Recovery and stable persistence of chloroquine sensitivity in Plasmodium falciparum parasites after its discontinued use in Northern Uganda

Background Usage of chloroquine was discontinued from the treatment of Plasmodium falciparum infection in almost all endemic regions because of global spread of resistant parasites. Since the first report in Malawi, numerous epidemiological studies have demonstrated that the discontinuance led to re-emergence of chloroquine-susceptible P. falciparum, suggesting a possible role in future malaria control. However, most studies were cross-sectional, with few studies looking at the persistence of chloroquine recovery in long term. This study fills the gap by providing, for a period of at least 6 years, proof of persistent re-emergence/stable recovery of susceptible parasite populations using both molecular and phenotypic methods. Methods Ex vivo drug-susceptibility assays to chloroquine (n = 319) and lumefantrine (n = 335) were performed from 2013 to 2018 in Gulu, Northern Uganda, where chloroquine had been removed from the official malaria treatment regimen since 2006. Genotyping of pfcrt and pfmdr1 was also performed. Results Chloroquine resistance (≥ 100 nM) was observed in only 3 (1.3%) samples. Average IC50 values for chloroquine were persistently low throughout the study period (17.4–24.9 nM). Parasites harbouring pfcrt K76 alleles showed significantly lower IC50s to chloroquine than the parasites harbouring K76T alleles (21.4 nM vs. 43.1 nM, p-value = 3.9 × 10−8). Prevalence of K76 alleles gradually increased from 71% in 2013 to 100% in 2018. Conclusion This study found evidence of stable persistence of chloroquine susceptibility with the fixation of pfcrt K76 in Northern Uganda after discontinuation of chloroquine in the region. Accumulation of similar evidence in other endemic areas in Uganda could open channels for possible future re-use of chloroquine as an option for malaria treatment or prevention.

of P. falciparum parasites resistant to chloroquine in Southeast Asia and South America. The resistant parasites that first emerged in Southeast Asia spread to East Africa (Tanzania and Kenya) by 1980 [1,2] and eventually across the malaria endemic regions of Africa [3]. Chloroquine was, therefore, withdrawn/discontinued for routine treatment of P. falciparum malaria in nearly all malaria endemic regions. However, with widespread discontinued use, numerous molecular-epidemiological studies showed that there was return of chloroquine susceptibility in P. falciparum field isolates [4]. This is supported by ex vivo [5][6][7][8][9][10][11][12][13][14][15][16] and in vivo drug-susceptibility studies [5,17,18]. Findings suggest that chloroquine might be re-used in the future as an option for the treatment and/or chemoprophylaxis on the condition that chloroquine sensitivity is maintained in the area. As the parasite is "skilled" in evading anti-malarial treatments, continuous surveillance on longitudinal persistence of chloroquine susceptibility by molecular and phenotypic analysis [8,16,18,19] is needed.
In Uganda, first-line treatment for uncomplicated malaria was changed from chloroquine to chloroquine plus sulfadoxine/pyrimethamine in 2000, then again changed to artemether-lumefantrine in 2006 [20]. Several studies in different regions in Uganda reported high prevalence of ex vivo chloroquine resistant parasites (IC 50 s ≥ 100 nM) accompanied with high prevalence of lysine to threonine change at position 76 (K76T) in pfcrt [21][22][23][24]. However, a recent study showed recovery of chloroquine susceptibility in Eastern Uganda, Tororo; average IC 50 s decreased from 248 nM in 2010-2013 to 33 nM in the community and 57 nM in the hospital setting in 2016 [10]. Here, to investigate whether chloroquine sensitivity also recovered in other regions in Uganda and, if so, to examine the persistence of chloroquine sensitivity, ex vivo drug susceptibility studies for a 6-year period since 2013 in Gulu, Northern Uganda were performed. Results show that chloroquine susceptibility stably persisted during the study period with a significant decrease and eventual absence of the chloroquine resistant K76T alleles in pfcrt.

Study site
A comprehensive drug susceptibility assessment was conducted at St. Mary's Hospital Lacor in Gulu, Northern Uganda (Fig. 1) from 2013 to 2018 [25,26] Average temperature in the studied area is 24.6 °C and average annual rainfall is about 1507 mm with two rainy seasons; a smaller peak in April-May (average rainfall 150 mm) and a heavier peak in August-September (average rainfall 234 mm) [27]. Plasmodium falciparum is the most prevalent species and is mainly transmitted by Anopheles funestus and Anopheles gambiae as the major vectors.
Malaria control programmes in the studied region include vector control by long-lasting insecticidal nets (LLINs) and indoor residual spraying (IRS), artemisininbased combination therapy (ACT) together with improved diagnosis, management of severe malaria, and intermittent preventive treatment of malaria during pregnancy. These control measures were performed with funding from the Global Fund, USAID/PMI, DFID, World Vision and other partners [28]. Mass distribution of LLINs was first implemented in 2009-2010, which continued until 2013-2014. Through these extensive efforts, malaria burden was effectively reduced from 72% in 2009 to 29% in 2014 [27].

Patients
Initial screening was performed using RDT (SD BIOLINE Malaria Ag P.f/Pan test, Abbott, USA) for 1575 symptomatic patients who visited St. Mary's Hospital Lacor. Inclusion criteria were: (a) patients that are P. falciparum positive by RDT and microscopy, (b) aged ≥ 6 months, and (c) with no history of taking anti-malarial drug(s) within 2 weeks before enrollment. Patients fulfilling the inclusion criteria were enrolled after obtaining written informed consent from the patients or parent/ guardian(s). For children aged 7 to 17 years, separate assent was also obtained. Ethical

Sample collection and ex vivo susceptibility assay for chloroquine and lumefantrine
Blood samples of approximately 100-500 µL (< 2 years old), and 1 mL (≥ 2 years old) were collected by peripheral venipuncture or finger prick and immediately transferred to the laboratory adjacent to the hospital. Thick and thin blood smears stained for 30 min with 2% Giemsa solution were used to determine parasitaemia.
At every visit from 2013 to 2018 (a total of nine times sampling period), ex vivo drug susceptibility studies were performed. Ex vivo susceptibility was evaluated for chloroquine and lumefantrine for the samples with parasitemia ≥ 0.05% as previously reported [29]. Parasite culture was incubated in the presence of chloroquine (25-1600 nM) or lumefantrine (1.25-80 nM) at 37 °C for 72 h in a gas atmosphere of 5% CO 2 , 5% O 2 (Anaer-oPack malaria culture system, Mitsubishi Gas Chemical Co. Inc., Tokyo, Japan). Laboratory-maintained 3D7 clone was used for quality evaluation of pre-dosed drug plates. Parasite culture without antimalarials served as control. To evaluate parasite growth, thick smears were made from drug free culture after 72 h of incubation and the number of schizonts counted. If less than 5 schizonts were seen per field, the test samples included in that plate were not used for further analysis. Drug sensitivity was assessed using an enzyme-linked immunosorbent assay (ELISA) that quantifies parasite histidine-rich protein-2 (HRP-2) [30]. The effective concentration needed to inhibit P. falciparum growth by 50% (IC 50 ) was determined by non-linear regression using an online ICEstimator software (http://www.antim alari al-icest imato r.net) [31]. The quality of the ex vivo drug assay was evaluated based upon the level of fitness to the expected shape of the curve obtained by the inhibitory sigmoid Emax model [31].

pfcrt and pfmdr1 genotyping
Polymorphisms at amino acid positions 72-76 in the P. falciparum chloroquine resistance transporter gene (pfcrt) were determined by direct sequencing. In the P. falciparum multidrug resistance-1 (pfmdr1) gene, polymorphisms at codons 86, 184, 1034, 1042 and 1246 were determined by direct sequencing and/or restriction fragment length polymorphism (RFLP) analysis, as previously described [29,32]. For direct sequencing, initial and nested PCR was done with PrimeSTAR Max DNA Polymerase (Takara Bio Inc., Japan) in 10 μL

Full sequencing of pfcrt
The full sequence of pfcrt was obtained either by whole-genome sequencing (n = 17) or target sequencing (n = 39). Whole genome sequence data was previously reported [26]. In brief, Acrodisc filters (Pall Corporation, New York, NY, USA) was used to reduce the extent of human DNA contamination from blood samples. Approximately 1-1.5 Gb of data per sample were obtained using Illumina instruments (Miseq and Hiseq 2000). Single-nucleotide polymorphisms were called at all genomic positions with > 80% frequency of > 10 reads support. For target sequencing, the DNA fragment of a genomic region coding for pfcrt gene was amplified by PCR with primers (Pfcrt-F: 5′-TAC TTT CCC AAG TTG TAC TGC TTC TAA GCT-3′, Pfcrt-R: 5′-TTT ACC TAT TTA TCA AAA CAC CAA AAG GGA-3′), which covers the whole DNA sequence of pfcrt gene. PCR was performed with PrimeSTAR GXL DNA Polymerase (Takara Bio Inc., Japan) in 5 μL reaction mixture containing 1 μL of DNA solution and 0.25 μM of primer set. PCR conditions consisted of denaturation at 98 °C for 10 s, followed by 40 cycles of amplification (98 °C for 10 s, 60 °C for 15 s, and 68 °C for 5 min), with a final elongation period of 68 °C for 5 min. PCR products were diluted with 5 μL of pure water, electrophoresed in 2% agarose gel and stained with ethidium bromide. The PCR products were then purified with the ExoSAP-IT reagent (Affymetrix, USA). Libraries were prepared from the purified PCR products with Nextera XT DNA Library Prep Kit (Illumina, USA). The libraries were sequenced by MiSeq (Illumina) with the paired-end method and 250 bp of read length. The reads were also used to map the pfcrt gene sequence of P. falciparum 3D7 as a reference and assembled a single contiguous sequence by CLC Genomics Workbench (Qiagen). All sequences were deposited in the DNA Data Bank of Japan (DDBJ) with accession numbers LC498195-LC498250.

Statistical analysis
All statistical analyses were performed using R software (version 3.6.1). Data was analyzed using Kruskall Wallis test, Wilcoxon rank sum test, and Jonckheere-Terpstra test. p values < 0.05 were considered statistically significant.

Ex vivo drug susceptibility of chloroquine and lumefantrine
Of 1575 patients who visited St. Mary's Hospital Lacor, 793 patients were enrolled based on P. falciparumpositive results by RDT (Fig. 2). The rest were excluded because of (a) absence of P. falciparum by microscopic examination (n = 535), (b) use of anti-malarial drug(s) within the last 2 weeks before enrollment (n = 198), or (c) other reasons (n = 49) (Fig. 2). The commonly used antimalarial for pretreatment was artemether-lumefantrine (77%) ( Table 1). Chloroquine use was confirmed in only 3 patients in 2013 and one patient in 2014.
Among 793 blood samples obtained from enrolled patients, 203 were excluded because of very low parasitaemia (< 0.05%) or insufficient amount of blood, resulting in 590 samples used for ex vivo drug susceptibility assays. Also, ex vivo study for chloroquine in 2016 and lumefantrine in 2015 was not performed because of inadequate quality of pre-dosed drug plates. Thus, in total, 319 and 335 ex vivo drug-susceptibility assays for chloroquine and lumefantrine respectively were available for analyses. Background information of patients who participated in the study per year is shown in Table 2. Median age was 3.5 years (IQR 2.0-4.8) and hemoglobin level < 10 g/dL was observed in 27% of patients. Median parasitemia at enrollment were 0.2-3.5%, which significantly varied between studied years (p-value = 3.9 × 10 −15 , Kruskal-Wallis test). Beside parasitaemia at enrollment, no significant difference was observed in background factors among the studied years.
Of 319 and 335 ex vivo drug-susceptibility assay for chloroquine and lumefantrine, respectively, 42 chloroquine and 48 lumefantrine assays were performed with O instead of AB blood-group serum due to the latter's unavailability during the sampling period. These samples were excluded from further analysis. Thus, in summary, ex vivo drug study was successfully conducted for 239/277 samples (86.3%) for chloroquine and 168/287 for lumefantrine (58.5%) (Fig. 2). For chloroquine, only 1.3% (3/239) fulfilled the criteria for chloroquine resistance (IC 50 > 100 nM) (Fig. 3a). From 2013 to 2018, the geometric means of the IC 50 s (17.4-24.9 nM) were much lower than the threshold for chloroquine resistance and was stable without any significant decrease 782 Excluded. 535   . In all studied parasites, IC 50 values were below the conservative cutoff of 50 nM for lumefantrine resistance [33], and lower than the 150 nM value [34] (Fig. 3b).

Allele prevalence and frequency of pfcrt and pfmdr1 polymorphisms
In pfcrt, the prevalence of chloroquine-resistant alleles (CVIET allele; amino acid position 72-76, mutation underlined) significantly decreased from 28.8% in 2013 to 1.1% in 2016 and finally could not be detected in 2017 (Fig. 4). Besides genome sequencing, the absence of minor alleles bearing CVIET was further confirmed by pfcrt target sequencing. In pfmdr1, chloroquine sensitive N86 allele was fixed or nearly fixed throughout the study period. Prevalence of the mutant allele at position 184 (Y184F) gradually increased from 2.4% in 2013 to 48.5% in 2018, albeit this trend was not significant (p-value = 0.13, Jonckheere-Terpstra trend test). Wildtype alleles were nearly fixed at other loci in pfmdr1.

Association between ex vivo drug sensitivity and alleles in pfcrt and pfmdr1
In pfcrt, parasites carrying wild-type alleles showed significantly lower IC 50 s to chloroquine than those carrying mutant allele (geometric mean, 21.4 vs. 43.1 nM, p-value 3.9 × 10 −8 , Wilcoxon rank sum test) (Fig. 5). To see whether other polymorphism(s) besides those at position 72-76 played a role in the recovery of chloroquine susceptibility, entire sequences of pfcrt in 56 samples were analysed. IC 50 s were also successfully obtained in 44 samples (Table 3), in which 31 (71%) bore only wild-type alleles of the gene. The second (n = 5, 11%) most common haplotype, HP-4, corresponds to a prevalent mutant haplotype (CVMNT + A220S + Q271E + R371I) in Africa [3,35]. Remaining eight parasites harbored minor haplotypes, all with wild type allele K76 and displayed IC 50s to chloroquine of 18-35 nM. A previous study (16), has implicated C356R in the sensitivity recovery of chloroquine in the K76T-harbouring parasites, however this mutant allele was not found in the study area. These results further confirm the expansion of wild K76-harbouring parasites as the cause of chloroquine susceptibility reversal in the studied area, rather than additional mutation in the pfcrt gene. Association between ex vivo drug sensitivity and respective pfmdr1 alleles, however, could not be properly  (Fig. 5). Analysis of susceptibility to lumefantrine indicated that parasites carrying pfcrt CVIET exhibited significantly lower IC 50 s than those with CVMNK (13.4 vs. 28.0 nM, p-value = 8.1 × 10 −5 , Wilcoxon rank sum test) (Fig. 5).
No significant difference in IC 50 s for lumefantrine was found between parasites with pfmdr1 Y184 (25.0 nM) and Y184F (30.0 nM).

Discussion
In Uganda, chloroquine was officially withdrawn and replaced with artemether-lumefantrine in 2006. The present analysis revealed that chloroquine susceptibility has returned and is stably maintained for at least 6-years in Gulu, Northern Uganda. This is the first report in Uganda to demonstrate stable and persistent recovery of chloroquine sensitivity using both phenotypic and genotypic approaches. The prevalence of K76 allele in pfcrt rapidly increased from 67% in 2013 to complete fixation in 2017. This is most probably due to K76-harbouring parasites outcompeting the K76T-harbouring parasites because of fitness advantage in the absence of chloroquine selection pressure, as previously observed in Malawi [36,37]. Indeed, recent transfection studies also show that K76T confers a substantial fitness cost to parasites [35,38]. This fitness cost can be partly explained by a functional impairment in haemoglobin digestion and subsequent reduction in the supply of amino acids in K76T harbouring parasites [39]. In another possible scenario, a back mutation from T to K at position 76 in pfcrt may potentially induce chloroquine sensitivity. As an example, the chloroquinesusceptible 106/1 clone had mutant alleles at positions 74 and 75 but showed wild-type K allele at position 76 (CVIEK) [40]. The only difference in the pfcrt haplotype at position 72-76 in this clone and the widely prevalent chloroquine-resistant haplotype (CVIET) is at amino acid position 76. In the present analysis, however, no such haplotype (CVIEK) was found, negating this possibility. Also, no evidence of additional mutation in pfcrt such as C350R which has been reported to be associated with the restoration of chloroquine susceptibility [16] was obtained. Taken together, these results strongly suggest that neither back mutation nor additional mutations  in pfcrt was associated with the observed recovery of chloroquine sensitivity in the study area. It is of note that recovery of chloroquine sensitivity after its withdrawal occurred much earlier in Gulu than in other regions in Uganda [9,[22][23][24]41]. In 2013, as much as 65% of parasites displayed ex vivo chloroquine resistance [24] and 60-80% carried K76T allele in Tororo, Eastern Uganda [42]. In contrast, present results revealed that prevalence of ex vivo chloroquine-resistance and K76T alleles were already 6% and 29%, respectively, in 2013, indicating a faster recovery or re-emergence of chloroquine sensitive strains in the region. Despite the government's move to change the national treatment policy, chloroquine might be used as self-treatment and/ or prophylaxis. Such chloroquine use potentially creates various levels of chloroquine selecting pressure in the region, which would be one of the important factors that affects the speed of recovery of susceptible parasites [43]. However, countrywide anti-malarial survey reported no considerable difference of chloroquine use between Tororo and this study area [44][45][46], suggesting that this could not be a main factor for the observed findings.
Piperaquine use in Tororo might explain the observed difference to some extent. Dihydroartemisinin-piperaquine has been used as a second line treatment for uncomplicated malaria in Uganda. In Tororo, this regimen was widely used in different drug trials for malaria treatment [47,48] and chemoprevention [24,[49][50][51][52]. Previous studies indicated that dihydroartemisinin-piperaquine treatment selected the N86Y allele in pfmdr1 in Uganda [48,[52][53][54], albeit one study questioned this association [55]. Approximately one-third of parasites possessed the N86Y in 2010-2013 in Tororo [9], which was much higher than that found in the studied area (2%). Since N86Y is associated with chloroquine resistance, it might be plausible that the N86Y mutation selected by piperaquine played a role in the slower recovery of chloroquine sensitivity in Tororo.
In contrast to chloroquine, average lumefantrine IC 50 values in the present analysis (21-29 nM) were considerably higher than those in Eastern Uganda (3.0-5.4 nM) [9,24]. It has been suggested that decrease in lumefantrine susceptibility are associated with wild type alleles in pfcrt and pfmdr1 [48,[56][57][58][59]. In this study higher IC 50 s to lumefantrine were observed in K76 sequences than those with K76T. The higher prevalence of K76 allele in our study area than that in Eastern Uganda [9,23,42] may partly explain the observed lumefantrine susceptibility.
Molecular epidemiological analysis displayed a large increase in the proportion of parasites carrying the Y184F mutation in pfmdr1. This occurred at the same time when substantial reduction on the pfcrt K76T mutation was observed, particularly between 2014 and 2015.
One possible explanation would be an increase in lumefantrine usage results in the selection of these alleles in this area. Indeed, previous in vivo studies have demonstrated that artemether-lumefantrine treatment selected for these alleles (pfcrt K76 and Y184F) in Africa including Uganda [24,60,61]. In the present study, IC 50 s for lumefantrine were significantly higher in pfcrt K76 harboring parasites than those harboring K76T; while, Y184F allele did not show significantly high IC 50 s for lumefantrine, consistent with previous transfection study that revealed no association of Y184F mutation to in vitro lumefantrine susceptibility [59]. Thus, in vivo selection for Y184F allele after artemether-lumefantrine treatment may be because of mechanisms other than susceptibility to lumefantrine.

Conclusions
The study shows the stable persistence of chloroquinesusceptibility with the fixation of pfcrt K76 in Northern Uganda. This observation implies the possibility of future clinical trials for potential re-use of chloroquine as an option for malaria treatment or prevention. Such trial was performed in Malawi where longtime stable chloroquine susceptibility has been evidenced and has revealed that weekly chemoprophylaxis with chloroquine showed 78% lower risk of clinical malaria than intermittent sulfadoxine-pyrimethamine [62]. Similar trial in Uganda would provide an insight into potential re-introduction of chloroquine. However, further evidence of long-time persistence of return of chloroquine susceptibility is warranted in other endemic areas in Uganda before the implementation of clinical trials.