Genetic diversity and drug resistance surveillance of Plasmodium falciparum for malaria elimination: is there an ideal tool for resource-limited sub-Saharan Africa?

To pre-empt the emergence or spread of resistance to anti-malarial drugs in Africa, the genotype and phenotype of parasites from all malaria endemic populations need to be regularly monitored. Accurate molecular surveillance in elimination settings require sensitive diagnostics of low grade parasitaemia and unbiased sampling strategies that enables resolution of temporal and geospatial patterns. Thus, the detection of parasitaemia by histidine rich protein (HRP)-based rapid diagnostic test (RDT), for instance, will be insensitive and potentially exclude P. falciparum parasites lacking HRP2 and HRP3 genes [42]. Given available resources, three common approaches can provide samples; (a) community-based cross-sectional surveys designed to capture the demography and spatial specifications of the target population, (b) clinical surveys, in which samples from patients suspected or diagnosed of malaria at health facilities are collected and (c) therapeutic efficacy studies (TES) engaged by research groups or the National Malaria Control Programme (NMCP). Each of these sampling approaches can be biased by being non-random and unrepresentative of the population. This can be ameliorated by increasing the density of spatial and temporal coverage or analysing all positive samples for populations where the prevalence of P. falciparum is low [43].

The choice of drug resistance surveillance assay or combination of phenotypic and genotypic analyses depends on resources and the primary goal of the study and NMCPs [44‐47]. Therapeutic efficacy studies remain the gold standard for anti-malarial susceptibility testing, but needs significant logistic organization [48]. The health budget, infrastructure and trained human resources in most sub-Saharan African countries are insufficient to support routine and wide application of TES. Typically, most NMCPs in sSA are supported by World Health Organization (WHO) and/or the Global fund for in vivo TES in limited number of sites. These provide samples for PCR correction and global genetic surveillance coordinated by the WHO. In vitro/ex vivo testing and genotyping of all tested samples will enrich drug efficacy data. Nevertheless, a subset of samples collected either from TES or cross-sectional surveys could be analysed by ex vivo/in vitro assays because of logistical constrains.

Samples from community or clinical surveys or from TES can be tested ex vivo or in vitro for susceptibility to different anti-malarial compounds. In vitro assays assess cultured parasites, meaning that only a subset of the parasites collected in the sample that are adapted to laboratory culture conditions are analysed, in contrast to ex vivo for which fresh samples are immediately tested. These drug susceptibility tests offer a convenient approach to determine sensitivity to anti-malarials, as they are based on direct contact between parasites and test drugs over a range of concentrations. In addition, several drugs (including those at experimental stages) can be tested against the same parasite isolate. A major limitation is the need for functional cell culture facilities and equipment for measuring parasite growth, such as microscope, spectrophotometer, fluorimeter or cytometer. Most tests determine the inhibitory concentration that kills half of the cultured parasites (IC50). IC50 assays have only recently been standardized by World Wide Antimalarial Resistance Network (WWARN) and are not suitable for fast acting artemisinin derivatives. Hence, the in vitro ring-stage survival assay (RSA) and piperaquine survival assay (PSA) were developed for testing ART-derivatives and piperaquine, respectively [45, 49, 50]. RSA together with Pfk13 molecular characterization is recommended in surveillance of ART-R [48]. These are highly technical assays targeting P. falciparum early ring-stages with a burst of high drug concentration in vitro. They are also yet to be developed for low grade parasitaemia often found in asymptomatic malaria parasite infection. Like IC50 assays, they are largely unstandardized and require specialized cell biology skills and techniques to distinguish parasite phenotypes such as dormancy. Although it is impossible to standardize in vitro assays, a data analytical tool (IVART) has been developed by WWARN to standardize the analyses process and facilitate future comparability of data across different studies [51]. The lack of drug resistant standards for current and candidate drugs, the substantial laboratory infrastructure and need for highly trained personnel in cell biology and data analysis limits the wide adoption of in vitro testing across Africa [44]. The future development of short culture-independent bench top tests, which target parasite metabolic markers from pathways of drug resistance, could allow for wider adoption.

Trends in anti-malarial drug efficacy and resistance can also be assessed using molecular markers (Table 2). These can be good predictors of treatment outcome patterns at a population level [52‐54]. Molecular markers have several practical advantages over in vitro tests, including the possibility of studying many isolates within a short time with DNA from dried blood spots (DBS), which are easy to collect, transport and store. There exists a wide panel of molecular markers, mostly associated with widely used previous first line anti-malarial drugs, quinolines and sulfadoxine–pyrimethamine (SP) [44]. These previous first-line drugs are still used for intermittent preventive treatment in pregnancy (IPTp) and seasonal malaria chemoprevention (SMC), or against co-transmitted Plasmodium vivax. Hence, their resistance markers remain relevant, especially as some alleles modulate efficacy of ACT [55]. Most marker alleles can be typed using technological variants of various methods, including; dot blotting, restriction fragment length polymorphism (RFLP), PCR/qPCR, mini/microarrays, targeted/whole-genome sequencing. Sample pooling strategies, advances in PCR and DNA sequencing technologies promise to improve the scalability of genotyping resistance marker and wider population coverage [56]. Next Generation Sequencing (NGS) in particular can now be achieved with as low as $22 per Gigabyte (Gb) on Illumina HiSeq 3000/4000 while Third Generation Sequencing with Pacific BioSciences, the most widely used long-read platform costs $1000 per Gb [57].

Emerging nucleic acid sequencing technologies such as the portable nanopore devices also have the potential to revolutionize next-generation sequencing approaches towards field-based genotyping in resource-constrained regions in sSA. However, nanopore assays specific to malaria parasite diversity and drug resistance markers are unavailable or only recently been developed despite clear applicability of this technology in outbreaks such as Ebola and Lassa fever in Africa [58, 59]. Developing field deployable assays for markers of ACT component drugs (artemisinin derivative in combination with either lumefantrine, amodiaquine, mefloquine, quinolines or sulfadoxine–pyrimethamine) in Africa would be most useful.

Mutations in the propeller domain of the Pfk13 gene have been validated as P. falciparum ART-R markers in SEA [34, 54]. SNPs in four other genes, namely, ferredoxin (PF3D7_1318100, fd), apicoplast ribosomal protein S10 (PF3D7_1460900, arps10), multidrug resistance protein 2 (mdr2) and chloroquine resistance transporter (crt) are also thought to be the backbone loci on which ART-R associated kelch13 mutations are most likely to arise [54]. Though key Pfk13 ART-R associated mutations are absent in sSA [38], typing panels could include known and candidate drug resistance loci whose allele frequencies are changing in some parts of sSA following ACT implementation [29, 60]. Artemether–lumefantrine (AL), for instance, is thought to select for Pfcrt and Pfmdr1 wild type variants following treatment. Other markers could include variants of Pfap2mu (encoding clathrin-associated AP2 adaptor protein, µ subunit) and Pfubp1 (encoding ubiquitin carboxyl-terminal hydrolase 1) found more frequently in P. falciparum isolates post-ACT [61]. Indeed, candidate SNP variants in Pfap2mu should be included in SNP panels for continental surveillance during ACT and for validation as a marker for ACT tolerance in African parasite populations.

The choice of genotyping panels is complicated by the variety of recommended ACT medicines, heterogeneously available across Africa. This presents a threat of partner drug resistance across a broad spectrum of anti-malarials. Already, mutations in Pfcrt and Pfmdr1 are being selected by the most common ACT, artemether–lumefantrine (AL) and artesunate–amodiaquine (AS–AQ) [62]. Polymorphisms in both genes are also associated with structurally similar partner drugs; mefloquine, amodiaquine, piperaquine [63]. Moreover, Pfcrt mutations as well as plasmepsin 2/3 copy number variations have been shown to confer P. falciparum resistance to piperaquine [64, 65]. The presence of these in African populations would affect deployment and efficacy of piperaquine-based ACT [66]. Monitoring of partner drug resistance markers is, therefore, particularly key for parasite elimination. When such markers are detected in the circulating parasite populations, treatment and preventative chemotherapy policy could be revised to switch to other drug combinations. Recommended drug-based preventive malaria interventions in Africa include; intermittent preventive treatment in infants (IPTi) and pregnant women (IPTp) with SP in areas of moderate-to-high malaria transmission in sSA and amodiaquine in combination with SP for SMC in the Sahel [16, 67]. However, resistance to SP is widespread and marked by mutations in the P. falciparum dihydropteroate synthetase (Pfdhps) and dihydrofolate reductase (Pfdhfr) genes. Though chloroquine, amodiaquine and SP are no longer first-line malaria treatment, mutations in Pfmdr1, Pfdhfr, Pfdhps, Pfcrt and candidate loci for artemisinin and piperaquine resistance are needed to assess the impact of current drug interventions.

The design of surveillance panels should consider the complex relationships between individual mutations and haplotypes with sensitivity to specific anti-malarials. For instance, the Pfcrt 76T mutant is essential for chloroquine resistance but is selected as Pfcrt CVIET or Pfcrt SVMNT haplotypes that confer fitness against chloroquine pressure [68]. Similarly, SP resistance is linked to the Pfdhfr IRN triple and Pfdhps double GE haplotypes selected by SP interventions [69]. Medium SNP typing panels could include these linked loci, while small panels could combine markers that maximize information for key resistance loci of anti-malarial treatments or chemoprevention. For example, countries using AL, could employ a small panel consisting of N86Y, K76T and C580Y for Pfmdr1, Pfcrt and Pfk13, respectively. This can be expanded to include Pfdhfr K540E and A581G to cover selection by IPTi/IPTp and SMC in the Sahel. These panels will gain from assays such as RFLP and allelic discrimination real time PCR by high resolution melting (HRM), which do not rely on heavy and expensive equipment. These techniques can achieve higher coverage given the wide availability of restriction and amplification reagents including isothermal polymerases. They could be extended to detect copy number variants, such as those in plasmepsin genes associated with sensitivity to piperaquine. The small/medium panels would be used at peripheral labs with limited capabilities, but all the markers should be assessed in CEAs.

Single nucleotide polymorphisms and structural genetic variants of P. falciparum can serve to fingerprint infecting parasites and inform the effectiveness of drug treatments or vector interventions. Four major strategies have mostly been explored, each with advantages and limitations for wide adoption across sSA. These include: msp/glurp typing, microsatellite analysis, molecular (DNA) barcodes, targeted deep sequencing and genome-wide variation analysis.

Viriyakosol et al. [70] were the first to deploy a tool that exploits length polymorphisms in merozoite surface protein (msp) and glutamine rich protein (glurp) genes to assess the identity or genotype of infecting parasites. Two different msp markers are often used, msp1 and 2. This technique is particularly useful in determining the complexity of infection (COI), a measure of the effectiveness of intervention programmes. Msp/glurp typing are widely used in anti-malarial drug efficacy trials to distinguishing recrudescent parasites from new infections [71, 72]. It has been one of the most widely adopted techniques because of the availability of PCR and DNA electrophoresis equipment which are now portable. Furthermore, its sensitivity and reliability are improved in specialized laboratories by incorporating capillary electrophoresis and more recently qPCR-HRM [71, 73].

A major limitation of msp/glurp typing is lack of standardization of scoring and reporting formats that can allow for the comparison of results across different endemic site laboratories. The msp/glurp genotyping protocol is also labour-intensive, depends on the sensitivity of PCR, which may not amplify low abundance variants or result in artefacts [74]. In addition, the sensitivity is low when using agarose gels and the interpretation can be subjective, especially in high transmission areas in Africa where polyclonal infections will lead to multiple bands. Furthermore, msp/glurp genes are under immune selection pressure, which can skew the frequency of some allelotypes [30]. This could affect the accuracy of population structure and transmission patterns inferred from these loci. Notwithstanding the above limitations of msp/glurp typing, it remains a popular method for fingerprinting across endemic regions in sSA. Its continuous use will benefit from allele size reference standards derived from culture adapted reference isolates from major P. falciparum populations across Africa.

Microsatellites are tandem repeats of one to six base pairs (bp) that are highly polymorphic. They are abundant in the P. falciparum genome, mainly as [TA]n, [T]n, and [TAA]n repeats [75]. Anderson and colleagues pioneered assays of twelve microsatellite loci as sensitive tools for parasite genetic structure and differentiation analyses [76]. These twelve loci have been widely used for assessment of genetic diversity and population structure across many studies. In the context of malaria elimination, microsatellites can be used to fingerprint parasites and resolve relatedness of infections at high spatial resolutions especially in settings with low prevalence and high levels of monoclonality [77‐79]. Also, the combination of P. falciparum drug-resistance markers and flanking microsatellite loci can be employed to assess the genetic diversity and evolution of selective signatures around drug resistance genes [39].

Similar to msp/glurp typing, microsatellite typing depends on accurate DNA fragment amplification by PCR, which can result in artefacts. Accurate allele sizing requires expensive capillary electrophoresis equipment that are largely unavailable. Moreover, there are no methods for phase resolution of parasite haplotypes in mixed infections. However, thousands of P. falciparum genome sequences are now available for mining new microsatellite loci which can be developed into new assays for population diversity analyses. Low cost thermocyclers combined with HRM for determining sizes of DNA fragments can improve the availability of microsatellite typing to less specialized laboratories [80].

SNP barcoding has been shown to be robust for evaluating parasite genotypes derived from communities, malaria patients or laboratory strains [81]. It provides a fingerprint for each infection, distinguishing the haplotype signature for single clone infections from those with mixtures of parasite genomes [82]. SNP barcoding are also sensitive for detecting and genotyping sub-microscopic parasitaemia in low malaria transmission areas even when RDTs are negative [83]. In addition, the technique has the potential for identifying the sources of epidemics [84]. The choice of a SNP panel and density within a barcode can be guided by the level of sensitivity required for distinguishing populations or detecting alleles associated with phenotypes such as drug resistance. Early approaches were based on qPCR to distinguish alleles of a panel of 24 SNPs that represented unique signatures for P. falciparum isolates [82]. These panels have been developed for qPCR-HRM and the Sequenom Mass Array assays [84, 85]. Only a small number of specialized labs in Africa have successfully implemented and used these assays due to the absence and cost of advanced PCR or array technologies. Several other SNP barcode combinations have now been developed for different types of analyses:

While the above techniques and platforms overcome the limitations of traditional msp/glurp and microsatellite genotyping, they remain mostly applicable only in specialized northern labs. Furthermore, their sensitivity across populations of different effective sizes and recombination rates, which disrupts association among SNPs panel, remain unevaluated. These techniques are also most suitable for single clone infections. Even with the application of algorithms such as RealMcCOIL to estimate the number of infecting genomes in mixed infections, reconstruction and analysis of individual haplotypes from infections with greater than two clones remain almost impossible [89]. With thousands of P. falciparum genomes now available, new barcode panels with population specific alleles combined with new computational models should allow for high resolution determination of origin and diversity of infections in elimination settings and connectivity between spatial transmission hotspots.

This is a powerful approach for variant discovery and detection in target genes or genomic regions. TDS can assess the relative distribution of variants of specific genes at individual patient level, or at population level with variance in environment; heterogeneous human and vector hosts as well as interventions. Coupled to long read technologies, it can also permit the reconstruction of haplotypes for relevant genes such as drug resistance loci and vaccine candidates. Applied to known or candidate drug resistance gene loci, it can elucidate the landscape of selection of alleles in such targets as parasites adapt against drugs [90]. In an observational surveillance study of potential drug resistance loci in Senegal and Thailand, TDS identified dozens of previously undescribed mutations in Pfk13 [91, 92]. TDS can, therefore, be an important public health tool, enabling the detection of low frequency and potential drug resistance loci that may be selected under increased drug pressure.

TDS of P. falciparum also provides a high-throughput, highly sensitive approach for detecting minority clones in polyclonal infections and more accurate quantitative estimates of clonal frequency [10, 93]. By determining the frequency of parasite gene haplotypes in individual infections from a parasite population, the tool can enable the assessment of within-host diversity, recombination events and relatedness between infections. Such applications are important in the context of Africa where infections are often mixed, and the ecology of mixed genotypes might influence the transmission of drug resistance loci, the backbone on which resistance may emerge or clones selected by vaccines [94]. However, TDS suffers from ascertainment bias of gene panels. It requires next-generation sequencers, powerful computation and well-trained biostatisticians for data analysis that are not readily available in malaria endemic countries. As this can produce highly relevant data for diversity and drug resistance monitoring in elimination settings, centralized analyses can be done at CEAs and dissemination of simplified reports such as heatmaps of allele and haplotype frequencies could facilitate translation by control programmes.

Global surveys of P. falciparum genomic variations has revealed thousands of SNPs, indels, and structural variants, typed across populations using microarray or next-generation sequencing approaches [10, 31]. Genome-wide SNP loci have mostly been characterized and used to scan for and differentiate signatures of natural selection or employed in genome-wide association studies to identify loci underlying phenotypes such as drug resistance [25, 30, 32, 33]. Genome-wide variation typing has evolved from microarray technologies, which suffer from ascertainment bias, to whole genome sequencing (WGS) by short read technologies. Theoretically, the wide variety and density of polymorphisms from WGS should improve the resolution of studies on diversity, the tracking of known functional variants and discovery of novel markers that may be associated with new adaptive parasite phenotypes. WGS data can also be employed in development of new population genetic and diagnostic tools.

Unlike the analysis of individual loci, which may be the target of strong natural selection, inferences from genome-wide diversity are less subject to target biases and, therefore, more accurately reflect overall patterns of P. falciparum genetic variation. Nevertheless, the current costs of next-generation sequencers, sample processing reagents, infrastructure for data storage and analysis and need for trained lab and data analyses scientists limits its wide application across most of Africa. In addition, WGS may be limited by the lack of good quality DNA and the need for enrichment for blood samples that contain mostly human DNA. Furthermore, it is difficult to construct haplotypes for complex infections and the technique has a low sensitivity to detect minority clones. With improved error rates from cheaper portable sequencers, such as the nanopore, WGS and GVA will most likely become more translational for malaria elimination. A model where a limited number of CEA laboratories can be equipped to receive and process samples from a network of peripheral laboratories in the continent at high throughput is proposed. Already, WGS is possible in a number of centres in sSA such as the Medical Research Council Unit The Gambia at LSHM. Human capacity for bioinformatics is also being built through African-led networks; Developing Excellence in Leadership and Genetics Training for Malaria Elimination (DELGEME, www.​delgeme.​org) and the Human Hereditary and disease in Africa (H3Africa) Pan-African Genetic Epidemiology Network (PAMGEN). These programmes will expand access and application of WGS/GVA approaches to malaria elimination research in sSA. Effective deployment and translation could be further facilitated by sample pooling and employing cheaper but sensitive Real time PCR genotyping methods to pre-screen and select samples in peripheral laboratories prior to genome wide analyses at CEAs. Pre-screening could prioritize loci for diversity, drug or vaccine resistance most relevant to the local control and intervention strategies.