially in the Central Amazon region. Simultaneously, changes in drug policy seem to have been a critical selective force. We argue that the combination of increased opportunities for outcrossing and changes in drug policy strongly influenced population structure and parasite evolution between the study time points.
Previous studies have suggested that low transmission could maintain linkage disequilibrium in P. falciparum 
. However, the mechanisms underlying population differentiation in low transmission areas have not been explained, beyond invoking genetic drift 
or, rarely admixture 
. Our study clearly demonstrates how these factors have influenced parasite populations. We demonstrate how clonet expansion and migration can alter parasite population structure in a region after the era of eradication. We show that multiple clonets can be maintained in sympatry if there is not sufficient transmission for outcrossing.
Historical malaria eradication may have reduced parasite populations to a patchy network of refuges in areas of low P. falciparum
transmission like South America 
. These refuges could be interpreted as the kind of islands that Sewall Wright addressed with his island model of population structure 
. In the island model, the population is subdivided into subgroups with each breeding within itself, except for migrants 
. These subgroups can be due to geography, ecology, or time 
. In the case of South America, malaria eradication would have led to separated, allopatric parasite refuges. These, in turn, would have eventually led to clonal lineages developing through genetic drift and inbreeding. After malaria eradication efforts waned, such populations would have expanded locally and begun to migrate to other islands of inbred parasites. The opportunities for outcrossing between these coexisting, sympatric, clonets would have been limited because of the low frequency of multiple infections in a low transmission environment.
The demonstration that Peru had at least five clonal lineages of P. falciparum
in the post eradication era is consistent with other recent studies around Iquitos 
. In one, whole genome sequences of 14 P. falciparum
isolates led to the conclusion that the isolates were largely identical, with at most four parental haplotypes 
. Another argued that five independent clusters of related subpopulations exist based on microsatellite data collected between 2003–2007 
. Our study was able to draw more general conclusions because it utilized samples from across the country. Furthermore, many samples were collected around 1999, after the peak of the malaria resurgence, which suggests that only a few lineages were involved.
Genetic drift caused by self-mating, mitotic replication, or other processes is an important source of genetic diversity in areas of clonal propagation like South America. Over the short term, however, clonets might have been maintained if there was only one clonet present in each locality, as on the Pacific coast of Peru, or had there been insufficient transmission for outcrossing to occur among clonets. We argue that the increase in malaria transmission in the late 1990s led to admixture and cryptic parasite population substructure (). Furthermore, we argue that the rapid increase in malaria transmission during the late 1990s led to sufficient multiple infections for clonet outcrossing to occur and that this was a greater influence on genetic diversity than the genetic drift that likely generated the clonets. Another recent study has confirmed the frequent occurrence of recombination in this region as 33.6% of infections examined in a cohort study involving villages near Iquitos between 2003–2007 were mixed, with a decrease in LD over time, and recombination in later years 
Clonal substructure was found at all of the collection sites in the Amazon interior (clonets A, B, C, D, and E) and a single clonet expansion was found on the northern Pacific coast (clonet E). The Andes appeared to act as a semi-permeable barrier to gene flow, but, within the Amazon interior no barrier was found. Though the coastal (clonet E) and western Amazon sites (C and E vs. D and C) do not share many clonets, Caballococha (A, B, C) and Padre Cocha (A, B, C, D) shared three. The presence of a few clonet E isolates in Pampa Hermosa suggests a recent introduction, perhaps in the early 1990s, to the Western Amazon by way of Andean roads that terminate near Pampa Hermosa 
. The comparative genetic diversity seen in Padre Cocha is not surprising given its close proximity to Iquitos, where ~42% of Loreto's population lives 
. Iquitos has a large enough population to support multiple lineages, is a hub of human movement, and many P. falciparum
cases were reported there, allowing for potential multiple infections and recombination.
Historical facts about the origins of CQ and SP resistance in different time points allow us to speculate when different clonets may have migrated to Peru. The first reports of CQ resistance in Peru occurred in the eastern Amazon in 1979–1980, while parasites to the south remained CQ-sensitive 
. Peruvian reports occurred decades after the first continental reports from Venezuela (1959; 
) or Rondônia, Brazil (1960; 
). It was suggested that resistance may have spread from coastal and interior Ecuador, where it was reported as early as 1976 
. However, nearby Rondônia, Brazil might have also been a source of Peruvian CQ-resistant parasites. Molecular data suggested at least two independent origins for CQ resistance in South America 
, which we identified as the pfcrt
CVMNT-B allele (coastal and western regions of South America) and the CVMNT-A/SVMNT-A alleles (Amazon). Our data indicates that these two CQ resistant lineages have colonized Peru with the CVMNT-A/SVMNT-A alleles mostly occurring in the eastern and central Amazon region and the CVMNT-B allele on the Pacific coast and in several sites in the Amazon 
. SP resistance was proposed to have developed in the Amazon as early as the 1970s and spread regionally 
, but was not reported in Peru until the 1990s.
We were able to hypothesize when the five clonets were introduced into Peru based on their drug resistance profiles. Clonets A and B always had the Stct
allele associated with highly resistant dhfr
alleles (noted in the Amazon by others as well 
), along with the pfmdr1
α lineage. Therefore, it appears clonet A and B may have swept through the Amazon basin from Brazil and expanded during the late 1990s when SP was introduced for primary treatment of malaria in Peru (). The absence of the A and B clonets in the western Peruvian Amazon and the coast may be due to their recent introduction to Peru, limited internal migration, lack of widespread SP use, control efforts, the Andes Mountains, and/or differences in vector populations. Our interpretation is consistent with our data suggesting recent population expansion of clonet B. We stress that testing for MDE, bottlenecks, and population expansions requires polymorphic markers, which are difficult to locate in clonets. Clonet A did not share a similar signature of recent expansion, which might have been due to an earlier introduction to the region, a more diverse founding population, or outbreeding with clonets B and/or C based on shared neutral markers.
Hypothesized Spread of Clonets Across Peru.
Clonet C carried the CVMNT-A allele, lacked highly resistant SP- resistant genotypes, and was only found in Padre Cocha and Pampa Hermosa. Indeed, a previous study using samples from across South America only reported this allele in Padre Cocha 
. Furthermore, a CVMNT allele that grouped with SVMNT was reported among 12 samples collected in Iquitos and two samples from Tabatinga, Brazil, which neighbors Caballococha 
, suggesting clonet C may have also been found in the Amazon basin bordering Peru. Clonet C carried mostly SP sensitive dhfr
genotypes, which may share ancestry with the alleles found in clonet A and B. Therefore, it is reasonable to argue that clonet C may represent a vestige of an ancestral Amazonia lineage which has been present in Peru for a sufficient period to allow recombination with other clonets. It may be a remnant of the CQ-resistant lineage hypothesized to have developed in southern Rondônia. This would imply it entered eastern Peru from Brazil sometime after the development of CQ resistance in Rondônia in 1960, but prior to the development of SP resistance ().
Clonets D and E had divergent neutral backgrounds and were associated with the CVMNT-B allele, the pfmdr1
β lineage, but not the highly resistant SP genotypes. Based on these commonalities between the clonets D and E, we propose that these two clonets may have spread from the Colombia to Ecuador and then Peru (). Clonet D carried a unique 108-C dhfr
haplotype and a unique synonymous 540 dhps
mutation not reported elsewhere in South America, potentially indicating genetic drift. Clonet D may represent the CQ-resistant coastal lineage that was argued to have spread from the coast of Ecuador between 1976 and 1980 
. It may have spread from Padre Cocha and Caballococha into Ullpayacu or from a bordering Ecuadorian site 
. Our findings suggest that clonet C and D may have been in Peruvian Amazon longer than clonet A or B, though the recent history of clonet E is ambiguous.
Clonet E was likely introduced to coastal Peru from a more direct coastal migration sometime after 1976 and recently spread into the Peruvian Western Amazon. Clonet E had the least microsatellite variation of the clonets and our analysis suggests this clonet has undergone a bottleneck. However, there was a rapid increase in malaria cases in the region during the 1990s and our study suggests that only clonet E was present. The lack of a statistically significant population expansion might have been caused by more than one clonet E-like lineage invading Peru since 1976 or the sheer lack of genetic diversity overwhelming the statistical test. However, there was a rapid increase in malaria cases in the region during the 1990s and our study suggests that only clonet E was present.
Our findings suggest that the two pfmdr1
lineages, α and β, have different geographical distributions. Based on our data we predict the α haplotype (found in clonets A, B and a major subset of C, and a few from D) evolved in the Amazon interior of South America while the β pfmdr1
haplotype (clonets D, E, and to a lesser extent C) evolved on the Pacific Coast or the nearby western Amazon interior. The breakdown in this pattern in clonet C and D is presumably due to outcrossing. This is consistent with a study of pfmdr1
from Colombia, Brazil, and Guyana, which found that pfmdr1
haplotypes from Colombia and Guyana were quite distinct 
Clonet breakdown in Iquitos (2006–2007) indicates recombination between four different clonets multiple times over the preceding 7–8 years due to increased transmission. There were only a small number of original clonets seen in this region in 1999 (clonets A,B,C, and D) that persisted until 2006–2007 (). The remaining isolates appeared to be recombinants of clonets B and C, C and D, or, in a few isolates, recombinants of clonets B, C, and D or A, B, and C. We suggest the frequencies of these recombinants are at least partly influenced by the replacement of SP with ACT in this region in 2001. In support of this argument, the SP resistant clonets A and B rapidly declined by 2006–2007 and we showed earlier that the SP resistant genotypes they carried had significantly declined during this period 
. But beyond this, there was a distinct absence of clonet hybrid offspring which carried SP resistant alleles in 2006–2007 (), even though parasites carrying these alleles had been prevalent in 1999. If such sexual recombination was just as likely to have occurred, the marked decline of offspring with SP resistant profiles suggests they had lower fitness after SP was removed. These observations illustrate that the parasite clonal lineages underwent rapid changes in population structure due to the increase in malaria, which allowed for outbreeding at the same time that new drug policies altered selection pressure.
Given the large scale migrations within the Amazon basin, periodic epidemics 
, and changes in drug policy there may have been opportunities for outcrossing to occur outside of Peru along with selection pressure from various drugs. Yet it appears that SP resistance may be fixed in the remaining Amazon basin 
and therefore, multiresistant parasites are likely to persist. This is true in Venezuela, where pfcrt
, and dhps
were linked 
and other parasites were in linkage for various antigenic genes 
. However, SP sensitive parasites could disperse into the greater Amazon from Peru due to the development of the Interoceanic Highway, which will connect Atlantic ports of Brazil with Pacific ports of Peru. As countries in the region have moved away from SP use, these SP sensitive parasites may outcompete the already established resistant parasites. Indeed, such migration may have already occurred based on the presence of SP sensitive dhfr
alleles in the Colombian Amazon 
. In summary, our study suggests that the molecular characterization of population structure and drug resistance profiles of P. falciparum
will provide valuable insight into how control programs influence the underlying dynamics and evolution of parasites. Such studies may help to predict the genetic profiles (eg: drug resistant profile) of vestigial parasite populations during and after malaria elimination programs and predict the genetic profiles of parasites that may reappear in subsequent outbreaks.