Background
Falciparum malaria remains one of the most important infectious diseases in the world, causing significant morbidity and mortality. More than 90 % of malaria deaths occur in sub-Saharan Africa, with the majority of these in children less than 5 years old [
1]. Considerable progress has been made in the past decade in reducing the burden of malaria in Africa, largely due to interventions such as long-lasting insecticidal nets (LLINs), indoor residual spraying (IRS), and the use of artemisinin-based combination therapy [
2]. However, the effectiveness of such large-scale interventions is non-uniform in part due to significant heterogeneity in baseline malaria transmission, demonstrating the need for targeted control and elimination programs that prioritize the most successful techniques [
3]. To most efficiently target interventions, the ability to stratify risk within a population is necessary, potentially including an understanding of fine-scale transmission. Traditional methods of evaluating transmission lack the resolution to differentiate the burden of disease over small spatial scales, especially in high transmission areas like Uganda, where an estimated 12 million clinical cases are treated annually in the public health system alone [
4].
As with most infectious diseases, malaria transmission within a geographical area is heterogeneous and can vary greatly between villages and between households in a village [
5‐
10]. Variations in malaria burden may reflect differences in vector distribution by habitat, human-vector contact, and human host factors [
5,
8]. More specifically, spatial clustering of malaria cases is well-described, which can be observed at scales as small as the household level [
8,
9]. While it has been proposed that parasite genetic data can be used to identify active clusters of infection, or “hotspots”, and to assess their contribution to onward transmission, these techniques are best established in areas of relatively low transmission, and it is less clear if these techniques can identify clustering of related infections in moderate to high transmission settings. To evaluate whether parasite genetic data can be used to elucidate heterogeneity in malaria transmission at fine spatial scales in a moderate transmission area, this study analysed data from 26 microsatellites in 349
Plasmodium falciparum infections identified in participants living in 80 households in Kihihi, Uganda.
Discussion
To evaluate the potential to detect genetic relatedness of malaria parasites in a moderate malaria transmission setting, relatedness of infections was measured using 26 microsatellites in Kihihi sub-county, southwestern Uganda. As expected, genetic diversity was high and the majority of samples were polyclonal. Despite the high diversity, spatiotemporal structure in genotypes was detectable, with significant spatial and temporal clustering of highly related infections identified in both higher and lower transmission areas of Kihihi, including marked clustering of related infections within households. Despite modest differences in malaria incidence found at higher vs. lower elevations within the study site, there were differences in MOI and spatial clustering of infections, reflecting underlying differences in transmission identified using detailed entomologic measures. Of note, statistical power for both of these molecular metrics was higher than that observed for malaria incidence. Thus, these data demonstrate the potential utility for genetic measures of parasite diversity in stratifying malaria transmission, even in moderate to high transmission settings where the majority of infections are polyclonal.
While it is clear that there is heterogeneity in malaria transmission at all levels of endemicity, this phenomenon is most obvious in areas of low transmission [
8,
22]. Therefore, genetic epidemiology of malaria often focuses on low transmission settings. For example, to look at the contribution of importation
versus local transmission in Namibia, Tessema et al.. showed that microsatellite data could capture strong spatial signals over both local and regional scales in a study with limited travel history and without cross-border mobile phone data. In a study conducted along the Thai-Myanmar border, single nucleotide polymorphism (SNP) barcode data were used to show that identity by descent (IBD), a measure of parasite relatedness, declined with increasing inter-clinic distance [
23]. In Zambia, a country targeted for elimination, Pringle et al. used microsatellite genotyping of samples collected through reactive case detection (RCD) to show that participants from the same RCD event harboured more genetically related parasites than those from different RCD events [
20]. Two other studies in moderate to high transmission areas showed clustering of infections at distances of < 1km [
24] or at the household level [
25], using SNP genotyping and amplicon sequencing, respectively. Other molecular epidemiology studies of malaria in settings of various endemicity have failed to observe any parasite population genetic structure, either geographically or temporally [
26‐
28]. Notably, using only 349 samples from 80 households, within-household clustering of infections was detected, suggesting that genetic data in this setting reflect what is known about clustering of malaria at the household level from epidemiological studies [
22]. These data show that by focusing on the most highly related infections in a population, as prior studies have done in lower transmission settings, micro-epidemiological patterns can be revealed even in a moderate transmission setting.
Additionally, this study showed that relatedness of infections between children was greater than that of infections between children and adults or infections between adults. This pattern was evident both between and within households, but was much more pronounced within households. This pattern may be due to age-related immunity, since adults may more readily suppress or clear a proportion of parasites to which they had been previously exposed. Alternatively, because children harbour higher parasite densities than adults, they may be more likely to transmit to other children, including those in the same household [
29]. A recent study by Nelson et al. that employed targeted amplicon deep sequencing to genotype parasites found that symptomatic children more commonly shared haplotypes with asymptomatically infected household members; however, household members were not stratified by age in this study [
25]. Understanding fine-scale transmission networks and whether household clustering is driven by transmission between household members or whether household members acquire related infections from exposure to the same vector would be helpful in determining the best method for spatially targeting malaria interventions in lower transmission settings, including interventions targeting the human parasite reservoir.
The study had some limitations. For example, IBS is a biased measure of parasite relatedness if one is attempting to compare IBS measures across different studies or populations, because it is sensitive to allele frequencies [
30]. An advantage of this measure, however, is its simplicity and applicability to polyclonal samples, and it has been used successfully in other studies [
20,
21]. Ideally, computationally tractable but more generalizable metrics of relatedness applicable to multiallelic, polyclonal samples would provide more robust analyses from this type of data. Unfortunately, methods to calculate such metrics, e.g. unbiased estimates of identity by descent (IBD) from these types of data are not currently available. Another limitation is the small sample size, with only 349 genotyped infections, which did not allow analyses to determine the dynamics of within-household transmission. However, this sample size was sufficient to observe clustering of highly related infections and differentiate neighbouring geographic regions that had only modest differences in transmission intensity. Notably, the ability to detect signal in small sample sizes is an advantage in studies that aim to utilize genetic data for surveillance due to lower cost and easier collection. Another strength of this study was the ability to capture information from polyclonal infections, leveraging the high diversity and multiallelic nature of the microsatellite loci. Each time an analysis in this study was replicated using only monoclonal samples, a similar but less well-defined pattern was found, suggesting that polyclonal samples contribute important information and help to better define aspects of population structure relevant for transmission. However, while the multiallelic, high-diversity nature of microsatellite loci has some advantages, this type of marker can be cumbersome to evaluate, particularly when considering larger numbers of loci to more accurately define relatedness between infections. Other genotyping techniques that are able to resolve multiallelic loci, such as multiplex amplicon or “microhaplotype” sequencing, can provide rich data from polyclonal infections and leverage current sequencing technologies to more easily allow evaluation of larger numbers of loci, providing higher resolution to evaluate transmission particularly in moderate-high transmission settings [
25,
31‐
34]. While these methods are more expensive and require more extensive laboratory and bioinformatics structure than the methods used in this study, they may be preferred techniques in the future as sequencing costs continue to decline.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.