Climate determines transmission hotspots of Polycystic Echinococcosis, a life-threatening zoonotic disease, across Pan-Amazonia

Abstract

Polycystic Echinococcosis (PE), a neglected life-threatening zoonotic disease caused by the cestode Echinococcus vogeli, is endemic in the Amazon. Despite being treatable, PE reaches a case fatality rate of around 29% due to late or missed diagnosis. PE is sustained in Pan-Amazonia by a complex sylvatic cycle. The hunting of its infected intermediate hosts (especially the lowland paca Cuniculus paca) enables the disease to further transmit to humans, when their viscera are improperly handled. In this study, we compiled a unique dataset of host occurrences (~86000 records) and disease infections (~400 cases) covering the entire Pan-Amazonia and employed different modeling and statistical tools to unveil the spatial distribution of PE's key animal hosts. Subsequently, we derived a set of ecological, environmental, climatic, and hunting covariates that potentially act as transmission risk factors and used them as predictors of two independent Maximum Entropy models, one for animal infections and one for human infections. Our findings indicate that temperature stability promotes the sylvatic circulation of the disease. Additionally, we show how El Niño-Southern Oscillation (ENSO) extreme events disrupt hunting patterns throughout Pan-Amazonia, ultimately affecting the probability of spillover. In a scenario where climate extremes are projected to intensify, climate change at regional level appears to be indirectly driving the spillover of E. vogeli. These results hold substantial implications for a wide range of zoonoses acquired at the wildlife-human interface for which transmission is related to the manipulation and consumption of wild meat, underscoring the pressing need for enhanced awareness and intervention strategies.

Keywords: ENSO; climate change; modeling; zoonotic diseases; zoonotic spillover.

Fig. 1.

(A) Diagram illustrating the epidemiological cycle of E. vogeli. It is a two-host cycle that requires intermediate hosts (in the wild, mainly the C. Paca rodent or the Dasypus novemcinctus) and final hosts (in the wild, the dog Speothos Venaticus). Humans can also act as intermediate hosts and domestic dogs as final hosts, thus creating an alternative, communal, cycle. (B) Distribution of available evidence of E. vogeli infections. Human infections are shown in purple and animal infections in blue. The background depicts evidence of sylvatic hosts, presented as dimmed presence points.

Fig. 2.

Graphical summary of the covariate construction process. (A) Construction of key wild hosts habitat suitability maps: (Up) Presence records of the four main wild animals involved in the sylvatic circulation of PE. (Down) Derived map of E. vogeli natural suitability, which accounts for the common presence of intermediate and final hosts and thus shows which places are more suitable for the parasite to complete its lifecycle. Overlaid in blue are animal infection records, which are seen to align with the high suitable (green) areas. The natural suitability map was generated using Ecological Niche Models run on the shown presence records. (B) Construction of C. paca’s hunting pressure map. (Up) Available information on paca’s hunting pressure, assembled by the authors of this study (Down) Continuous regional C. paca’s hunting pressure across Pan-Amazonia obtained through a kriging algorithm (C) Construction of climate variability and climate change variables. (Up) Monthly temperature data in the Peruvian amazon (longitude -70, latitude -3) spanning from 1991 to 2021. A Singular Spectrum Analysis (SSA) decomposition into trend, QB and QQ is shown on top. (Down) Temperature trend of the last 30y across Pan-Amazonia obtained with an SSA decomposition run for all pixels in the region. This was performed for both Temperature and for Absolute Humidity and a map of trend, QB and QQ was derived for each.

Fig. 3.

Animal model results averaging results of the five best-performing sub-models. (A) Model predictions overlaid on the Natural Suitability map. Most of the areas selected by the sylvatic circulation model coincide with the suitable areas (shown in light green). The dark green areas represent locations with the highest risk of E. vogeli expansion, as they harbor the appropriate conditions for its circulation and thus require special attention. Some areas in southeastern Amazon are selected by the sylvatic circulation model but are not part of the main suitable areas for the disease's circulation (indicated in yellow) (B) Variable importance derived from a permutation test. (C) Response functions that illustrate the variation in model predictions as we sweep through the full range of values of the variable under study while keeping the rest at their mean or modal value.

Fig. 4.

Human model results averaging the results of the five best-performing submodels. (A) Model predictions (B) Variable importance derived from a permutation test. (C) Response functions illustrate the variation in model predictions as we sweep through the full range of values of the variable under study while keeping the rest at their mean or modal value.

Fig. 5.

Compressed quasibiennial (QB_T) and quasiquadrennial (QQ_T) modes of temperature. Each pixel represents the sum of the absolute value of the quasibiennial (A) and quasiquadrennial (B) modes of temperature throughout the period 1991 to 2020.

Fig. 6.

ENSO effect on hunting practices across Pan-Amazonia. (A) Mean monthly hunting offtake across studied sites: Original time-series and low-frequency component identified using SSA. (B) Scale-dependent correlation analysis between the low-frequency component of the hunting records and the ENSO3.4 index. Consistent strong couplings (with a correlation coefficient of approximately −1) are observed during the main ENSO extreme events (both warm and cold), as indicated by the blue diagonal. Analyses are performed using windows of 3 y. (C) Lags at which the maximum absolute correlation is achieved in the 3-y windows. The coupling during the main ENSO events shows maximum values at lags of around 8 to 10 mo. Notice that the 2009-2010 was a Modoki-type EN (a central-pacific EN) and therefore yields different lags and correlation coefficients in the analyses.