Artemisinin Resistance at the China-Myanmar Border and Association with Mutations in the K13 Propeller Gene

Abstract

Artemisinin resistance in Plasmodium falciparum parasites in Southeast Asia is a major concern for malaria control. Its emergence at the China-Myanmar border, where there have been more than 3 decades of artemisinin use, has yet to be investigated. Here, we comprehensively evaluated the potential emergence of artemisinin resistance and antimalarial drug resistance status in P. falciparum using data and parasites from three previous efficacy studies in this region. These efficacy studies of dihydroartemisinin-piperaquine combination and artesunate monotherapy of uncomplicated falciparum malaria in 248 P. falciparum patients showed an overall 28-day adequate clinical and parasitological response of >95% and day 3 parasite-positive rates of 6.3 to 23.1%. Comparison of the 57 K13 sequences (24 and 33 from day 3 parasite-positive and -negative cases, respectively) identified nine point mutations in 38 (66.7%) samples, of which F446I (49.1%) and an N-terminal NN insertion (86.0%) were predominant. K13 propeller mutations collectively, the F446I mutation alone, and the NN insertion all were significantly associated with day 3 parasite positivity. Increased ring-stage survival determined using the ring-stage survival assay (RSA) was highly associated with the K13 mutant genotype. Day 3 parasite-positive isolates had ∼10 times higher ring survival rates than day 3 parasite-negative isolates. Divergent K13 mutations suggested independent evolution of artemisinin resistance. Taken together, this study confirmed multidrug resistance and emergence of artemisinin resistance in P. falciparum at the China-Myanmar border. RSA and K13 mutations are useful phenotypic and molecular markers for monitoring artemisinin resistance.

INTRODUCTION

Artemisinin (ART)-based combination therapies (ACTs) have played an indispensable role in reducing global malaria-associated mortality and morbidity. However, these achievements are threatened by the recent emergence of ART resistance in Plasmodium falciparum in the Greater Mekong Subregion (GMS) of Southeast Asia. Clinical ART resistance was first detected in western Cambodia (1–3) but is now detected in at least six areas of the GMS due to spread, independent emergence, or both (4–8). ART resistance is associated with a parasite clearance half-life of >5 h, whereas the normal value is ∼2 h (6, 7). Accurate determination of parasite clearance half-life requires monitoring parasitemia in patients' peripheral blood every 6 h (9). In practice, the proportion of patients remaining parasite positive 3 days after treatment with ART drugs is an alternative indicator for delayed parasite clearance (10). A working definition proposed by the WHO for suspected ART resistance uses “>10% of cases with parasites detectable on day 3 after treatment with an ACT.” The delayed-clearance phenotype cannot be measured using standard in vitro assays but corresponds to decreased susceptibility of ring-stage parasites to ARTs in an in vitro ring-stage survival assay (RSA) (11). However, this novel method has been applied to only a small number of field parasites from Cambodia with defined parasite clearance rates and to several genetically manipulated laboratory strains (11–14).

Recently, mutations in a kelch propeller gene (K13) on P. falciparum chromosome 13 (PF3D7_1343700) were associated with ART resistance in clinical parasites from Cambodia (12). Genetic manipulations conclusively showed that certain K13 propeller mutations conferred in vitro ART resistance (14, 15). Mechanistic studies indicated that resistant parasites upregulated the unfolded protein response pathway (16), and the K13 mutations probably mediated ART resistance via phosphatidylinositol-3-kinase signaling (17). A large multicenter clinical study further demonstrated that point mutations in the K13 propeller domain are collectively associated with the slowly clearing infections (6). Molecular surveillance showed that K13 mutations associated with ART resistance were restricted to certain areas of the GMS, with C580Y being the predominant mutation and approaching fixation in western Cambodia (12, 18–20), but such mutations were not identified in Africa (6, 19, 21–24). Even within the GMS, there is significant geographic heterogeneity in both the patterns of the K13 mutations and their prevalence (25–27), possibly reflecting different drug histories and evolutionary origins (28). To date, the correlations of K13 mutations with delayed parasite clearance have been demonstrated only in limited studies (6, 12, 18, 19).

In China's Yunnan province, ARTs have been used since the late 1970s (29), but clinical studies clearly documenting ART resistance in this region are lacking (5, 30, 31). Here, we aim to provide conclusive evidence of ART resistance in the China-Myanmar border area and to validate RSA and K13 for monitoring ART resistance. By performing correlation analysis between day 3 parasite positivity from clinical studies, RSA results with culture-adapted parasites, and K13 polymorphisms, we provide strong evidence indicating the emergence of ART resistance at the China-Myanmar border and validated RSA as a phenotypic measure and K13 as a promising molecular marker for monitoring ART resistance.

MATERIALS AND METHODS

Study design and participants.

We analyzed three efficacy studies of ART family drugs conducted previously in Kachin State of northeast Myanmar (97.56°E, 24.75°N) and adjacent Yingjiang County, Yunnan Province, China. Uncomplicated P. falciparum malaria was identified by microscopy after screening febrile patients attending local clinics (Table 1). The two studies conducted in 2008 and 2012 to 2013 were 2-day dihydroartemisinin-piperaquine (DP) treatment (30, 32), whereas the 2009-2010 study used 7-day artesunate (ATS) monotherapy (31). Patients with P. falciparum monoinfection and no signs of severe malaria were enrolled. Prior to treatment, 2 to 5 ml of venous blood was obtained from patients and cryopreserved in liquid nitrogen for laboratory culture adaption. Parasite density was determined as described earlier (32). After treatment, patients were followed up on days 1, 2, 3, 7, 14, and 28 (and 42 in 2012 to 2013). Informed consent/assent was obtained from all patients or legal guardians of participants prior to enrollment. All studies were approved by the ethical committees from the participating institutions and the local health department in Kachin.

TABLE 1.

Demographic information of P. falciparum patients involved in three clinical efficacy studies of artemisinin drugs conducted between 2008 and 2013 at the China-Myanmar border

Parameter	Value(s) for study:
	1 (2008; DPa)	2 (2009–2010; artesunateb)	3 (2012–2013; DPa)
Patient data
Febrile patients screened (no.)	1,073	1,095	9,044
P. falciparum patients identified (no.)	91	65	142
P. falciparum patients recruited (no.)	74	65	109
Completed 28-day follow-up (no.)	64	47	71
Age (yr)	2–60	3–58	1–60
Male:female	54:20	48:17	66:43
Mean (range) day 0 asexual parasite density (/μl)	16,666 (560–625,040)	60,351 (1,004–657,818)	3,905 (140–147,600)
Day 28 ACPR (%)	100	95.9 (1 LCF, 1 LPF)	100c
Parasite positivity,d n (%)
Day 1	21 (32.8)	52 (80)	32 (45.1)
Day 2	11 (17.2)	27 (41.5)	18 (25.4)
Day 3	4 (6.3)	15 (23.1)	5 (7.0)

^aA concentration of ∼6 mg/kg of body weight of DHA in 4 doses in 2 days was used.

^bArtesunate was used at 16 mg/kg for 7 days.

^cDay 42 follow-up with 100% ACPR.

^dParasite positivity rates were calculated using patients available on each date instead of patient numbers who completed the 28-day follow-up.

Laboratory procedures.

Thin and thick blood smears were stained with 2% Giemsa for 30 min and read by two experienced microscopists. Parasites on thick blood smears were counted against 200 leukocytes, and parasite density (number of parasites per microliter of blood) was calculated assuming 8,000 white blood cells per milliliter of peripheral blood. Parasite genomic DNA was isolated from blood spots on filter paper or cultured parasites using a QIAamp DNA minikit (Qiagen). Parasites were genotyped at three polymorphic antigenic markers, merozoite surface protein 1 (msp1), msp2, and glutamate-rich protein (glurp), as described previously, and only monoclonal infections were used for culture adaptation (33).

Parasite culture and drug assays.

For culture adaptation, frozen clinical samples were thawed in saline processing solutions, washed twice with RPMI 1640 medium, mixed with fresh type O⁺ human red blood cells (RBCs), and suspended at 5% hematocrit in complete medium supplemented with 6% AB human serum (33). The culture was maintained under an atmosphere of 90% N₂, 5% O₂, and 5% CO₂. On average, parasites were continuously cultured for 3 to 4 weeks before drug sensitivities were measured.

Parasite cultures were assayed for their in vitro susceptibilities to eight antimalarial drugs, chloroquine (CQ), piperaquine (PPQ), mefloquine (MQ), quinine (QN), lumefantrine (LMF), pyronaridine (PND), artesunate (ATS), and dihydroartemisinin (DHA), using a SYBR green I-based assay (34). CQ, MQ, QN, and ART drugs were purchased from Sigma (St. Louis, MO); PPQ was from Chongqing Kangle Pharmaceutical Co. (Chongqing, China), while LMF and PND were from Kunming Pharmaceutical Co. (Kunming, Yunnan, China). The stock solutions of CQ and PND were prepared in distilled water, PPQ in 90% methanol and 10% 1 M hydrochloric acid, while other drugs were dissolved in dimethyl sulfoxide (DMSO). Cultures were synchronized by 5% d-sorbitol treatment (35), and ring-stage parasites were diluted with fresh complete medium to 1% hematocrit and 0.5% parasitemia. In vitro drug assays were performed in 96-well microtiter plates with serially diluted concentrations of each drug. Three technical repeats and two biological replications were performed for each drug concentration and each parasite isolate. For consistency, 3D7 was included throughout the study as an internal reference strain.

RSA was performed on culture-adapted parasite isolates by following the established procedure (11). Briefly, tightly synchronized young ring-stage parasites (0 to 3 h) were exposed to 700 nM DHA or 0.1% DMSO for 6 h, washed with RPMI 1640, and continuously cultured for 66 h, and then ∼10,000 RBCs on thin blood smears of the culture were assessed independently by two microscopists to count viable parasites. A third microscopist read the slides in cases of >20% discrepancy between the two readings. Survival rates were expressed as ratios of viable parasites in DHA-exposed and DMSO-exposed samples.

Sequencing of genes associated with drug resistance.

Polymorphisms in candidate genes associated with drug resistance were determined by PCR and sequencing using previously described protocols. These include two pfcrt fragments covering codons 72 to 76 and 220 and two pfmdr1 fragments including codons 86, 184, 1034, 1042, and 1246 (33, 36) and the pfmrp1 gene (37), pfdhfr fragments containing codons 51, 59, 108, and 164, pfdhps fragments containing codons 436, 437, 540, and 581 (36), a pfnhe1 fragment containing the ms4760 minisatellite (33), and the full-length K13 gene (27). The haplotype network was constructed by Network software using the median joining algorithm (38).

Statistical analyses.

Statistical analyses were performed using GraphPad Prism 6.0. The geometric mean 50% inhibitory concentration (IC₅₀) and 95% confidence interval (CI) were calculated by fitting the drug response data to a sigmoid curve. Medians and interquartile ranges (IQR) were calculated if the data were not normally distributed. IC₅₀s between the groups were compared by Mann-Whitney U test. Correlations between IC₅₀s of drugs were determined using the Spearman's test. Mutation frequencies between two groups were compared using Fisher's exact test. Associations between assay results and mutations were investigated by multiple t tests.

Nucleotide sequence accession numbers.

K13 gene sequences reported in this study were deposited in GenBank under accession numbers KT328077 to KT328133.

RESULTS

Clinical efficacies of ART drugs.

We analyzed parasites from three recent efficacy studies conducted in 2008, 2009 to 2010, and 2012 to 2013 on ART drugs (ATS monotherapy and DP) in the China-Myanmar border area. Collectively, >11,000 febrile patients were screened, and 248 patients with single P. falciparum infections were enrolled. ART drugs remained highly efficacious in treating falciparum malaria with 28-day adequate clinical and parasitological responses (ACPR) above 95.9% (Table 1). Notably, DP retained 100% day 28 ACPR and 100% day 42 ACPR for the two DP studies, although the populations in both studies were relatively small. Further, the two DP studies also showed day 3 parasite-positive rates below 10% (6.3 and 7.0%, respectively). In contrast, the day 3 parasite-positive rate for the ATS study reached 23.1%. Interestingly, the one case with late clinical failure (LCF) (fever and parasitemia on day 20) and the one with late parasitological failure (LPF) (parasitemia on day 27) in the ATS monotherapy study were both day 3 parasite negative. When parasite density data from three trials were analyzed together, the day 0 parasite density was significantly higher in the day 3 parasite-positive group (P = 0.0114 by Mann-Whitney U test).

K13 mutations and day 3 parasite positivity.

To validate the predictability of K13 mutations for potential ART resistance, we selected 24 and 33 day 3 parasite-positive and -negative cases, respectively, from the three efficacy studies. The day 3 parasite-negative cases were chosen to match each of the day 3 parasite-positive cases from each of the efficacy studies. Sequencing 57 full-length K13 sequences identified 9 nonsynonymous mutations in 38 (66.7%) samples, of which 7 mutations were located in the propeller domain (i.e., after amino acid 440) (Table 2). The mutations are very distinctive from those found in other regions of the GMS. F446I was the predominant mutation, with 49.1% prevalence, whereas other mutations were at low frequencies (1.8 to 5.3%). Of the four mutations correlated with ART resistance in Cambodia, only R539T was detected in one isolate (1.8%). Besides point mutations, an NN insert between amino acids 136 and 137 in the 3D7 sequence was identified in 49 (86.0%) samples. There were significant differences in the prevalence of F446I mutation and NN insert between day 3 parasite-positive and -negative cases (P < 0.05 by Fisher's exact test) (Table 2). The total mutations in the propeller domain, the predominant F446I mutation, and the NN insert all were significantly associated with day 3 parasitemia in patients (odds ratios of 5.30, 4.86, and 16.33, respectively; P < 0.05). It is worth mentioning that the day 0 parasite in the LCF case had a wild-type K13 gene, whereas the day 0 parasite in the LPF case had both an NN insert and the F446I mutation in the K13 gene.

TABLE 2.

Prevalence of K13 gene mutations in P. falciparum clinical isolates from day 3 parasite-positive and -negative patients after treatment with artemisinin family drugs

Mutation	Prevalence of mutations in day 3 samples (no. [%])			Odds ratio (95% CI)	P valueb
	Parasite positive (n = 24)	Parasite negative (n = 33)	Total (n = 57)
NN insertiona	24 (100.0)	25 (75.8)	49 (86.0)	16.33 (0.9–298.7)	0.0158
E252Q	0	1 (3.0)	1 (1.8)
R255K	0	1 (3.0)	1 (1.8)
F446I	17 (70.8)	11 (33.3)	28 (49.1)	4.86 (1.6–15.2)	0.0074
N458Y	0	1 (3.0)	1 (1.8)
C469Y	1 (4.2)	0	1 (1.8)
R539T	1 (4.2)	0	1 (1.8)
P574L	1 (4.2)	2 (6.1)	3 (5.3)
A676D	0	1 (3.0)	1 (1.8)
H719N	0	1 (3.0)	1 (1.8)
After 440	20 (83.3)	16 (48.5)	36 (63.2)	5.3 (1.5–19.0)	0.0116

^aPresence of NN insertion between amino acids 136 and 137.

^bComparison between the two groups (Fisher's exact test).

Correlation of the ring-stage survival phenotype with day 3 parasitemia.

We culture adapted 44 clinical parasite isolates (19 from day 3 parasite-positive and 25 from day 3 parasite-negative cases) from the three efficacy studies and measured their ring survival by RSA. Ring survival rates of these parasites varied drastically, ranging from 0 to 61.1% (Table 3 and Fig. 1). Remarkably, the median percent ring survival rate of the day 3 parasite-positive group (5.0%) was ∼10 times greater than that of the day 3 parasite-negative group (0.5%) (P < 0.0001) (Table 3), indicating that parasites having higher ring survival rates tended to persist through day 3 in vivo. Despite this general consistency, three parasite isolates in the day 3 parasite-negative group had ring survival rates (7.4%, 21.6%, and 61.1%) radically different from those of the rest of the samples in this group (Fig. 1). K13 sequencing revealed that these three parasites all carried a mutation in the K13 propeller domain (N458Y in one isolate and F446I in two isolates), indicating that RSA results and K13 genotypes were more compatible with each other. Furthermore, parasites with K13 propeller mutations had an ∼8-fold higher ring-stage survival rate than wild-type parasites (P = 0.0032 by Mann-Whitney U test) (Fig. 1). Similarly, parasites with the 446I mutation also had a significantly higher ring survival rate than the F446 parasites (P = 0.0326 by Mann-Whitney U test) (see Fig. S1 in the supplemental material). Interestingly, the K13 NN insert also was significantly associated with increased ring survival (P = 0.0046 by Mann-Whitney U test).

TABLE 3.

In vitro IC₅₀ values and ring-stage survival of culture-adapted P. falciparum from day 3 parasite-positive and -negative patients

Druga	Parasite positive (n = 19)			Parasite negative (n = 25)			P valueb	Total (n = 44)			3D7 (mean)	P valuec
	Median (IQR)	Mean (95% CI)	Range	Median (IQR)	Mean (95% CI)	Range		Median (IQR)	Mean (95% CI)	Range
DHARSA	5.0 (2.2–7.7)		1.0–19.8	0.5 (0–1.9)		0–61.1	<0.0001	2.0 (0.1–6.1)		0–61.1	0	<0.0001
ATS	1.6 (1.2–2.2)		1.1–4.8	2.3 (1.6–3.2)		1.1–6.3	0.0155	1.8 (1.4–2.8)		1.1–6.3	1.7	0.0040
DHA	2.2 (1.8–3.0)		0.8–6.4	2.1 (1.6–2.7)		1.1–3.6	0.3996	2.2 (1.6–2.8)		0.8–6.4	1.5	<0.0001
PND	9.1 (7.0–11.8)		4.5–34.8	7.2 (6.0–8.4)		5.1–14.5	0.0642	7.6 (6.3–9.8)		4.5–34.8	4.1	<0.0001
CQ		1,567.0 (1,155.0–1,980.0)	478.8–3,278.0		1,606.0 (1,118.0–2,094.0)	296.2–4,097.0	0.6386		1,589.0 (1,277.0–1,901.0)	296.2–4,097	17.6	<0.0001
PPQ		23.8 (18.9–28.8)	8.9–46.4		20.0 (17.3–22.7)	10.5–34.5	0.1305		21.7 (19.1–24.3)	8.9–46.4	14.7	<0.0001
MQ		34.1 (21.5–46.7)	2.0–93.8		23.0 (19.0–27.0)	7.1–44.7	0.2346		28.0 (22.0–34.0)	2.0–93.8	17.2	0.0008
QN		47.2 (38.4–55.9)	21.1–79.2		85.1 (57.0–113.2)	28.3–207.2	0.0563		68.1 (51.5–84.7)	21.1–207.2	15.1	<0.0001
LMF		4.8 (3.0–6.6)	1.4–12.5		9.0 (6.4–11.6)	1.4–19.6	0.0247		7.1 (5.4–8.8)	1.4–19.6	5.2	0.0310

^aDHA^RSA was performed by exposing tightly synchronized young ring-stage parasites (0 to 3 h) to a bolus DHA treatment for 6 h and measuring parasite survival (as percentages) 66 h later. The rest of the assays measuring IC₅₀ values (in nM) were the standard drug assay where parasites were exposed to continuous drugs for 72 h, and parasite growth was measured by SYBR green I fluorescence.

^bMann-Whitney U test.

^c Student's t test.

FIG 1.

Comparison of ring-stage parasite survival rates with in vivo parasite clearance and K13 genotypes. Ring-stage survival rates of the culture-adapted parasites measured by the RSA were compared between day 3 parasite-positive and -negative samples, between parasites with and without the NN insertion at the N terminus, and between parasites with and without mutations in the K13 propeller domain. Wild-type (WT) parasites are shown as open circles, whereas parasites carrying K13 mutations are color coded. Statistical analysis was performed using a Mann-Whitney U test, and P values are shown.

In vitro sensitivities of parasites to eight antimalarial drugs.

The IC₅₀s of eight antimalarials were determined for 44 culture-adapted parasites and 3D7. All parasites were highly resistant to CQ, with IC₅₀s exceeding 296 nM (Table 3). The ranges of IC₅₀s to PPQ, MQ, and QN were relatively wide, whereas the median or mean IC₅₀s to LMF, PND, DHA, and ATS were below 10 nM. Nonetheless, the median or mean IC₅₀s of the field isolates to all tested drugs were significantly higher than those of 3D7 (P < 0.0001 by Student's t test). Pairwise comparison showed that there were highly significant, positive correlations between ATS and DHA and between ATS and LMF (P < 0.005 by Spearman's test) (Fig. 2). In addition, positive correlations also were found between the RSA result and MQ and between ATS and QN (P < 0.05). In comparison, significant, negative correlations were noticed between ATS and PND, LMF and PND, and LMF and MQ (P < 0.05). Surprisingly, the ring-stage survival rates and IC₅₀s to ARTs did not show any significant correlation.

FIG 2.

Correlations between IC₅₀s of cultured parasite isolates to eight antimalarial drugs. The correlations between IC₅₀s were analyzed by Spearman's test. The degree of correlation between IC₅₀s of two drugs is color coded, and coefficients are shown in the upper diagonal. One and two asterisks show the significance levels. Except for DHA^RSA, which is the result from the RSA, susceptibilities (IC₅₀s) to the rest of the drugs were measured using the SYBR green I-based assay. ATS, artesunate; DHA, dihydroartemisinin; CQ, chloroquine; PPQ, piperaquine; QN, quinine; PND, pyronaridine; LMF, lumefantrine; MQ, mefloquine.

The IC₅₀s for CQ, PPQ, MQ, QN, PND, and DHA were not significantly different between day 3 parasite-positive and -negative isolates (Table 3). However, parasites from day 3 parasite-negative cases had higher IC₅₀s for ATS and LMF than those from day 3 parasite-positive cases (P < 0.05 by Mann-Whitney U test).

Molecular polymorphisms of other genes associated with drug resistance.

Polymorphisms in pfcrt, pfdhfr, pfdhps, pfmdr1, pfmrp1, and pfnhe-1 were genotyped and are shown in Table 4 (also see Table S1 in the supplemental material). All parasites carried CVIET at positions 72 to 76 of pfcrt, and both 76T and 220S reached fixation (see Table S1). Similarly, C59R, S108N, and I164L in pfdhfr and A437G and K540E/N in pfdhps approached fixation. The quintuple mutants with pfdhfr (N51I, C59R, and S108N) and pfdhps (A437G and K540E) exceeded 50%, indicating high resistance to antifolates. The frequencies of the seven SNPs in pfmrp1 ranged from 13.2% to 71.1%, with mutations H191Y, S437A, and I876V occurring in >50% isolates. Parasites harboring I876V exhibited higher IC₅₀s to most of the tested drugs, whereas those with H191Y, S437A, and H785N had decreased susceptibilities to DHA, ATS, CQ, PPQ, and LMF (see Fig. S1). Sequencing pfnhe-1 found six ms4760 alleles (39), with allele 7 prevailing in both day 3 parasite-positive and -negative groups (>50%) (see Fig. S2). In agreement with our previous study (33), parasites with higher R₁ (DNNND) copy numbers displayed reduced sensitivity to QN, whereas an increased R₂ (NHNDNHNNDDD) copy number was significantly associated with increased QN susceptibility (P = 0.0199). Interestingly, the NN insert and F446I mutation in the K13 gene were associated with significantly increased sensitivity to QN (P < 0.05) (see Fig. S1 in the supplemental material). Except for the prevalent K13 mutations (NN insert and F446I), none of the mutations in other genotyped genes showed significant differences between day 3 parasite-positive and -negative groups (see Table S1).

TABLE 4.

Comparison of frequencies of major haplotypes of five genes associated with drug resistance between day 3 parasite-positive and -negative isolates

Gene	Haplotypea	Presence of parasite (no. [%]) on day 3		P valueb
		Positive	Negative
pfdhfr	NRNL	2 (11.8)	7 (33.3)	0.1484
	IRNL	13 (76.5)	12 (57.1)	0.3068
pfdhps	AGEA	10 (58.8)	10 (47.6)	0.5318
	SGEG	1 (5.9)	5 (23.8)	0.1965
pfmrp1	HNSHITF	6 (35.3)	2 (9.5)	0.1066
	YSAHITF	1 (5.9)	4 (19.0)	0.3551
	YNAHVTF	1 (5.9)	5 (23.8)	0.1965
	YNANVMI	2 (11.8)	0 (0)	0.1935
	YNANVMI	2 (11.8)	0 (0)	0.1935
pfmdr1	NYN	10 (58.8)	15 (71.4)	0.5017
	NFN	7 (41.2)	6 (28.6)	0.5017
pfnhe-1	6	2 (11.8)	5 (23.8)	0.4267
	7	12 (70.6)	11 (52.4)	0.3264

^aPoint mutations are shown in boldface.

^bComparison between the two groups was done by Fisher's exact test.

The SNPs in the seven genotyped drug resistance genes gave rise to 34 haplotypes in the 38 parasite isolates (see Fig. S3 in the supplemental material), demonstrating high-level genetic diversity in this region. Four and two samples showed haplotypes 1 and 2, respectively, while each of the remaining isolates was unique (see Fig. S3). It is noteworthy that parasites with haplotypes 13 and 25 showed reduced sensitivities to all tested drugs. To examine the phylogenetic relationship of the parasites, a haplotype network was generated based on SNPs observed in the seven genes. Four different branches of the network (groups 1 to 4) were connected through a linker group (Fig. 3A). Day 3 parasite-positive isolates appeared to be polyphyletic and were distributed across the entire network, suggesting independent origins of resistant mutations in different genes on diverse backgrounds. However, we did observe some clustering of parasite isolates based on day 3 parasitemia. Six haplotypes within group 1 contained only day 3 parasite-positive isolates and were separated from group 2, which primarily contained day 3 parasite-negative isolates. This haplotype grouping was strongly influenced by the F446I mutation, whereas all other mutations possessed common alleles between groups 1 and 2 (Fig. 3B). This suggests that some parasites with the F446I mutation shared a common origin, possibly resulting from a recent spread of parasites carrying the F446I mutation. In comparison, groups 3 and 4 were separated mostly based on mutations in pfmrp1 and pfdhps.

FIG 3.

Median-joining haplotype network of P. falciparum isolates harboring mutations in six drug resistance-associated genes. (A) The haplotype network was constructed for P. falciparum isolates using 34 haplotypes obtained from amino acid changes observed in pfdhfr, pfdhps, pfcrt, pfmdr1, pfmrp1, and K13. The size of each circle corresponds to the number of samples sharing the same haplotype, and the length of an edge is proportional to the number of variations between two haplotypes. Four branches of the network are circled and named groups 1, 2, 3, and 4, and the connecting group is named Link. (B) Consensus sequence corresponding to each group. Point mutations are in boldface and underlined, with the locations of mutations being 51/59/108/164 in pfdhfr, 436/437/540/581 in pfdhps, 72/74/75/76/220 in pfcrt, 86/184/1042 in pfmdr1, 191/325/437/785/876/1007/1390 in pfmrp1, and 446/252/255/458/469/539/574/676/719 in K13. Dashes indicate the presence of both alleles.

DISCUSSION

In the China-Myanmar border area where ART family drugs have the longest history of deployment, ART resistance has not been confirmed. Our studies showed 100% efficacies with DP (30, 32), compared to one LCF and one LPF out of 65 patients enrolled in ATS monotherapy (31). Consistently, <10% of patients receiving DP treatment had parasitemia by day 3, but such patients in the ATS study reached over 20%, suggesting that increased efficacies in DP treatment relied on the partner drug PPQ. ART resistance manifested as delayed parasite clearance rate is due largely to prolonged and dormant ring stages (13, 40). This resistance phenotype cannot be captured with the conventional drug assays measuring IC₅₀s. Indeed, none of the eight drugs tested showed significantly higher IC₅₀s in the day 3 parasitemic group. In contrast, RSA with 44 culture-adapted clinical isolates showed a >10-fold increase in ring survival in the day 3 parasitemic group. The suitability of K13 gene polymorphisms as a more convenient and accurate marker for rapid epidemiological assessment of ART resistance was first evaluated in a multicenter study (6). In our study, patients with parasites carrying K13 propeller mutations had >5 times odds to remain parasitemic on day 3 after ART treatment. Moreover, parasites with these K13 mutations also had ∼8 times higher levels of ring-stage survival than those with wild-type K13. Furthermore, K13 propeller mutations could even resolve the discrepancy between the RSA result and day 3 parasitemia data. Altogether, our study supports the use of RSA and K13 genotyping as complementary methods for epidemiological assessment of ART resistance.

At the China-Myanmar border, parasites with the K13 mutations have reached 66.7%. F446I is the most prevalent point mutation in this area (26, 27), and it is also prevalent in northern Myanmar (25). In southwestern China, this mutation was detected in significantly more isolates in day 3 parasite-positive groups and was associated with delayed parasite clearance (26). The distinct patterns of K13 mutations in the GMS suggest their independent emergence, which might reflect different drug histories. Haplotype network analysis also indicated a potential spread of parasites carrying the F446I allele in this region. Another point mutation, N458Y, detected earlier in Cambodia and Myanmar (6, 12, 25), was significantly associated with prolonged parasite clearance half-lives in clinical studies (6). Remarkably, a single isolate from our samples with N458Y showed a >60% ring-stage survival rate, whereas the median was 3.77% in all parasites carrying K13 propeller mutations. While most K13 genotyping studies so far focused on the propeller domain, our analysis of the full-length K13 sequences led to another intriguing finding. An NN insert in the N terminus of K13 protein was present in all day 3 parasite-positive isolates and was associated with increased ring survival. Since F446I was detected only in parasites carrying the NN insert, it is possible that the NN insert acts cooperatively with F446I in conferring ART resistance or is the result of hitchhiking with the spread of F446I. In Cambodia, the increased prevalence of K13 propeller mutations, especially C580Y, closely mirrored the emergence and spread of ART resistance (12). In parallel, significant increases in the prevalence of K13 propeller mutations in the last 7 to 8 years also suggest the emergence of ART resistance at the China-Myanmar border (27).

Both in vitro drug assay and genotyping of seven genes associated with drug resistance suggested the presence of multidrug-resistant parasites at the China-Myanmar border. Although standard in vitro assays of the 44 culture-adapted parasites did not detect significant differences in sensitivity between day 3 parasite-positive and -negative parasite groups to most drugs, 34.2% of the parasites showed higher IC₅₀s than 3D7 to all drugs tested. In addition, the range of IC₅₀s to several drugs was relatively wide, indicating the presence of parasite isolates with potential resistance to these drugs. Positive correlation was detected between several aminoalcohol drugs, suggesting a shared mechanism of resistance. It is noteworthy that the lack of correlation between the DHA RSA and ATS IC₅₀ results is arguably counterintuitive, but it highlights the radical difference in the principles of the two assays. This study provided further support for the use of in vitro RSA to capture the in vivo delayed parasite clearance phenotype. Despite reduced P. falciparum prevalence in recent years, parasites still displayed extraordinarily high haplotype diversity at the seven genes analyzed. Genotypes of resistance genes also indicated high resistance to 4-aminoquinolines, antifolates, and quinine. In particular, 44.7% of parasites harbored haplotypes suggestive of resistance to multiple drugs. Thus, the evidence presented here showing the emergence of ART resistance is worrisome. Since day 3 parasitemia as a proxy measure of parasite clearance rate is imprecise and can be influenced by factors such as the initial parasite densities, host immunity, and drug pharmacokinetics in individual patients, the validation of a phenotypic assay (RSA) and K13 polymorphisms as predictive markers of ART resistance offers more convenient surveillance tools for monitoring drug resistance in P. falciparum.