Cross-species transmission and emergence of zoonotic-origin diseases occur at complex animal-human-environment interfaces, within dynamic social-ecological systems. These interfaces represent significant One Health challenges. Here, we present the first attempt to analyse nearly 50 years of data on naturally occurring zoonotic infections (or contaminations) in Austria, leveraging an original One Health approach, based on a novel dataset and network theory. This work demonstrates that most zoonotic agents are capable of infecting both human and diverse animal species across various taxa, while evolving within multi-source multi-agent ecological communities, consistent with the established principles in parasite community ecology31. We argue that the comprehensive analysis of the zoonotic web holds greater value when studying zoonotic transmission chains compared to the commonly employed host-pathogen network approach, as it offers a broader epidemiological perspective and more analytical flexibility. Notably, we studied the centrality of zoonotic sources, including hosts, vectors, foodstuffs, and environmental matrices, within the network of zoonotic agent sharing, and evidenced that certain sources play a disproportionate role in the sharing of zoonotic agents. Specifically, we underscored the crucial role of arthropod vectors and foodstuffs (typically omitted in host-pathogen networks) in the risk of zoonotic disease emergence and transmission through the zoonotic web, pinpointing potential targets for One Health surveillance programmes.
Ten genera of zoonotic agents constituted 41% of the published research on zoonotic diseases in Austria, with seven of them involving agents subjected to compulsory surveillance and reporting in humans and/or animals32. This outcome underscores an imbalance in research interest, likely influenced by funding opportunities as well as global- and national-level prioritisation, typically based on known incidence and impact on human populations. Such a bias may lead to a skewed assessment of the overall zoonotic risk, especially concerning potentially "neglected" zoonoses such as certain helminth infections (e.g., dirofilariosis, dicrocoeliosis, hepatic capillariosis). Moreover, research trends show that very few publications in Austria address the environmental compartment, aligning with global observations33.
From 1975 to 2022, Austria saw the emergence of eight zoonotic agents, averaging one EID every six years. While there is often an emphasis on viral emergence, particularly considering that RNA viruses pose the most significant threat34, our findings challenge this perspective. We have demonstrated that most emerging pathogens (six out of eight) in Austria are bacteria and helminths. Notably, two of the emerging bacteria belong to the genus Rickettsia, aligning with the findings of Jones, et al. 1. This highlights the importance of broadening our focus beyond viral threats and acknowledging the substantial role that bacterial and helminthic pathogens play in the landscape of emerging diseases. Moreover, four emerging zoonoses are transmitted by arthropods vectors (WNV, USUV, R. helvetica, R. conorii subsp. raoultii). As a result of climate change and globalisation, there is a growing likelihood of new arthropod species populations becoming established in Austria, increasing the risk of future EID events35. Surprisingly, despite SARS-CoV-2 being notifiable for both humans and animals36,37, none of the COVID-19-related publications concerning human cases refer to it as a zoonotic disease. Likewise, the sole publication investigating SARS-CoV-2 in Austria animals fails to mention its zoonotic potential38.
Within the zoonotic web, multiple zoonotic sources contribute to the maintenance and spread of zoonotic agents. However, many sources found (sero)positive for a zoonotic agent, may not, when taken individually, be able to maintain a sustained persistence of the agent within the network39. Nevertheless, as members of a zoonotic source community, interacting with maintenance and non-maintenance sources, they potentially play a role in the zoonotic agent ecology40. We demonstrate that the zoonotic agent sharing network in Austria is organised into six communities. Our results indicate that the community including human, the oldest domesticated species (e.g., dog, cat, sheep, cattle, pig41), and synanthropic species (e.g., Norway rat, house mouse) shares the most zoonotic agents. These national-level findings align with results from global studies3,42. Additionally, human-modified environments, such as sandboxes, cluster with humans, domesticated and commensal species, highlighting the role of the shared ecosystem and environmentally persistent stages in the ecology of certain zoonoses43. The determinants of the zoonotic source community assembly and composition remains a challenge in disease ecology5,44. We found evidence that a limited number of highly connected zoonotic agents in the bipartite zoonotic web, such as USUV, S. enterica, WNV, and Influenza A, may, at least partly, drive zoonotic agent sharing community assemblage. The grouping of most food products into one community implies that anthropogenic activities, particularly those related to food processing and transformation, may further influence the pattern of assembly within zoonotic source communities. These findings suggest that a combination of local epidemiological, ecological, human-related, and behavioural (e.g., relationship and proximity to human)3 factors play a key role in shaping zoonotic agent sharing community patterns.
Our findings underscore the presence of central zoonotic sources in the network, demonstrating robust results across three to four centrality metrics after controlling for research effort. These central zoonotic sources have a higher number of interactions with zoonotic agents, acting as hubs, or bridge different zoonotic sources communities in the network, acting as connectors45. In particular, some livestock species (e.g., cattle, chicken), companion animal (e.g., dog, cat, turtles), wildlife (e.g., yellow-necked field mouse, wild boar), and vectors (Ixodes) play a crucial role as bridge hosts, through which zoonotic agents can potentially spillover from maintenance (generally wild) host populations or communities to target populations (generally domesticated species or humans) that are usually “protected” through public health or biosecurity measures25,39,46,47. Notably, Ixodes ticks are pivotal in the epidemiology and zoonotic spillover of bacteria from the genera Rickettsia, Borrelia, and Babesia. Furthermore, the two communities involving USUV and WNV hosts illustrate the maintenance of zoonotic viruses within partially overlapping host communities. In this subsystem, mosquitoes of the genus Culex play a central role, serving as primary amplification vectors for WNV and USUV within each bird community. Additionally, Culex mosquitoes act as bridge vectors between both avian maintenance communities and between these communities and potential mammalian hosts, including humans48. These results emphasise the importance of both vector monitoring and testing for pathogens as an essential component for early detection of emerging zoonoses and the establishment of early warning systems.
We present a novel approach based on the identification and quantitative characterisation of specific network structures, named One Health 3-cliques, for estimating the likelihood of zoonotic spillover at human-animal-environment interfaces. This method is flexible and can be applied to any zoonotic web. Our findings demonstrate that there is an increased probability of zoonotic spillover at human-cattle and human-food interfaces. Notably, human zoonotic infection through consumption of contaminated food is a major public health risk, with Listeria, Salmonella, and Escherichia being the most frequently reported agents in food products across the included publications. Our results further emphasize the critical importance of monitoring zoonotic agents in food-processing environments.
A crucial challenge in formulating One Health surveillance and primary prevention strategies (i.e., at source)17 for multi-source zoonotic agents, in particular emerging ones, is identifying what is the reservoir of infection46, i.e., characterising, within a given context, the “ecologic system in which an infectious agent survives indefinitely”49 and from which it can be sustainably transmitted to the target population39. The goal is to define what could be an optimal (high specificity and sensitivity) sentinel50 to detect the circulation of a specific zoonotic agent above an acceptable threshold posing a potential transmission risk to the target population (typically human). Identifying sentinels through network metrics should depend on the topology of the network, the infectious agent to be monitored (e.g., endemic versus emerging, transmission route(s)), the (estimated) infection rate, the target population, the objective of the surveillance (e.g., early detection versus prevalence estimation)51,52, and the specific epidemiological, ecological, and socio-cultural-economic (e.g., what resources are available, what measures are acceptable) context. Notably, selecting sentinels that are distant from each other in the network proved to enhance the overall probability of one sentinel being in proximity to an outbreak, thereby increasing the likelihood of detection53. For example, distributing the sentinels in different communities52 and prioritising surveillance of highly connected nodes in the network29 (e.g., via regular sampling) would achieve higher performance than randomly selected nodes.
Nodes to be prioritised for surveillance may be different than those used for disease control53. Removing central nodes in the network, e.g., via vaccination or culling targeting “bridge” zoonotic sources, can significantly reduce the connectivity of the zoonotic web29, therefore decreasing the likelihood of zoonotic spillover into the human population. However, betweenness centrality fails to discriminate between zoonotic sources that have high betweenness because they have a lot of connections in the network (i.e., high degree centrality), such as human and cattle, or sources that really connect two communities, serving as bottlenecks for zoonotic transmission flow29 (e.g., Ixodes). Nevertheless, the effectiveness of interventions is intricately connected to the specific system under study and must be tailored to the context. For example, badger culling, equivalent to removing the badger node in the zoonotic web, has shown contrasting results on the prevalence of tuberculosis in cattle in the UK54,55.
Our study acknowledges several limitations. First, poorly described taxonomic names hinders precise identification of zoonotic agents or vertebrate hosts at the species level. Likewise, the unspecific description of food origin (e.g., “unspecified” animal), alongside our conservative approach to data validation/cleaning and adherence to authors' terminology, may have resulted in inaccurate assessment of the degree centrality for some nodes. For example, shiga toxin-producing E. coli (STEC) strains could refer to both VTEC and EHEC56; similarly, in the case of a host linked to both Listeria and L. monocytogenes, Listeria could potentially be L. monocytogenes. Imprecise description of the samples and zoonotic agents in publications represents a major limitation to the estimation of the zoonotic risk. Moreover, the single species-single pathogen approach, especially dominant in human medicine11, and the tendency to disproportionally investigate zoonotic sources that are closer to humans can results in sample bias and in a skewed distribution of the number of zoonotic agents recorded per sources, with human showing the highest number of zoonotic agents, followed by domesticated species. Ultimately, expanding the dataset by including additional data on natural infections from diverse laboratories (e.g., university laboratories that often investigate a broader range of sources and agents compared to national reference labs) and incorporating a temporal dimension to zoonotic source-agent interactions would allow for a more comprehensive and dynamic assessment of the zoonotic transmission chain within and between the communities. This approach could unveil seasonal variations in spillover events57 as well as mechanisms that link host diversity to disease spread and emergence58. Moreover, as data on directionality in transmission is largely unavailable, we used a non-directed network and assumed a symmetrical process in interspecies transmission. This simplification of the spillover process may have limitations in capturing nuances in the dynamics of zoonotic transmission59 (e.g., WNV can be transmitted from birds to humans via mosquitoes but this transmission process is not reciprocal). Furthermore, our data provides information on infection solely at the species level, overlooking individual variations in shedding, and potentially missing key individuals acting as hubs (“superspreaders”). Finally, controlling for detection method stringency8, such as PCR (or other direct detection methods) versus serology, could further refine our findings, allowing to adjust edge weight within the network.