Remote native communities represent a challenge for the malaria elimination strategies currently deployed in the Peruvian Amazon region1. In this study, we provided insights on the microepidemiology, resistance markers, and pfhrp2/3 gene deletion in NJ. To our knowledge, this is the first report of malaria genomic surveillance in a native community in this region.
NJ has an exceptionally high malaria prevalence and persistence, exceeding previous reports in other endemic communities in the Peruvian Amazon46. Despite continuous diagnosis, the number of malaria cases remained unchanged during the study period. This result contrasts with a previous report in Mazan, where four consecutive ACD visits quickly reduced the disease's prevalence20.
Population genetics analysis showed that the Pv population in NJ had moderate diversity, with a high proportion of polyclonal infections, without differences between 2019 and 2020. Several Pv lineages were present in NJ during the study period, commonly observed in high-transmission populations, favoring genetic recombination and increasing genetic diversity47,48. Similar to NJ, previous reports featured the Pv population in Peru as heterogeneous, with moderate to high levels of genetic diversity and a high proportion of polyclonal infections12–14, although a recent report showed a monoclonal population with low diversity in Iquitos and Mazan5.
The MINSA intervention in 2019 did not reduce the Pv burden in NJ. Only the ACD visits failed to decrease the cases and did not affect complexity of infection. We list here some hypotheses related to this observation, i) most adults in NJ travel out of the community for field work or social activities to nearby areas, ii) at the same time, children and their caregivers stay in the community, reflected in the low coverage in each ACD visit (Fig. 3), iii) the continuous mobility negatively affected treatment compliance and increases the likelihood of parasite importation into NJ, which can enrich the genetic diversity and the persistence of the disease in remote communities, as previously reported for the communities of the Alto Juruá river in Brazil49.
The transmission dynamics of Pf in NJ were different compared to Pv. The analysis showed a population substructure with 3 clusters reflecting temporal changes. Cluster 2 (February 2020) clearly differentiated from parasites detected earlier and later in the community. In addition, the IBD network showed that some parasites from this cluster were related to parasites in Andoas, a city close to NJ. This suggests that this cluster corresponds to a Pf outbreak that appeared in February 2020 in NJ. We hypothesize that new parasites were introduced (potentially from Andoas or the vicinity and were not related to local NJ parasites), rapidly spread in NJ (increase of samples collected by PCD at that time, high proportion of polyclonal infections), and then were successfully controlled (not related with later parasites). We cannot further assess the hypothesis as we lack mobility data. Nevertheless, Pf outbreaks related to similar introduction events have been reported in non-endemic regions in Peru using molecular surveillance7,8, and in an indigenous area in Amazonas, a Peruvian region next to Loreto, using epidemiologic data9.
On the other hand, Cluster 3 (March to May 2020) showed low diversity, high differentiation to other clusters, and unique features compared to the other clusters. Following our hypothesis, controlling the outbreak in Feb 2020 generated a bottleneck event in the Pf population in NJ, like those previously reported in Colombia and Honduras-Nicaragua50,51. Expert microscopists were present at the health post during the study period, improving diagnosis performance in the community, followed by treatment administration following national guidelines. Opportune diagnosis and treatment shut down this local outbreak.
Given NJ's heterogeneous Pf and Pv transmission dynamics, tailored strategies should be proposed to control malaria in this scenario. In addition to better educational campaigns and vector control measures, the focus should be on working adults (the mobile population) in NJ, which must be diagnosed and treated promptly when returning to the community. This would prevent the potential spread of new parasite populations and reduce malaria transmission in the community. Malaria treatment lasts 3–7 days in Peru18, which may represent a problem for monitoring treatment compliance in mobile populations. Considering the limited availability of expert microscopists in remote areas, administering one standard treatment for any malaria infection is an option52. A single-dose tafenoquine scheme, along with glucose-6-phosphate dehydrogenase (G6PD) activity monitoring, has been tested recently in Peru and other countries and showed efficacy for the radical cure of Pv malaria53,54 and can be potentially used in the future. In addition, similar studies should work in testing new single-dose treatments for Pf55.
MINSA has implemented the training of community health agents (CHAs), people chosen by their community to develop disease prevention actions and support for the diagnosis and treatment, into the current NMEP1. Similar approaches have been used to control malaria in the mobile population in Myanmar56 and in an indigenous area in Panama57, showing promising results in each case. In Peru, CHAs better understand the dynamics of human mobilization, facilitating the work with the population. However, an adequate number of CHAs, enough resources, and continuous monitoring of their work are necessary to ensure the success of this strategy to control malaria. In addition, due to the geographical location and the socio-cultural context, the participation of the different stakeholders is also crucial58. We also recommend conducting qualitative studies in indigenous areas to understand their point of view about malaria and their mobility behavior59 for a better strategies adaptation.
For Pf, the most frequent haplotypes were associated with drug resistance: RICNI (pfdhfr) and SGEGA (pfdhps) to SP60, NDFCDY (pfmdr1) and SVMNT (pfcrt) to CQ61,62. An increase in the frequency of these haplotypes has been reported in recent years5,11. Fortunately, none of these are drugs used in the current treatment scheme against Pf in Peru18. In this regard, no validated mutation for artemisinin resistance has been found in this study, being concordant with previous studies5,11,63.
The Pv AmpliSeq assay includes putative resistance-associated markers. Most correspond to orthologs to genes in Pf associated with resistance13. Only two markers are validated for Pv: pvdhps (positions 383, 553) and pvdhfr (positions 57, 58, 61, 117, 173) for SP resistance64. The A553G mutation in pvdhps and the S58R, S117N, and I173L mutations in pvdhfr were present in 33–95% of the samples in NJ and Mazan, comparable to a previous report5. The presence of these mutations in Pv may be explained by exposure to SP due to mixed infections or missing diagnosis65, although SP ceased to be used against Pf more than 20 years ago66. In the case of pvmdr1, the LMYF haplotype was the most abundant (69–77%) in NJ and Mazan, which is consistent with previous surveillance reports on this gene5,67.
The first report of the deletion in the pfhrp2/3 genes and its relationship to failure in rapid diagnostic tests was in Peru16. Since that report, the frequency of these deletions in the country has increased after 2012 up to 70%10,11,17. In contrast, the moderate proportion of samples with pfhrp2/3 genes (50–67%) showed that deletions have not yet spread to hard-to-reach communities such as NJ. In our study, the double deletion was only predominant in Mazan.
In this study, we found different results in pfhrp2/3 genotyping between PCR and the Pf AmpliSeq assay. There are also some differences in the regions targeted by these two assays that could, for a large part, explain these contrasting results. The Pf AmpliSeq assay targets the genes with several amplicons that together span the full length of the genes. In the analysis applied here, we determined a ratio of the depth of the pfhrp2 or pfhrp3 amplicons compared to all other amplicons targeted in the assay, and a deletion is determined when multiple amplicons have decreased depth11. When amplicons for one gene yield contrasting results (which could occur when only part of the gene is deleted), an inconclusive result is generated. The PCR used in this work only targets only exon 2 of each gene in a conventional PCR protocol16,44.
In addition, there are reports that showed different breakpoints for the deletions in both genes, that can impact the accuracy of both analysis approaches applied here68. In the case of pfhrp2, the breakpoint often encountered in Peru are in exon 2 or the neighboring intron, and therefore partial deletion of pfhrp2 gene is expected, especially deletion of exon 1. On the other hand, for pfhrp3, the deletion in Peru is explained by a recombination with chromosome 11, thus a deletion of the entire gene is expected. Further work is needed to structurally characterize pfhrp2/3 deletions in Peru, particularly in remote areas not previously included in studies characterizing these genes. Here, we have detected different Pf clusters than those from previously studied areas5,11. Using updated structural variation information from more in depth studies in these regions, we can re-evaluate and optimize the approach of which amplicons to use in the determination of deletions based on the read depths in the Pf AmpliSeq assay. Besides that, homology, repetitiveness, and high %AT sequences of both genes could have complicated alignment of reads in certain regions of the genes.
This work has some limitations. Initially, we planned to carry out a treatment efficacy study in the community, following a cohort for two years. At the beginning, samples were collected as part of MINSA interventions, and no additional epidemiological data were collected. Unfortunately, all work stopped due to the COVID-19 pandemic. Second, we used sampling by convenience, leading to the analysis of different sample sizes by place and samples with different collection methods and times. Because of high sequencing costs11, we only performed one run per species assay. Today, collaborating with international consortia (such as MalariaGEN) for sequencing is more common in countries with high costs and scarce infrastructure. In addition, the heterogeneous malaria transmission in Peru must be considered to make molecular surveillance representative of the country11,13. Moreover, the pfcrt gene could not be genotyped in 35% of samples. Since in previous studies, this gene performed well, we should investigate the genetic variability in the primer regions of the amplicons.
In conclusion, NJ, an indigenous remote community in the Peruvian Amazon with high transmission and persistent malaria, has local Pv parasites with modest genetic diversity. In contrast, the Pf population has low genetic diversity and temporal clustering. An outbreak represented by a unique genetic cluster was detected in February 2020, which was controlled through timely microscopy diagnosis by expert personnel and treatment administration. The Pf population carried mutations associated with CQ and SP resistance but not with artemisinin. Additionally, the presence of pfhrp2 and pfhrp3 genes was frequent in NJ and other remote areas, except in Mazan, where the double deletion predominated.
This information highlights the importance of regular molecular surveillance in indigenous remote communities, preferably linked to MINSA activities, regarding imported cases, transmission dynamics, and spread of resistance, to promptly adapt malaria elimination strategies in Peru. In this regard, we showed that AmpliSeq assays, with some improvisations, can be helpful for NMEP.