A retrospective analysis of P. falciparum drug resistance markers detects an early (2016/17) high prevalence of the k13 C469Y mutation in asymptomatic infections in Northern Uganda

ABSTRACT The emergence of drug-resistant Plasmodium falciparum parasites in sub-Saharan Africa will substantially challenge malaria control. Here, we evaluated the frequency of common drug resistance markers among adolescents from Northern Uganda with asymptomatic infections. We used an established amplicon deep sequencing strategy to screen dried blood spot samples collected from 2016 to 2017 during a reported malaria epidemic within the districts of Kitgum and Pader in Northern Uganda. We screened single-nucleotide polymorphisms within: kelch13 (Pfk13), dihydropteroate synthase (Pfdhps), multidrug resistance-1 (Pfmdr1), dihydrofolate reductase (Pfdhfr), and apical membrane antigen (Pfama1) genes. Within the study population, the median age was 15 years (14.3–15.0, 95% CI), and 54.9% (78/142) were Plasmodium positive by 18S rRNA qPCR, which were subsequently targeted for sequencing analysis. We observed a high frequency of resistance markers particularly for sulfadoxine-pyrimethamine (SP), with no wild-type-only parasites observed for Pfdhfr (N51I, C59R, and S108N) and Pfdhps (A437G and K540E) mutations. Within Pfmdr1, mixed infections were common for NF/NY (98.5%). While for artemisinin resistance, in kelch13, there was a high frequency of C469Y (34%). Using the pattern for Pfama1, we found a high level of polygenomic infections with all individuals presenting with complexity of infection greater than 2 with a median of 6.9. The high frequency of the quintuple SP drug-resistant parasites and the C469Y artemisinin resistance-associated mutation in asymptomatic individuals suggests an earlier high prevalence than previously reported from symptomatic malaria surveillance studies (in 2016/2017). Our data demonstrate the urgency for routine genomic surveillance programs throughout Africa and the value of deep sequencing.

G lobally, Plasmodium falciparum is a leading cause of morbidity and mortality, particularly in sub-Saharan Africa (1,2).The burden of malaria has steadily declined over the past decades with extensive application of control measures (3).However, more recently, progress appears to have stalled (4), and the WHO reports an estimated 249 million cases with 608,000 malaria deaths in 2022 (2).A major concern is the emergence of artemisinin drug-resistant parasites (5)(6)(7).Malaria treatment, particularly with artemisinin-based combination therapies (ACTs), has been a cornerstone of malaria control (8,9).In most of sub-Saharan Africa, the first-line treatment for uncomplicated malaria is artemether-lumefantrine (AL) or artesunate/amodiaquine, while IV artesunate is used for severe malaria (10).Molecular genetic surveillance can play an important role in tracking the spread of drug-resistant parasites (11)(12)(13).This approach may guide the judicious application of available control interventions, potentially halting the spread of resistant parasites (12).
Previously, chloroquine and sulfadoxine-pyrimethamine (SP) resistance arose first in southeast Asia and spread later to Africa (14,15).Due to the use of SP in intermittent preventive treatment in pregnancy, it is still important to monitor the resistance-confer ring genes, P. falciparum dihydrofolate reductase (Pfdhfr) and dihydropteroate synthase (Pfdhps).Similarly, over the past 10 years, artemisinin-resistant parasites have emerged and spread in the Greater Mekong sub-region of southeast Asia and now appear to be emerging in sub-Saharan Africa (15)(16)(17)(18)(19)(20).Furthermore, in the P. falciparum multidrug resistance 1 (Pfmdr1) gene, linkage has been observed between N86 (wild type) and reduced susceptibility to lumefantrine, which is widely used as the artemisinin partner drug in East Africa (21).It is important to note that the increased risk for AL resistance selection is a result of the current high and historical use of AL.In Africa, AL accounts for 85% for all ACTs procured for Africa and indeed Uganda.
Mutations in the propeller domain of the kelch (K13) gene remain the primary molecular markers for artemisinin resistance (22).Although genetic surveillance studies across Africa have detected various mutations associated with resistance, only recently were parasites associated with delayed artemisinin parasite clearance identified in clinical studies from Rwanda (19) and Uganda (18).Important to this report is the study in North Uganda between 2015 and 2019, which assessed the frequencies of two mutations C469Y and A675V using symptomatic samples from a health facility in Gulu district, showing a rising (from about 4% to 20%) frequency of these mutations since 2018 and 2016, respectively (18).Therefore, evaluating the frequency of genetic resistance markers within different populations remains vital (11).
In malaria endemic regions, such as in sub-Saharan Africa, mixed infections are common (23).In a study of 15 countries across Africa, about 55% of the 2,263 whole-genome sequenced P. falciparum isolates were polygenomic with up to 9 clones in an infection.High malaria transmission drives high parasite mixing and a high complexity of infection (COI), thus mixed infections are likely key to the genesis and transmission of new mutations within a population.Notably, a high inbreeding coefficient (Fws) was observed in samples from Western Kenya, which is a region of holoendemic malaria transmission (23), a situation most likely maintained within asymptomatic infections that often go undetected by passive case detection surveillance studies in malaria endemic populations.
Here, we screened dried blood spot samples (DBS) collected from children and adolescents with asymptomatic infections from two districts, Kitgum and Pader, in Northern Uganda during a period of heightened malaria transmission (2016-2017).A study from Kitgum General Hospital showed a rise in malaria cases and hospitalization, the malaria test positivity rate increased from 10.5% to 54.6% between 2014 and 2016, and similarly, malaria accounted for over 40% of inpatient admissions (24).In malaria endemic areas, asymptomatic infections are common and are a major concern for elimination efforts (25).Patients with asymptomatic infections are typically older with some level of naturally acquired immunity (26).In the absence of symptoms such individuals do not seek care and, in turn, are silent reservoirs perpetuating malaria transmission (27).We aimed to estimate the presence of drug-resistant markers in asymptomatic P. falciparum infections using a targeted amplicon deep sequencing (AmpSeq) method.Furthermore, we assessed the level of genetic diversity and COI using the P. falciparum apical membrane antigen 1 (Pfama1) gene-a highly polymorphic merozoite surface antigen (28).
For the Pfmdr1 gene, NFT haplotype was observed at high frequency (44.1%), while the wild-type NYT haplotype was observed at 42%, as mixed infections in the popula tion.Their adjusted frequencies were 59% and 37%, respectively (Table S1).The newly described haplotype (NYS) was observed at 13.8% frequency (Table 2), with a much lower prevalence of 4% (Table S1).

A high frequency of artemisinin resistance markers
Pfkelch13 propeller domain genotyping was successful from 50 samples.Two Pfkelch13 mutations (A578S and C469Y) were detected at 4% and 34%, respectively.Surprisingly, we observed the presence of both the C469Y and rare E507G mutations in two samples.

Asymptomatic individuals present high levels of complexity of infection
The highly polymorphic Pfama1 gene was used to investigate the complexity of infection.Among the 78 samples positive by the 18S rRNA qPCR, 70 samples (90%) were successfully sequenced.A total of 74 unique haplotypes were identified, and all 70 individuals harbored polyclonal infections (COI >2).The mean COI was 6.97 (range 2-19) clones per individual (Fig. 1).Pfama1 showed more evenly distributed combina tions where minor alleles were frequently present without mixture.Some Pfama1 loci contained all three combinations per locus of wild type, mutant, and mixed, e.g., codons 200, 204, 225, 230, 242, 243, 267, 283, and 285 (Table 3).

DISCUSSION
The drug resistance markers were primarily mixed genotype infections with double, triple, and quadruple Pfdhfr mutants, demonstrating the extent of mutations within individual asymptomatic infections.Of note, quadruple Pfdhfr mutations, the inclusion of codon I164L, and Pfdhps A581G mutations were present but at much lower frequencies of <15%.They remain important loci to monitor and determine the extent of high levels of SP resistance.As expected, the triple Pfdhfr and double Pfdhps mutations continue to be maintained at high frequencies (29,30).Despite the high levels of genetic resistance markers to SP, this drug is still recommended for intermittent preventive treatment in pregnancy (IPTp) in malaria-endemic areas of Africa (31).Recent evidence demonstrates that the impact of the chemoprophylactic use of SP in pregnancy may not be through the prevention of malaria infection specifically, but it appears to be associated with improved birth outcomes, i.e., a reduction in the risk of low birth weight (32) .Studies conducted around the districts of Gulu, Lamwo, and Agago, which are all within 100 km of both Pader and Kitgum (the community sites for this study), described an increasing presence of the C469Y mutation in about 40 symptomatic cases at a The number of samples genotyped (dhfr = 58, dhps = 71, k13 = 50 while mdr1 = 67 samples).The frequency percentage of each haplotype is determined by the number of occurrences of a particular haplotype divided by the total number of haplotypes, denoted as "n".
prevalence of about 2% in 2017 (18) to >15% in 2019 (33) in Northern Uganda.A more recent report suggests stable circulation at approximately 20% prevalence of these mutations in 2022 among uncomplicated infections (34).These concerning findings highlight the high prevalence of the C469Y mutation beyond the routine detection of the easily identifiable malaria cases due to the presence of symptoms, while the reservoir of malaria transmission continues to go undetected, even though they make up a large percentage of malaria infections in malaria-endemic regions.In Pader and Kitgum, in 2016/17, asymptomatic RDT-positive malaria infections were at a prevalence of 55%.Furthermore, the independent emergence and local spread of clinically artemisi nin-resistant P. falciparum have recently been documented in Africa (2, 5).In Uganda, countrywide studies done from 2016 to 2019 suggested that these alleles were restricted to the northern half of the country (30).Recently, the C469Y mutation was observed at a frequency of 34% in Northern Uganda compared to 3% in Eastern Uganda.Furthermore, isolates with this mutation were significantly less susceptible to lumefantrine, even in mixed, C469, and 469Y infections (35).Perhaps, the presence of the C469Y and A675V [a mutation described by Balikagala and colleagues (18)] may have been a factor that reduced ACT drug efficacy in the districts of Kitgum and Pader in Northern Uganda during the period of heightened transmission (24,36,37).This artemisinin resistance mutation emphasizes the need for continuous, longitudinal genomic surveillance for the early detection and monitoring of drug resistance.This study further strengthens the requirement of active genomic surveillance endeavors in the community (12,38).This could allow for the early detection of important resistant mutations (18,39).The asymptomatic infections assessed in Pader and Kitgum were primarily polyclo nal infections, reflecting the high and perennial malaria transmission in this region of Northern Uganda.This is similar to other reports in high malaria burden areas that have high levels of polyclonal infections and genetic diversity (28,40).Much of the data on genomic markers of parasite resistance comes from research groups (41) and has been opportunistic, resulting in a substantial delay in the identification of increasing prevalence of drug resistance.Our deep sequencing approach showed a higher-thanexpected frequency of mixed infections for these alleles encoding drug resistance, in contrast to the pattern for AMA1, where minor alleles were often present in pure infections.This high frequency of mixed infections not only reflects the high COI in this population but provides an insight into how resistance mutations emerge, particularly in asymptomatic individuals who are not likely to be treated.Infections are initially, purely wild type, present as mixed infections and thereafter purely drug resistance when the prevalence of the mutation is high in the population due to a high selection pressure.The mixed infections would usually be missed by capillary sequencing approaches; the short-fragment deep sequencing approach increases the accuracy in calling haplotypes and mixed infections.Taken together, our data demonstrate the urgent need for the inclusion of asymptomatic infections in programmatic genomic surveillance to guide public health.Molecular surveillance is an important tool for tracking the prevalence of mutations that may affect the efficacy of anti-malarial drugs (12,13).Furthermore, multi-strain infections may impact malaria disease management and transmission (40).It has also been proposed that multiclonal malaria infections can influence clinical outcomes (40) and may have a negative impact on an individual's response to anti-malarial drug treatment (42).Sensitive detection methods, such as amplicon-based deep sequencing, could enable a better understanding of malaria strain diversity in relation to potential malaria vaccine antigen candidates and track changes in parasite genetic diversity (43).
This study had some limitations.The sample size following sequencing was rela tively small due to the low parasitemia in asymptomatic infections.However, the detection of these mutations gives an early indication of the presence of the drug resistance mutations in a previously and not routinely sampled population.Furthermore, the analysis of the microhaplotype frequencies by two approaches demonstrates the diversity of the infections based on the number of observations per infection and the frequencies of the microhaplotypes in the population based on a weighted within-host frequency.This study does provide a valuable baseline for future studies within the districts of Kitgum and Pader in Northern Uganda, a region where clinical evidence of prolonged parasite clearance (18) and an increase in the C469Y mutation frequency have been described.Second, this study did not follow up with any individuals and does not demonstrate clinical evidence of resistance.Finally, we did not investigate the prevalence of the A675V mutation associated with artemisinin resistance, which indeed may be circulating at a higher frequency.Studies in districts surrounding the study area estimate prevalence at approximately 41% (33).The greater resolution provided by AmpSeq distinguishes loci under positive selection in the population, such as those of drug resistance markers, that were classified as either mutant and mixed or wild-type and mixed genotypes.In great contrast, for genes under balancing selection, such as Pfama1, loci present with a combination of all three, wild type, mutant, and mixed in the population.

Conclusions
This study shows, for the first time, candidate artemisinin resistance markers at a high frequency in asymptomatic infections, at a time prior to its detection in symptomatic infections, adding to the growing body of evidence of this malaria transmission reservoir harboring drug-resistant parasites that may facilitate their spread.The quintuple Pfdhfr and Pfdhps mutations remain at a high frequency, and Pfmdr1 is 100% N86 in this population.Therefore, molecular surveillance is vital and should be sustained in view of reported clinical evidence of prolonged clearance of parasites harboring the C469Y mutation.Importantly, asymptomatic individuals should be considered during control efforts to reduce the prevalence and spread of drug resistance, particularly in high malaria transmission areas.

Study design and setting
This study was nested within a larger case control study investigating the pathogene sis of nodding syndrome-a poorly understood neurological disorder thought to be caused by an infection with the filarial worm Onchocerca volvulus (44).In the parent study, healthy individuals, 8 years or older, from the nodding syndrome-affected districts of Kitgum and Pader in Northern Uganda (Fig. 2), were recruited as healthy controls between 2016 and 2017 (44).In this area, inhabited by the Acholi people, malaria is a major cause of morbidity and mortality (24,45).During the study period, the region was experiencing an upsurge in malaria cases, and a study from Kitgum general hospital found that malaria was the leading cause of hospitalization, accounting for 40% of inpatient admissions (24).
Between September 2016 and September 2017, 154 children and adolescents were enrolled as age-matched healthy controls into the parent study from within the sur rounding communities.All enrolled study participants underwent a clinical examination with a detailed clinical history reported.A blood sample was collected from each participant, and dried blood spot prepared on Whatman filter paper.All were screened for malaria using P. falciparum HRP2 rapid diagnostic tests (CareStart, 2016).For this analysis, 142 samples collected from healthy community controls with no clinical symptoms were available.

Laboratory procedures for DNA extraction, amplification, and sequencing
To extract parasite DNA from the RDT-positive DBS, a previously published protocol (12) was used.Briefly, two punches each 2.5 mm were punched from two locations (at the center and periphery) of the DBS and placed into a 1.5-mL Eppendorf using sterile tweezers.DNA extraction was done using the Chelex saponin method (46).Parasite DNA was amplified using 18S rRNA Plasmodium falciparum qPCR assay (47).Samples above the median cycling threshold of 35 were considered as low parasite mia, and they were excluded from sequencing.Using a previously published nested PCR approach (12), amplicons were generated in duplicates for the following genes, using molecular identifiers (MID) labeled primers (12) for P. falciparum apical membrane antigen 1 gene (Pfama1: PF3D7_1133400), P. falciparum dihydrofolate reductase (Pfdhfr: PF3D7_0417200), P. falciparum dihydropteroate synthase (Pfdhps: PF3D7_0810800), P. falciparum kelch13 (Pfk13: PF3D7_1347700), and P. falciparum multidrug resistance 1 (Pfmdr1: PF3D7_0523000).The ama1 primers spanned nucleotides 441-946, which overlap with domain I, a region with the highest nucleotide diversity in ama1 (28).The PCR products were individually purified using the AMPure XP beads (Beckman Coulter, Inc.) as per the manufacturer's instructions.Thereafter, the quantification of the purified DNA was done using a Qubit double-strand DNA (dsDNA) High-Sensitivity (HS) Assay Kit, according to the manufacturer's instructions.Library preparation was done using the KAPA Kit, while a size selection clean-up was done using 0.8× AMPure XP beads.The adapter-ligated libraries were then amplified using Illumina primers and cleaned with 0.8× AMPure.The libraries were quantified using a Qubit dsDNA HS Kit, and sizes were verified by the DNA 1000 Assay Kit using the 2100 Bioanalyzer (Agilent).The libraries were mixed in equimolar concentrations, denatured, spiked with 8% PhiX DNA, and finally sequenced using a MiSeq Reagent Kit v3 (Illumina).

Sequence data analysis
SeekDeep v3.0.1 (48) was utilized for data analysis and involved three main algo rithms for analyzing amplicon sequence data.The first algorithm, Extractor, performed demultiplexing of sequence data using the MIDs and primers utilized for amplification.Additionally, Extractor performed read filtering based on target-specific read lengths and trimmed bases with quality scores below 20.This ensured that only high-quality bases were included in subsequent analyses.
The second algorithm, Qluster, refined the microhaplotype identification process by collapsing amplicon reads by integrating sequencer-generated quality values and kmer frequencies.Amplicon sizes in the panel ranged from 500 to 520 base pairs, and microhaplotypes were constructed based on reads spanning these read lengths.Overlapping reads were stitched to ensure the continuity of the sequence data, contributing to accurate microhaplotype reconstruction.This approach effectively differentiated true biological variations from sequencing errors and PCR artifacts, thereby improving the reliability of microhaplotype inference.The clustering process involved an initial stage where reads with no differences were collapsed into identical sequence clusters indexed by their k-mers for efficient processing.This was followed by an iterative clustering stage, where clusters were compared iteratively with progressively stringent merging criteria, allowing for an adaptive response to the sequence diversity in the samples.Each mismatch was evaluated based on mismatch quality and the quality of adjacent bases, enhancing the accuracy of error distinction from true variants.To identify and mitigate PCR chimeras, sequences were examined for unusual patterns or breakpoints that typically indicate chimera formation.These chimeric sequences were flagged and excluded from the final microhaplotype assembly.
The third algorithm, process Clusters, further refined the microhaplotype identification by re-evaluating clusters for residual errors.This step involved re-processing clustered reads to remove any remaining low-frequency errors or artifacts, ensuring the final microhaplotype calls were highly accurate.Microhaplotypes were discarded if they did not occur in duplicate samples and if their combined relative frequency was <1%.A conservative cutoff of 1% was set based on the lowest lab isolate mixture (0.5:0.49:0.01;i.e., 50% 3D7, 49% HB3, 1% 7G8), unless the microhaplotype was independently detected in other samples at >1%.The SNP frequencies in the population were calculated as the number of samples that contained the SNP over the total number of samples genotyped to determine the total number of samples with wild-type mutant or mixed genotypes.For the microhaplotype frequencies, two approaches were used.The first is based on regarding each instance of a microhaplotype as an observation, determined by dividing the count of each microhaplotype's occurrences by the total number of microhaplotype occurrences.The second approach used the within-host frequency of the haplotype as a weighting to adjust the haplotype frequency reported.All statistical analysis was performed in R v4.0.3 (49).

FIG 1
FIG 1 Distribution of individuals with genetically distinct parasite infections using Pf-AMA1 genotyping.The y-axis is the frequency of individuals, and on the x-axis is the complexity of infections.

FIG 2
FIG 2 Map of Uganda showing the area where the samples were collected.The area in brick-red shows the districts of Kitgum and Pader in Northern Uganda, while the area in gray represents the surrounding district of Lamwo, Agago, and Gulu.The areas in blue shows water bodies.The maps were generated using QGIS.

TABLE 2
Frequency of resistance haplotypes for each gene and of infections containing mixed haplotypes b a Wild type, bold text indicates the mutant codon.The k13 482 H and 507G mutants were observed in two samples.b

TABLE 3
Frequency of mutations from the Pfama1 gene in the population a a Total number of samples sequenced = 78.