Framework for heterogenous contact structures in bat-pathogen interactions
Taken together, the information in this study emphasises that models of bat disease dynamics that assume contact rate is density-dependent, but assume transmission scales with total roost abundance, may not represent actual contact structures. Such inadequate specification of transmission may produce substantially biased estimates of the basic reproductive number (R0) and propagate error to model predictions like the probability of pathogen invasion and persistence, predicted peak and timing of epidemics, and estimates of the force of infection (Borremans et al. 2017; Hopkins et al. 2020).
Intermediate, non-linear or hybrid transmission functions are a possible alternative to standard density-dependence (e.g. Antonovics, Iwasa & Hassell 1995; Ryder et al. 2007; Cross et al. 2013; Orlofske et al. 2017), but these may not reveal underlying mechanisms for the relationship, and as a result, may be hard to selecta priori based on ecological information, and may not be generalisable or predictive between bat roosts of the same species (Smith et al. 2009; Ferrari et al. 2011). Instead of modifying the transmission function, it may be better to investigate alternative approaches to integrating contact structure within host-pathogen models at ecologically relevant scales (De Jong 2002). We therefore propose a framework (Fig. 3) to help guide the incorporation of heterogenous contact structures into infectious disease models of bats in ecologically relevant ways – for example by structuring groups within roosts as metapopulations, with separate ecological processes defining contacts within and between groups. Our framework prompts ecological questions that may be relevant for specifying transmission within wildlife disease models. They include whether hosts mix homogenously throughout the roost or mix within smaller subgroups; how population or group contacts are expected to change with increasing abundance; and whether roost or group area fluctuates with abundance.
Given a roost has fluctuating abundance or density, the first step in the framework is to consider the nature of mixing in bat roosts. That is, whether bats mix evenly/randomly throughout the roost, mix within smaller subgroups, or have other structured contact networks. This will determine what scale is ecologically relevant for transmission, and so, what scale(s) the model should consider. If bats mix throughout the roost (i.e. all individuals have equal likelihood of coming into contact) the mechanisms driving contact rate will fall more simply to a choice between density-dependent and frequency-dependent expected dynamics. If occupied area changes with abundance, models will be best parameterised by density at this roost scale, otherwise, by either abundance or density.
In cases where individuals interact within aggregate groups that include only a proportion of the population, transmission mechanisms may need to be more nuanced to include special structuring within the roost. This is because the structure of the host population (and the strength of coupling between local groups) may drive transmission between groups, and be different to (and/or independent of) the nature of within-group contacts (Jong, Diekmann & Heesterbeek 1995; Ferrari et al.2011). In other mammal systems, this paradox has led to cases where dynamics appear to be density-dependent at the within-group scale, but frequency dependent at the between-group scale (Ferrari et al.2011; Cross, Caillaud & Heisey 2013). In these cases, models that can distinguish within- and between-group transmission pathways may be useful (e.g. metapopulation models). If mixing is non-random and based on individual contact networks, individual based models may provide a good framework. Of course, the complexity of adopted models should be driven by the objective of the investigation, and reflect a parsimonious attempt to reproduce transmission patterns relevant to the system and question. This need not necessarily capture every single mechanism in the real system.
Consideration of these questions will provide a more ecologically informed, mechanistic basis for specifying transmission, but will require more data and more computational power. This may or may not be achievable for many host species, for which basic ecological information is lacking. Even if ecologically informed specification of transmission is not possible, consideration of our framework will help to highlight cases where traditional density-dependent transmission may fail to reproduce data, and why. If integrated into research programmes, this could create the opportunity for a model guided fieldwork approach (Restif et al. 2012) and represent bat-disease systems in a more holistic approach. This framework also assumes transmission between bats is direct and occurs predominantly within the roost. This is consistent with our knowledge of bat-virus systems of zoonotic importance (Plowright et al. 2015). Nevertheless, understanding the nature of density at transmission-relevant scales, and building this into transmission dynamics, will be important to gain more realistic predictions of pathogen invasion and persistence in bat populations. This will be crucial for accurately forecasting disease risk from these animals.

Conclusion

Transmission is the focal process in host-pathogen interactions. The nature of infectious contacts, and how transmission scales with animal density, is complex for host species whose population structures are heterogenous and underpinned by ecological processes across different scales. Using a high-profile bat-virus system, we show that basic bat population measures from larger scales were not strongly predictive of local scale measures where viral transmission occurs. We also suggest that the highly aggregative spatial structuring of bats is likely to add substantial heterogeneity to the contact structure of roosting populations, further complicating models of pathogen transmission. We urge researchers to carefully consider which scale and modelling method is most relevant for transmission in bat-virus models. More broadly, we propose a framework to guide the structuring of transmission in more ecologically relevant contexts. This approach can apply to many species that occupy communal breeding or resting sites, and has an advantage over other statistically based approaches by allowing selection of scale and transmission structurea priori based on ecological information. Outputs using this ecologically informed approach will be more generalisable and predictive of infection patterns, and can be used to gain mechanistic insight into the drivers of transmission, local epidemics and pathogen spillover risk.

Authors contributions

TJL conceived and designed the research, acquired funding and led project administration; TJL and RB collected and curated the data; TJL, AJP and HM analysed and visualised the data; AJP, HM, RKP and PE provided supervision; TJL drafted the manuscript, and all authors participated in review and editing of drafts.

Data Availability Statement

Summarised data will be made available on GitHub at: < https://github.com/TamikaLunn/FF-roost-structure >. Data will also be made available from the Dryad Digital Repository upon publication.

Acknowledgements

We would like to thank Beccy Abbot, Kirk Silas, Devin Jones, Liam Chirio, Rachel Smethurst and Cara Parsons for their assistance in the field. We acknowledge the Danggan Balun, Kabi Kabi, Turrbal, Widjabul Wia-bal, Yugambeh and Yuggera Ugarapul people, who are the Traditional Custodians of the land upon which this work was conducted. Fieldwork for this work was supported by the Paddy Pallin Foundation, The Royal Zoological Society of NSW, The Foundation for National Parks and Wildlife, the National Science Foundation (a Dynamics of Coupled Natural and Human Systems grant DEB1716698) and a DARPA PREEMPT program Cooperative Agreement (#D18AC00031). TJL was supported by an Endeavour Postgraduate Leadership Award and a Research Training Program scholarship sponsored by the Australian Government, AJP was supported by an ARC DECRA fellowship (DE190100710) and a Queensland Government Accelerate Postdoctoral Research Fellowship, RB was supported by a Griffith University Honours Scholarship and an EFRI Thesis Write-Up Scholarship, and RKP was supported by USDA National Institute of Food and Agriculture (Hatch project 1015891). This research was conducted under a Griffith University Animal Research Authority permit (DEB-1716698), a Scientific Purposes Permit from the Queensland Department of Environment and Heritage Protection (WISP17455716), a permit to Take, Use, Keep or Interfere with Cultural or Natural Resources (Scientific Purpose) from the Department of National Parks, Sport and Racing (WITK18590417), a Scientific Licence from the New South Wales Parks and Wildlife Service (SL101800) and general and products liability protection permit (GRI 18 GPL), and with permission to undertake research on council and private land. The content of the information does not necessarily reflect the position or the policy of the U.S. government, and no official endorsement should be inferred.

Tables and Figures

Table 1: Model comparison of candidate model set. Best candidate models, as given by Akaike information criterion (AIC), are bolded (ΔAIC<2).