Multilevel Models for the Distribution of Hosts and Symbionts

Abstract

Symbiont occurrence is influenced by host occurrence and vice versa, which leads to correlations in host-symbiont distributions at multiple levels. Interactions between co-infecting symbionts within host individuals can cause correlations in the abundance of two symbiont species across individual hosts. Similarly, interactions between symbiont transmission and host population dynamics can drive correlations between symbiont and host abundance across habitat patches. If ignored, these interactions can confound estimated responses of hosts and symbionts to other factors. Here, we present a general hierarchical modeling framework for distributions of hosts and symbionts, estimating correlations in host-symbiont distributions at the among-site, within-site, among-species, and among-individual levels. We present an empirical example from a multi-host multi-parasite system involving amphibians and their micro- and macroparasites. Amphibian hosts and their parasites were correlated at multiple levels of organization. Macroparasites often co-infected individual hosts, but rarely co-infected with the amphibian chytrid fungus. Such correlations may result from interactions among parasites and hosts, joint responses to environmental factors, or sampling bias. Joint host-symbiont models account for environmental constraints and species interactions while partitioning variance and dependence in abundance at multiple levels. This framework can be adapted to a wide variety of study systems and sampling designs.

Using multi‐response models to investigate pathogen coinfections across scales: Insights from emerging diseases of amphibians

“…Model parameters were estimated by drawing 1000 samples from their joint posterior distributions using the Markov Chain Monte Carlo (MCMC) algorithm implemented the MCMCglmm package (Hadfield 2010) in R (R Core Team 2013) (see prior distributions and MCMC chain specifications in Appendix S1). Although the outlined framework does not depend on Bayesian model-fitting, current limitations of existing likelihood-based implementations for modeling correlations and associations emphasize the additional utility of Bayesian procedures for assessing parasite correlations across scales (see Joseph et al 2016).…”

“…Many factors influence parasite exposure and susceptibility both among host individuals (e.g., age, sex, diet, genetics) and across host populations or species (e.g., environmental, spatial, or epidemiological factors) (Wilson et al 2002; Hawley and Altizer 2011; Hellard et al 2015). In some cases, parasite correlations observed at one scale (e.g., infections within individual hosts) maybe absent or reversed at other scales (e.g., average infection load among sites), emphasizing the importance of explicitly considering scale (the inappropriate application of correlations at one scale to another scale is sometimes referred to as the ‘ecological fallacy’) (Joseph et al 2016). Thus, even in the absence of interspecific parasite interactions, spatial or temporal non-independence of the factors influencing infection can generate correlations in parasite occurrence or abundance.…”

Positive biotic interactions in freshwaters: A review and research directive

“…Species distribution modelling is a suite of methods for predicting how species occur across landscapes at different spatial scales (Elith & Leathwick, 2009). Biotic interactions are being increasingly incorporated into species distribution models (Joseph, Stutz, & Johnson, 2016;Wisz et al, 2013), but only recently have positive interactions been considered (Afkhami, McIntyre, & Strauss, 2014;Duffy & Johnson, 2017;Filazzola et al, 2017). These models provide a framework to directly test for positive spatial correlations between species, even when those associations and their outcomes are context dependent (Tikhonov, Abrego, Dunson, & Ovaskainen, 2017).…”

An inverse latitudinal gradient in infection probability and phylogenetic diversity for Leucocytozoon blood parasites in New World birds

“…Climate change impacts parasite infection rates because spatiotemporal variation in host community structure, caused by shifts in temperature and precipitation, alters host-parasite contact rates (Canard et al, 2014). Host traits that limit infection can also fluctuate in response to environmental conditions (Joseph, Stutz, & Johnson, 2016). Warming temperatures, for instance, can alter the phenology of avian hosts by changing the onset of their breeding period (Walther et al, 2002).…”