Seasonal host community dynamics constrain the risk of parasite transmission between migrant and resident species

Seasonal migration shapes the community dynamics that influence pathogen transmission between migrants and resident species. While theoretical and empirical evidence has accumulated, whether migration increases or decreases the risk of cross-species infection remains inconclusive. We studied how the seasonal arrival and departure of a single avian migrant species change the composition of local communities in breeding areas affecting the haemosporidian infection dynamics. The seasonal reordering of resident species abundances induced by migrants, minimizes the opportunities for contact between highly infected hosts and susceptible species, either migrants or residents, thereby limiting the transmission of parasites to occasional spillover events. The occurrence of spillover dynamics during the seasonal sympatry between migrants and residents provides a plausible explanation that reconciles empirical inconsistencies in the intersection of animal migration and infection risk at the host community level. Our findings underscore the critical role played by seasonality in shaping infection dynamics in migratory systems.

Methods
Between October 2015 and September 2019, we studied local communities of closely related bird species (Order: Passeriformes) and their haemosporidian parasites at five study sites in Southern Chile (Fig. 1A). Each of these sites represents different habitat types, where the long-distance migrant Elaenia albiceps occurs in sympatry with local resident species during the austral spring and summer, between October and March (hereafter, breeding season). The local absence of E. albiceps occurs during the austral fall and winter, between April and September (hereafter, non-breeding season). Llancahue (39°84’ S; 73°14’ W) is an evergreen forest, supporting remnants of old-growth temperate rainforest and secondary forests (Biscarra et al. 2021). Quempillén (41°87’ S; 73°77’ W) is a secondary forest with dense understory vegetation covered by a mixture of native berry trees, native bamboo, fern species, and exotic shrublands (Díaz & Armesto 2007). San José (39°55’ S; 72°98’ W) is an anthropogenic habitat in which native temperate rainforests have been largely replaced by agricultural lands and exotic tree plantations, followed by recent natural regeneration (Zamorano-Elgueta et al. 2015). Santa María (39°68’ S; 73°19’ W) and Teja Norte (39°79’ S; 73°26’ W) are riparian habitats within the Río Cruces wetland. While Santa María is located on the border of a temperate forest patch and presents mainly low marsh vegetation maintained by cattle grazing, Teja Norte has a dense reedbed surrounded by scrubs. For logistical reasons, Llancahue was subsampled during winter and San José was sampled for only one year (Table S1; Fig. S1).

Individual birds were systematically trapped once per month with mist nets operating for 5 h each, beginning at sunrise. Individuals were identified at the species level and marked with numbered aluminum leg rings to record individual-based abundance. Blood samples were obtained from a random subset of the trapped bird species by puncturing the brachial vein and were deposited in a 2 ml tube with 96% ethanol for further molecular analyses. Additionally, a drop of each blood sample was used for parasite quantification in the blood smear, fixed with 100% methanol, and stained with a three-step quick stain (Differential Quik Stain Kit; Polysciences, Warrington, PA, USA). In infected individuals, the intensity of infection was quantified using optical microscopy with 1000X amplification and was reported as the number of haemosporidian gametocytes in 10,000 erythrocytes. DNA was extracted from blood samples using an E.Z.N.A. Blood DNA Kit (Omega Bio-tek, Norcross, GA, USA) according to the manufacturer’s protocol. We screened all DNA samples for infection with either of the haemosporidian genera Plasmodium, Haemoproteus, and Leucocytozoon using nested PCR that targets a ~1,068 bp fragment of the cytochrome b gene (cyt b) of the parasite mitochondrial DNA, using the primers DW2 and DW4, followed by DW1 and DW6 (Bensch et al. 2009; Perkins & Schall 2002). All PCR runs were performed using positive and negative controls. PCR products were purified (E.Z.N.A. Cycle Pure Kit, Omega Bio-tek) and then sequenced in both directions with an ABI3100 Genetic Analyzer (AustralOmics, Valdivia, Chile). All haemosporidian sequences obtained were submitted to GenBank (accession numbers are pending upon publication). Individual sequences were compared with the BLAST database and grouped by haemosporidian genus. Sequences were aligned using MAFFT v7 and evaluated in a median-joining haplotype network to distinguish any host or seasonal distribution using Network 5.0.1.0 (Fluxus Technology Ltd., Suffolk, England) and Network Publisher v.2.1.2.3 (Fluxus Technology Ltd.).

Seasonal host community dynamics
The dynamics of host communities were analyzed by considering the accumulated abundance of each species in a local community over a three-month capture period. Breeding communities were formed by individuals captured between October and December and between January and March, while non-breeding communities were formed by individuals captured between April and June and between July and September. To avoid pseudoreplication in community estimates, individuals recaptured within the same three-month period in an annual cycle were excluded from the analyses (Hurlbert 1984; Schwarz 2002). To visualize the two-dimensional distribution of pairwise Bray-Curtis dissimilarities between samples, nonmetric multidimensional scaling (NMDS) with relative abundance transformation was performed. Permutational multivariate analysis of variance (PERMANOVA) was used to test for differences in community composition using the “vegan” R package (Oksanen et al. 2022). To gain a deeper understanding of the processes underlying changes in community composition across seasons in the presence of E. albiceps, we calculated two community-change metrics based on rank abundance curves (RACs) that incorporate both species identities and abundances (Avolio et al. 2015, 2019). We used Evar as a measure of variance in the relative abundance of species in a community (Smith & Wilson 1996) and calculated the change in Evar between seasons (ΔE) as the difference in evenness between the breeding and non-breeding seasons. Additionally, we calculated the average change in the rank of each species in the community (ΔR), which considered the gains and losses of species in each evaluated season, incorporating species turnover. All calculations of changes in RAC-based measures between seasons were conducted using the “codyn” R package (Avolio et al. 2019; Hallett et al. 2016). We tested for statistically significant differences in the drivers of seasonal community changes among communities using Welch’s two sample t-tests.

Host contribution to parasite transmission
To determine the potential reservoirs of infection for each lineage of haemosporidian circulating within the communities, we estimated the relative contribution of each host species to the overall infectious pool (π) following Streicker et al. (2013). Host species were categorized as key hosts, indicating their role as infection reservoirs if π ≥ 0.5. Conversely, values of π < 0.5 suggested that these host species were only occasionally infected through spillover dynamics. To understand how host species contribute to parasite transmission and persistence, we explored three asymmetries that underlie host heterogeneity: the degree of asymmetry in host abundance (θA) which serves as a proxy for encounter opportunities between host species, the degree of asymmetry in the prevalence of infection (θI) which represent the exposure of host species to infected vectors, and the degree of asymmetry in infection intensity (θS), which act as a proxy for the susceptibility to infection and the capacity to shed parasites (Streicker et al. 2013). In order for a host species to be classified as a key host, at least one of these asymmetries must be significantly greater than 1.