Decreased In Vitro Artemisinin Sensitivity of Plasmodium falciparum across India

Artemisinin-based combination therapy (ACT) has been used to treat uncomplicated Plasmodium falciparum infections in India since 2004. Since 2008, a decrease in artemisinin effectiveness has been seen throughout the Greater Mekong Subregion. The geographic proximity and ecological similarities of northeastern India to Southeast Asia may differentially affect the long-term management and sustainability of ACT in India.

unknown (16)(17)(18)(19). A continuing challenge for India is to determine whether the decreased effectiveness of the artesunate-sulfadoxine-pyrimethamine (AS-SP) regimen is restricted to the east and northeast regions or whether it can also be found in other parts of the country and if reduced effectiveness exists against other ACT regimens.
The gold standard of ACT resistance is delayed in vivo parasite clearance after drug treatment, ideally assessed by 42 or 63 days of clinical follow-up. Such longitudinal clinical evaluations can be challenging. In this respect, the method of in vitro RSA (RSA 0 -3hr ) (20) offers some advantages; a sample drawn at a single time point per patient and cryopreserved can be tested in vitro, at a later time, in a specialized malaria lab (18).
In addition to high RSA 0 -3hr values, some mutations in P. falciparum kelch (Pfkelch; Pf3D7_1343700) have been associated with decreased in vivo parasite clearance rates (20,21). Previously published sequence data for Indian isolates in the east and northeast regions show mutations in the kelch gene; however, the ability of kelch mutations to predict decreased ACT efficacy in Indian parasites has been inconsistent (22)(23)(24).
In the present study, clinical P. falciparum samples from two geographically distant settings in India, one in southwest (SW) India (Goa State) and one in northeast India (Assam, Arunachal Pradesh, and Tripura States), were studied. Artemisinin sensitivity was assessed in vitro using RSA 0 -3hr , and parasite DNA sequences at the kelch locus were studied. (

RESULTS
Demography and clinical history of Indian samples. The samples from northeast India came from the border states of Assam, Arunachal Pradesh, and Tripura. The samples from southwest India were collected in the state of Goa ( Fig. 1 and Table 1). The demographics of local inhabitants from Assam, Arunachal Pradesh, and Tripura constitute the demographics of northeast India, whereas the samples from the southwest were collected from residents of Goa as well as individuals who had recently moved from the neighboring states of Karnataka and Maharashtra to Goa. In both regions, northeast and southwest, transmission undergoes a seasonal peak during the monsoons and continues at a reduced intensity throughout the year.
Culture-adapted samples from the northeast (NE; n ϭ 12) and southwest (SW; n ϭ 10) were used for RSA (Table 1). Both study groups (NE and SW) had similar proportions of severe malaria cases. The treatment regimen at the time of sample collection comprised artemether-lumefantrine (AM-LF) in northeast India and artesunate-mefloquine (AS-MQ) in southwest India. The AS-MQ regimen prescribed in the Goa Medical College and Hospital (GMC) during the study period was different from the standard artesunate-sulfadoxine-pyrimethamine (AS-SP) regimen used in all parts of India except the northeast states (25). The ACT regimen in the northeast states was changed from AS-SP to AM-LF in 2013 (25) due to the sulfadoxine-pyrimethamine (SP) resistance reported in this region (26)(27)(28).
Classification of morphological changes in the DHA-treated parasites. Ringstage parasites from southwest and northeast India treated with dihydroartemisinin (DHA) for 6 h exhibited distinct phenotypes ( Fig. 2A). Viable rings, along with parasites that displayed an arrested ring morphology, were observed. The arrested ring-stage parasites had a round morphology with distinct dark red/pink-stained chromatin and a light blue-stained cytoplasm, similar to the findings for dormant parasites described in previous reports (29,30). The majority of treated parasites were pyknotic, as defined by collapsed nuclei that stained dark purplish red with Giemsa and no significant cytoplasm. Other nonviable parasites showed a ring-like chromatin morphology but lacked a regular cytoplasm. Parasites with another deformed phenotype included parasites that were morphologically similar to mature stages but that lacked a distinct chromatin organization. Viable parasites in the DHA-exposed cultures exhibited an arrested ring or dormant morphology, along with small populations of viable rings (Fig. 2B). The proportion of viable parasites among DHA-treated southwest isolates (median ϭ 2.1%) was lower than that among the Cambodian resistant lines (median ϭ 36.7%). Within the two Indian groups, the northeast isolates, on average, had a higher proportion of viable parasites (median ϭ 4.3%).
Ring-stage survival after DHA exposure. The tolerance to artemisinin in the cultures exposed to DHA for 72 h varied considerably between the two groups of Indian parasites (Fig. 3). The RSA 0 -3hr survival rate of parasites from the northeast (median, 1.7%) was three times higher than that of parasites from the southwest (median, 0.4%). In total, 8 of 12 (66%) northeast isolates and 3 of 10 (30%) southwest isolates had a survival rate of greater than 1%. The 1% or higher RSA 0 -3hr survival rate is considered indicative of artemisinin tolerance (10). Both Indian groups had a lower survival rate than the artemisinin-resistant controls from Cambodia obtained from BEI Resources (median survival rate, 17.1%; range, 3.3% to 31.9%). Parasite strains in the negative- control group (artemisinin-sensitive strains) had a survival rate range between 0% and 1.7%, with the median survival rate being 0.5%. This group comprised the BEI Resources artemisinin-sensitive Cambodian strains and standard laboratory reference lines 3D7 and Dd2.
Compared to the survival rates for the controls, three northeast Indian isolates had survival rates above 3.4%, which was the 0th percentile (P0) or lowest survival rate of the Cambodian resistant strains. No southwest Indian isolate reached this threshold. Six northeast isolates and two southwest isolates had survival rates higher than 1.8%, the 100th percentile (P100) for the negative controls. The Cambodian strains with verified artemisinin resistance status served not only as reference points for survival rate comparison but also as internal controls for the RSA experiment itself under our test conditions. A statistically significant difference (P ϭ 0.0025, Kruskal-Wallis test for the location of sample collection variable) was noted between the three groups of northeast and southwest Indian isolates and the Cambodian positive controls. The southwest isolates were statistically significantly different from both the northeast isolates (P ϭ 0.0208) and the Cambodian artemisinin-resistant strains (P ϭ 0.0002). In comparison, northeast isolates and resistant strains from Cambodia were statistically significantly similar (P ϭ 0.0994). Statistical analysis of the RSA 0 -3hr survival rate by two microscopists showed good agreement (Pearson correlation r ϭ 0.75, P Ͻ 0.0001).
Molecular characterization of kelch from northeast and southwest Indian isolates. DNA sequencing of the Indian isolates revealed two types of mutations in the kelch gene, a common insertion and a nonsynonymous mutation. The insertion of 6 nucleotides (coding for Asn-Asn [NN]) was observed in patient isolates SW24 and SW46, as well as patient isolates NE10, NE16, NE20, NE27, NE33, NE39, NE45, and NE53, at nucleotide position 407 (between codons 142 and 143) in the BTB/POZ (broadcomplex, Tramtrack, and Bric-à-brac/poxvirus and zinc finger) domain of Kelch (Fig. 4A) (GenBank accession number MK949521). The contribution of this insertion to artemisinin resistance is unknown. However, it is notable that this insertion was relatively frequent in Indian isolates (unpublished data). It was more prevalent in samples from northeast India (66%) than in those from southwest India (20%). Out of 11 samples  showing a Ͼ1% survival rate, 7 (63%) had this insertion. Among the 10 samples that had the insertion, 3 (30%) had a survival rate lower than 1%. The only nonsynonymous mutation in the Kelch propeller domain was noted in patient isolate NE53, in which an alanine-to-valine mutation at position 675 (A675V) was seen (Fig. 4B) (GenBank accession number MK949522). The corresponding RSA survival rate was 2%.

DISCUSSION
Altered artemisinin sensitivity in Indian P. falciparum isolates. The RSA 0 -3hr survival rate of Indian isolates varied widely (0.2% to 8.1%), reflecting vastly different in vitro sensitivities to artemisinin (Table 1). These survival rates were similar to recently reported survival rates from a treatment study in east India (14,15) and between the survival rates of Southeast Asian and African isolates reported elsewhere (10,20,(31)(32)(33). None of the Indian groups exhibited consistently high survival rates like the isolates from Pailin, Cambodia (median, 14.9%), where artemisinin resistance is entrenched (10). However, they also do not resemble the survival rates of Ugandan and Gambian isolates from the same period when the Indian isolates were recovered. The Ugandan and Gambian isolates from that period showed consistently low survival rates (0% to 1%) (31, 34), though two very recent studies have reported elevated RSA 0 -3hr survival rates (0.2% to 34.3%) in African parasites (32,35). In the future, recording of in vivo parasite clearance data for Indian and African isolates, alongside in vitro RSA survival data, will had ring-like structures, while others were extracellular or devoid of regular cytoplasm. (B) The relative proportion of different morphological stages in the treated parasite population show that a higher proportion of viable parasites was found among the isolates from the northeast than among the isolates from the southwest. However, both these groups had a much lower proportion of viable parasites than the positive controls (Cambodian artemisinin-resistant isolates). further help define the best ways to assess the status of artemisinin sensitivity in these regions.
The RSA 0 -3hr survival rate has been a reliable in vitro indicator of the clinical correlate of artemisinin resistance measured by parasite clearance after ACT treatment (10,36). The RSA 0 -3hr threshold of 1% was originally promoted on the basis of the corresponding parasite clearance half-life data for Cambodian isolates. Isolates with a clearance half-life below the threshold of 5 h were considered fast clearing, and those with a clearance half-life above that threshold were considered slow clearing (20). A 1% survival rate has since been adopted as the RSA threshold in subsequent in vitro resistance studies across different study sites in Southeast Asia and Africa. With an initial 77% accuracy rate in identifying slowly clearing parasites (10), the 1% threshold can capture the bulk of resistant isolates. In the absence of large-scale parasite clearance data collected from every region, which is needed in India, it seemed logical to apply a second, empirically derived threshold level to obtain a more conservative estimate of resistant isolates.
Two empirical thresholds were considered for the current study: the first was the 100th percentile (P100) for the negative control, which considers the highest RSA value for negative controls (1.8%) to be the threshold. The second threshold considered was the 0th percentile (P0) for positive controls, i.e., the lowest RSA value for positive controls (3.4%). Considering these most conservative estimates of threshold values, at least 3 northeast Indian isolates (NE17, NE20, and NE45) out of a total of 22 (13%) could be categorized as having a decreased artemisinin sensitivity phenotype. RSA 0 -3hr survival rate cutoffs like these, based on actual experimental data, may also be a better indicator of region-specific artemisinin tolerance trends than the generic 1% threshold.
Decreasing artemisinin sensitivity in India without kelch mutations. The single nonsynonymous mutation, A675V, found in Kelch in the isolate from one patient in this study has been considered a candidate marker for artemisinin resistance by WHO (1). It is reported here from India for the first time. A675V was previously seen in six different areas of the Greater Mekong Subregion and in east Africa: on the Indonesia-Myanmar border (37), in southern Myanmar (38), at the China-Myanmar and Thailand-Myanmar borders (33,39), in Rwanda (40), and in Uganda (35). The mutation was associated with slowly clearing clinical infections in Thailand (39) and delayed in vitro parasite clearance in Southeast Asian samples (30). However, A675V was also found in an artemisinin-sensitive clinical sample in Myanmar (38), suggesting that this mutation in itself cannot be the sole cause of artemisinin tolerance. In the current study from India, no significant association between the RSA 0 -3hr rate survival and kelch mutations was observed ( Table 1). The only Kelch mutation seen, A675V in sample NE53, was associated with a 2% survival rate. However, for all other samples with a Ͼ1% RSA 0 -3hr survival rate, there were no kelch mutations. Interestingly, such discordance (albeit in a smaller proportion) was noted in culture-adapted western Cambodian isolates, with kelch mutations being absent in about 11% of the isolates exhibiting a Ͼ1% RSA 0 -3hr survival rate (41).
The Asn-Asn (NN) insert between codons 142 and 143 of kelch reported here is the first of its kind seen in India (Table 1). However, previous studies have reported one or two NN insertions at codon 142 of kelch in isolates from Senegal (42) and in Cambodia (41) and between codons 136 and 137 of kelch in isolates recovered on the China-Myanmar border (43). The NN kelch insert between codons 142 and 143 was observed in 45% of the Indian samples, and it was most prevalent in the northeast isolates (66%). Within the northeast India group of isolates, the NN insert was associated with seven out of eight samples with a Ͼ1% RSA survival rate. None of the southwest India samples with this insert (20%) had a Ͼ1% survival rate. Although these findings are interesting, on the basis of the number of samples with the NN insert between codons 142 and 143 in this study, it is not possible to infer any association of this insert with changes in artemisinin sensitivity.
Overall, the molecular data presented in this study from India and the historical prevalence of kelch polymorphisms suggest that the effectiveness of the kelch locus as a marker for artemisinin resistance surveillance, by itself, is not robust. This is backed by observations in western Myanmar, where the proportion of samples positive on day 3 was much less than the proportion of isolates with K13 propeller mutations (37). Even in the lower Mekong region, where the proportion of patients positive on day 3 broadly matches that of isolates with K13 propeller mutations, only certain Kelch propeller domain mutations are associated with higher RSA 0 -3hr survival rates (44). Mutations in the Kelch propeller domain do not always confer reduced clinical sensitivity to artemisinin (33), and links of kelch mutations to artemisinin resistance are dependent on the larger population structure, as detailed by the MalariaGEN Plasmodium falciparum Community Project (45).
Conclusion. The current WHO definition of artemisinin resistance rests on two primary indicators: (i) a high prevalence of isolates with a delayed parasite clearance phenotype in a population and (ii) a high prevalence of kelch gene mutants (Ͼ10% of the population). A confirmed case of artemisinin resistance involves both these components in the same patient (4). In a large and populous country like India, with a high absolute number of cases of malaria but a low prevalence due to the large denominator, considerable resources will be required to perform large-scale clinical assessments of artemisinin resistance across the country. Also, in a study like ours, the size of the sample set is limited by the number of successfully lab-adapted clinical isolates at a time point. So, while the current sample set provides an account of the in vitro artemisinin sensitivity in Indian isolates during the study period, it is underpowered to formally infer wider conclusions regarding possible artemisinin resistance in India.
However, the elevated RSA 0 -3hr survival rate and the concomitant presence of kelch mutations in this preliminary sample set, along with clinical reports of delayed parasite clearance from eastern India (14,15), point toward changing artemisinin sensitivity and warrant wider surveillance of the resistance-associated phenotype and genotype for artemisinin as well as partner drugs in this region.
Finally, our data suggest that the surveillance priority for artemisinin resistance in India should not be determined purely on the basis of the geographic proximity to Southeast Asia or kelch mutations alone. A countrywide surveillance of ACT efficacy that incorporates clinical, genotypic, and phenotypic indicators is needed to obtain a complete and accurate picture of possible decreasing artemisinin sensitivity throughout India. The results from the present study justify such strategies and investments.

MATERIALS AND METHODS
PicardTools (https://github.com/broadinstitute/picard) were used to sort the aligned reads and mark the duplicates, respectively. GATK tools were used to realign around indels and recalibrate the quality scores. Mpileup from Samtools (53) was then used to call variants from all samples taken together. This gave a single variant call format (VCF) file that contained all variants from the entire sample set. The SnpEff tool (54) was then used to add annotations to the VCF file. Paired-end 100-bp reads from the sequencing were aligned to the P. falciparum reference genome (PlasmoDB Pfv9). Custom R scripts that leveraged the VariantAnnotation package from the Bioconductor suite (55) were used to filter variants on the basis of mapping quality and read depth. Only those which had read depths of greater than 10 and a quality score above 100 were selected. Variants in intergenic regions and on var genes were removed. Filtered variants were analyzed using custom R scripts.
Data availability. Sequences have been deposited in GenBank under accession no. MK949521 and MK949522.