K13-propeller gene polymorphisms in Plasmodium falciparum parasite population in malaria affected countries: a systematic review of prevalence and risk factors

Background Efficacy of artemisinin (ART) agents, a critical element of current malaria control efforts is threatened by emergence and spread of resistance. Mutations in pfkelch13 gene associated with ART-resistance evolved in Southeast Asia (SEA). k13 mutations whose role in ART-resistance remains unknown, have subsequently emerged independently across all malaria-affected regions. The aim of this systematic review was to determine the prevalence and identify risk factors of Plasmodium falciparum k13 mutations in malaria-endemic countries. Methods An electronic search of studies from 2014 to date was done in MEDLINE via PubMED, SCOPUS, EMBASE and LILACS/VHL databases. Mesh terms and Boolean operators (AND, OR) were used. Two librarians independently conducted this search (RS and AK). The articles were screened for inclusion using a priori criteria set following PRISMA-P and STREGA guidelines. Three independent reviewers (NL, BB, and OM) extracted the data. Data analysis was performed in Open Meta Analyst software. Random effects analysis (DL) was used and heterogeneity established using I2-statistic. Results A total of 482 articles were retrieved from Pubmed = 302, Lilacs/Vhl = 50, Embase = 80, and Scopus = 37; Bibliography/other searches = 13, of which 374 did not meet the inclusion criteria. The aggregate prevalence of single nucleotide polymorphisms (SNPs) in pfkelch13 gene was 27.6% (3694/14,827) (95% CI 22.9%, 32.3%). Sub-group analysis showed that aggregate prevalence of non-synonymous SNPs in pfkelch13 gene was higher, 45.4% (95% CI 35.4%, 55.3%) in Southeast Asia as opposed to 7.6% (95% CI 5.6%, 9.5%) in the African region. A total of 165 independent k13 mutations were identified across malaria-affected regions globally. A total of 16 non-validated k13 mutations were associated with increased ART parasite clearance half-life (t1/2 > 5 h). The majority, 45.5% (75/165), of the mutations were reported in single P. falciparum parasite infections. Of the 165 k13-mutations, over half were reported as new alleles. Twenty (20) non-propeller mutations in the pfkelch13 gene were identified. Conclusion This review identified emergence of potential ART-resistance mediating k13 mutations in the African region. Diversity of mutations in pfkelch13 gene is highest in African region compared to SEA. Mutations outside the pfkelch13 propeller region associated with increased ART parasite clearance half-life occur in malaria-affected regions.


Background
The useful life of artemisinin (ART) agents in malaria treatment is at risk due to emergence of several resistance alleles in pfkelch13 gene (k13) [1]. This is further worsened by the current lack of effective vaccine and alternative medicines to ART agents in malaria management. Decreased Plasmodium falciparum sensitivity to ART agents is rapidly evolving and expanding globally since being first reported in southeast Asia (SEA) [2,3]. Widespread prevalence of parasite ARTresistance threatens current gains in malaria control and eradication efforts. This is especially the case as ART agents remain the only effective malaria treatment to date [4]. Diversity in pfkelch13 gene polymorphisms hinders efforts to track and contain the threat of parasite ART-resistance especially in high malaria transmission regions. In addition, the role of a majority of reported African k13 mutations in causing ART-resistance remains unknown [5].
An increase of 2 million malaria cases was observed in 2017 when 219 million cases were reported compared to 217 million in 2016 [4]. With malaria being one of the leading causes of morbidity and mortality especially in tropical countries, a further increase of an already high burden of malaria is a cause for concern. It is an indication of the need to rethink current malaria control efforts. With no widely effective vaccine, ART agents and their derivatives remain the only effective agents currently used in malaria control efforts. However, decreased P. falciparum parasite sensitivity that has emerged in SEA presents a threat to future efficacy of these agents.
Due to occurrence of multiple polymorphisms in pfkelch13 gene, there is a need to regularly keep track of various emerging alleles in different malaria-affected regions globally. In this systematic review, the aim was to describe potential emergence and/or spread of ARTresistance alleles in pfkelch13 gene across malariaaffected regions. The findings of this review are useful to malaria policy makers and research scientists as it provides a guide on potential k13 mutations whose role in ART-resistance needs to be urgently established, especially in malaria-endemic regions outside SEA.

Protocol development
In 2017, a review protocol (#CRD 42018084624) was developed and registered in International Prospective Register of Systematic Reviews (PROSPERO: http:// www.crd.york.ac.uk/prosp ero) and thereafter published it in peer review journal [6]. The recommendations of PRISMA-P [7] and STREGA guidelines [8] were observed during protocol development and review conduct.

Review question
The review intended to answer the question: What is the aggregate prevalence of pfkelch13 gene mutations in malaria-affected countries since change in malaria treatment policy to ART agents? [6].
Searching for articles in selected databases using the stated search terms was restricted to title or abstract. To maximize outcomes of the search, specific terms were not included such as those that focus on study design. The review found articles in French and Spanish, in which case Google translator (https ://trans late.googl e.com/) was used for translation to English prior to screening of the articles for inclusion.

Additional searches
Screening reference lists of included studies was a key targeted article search approach. Further, a researcher in malaria (PR) was contacted for recent publications on k13 gene polymorphisms in P. falciparum in different malaria-affected countries. There was no language restriction in article search and no search of grey literature.

Data management
EndNote software version X7 (Thomson Reuters, 2015) was used in data management for the review articles. All identified titles of articles were imported into the software and duplicates removed. Article titles were then screened and grouped into different eligibility categories (included or excluded) following a priori criteria. Data extraction form was developed in Excel 2007, pre-tested on 5 articles and adjusted using findings of the pilot test. Data were extracted in triplicate and kept using the adjusted tool.

Eligibility criteria for article selection
The review included articles from studies that investigated prevalence of K13-mutations among P. falciparum parasites in all malaria-affected regions/countries. Studies included were those that assessed for molecular markers of ART-resistance in pfkelch13 gene. Studies that reported both molecular and phenotypic ART-resistance were included. Additionally, studies that assessed for k13 mutations both before and after official introduction of ART combination therapy (ACT) in malaria treatment were included. Studies that analyzed for k13 gene mutations using sequencing techniques were included. All studies that used cross-sectional and clinical trial designs in data collection were included.

Exclusion criteria for ineligible articles
Studies that were excluded focused on dhfr, dhps, pfAP2, PfATPase6/SERCA , and PvK12 resistance alleles, malaria treatment, malaria epidemiology and/or transmission, methods/drug targets, non-Plasmodium malaria parasites, pharmacogenetics, review/opinion papers, non-human participants, mathematical modelling of resistance, studies primarily done prior to official introduction of ART agents in malaria treatment, articles which reported only synonymous k13 mutations, and citations whose full text articles could not be retrieved.

Data extraction
Data extraction form was developed in Microsoft Excel 2007 (Microsoft Corp, Washington, USA) and extracted the following information from included articles: author, year of study, country/region, study design, sample collection method, laboratory where analysis was performed, allele calling algorithm, source of DNA, year when ACT was initiated, period/years covered by study, number of parasite DNA where genotyping was attempted, number of parasite DNA samples where genotyping was successful, k13 mutations confirmed to cause delayed ART parasite clearance (validated SNPs in k13 propeller mutations), non-validated SNPs in k13 gene (k13 mutations that have not been confirmed to cause delayed ART parasite clearance), non-propeller k13 mutations, non-validated k13 mutations associated with delayed ART parasite clearance, nucleotide changes in k13 mutations, genotyping method used, duration of ART use, nature of ART use (monotherapy/combination) and source of k13 mutation.

Minimizing bias in article identification, selection and data extraction
A second librarian (RS) validated electronic search in PubMed by performing an independent and duplicate search. A second reviewer (EAO) screened all full text articles excluded by the first reviewer (MO). Any discrepancies among the reviewers were resolved by discussion and consensus. Two reviewers (NL and BB) performed duplicate and independent data extraction. Any disagreement between the reviewers was resolved by discussion and consensus. Any further disagreement between the two reviewers was referred to the third reviewer for a final decision (MO).

Data synthesis
Extracted data were transferred to Open Meta Analyst software for analysis [9]. Both structured narrative and quantitative syntheses of extracted data were applied as appropriate. In the structured data synthesis descriptive summaries of outcomes of interest were generated. These included: non-synonymous SNPs in pfkelch13 gene, duration of ART use prior to data collection, source of mutations (indigenous/introduced), nature of ART use (combination or monotherapy), allele calling algorithms, nucleotide changes in k13 mutations, spread of k13 mutaions, pfkelch13 mutations not confirmed to cause delayed ART parasite clearance (non-validated) but associated with delayed P. falciparum parasite ART clearance, genotype success rate, frequency of k13 mutations.
DerSimonian-Laird (DL) random effects analysis was used in establishing summary estimate of prevalence of non-synonymous SNPs in pfkelch13 gene in malariaaffected regions using Open Meta Analyst software. Sub-group analysis was performed, region (SEA, Africa, India), and study design (cross-sectional, clinical trial designs). Heterogeneity in included articles was inferred from the summary estimates of I 2 -statistic. The I 2 statistic was used to indicate percentage (%) heterogeneity that could be attributed to between-study variance. Interpretation: I 2 = 25% (small heterogeneity), I 2 = 50% (moderate heterogeneity), I 2 = 75% (large heterogeneity) [10].

Missing data and risk of bias assessment
The variables that were missing from included articles were recorded as not reported. Statistical tests were not used in handling missing data. However, authors whose articles had some missing variables were contacted but did not get any feedback.
The risk of publication bias was assessed using indirect assessment of rank correlation between effect size and sample size (Kendall's tau) method. In this method correlation of the articles is interpreted from the analysis output in Open Meta (Analyst) software where 1, represent perfect correlation and 0 no correlation.

Study identification and selection
The search of Medline via PubMed, Embase, Lilacs/Vhl and Scopus databases yielded a total of 469 citations on P. falciparum k13 gene single nucleotide polymorphisms (SNPs). Additionally, 12 citations were obtained from bibliography search and one article was obtained through contact with a malaria researcher (PR). After adjusting for duplicates, 409 citations remained. Of these, 296 articles were discarded after reviewing the titles and abstracts as they did not meet a priori inclusion criteria. Five citations were discarded as the full text articles could not be obtained. A total of 113 full text articles were screened in detail using pre-determined inclusion criteria. A total of 50 articles were included in the final full text review (Fig. 1).

Characteristics of included studies
Twenty-five articles were from studies in SEA, 2 from India, 21 from Africa and one from studies both in Africa and China. Two studies, by Ashley et al. [11] and Menard et al. [12] were performed in multiple centers both in  (2), non-randomized trials (2), and cross-sectional studies (43). Over half, 52% (26/50) of included articles did not report the year of official initiation of ART use in malaria treatment in the region/country where data were collected. While in 24 studies that reported years of official introduction of ART use, the mean duration of time from official initiation of ACT use to data collection was 9.3 ± 7.2 years. In addition, countries where studies were conducted, ART agents had been in use in malaria treatment for on average 15.4 ± 7.7 years from official introduction to date (2018). The majority of studies on P. falciparum k13 mutation were done after official introduction of ART agents in malaria treatment.
One study by Ariey et al. [1] used laboratory-adapted P. falciparum parasites from both SEA and sub-Saharan Africa while all other studies collected P. falciparum parasite isolates from symptomatic patients. ART agents were reportedly used as combination therapy in malaria treatment in all malaria-affected countries from where data were collected.

Genotyping errors
Genotyping was conducted in total of 31,655 P. falciparum DNA isolates of which 27,667 (87.4%) were successfully genotyped ( Table 1). The genotypes were assigned to successfully sequenced parasite DNA in small batches in all studies. Twenty-five studies collected P. falciparuminfected blood samples using filter papers (Whatman 903), EDTA vacutainer tubes (14 studies) while 11 studies did not report how samples were collected. The allele calling algorithms used included: Mega software ver 6.06, Bioedit ver 7.0.1, Sequencer software ver 5.4.5, Jalview software ver 5.1, Lasergene Genomic Suite, DnaSP software ver 5.50, Genome Assembly Program GAP4, Genescan, PROVEAN software tool, R software, Genotyper software and CEQ 2000 genetic Analysis System Software.

Overall prevalence of non-synonymous SNPs in pfkelch13 gene
The overall prevalence of non-synonymous SNPs in pfkelch13 gene (validated and non-validated) in reviewed articles from malaria-endemic countries is 27

Epidemiology of non-synonymous SNPs in pfkelch13 gene
Of 165  The k13 mutation P574L that commonly occurs in SEA was reported in Rwanda [32] and in a study on immigrants from Africa to China [31]. The other mutations reported in low frequency in both Africa and SEA include A675V, P584L, R575K, 522C 189T, R561H, M476I, F446I, N554S, and A578S. Two studies by Feng et al. [33] and Yang et al. [31] in both Africa and China, reported a total of 21 different k13 gene mutations of which 20 were found in only one parasite infection. Twenty of the 74 (27%) non-synonymous k13 mutations in Africa were also reported in SEA. In India, of 5 reported non-synonymous k13 mutations, 4 were found in only one P. falciparum parasite infection while a mutation, A578S was found in 2 P. falciparum infections in India. All the mutations found in India were also reported in other malariaendemic regions. In China, the most commonly reported k13 propeller gene mutations are F446I and P574L. The other SNPs reported in low frequency in China include H719N, M579I, F495L, L492S, C469F/Y, and N458Y ( Table 1).

Occurrence of k13 gene mutations and ART use
In some studies in Africa, k13 gene mutations were reported among P. falciparum parasites prior to official introduction of ACT in malaria treatment. These mutations include: E602D (Djaman et al. [34]: Ivory Coast); Y588C, E556K (Guerra et al. [27]: Equatorial Guinea). A study in Thailand by Putaporntip et al. [35] also reported existence of E605G and N609S mutations prior to introduction of ACT in malaria treatment. Of these mutations only one SNP at codon position E556K of pfkelch13 gene persisted even after introduction of ART use in malaria treatment [27,31]. A different allele at the same codon position 556 (E556D) was reported in SEA [29,[35][36][37].

Nucleotide changes of most reported non-validated k13 gene SNPs
Single nucleotide polymorphism at codon position 574 (CCT → CTT) of the k13 gene was reported by up to 5 different studies. The majority of frequently reported non-validated non-synonymous SNPs in k13 gene occurred among African P. falciparum parasites (Table 3).

Non-synonymous SNPs in non-propeller region (< 440) of pfkelch13 gene
The most frequently reported non-propeller k13 mutation occurred at codon position K189T. The other common non-propeller k13 mutations are located at codon positions 255, 252 and 352 of the pfkelch13 gene. Nonpropeller SNP at codon position K189T has been associated with delayed ART P. falciparum parasite clearance (t 1/2 > 5 h) [11]. Mutation at codon position K189T was reported in both Africa and SEA (Table 4).

Discussion
Malaria incidence rate reduced by 18% globally, from 76 to 63 cases per 1000 population at risk between 2010 and 2016 [4]. However, potential widespread emergence of P. falciparum parasite, mosquito resistance to ART agents and insecticides, respectively, is a threat to these gains in malaria control and elimination [4]. The review observed that overall prevalence of non-synonymous SNPs in propeller region of pfkelch13 gene (both validated and non-validated) in malaria-affected regions is 27.6%. k13 mutations confirmed to cause delayed ART parasite clearance (validated k13 mutations) were observed in only SEA with aggregate prevalence of 45.4%. The observed prevalence of non-synonymous k13 mutations varied in different regions with SEA, having the highest prevalence of k13 mutations, while African region had higher diversity and low frequency of reported polymorphisms.
Since the first discovery of k13 mutations as independent markers of ART-resistance in Western Cambodia [2] and later in Thai-Myanmar [40], there has been an overall upward trend in allele frequency globally [14]. In this review up to 165 k13 mutations are reported across malaria-affected regions with 110 being reported in only one geographical location (SEA). Only 20% of the mutations reported in Africa were also detected in SEA. In addition, new mutations were reported each time in over 50% of the studies. This review highlights presence of diverse but low frequency occurrence of non-synonymous SNPs in pfkelch13 gene across malaria-affected regions globally. In addition, the role of a majority of these alleles in parasite ARTresistance remains unknown.
Mutations R539T, C580Y (validated) and P574L (candidate) were reported in China-Africa studies. A study by Yang et al. [31] reported R539T and P574L among parasites from Angola and Equatorial Guinea, while Huang et al. [36] found C580Y and R539T among P. falciparum parasites isolated among immigrants to China from Ghana. These mutations could have been imported to China from Africa and is not clear whether they spread to Africa from SEA or emerged independently. A majority of the k13 mutations arise independently as indicated by studies done in SEA [14,41]. The current reported SEA alleles associated with ARTresistance among P. falciparum parasites observed in Africa and China could be due to independent emergence. Independent emergence of resistance alleles in pfkelch13 gene potentially responsible for the observed variations in k13 mutations across malaria-affected regions is influenced by several factors. These include widespread uncontrolled use of ART agents in addition to occurrence of anti-malarial medicines with sub-therapeutic amounts of active agents, especially in low and middle income countries. The extent to which different regions/countries are affected by these factors differs and could contribute to observed variations in k13 mutations. Furthermore, existence of different parasite founder populations (genetic background) [42] in addition to variations in environmental conditions influences independent emergence of k13 mutations.   Over half of the observed k13 mutations were each reported in a single P. falciparum parasite infection. A study by Straimer et al. [43] showed that each single nucleotide change in pfkelch13 gene is independently associated with an increase in parasite survival during ART exposure. Different k13 propeller mutations mediate different levels of ART-resistance [16,40]. Currently 6 ART-resistance alleles (N458Y, Y493H, R539T, I543T, R561H, C580Y) are confirmed (validated) as causes of ART parasite resistance [1]. Overtime a mutation at codon position C580Y has emerged as the most dominant in SEA and is associated with highest parasite clearance half-life [14]. The variation in extent to which these mutations affect ART parasite clearance half-life may be attributed to involvement of other loci in the genome [41,42]. The numbers of k13 mutations that are associated with increased clearance half-life of P. falciparum parasites by ART agents are rising over time. This has presented a challenge to policy makers and scientists who need to keep up with frequently validating different mutations. The review reports 16 k13 mutations that are not validated (not confirmed) despite being associated with an increase in ART parasite clearance half-life. Some of the 16 potential ART-resistance alleles occur among African P. falciparum parasites (K189T, P553L, P574L, A675V, M476K, and Q613H). Despite the current association of some of these alleles with increased parasite clearance half-life, there is a need to validate all frequently reported SNPs in the k13 gene among P. falciparum parasite populations in different regions, especially in Africa, which bears over 90% of malaria burden. A recent study by Mukherjee et al. showed that parasites carrying k13 mutation D584V had increased in vitro RSA 0-3 h survival. The mutation D584 V was reported in SEA [17,44] and among African migrants to China [36].
In the current review, 20 of 165 observed k13 mutations occur outside propeller region of pfkelch13 gene. Two of these mutations (A189T and E252Q) have been associated with increased ART parasite clearance halflife [42]. A mutation in codon position K189T was reported in both SEA and Africa. Molecular screening for ART-resistance currently is based on sequencing propeller region of pfkelch13 gene (1, 725,980-1726, 940 bp; amino acid positions 419-707) [1]. However, this review shows that studies have recently established existence of non-synonymous mutations outside the propeller region of pfkelch13 gene that are associated with increased ART parasite clearance half-life. In Cambodia, a mutation E252Q was reported by Wang et al. [19], Tun et al. [30], Ashley et al. [11], while D281V and K189T were in a study by Ashley et al. [11]. In Africa, mutation K189T was reported by Boussaroque et al. [22] (Senegal) and Conrad et al. [20] (Uganda). These mutations may provide valuable clues for the existence of gene polymorphisms potentially associated with ART-resistance outside the propeller region of pfkelch13 gene [42].
On the Thai-Myanmar border, ACT was first introduced in malaria treatment in 1994 with resistance suspected just over one decade later, in 2008, and subsequently confirmed in a study by Ariey et al. [1]. The review observed non-synonymous k13 mutations among P. falciparum parasites that occurred prior to official introduction of ART agents in malaria treatment. These included: E602D, I588C, E252Q, and E556K and were all reported in African P. falciparum parasites. Only one of these mutations, E556K, persisted in parasite population even after official introduction of ART agents in malaria treatment. A study by Hermission [45] indicated that ART-resistance alleles in k13 gene became established within a year of official ART deployment in malaria treatment in SEA. In malaria-affected countries outside SEA however, after over a decade of ART use in malaria treatment, there is still no confirmed k13 resistant alleles among parasite population. In Africa, which bears over 90% of malaria burden, this has been welcome news. However, current resistance surveillance reports have indicated potential emergence of parasites with increased ART clearance half-life in the region.
Although ART agents with effective partner drugs still cure patients harbouring slow ART clearing parasites [14,46], resistance development to ART component of the combination will expose more parasites to partner drug alone, increasing likelihood of ACT failure. The current failure by dihydroartemisinin-piperaquine combination to cure patients in Western Cambodia is a demonstration of this phenomenon [14]. The review observed that ART is used as combination agents in malaria treatment across malaria-affected countries. The use of combination ART agents (ACT) in malaria treatment potentially reduces resistance development [47]. However, eradication of ART-resistant malaria parasites necessitates complete elimination of all malaria since, as malaria incidence reduces, the few infections which remain to be cleared are almost all resistant [47]. Using long-lasting insecticide-treated mosquito nets alongside ACT could go a long way in eradicating malaria disease due to the synergistic effect on malaria incidence [47].
Use of ART agents in malaria treatment influences development of parasite resistance. The review observed emergence of multiple k13 mutations across malariaaffected countries following introduction of ART agents. A study by Amaratunga et al. [48] showed that ART agents select for resistance among P. falciparum parasites. In addition, it is probable that use of partner drugs such as mefloquine [49], piperaquine [14], and lumefantrine [50] to which the parasites are resistant may be influencing ART-resistance development. The magnitude of ACT selective pressure on development of parasite ART-resistance is thus likely to be greater, especially in infections with concomitant resistance to partner drugs [47]. The current malaria combination therapy may benefit from having more than two active pharmaceutical ingredients. The use of combination therapy having more than two active ingredients has proved effective in restraining HIV resistance development [51]. However, treatment of infectious diseases in most low and middle income countries is faced with the challenge of limited therapeutic options affecting potential use of more than two active ingredients in malaria combination therapy.
The review had some limitations, few studies were retrieved and reviewed from India and none from Central Asia and South America. This was mainly due to poorly designed studies, which were excluded. However, the application of extensive literature search and a priori selection criteria reduces the potential effect of this limitation on the overall review outcome. Most of the articles poorly reported their findings with some omitting a few review variables. To reduce the effect of missing variables, outcomes were recalculated for the review variables using data reported in the articles. In addition, corresponding authors of the articles were contacted for missing information, although no response was received.

Conclusion
The number of potential ART resistance alleles in the pfkelch13 gene is rising across malaria-affected regions. The k13 mutations with potential role in ART-resistance have been reported to occur outside the propeller region of pfkelch13 gene. Among African P. falciparum parasites although not validated, k13 mutations associated with delayed parasite clearance have been reported. Overall prevalence of non-synonymous SNPs in the k13 gene is higher in SEA with higher diversity observed among African parasites. k13 mutations occurring in more than one geographical region affected by malaria is increasing, however, polymorphisms associated with ART resistance have not yet been observed outside SEA.
Abbreviations PRISMA: preferred reporting items for systematic reviews and meta-analysis; STROBE: strengthening reporting of observational studies in epidemiology; PROSPERO: International Prospective Register of Systematic Reviews; STREGA: Strengthening the Reporting of Genetic Association Studies.