Malaria outbreak in Laos driven by a selective sweep for Plasmodium falciparum kelch13 R539T mutants: a genetic epidemiology analysis

Summary Background Malaria outbreaks are important public health concerns that can cause resurgence in endemic regions approaching elimination. We investigated a Plasmodium falciparum outbreak in Attapeu Province, Laos, during the 2020–21 malaria season, using genomic epidemiology methods to elucidate parasite population dynamics and identify its causes. Methods In this genetic analysis, 2164 P falciparum dried blood spot samples were collected from southern Laos between Jan 1, 2017, and April 1, 2021, which included 249 collected during the Attapeu outbreak between April 1, 2020, and April 1, 2021, by routine surveillance. Genetic barcodes obtained from these samples were used to investigate epidemiological changes underpinning the outbreak, estimate population diversity, and analyse population structure. Whole-genome sequencing data from additional historical samples were used to reconstruct the ancestry of outbreak strains using identity-by-descent analyses. Findings The outbreak parasite populations were characterised by unprecedented loss of genetic diversity, primarily caused by rapid clonal expansion of a multidrug-resistant strain (LAA1) carrying the kelch13 Arg539Thr (R539T) mutation. LAA1 replaced kelch13 Cys580Tyr (C580Y) mutants resistant to dihydroartemisinin–piperaquine (KEL1/PLA1) as the dominant strain. LAA1 inherited 58·8% of its genome from a strain circulating in Cambodia in 2008. A secondary outbreak strain (LAA2) carried the kelch13 C580Y allele, and a genome that is essentially identical to a Cambodian parasite from 2009. A third, low-frequency strain (LAA7) was a recombinant of KEL1/PLA1 with a kelch13 R539T mutant. Interpretation These results strongly suggest that the outbreak was driven by a selective sweep, possibly associated with multidrug-resistant phenotypes of the outbreak strains. Established resistant populations can circulate at low frequencies for years before suddenly overwhelming dominant strains when the conditions for selection become favourable—eg, when front-line therapies change. Genetic surveillance can support elimination by characterising key properties of outbreaks such as population diversity, drug resistance marker prevalence, and the origins of outbreak strains. Funding Bill & Melinda Gates Foundation; The Global Fund to Fight AIDS, Tuberculosis and Malaria; Wellcome Trust. Translation For the Lao translation of the abstract see Supplementary Materials section.


Introduction
The parasite Plasmodium falciparum is the causal pathogen of the deadliest form of malaria, which causes hundreds of thousands of deaths every year, the majority of which are children in sub-Saharan Africa.Over the past two decades, substantial progress has been made towards reducing this burden, largely through public health interventions. 1,2Mortality has been greatly reduced thanks to highly efficacious artemisinin-based combination therapies (ACTs), which combine artemisinin derivatives, which are fast-acting parasiticidals, with a longer-lasting partner drug that clears surviving parasites.][5][6][7] Some of these strains are also resistant to partner drugs (mefloquine and piperaquine), [8][9][10] reducing ACT efficacy and forcing public health authorities to change front-line therapies.To protect ACT efficacy globally, and particularly in Africa, intense efforts have been undertaken to contain the spread of resistant strains and to eliminate P falciparum from the Greater Mekong Subregion within the next decade.Although these efforts have led to a reduction of malaria cases across the region, the resulting increased drug pressure has caused the emergence of new resistant forms, which present new public health challenges.It has become essential for Greater Mekong Subregion public health authorities to monitor epidemiological changes, such as localised malaria outbreaks, and respond appropriately to combat the resurgence of P falciparum.
Malaria outbreaks are generally thought to be caused by increased transmission, which can be driven by one or more factors.These factors include human behaviour, such as increased travel to higher-transmission forested areas; changing ecology or climate affecting the mosquito population; or importation of parasites into areas where elimination had been achieved. 11A less frequently considered factor is genetic selection, which might occur when a parasite strain expands in numbers after acquiring a genetic trait that increases its chances to survive or to transmit.Genetic traits might favour transmission in various ways-eg, by improving adaptation to local vector species, or by increasing gametocyte production. 12Human populations in lowtransmission regions such as the Greater Mekong Subregion develop lower levels of immunity than populations who are frequently bitten by infected mosquitoes.Hence, a higher proportion of infections progress to a symptomatic state and are treated with antimalarials, reducing the parasites' chances to transmit when the drugs are efficacious.Conversely, reduced drug sensitivity could allow more time for gametocytes to be produced before the infection is cleared. 3,12,13nalyses of genetic patterns in outbreak populations can be informative about the factors driving the outbreak.When genetic selection is the driver of an outbreak, offspring produced by self-fertilisation inherit the selected mutations and are more likely to survive, resulting in rapid expansions of identical individuals, and a collapse of genome-wide variation. 14Thus, an outbreak driven by selection might be characterised by reduced population diversity, decreased heterozygosity, and large clonal clusters. 5,15In contrast, outbreaks driven by ecological or human factors would tend to maintain or increase the outcrossing rate, maintaining typical or increased heterozygosity and allelic diversity in the population. 16herefore, in-depth analyses of parasite population structure changes can discriminate between different outbreak drivers, helping public health authorities decide on the most effective response-eg, whether to target forest workers, or investigate treatment efficacy.
Laos, a country in the Greater Mekong Subregion, is aiming to eliminate transmission of P falciparum by 2023.Since 2004, artemether-lumefantrine has been the firstline treatment for P falciparum, 17 which is only endemic in five southern provinces, and case numbers have been steadily decreasing.However, during the 2020-21 malaria season, the province of Attapeu had a sudden increase in

Research in context
Evidence before this study This study builds on previous work on the spread of multidrugresistant Plasmodium falciparum strains in the Greater Mekong Subregion.Here, we sought to identify the driving forces of a recent outbreak in Attapeu Province, Laos.We searched PubMed, without language restrictions from database inception to June 30, 2022, for literature pertaining to the underlying characteristics of malaria outbreaks.A search for outbreaks associated with human movement using the terms: "malaria", "outbreaks", "transmission", ("migration" OR "imported cases"), and "diversity" yielded 47 publications, including reports on high-diversity outbreaks caused by imported cases and recombination, and on parasite population shaped by migration of vectors and humans that maintain gene flows.Another search for the effects of genetic selection using the terms "malaria", "selective sweep", and "drug resistance" yielded 37 publications, including reports of reduced genetic diversity caused by selective sweeps of multidrug resistant malaria parasites.

Added value of this study
We analysed the genetic characteristics of outbreak parasites, using samples collected in Attapeu by routine surveillance.The outbreak was driven by the clonal expansion of LAA1, an artemisinin-resistant parasite population.Differently from the previously dominant KEL1/PLA1 strain, LAA1 carries the kelch13 R539T allele and no markers of piperaquine resistance.The expansion of LAA1 bears the hallmarks of a selective sweep, coinciding with regional changes in drug pressures.Despite the presence of two other minor strains, LAA1 parasites did not outcross, suggesting their selective advantage might be diminished through recombination.A large portion of the genomes of the three outbreak strains, including the regions containing kelch13, can be traced back to parasites circulating in western Cambodia around 2008-10.

Implications of all the available evidence
This study shows that malaria outbreaks, frequently ascribed to environmental or behavioural causes, can be driven by genetic selection.The clonal expansion causing the Attapeu outbreak coincided with the end of dihydroartemisinin-piperaquine use in neighbouring countries, which weakened the drug pressure sustaining the KEL1/PLA1 strain.The Attapeu outbreak populations are close descendants of artemisinin-resistant lineages found before 2010 in Cambodia, which persisted over long periods at very low frequency.These lineages have been very resilient over time, and can rapidly dominate when finding a fitness niche such as a favourable change in front-line therapy.Therefore, it remains crucial to monitor all kelch13 mutants; we have shown that genetic surveillance is a powerful tool to track these variants and explain the cause and dynamics of malaria outbreaks.
P falciparum malaria case numbers (130% increase; 378 cases in 2019 and 868 cases in 2020), at a time when other endemic provinces had a marked reduction (mean reduction of 65%).Attapeu is a largely forested province with low population density that borders with Cambodia and Viet Nam, with little international traffic.Malaria is mostly transmitted in the forest areas where Anopheles vectors circulate, and primarily affects forest workers and their communities.The GenRe-Mekong project has been conducting routine genetic surveillance of malaria in southern Laos since 2017 in collaboration with the Centre of Malariology, Parasitology, and Entomology of Laos, and collected 249 samples from Attapeu Province during the outbreak.These samples were genotyped with a broad set of markers, 18 and their genetic signatures were compared with those of parasites from previous seasons and other geographical locations.In this study, by use of genetic analyses we aim to investigate the outbreak parasite population dynamics, characterise the Attapeu outbreak strains, and reconstruct their origins.

Sample collection
In this genetic analysis, between Jan 1, 2017, and April 1, 2021, 2078 P falciparum dried blood spot samples were collected at 55 local public health sites in the southern Laos provinces of Savannakhet, Salavan, Sekong, Champasak, and Attapeu.Samples were collected from symptomatic patients of all ages who had been confirmed positive for P falciparum, using standard procedures of the GenRe-Mekong genetic surveillance project. 18Samples from Attapeu collected by the TRAC study in the 2011-12 malaria season were also included; the methods for collection of these samples have been described previously. 3No additional inclusion and exclusion criteria were applied to the selection of the patients.No additional inclusion or exclusion criteria were applied to the selection of dried blood spot samples.Patients signed a written consent form allowing the collection of blood spots for further analysis.For ancestry analyses, we used wholegenome sequencing (WGS) data from parasites included in the public release (Feb 24, 2021) of the MalariaGEN P falciparum Community Project v6.0, 19

Genotyping and WGS
Parasite genomic DNA was extracted from the dried blood spot sample, amplified, and processed on MiSeq sequencers (Illumina, San Diego, CA, USA) using the SpotMalaria v2 amplicon sequencing and genotyping platform.The genotyping process produced a standardised set of variants and haplotypes relevant to drug resistance, as well as a 101-variant genetic barcode.The genetic barcode provides a genome-wide summary of a sample's genetic variants and is used when comparing samples in epidemiological analyses. 18We filtered genetic barcodes to reduce errors due to data missingness, by removing barcode variants with more than 25% of missing calls, and samples with more than 25% of missing calls in the remaining variants.
For WGS samples, genotypes were called from Illumina short reads at 1 042 396 quality-filtered nuclear biallelic single-nucleotide polymorphisms (SNPs) using the standardised MalariaGEN P falciparum Community Project genotyping pipeline v6.0. 19Copy number estimations for the plasmepsin2/3 (pm23) or multidrug resistance 1 (mdr1) genes were performed by detection of a duplication breakpoint 19 or by quantitative PCR estimation. 20

Population structure analyses
Pairwise genetic distances (a measure of sample differentiation) were computed from genetic barcodes.Heatmaps of pairwise distance matrices were produced using the ComplexHeatmap R package (version 2.8.0). 21he same matrices were used as input to principal coordinate analysis, a technique that allows the grouping of samples by similarity.
Samples collected in Attapeu between Jan 1, 2011, and April 1, 2021, were clustered by creating a graph based on barcode genetic distances and applying a community detection algorithm to partition the sample set.The same method was used to cluster 9609 additional samples collected from eight countries by GenRe-Mekong. 18patial distribution of clusters was visualised using bespoke R code and map data from Natural Earth.
Population-level genetic diversity was assessed by the mean expected heterozygosity across all barcode SNPs, computed from allele frequency estimates after removing missing allele calls.

Statistical analysis
All statistical analyses were performed using R (version 4.2.0).For pairwise comparison between groups we used Wilcoxon rank-sum tests with a method to adjust the p value to control for false discovery rate, using the function pairwise.wilcox.test(x,p.adjust.method="fdr").For comparison between multiple populations, we used the kruskal.testfunction to perform Kruskal-Wallis tests.p<0•05 was considered statistically significant.Associations of kelch13 genotypes with the outbreak in Attapeu were tested by a logistic regression model.

Ancestry analysis
Pairwise identity by descent (IBD) analysis was performed on genotypes with 53 150 SNPs filtered by variability using hmmIBD. 22Visualisations of recombination patterns were produced by bespoke R scripts, which are available from the authors on request.

Role of the funding source
The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results
The  3 To support analyses of diversity and genetic distance, we analysed genetic barcodes comprising 101 SNPs selected for their variability and low mutual linkage. 18We excluded samples with high genotype missingness (>25%), resulting in a final set of 1787 samples, including 190 collected during the outbreak in Attapeu (figure 1; appendix 2 p 3).As a first step, we compared the diversity of genetic barcodes in outbreak parasites with that of parasites circulating in Attapeu in previous years, using the average SNP heterozygosity across all barcoding loci as a measure.In the 2017-19 period, diversity remained stable, at similar levels to those in 2011-12, with mean heterozygosity in the range of 0•39 (SD 0.11) to 0•42 (0•13).Starting in 2020, however, diversity collapsed, with heterozygosity reaching 0•12 (0•13) in 2021 (p<0•0001 when comparing 2020-21 vs 2012-19; figure 2A).Before the outbreak, diversity in other provinces was lower (range of 0•36 [0 •11] to 0•38 [0•14]) than in Attapeu (0•41 [SD 0 •11]; p=0•040) but significantly higher than in the most recent Attapeu samples (p<0•0001; appendix 2 p 7).We can conclude that the outbreak was characterised by a massive loss of diversity, unprecedented for Attapeu and for Laos in general.It is worth noting that Savannakhet Province had seasonal case number rises in three previous seasons (appendix 2 p 6), but these epidemics were not accompanied by a big reduction in diversity, suggesting that their drivers differed in nature from those of the Attapeu outbreak.
Expecting the reduction in diversity to be caused by expanding populations, we sought to investigate the population structure changes that accompanied the outbreak.To identify populations of highly related individuals, we estimated pairwise genetic distances and used these to cluster parasites with at least 95% barcode identity.This analysis identified 30 clusters with more than two members within the Attapeu sample set.In 2011-12, before the multidrug-resistant KEL1/PLA1 parasite migrated into the southernmost provinces of Laos, 18,23 2B).In 2017-18, considerably more population structure emerged, as parasites formed several clusters, and the proportion of parasites carrying unique barcodes dropped to 40% (34/86) in 2017 and 31% (59/192) in 2018.Several of these clusters might represent sub-strains of KEL1/PLA1, as previously described. 23Starting in 2019 and leading into the outbreak period, KEL1/PLA1 clusters disappeared and three clusters (labelled LAA1, LAA2, and LAA7) grew to dominate the population.Some members of these clusters were circulating at low frequency before 2019, but subsequently expanded so aggressively that they constituted over 94% (48/51) of samples in 2021.The most rapidly expanding cluster (LAA1) rose to 84% (43/51) prevalence in 2021, and was clearly responsible for most of the diversity reduction.The second largest cluster

A B
(LAA2) expanded more gradually, reaching a frequency of 8% (4/51) in 2021, and the third cluster (LAA7) appeared to shrink in size as LAA1 expanded.To summarise, P falciparum underwent a progressive increase in population structure in Attapeu, likely to have been driven initially by imported artemisinin-resistant strains.However, in the 2019-21 period, a small number of clonally expanding parasite clusters effectively replaced the previous populations, strongly suggesting that the outbreak resulted from a selective sweep.
To identify possible selection drivers, we characterised clusters according to the genotypes at key drug resistance markers (appendix 2 pp 3-4).As expected, in 2017-18, nine clusters comprised multidrug-resistant parasites likely to be variants of the KEL1/PLA1 strain, as they carried the kelch13 Cys580Tyr (C580Y) mutation (a marker of resistance to artemisinin) and the pm23 amplification (a marker of resistance to piperaquine). 23,24KEL1/PLA1 proliferated in several Greater Mekong Subregion countries where dihydroartemisinin-piperaquine was the front-line ACT, and is known to have penetrated Laos, although dihydroartemisinin-piperaquine was not used in Laos. 18During the outbreak, KEL1/PLA1 rapidly disappeared within a year and was replaced by the three outbreak clusters (appendix 2 p 8), all carrying alleles asso ciated with resistance to chloroquine, sulfadoxine, pyrimethamine, and artemisinin (appendix 2 pp 3-4).LAA2 parasites the C580Y mutation, although the absence of pm23 amplifications suggests that they were not closely related to KEL1/PLA1 parasites.The LAA1 and LAA7 clusters both comprised parasites carrying the kelch13 Arg539Thr (R539T) mutation, without pm23 amplification.Therefore, a rapid switch in prevalence of kelch13 mutations occurred in 2020, with R539T mutants prevailing over those carrying C580Y, causing the excess of cases (figure 3A).We found the R539T mutation to be very strongly associated with the outbreak onset (p<0•0001; appendix 2 p 5).Because LAA1 and LAA7 parasites were only found in Attapeu Province (appendix 2 pp 8-9), it is probable that these lineages have emerged locally.The R539T mutation was not detected in other provinces of Laos (figure 3B) but was present at low frequency in Attapeu before the outbreak.Most (48 [87%] of 55) of the LAA1 parasites lacked mdr1 amplification, a molecular marker associated with mefloquine resistance. 25All LAA2 and LAA7 carry a single copy of md1, predicted to be sensitive to the partner drug mefloquine.
To better understand the relationship between parasite clusters, we constructed a heatmap of genetic distances between pairs of individuals (figure 4).The most prominent feature on the heatmap is the extremely high degree of similarity among members of the LAA1 cluster (median distance=0•0 [IQR 0•016]), showing very little evidence of recombination between this essentially clonal group and other clusters.LAA1 also appear to diverge from other Attapeu parasites (median pairwise distance between LAA1 and non-LAA1 parasites=0•42 [IQR 0•030]; p<0•0001).Although clonality and differentiation are typical features of founder populations, LAA1 parasites circulated in small numbers for several seasons before their expansion, supporting the notion that the expansion is the result of a selective sweep.
The characteristics of LAA1 can be contrasted with those of KEL1/PLA1 clusters, which form a fragmented group of related clusters in the heatmap, consistent with differentiated sub-lineages.LAA7 is also located within this group, although it carries an R539T allele rather than C580Y, and does not possess the pm23 amplification (two key features of KEL1/PLA1).This finding suggests that LAA7 might have originated from KEL1/PLA1 parasites and, through recombination, acquired a different kelch13 variant and lost the pm23 amplification.The LAA2 cluster, in contrast, carries the C580Y allele but does not group with the KEL1/PLA1 clusters, suggesting that it derives from a separate C580Y lineage.Additional analyses using principal coordinates analysis cor ro borated these observations, showing that LAA1 sets itself apart from other clusters; that LAA2 is distinct from earlier KEL1/PLA1 clusters; and that LAA7 is more closely related to the C580Y parasites than to LAA1 (appendix 2 p 10).These findings raise questions about the origins of the outbreak clusters, particularly those carrying the kelch13 R539T mutation.Did they emerge because of independent mutation events, or are they descendants of previously identified R539T strains? 15,26,27Can recombination events explain the similarities between LAA7 and KEL1/PLA1?To answer these questions, we reconstructed the ancestry of the outbreak strains using IBD analyses, which identify genome stretches in a sample pair that are statistically likely to have inherited from a common ancestor.We used WGS data from Laos and its neighbouring countries, contributed to the MalariaGEN P falciparum Community Project v6.4,including a small number of LAA1, LAA2, and LAA7 samples collected by the GenRe-Mekong surveillance project (appendix 2 p 5).
We considered some of the oldest kelch13 mutants available, particularly the early founder populations circulating in western Cambodia in 2008 and 2009.Two of these populations, labelled KH2 and KH3 in previous work, 15 were found to harbour C580Y and R539T mutants respectively. 27When comparing LAA1 with one of the R539T mutants (KH3, collected in western Cambodia in 2008), we found 58•8% IBD, including a region containing the kelch13 gene (appendix 2 p 11; figure 5).This finding indicates that the artemisinin resistance variant carried by LAA1 did not emerge independently in Laos, but was inherited from strains circulating over a decade ago.The large IBD stretches observed indicate limited linkage breakdown, implying that very few recombination events occurred during that time.This hypothesis was confirmed by an analysis of R539T mutants circulating in Laos in 2017 (KH3-LA), which were nearly identical to Cambodian KH3 (IBD around 100%).Hence, this R539T lineage circulated for almost a decade essentially unchanged, spreading across national borders, before contributing to recombination events that produced LAA1, probably around 2017 when the earliest LAA1 members were sampled.Although no donor was identified for the remainder of the LAA1 genome, a kelch13 wild-type parasite from Champasak (LAA1-preWT) had 27•1% IBD with LAA1, mostly comple mentary to the KH3 contributions, suggesting that the recombination events that produced LAA1 involved local wild-type parasites.In contrast, a 2009 Cambodian KH2 C580Y mutant (KH2A) only shared 8•3% IBD with LAA1 ancestry, partly overlapping with the KH3 contribution (appendix 2 p 11).
The ancestry of LAA2 was simpler to reconstruct: its genome is essentially identical to that of the KH2A parasite (IBD >99%), suggesting that LAA2 expanded from a strain that has spread regionally, since 2009 or earlier, probably from western Cambodia, approximately 500 km from Attapeu.These parasites carry the C580Y mutation, but not the pm23 amplification which characterises KEL1/PLA1, as confirmed by the IBD levels (36•9%) when comparing LAA2 with an early Cambodian C580Y mutant carrying the amplification (KH2B, collected in 2010), suggesting that LAA2 derived from a lineage related to but distinct from KEL1/PLA1.IBD analyses showed that the LAA7 cluster probably originated from the recent recombination of two local strains, an R539T mutant (LAA7-preA, Attapeu, 2017), which contributed 49•0%, and a C580Y mutant Heterozygous means a mixed sample containing parasites carrying more than one kelch13 genotype.Heatmap cell colours reflect the pairwise distances between samples: shades of blue represent high pairwise similarity, the darkest blue corresponding to identical genetic barcodes; shades of red signify low barcode similarity, with deepest red representing a distance of 0•6 (ie, 60% of the barcode alleles differ between a pair of samples).pm23=plasmepsin2/3.(LAA7-preB, Attapeu, 2018), which contributed 45•1% (appendix 2 p 11).The two donor strains are themselves recombinants showing substantial IBD with early Cambodian mutants: LAA7-preA shares 23•1% of its genome with KH3, including the R539T kelch13 variant region, eventually inherited by LAA7, and LAA7-preB appears to be a KEL1/PLA1 strain closely related to earlier Cambodian strains (65•5% IBD with KH2B).The contribution of LAA7-preB has resulted in a high level of IBD between LAA7 and KEL1/PLA1, explaining the relationship observed in the genetic distance heatmap.

Discussion
In regions approaching malaria elimination, outbreaks threaten progress and can result in disease resurgence.
We have shown that genetic surveillance offers informative new perspectives on outbreaks, by studying three key aspects: genetic diversity at population level, markers of drug resistance, and ancestry of the outbreak strains.
The study of genetic diversity showed that the case number increase in the Laos outbreak was primarily caused by the clonal expansion of the LAA1 multidrug-resistant strain, carrying the kelch13 R539T mutation.This expansion occurred very rapidly, effectively replacing the dominant KEL1/PLA1 population in a single season.KEL1/PLA1 strains had become established in Champasak and Attapeu Provinces, probably imported from neighbouring countries where dihydroartemisinin-piperaquine was the ACT of choice.Although LAA1 parasites were present since 2018, they remained at low frequency while KEL1/PLA1 dominated.The KEL1/PLA1 population in Attapeu collapsed in 2019-20, coinciding with the end of dihydroartemisinin-piperaquine usage in neighbouring northeast Thailand and Viet Nam, soon after Cambodia ceased use of this treatment.These policy changes most probably reduced selective pressures on the parasite population, and it is also possible that movement restrictions in 2020, due to the COVID-19 pandemic, reduced the number of imported parasites.The demise of KEL1/PLA1 left a vacuum that was rapidly filled by the rise in frequency of LAA1.
The mainly clonal nature of the LAA1 expansion, with little evidence of interbreeding, is characteristic of populations under strong selective pressure.In outbreaks driven by increased transmission due to environmental The diagram shows the relationship between strains analysed for identity by descent (appendix 2 p 5).Each rectangle represents a sequenced sample: the three samples at the top represent older kelch13 mutants from Cambodia, the three samples at the bottom represent the Attapeu outbreak clusters; the remaining rectangles represent intermediate descendants and other parasite strains.Each rectangle is filled with coloured segments that show the proportion of contribution from the older Cambodian strains; white segments represent contributions from unknown sources.Next to each rectangle, we indicate: the sample label, the kelch13 mutation and the amplification status of pm23 or mdr1, and the place and year of sampling.Dashed arrows suggest putative inheritance paths, derived from observing identity by descent patterns, but might not denote an actual direct descendance of one sample from another (as opposed to both having common ancestry).Furthermore, numerous intermediate ancestry steps are not shown in this highly simplified depiction.mdr1=multidrug resistance 1. pm23=plasmepsin2/3.human factors, higher levels of recombination and greater diversity would be expected.The very low diversity seen in this outbreak was unusual when compared both with other provinces of Laos, and to previous seasons in Attapeu.It is improbable that the scarcity of parasite variety was a limiting factor, because we observed two additional clusters and some wild-type parasites circulating during the outbreak.Such low levels of recombination could occur if LAA1's selective advantages were conferred by a set of variants that would be disrupted through recombination with different genetic backgrounds, so that offspring from selfing (mating with genetically identical individuals) would retain the selective advantage and have a greater chance of survival than recombinants.This scenario could also explain how the LAA2 strain preserved its genome unchanged for at least 12 years since it was first sampled in Cambodia.This lineage appears to have retained its competitiveness through selfing for more than a decade and had a revival hundreds of km from its place of origin.This is not a unique case: essentially identical KH3 parasites were sampled in different countries, almost a decade apart, before their contribution to the LAA1 genome.Overall, IBD patterns indicate that successful recombinants of these multidrug-resistant strains are infrequent, as evidenced by the limited linkage breakdown in the LAA1 and LAA7 genomes.
All genetic evidence points to a selective sweep being the cause of the Attapeu outbreak.Additionally, we note that the KEL1/PLA1 population also waned in neighbouring Champasak, but only Attapeu had a surge in case numbers, which further suggests that the LAA1 population, local to this province, had a selective advantage.Unfortunately, our data cannot pinpoint the exact nature of this selective advantage.Selection can be driven by any factor that increases a parasite's chance to transmit-eg, increased gametocyte production or adaptation to vectors.However, given the recent history of malaria in the Greater Mekong Subregion, antimalarial drug resistance seems the most probable driver; hypothetically, LAA1 parasites might have reduced sensitivity to lumefantrine, the front-line ACT partner drug in Laos.This hypothesis cannot be confirmed by our data because there is no widely accepted marker of lumefantrine resistance; LAA1 parasites mostly do not have the mdr1 gene amplification, a proposed candidate marker, 25 which was a feature of their R539T mutant precursors.Still, it is possible that LAA1 inherited unknown drug resistant mutations from unknown contributors.We strongly recommend that artemetherlumefantrine efficacy be confirmed in Attapeu Province, where it would also be valuable to test the in vitro response to partner drugs.In response to the outbreak, the Centre of Malariology, Parasitology, and Entomology of Laos has conducted active case detection in the community at high risk and started a therapeutic efficacy study of artemetherlumefantrine in Attapeu, Savannakhet, and Sekong.
Our IBD analysis indicates that novel artemisininresistant lineages have inherited kelch13 haplotypes from early mutants circulating in western Cambodia.In other words, the kelch13 mutations circulating in the eastern Greater Mekong Subregion originate from a small number of mutation events, 27 in contrast with the many independently emerged mutations observed in Myanmar. 28,29Perhaps, the small number of founder popu lations described a decade ago 15 had already undergone a selection process that weeded out less fit mutants, and these populations and their derivatives have been very resilient over time.Even while KEL1/PLA1 parasites seemed to head for complete domination, 30 R539T parasites and C580Y parasites without pm23 amplification remained in circulation.This alternation in prevalence was probably triggered by human behaviours affecting selective pressures, such as changes in the choice of front-line treatment.Therefore, it remains crucial to monitor all kelch13 mutant varieties in this region, rather than focus on variants whose dominance might be temporary.Using WGS data, drug resistant strains can be systematically catalogued, and tracked by genetic markers.Here, we have shown that ancestry reconstruction methods can be powerful tools for this type of investigation.
As elimination approaches, public health interventions will apply increasingly stronger selective pressures on the parasite population, which might lead to new emerging forms of drug resistance, rapid sweeps, and unexpected outbreaks and resurgence.To maintain ACT efficacy, it will be crucial for national malaria control programmes to be empowered with an arsenal of tools with which to respond to these phenomena.Routine genetic surveillance, which is relatively straightforward and affordable to implement, 18 can address important use cases that go beyond the simple monitoring of drug resistance markers; here, we have shown it is a powerful tool for analysing the causes and dynamics of outbreaks.Although the development of a generalised methodology to address this use case will require comparative analyses of other outbreaks, which are outside the scope of this work, we believe that the present work provides the essential templates for future work.Using simple 101-SNP barcodes, we identified and characterised three different expanding strains and showed an abnormal drop in diversity, indicative of a selective sweep.Crucially, rapid processing enabled us to communicate findings to the National Malaria Control Programme before the beginning of the following season, so that they could plan their response.The addition of WGS data enabled a study of the origins of the outbreak populations, and their relationship to well known resistant strains and local parasites.There is clear value in analysing genomic epidemiology at different levels of resolution, using simpler techniques to answer the pressing questions rapidly, and studying the underlying phenomena and processes using more methods.We encourage the genomic epidemiology and surveillance communities to continue contributing to shared genomic datasets that support tackling these and other important public health use cases.

Contributors
VW, AD, KC, VS, MM, and OM conceived and designed the study.KC, VS, MM, SP, and SX coordinated sample and data collection.NT-N, SG, RDP, and OM processed samples, and generated and curated genomic data.VW, AD, and OM developed methods, and analysed and interpreted data.PNN, EAA, RJM, NPD, DPK, AMD, and OM were involved in the management and coordination of the genetic surveillance project.VW, AD, and OM wrote and revised the manuscript.VW and OM accessed and verified all the data.All authors provided critical revision of the manuscript.All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Declaration of interests
We declare no competing interests.

Data sharing
This publication uses data from the MalariaGEN P falciparum Community Project. 19Genetic Report Card data for the 2164 samples from Laos used in the present analyses are provided in appendix 3, which contains a data dictionary sheet (README) that details the fields included.

Figure 1 :
Figure 1: Sample selection The diagram shows how the analysed sample set was derived from Laos samples contributed by GenRe-Mekong (A) and by the earlier TRAC study (B), and the composition of this sample set by province and time period.

Figure 2 :Figure 3 :
Figure 2: Temporal changes in the population diversity in Attapeu province(A) Changes in population diversity over time.Mean single-nucleotide polymorphism heterozygosity across all genetic barcode loci is estimated as a measure of diversity for each year.The dashed line indicates a 5-year gap where data is not available.Error bars show standard deviation.(B) Increasing prevalence of clusters with highly related parasites (≥95% identity) and decreasing prevalence of parasites with unique barcodes (not belonging to a cluster).Bar segments represent the clusters identified each year, their size being proportional to the number of cluster members; parasites with unique barcodes are grouped in a grey segment at the bottom of the bar.Clusters that dominate the population during the 2020-21 outbreak (LAA1, LAA2, and LAA7) are shown in colour.The total number of samples in each year is shown.

Figure 4 :
Figure 4: Heatmap of relatedness between populations in Attapeu based on pairwise genetic barcode distancesA pairwise distance matrix was calculated from a set of 610 samples from Attapeu, using 96 quality-filtered barcode single-nucleotide polymorphisms.The arrangement of samples on the heatmap is determined by hierarchical clustering; the dendrogram on the left shows the clustering hierarchy.Colour bars at the top of the heatmap indicate for each sample: the year of sampling; the kelch13 genotype; the pm23 amplification status; and the cluster to which the sample belongs.Heterozygous means a mixed sample containing parasites carrying more than one kelch13 genotype.Heatmap cell colours reflect the pairwise distances between samples: shades of blue represent high pairwise similarity, the darkest blue corresponding to identical genetic barcodes; shades of red signify low barcode similarity, with deepest red representing a distance of 0•6 (ie, 60% of the barcode alleles differ between a pair of samples).pm23=plasmepsin2/3.

Figure 5 :
Figure 5: Reconstruction of ancestry of the Attapeu outbreak strainsThe diagram shows the relationship between strains analysed for identity by descent (appendix 2 p 5).Each rectangle represents a sequenced sample: the three samples at the top represent older kelch13 mutants from Cambodia, the three samples at the bottom represent the Attapeu outbreak clusters; the remaining rectangles represent intermediate descendants and other parasite strains.Each rectangle is filled with coloured segments that show the proportion of contribution from the older Cambodian strains; white segments represent contributions from unknown sources.Next to each rectangle, we indicate: the sample label, the kelch13 mutation and the amplification status of pm23 or mdr1, and the place and year of sampling.Dashed arrows suggest putative inheritance paths, derived from observing identity by descent patterns, but might not denote an actual direct descendance of one sample from another (as opposed to both having common ancestry).Furthermore, numerous intermediate ancestry steps are not shown in this highly simplified depiction.mdr1=multidrug resistance 1. pm23=plasmepsin2/3.
southern Laos province of Attapeu had a 130% increase in case numbers during the 2020-21 malaria season (378 cases in 2019 and 868 cases in 2020), in contrast with a low P falciparum incidence in other endemic provinces, including Savannakhet where the highest proportion of cases had been seen in previous years (2017-19).These changes in case numbers were reflected by the number of dried blood spot samples collected by GenRe-Mekong (appendix 2 p 6).To investigate this outbreak, we analysed genotypes from 2078 dried blood spot samples collected by the GenRe-Mekong project at public health facilities in southern Laos, which included 249 samples collected from cases in Attapeu Province during the outbreak, between April 1, 2020, and April 1, 2021, before COVID-19 pandemic lockdowns came into effect.These samples were compared to the remaining 1829 samples, collected by routine surveillance in five provinces (including 622 in Attapeu Province) between Jan 1, 2017, and March 31, 2020; and to 86 samples from Attapeu, collected by the TRAC study between Jan 1, 2011, and March 1, 2012.