Malaria elimination requires the identification of clearly defined spatial malaria transmission units where the control effort can by directed in cost-effective ways and where progressive, feasible goals towards elimination can be established. This requires an understanding of the dynamics of malaria transmission and the barriers hindering elimination goals within the spatial unit. Here, we initially defined a spatial malaria transmission unit by carrying out passive case collections and taking travel histories at local health posts. These revealed that infections diagnosed at the Public Malaria Facility in the town of Guapi, in the western Pacific region of Colombia, originated mostly (>95%) in 33 small rural settlements and in peri-urban neighborhoods of the town of Guapi. The area covered by the catchment facility is an area of 25 km radius, having the town of Guapi as its centre. Cases originating outside this spatial unit represented less than 5% of the cases and they originate from as far as Venezuela (Fig. 4). This has important implications for the malaria control programme: 1) the spatial size covered by the catchment facility is small enough to implement focussed control measures aimed at the elimination of malaria transmission in the area and 2) it reveals that urban malaria transmission is much lower than predicted by the current reporting system, suggesting that if appropriate malaria control measures are provided, urban malaria transmission can be eliminated within a reasonable timeframe.
A characterization of malaria transmission within this spatial unit showed that malaria transmission is heterogeneous in time and space. Three transmission patterns were observed: peri-urban transmission occurring close to mosquito breeding sites and representing a relatively small proportion of the total number of cases (less than 20% of overall cases); low endemic transmission (24.5% of total cases) occurring mainly in small settlements (60–200 inhabitants) in the rural areas of the municipalities of Santa Bárbara de Iscuandé, Guapi and Timbiquí, and epidemic transmission (56.7% of total cases), occurring in rural settlements (Limones, Carmelo, Quiroga, San José de Guare) of relatively high human population density (densely populated human settlements of approx 1,000 inhabitants), connected through riverways and located in close proximity to sites where gold mining activities were taking place at the time of the study. This coincided with a weakening of the malaria microscopy network in the Pacific region, the arrival of migrants from Venezuela to Colombia, the increase of malaria cases in the Pacific region associated to P. falciparum infections and an increase of mainly illegal-gold mining activities in large areas of Colombia.
Diagnostic SNP combinations defining parasite subpopulations can be used to infer sources, routes of importation and mobility patterns that can constitute a useful tool when difficulties arise in using other data to quantify the flow of individuals between locations (e.g. scant data on commercial boat mobility or low mobile phone coverage and use)31. We assessed the population structure of P. falciparum populations circulating in the area using a SNP barcode corresponding to 49 independent loci (see Methods), taking advantage of the long term persistence of clonal lineages in South American populations of P. falciparum28 which allows tracking parasite populations in time and space. Long term persistance of clonal lineages is probably due to the low diversity of P. falciparum populations in the Americas27 as a consequence of the bottleneck resulting from its relatively recent introduction during the slave trade45,46, the strong selective pressure by antimalarials47,48 and the relatively low mosquito Blood Index Feeding rates in South American anophelines as compared to high transmission areas in Africa49,50,51,52,53. This results in a low proportion (8% in this study) of polyclonal infections, low MOIs and highly clonal lineage infections persisting over time28.
We identified three P. falciparum subpopulations (A, B and C), circulating and coexisting in the area, of which two, B and C, were highly clonal. This may be the result of low effective recombination frequency and past waves of clonal expansion. Previous studies using a set of 250 SNPs have found four populations circulating in Guapi between 1999-2003, of which two lineages constituted 98% of the samples28. Using a set of six microsatellite loci we only detected two parasite populations. The higher resolution provided by SNP data may be due to the fact that three of the six microsatellite loci used (TA1, TA109, C2M34) have low diversity in these populations43,54. We used the Evanno method to estimate the optimal number of clusters detected by STRUCTURE without making any assumption on the demographic history of parasite populations, i.e. the contribution of selection, drift or past introduction of new genotypes. However, there are limitations to the method: it is sensitive to the sample size and it may not detect sub-structuring of the populations55. Here, for example, it cannot discriminate parasites originating from Venezuela from other populations which can otherwise be visualised in the distance-based dendrogram (Fig. 3B) or in PCA analysis (Fig. 3C). Since genetic distance between clusters may be due to divergence, admixture between the observed populations or with unknown “ghost” populations, a more detailed sampling of parasites populations from other regions outside the spatial transmission unit is needed, including a higher density of SNPs that would be required to evaluate demographic and epidemiological, or ancestral population admixture history hypotheses56.
The spatial frequency distribution of parasite populations shows that all three populations are observed in the town of Guapi, which therefore constitutes a sink for cases across the rural area of the Guapi municipality. At the extremes of the spatial transmission unit some populations are more abundant (Fig. S1 3) suggesting that parasite populations originate in different geographical areas. Evidence for this has been previously provided, showing the differential distribution of 4 different parasite populations along the Colombian Pacific coast28. The fact that all parasite populations are observed in the town of Guapi reveals its nature as a receptor of individuals seeking malaria diagnosis and treatment. With a population of nearly 18,000 inhabitants, it is the major human centre in the municipality and constitutes a hub for commercial activities providing different facilities such as education centres for children and the youth and health services through private and public providers. Private and public malaria diagnostic posts in Guapi report malaria cases to the regional and national health authorities. Critically, these reports assume all cases as being the result of local transmission, an assumption that this work clearly proves is not valid.
In this work we defined imported cases in the town of Guapi as those cases where the individual has been continuously present at a site outside urban Guapi in the previous two weeks before diagnosis. There are two limitations to this definition: first, individuals may have acquired the infection in places close, but not exactly at, to the indicated site of residence (e.g. at nearby mining sites) and second, infections may become patent after acquisition in the previous three to four weeks. Although these caveats mean that precision may be lacking with respect to the exact time and site of infection acquisition, the general conclusion that the majority of cases diagnosed in Guapi are imported and originate in defined areas in the surroundings is supported by their association to mining sites, known foci of infections. The large number of sites outside urban Guapi, suggests that if anything, the analysis underestimates the number of cases classified as rural or peri-urban.
TDA combines genetic and epidemic variables to explore transmission patterns over space and time. As mentioned above, we observe heterogeneous transmission in the Guapi area and surroundings. The town of Guapi is highlighted by all descriptive measures, as is Bagrero (geographic Pagerank, case Pagerank and betweenness) during epidemic and inter-epidemic intervals. The area of El Cuerval at the northen part of the malaria transmission unit is highlighted by high betweenness centrality in both networks. A strong interaction between Guapi and Carmelo is emphasised by the TDA, suggesting that epidemic areas such as Carmelo and Quiroga are connected by their epidemic dynamics. Bagrero and El Cuerval appear as secondary in the map, however, their betweenness centrality suggests they both play an important role in connecting regional parasite populations.
This suggests that El Cuerval and Bagrero possibly constitute semi-independent hubs within the transmission unit, which play a role in the reintroduction of parasites (due to the high betweenness centrality of the cases during non-epidemic years). A third transmission hub in the area comprised by Carmelo and Quiroga is characterised by high Pagerank values in the geographic network, with similarly high values in Pagerank and betweenness centrality in the case network, particularly during epidemic years (2017 mainly).
It is important to note that network descriptive statistics is highly susceptible to the projection used to express the output of TDA onto a map and to the choice of parameters in the TDA algorithm. Here, we propose a method to interpret TDA in the context of genetic and epidemic variables for georeferenced cases. However, other possible projections should be explored in further work, which will also confirm whether the observed patterns of case betweenness and Pagerank centrality and epidemic and inter-epidemic intervals adequately describe the dynamics of the malaria transmission unit. Furthermore, as data become available, studies with larger sample sizes will confirm whether our results are biased by the limited number of sequenced infections.
Interestingly, epidemics were not found to be due to the expansion of a single clonal population57,58 but to an increase of all three parasite populations circulating in the area (Fig. 4). This suggests that epidemic peaks are due to an increase in mosquito density due to the opening of breeding sites in mining areas51,59, an increase in human population densities at mining sites and the human mobility across the mining area and the populated settlements.
Although the number of cases originating outside the defined 25 km radius transmission unit is small (3%), its importance is revealed by the presence of cases originating in Venezuela. Interviews with patients at the diagnostic post revealed that infected individuals acquired the infection at gold mining areas in Venezuela. Parasites derived from those individuals carried genotypes associated with resistance to chloroquine, pyrimethamine and sulfadoxine that were previously unseen in the Pacific coast of South America57,60,61. Data from Colombia’s National Statistics Department show that in 2005, 2.8% of households in Guapi had some type of experience as international migrants, of which 47.2% had Venezuela as destination62. Colombian migration to Venezuela was mainly due to economic or humanitarian reasons. However, due to Venezuela’s ongoing crisis more than 20,000 Colombian residents have returned from Venezuela, in particular since 2015. In addition, between 2015 and 2017 more than 550,000 Venezuelan citizens migrated to Colombia, posing an important challenge to the health system and to Colombia’s malaria control effort9.
We observe that more than 80% of multiclonal infections (15 out of 18) occurred during epidemic peaks. This suggests that under this transmission pattern, epidemic peaks constitute potential recombination hotspots. If this is the case, we would expect successive waves of clonal expansion followed by recombination events. Since the net effect of recombination between two different clones would be a decrease in the branch length linking the two clones and an increase in the branch length within a particular clonal population, the dendrogram in Fig. 3 suggests this possibility. High density genotyping to infer segmental Identity by Descent (IBD) would be desirable to map recombination events that may prove to be hallmarks of epidemic events.
These results may have important consequences for planning control interventions. For example, Guapi operates as a receptor of infections, maintains some local transmission and constitutes a hub for the dissemination of infections. However, transmission is localised to a few neighbourhoods and vector control may constitute an efficient control measure (e.g. identification and control or elimination of mosquito breeding sites). For sites with high Pagerank centrality, where localised epidemics occur, timely diagnosis and treatment may be required to suppress further transmission while sites with high betweeness but low Pagerank centrality may require continuous surveillance of cases and active case detection. Also, as epidemics occur at sites that are the most populated in the area, close to sites where mining activities take place and in areas that are connected through river or seaways, which allows for the dissemination of infections, microscopy diagnostic posts should be considered as well as a system capable to reach the mining population.
An operational definition of a malaria control unit requires an evaluation of the efficacy of control measures and the factors, intrinsic (biological) and extrinsic (social, political, demographic, etc), hampering the malaria control effort. This would allow determination of the feasibility and cost of gradual and attainable goals towards malaria elimination. As an example of the difficulties encountered by the control system we estimated the size of the asymptomatic reservoir in the neighbourhood of Guapi that has historically contributed most of the malaria cases and in a hamlet representative of many of the locations where endemic malaria transmission occurs. We also determined the nature of the genotypes associated to drug resistance circulating in the area and the frequency of the Pfhrp2/3 deletion that contributes false negative diagnostic results when using RDTs.
The number of identified asymptomatic individuals was low (<5%) as compared with previous studies in the Colombian Pacific coast (>20%). This may be due to the different sampling methods. Here, we actively searched for asymptomatic individuals from a randomly chosen sample in Santa Mónica, Guapi or from the total population of El Cuerval, Timbiquí, at different time points. Others used a reactive active search strategy whereby asymptomatic individuals were searched for in a given perimeter from an identified symptomatic index case during an epidemic peak63. Here we show that the number of asymptomatic individuals decreased as malaria incidence decreased. The epidemiological relevance of asymptomatic individuals may therefore depend on the rate of natural clearance of parasite infections (average persistence of asymptomatics) and the mosquito biting rate.
All parasites displayed genotypes associated to chloroquine resistance and 80% displayed genotypes associated with resistance to pyrimethamine and sufadoxine. This is important in terms of the efficacy of the control measures since, as recently documented, 32% of individuals in a location in the south Pacific coast were found to use chloroquine as auto medication according to Saker-Solomon tests directed at detecting chloroquine in urine38. Likewise we have surveyed local pharmacies in Guapi and found that both chloroquine and sulfadoxine-pyrimethamine can be obtained over the counter outside the health providers’ system.
Finally, 6% of parasite samples analysed exhibited a deletion in both the Pfhpr2 and 3 genes suggesting that at least 6% of diagnostic tests performed with RDTs within the transmission unit are false negatives.
An understanding of malaria transmission at the micro epidemiological level can define a malaria transmission unit by assessing the effective range of the catchment facility, the heterogeneity of transmission and identifying sources and sinks of infection. Such information is necessary to direct the malaria control effort, but in order to build an effective, spatially defined, operational malaria control unit, an assessment of the efficacy of the individual control measures that takes into account entomological, environmental, social, and demographic factors, as well as factors limiting the accessibility to the health system or limiting the establishment of efficient control measures is also required. This work is an important step to defining such a control unit for the Guapi region, and some of the findings can be extrapolated to similar urban areas along the Pacific Coast.
Source: Ecology - nature.com
