Genomic analysis reveals independent evolution of Plasmodium falciparum populations in Ethiopia

Plasmodium falciparum malaria remains one of the major public health problems worldwide accounting for 228 million cases in 2018 compared to 231 million in 2017, while the number of deaths due to malaria decreased by just 2.5%, from 415,000 to 405,000 during the same period [1]. Sub-Saharan Africa (sSA) still accounts for 94% of global death. In Ethiopia, more than 75% of the total area is malarious, and P. falciparum and Plasmodium vivax co-exist [2] making malaria control more complicated than in other African countries.

Across malaria-endemic regions, large-scale deployment of anti-malarial drugs has led to the emergence of drug resistance to chloroquine (CQ) and sulfadoxine/pyrimethamine (SP) antifolate drugs [3‐5]. Like many other countries, Ethiopia has switched from CQ to SP in 1998 and from SP to artemether–lumefantrine (AL) in 2004 [6] for the treatment of uncomplicated P. falciparum malaria in response to the development of parasite resistance. However, CQ remained the first-line drug for P. vivax treatment in the country [7], leading to a continued selection of CQ-resistance markers in P. falciparum as the result of indirect pressure from CQ and the presence of mixed infections of P. falciparum and P. vivax. Similar to CQ and SP, P. falciparum developed resistance to AL first at the Thai-Cambodian border [8] and recently in East Africa [9]. The rapid development of resistance in P. falciparum to series of the first-line anti-malarial hinders malaria prevention, control, and elimination efforts.

Anti-malarial drugs are known to pose tremendous selective pressure on P. falciparum leading to the worldwide spread of resistant parasites [10]. It was well noted that P. falciparum resistance to the two conventional anti-malarial drugs, CQ and SP, has resulted in increased malaria morbidity and mortality across endemic settings. Apart from increased morbidity and mortality, selective sweeps of drug resistance mutations have reduced levels of polymorphism in P. falciparum as these resistant and sensitive strains continue to recombine in mosquitoes [11, 12], with perhaps a reduced diversity around the selected loci. However, the greatly reduced level of diversity across the entire P. falciparum genome most likely resulted from a severe population bottleneck as has been observed in gorilla-to-human cross-species transmission [13].

Based on the analysis made on 12 strains collected from different countries in Africa and Asia, the average diversity of P. falciparum at fourfold degenerate sites was estimated to be 8 × 10⁻⁴ per site [13]. However, published mutation rates for P. falciparum were in the range of 1–10 × 10⁻⁹ mutations per site per replication cycle [14, 15]. Depending on P. falciparum life cycle and assuming varying lengths of time that the parasites spend either in the vector or in the mammalian host, the parasites are likely to undergo at least 200 replication cycles per year suggesting that the observed level of genetic diversity in P. falciparum could have readily accumulated within the past 10,000 years [10].

Plasmodium falciparum parasite change and select its new genetic variants owing to drug exposure to cause disease and overcome challenges from host immunity and therapeutic interventions [16]. Indeed, high pressure from immunity and drugs are known to select adaptive parasite strains that maintain transmission [17] and, therefore, many P. falciparum genes encoding immune and drug targets are under natural selection and show signatures of balancing or directional selection [4, 17‐21]. This selection may vary due to differences in innate susceptibility of human populations, variations in ecological transmission, resulting in varying degrees of acquired immunity and/or drug pressure. Malaria parasites from low and high endemic regions have a distinct opportunity for transmission and host acquired immune responses [22]. For effective management of malaria control and intervention strategies, it is important to determine genetic variation patterns due to parasite adaptation to host environments and drug interventions. Balancing selection brings the favoured alleles of parasites to an intermediate equilibrium where they are maintained as genetic polymorphisms, while the directional selection forces cause the parasite’s genetic variants to increase in frequency and facilitate the occurrence of selective sweeps around the affected loci [23].

Plasmodium falciparum population genomics has been highly studied in West African populations and showed signatures of balancing selection on multiple candidate vaccine antigens and strong directional selection around known drug resistance genes [19, 24]. In contrast, there is little information about the genomic variations of P. falciparum populations in the horn of Africa, including Ethiopia, where P. falciparum and P. vivax malaria co-exist and are heterogeneously distributed. A recent study reported P. falciparum populations in the horn of Africa, specifically in Ethiopia, are unique and structurally diverged from other West, East, and central African P. falciparum populations [25]. These parasite populations share a chunk of genes with other sub-Saharan African P. falciparum populations across drug and immune targets and facilitate the spread of drug-resistant strains [25]. These studies call for in-depth analysis of Ethiopian parasite genomes to deepen understanding of genome diversity and natural selection in Ethiopia’s unique human populations with co-species transmission dynamics.

Understanding the population genetic diversity of P. falciparum strains circulating in the specific region of central Ethiopia is very important to monitor the effectiveness of control schemes and provide baseline information for making informed decisions by the national malaria control programme [26]. This study aimed to characterize the P. falciparum populations in West Arsi, in central Ethiopia, using whole-genome analysis of data generated by Illumina next-generation sequencing.

The transmission dynamic coupled with the unique history, ecology, and demography of Ethiopia raises interest in the genetics of its parasite population. High-resolution whole-genome SNP data was used to analyse P. falciparum parasite genetic diversity in the central region of Ethiopia and compared with similar parasite data from mainland Africa (DR Congo and Malawi) and Southeast Asian parasites, from Cambodia and Thailand. In this analysis, similar MAF across all five parasite populations with over-representation of low frequency (< 5%) variants was observed as previously reported [19, 21]. Interestingly, mean F_WS values were significantly higher in the Ethiopian parasite isolates as compared to the other African populations, but not the Southeast Asian parasite populations. F_WS is a genome-wide metric that averages heterozygosity across the genome in comparison with heterozygosity within the local parasite population [21]. Hence, it is a measure of within-host diversity of infections that helps to gauge the potential for inbreeding (or outcrossing). The higher F_WS values (> 0.95) in Ethiopia (P. falciparum prevalence of 0.02) [36] and East Asian infections is underscored by the low malaria transmission rates in these settings which supports a higher inbreeding and clonal propagation of infections (Fig. 1; Additional file 3). Unlike the other East African countries (DR Congo and Malawi) where transmission intensities are higher [20, 35], and there was a good distribution of F_WS values with the majority of infections being polyclonal with high potential for outcrossing (Fig. 1). These findings are supported by similar studies that link lower F_WS values to in west African where transmission is high [18]. However, of note, is the possibility for high F_WS values to occur in areas of high transmission intensity if P. falciparum circulates in a geographically isolated community which limits the chance of outcrossing with other genetically distinct P. falciparum parasites as observed in the previous study [21].

An analysis of parasite population structure within and between continents revealed a higher degree of population structure between Ethiopian parasites and other East African parasites and between Southeast Asia and East Africa. However, neither PCA (Fig. 2) nor admixture analysis (Fig. 3) could resolve parasite populations in DR Congo and Malawi. These observations are corroborated by several studies that report regional and inter-continental level structure in global P. falciparum parasite populations [21]. However, the separation of Ethiopian parasites from the two East African populations is worth noting. Notwithstanding the increased human mobility between Addis Ababa and the rest of Africa, particularly East Africa, there remain important barriers to gene flow between parasite populations in central Ethiopia and the rest of the sub-region. Indeed, one possible factors that severely limit gene flow between Ethiopia and its neighbours is the local malaria transmission intensity as a function of poor vectoral capacity determined by the ecological landscape (highlands).

Against the backdrop of this unique eco-epidemiology of P. falciparum malaria in Ethiopia, Tajima D and IHS was used to explore the mechanisms of natural selection in the country. However, identification of many antigenic genes under balancing selection with Tajima D value greater than one were observed in Ethiopia. These genes included known vaccine candidates, such as ama1, trap, msp1, eba175, and clag2 (Additional file 4), which were previously identified in different populations that vary in transmission intensity [24, 33, 37], to be under balancing selection. Besides, 15 genes under positive directional selection by IHS were identified, which includes SURFIN and PHIST families previously suggested to be targets of immunity [24]. It can be hypothesized that the low seasonal transmission in Ethiopia maintains significant immune selection pressure on the infection reservoir than drug pressure due to clinical malaria. Therefore, the candidate vaccine antigen loci under balancing selection may be largely due to immune modulation and not positive adaptive selection influenced by drug pressure. This is supported by our failure to detect selection signatures in known drug target genes such as Pfcrt, Pfmdr1, Pfdhfr, Pfdhps, and Pfkelch-13. The ability of IHS to detect selection in these drug resistance genes in Ethiopia may be because the frequency of polymorphisms in these loci are either fixed or near fixation in the Ethiopian population (Table 5). These findings are supported by a previous study in Ethiopia which showed that the CQ-resistant haplotype (CVIET) was fixed [7]. The continued use of CQ in Ethiopia for the treatment of P. vivax malaria may account for the high prevalence of CQ resistant markers. Also, Pfmdr 1 mutations have been demonstrated to mediate AL resistance. Therefore, the high prevalence of Pfmdr 1 mutations may signal poor efficacy of AL as the first treatment for P. falciparum malaria in Ethiopia. Variable prevalences of CQ-resistant polymorphisms were observed only in DR Congo and not in Malawi, evidence that supports the complete reversal of CQ susceptibility in Malawi as reported by Ochola et al. [20].

Undoubtedly, artemisinin resistance has taken root in Southeast Asia. Despite 36% and 26% prevalence of PfKelch13-C580Y mutation in Cambodia and Thailand samples, respectively, no validated PfKelch13 mutation was found in the African samples. However, an uncharacterized Pfkelch13 mutation (PfK13-K189T) found at prevalence > 10% in all the African datasets and previously reported in other studies [35], may be important, but its role in artemisinin resistance is unknown. One study [8] reported that mutation in Pfkelch13 at amino acid positions less than 441 may not play any role in mediating artemisinin resistance. A validated Pfkelch13-R561H mutation for artemisinin resistance was recently reported in other East African P. falciparum populations [9].

Overall, this study reveals the presence of a comparably low genetic diversity of P. falciparum parasites in Ethiopia. The majority of infections were of low complexity, demonstrated significant population structure with Ethiopian parasites diverged from parasite populations within the sub-region. Based on the analysis made it is suggested the presence of limited gene flow between parasite populations in the East African sub-region and Ethiopia. More importantly, apparent balancing selection in antigenic loci known to be targets of immunity and adaptive positive selection in SURFIN and PHIST gene families that are potential vaccine antigens. Though selection analysis did not pick up any adaptive mutations in known drug-resistant genes, CQ-resistance Pfcrt-K76T genotype seems fixed in Ethiopia like the wild-type genotype (K) in Malawi. In this analysis no PfKelch13 validated mutations were reported in Ethiopia, DR Congo, and Malawi except a PfK13-K189T African specific uncharacterized mutation. Further molecular studies involving deeper sampling of Ethiopian parasite populations are essential to understand the genetic diversity, gene flow, and temporal evolution of drug resistance loci within Ethiopia. Furthermore, such findings can be used to support national malaria control decision-making for optimal impact in further reducing malaria transmission in Ethiopia.