Our large genomic epidemiology investigation of P. vivax in Colombia, demonstrates patterns of within-host and population diversity consistent with low endemicity between 2013 and 2017. Although this level of endemicity is conducive to timely P. vivax elimination, there is evidence of low-level shared reservoirs of infection with neighbouring countries, and presence of known drug resistance-associated variants as well as several potential new adaptations. Further details on the observed genetic patterns and their implications for P. vivax surveillance and treatment efficacy are discussed.
A major concern for the Colombian National Malaria Control Program (NMCP) is whether local interventions are effective in reducing parasite transmission. Whilst traditional entomological and parasitological measures of infection prevalence and incidence are critical, asymptomatic and subpatent reservoirs are impossible to capture with these approaches. The dormant liver stages of P. vivax further complicate measures of the burden of P. vivax46. Parasite population genetic features provide useful insights on local transmission to complement more traditional measures. For instance, high within-host infection diversity is one such feature, generally reflecting high endemicity 46,47. Our finding that polyclonal infections occurred in only 12% of Colombian isolates aligns with low rates of P. vivax transmission, comparable to Malaysia (16% polyclonal infections) during its malaria pre-elimination phase 48. To put these figures in context, high transmission regions of Papua Indonesia, Papua New Guinea, the Greater-Mekong subregion, and Ethiopia report between 30 and 60% polyclonal infections 5,6,34,49−51. We also observed that over 60% of the polyclonal infections in Colombia comprised closely related clones such as siblings or half-siblings (> 25% genomic IBD), suggesting recent shared parentage. These results infer a higher frequency of co-transmission as opposed to superinfection events in Colombia, as might be expected as transmission declines. However, using similar methods, a study in a moderately high transmission region of Ethiopia reported a similar frequency (57%) of co-transmission events 34. With limited information from other endemic regions, it’s unclear the degree to which reactivated hypnozoites, as opposed to reinfections, determine within-infection relatedness patterns in P. vivax. Further research in this area using genome phasing approaches or single cell sequencing, will enable a more contextualised view of the relationship between within-host relatedness and P. vivax endemicity 52. The current data contribute an important baseline for Colombia against which future data can be evaluated.
A wide variety of ecological patterns and associated malaria epidemiology have been described in Colombia, highlighting the importance of subnational interventions adapted to local needs 1,12. Our investigations of within-host diversity at the departmental scale found modest evidence of heterogeneity between sites. Chocó has historically harboured high levels of malaria and is one of the priority areas for malaria elimination in Colombia. Over 60% of the polyclonal infections in Chocó appeared to reflect superinfections in contrast to 0% in Córdoba. Chocó also presented a higher frequency of polyclonal infections (23%) than Córdoba (7%), but the sample size in Chocó was constrained (n = 13) and the difference was not statistically significant. Nonetheless, these trends call for ongoing monitoring of transmission reduction efficacy in Chocó. Whilst our study was moderately granular in spatial scale, our findings highlight the potential for genetic epidemiology approaches using more high-throughput methods such as barcode genotyping to capture important infection dynamics.
Population structure and relatedness provide an additional genetic measure yielding insights into local malaria transmission dynamics 46,53. A notable genetic feature in Colombia was the high frequency of closely related infections, including the large K2 cluster (n = 14 infections with IBD ≥ 50%) observed in Córdoba. These patterns indicate a high degree of inbreeding and have been observed in other low endemic P. vivax populations, including Malaysia and Panama as infection prevalence declined and opportunities for outcrossing diminished 48,54. Similar patterns have also been described in the P. falciparum population in Colombia during a period when cases were declining 8,55. The high degree of inbreeding, particularly in Córdoba, is therefore promising regarding Colombia’s goals of achieving malaria elimination certification. However, our data is from 2013-15 and ongoing surveillance will be critical to avoid resurgence from highly resilient or adaptive strains. Parasite populations with low levels of outcrossing may be more amenable to the emergence of drug-resistant malaria strains 56. Infrequent outcrossing is favourable when supporting variants are required to overcome fitness costs of drug resistance-conferring mutations. Our and similar P. falciparum studies in Colombia demonstrate a high potential for infection lineages to persist over multiple years 55. P. vivax studies in Malaysia and Panama have reported similar trends, with infections persisting up to a decade 48,54.
Surveillance of P. vivax drug resistance is challenged by limited information on the molecular markers 31. Amongst the few described markers of resistance, we found no evidence of the pvmdr1 copy number duplication (0% across Colombia) that has been associated with mefloquine resistance, suggesting that it may be a suitable alternative to chloroquine (as a partner drug in artemisinin combination therapies, ACTs) 37,38. However, the number of samples in our study that were suitable for copy number evaluation was constrained (n = 22, 39%) and hence further surveillance is warranted. Of note, investigations of P. falciparum infections collected at a similar time and location (enrolments in the Pacific coast and Cauca River regions in 2015) identified ~ 7% prevalence of pfmdr1 copy number amplifications 57. The selective pressure on pfmdr1 justifies continued close surveillance of pvmdr1 amplification in the P. vivax population, which may receive inadvertent drug pressure.
Although sulfadoxine and pyrimethamine (SP) are not currently recommended as first-line policy for any malaria species in Colombia, SP has been used to treat P. falciparum in the past, in combination with CQ (from 1981–1998) and AQ (from 1998–2008). In 2008, policy was changed to ACTs owing to widespread therapeutic failures 58,59. As SP has been shown to have a positive impact on infant birth weight when used in intermittent preventive treatment in pregnancy (IPTp), even to some degree when P. falciparum SP resistance variants are present, there is a potential public health application for SP 60. Our results show evidence of reduced SP efficacy in P. vivax in Colombia. Frequencies of the pvdhfr S58R + S117N double mutant ranged from 80–100%, and the pvdhps A383G mutant ranged from 66–100% across Colombia. However, there were no triple or quadruple pvdhfr mutants and no double pvdhps mutants, suggesting that full-grade SP resistance is not common. Similar patterns have been reported in P. falciparum in Colombia, with a 2018 study in Chocó reporting 100% double pfdhfr mutants, and 43% single pfdhps mutants 61. Our genomic data also showed evidence of similar trends in P. vivax in neighbouring Brazil and Peru, and no SP resistance variants in Mexico. In contrast to the Americas, Thailand and Indonesia exhibited high mutation rates, potentially reflecting differences in drug use between these regions. In addition to the known SP resistance candidates, we identified a pvdhfr T98_N108del that was prevalent (~ 60%) in Colombia but absent in the other populations. The functional impact of the pvdhfr T98_N108del remains unclear but its high prevalence in Colombia justifies further investigation. Further investigation is also required to understand the utility of the antifolates in IPTp for P. vivax amidst different pvdhfr and pvdhps backgrounds.
Chloroquine (CQ) remains the frontline treatment for blood-stage P. vivax infections in Colombia, but no validated markers of clinical efficacy have been identified for this species 31. Following evidence of clinical failures, CQ was removed from P. falciparum treatment policy in Colombia in 2008, but the resistance-conferring pfcrt K76T mutation remains prevalent 61. In addition to pfcrt, hard selective sweeps postulated to reflect CQ pressure have also been reported at pfaat1 in the Pacific Coast of Colombia 8. In contrast to P. falciparum, clinical surveys of CQ efficacy against P. vivax have shown failure rates below 3% in Colombia 62. There was no evidence of selection in either of pvcrt-o or pvaat1 in our study. For reference purposes, we documented the frequencies of the pvmdr1 Y976F and F1076L mutations that have been reported to be minor determinants of CQ resistance in P. vivax, finding 0–20% frequency across Colombia. These frequencies are comparable to Thailand, where ~ 10% infections carried the Y976F mutation, and < 10% CQ failure rate was reported 13. In contrast, 100% of the Papua Indonesian infections carried the Y976F mutation; a population where > 60% chloroquine failure was reported by day 28 in the early 2000s 15. However, the interpretation of these findings in the context of CQ’s efficacy in Colombia is constrained and will require validated markers.
Using information on orthologous genes involved in drug resistance in P. falciparum, and hypothesis-free methods to detect signals of selection, we identified several candidate markers of resistance and other adaptations. The candidates include non-synonymous variants in orthology-based drug resistance candidates pvmdr2, pvmrp2, and plasmepsin IV noted for substantial inter-population differences in frequency. Using haplotype-based signals of selection, putative adaptations were identified in genes involved in a range of other functions including vitamin B6 synthesis (pyridoxine biosynthesis protein PDX2), intracellular transport (oligomeric Golgi complex subunit 4), immune evasion (6-cysteine proteases P12 and P47, and PIMMS43), and malaria transmission (metacaspase-2) 43–45. However, signals of extended haplotype homozygosity can be complex to decipher, and these candidates will require further exploration in functional studies.
As Colombia invests efforts towards malaria elimination, detection of imported cases and other key reservoirs will be critical to avoid resurgence. Imported malaria is a particular challenge for P. vivax, where the dormant liver stages and highly persistent subpatent and asymptomatic infections can promote infection spread and confound the accuracy of travel histories. Our study identified two Colombian infections that exhibited higher genetic relatedness to infections from neighbouring Peru than to the local Colombian population, suggestive of cross-border spread. Our data also shows comparable prevalence between Colombia and Peru at most of the putative resistance-conferring mutations in pvdhfr, pvdhps and pvmdr. However, the evidence for other P. vivax adaptations that differ between Colombia and Peru, highlight the risk of introducing new variants into either population that may undermine local interventions.
Our data demonstrate the potential of molecular approaches to capture new insights on local P. vivax transmission and adaptations within Colombia, as well as cross-border spread. Such clinically relevant information can support the NMCP in their surveillance and response strategies to fast-track P. vivax elimination.