A Novel Plasmodium falciparum Kelch13 A675T Mutation and High Levels of Chloroquine and Sulfadoxine-Pyrimethamine Resistance in Burundi

Abstract

Antimalarial resistance, including failure of artemisinin combination therapy, is increasing in Africa. Molecular surveillance of markers of drug resistance is essential to monitoring drug resistance. Plasmodium falciparum isolates from Cibitoke Province, Burundi were sequenced for ten markers of resistance (including potential background markers conferring kelch-13mediated artemisinin partial resistance) by Illumina or Nanopore sequencing. Plasmepsin-2 and mdr1 copy number variations were typed by digital PCR. Among 157 isolates, no validated pfkelch13 mutations linked to artemisinin partial resistance (ART-R) were detected. However, a novel pfkelch13 mutation, A675T, was identified. A mutation in the same position (A675V) had been found and validated in neighboring Rwanda previously. Background mutations were frequent, including pffd D193Y (45.9%) and pfarps10 V127M (11.5%). Sulfadoxine-pyrimethamine (SP) resistance was widespread, with quintuple haplotypes detected in 82.3% and sextuple haplotypes in 12.3% of mono-infections. The pfcrt IET triple mutant haplotype (94.9%) and pfmdr1 N-F-D mutant haplotype (57.9%) indicate sustained CQ resistance and efficacy of artemether-lumefantrine (AL). No pfpm2 duplications but rare pfmdr1 amplifications (1.8%) were identified. These findings indicate substantial SP resistance, posing a threat to the efficacy of IPTp and SMC strategies in Burundi. The findings also show potential ART-R and partner drug resistance in Burundi, highlighting the need for strengthened surveillance and coordinated regional efforts to mitigate resistance and sustain malaria control.

INTRODUCTION

Malaria remains a global health burden, with 263 million cases and 597,000 deaths reported in 2023[1]. Sub-Saharan Africa bears the highest burden, accounting for 94% of all malaria cases and 95% of deaths, with children under five disproportionately affected [2]. Despite the significant progress in reducing malaria incidence over the past two decades, challenges such as limited healthcare access, drug resistance, and vector control failures threaten to reverse these gains [3–6]. Among these challenges, the emergence of drug-resistant Plasmodium falciparum represents one of the most significant obstacles to global malaria control[7]. Burundi exemplifies these challenges, with nearly 80% of the population at risk and malaria cases rising from 6.8 million in 2021 to over 8.2 million in 2022[8].

Antimalarial drugs remain central to malaria prevention and treatment. Artemisinin-based combination therapies (ACTs) are the first-line treatment of uncomplicated malaria, combining the rapid parasite clearance of artemisinin derivatives with the sustained efficacy of a partner drug[9,10]. Sulfadoxine-pyrimethamine (SP) is widely used for intermittent preventive treatment during pregnancy (IPTp) and seasonal malaria chemoprevention (SMC) in children[11]. Resistance to artemisinin, partner drugs, and SP threatens the effectiveness of these interventions[12,13]. Mutations in pfkelch13 are key mediators for artemisinin partial resistance (ART-R). Yet, not all mutations in pfkelch13 contribute to resistance, and resistance has been observed in the absence of pfkelch13 mutations[14]. The impact of mutations on resistance can be validated by following parasite densities in patients after treatment to assess clearance rates [15], through in vitro ring-stage survival assays (RSA) [14] or generating isogenic parasite lines harboring candidate mutations via genome editing (e.g., CRISPR/Cas9) and subsequently evaluating their drug susceptibility using the RSA [16,17]. pfkelch13 mutations such as N458Y, M476I, Y493H, R539T, I543T, P574L, and C580Y are well-documented drivers of ART-R in Southeast Asia [18]. In Africa, A469Y, P553L, R561H, R622I, A675V, and others have increasingly been reported [19–26]. Recently, the pfkelch13 R561H mutation, validated as a driver of ART-R through its association with delayed parasite clearance in laboratory and clinical studies, has been observed in Rwanda[19,27], Tanzania[21,22] and Eritrea[28]. Meanwhile, C469Y and A675V, linked to prolonged parasite clearance times, have been detected in Uganda[23].

Mutations in background genes, such as pfarps10, pffd, pfcrt, pfap2mu, pfatg18, pfpx1, and pfmdr2, are thought to facilitate the emergence of ART-R, particularly in Southeast Asia [29,30]. While some of these mutations have been reported in Africa[19,31–34], their role in facilitating pfkelch13-associated resistance in this region remains uncertain. Differences in transmission dynamics, genetic background of the parasite population, and drug selection pressures may influence how resistance evolves in different parts of the world. Asides pfkelch13, ART-R has been linked to mutations in other genes, such as the sarcoplasmic-endoplasmic reticulum Ca2 -ATPase-type protein (pfSERCA) and the actin-binding protein pfcoronin[35,36].

Resistance to ACT partner drugs further complicates malaria treatment. Mutations in the pfcrt gene are associated with resistance to chloroquine and piperaquine, while pfmdr1 polymorphisms have been suggested to play a role in lumefantrine and amodiaquine susceptibility[37–40]. Amplifications in plasmepsin 2/3 genes are strongly associated with piperaquine resistance, particularly in Southeast Asia [41]. In East Africa, SP resistance is driven by quintuple and sextuple pfdhfr/pfdhps haplotypes, which are approaching fixation in East African countries such as Tanzania, Kenya, and Uganda[42–46].

Burundi’s malaria control strategy includes SP for IPTp and ACTs (AL) for uncomplicated malaria [8,47]. Its proximity to regions with well-documented ART-R, such as Rwanda[19], Uganda[23], the Democratic Republic of Congo (DRC)[48] and Tanzania[21], increases the risk of resistant strains spreading into the country. No data is available on the prevalence and distribution of pfkelch13 mutations and other resistance-associated markers in Burundi.

This study employs multiplexed PCR targeted amplicon sequencing and digital PCR (dPCR), to investigate the prevalence of SNPs in resistance markers and conduct copy number variation (CNV) analysis in P. falciparum isolates from Burundi. Our findings revealed the presence of novel pfkelch13 mutations, and high SP and chloroquine resistance. This research provides baseline drug resistance data, offering critical insights into the prevalence of resistance-associated mutations in P. falciparum in Burundi.

MATERIALS AND METHODS

Ethical approval

This study was conducted in compliance with ethical standards and received approval from the Comité National d’Ethique pour la protection des êtres humains sujets de la recherche biomédicale et comportementale of Burundi (approval reference: CNE/03/2021) and the Institutional Review Board of the University of Notre Dame (approval reference: 21-02-6446). All adult participants gave informed written consent prior to sample collection, and minors’ legal guardians provided consent in writing.

Study Area, sample collection, and molecular screening

Burundi experiences high, year-round transmission of P. falciparum malaria, with moderate seasonal fluctuations. Samples for this study were collected during the dry season in June 2021 from Cibitoke Province, located in northwestern Burundi near the borders with Rwanda and the DRC. Approximately 200 μL of blood was obtained via finger prick into EDTA tubes and stored at −20°C until molecular analyses were conducted. The sample collection was part of a previous evaluation of rapid diagnostic tests (RDT)[49]. Molecular screening for P. falciparum was conducted using the varATS qPCR assay, as described previously [49].

Multiplexed PCR for marker amplification

Multiplexed PCR was employed to amplify gene targets associated with drug resistance in P. falciparum. The target regions included codons in pfkelch13 (e.g., R561H, C580Y, A675V), pfdhfr (e.g., C59R, S108N, I164L), pfdhps (e.g., A437G, K540E, A581G), pfcrt (e.g., K76T, T93S, H97Y), and pfmdr1 (e.g., Y184F, N1042D, F1226Y). Less-studied markers such as pfarps10 (V127M, V128Y), pfatg18 (T38I), pfmdr2 (T484I, I492V), pfpx1 (C1484F), and pffd (D193Y) were also included. A comprehensive list of target codons and primer sequences is provided in Supplementary Table S1.

To amplify these markers, three separate PCR reactions were performed using distinct primer mixes: PM 1, PM 2, and PM 3 (Supplementary Table S2). Each reaction was prepared in a total volume of 25 μL, consisting of 14.25 μL of PCR-grade water, 5 μL of 5X KAPA HiFi Buffer (1X final concentration), 0.75 μL of 10 mM KAPA dNTP Mix (0.3 mM final concentration), 0.5 μL of a primer mix (PM 1, PM 2, or PM 3), 0.5 μL of HiFi HotStart DNA Polymerase (1 U/μL, final concentration 0.5 U), and 4 μL of DNA template.

PCR amplifications were conducted on a BioRad T100^™ Thermal Cycler. The thermal cycling program consists of an initial denaturation at 95°C for 3 minutes, followed by 35 cycles of denaturation at 98°C for 15 seconds, annealing at primer-specific temperatures (58°C for PM 1, 62°C for PM 2, and 51°C for PM 3) for 15 seconds, and extension at 72°C for 30 seconds. A final extension was performed at 72°C for 2 minutes.

Library Preparation and Sequencing

Amplicons from the three PCR reactions were pooled in equimolar concentrations to create a master amplicon plate for downstream sequencing library preparation. Library preparation was conducted using the Oxford Nanopore Technologies (ONT) SQK-NBD114.24 and 96 kit, following the manufacturer’s protocol (version NBA_9170_v114_revL_15Sep2022) with modifications as described earlier [50]. Briefly, for the end-prep step, incubation times were extended to 15 minutes each at 20°C and 65°C. End-prepped DNA was purified using a 1.8X bead-to-sample ratio of AMPure XP beads (Beckman Coulter) and resuspended in 15 μL of nuclease-free water. 3.75 μL of end-prepped DNA was used instead of the standard 0.75 μL for the native barcoding ligation, and the incubation time was extended to 30 minutes from 20 minutes. Barcoded samples were purified using a 1.4X bead-to-sample ratio of AMPure XP beads, replacing the standard 0.4X ratio. During the adapter ligation and clean-up step, the incubation time was increased to 30 minutes, and pooled barcoded samples were purified using 0.8X AMPure XP beads. Short fragment buffer was used for all wash steps during purification.

The final pooled library was quantified using a Qubit fluorometer (ThermoFisher) and diluted in ONT elution buffer to approximately 130 fmol before being loaded onto R10.4.1 flow cells. Sequencing was conducted on a MinION Mk1C instrument (ONT) using MinKNOW software (distribution version 23.07.12, core version 5.7.5, and configuration version 5.7.11).

For Illumina sequencing, a nested PCR was performed on primary amplicons from PM1, PM2, and PM3 to append overhang linker sequences. The forward and reverse primers used in the nested PCR contained 5′ overhangs (5′-GTGACCTATGAACTCAGGA-3′ and 5′-CTGAGACTTGCACATCGCAGC-3′, respectively) to facilitate subsequent adapter addition[51]. Nested PCR products were purified using a 1.8X bead-to-sample ratio of AMPure XP beads and pooled in equimolar concentrations. Unique dual-index barcodes were added via adapter PCR for sample multiplexing[51]. Barcoded products were then purified using a 1.4X bead-to-sample ratio, normalized, and pooled into a single sequencing library. The final library was sequenced on the Illumina NextSeq2000 platform using a P1 flow cell and XLEAP-SBS 600-cycle kit (2 × 300 bp reads), with a 50% PhiX v3 spike-in to enhance base diversity and optimize cluster generation for low-diversity amplicon libraries.

Bioinformatics Analysis of Sequencing Data

Nanopore sequencing data (*.pod5 files) were basecalled using Dorado (v0.8.3; https://github.com/nanoporetech/dorado) with the super accuracy model dna_r10.4.1_e8.2_400bps_sup@v5.0.0. Previously established thresholds and flags were used[50,52]. Briefly, the --no-trim flag was applied to retain full-length reads. To ensure high data quality, a minimum Q-score threshold of 20 (--min-qscore 20), corresponding to an accuracy of ≥99%, was applied to filter reads for downstream analysis. Quality-filtered reads were demultiplexed using Dorado with double-ended demultiplexing enabled (--barcode-both-ends) to minimize false-positive assignments. A summary of the basecalling results, including reads below the Q-score threshold, was generated using the Dorado summary command, and quality metrics were further assessed with NanoStat.

Nanopore sequencing data were processed as described earlier [50,52]. In short, haplotypes were inferred from the resulting FASTQ files using the R packages HaplotypR (v0.5.0; https://github.com/lerch-a/HaplotypR) and DADA2 (v1.26.0) [53,54]. Reads were demultiplexed by marker using HaplotypR’s demultiplexByMarkerMinION() function, and sequences with ambiguous bases or incorrect lengths were excluded. Haplotypes were inferred using DADA2’s learnError() and dada() functions, with sequence abundance tables generated using makeSequenceTable(). To ensure the integrity of the data, chimeric haplotypes were identified and removed using removeBimeraDenovo(). Additional filtering excluded sequences with singletons or ambiguous bases. For each sample, markers were included if they had a minimum coverage of 10 reads. Haplotype inference required at least 50 supporting reads per haplotype and a within-host frequency of ≥1%

Illumina paired-end data were processed as described earlier [51,53]. Forward (R1) and reverse (R2) reads were merged using HaplotypR’s bindAmpliconReads() function, with R1 trimmed to 230 bp and R2 to 210 bp to reduce noise and standardize the dataset. Low-quality reads and those containing ambiguous bases were filtered using the FilterFastq() function. Haplotypes were inferred using the DADA2 pipeline, with the same error modeling, chimera removal, and abundance table generation steps as described for Nanopore data. All haplotype calls adhered to the same thresholds for within-host frequency and minimum coverage as for the Nanopore data.

Haplotypes, defined as specific combinations of alleles across multiple SNP positions, were analyzed via visual inspection and alignment against annotated P. falciparum reference genomes using Geneious Prime (v2025.0.3). Reference sequences included pfkelch13 (PF3D7_1343700), pfdhfr (PF3D7_0417200), pfmdr1 (PF3D7_0523000), pfcrt (PF3D7_0709000), pfpx1 (PF3D7_0720700.1), pfdhps (PF3D7_0810800), pfatg18 (PF3D7_1012900), pffd (PF3D7_1318100), pfmdr2 (PF3D7_1447900), and pfarps10 (PF3D7_1460900). Haplotypes that failed to align to their respective reference genome were excluded. In some targets (pfarps10, pfatg18, pffd, and pfcrt codons 343–356), insertions or deletions (indels) were observed adjacent to the target codons, due to unavoidable flanking during primer design. While these indels did not affect SNP-based classification, they contributed to haplotype diversity (Supplementary Figures S1–S4).

Pfkelch13 Mutation Validation

To validate detected pfkelch13 mutations, multiplex PCR was used to amplify three overlapping regions spanning codons 400–699. Each region was amplified in duplicate per sample and sequenced on a MinION Mk1C platform. A second PCR using primers spanning codons 400–699 was also performed in duplicate and sequenced (Supplementary Table S3). Variants were classified as confirmed only if the same SNPs were consistently detected across all runs and replicates.

Digital PCR Analysis of Gene Copy Number Variations in pfmdr1 and pfpm2

Digital PCR (QIAcuity One, Qiagen, Germany) assays were used to quantify pfpm2 and pfmdr1 alongside the β-tubulin housekeeping gene in 164 clinical samples. Primer and probe sequences are listed in Supplementary Table S4. Each reaction was prepared in a total volume of 12 μL, comprising 3 μL of QIA-PCR Master Mix (1X final concentration), 1.08 μL each of 10 μM forward and reverse primers, 0.3 μL of 10 μM probes, 1 μL of restriction enzyme, and 2 μL of DNA template. For the pfpm2 assay, HinfI (10 U) was used, while AluI (10,000 U/mL) was included for the pfmdr1 assay. The final reaction volume was adjusted with 1.08 μL of nuclease-free water. Cycling conditions for both assays included an initial denaturation at 95°C for 2 minutes, followed by 49 cycles of 95°C for 15 seconds and 60°C for 1 minute. Density for each target gene was estimated based on fluorescence signals generated during the duplex dPCR assays. Copy number variation (CNV) was calculated as the ratio of the parasite density of the target gene to that of the β-tubulin housekeeping gene. A CNV ratio of ≥1.5 was considered indicative of gene duplication.

RESULTS

Reads Distribution and Sequencing performance

Supplementary Table S5 summarizes the sequencing results, providing details on read counts, identified haplotypes, and observed mutations across all screened gene targets. Sequencing read numbers for each marker are shown in Fig. 1A, with all markers exhibiting sufficient read coverage across samples. Most samples exceeded the minimum threshold of 10 reads per marker, with 86–100% of targets achieving at least 50 reads (Fig. 1B). Read counts were generally robust across samples, including those with lower parasite densities. While several markers showed minimal variation in read depth across density groups, moderate positive correlations between parasite density and read counts were observed for a subset of targets (e.g., PfK13, Pfcrt, Pfmdr1), suggesting density-responsive amplification. These relationships were visualized and quantified in Supplementary Figure S5.

Figure 1. Read coverage and distribution across genetic markers.

A) Illustrates the Log10-transformed read distributions for each marker. pfcrt [Codon 74–97], pfcrt [Codons 343–356], Pfatg18 [Codon 38] and pfpx1 [Codon 1484] contain reads obtained from MinION and Illumina sequencing, while the rest contained reads obtained from MinION sequencing only. B) Presents the percentage distribution of samples with reads ≥50 (Blue) and <50 (Orange) for each marker. Across all markers, the vast majority of samples achieved this threshold, with performance ranging from 86% to 100%.

Polymorphisms in pfkelch13 and Artemisinin-Predisposing Background

All 157 P. falciparum samples were successfully genotyped for pfkelch13 mutations (Table 1), revealing a novel A675T and two synonymous mutations (S477S and T684T) (Supplementary Figure S6 and S7). Sequencing of samples with mutations was repeated and these SNPs were confirmed. While the A675T (GCC>ACC) mutation has not been observed before, a substitution at the same codon, A675V (GCC>GTC), is a WHO-validated pfkelch13 mutation associated with ART-R. A675V has been documented in Kenya [23,55–57].

Table 1.

Frequency of K13 propeller variants screened in this study

Variant types	Base substitution	Variant Frequency	Wild type	Variant type
Validated
F446I	TTT>ATT	0	157/157	NS
N458Y	AAT>TAT	0	157/157	NS
M476I	ATG>ATA	0	157/157	NS
Y493H	TAC>CAC	0	157/157	NS
R539T	AGA>ACA	0	157/157	NS
I543T	ATT>ACT	0	157/157	NS
P553L	CCG>CTG	0	157/157	NS
R561H	CGT>CAT	0	157/157	NS
P574L	CCT>CTT	0	157/157	NS
C580Y	TGT>TAT	0	157/157	NS
A675V	GCC>GTC	0	157/157	NS
Candidate
P441L	CCA> CTA	0	157/157	NS
G449A/D	GGT>GC/AT	0	157/157	NS
C469Y/F	TGC>TA/TC	0	157/157	NS
A481V	GCT>GTT	0	157/157	NS
R515K	AGA>AAA	0	157/157	NS
P527H	CCT>CAT	0	157/157	NS
N537I	AAT>ATT	0	157/157	NS
G538V	GGT>GTT	0	157/157	NS
V568G	GTG>GGG	0	157/157	NS
Detected in study
S477S	TCT>TCG	2/157	155/157	S
A675T*	GCC>ACC	1/157	156/157	NS
T684T	ACA>ACT	2/157	155/157	S

^*Novel mutation reported in this study, NS-non-Synonymous, S-Synonymous

Background mutations previously implicated in contributing to ART-R were screened. No mutations were detected in pfpx1 (C1484; n = 157) and pfmdr2 (T484I; n = 153). For pfmdr2 codon I492V, 41.2% (63/153) of isolates were wild-type, 35.3% (54/153) carried the mutant allele, and 23.4% (36/153) samples displayed mixed alleles. In pffd (D193Y), 54.1% (85/157) were wild-type, 7.0% (11/157) were mutants, and 38.9% (61/157) samples showed mixed alleles. Most isolates were wild-type for pfarps10 codons V127M and D128H (88.5%, 139/157), with 3.2% (5/157) mutant and 8.3% (13/157) had mixed alleles. Similarly, 88.5% (139/157) of isolates were wild-type at pfatg18 (T38I), with 3.2% (5/157) mutant and 8.3% (13/157) mixed.

Polymorphism in pfcrt and pfmdr1

Among the 157 P. falciparum samples genotyped for the pfcrt gene, 94.9% (149/157) carried the IET triple mutant haplotype (M74I, N75E, K76T), comprising 89 pure mutants and 60 mixed alleles. All samples were wild-type at codons C72S, T93S, M343L, C350R, and G353V. 10.2% (16/157) harbored the 356T mutant allele (Table 2).

Table 2.

Mutations detected in pfcrt, pfmdr1, pfdhps, and pfdhfr markers

Marker		N (%)	Status	n	%
pfcrt
	C72S	157 (100)	Wild-type	157	100
	M74I/ N75E/K76T	157 (100)	Wild-type	8	5.1
			Mutant	89	56.7
			Mixed	60	38.2
	T93S	157 (100)	Wild-type	157	100
	H97Y/L	157 (100)	Wild-type	156	99.4
			Mixed	1	0.6
	M343L/C350R/G353V	157 (100)	Wild-type	157	100
	I356T	157 (100)	Wild-type	141	89.8
			Mutant	15	9.6
			Mixed	1	0.6
pfmdr1
	N86Y	157 (100)	Wild-type	133	84.7
			Mutant	4	2.5
			Mixed	20	12.7
	Y184F	157 (100)	Wild-type	65	41.4
			Mutant	45	26.7
			Mixed	47	29.9
	S1034C	157 (100)	Wild-type	154	98.1
			Mixed	3	1.9
	N1042D	157 (100)	Wild-type	147	93.6
			Mixed	10	6.4
	F1226Y	157 (100)	Wild-type	157	100
	D1246Y	157 (100)	Wild-type	152	96.8
			Mutant	1	0.6
			Mixed	4	2.5
pfdhfr
	C50R	157 (100)	Wild-type	157	100
	N51I	157 (100)	Mutant	153	97.5
			Mixed	4	2.5
	C59R	157 (100)	Wild-type	2	1.3
			Mutant	120	76.4
			Mixed	35	22.3
	S108N	157 (100)	Mutant	153	97.5
			Mixed	4	2.5
	I164L	157 (100)	Wild-type	155	98.7
			Mixed	2	1.3
pfdhps
	S436A	157 (100)	Wild-type	154	98.1
			Mutant	1	0.6
			Mixed	2	1.3
	A437G	157 (100)	Wild-type	10	6.4
			Mutant	128	81.5
			Mixed	19	12.1
	K540E	157 (100)	Wild-type	2	1.3
			Mutant	115	73.2
			Mixed	40	25.5
	A581G	157 (100)	Wild-type	75	47.8
			Mutant	8	5.1
			Mixed	74	47.1
	A613T	157 (100)	Wild-type	157	100

In the pfmdr1 gene, mutations were observed at codons N86Y (15.3%, 24/157), Y184F (71.3%, 112/157), S1034C (1.9%, 3/157), N1042D (6.4%, 10/157), and D1246Y (3.2%, 5/157), including mixed alleles (Table 2).

Polymorphism in pfdhfr and pfdhps

All 157 P. falciparum samples were successfully genotyped for the pfdhfr gene. Mutations were detected at codons N51I and S108N in 100% of samples (153 pure mutants, 4 with mixed-allele infections), and at C59R in 98.7% of samples (120 pure mutants, 35 with mixed-allele infections) (Table 2). For the pfdhps gene, mutations were observed at codons S436A (1.9%; 1 pure mutant, 2 mixed), A437G (93.6%; 129 pure mutants, 18 mixed), K540E (98.7%; 115 pure mutants, 40 mixed), and A581G (52.2%; 8 pure mutants, 74 with mixed-allele infections) (Table 2).

Mutant alleles of pfdhfr and pfdhps were combined to define resistance haplotypes (Table 3). Among single-allele infections, the quintuple mutant haplotype was the most prevalent, observed in 82.3% (47/57) of samples, followed by the sextuple mutant haplotype in 12.3% (7/57). In samples with mixed-allele calls, 91% (91/100) harbored either the quintuple or sextuple mutant haplotype, indicating widespread circulation of highly resistant genotypes (Table 3).

Table 3.

Prevalence of combined pfdhfr and pfdhps mutant haplotypes

Mutation	Haplotype	n (%)
Single Allele Mono-infection [n=57]
Triple	pfdhfr I51R59N108	1 (1.8)
Quadruple	pfdhfr I51R59N108 + pfdhps G437	0
Quintuple	pfdhfr I51R59N108 + pfdhps G437E540	47 (82.3)
Sextuple	pfdhfr I51R59N108 + pfdhps G437E540G581	7 (12.3)
Others	pfdhfr I51C59N108 + pfdhps G437E540	2 (3.5)
	pfdhfr I51C59N108 + pfdhps G437E540 G581	0
Mixed Alleles Infection [n=100]		n (%)
	pfdhfr I51C/R59N108+ pfdhps A/G437E540A/G581	1 (1.0)
	pfdhfr I51C/R59N108+ pfdhps A/G437K/E540A581	1 (1.0)
	pfdhfr I51C/R59N108+ pfdhps A/G437K/E540A/G581	4 (4.0)
	pfdhfr I51C/R59N108+ pfdhps A437K540A581	1 (1.0)
	pfdhfr I51C/R59N108+ pfdhps G437E540A581	9 (9.0)
	pfdhfr I51C/R59N108+ pfdhps G437E540A/G581	13 (13.0)
	pfdhfr I51C/R59N108+ pfdhps G437E540G581	1 (1.0)
	pfdhfr I51C/R59N108+ pfdhps G437K/E540A/G581	1 (1.0)
	pfdhfr I51R59N108+ pfdhps A/G437K/E540A581	6 (6.0)
	pfdhfr I51R59N108+ pfdhps A/G437K/E540A/G581	5 (5.0)
	pfdhfr I51R59N108+ pfdhps A437K/E540A581	4 (4.0)
	pfdhfr I51R59N108+ pfdhps A437K/E540A/G581	4 (4.0)
	pfdhfr I51R59N108+ pfdhps G437E540A/G581	34 (34.0)
	pfdhfr I51R59N108+ pfdhps G437K/E540A581	2 (2.0)
	pfdhfr I51R59N108+ pfdhps G437K/E540A/G581	10 (10.0)
	pfdhfr N/I51C/R59S/N108+ pfdhps A/G437K/E540A581	1 (1.0)
	pfdhfr N/I51C/R59S/N108+ pfdhps A/G437K/E540A/G581	1 (1.0)
	pfdhfr N/I51C/R59S/N108+ pfdhps G437E540A581	1 (1.0)
	pfdhfr N/I51C/R59S/N108+ pfdhps G437K/E540A/G581	1 (1.0)

Mutation Prevalence and Co-occurrence

Figure 2 summarizes the distribution and co-occurrence of resistance-associated mutations across four key genes: pfcrt, pfdhfr, pfdhps, and pfmdr1. The most prevalent mutations included pfdhfr N51I, C59R, and S108N, pfdhps A437G and K540E, and pfcrt K76T, all detected in the majority of samples (Figure 2A). In contrast, pfdhps A581G and pfmdr1 N86Y and D1246Y were found in a smaller subset of isolates. The most frequent mutation combinations featured the pfdhfr triple mutant (N51I, C59R, S108N) alongside pfcrt K76T, with or without the pfdhps double mutant (A437G and K540E). The co-occurrence heatmap (Figure 2B) shows frequent co-detection of mutations within individual samples, particularly among the pfdhfr triple mutant and between pfdhps A437G and K540E. Pfcrt K76T also commonly co-occurred with both pfdhfr and pfdhps mutations, reflecting the predominance of multi-locus resistance profiles in the population.

Figure 2. Prevalence and Co-occurrence of Drug Resistance Mutations in Plasmodium falciparum Isolates.

A) UpSetR plot showing combinations of resistance-associated alleles in pfdhfr, pfdhps, pfmdr1, and pfcrt genes. Bars indicate the frequency of samples with each mutation profile; the matrix below shows which mutations are involved. The left bar chart indicates overall mutation prevalence. B) Heatmap illustrating co-occurrence frequencies of mutations in the same four genes. Darker shades represent higher co-occurrence between mutations (i.e., codons), while lighter shades indicate lower or no co-occurrence. In both UpSetR and heatmap plots, samples with mixed allele calls (both mutant and wild-type at a locus) were classified as mutant, reflecting potential subclonal resistance. Mutation combinations represent presence within individual samples and do not imply genomic linkage.

Gene Duplication in Pfmdr1 and Pfpm2

Copy number variations (CNVs) for plasmepsin II (pfpm2) and Multidrug resistance gene 1 (pfmdr1) were successfully assessed in all 164 samples by digital PCR. No parasites exhibited gene duplication (CNV > 1.5) for pfpm2 (Figure 3A). In contrast, CNVs were observed in pfmdr1 in three samples (1.8%), indicating gene amplification (Figure 3B).

Figure 3. Copy number variations in (a) plasmepsin 2 and (b) Multidrug resistance gene 1.

None of the 164 samples exhibited gene duplication for pfpm2, while three samples showed copy number variations (CNVs) exceeding 1.5 in pfmdr1. Purple dots in the pfpm2 and pfmdr1 assays represent positive controls (KH004 and Dd2).

DISCUSSION

This study provides critical molecular insights into the evolving drug resistance landscape of P. falciparum in Cibitoke Province, Burundi, revealing widespread resistance to CQ and SP, alongside concerning trends of partner drug efficacy in ACTs. No validated or candidate pfkelch13 mutations associated with ART-R were detected. However, the identification of a novel pfkelch13 mutation, A675T, is notable, as it occurs at the same codon as the well-characterized A675V mutation, which has been linked to delayed parasite clearance in Rwanda[27,56], Uganda[58,59], Kenya[57] and Tanzania[60]. While the functional significance of A675T remains unclear, its presence warrants further surveillance and functional studies to assess its potential role in resistance evolution.

Background mutations previously implicated in contributing to the emergence of artemisinin resistance in Southeast Asia[29], such as pffd (D193Y), pfarps10 (V127M), and pfcrt (I356T), were detected at varying frequencies. The prevalence of pffd D193Y (45.9%, including samples with mixed-allele) was significantly higher than that reported in Rwanda[19] whereas this mutation was not found in Ghana [34], Uganda[31,32,61], and Mozambique[62]. This may reflect localized dynamics driven by unique transmission patterns, drug selection pressures, or parasite population genetics in Burundi. Additionally, pfmdr2 I492V (41.2%) pfarps10 V127M (11.5%), pfcrt I356T (10.2%) and pfatg18 T38I (11.5%) were observed. Similar frequencies for some of these markers have been reported in Uganda[31,32], Tanzania[33], Ghana[34]. Studies showed that these mutations facilitate pfkelch13-mediated resistance in Southeast Asia [29,30]; their impact on sub-Saharan Africa is unknown.

AL remains the first-line treatment for uncomplicated malaria in Burundi. However, the high prevalence (57.9%) of the pfmdr1 N86-F184-D1246 mutant haplotype, including mixed alleles, raises concerns about emerging AL tolerance. This haplotype is associated with AL selection pressure and may confer a fitness advantage to P. falciparum under AL drug pressure [63–68]. These findings mirror patterns observed in other regions, where sustained AL use has driven the increasing prevalence of the N-F-D haplotype[40,69–71].

Amid growing concerns over artemisinin resistance, the potential reintroduction of former first-line therapies, such as CQ, has been proposed for the treatment of P. falciparum malaria [40,72]. In several endemic countries, CQ withdrawal has led to the re-emergence of CQ-sensitive P. falciparum parasites [73–77]. However, in Burundi, despite CQ being discontinued over two decades ago, 94.9% of isolates analyzed in this study harbored the CQ resistance-associated pfcrt I74-E75-T76 triple mutant haplotype. This persistence reflects trends observed in other East African countries [78–83]. The continued use of artesunate–amodiaquine (ASAQ) as first-line therapy in Burundi until 2019 may have contributed to the maintenance of these pfcrt mutations through sustained selection pressure from amodiaquine [84].

SP remains a cornerstone of IPTp and SMC strategies in Burundi. This study reveals a high prevalence of SP resistance, with the quintuple mutant haplotype detected in 82.3% of mono-infections and the sextuple haplotype in 12.3%. Additionally, 91% of mixed alleles carried these resistant haplotypes. While comparably high levels of SP-resistant P. falciparum have been reported in neighboring East African countries [85–90], resistance remains less prevalent in West Africa[11,91,92]. These findings pose significant risks to maternal and fetal health, reinforcing the urgent need for alternative chemoprevention strategies.

While Nanopore sequencing facilitates genomic surveillance, stringent quality control is crucial. One sample initially appeared to carry multiple pfkelch13 mutations; however, these mutations were not detected when sequencing was repeated twice. This indicates that the initial mutations stemmed from errors introduced during PCR or sequencing.

The identification of novel pfkelch13 mutations and the widespread prevalence of resistance markers underscore the need for longitudinal surveillance, expanded geographic sampling, and functional validation of candidate mutations. These efforts are crucial for refining interventions and optimizing malaria control strategies. Regional collaboration with neighboring countries facing similar resistance challenges is essential to enhance coordinated surveillance and inform adaptive treatment policies.

In conclusion, this study establishes a critical molecular baseline for resistance monitoring in Burundi, emphasizing the urgency of re-evaluating treatment protocols, strengthening surveillance systems, and fostering regional cooperation. Addressing these challenges will help safeguard Burundi’s malaria control efforts, mitigate the spread of resistance, and contribute to global malaria elimination initiatives.

Funding Statement

This work was supported by funds provided by the University of Notre Dame to C. K. C.K. was supported by the Bill & Melinda Gates Foundation (Grant INV-005898). T.K. received support from the Indiana Clinical and Translational Sciences Institute, funded in part by Grant Number UM1TR004402 from the National Institutes of Health, National Center for Advancing Translational Sciences, through the Clinical and Translational Sciences Award. T.K. is also supported by a postdoctoral fellowship from the Environmental Change Initiative at the University of Notre Dame, which provides partial salary support.

DATA AVAILABILITY

All nucleotide sequences generated in this study have been deposited in GenBank under accession numbers PV548606–PV548676. Due to GenBank length restrictions, short sequences (<50 bp) corresponding to pfdhps (codons 436–437) and pfmdr1 (codons 1034–1042) are provided in Supplementary Table S6.