Language Selection

Get healthy now with MedBeds!
Click here to book your session

Protect your whole family with Orgo-Life® Quantum MedBed Energy Technology® devices.

Advertising by Adpathway

         

 Advertising by Adpathway

Viral disease outcomes are indistinguishable between experimentally infected bats and rodents

3 days ago 6

PROTECT YOUR DNA WITH QUANTUM TECHNOLOGY

Orgo-Life the new way to the future

  Advertising by Adpathway

  • Loading metrics

Open Access

Peer-reviewed

Research Article

?

This is an uncorrected proof.

Abstract

A common explanation for bats being a conspicuous source of zoonotic viruses is their purported ability to coexist with viruses without suffering overt disease. Widespread acceptance of the idea that bats have unusual antiviral traits has catalyzed the discovery of unique features of bat immune systems which may have translational value as broad-spectrum antivirals, particularly if they evolved as a byproduct of bats’ unique life history rather than through conventional co-evolutionary processes. Surprisingly, whether bats compared to other host groups suffer less disease from co-evolved viruses or from viruses generally has not been formally assessed. Here, synthesizing 86 years of experimental infections, involving 54 viruses, 85 host species, and over 5,600 animals, we document disease in bats following inoculation by taxonomically diverse viruses, including ones that are relatively benign in humans. The occurrence of overt disease, the likelihood of mortality, and the severity of disease in bats were indistinguishable from those experienced by rodents, another group associated with many zoonotic viruses. These patterns were consistent when considering only bat-associated viruses inoculated into bats and rodent-associated viruses inoculated into rodents and among inoculations which lacked shared co-evolutionary history. Instead, disease outcomes following infection were shaped by experimental design, viral host range, and evolutionary context. Unexceptional disease in bats from novel or co-evolved infections suggests that order-level host life history traits such as flight have inconsistent consequences for antiviral immunity and highlights the need to evaluate the functional properties of putatively unique features of bat immunity in vivo. While these results do not exclude the possibility of developing broader-acting or more potent antivirals from bat immune systems or diminish the potential value of mechanistic insights into bat immunity, they suggest that bat immune systems were shaped by conventional co-evolutionary processes rather than through unique circumstances that endow exceptional antiviral defenses.

Citation: Farrell MJ, Tucker SK, Mollentze N, Streicker DG (2026) Viral disease outcomes are indistinguishable between experimentally infected bats and rodents. PLoS Biol 24(10): e3004053. https://doi.org/10.1371/journal.pbio.3004053

Academic Editor: Andrea L. Graham, Princeton University, UNITED STATES OF AMERICA

Received: August 19, 2026; Accepted: September 28, 2026; Published: October 29, 2026

Copyright: © 2026 Farrell et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: Data and scripts to reproduce analyses are available at https://github.com/maxfarrell/bat_rodent_inoculations and archived at Zenodo [35] (https://doi.org/10.5281/zenodo.15195673).

Funding: This work was supported by a Philip Leverhulme Prize in Biological Sciences (PLP-2020-362; https://www.leverhulme.ac.uk), a Wellcome Trust Senior Research Fellowship (217221/Z/19/Z; https://wellcome.org), the NSF/BBSRC Ecology and Evolution of Infectious Diseases Program (DEB 2011069, BB/V003798/1; https://www.nsf.gov, https://www.ukri.org/councils/bbsrc) awarded to D.S., and an NSERC Postdoctoral Fellowship (PDF – 557470 - 2021; https://nserc-crsng.canada.ca) awarded to M.J.F. S.T. was funded by the Wellcome Trust Integrative Infection Biology PhD Programme at the University of Glasgow (218518/Z/19/Z; https://wellcome.org). Additional funding was provided by the UK Medical Research Council (MRC) through core support to the MRC-University of Glasgow Centre for Virus Research (Virus Cross Species Transmission Programme: MC UU 00034/3; https://www.ukri.org/councils/mrc). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: DGS is a member of PLOS Biology’s Editorial Board. The other authors declare that no competing interests exist.

Abbreviations: CFR, case fatality rates; FFU, focal forming units; PFU, plaque forming units; WOS, Web of Science.

Introduction

The purported rarity of overt disease in bats infected with viruses that are pathogenic in other mammals has fueled the widespread perception that bats have enhanced capabilities to limit the severity of active viral infections [1–7]. Efforts to explain this phenomenon have identified numerous features of bat immune systems that appear to differ from other mammals [8–10]. Most enticingly, some aspects of bat immune systems such as their capacity to limit inflammation are hypothesized to have evolved as a byproduct of the energetic demands of flight which are unique to bats among mammals. If so, these features might protect against diverse viruses, potentially representing candidates for breakthrough, broad-spectrum antivirals [4,5,11,12]. Supporting this possibility, bat derived ASC2 (apoptosis-associated speck-like protein containing a CARD) appears to dampen inflammation caused by both coronaviruses and orthomyxoviruses, and reduced inflammatory cytokines caused by orthoreovirus and flaviviruses [13]. Alternatively, apparent limitation of virus-induced disease in bats may largely reflect conventional adaptations to the specific viral groups that bats co-evolved with. Indeed, neither taxon-specific immune features nor sub-clinical infections from viruses that cause disease in other species are unique to bats [14,15]. In this case, interactions between bat life history and their co-evolved viruses might still lead bats to have more potent disease control mechanisms against these particular viruses when compared to other host groups, but these defenses would be unlikely to offer protection against a broad-spectrum of viruses.

Distinguishing whether bats exhibit more potent or taxonomically comprehensive suppression of viral disease compared to other mammals underpins research into the therapeutic potential of bat immune systems and the nature of bats as special zoonotic reservoirs, but remains untested. Observations of infection outcomes in wild bats are largely anecdotal; for example, viruses detected following mortality events with unknown etiology, or viruses detected in apparently healthy bats whose ultimate fates are unobservable [16,17]. Field observations are further complicated by the small body sizes and cryptic life histories of bats, which limit detections of virus-induced morbidity and mortality to the most conspicuous events. In contrast, experimental inoculations of captive animals can confidently monitor infection outcomes following known exposure events. Such infections have revealed limited disease during infection by several viruses that can cause high mortality in humans (e.g., Marburg virus, Nipah virus, and Hendra virus) [1,3,18]. Determining whether these observations represent the exception or the rule requires synthesizing outcomes across a broader range of viruses. Until now, such analyses have been precluded by the absence of systematic datasets on the disease outcomes of experimental infections.

We examine the current state of evidence that bats experience dampened virus-induced disease by extracting infection outcome data from 86 years of experimental viral infection studies. The severity of disease outcomes may be influenced by both tolerance (i.e., the ability to limit disease at a given viral load) and resistance (i.e., the capacity to inhibit infection upon exposure). Both phenomena likely occur in bats and may select for increased viral pathogenicity. Tolerance has been of particular interest since it has been hypothesized to be heightened in bats relative to other mammals and to promote zoonotic transmission by enabling bat survival and continued viral shedding despite high viral loads [9,18–20]. We note that despite the informal use of the term “tolerance" in the bat virus literature when referring to the outcomes of experimental and natural viral infections, formally measuring tolerance would require noninvasive quantification of viral loads in physiologically relevant tissues and prohibitively expensive long-term monitoring of fitness costs in infected hosts. It is therefore unclear that tolerance has or can be measured from experimental or natural infections [21]. We therefore frame our analysis around the occurrence and severity of outcomes following viral infection based on the extent of author-reported disease (including clinical and sub-clinical pathologies). Because we are interested in disease outcomes in compatible host-virus pairs, we restrict our analysis to inoculations in which the original authors deemed hosts to be susceptible (See Section A in S1 Appendix for detailed methods).

We contrast bats with rodents, another species-rich mammalian order [22] that hosts a large number of zoonotic viruses (some asymptomatically [23]). However, on average rodents have shorter life expectancies and higher reproductive rates than bats, which some theory has suggested may make them less likely to evolve mechanisms to suppress viral disease, though other models have suggested that these relationships may be complex [24–29]. Importantly, our dataset included viruses that are presumed to have co-evolved with bats and rodents (homologous inoculations), and those that are currently unlinked to either mammalian order (heterologous inoculations). This allowed us to define and test two main hypotheses on how disease outcomes in bats differ from other groups: that nonspecific antiviral mechanisms enable bats to reduce disease severity irrespective of co-evolutionary relationships (generalized response) and that bats suppress disease from their co-evolved viruses better than other host groups due to the combined selection pressures of bat life history constraints and infection by specific viral groups (heightened co-evolutionary response).

Beyond intrinsic host properties, variation in disease outcomes may reflect eco-evolutionary characteristics of viruses [30–32]. For example, viruses that infect a broader diversity of host species may replicate poorly in any single species, making them less pathogenic than host-specific viruses [31,33]. Viruses may also cause increased pathogenicity in novel hosts which lack co-evolved defenses, or alternatively disease severity may be reduced in novel hosts due to poor adaptation and reduced replication [31,33]. Therefore, in our models of disease outcomes we included measures of viral host range and proxies for the extent of co-evolution with inoculated host groups. Experimental design choices such as the dose and route of inoculation, the method and extent of virus passage prior to inoculation, and the number of individuals included in each study may also influence the likelihood that an inoculation leads to detectable disease. We therefore adjusted for these experimental details in our models where possible, or carried out independent analyses to ensure they were unlikely to explain our main findings. Finally, we investigated the relationship between human case fatality rates (CFR) and disease outcomes in experimentally inoculated bats and rodents to assess whether either host group was systematically inoculated with viruses that are more pathogenic in a third host group, and to explore the suggestion that bats consistently experience reduced disease severity from viruses that are highly pathogenic to humans.

Results

Diverse viruses cause overt disease in bats and rodents

We identified 112 studies published between 1936 and 2022 that reported disease outcomes following experimental inoculations (Section D in S1 Appendix) of 54 viruses from 15 viral families (53 RNA viruses; 1 DNA virus) into 85 susceptible host species (Fig 1; Fig H and Table C in S1 Appendix). These represented 149 host-virus pairs and a total of 5,622 inoculated animals. We excluded laboratory-bred rodents as they may lack the full immune competency of their wild counterparts [34]. Across host-virus combinations, bat-virus pairs included a larger number of viruses than rodent-virus pairs, but fewer bat species were subjected to experimental inoculation, indicating frequent use of a smaller number of model bat species (Fig I in S1 Appendix). Nevertheless, while some host clades were not subjected to experimental inoculations, our dataset comprised a phylogenetically representative sample of both bats and rodents (Fig H in S1 Appendix), capturing variation in host life history traits hypothesized to influence disease outcomes ([25]; Fig J in S1 Appendix). For each host-virus pair per study, we recorded author-reported signs of disease and the occurrence of virus-induced mortality. We further converted signs of disease into a five point severity scale which ranged from no overt disease to death. We therefore consider three distinct disease outcome metrics: disease presence, mortality, and severity. Across this dataset, disease occurred in 48% (40/83) of rodent-virus pairs and 73% (48/66) of bat-virus pairs (Fig 1). Among all host-virus pairs, viruses from 15 families caused disease (Table E in S1 Appendix), with severe disease observed across the host phylogeny rather than being clustered within select clades (Fig 1). Filoviruses and paramyxoviruses are commonly cited examples of highly pathogenic viruses that rarely cause clinical disease in bats. Consistent with this idea, all tested filoviruses caused at most subclinical disease in bats and most paramyxoviruses caused no disease or subclinical disease in bats. However, two out of six bat-paramyxovirus pairs (specifically Sosuga pararubulavirus in Rousettus aegyptiacus and Newcastle disease virus in Myotis lucifugus) caused severe clinical disease and death, respectively. Other notable virus-bat pairs that resulted in severe or lethal disease included Foot and mouth disease virus inoculated into Desmodus rotundus, Cocal vesiculovirus inoculated into Myotis lucifugus, Kaeng Khoi virus inoculated into Chaerephon plicatus, Tacaribe virus inoculated into Carollia perspicillata and Artibeus jamaicensis, Kyasanur Forest Disease Virus inoculated into Rousettus leschenaultii and Cynopterus sphinx, Tick-borne encephalitis virus inoculated into Barbastella barbastellus and Plecotus auritus, Dengue virus inoculated into Artibeus literatus, West Nile Virus inoculated into Eptesicus fuscus and all 9 included lyssavirus species inoculated into a variety of bat species (Table E in S1 Appendix).

thumbnail

Fig 1. Maximum disease severity observed in experimental infections for each host-virus combination.

The five point severity scale is calculated based on author reports of infection-induced disease and death (Section D in S1 Appendix). Blank spaces indicate untested interactions. Hosts (rows) are grouped according to the [22] mammal supertree. Viruses (columns) are grouped following the ICTV taxonomy. The barchart on the right shows the total number of host individuals per species included in experimental infections. For an alternative version depicting mean observed severity, see Fig M in S1 Appendix. The data underlying this figure can be found in [35] (https://doi.org/10.5281/zenodo.15195673). This figure includes a derivative of “Desmodus” by Melissa Ingala (via PhyloPic), used under CC BY 3.0.

https://doi.org/10.1371/journal.pbio.3004053.g001

Disease presence, mortality, and severity are indistinguishable between bats and rodents

Although raw data suggested similar disease outcomes between host orders (Fig 1), this comparison did not control for differences in experimental design (e.g., dose, inoculation route, number of animals inoculated, etc.), host specificity, host-virus co-evolution, or the nonindependence among taxa due to shared evolutionary histories. Indeed, the number of individual bats per study tended to be higher and more variable than the number of rodents (bats: median = 14.5, range 1–493; rodents: median = 6, range = 1–61), which, if unaccounted for, could inflate observations of clinical disease in bats. We therefore used hierarchical Bayesian models to examine host taxonomic effects alongside eco-evolutionary metrics reflecting the potential degree of virus adaptation to inoculated hosts, and the details of experimental design. At the host species level (i.e., the occurrence or maximum severity of disease among all individuals inoculated with a given virus within a given study; n = 193 unique host-virus-study combinations), disease presence, mortality, and severity following viral inoculation remained indistinguishable between bats and rodents (the 80% credible interval for the effect of host order included zero; Fig 2). The absence of host order effects was robust to removing lyssaviruses (Section B.2.4 in S1 Appendix), which represented 23% (45/193) of host-virus-study combinations in our dataset and, due to causing mortality, are sometimes considered the “exception to the rule" as viruses that cause severe disease in bats [25]. In contrast, hierarchical effects describing nonindependence among hosts and viruses indicated that variation in disease outcomes was partially driven by variation among hosts and viruses that was not captured by other predictors in the model (Section B.4 in S1 Appendix).

thumbnail

Fig 2. Estimated effect sizes for predictors in species-level and individual-level models of A) disease presence, B) mortality, and C) disease severity.

Points represent posterior mean estimates, thick bars represent 80% credible intervals, thin bars represent 95% credible intervals. Predictors in both species- and individual-level models include host order (effect of Chiroptera with Rodentia as reference group), the standardized effect size of phylogenetic distance among all documented hosts per virus (“Host phylogenetic diversity”), the mean evolutionary distance from all documented host species to the inoculated host species (“Host evolutionary isolation”), and phylogenetic and nonphylogenetic hierarchical effects for host and virus species (Section B.4 in S1 Appendix). Species-level models additionally included the number of individuals per study (“Number of individuals”) and a hierarchical effect of article. Individual-level models additionally included inoculation dose adjusted for species body mass (“Dose/g”; see Section B.2.5 in S1 Appendix for alternative approaches to model inoculation doses) and hierarchical effects for inoculation route (e.g., direct inoculation into the brain, oronasal inoculation, etc.) and the dose measurement units (e.g., Mouse LD50, TCID50, and PFU, etc.). Positive host order effects would indicate more severe disease in bats compared to rodents. The data and scripts underlying this figure can be found in [35] (https://doi.org/10.5281/zenodo.15195673).

https://doi.org/10.1371/journal.pbio.3004053.g002

Modeling host-virus interactions at the species-level maximized the diversity of hosts and viruses in our analyses by including studies that did not report data for individual animals. However, this approach disguised individual animal-level variation in disease outcomes and required omitting experimental conditions that varied within studies (e.g., dose and inoculation route). We therefore analyzed infection outcomes for 1,620 individual animals, spanning 44 viruses, 75 host species, and 128 host-virus pairs from 84 studies which reported sufficiently resolved data. Using a similar structure to our species-level models, we fit individual-level models for the three measures of disease outcome, but replaced the study-level continuous predictor (number of inoculated host individuals) with body mass-corrected dose and inoculation route (see Section A.2 in S1 Appendix). These models again found no difference between bats and rodents in any disease metric (Fig 2).

Contexts of infection shape disease outcomes

While our data and models do not support a host order effect, we did find evidence for several other factors theoretically predicted to influence the outcomes of infection. Among eco-evolutionary traits of viruses, infection by phylogenetic generalists (i.e., viruses that infect evolutionarily diverse host species) was less likely to result in disease, mortality, and severe disease (Fig 2), consistent with the hypothesis that costs of generalism lead specialist viruses to cause more severe disease [31].

To test whether either host order dependably experienced reduced disease when exposed to potentially novel viruses, we included an effect of host evolutionary isolation (HEI) [15], calculated as the mean phylogenetic distance from all known hosts [36] to the inoculated host (see Section A.2.4 in S1 Appendix for details). As pathogens commonly infect closely related host species, this metric assumes that the phylogenetic centroid of known susceptible species reflects the most likely position of hosts to which a virus is best adapted, and that the distance from an inoculated host to this centroid approximates the extent of co-adaptation [15,33,37,38]. Although estimated effect sizes were uncertain, mean effects of HEI were positively associated with disease outcomes, with greater certainty in species-level models of mortality (80% CIs excluded zero; Fig 2; Section B.2 and Fig AA in S1 Appendix). Although tentative, these results suggest that virus inoculations into distant relatives of currently documented susceptible species may more often result in more severe disease, a pattern previously reported for infection-induced mortality in humans and domesticated animals [15,31,39,40].

As HEI does not differentiate whether documented host species are reservoirs to which the virus may be adapted, incidental hosts, or a combination of the two, it does not directly test whether heterologous inoculations caused more disease. We therefore constructed a smaller dataset of 36 viruses for which the reservoir host group was known at least to the level of taxonomic order, and introduced a binary effect (“reservoir match") using an existing dataset of virus-reservoir associations to assign the host provenance of viruses ([41]; Section B.2.2 in S1 Appendix). Across models, we found little compelling evidence for elevated disease outcomes in either homologous or heterologous infections (Fig T in S1 Appendix). A single exception was our individual-level model of disease presence, where inoculations into homologous hosts (e.g., bats inoculated with bat-reservoired viruses and rodents inoculated into rodent-reservoired viruses) increased the probability of disease (80% CI excluded zero; Fig T in S1 Appendix). This effect held in a sensitivity analysis that removed the HEI variable, since inoculations associated with low HEI tended to be homologous (Figs N and U in S1 Appendix). Whether including or excluding HEI, models with the reservoir match effect found no effects of host order (Figs U and T in S1 Appendix). Fitting these models with a reservoir match by host order interaction term similarly provided no evidence that bats and rodents differed in their ability to mitigate homologous versus heterologous infections (Figs V, W, and X in S1 Appendix).

Although a weak positive effect of reservoir match appears at odds with positive effects of HEI, it may be that for some host-virus interactions, adaptive virulence in homologous infections may outweigh maladaptive virulence in heterologous infections [42]. Alternatively, because our reservoir match metric was defined at the order and not species level, it may fail to capture the extent to which the inoculated host species experienced selection to reduce disease caused by infection with that particular virus. As a final sensitivity analysis and to separately test the hypotheses that bats have unusually potent disease suppression with either bat-associated viruses or evolutionarily novel viruses, we refit our disease outcome models to homologous and heterologous infections separately. Trends in eco-evolutionary variables and experimental design effects were similar to those observed in our main models, detecting no clear difference between bats and rodents for any measure of disease outcome (Section B.2.3 in S1 Appendix). This suggests that whether the virus involved is one for which bats are unlikely to have co-evolved defenses or one where virus-specific antiviral defenses may have putatively co-evolved, the frequency or severity of disease in bats cannot be demonstrated to be unusual using existing data from experimental infections.

We also detected influences of experimental design in both our individual- and species-level models. In the species-level models, all three disease outcomes were positively associated with the number of inoculated individuals (Fig 2), pointing to the possibility that experimental infection studies are often underpowered to determine whether a specific host-virus association causes disease. In our individual-level models, larger inoculation doses unsurprisingly increased the probability of disease and led to more severe disease (Fig 2). Inoculation doses ranged over six orders of magnitude in both groups (Figs D and E in S1 Appendix), but were not significantly different between bats and rodents (Fig F in S1 Appendix). While high doses are often used in experimental inoculations to increase the likelihood of disease (raising separate questions about how well these experiments recapitulate natural infection), this does not explain the lack of a host order effect in our models. Similarly, as inoculations directly into the brain or cranium are associated with increased severity, and are more commonly performed in bats compared to rodents (Fig G in S1 Appendix), we re-fit the individual-level full model removing inoculations via direct injection into the brain or crania. These models show no host order effects, indicating that the tendency for these inoculations to be associated with bats does not impact conclusions drawn from our main models (Fig AD in S1 Appendix). Thus, our failure to detect decreased disease in bats cannot be explained by systematic differences in dose or inoculation route between bats and rodents (Fig 2; Figs D, G, and F in S1 Appendix).

Taken together, our findings support the conclusion that specialist viruses cause more severe disease and highlight ways that experimental design choices may bias current perceptions of generalized disease suppression in prominent reservoirs of viral zoonoses. Impacts of host-virus co-evolution appear to be more complex. Overt disease may be more likely for viruses infecting evolutionarily distant hosts, perhaps reflecting maladaptive virulence [15]. At the same time, viruses infecting species closely related to their co-adapted reservoir hosts may be more likely to successfully infect and cause disease, in this case reflecting adaptive virulence [42], which may be exacerbated when viruses infect susceptible hosts with little to no evolved defenses.

Comparing disease severity among bats, rodents, and humans

In principle, inoculation of bats with disproportionately pathogenic viruses could have disguised true differences in disease severity relative to rodents. We explored this idea using human CFR [18] as a comparable and broadly-available measure of pathogenicity. As expected from the generally elevated CFRs of bat-associated zoonotic viruses [18], experiments inoculating bats more often used viruses with higher CFRs compared to experiments inoculating rodents (Panel A in Fig AF in S1 Appendix). However, this effect was not robust after controlling for virus-level hierarchical effects (Fig AI in S1 Appendix). We also failed to find a difference in the CFRs of viruses used in bats versus rodents when restricting the analysis to heterologous inoculations or when excluding bat-associated viruses (Fig AI in S1 Appendix). As such, if a genuine reduction of disease in bats was undetectable in our dataset due to biases in viruses selected for experiments, it would primarily reflect bats experiencing less-than-expected disease from bat-associated viruses—consistent with the expectation that evolutionary association leads to virus-specific adaptations.

We next related CFRs with disease outcomes in experimentally inoculated animals to examine whether bats consistently have reduced disease for viruses that cause severe disease in humans (Fig 3; Fig AG in S1 Appendix). Supporting this notion, we identified a putative region where viruses cause high mortality rates in humans but mild disease in bats. However, this region included only bat-borne filoviruses (which at most caused sub-clinical disease in bats, i.e., a severity score of 2; Fig 1) and henipaviruses (Fig AH in S1 Appendix). More surprising, nonbat-associated viruses which cause relatively low human mortality (i.e., flaviviruses, alphaviruses, and hantaviruses; CFRs < 20%), sometimes caused severe disease in bats, including mortality (severity scores of 5; Fig 3) and high infection-fatality rates (Fig AG in S1 Appendix). Disease outcomes in bats are therefore compatible with conventional expectations that vertebrate immune systems are tuned to suppress disease caused by co-evolved viruses, but that these defenses fail to reliably protect against evolutionarily novel viruses [15,43]. In both bats and rodents, inoculations of moderate to high human CFR viruses from heterologous hosts were virtually absent (Fig AH in S1 Appendix), highlighting a major data gap in current understanding of viral disease.

thumbnail

Fig 3. Mean human case fatality rates per virus vs. maximum observed disease severity per experimental infection study.

Colors and shapes indicate the taxonomic order (Chiroptera/Rodentia) of host species in the experimental infection data. See Fig AH in S1 Appendix for an alternate version indicating the taxonomic group of known reservoir hosts. Points are jittered horizontally to improve visibility of overlapping points. The data and underlying this figure can be found in [35] (https://doi.org/10.5281/zenodo.15195673).

https://doi.org/10.1371/journal.pbio.3004053.g003

Discussion

The hypothesis that bats have evolved mechanisms to suppress viral disease more effectively than other animals prompted discoveries of features of bat immune systems which are proposed to have translational medical value [13,11,8,9]. Synthesizing 86 years of experimental infections, we show that a wide diversity of viruses cause severe disease in bats and find no evidence for systematic differences in disease occurrence or severity between bats and rodents, whether considering all viruses, viruses that putatively co-evolved with bats, or viruses that are presumably novel to bats. Instead, disease outcomes are shaped by factors which would be expected in any vertebrate, including aspects of experimental design, the host specificity of viruses used, and to a less certain extent the degree of co-evolutionary association with the inoculated host. These findings support the conclusion that some bat species evolved defenses which limit disease from specific co-evolved viruses, but the current literature provides no evidence that the frequency, potency, or cross-virus generalizability of these defenses in bats should be expected to be unusual relative to other mammals.

Our results, constituting the first quantitative test of whether bats experience comparatively low rates of virus-induced disease, support the idea that bats are unlikely to possess a ‘one size fits all’ solution where life-history driven selection resulted in generalized mechanisms to reduce the frequency or severity of viral disease [44]. If unique traits of bats (increased longevity given body size, flight) led to the evolution of unique antiviral mechanisms, our results suggest that these are either insufficient to protect bats at the organismal level (versus effects observed in vitro) or only protect against certain viruses. For example, we report a variety of nonbat associated viruses which caused severe disease in bats and show that the co-evolutionary context of infection influences disease outcomes in bats, patterns which would not be expected if bats had near-universal protection against viruses. Equally surprising, disease outcomes remained unexceptional in bats even when considering putatively co-evolved infections. This suggests that the evolutionary pressures resulting from the unique interaction of bat life history (longevity, flight) with the viruses they maintain have not necessarily resulted in mechanisms that allow bats to survive infections by co-evolved viruses better than other host groups. We speculate that widespread acceptance of the hypothesis that bats have exceptional abilities to coexist with viruses without overt signs of disease became embedded due to the association of bats with viruses that are highly pathogenic in humans [18], the challenges of observing virus-induced mortality in most bat species under natural conditions due to their small body sizes and reclusive behaviors, and an extrapolation of theory founded on a small set of well-studied systems such as henipaviruses and filoviruses [1,3,46,45]. While we find no issue with these individual studies, our results point to the potential dangers of over-generalization from otherwise well-grounded inferences.

Synthesizing published experimental infections introduced several challenges for our analyses. While we adjusted for variation in experimental design choices such as number of inoculated individuals, dose, and inoculation route in our models of disease outcomes, we found large heterogeneity across study designs and other features of experimental design, such as the degree of prior viral passage may have altered infection outcomes. Although the degree and mechanisms of pathogen evolution following serial passage depend on many factors [47–49], serial passage often increases virulence in passaged hosts and attenuates virulence in original hosts [50–52]. The viruses in our study were most often passaged in either primate or rodent sources, rather than bats (Fig K in S1 Appendix), which might be expected to decrease, not increase the severity of disease outcomes in inoculated bats. Where available, the mean number of prior passages also did not differ for bat and rodent inoculations (Fig L in S1 Appendix), suggesting that our inability to recover effects of host order on disease outcomes is unlikely attributable to systematic biases in passaging practices. Nevertheless, future studies should examine how passage number, host origin, and passaging mechanisms affect disease outcomes. Our study was also limited to bats and rodents. While these mammalian groups are commonly used in experimental viral infections and—as the two largest mammalian orders—host the majority of zoonotic viruses [41], the omission of other host clades opens the possibility that both bats and rodents experience unusually mild disease. However, given that the life history strategy of rodents prioritizes reproduction over longevity [24–27], and mechanisms that limit the severity of viral infections in bats are suggested to be byproducts of adaptations to powered flight [1–7], this outcome seems unlikely. Although our results should be re-visited as inoculations into other host orders accumulate, we suggest that bats and rodents are well-characterized groups that represent broader phenomena of taxon-specific immunological traits [14] and context-dependent outcomes of host-virus interactions. It is also conceivable that certain bat clades have heightened tolerance or susceptibility to viruses, akin to results observed in experimental infections of invertebrates [40,53]. While the hierarchical effects of bat species in our models of disease outcome do not obviously support this conclusion (Section B.4 in S1 Appendix), we caution that sparse data for some bat clades makes these effects difficult to interpret. Another implicit challenge of using experimental infection data is that experiments are typically carried out over short timescales that preclude quantification of long-term fitness impacts, and they cannot quantify viral load in infected tissues without sacrificing individuals or using highly invasive sampling (both of which could affect host fitness). As these unmeasured outcomes define viral tolerance, both our results and outcomes of individual experiments may reflect either resistance, tolerance, or a combination of both. Irrespective of the mechanism, we find no evidence for unusual control of infection outcomes in bats. Another limitation of our analysis is that comparatively few viruses (8 out of 54) were inoculated into both rodents and bats. However, when subset to include only these viruses, our models still found no consistent difference between bats and rodents (Section B.2.7 in S1 Appendix), reinforcing the current limited evidence for enhanced tolerance to viral infection in either clade.

Implications for future research

The belief that bats evolved exceptionally generalized or unusually potent mechanisms to suppress virus-induced disease largely underpins the aspiration that bats will be an unusually rich source of broad-acting or potent antivirals. The incompatibility of the currently available data with this premise questions whether the translational value of bat immunity should be expected to be disproportionately high relative to other host groups. Nevertheless, our results do not diminish the potential biomedical value of uncovering the mechanisms by which animals evade severe disease, and useful antivirals may be developed from bat immune systems in the future. Indeed bat ASC2, though in early stages of development, may represent one such biomedical application of bat immunity [13]. Furthermore, the mechanisms that underpin putatively elevated tolerance observed in co-evolved host virus-relationships (e.g., filoviruses and henipaviruses) may yield useful therapeutics targeted to particular viral clades. However, by countering the current narrative that such biomedical breakthroughs will arise disproportionately from studying bats, our study points to the potential value of antiviral discovery in immune systems of other organisms.

Looking forward, it will be useful to identify how specific immunological traits protect across related viruses and, if some host taxa show heightened ability to reduce viral disease, to determine which viruses their immune features may suppress. To this end, we identify major evidence gaps in the current body of experimental inoculations: few bat-associated viruses have been inoculated into rodents, rodent-associated viruses with moderate human CFR have not been inoculated into bats or wild rodents, and relatively few human-associated viruses have been inoculated into either bats or rodents (Fig AH in S1 Appendix). Future comparisons of disease outcomes across host groups should inoculate representative species from each group with the same sets of viruses, comprising those that are both homologous and heterologous with a paired design involving the same doses and inoculation routes. Although it would incur major logistical challenges and costs, such experiments would ideally be carried out using noninvasive proxies for viral load (e.g., using bioluminescent viruses [54]) and would observe animals over sufficient timescales to quantify multiple impacts of infection on host fitness. When known, studies focused on comparing natural immune responses should also aim to use realistic doses and inoculation routes. Our finding that inoculation dose and study size influence observed disease outcomes also suggests that some studies showing limited disease may be artifacts of insufficient powering of experiments for disease detection. This emphasizes the need for larger experiments to detect disease outcomes, and the value of synthesizing data across multiple experiments to detect biases which would be missed by single studies and to make generalizable inferences despite them, as we have done here.

The recurrent emergence of high-impact zoonoses from bats, including ebolaviruses, henipaviruses, and coronaviruses, makes the need to study the relationship between bats and viruses unquestionable. The inconsistency between our results suggesting that disease outcomes in bats follow conventional evolutionary expectations and the pervasive notion of exceptional viral tolerance in bats emphasizes the need to enhance communication between virologists, immunologists, ecologists, and evolutionary biologists. In particular, we urge greater consideration of context dependence in bat-virus interactions through using analogous baselines (e.g., comparing bats and other wildlife rather than bats to humans or laboratory animals), explicit measurement of tolerance versus resistance mechanisms within host individuals, and incorporating robust quantitative consideration of host phylogeny, virus phylogeny, and co-evolutionary associations to differentiate specific outcomes from generalizable conclusions.

Materials and methods

Literature identification

To identify primary scientific articles describing experimental infections of viruses in bats and rodents, we first conducted a literature search via Web of Science (WOS) in the title, abstract, or author keywords (see Section A in S1 Appendix). Terms were selected to include publications mentioning common names for groups of bats and rodents (rodent names based on the NCBI Taxonomy), viruses in general, and terms related to experimental infection. The search conducted on April 14th 2022, via the University of Glasgow returned 2,955 articles. To supplement the WOS search results, we added publications describing virus associations with bats and rodents reported in the CLOVER database [55], which harmonizes data on documented host species from the Global Mammal Parasite Database 2.0 [56], the ENHanCEd Infectious Diseases Database (EID2; [57], the Host-Parasite Phylogeny Project dataset [58], and from [59]. To merge duplicated articles and limit results to those reporting experimental infections, we gathered titles and abstracts for CLOVER articles via PubMed (rentrez R package v1.2.3; [60]). We filtered the CLOVER articles to include those with the same experimental infection terms (Section A in S1 Appendix, lines 11–15) in the title or abstract, before merging them with the WOS articles. The combined set yielded 3,495 unique articles.

Because we sought to identify drivers of viral disease that were the natural product of ecological and life history differences among species, we limited our dataset to animals expected to be immunocompetent with a suite of immune features as indistinguishable from wild individuals as possible (see Section A.1.2 in S1 Appendix for details). Briefly, we excluded publications describing cell line experiments with no in vivo experiments, studies of only laboratory animals commonly available from commercial breeders (unless there was explicit mention of the animals being wild caught). We included experimental transmission studies (cases where a susceptible individual was infected by transmission from an infected individual in captivity), and vaccine studies if they included an experimental infection challenge as the control groups may contain useful information. This resulted in 240 core publications.

Among this core set, we identified 33 review papers. Since reviews are assumed to have conducted detailed literature searches and/or are guided by expert knowledge of specific fields, we used them to expand our set of core publications (see Section A.1.4 in S1 Appendix). First, for each review we identified any cited references to experimental infections and included them in our core set, if not already present. Next, we conducted a WOS search of literature citing these review papers to identify any recent publications that met our study criteria. This search returned 1,247 citations which were then filtered by searching for experimental infection terms (Section A in S1 Appendix, lines 11–15) in the Topic field, yielding 115 publications which were screened in the same manner as the core publications. Additional review papers identified in this step were also screened to identify any cited references to experimental infections.

In total, we identified 336 relevant articles. After excluding review papers (n = 50), articles without readily available full texts, articles in languages other than English, symposium abstracts, theses, and datasets unlinked to a paper, the literature search yielded 246 primary research articles which were screened for disease outcome data. We included publications in which authors reported the hosts as susceptible to infection (by any method of testing), and reported some aspect of disease outcome (clinical disease, mortality, histopathology, behavioral change, symptoms of infection). This yielded 137 articles reporting disease outcomes for susceptible host-virus combinations which were read in detail to create structured data. 105 of these contained individual-level host data, while an additional 32 articles reported data only at the host species level (see Section D in S1 Appendix for references).

Data extraction and harmonization

Extraction of individual-level data.

For studies reporting individual-level data, for each individual, the host species, virus species, virus strain, virus passage history, inoculation route, and inoculation dose were recorded. To document the outcome of infection, we recorded whether each individual was deemed susceptible to infection, was reported to show signs of clinical disease, the severity of disease as described by the authors of the study, whether there was any change in body temperature associated with infection, the degree of histopathology, individual organs or tissues that showed signs of damage, whether individuals died of infection, reported changes in behavior or other notable symptoms, and the amount of time individuals were observed post-inoculation.

Extraction of species-level data.

For the 32 studies where individual-level data were not reported in original articles, we extracted signs of disease mentioned across the entire study population, reporting the maximum of each disease category observed (for example, if infection resulted in mortality of any individuals, we set mortality to “True" for that host-virus-study combination). For articles that provided individual-level data, we collapsed disease outcomes across individuals to identify any instances of disease, mortality, and the maximum observed severity score (see below) for species-level models described below, but retained variation in disease outcome for individual-level models.

Harmonization of species names, known hosts, reservoirs, and host traits.

Virus names were harmonized to ICTV-accepted names via NCBI. Viruses without accepted ICTV names were excluded from further analyses. These included four novel viruses (TRVL 34053-1 virus, Turkmenia-1-74, Lyssavirus Eptesicus isolate, Lyssavirus Rousettus isolate), three unique reassortments of La Crosse and Snowshoe Hare viruses, one virus not identifiable to a contemporary accepted species (Korean Hemorrhagic Fever virus), and two viruses reported only at the genus level (a Lyssavirus and a Poxvirus). Host names were harmonized to the Upham mammal supertree [22] via the Mammal Diversity Database v1.9 [61]. Information on virus host ranges was taken from VIRION [36] and used to augment information on susceptible hosts from our experimental infection data. Information on the taxonomic orders of known reservoir hosts was taken from [41].

Harmonization of signs of infection.

We recorded whether the authors reported the presence of “clinical disease”, representing a binary subjective assessment of whether individuals showed overt signs of infection. We also recorded whether or not individuals died due to infection. Individuals who were culled as part of experimental design, died from anesthesia, or died of unknown causes not attributable to infection were recorded as unknown (NA) for death due to infection. We also recorded any reported behavioral or physical changes following infection, which we grouped into 11 broad categories (Table A in S1 Appendix). Some studies clearly described which aspects of disease manifestation were looked for and documented. For example [62], write “Post-inoculation, individual animals were observed daily for signs of morbidity, or malaise (inappetence, decreased activity, recumbence with reluctance to move, etc.) and clinical lesions or rash for up to 34–41 days depending on the progression of disease”. However, most studies did not clearly define which signs of disease were systematically monitored. As such, signs of disease may be regarded as akin to presence-only data since we do not always have information on when clinical signs were looked for but not found. It is also important to note that by documenting reported signs of disease, we found some individuals reported as showing no “clinical disease” by the authors, but displaying one or more of the signs of disease described above. This raises the issue of subjective assessments when reporting the presence of “clinical disease”. Therefore in our analyses, we construct a binary response variable “presence of disease” which is equal to 1 if there is author-reported clinical disease, disease-induced mortality, or any sign of disease reported.

Creation of disease severity measure.

To investigate disease severity, we created an ordinal rank from 1 to 5 based on a combination of author-reported “clinical disease”, mentions of severity, signs of disease, animal culling for ethical reasons, and death due to infection. We assign severity scores considering the potential impact of each criteria under natural conditions, assigning lower scores to outcomes that are likely to be survivable and higher scores to clinical signs (e.g., impaired movement, internal organ damage) that are likely to lead to death in natural settings. When signs of infection spanned multiple criteria (e.g., loss of appetite and death), we assigned the higher severity score (see Table B in S1 Appendix for specific criteria).

Harmonization of dose and inoculation route.

For studies reporting individual-level data, infectious dose and route of inoculation per individual were recorded. As dose information was reported in various units, we grouped dose units into four main categories: mouse LD50 (predominantly Mouse Intracerebral LD50, but including various mouse ages, mean and median measures, and unreported inoculation sites), Plaque Forming Units (PFU), Tissue Culture Infectious Dose 50 (TCID50, including median, mean, and different tissue cultures as this was rarely reported), and Focal Forming Units (FFU). Mouse LD50 was most commonly reported, followed by PFU, TCID50, then FFU. For each individual, dose per body weight was calculated using interpolated mean adult body weights per species from the COMBINE mammal trait database [63]. We normalize the inoculation dose by mean species body mass in order to allow for more meaningful comparisons across taxa as host body size has been hypothesized to affect viral infections through differences in metabolic rate, immune system responses, and host physiological factors such as organ size, blood volume, number of cells which influence disease progression following inoculation [64–66]. In a comparative study, the same number of viral particles in a given inoculation would result in a relatively lower inoculation dose for larger bodied animals—a minimum dose may lead to a productive infection in two differently sized species, but a larger dose may be needed to cause severe disease in a larger animal. As such, inoculation dose should be normalized to a host specific measure such as dose per gram of body weight. While there is debate about the exact shape of the allometric relationship to use when modeling inoculation-disease scaling [64–66], we argue that use of a simple dose-per-weight metric is necessary to adjust for proportional differences in physiological capacity among host species and the viral load per unit of tissue, offering a fairer comparison of disease severity than the raw estimate of viral particles per inoculation. To determine the sensitivity of our results to this normalization approach, we fit alternative individual-level models using dose uncorrected for body mass (Fig AB in S1 Appendix) and adding a separate body mass effect (Fig AC in S1 Appendix), both of which yielded qualitatively similar results to our main models dose divided by average body mass.

While some studies have conducted experiments to determine the relationship between units, such as [67] finding that PFU and MICLD50 were roughly equivalent for Western equine encephalitis virus strains, exact relationships between dose units were unavailable for most viruses, making it impossible to harmonize dosages into a single comparable scale. Instead, we standardized to units of standard deviation per dose category and include a hierarchical effect of dose category in models adjusting for relative dose. We also recorded inoculation routes when reported and grouped these into six major types: injection into the brain (i.e., intracerebral and intracranial), intramuscular, intraperitoneal, under the skin (subcutaneous, subdermal, or intradermal), oronasal inoculation (including oronasal, oral, nasal, or intranasal), and simulated “natural” infections (e.g., transplacental, contact, arthropod vectors, biting, or aerosols). Dose quantities were not reported for natural infections and treated and could not be included in individual-level models adjusting for relative dose.

Statistical analyses

Hierarchical Bayesian regressions were used to model three disease responses: presence/absence of any detectable disease (severity score >1; Bernoulli response with logit link), mortality (Bernoulli response with logit link), and disease severity (ordinal response with a cumulative probit link). To account for phylogenetic nonindependence and species-level effects, all models included hierarchical effects of host and virus species, separated into phylogenetic (assuming a Brownian motion model of evolution) and nonphylogenetic effects, following the additive quantitative genetic model adapted to interspecific data as in [68] and later recommended by [69] as best practice for meta-analyses in ecology and evolution. The correlation matrix for determining host species-level Brownian phylogenetic effects was calculated from the [22] mammal supertree. Because a single phylogeny for all viruses in the study is not available, we instead used a taxonomic tree representation for viruses, based on ICTV taxonomy as reported by NCBI on February 23, 2024, with branch lengths of unit 1 per taxonomic level, subsetting the taxonomy to include only Superkingdom, Realm, Kingdom, Phylum, Class, Order, Family, Genus, and Species, following [41]. For species-level and individual-level models, we additionally include a hierarchical effect of article to adjust for potential variation in study design unaccounted for by continuous predictors (e.g., numbers of individuals, doses, and inoculation routes).

To aid in the comparison of effect sizes across continuous and binary predictors, all continuous predictors were centered and scaled to a standard deviation of 0.5 [70], with number of individuals (species-level models) and dose/mass (individual-level models) log-transformed prior to scaling and centering. Models were fit in Stan version 2.21.0 [71] using the R package brms version 2.16.3 [72]. All models used Normal(0, 1.5) priors for regression coefficients and Normal(0, 1) priors for hierarchical variance parameters, as recommended by [73]. Full models were run across four independent chains, with a minimum of 4,000 iterations per chain. For each chain, the first 50% of iterations were used as burn-in and discarded, resulting in a minimum of 8,000 post-warmup posterior draws per model. We diagnosed convergence by observation of Rhat values equal to 1.0 for all estimated parameters, and ensured models fit well using posterior predictive checks.

Dose and inoculation route.

As the reported inoculation doses varied substantially across and within studies (Fig D in S1 Appendix), we included an effect of dose divided by the average body mass of the infected species in individual-level models. To include doses measured through different dose units (e.g., Mouse LD50, TCID50, and PFU), dose/g was log-transformed, then standardized and centered relative to the mean and standard deviation per dose unit. Models including dose included this continuous dose effect as well as a hierarchical effect of dose unit. In addition to dose, studies varied the inoculation routes used during experimental infection (Fig G in S1 Appendix). Since some inoculation routes, such as intracerebral infection, might lead to more severe disease compared to others, such as oronasal exposure, inoculation route was included as a hierarchical effect. In all models, dose/g showed a positive effect, indicating that increasing dose relative to body mass led to a higher probability of showing signs of disease, mortality, and disease severity.

Eco-evolutionary metrics.

To test whether virus host specificity and the evolutionary position of inoculated hosts influence the presence and severity of disease, we include three eco-evolutionary predictors. Hypothesizing that costs of generalism [31,33] lead specialist viruses to cause more severe disease, we included a measure of host phylogenetic diversity (more specifically the standardized effect size of phylogenetic distances among hosts—ses.pd; section A.2.4 in S1 Appendix), with larger values indicating viruses with greater phylogenetic host ranges. Across all models, host phylogenetic diversity showed a mean negative effect.

We also wanted to test the hypothesis that viruses inoculated into host species that are distantly related to their co-adapted hosts may cause more severe disease [15]. One measure of potential host adaptation is whether the inoculated host is related to known reservoir hosts. However, the reservoir hosts of many viruses included in our dataset were unknown [74]. We therefore used two proxy variables. The first, HEI [15], is calculated as the mean phylogenetic distance from all documented hosts to the inoculated host. As pathogens commonly infect closely related host species [33,37,38], this metric assumes that the phylogenetic centroid of known susceptible species reflects the most likely position of hosts to which a virus is best adapted, and that the distance from an inoculated host to this centroid provides a reasonable proxy for the relative extent of co-adaptation [15]. As this metric provides a continuous proxy of adaptation and can be calculated for all viruses in our dataset, we included it in all models. While estimated effect sizes showed high uncertainty with 95% credible intervals crossing zero, the positive effect of HEI is consistent with previous models of infection-induced mortality for domesticated-animal pathogens [15], indicating that inoculating hosts distantly related to other documented hosts may result in overt disease and more severe disease, potentially reflecting maladaptive virulence [15,31,40].

As HEI does not take into account information on the relative roles of host species in transmission (e.g., reservoir versus spillover host), which may reflect a different type of host adaptation, we fit a set of sensitivity models adding a binary effect of “reservoir match". Using the taxonomic orders of documented reservoirs from [41] available for 36 viruses in our dataset, we constructed a binary metric which is “True" if the inoculated host belongs to any of the known reservoir orders, and “False" if not. As this metric was not available for all viruses in our dataset [74], and captures a different aspect of host adaptation, we explored it’s effect through sensitivity models with it included as an additional predictor, and as a replacement for HEI (Section B.2.2 in S1 Appendix). The binary “reservoir match" also allowed us to run a further set of sensitivity models restricting our analysis to heterologous inoculations (i.e., viruses inoculated into species from nonreservoir groups) and homologous inoculations (i.e., viruses inoculated into species from presumed reservoir groups; Section B.2.3 in S1 Appendix). These analyses offer an opportunity to evaluate whether bats’ capacity to tolerate viruses extends to viruses for which they are unlikely to have evolved defenses. We find that while homologous infections include smaller distances of HEI, there is considerable overlap in HEI for homologous and heterologous inoculations (Fig N in S1 Appendix). Furthermore, as some viruses have presumed reservoirs in multiple host orders, plotting HEI across the presumed reservoir groups shows that inoculations associated with low HEI interactions tend to be homologous infections, but that some homologous infections include interactions marked by high evolutionary isolation.

Previous research found that increasing HEI is associated with increased mortality among domesticated animal pathogens, but this comes at the cost of decreased transmission [15], suggesting that similar trends in our data may represent maladaptive virulence [31,40]. Sensitivity models including a binary reservoir match effect (with or without HEI) showed some positive effects of reservoir match (Section B.2.2 in S1 Appendix). While this may be unexpected if long-term co-evolution leads to increased disease tolerance, these results may instead reflect adaptive virulence [42] in which some level of virulence is necessary for transmission and maintenance in nature. Alternatively, as our metric of reservoir match does not capture whether the exact host species being tested is a known reservoir (Fig O in S1 Appendix), positive effects on disease outcomes may be explained by viruses being well adapted to the physiology of the hosts in the same taxonomic order, but not to the specific host species used in experimental inoculations, which may have experienced little to no selection to tolerate disease caused by these viruses.

Comparison to human CFR

Using human CFR, reported by [18] as a comparable measure of pathogenicity, we evaluated whether inoculation of bats with disproportionately pathogenic viruses could have disguised true differences in tolerance relative to rodents. We find that while bats are more often inoculated with viruses that cause high mortality in human infections, there is a large variation in the maximum severity observed in inoculated host species, including large variation in bats among viruses responsible for high human CFR (Fig AF in S1 Appendix). Patterns are similar when considering infection-fatality rates per study combinations (Fig AG in S1 Appendix).

To investigate whether putatively heightened tolerance in bats may be obscured by bats being more often inoculated with highly pathogenic viruses, we fit a series of models predicting mean human case fatality rates (CFR) as a function of the order of inoculated hosts (Fig AI in S1 Appendix). We treated mean human CFR as a proportion (0.0–1.0) rather than a percentage and modeled it using zero-one-inflated beta models in brms. We used brms default priors and assessed convergence of all models using R-hat = 1.0 and visual inspection of model fits using posterior predictive checks. All models included a binary predictor of inoculated host order and a study-level hierarchical effect. The base model with no virus hierarchical effects and including all viruses in our study identified that bats are more often infected with high human CFR viruses, which aligns with the raw data presented in Fig AF in S1 Appendix. However, when we exclude bat-reservoired viruses, or model all viruses but include virus hierarchical effects, we find no difference between bats and rodents. Similarly, reducing models to include only heterologous host-virus pairs (viruses with known reservoirs inoculated into nonreservoirs), we find no difference between bats and rodents in mean CFR. As such, if genuinely heightened tolerance was undetectable in our dataset due to biases in highly pathogenic viruses being selected for experiments, it would mostly reflect bats experiencing less than expected disease from bat-associated viruses—consistent with the expectation that evolutionary association leads to virus-specific adaptations.

Acknowledgments

We thank members of the Streicker group, the MRC-University of Glasgow Centre for Virus Research, the University of Glasgow School of Biodiversity, One Health and Veterinary Medicine, and Nicole Mideo for valuable feedback.

References

  1. 1. Baker ML, Schountz T, Wang L-F. Antiviral immune responses of bats: a review. Zoonoses Public Health. 2013;60(1):104–16. pmid:23302292
  2. 2. Brook CE, Dobson AP. Bats as “special” reservoirs for emerging zoonotic pathogens. Trends Microbiol. 2015;23(3):172–80. pmid:25572882
  3. 3. Calisher CH, Childs JE, Field HE, Holmes KV, Schountz T. Bats: important reservoir hosts of emerging viruses. Clin Microbiol Rev. 2006;19(3):531–45. pmid:16847084
  4. 4. O’Shea TJ, Cryan PM, Cunningham AA, Fooks AR, Hayman DTS, Luis AD, et al. Bat flight and zoonotic viruses. Emerg Infect Dis. 2014;20(5):741–5. pmid:24750692
  5. 5. Shen Y-Y, Liang L, Zhu Z-H, Zhou W-P, Irwin DM, Zhang Y-P. Adaptive evolution of energy metabolism genes and the origin of flight in bats. Proc Natl Acad Sci U S A. 2010;107(19):8666–71. pmid:20421465
  6. 6. Wang L-F, Walker P, Poon LLM. Mass extinctions, biodiversity and mitochondrial function: are bats ‘special’ as reservoirs for emerging viruses? Curr Opin Virol. 2011;1(6):649–57.
  7. 7. Wong S, Lau S, Woo P, Yuen K-Y. Bats as a continuing source of emerging infections in humans. Rev Med Virol. 2007;17(2):67–91. pmid:17042030
  8. 8. Banerjee A, Baker ML, Kulcsar K, Misra V, Plowright R, Mossman K. Novel insights into immune systems of bats. Front Immunol. 2020;11:26. pmid:32117225
  9. 9. Irving AT, Ahn M, Goh G, Anderson DE, Wang L-F. Lessons from the host defences of bats, a unique viral reservoir. Nature. 2021;589(7842):363–70. pmid:33473223
  10. 10. Morales AE, Dong Y, Brown T, Baid K, Kontopoulos D-G, Gonzalez V, et al. Bat genomes illuminate adaptations to viral tolerance and disease resistance. Nature. 2025;638(8050):449–58. pmid:39880942
  11. 11. Baid K, Irving AT, Jouvenet N, Banerjee A. The translational potential of studying bat immunity. Trends Immunol. 2024;45(3):188–97. pmid:38453577
  12. 12. Dolgin E. Bat biotech takes flight. Nat Biotechnol. 2023;41(8):1047–52. pmid:37460678
  13. 13. Ahn M, Chen VC-W, Rozario P, Ng WL, Kong PS, Sia WR, et al. Bat ASC2 suppresses inflammasomes and ameliorates inflammatory diseases. Cell. 2023;186(10):2144-2159.e22. pmid:37172565
  14. 14. Shaw AE, Hughes J, Gu Q, Behdenna A, Singer JB, Dennis T, et al. Fundamental properties of the mammalian innate immune system revealed by multispecies comparison of type I interferon responses. PLoS Biol. 2017;15(12):e2004086. pmid:29253856
  15. 15. Farrell MJ, Davies TJ. Disease mortality in domesticated animals is predicted by host evolutionary relationships. Proc Natl Acad Sci U S A. 2019;116(16):7911–5. pmid:30926660
  16. 16. Kemenesi G, Kurucz K, Dallos B, Zana B, Földes F, Boldogh S, et al. Re-emergence of lloviu virus in miniopterus schreibersii bats, hungary, 2016. Emerg Microbes Infect. 2018;7(1):1–4.
  17. 17. Kohl C, Brinkmann A, Radonić A, Dabrowski PW, Nitsche A, Mühldorfer K, et al. Zwiesel bat banyangvirus, a potentially zoonotic Huaiyangshan banyangvirus (Formerly known as SFTS)-like banyangvirus in Northern bats from Germany. Sci Rep. 2020;10(1):1370. pmid:31992832
  18. 18. Guth S, Mollentze N, Renault K, Streicker DG, Visher E, Boots M, et al. Bats host the most virulent-but not the most dangerous-zoonotic viruses. Proc Natl Acad Sci U S A. 2022;119(14):e2113628119. pmid:35349342
  19. 19. Brook CE, Rozins C, Guth S, Boots M. Reservoir host immunology and life history shape virulence evolution in zoonotic viruses. PLoS Biol. 2023;21(9):e3002268. pmid:37676899
  20. 20. Mandl JN, Schneider C, Schneider DS, Baker ML. Going to bat(s) for studies of disease tolerance. Front Immunol. 2018;9:2112. pmid:30294323
  21. 21. Medzhitov R, Schneider DS, Soares MP. Disease tolerance as a defense strategy. Science. 2012;335(6071):936–41. pmid:22363001
  22. 22. Upham NS, Esselstyn JA, Jetz W. Inferring the mammal tree: species-level sets of phylogenies for questions in ecology, evolution, and conservation. PLoS Biol. 2019;17(12):e3000494. pmid:31800571
  23. 23. Pereira PDC, Diniz DG, da Costa ER, Magalhães NG de M, da Silva A de JF, Leite JGS, et al. Genes, inflammatory response, tolerance, and resistance to virus infections in migratory birds, bats, and rodents. Front Immunol. 2023;14:1239572. pmid:37711609
  24. 24. Albery GF, Becker DJ. Fast-lived hosts and zoonotic risk. Trends Parasitol. 2021;37(2):117–29. pmid:33214097
  25. 25. Hayman DTS. Bat tolerance to viral infections. Nat Microbiol. 2019;4(5):728–9. pmid:31015739
  26. 26. Han BA, Schmidt JP, Bowden SE, Drake JM. Rodent reservoirs of future zoonotic diseases. Proc Natl Acad Sci U S A. 2015;112(22):7039–44. pmid:26038558
  27. 27. Luis AD, Hayman DTS, O’Shea TJ, Cryan PM, Gilbert AT, Pulliam JRC, et al. A comparison of bats and rodents as reservoirs of zoonotic viruses: are bats special? Proc Biol Sci. 2013;280(1756):20122753. pmid:23378666
  28. 28. Boots M, Donnelly R, White A. Optimal immune defence in the light of variation in lifespan. Parasite Immunol. 2013;35(11):331–8. pmid:23869870
  29. 29. Miller MR, White A, Boots M. Host life span and the evolution of resistance characteristics. Evolution. 2007;61(1):2–14. pmid:17300423
  30. 30. Galen SC, Ray S, Henry M, Weckstein JD. Parasite-associated mortality in birds: the roles of specialist parasites and host evolutionary distance. Biol Lett. 2022;18(4):20210575. pmid:35414225
  31. 31. Leggett HC, Buckling A, Long GH, Boots M. Generalism and the evolution of parasite virulence. Trends Ecol Evol. 2013;28(10):592–6. pmid:23968968
  32. 32. Agueldo Romero P, Elena SF. The degree of plant resilience to infection correlates with virus virulence and host-range. Span J Agric Res. 2008;6:160–9.
  33. 33. Antonovics J, Boots M, Ebert D, Koskella B, Poss M, Sadd BM. The origin of specificity by means of natural selection: evolved and nonhost resistance in host-pathogen interactions. Evolution. 2013;67(1):1–9. pmid:23289557
  34. 34. Abolins S, King EC, Lazarou L, Weldon L, Hughes L, Drescher P, et al. The comparative immunology of wild and laboratory mice, Mus musculus domesticus. Nat Commun. 2017;8:14811. pmid:28466840
  35. 35. Farrell M, Tucker S, Mollentze N, Streicker D. Data and code to reproduce the analyses in “viral disease outcomes are indistinguishable between experimentally infected bats and rodents”. bat_rodent_inoculations; 2026. https://doi.org/10.5281/zenodo.15195673
    • 36. Carlson CJ, Gibb RJ, Albery GF, Brierley L, Connor RP, Dallas TA, et al. The Global Virome in One Network (VIRION): an Atlas of Vertebrate-Virus Associations. mBio. 2022;13(2):e0298521. pmid:35229639
    • 37. Davies TJ, Pedersen AB. Phylogeny and geography predict pathogen community similarity in wild primates and humans. Proc Biol Sci. 2008;275(1643):1695–701. pmid:18445561
    • 38. Park AW, Farrell MJ, Schmidt JP, Huang S, Dallas TA, Pappalardo P, et al. Characterizing the phylogenetic specialism-generalism spectrum of mammal parasites. Proc Biol Sci. 2018;285(1874):20172613. pmid:29514973
    • 39. Guth S, Visher E, Boots M, Brook CE. Host phylogenetic distance drives trends in virus virulence and transmissibility across the animal-human interface. Philos Trans R Soc Lond B Biol Sci. 2019;374(1782):20190296. pmid:31401961
    • 40. Longdon B, Hadfield JD, Day JP, Smith SCL, McGonigle JE, Cogni R, et al. The causes and consequences of changes in virulence following pathogen host shifts. PLoS Pathog. 2015;11(3):e1004728. pmid:25774803
    • 41. Mollentze N, Streicker DG. Viral zoonotic risk is homogenous among taxonomic orders of mammalian and avian reservoir hosts. Proc Natl Acad Sci U S A. 2020;117(17):9423–30. pmid:32284401
    • 42. Cressler CE, McLeod DV, Rozins C, Van Den Hoogen J, Day T. The adaptive evolution of virulence: a review of theoretical predictions and empirical tests. Parasitology. 2016;143(7):915–30. pmid:26302775
    • 43. Poulin R, Combes C. The concept of virulence: interpretations and implications. Parasitol Today. 1999;15(12):474–5. pmid:10557145
    • 44. Pei G, Balkema-Buschmann A, Dorhoi A. Disease tolerance as immune defense strategy in bats: One size fits all? PLoS Pathog. 2024;20(9):e1012471. pmid:39236038
    • 45. Williamson MM, Hooper PT, Selleck PW, Gleeson LJ, Daniels PW, Westbury HA, et al. Transmission studies of Hendra virus (equine morbillivirus) in fruit bats, horses and cats. Aust Vet J. 1998;76(12):813–8. pmid:9972433
    • 46. Guito JC, Prescott JB, Arnold CE, Amman BR, Schuh AJ, Spengler JR, et al. Asymptomatic infection of marburg virus reservoir bats is explained by a strategy of immunoprotective disease tolerance. Curr Biol. 2021;31(2):257-270.e5. pmid:33157026
    • 47. Tsugawa T, Tatsumi M, Tsutsumi H. Virulence-associated genome mutations of murine rotavirus identified by alternating serial passages in mice and cell cultures. J Virol. 2014;88(10):5543–58. pmid:24599996
    • 48. Elena SF, Olmo-Uceda MJ, Ambrós S. Host replacement dynamics shape virulence and genome evolution in a plant RNA virus. Virus Evol. 2025. https://doi.org/10.1093/ve/veaf062
      • 49. Peña J, Chen-Harris H, Allen JE, Hwang M, Elsheikh M, Mabery S, et al. Sendai virus intra-host population dynamics and host immunocompetence influence viral virulence during in vivo passage. Virus Evol, 2(1):vew008, 2016.
      • 50. Kubinak JL, Cornwall DH, Hasenkrug KJ, Adler FR, Potts WK. Serial infection of diverse host (Mus) genotypes rapidly impedes pathogen fitness and virulence. Proc Biol Sci. 2015;282(1798):20141568. pmid:25392466
      • 51. Lezcano OM, Fuhrmann L, Ramakrishnan G, Beerenwinkel N, Huynen MA, van Rij RP. Parallel evolution and enhanced virulence upon in vivo passage of an RNA virus in Drosophila melanogaster. Virus Evol. 2023;9(2):vead074. pmid:38162315
      • 52. Ebert D. Experimental evolution of parasites. Science. 1998;282(5393):1432–5. pmid:9822369
      • 53. Imrie RM, Wallace MA, Longdon B. Positive correlations in susceptibility to a diverse panel of viruses across host species. bioRxiv. 2024:2024.10.09.617458.
        • 54. Stading BR, Osorio JE, Velasco-Villa A, Smotherman M, Kingstad-Bakke B, Rocke TE. Infectivity of attenuated poxvirus vaccine vectors and immunogenicity of a raccoonpox vectored rabies vaccine in the Brazilian Free-tailed bat (Tadarida brasiliensis). Vaccine. 2016;34(44):5352–8. pmid:27650872
        • 55. Gibb R, Albery GF, Becker DJ, Brierley L, Connor R, Dallas TA, et al. Data proliferation, reconciliation, and synthesis in viral ecology. BioScience. 2021;71(11):1148–56.
        • 56. Stephens PR, Pappalardo P, Huang S, Byers JE, Farrell MJ, Gehman A, et al. Global mammal parasite database version 2.0. Ecology. 2017;98(5):1476. pmid:28273333
        • 57. Wardeh M, Risley C, McIntyre MK, Setzkorn C, Baylis M. Database of host-pathogen and related species interactions, and their global distribution. Sci Data. 2015;2:150049. pmid:26401317
        • 58. Olival KJ, Hosseini PR, Zambrana-Torrelio C, Ross N, Bogich TL, Daszak P. Host and viral traits predict zoonotic spillover from mammals. Nature. 2017;546(7660):646–50. pmid:28636590
        • 59. Shaw LP, Wang AD, Dylus D, Meier M, Pogacnik G, Dessimoz C, et al. The phylogenetic range of bacterial and viral pathogens of vertebrates. Mol Ecol. 2020;29(17):3361–79. pmid:32390272
        • 60. Winter DJ. Rentrez: an R package for the NCBI eutils API. The R Journal. 2017;9:520–6.
        • 61. Database, MD. Mammal diversity database; 2022
          • 62. Hutson CL, Gallardo-Romero N, Carroll DS, Clemmons C, Salzer JS, Nagy T, et al. Transmissibility of the monkeypox virus clades via respiratory transmission: investigation using the prairie dog-monkeypox virus challenge system. PLoS One. 2013;8(2):e55488. pmid:23408990
          • 63. Soria CD, Pacifici M, Di Marco M, Stephen SM, Rondinini C. COMBINE: a coalesced mammal database of intrinsic and extrinsic traits. Ecology. 2021;102(6):e03344. pmid:33742448
          • 64. Cable JM, Enquist BJ, Moses ME. The allometry of host-pathogen interactions. PLoS One. 2007;2(11):e1130. pmid:17987117
          • 65. Althaus CL. Of mice, macaques and men: scaling of virus dynamics and immune responses. Front Microbiol. 2015;6:355. pmid:25954270
          • 66. Banerjee S, Perelson AS, Moses M. Modelling the effects of phylogeny and body size on within-host pathogen replication and immune response. J R Soc Interface. 2017;14(136):20170479. pmid:29142017
          • 67. Hardy JL, Reeves WC, Rush WA, Nir YD. Experimental infection with western equine encephalomyelitis virus in wild rodents indigenous to Kern County, California. Infect Immun. 1974;10(3):553–64. pmid:4426699
          • 68. Schmidt JP, Davies TJ, Farrell MJ. Opposing macroevolutionary and trait-mediated patterns of threat and naturalisation in flowering plants. Ecol Lett. 2021;24(6):1237–50. pmid:33786974
          • 69. Cinar O, Nakagawa S, Viechtbauer W. Phylogenetic multilevel meta‐analysis: a simulation study on the importance of modelling the phylogeny. Methods Ecol Evol. 2021;13(2):383–95.
          • 70. Gelman A. Scaling regression inputs by dividing by two standard deviations. Stat Med. 2008;27(15):2865–73. pmid:17960576
          • 71. Stan Development Team. Stan modeling language users guide and reference manual; 2023.
            • 72. Bürkner P-C. brms: an R package for bayesian multilevel models using stan. J Stat Soft. 2017;80(1).
            • 73. McElreath R. Statistical rethinking: a bayesian course with examples in R and STAN. 2nd ed. Chapman and Hall/CRC; 2020.
              • 74. Babayan SA, Orton RJ, Streicker DG. Predicting reservoir hosts and arthropod vectors from evolutionary signatures in RNA virus genomes. Science. 2018;362(6414):577–80. pmid:30385576
              Read Entire Article
              1. Homepage
              2. Biology
              3. Viral disease outcomes are indistinguishable between experimentally infected bats and rodents

              Related

              Common pesticide linked to 2.5 times the risk of Parkinson’s disease

              Common pesticide linked to 2.5 times the risk of Parkinson’s...

              7 hours ago 10
              Scientists shrink Archimedes’ screw to the size of a single molecule

              Scientists shrink Archimedes’ screw to the size of a single ...

              8 hours ago 10
              Scientists use sound waves to deliver a new attack on deadly brain cancer

              Scientists use sound waves to deliver a new attack on deadly...

              9 hours ago 7
              Scientists say a forgotten particle may explain why the universe exists

              Scientists say a forgotten particle may explain why the univ...

              15 hours ago 14
              Learning the Sliding Filament Model of Muscle Contraction

              Learning the Sliding Filament Model of Muscle Contraction

              1 day ago 19
              Scientists discover metals that defy the deep freeze at -424°F

              Scientists discover metals that defy the deep freeze at -424...

              1 day ago 17
              Scientists discover a bacterial defense so strange they thought it was broken

              Scientists discover a bacterial defense so strange they thou...

              1 day ago 13

                       

               Advertising byAdpathway

              Trending

              Popular

              OrgoLife Quantum Energy MedBed Technology® — The Breakthrough Wellness Revolution of 2025 Arrives in Magog and Expands Worldwide

              OrgoLife Quantum Energy MedBed Technology® — The Breakthroug...

              10 months ago 51594
              EXCLUSIVE INVESTIGATIVE REPORT — 2025: “I Spent a Full Day Inside Canada’s Quantum MedBed Center… and My Life May Never Be the Same Again.”

              EXCLUSIVE INVESTIGATIVE REPORT — 2025: “I Spent a Full Day I...

              10 months ago 39882
              Journalisme poubelle : Disséquer le Journal JDM à Montréal *Auteur : Luke Denis, journaliste indépendant – 2025

              Journalisme poubelle : Disséquer le Journal JDM à Montréal *...

              10 months ago 23311
              Silence, on manipule : La machine médiatique contre la vérité *Auteur : Luke Denis, journaliste indépendant – 2025 TVPoubelle et le Journal à Montreal

              Silence, on manipule : La machine médiatique contre la vérit...

              1 year ago 6764
              9 Best Goldendoodle Immunity Supplements

              9 Best Goldendoodle Immunity Supplements

              10 months ago 3926
              Control of mitochondrial dynamics by the metabolic regulator dPGC1 limits Yorkie-induced oncogenic growth in Drosophila

              Control of mitochondrial dynamics by the metabolic regulator...

              10 months ago 3839
              First Due ranked #115 on the 2025 Deloitte Technology Fast 500

              First Due ranked #115 on the 2025 Deloitte Technology Fast 5...

              11 months ago 3780
              Flea & Tick Prevention for Border Collies: A Safe & Effective 4 Step Plan

              Flea & Tick Prevention for Border Collies: A Safe & Effectiv...

              10 months ago 3534
              Architects gain a new superpower for complex curved designs

              Architects gain a new superpower for complex curved designs

              10 months ago 2891
              Next gen cancer drug shows surprising anti aging power

              Next gen cancer drug shows surprising anti aging power

              10 months ago 2843
              Astronomers find a planet orbiting at a wild angle no one can explain

              Astronomers find a planet orbiting at a wild angle no one ca...

              10 months ago 2806
              New moonquake discovery could change NASA’s Moon plans

              New moonquake discovery could change NASA’s Moon plans

              10 months ago 2722

                       

               Ads byAdpathway

              English (US) English (US) · French (CA) French (CA) ·
              About Us · Contact Us · Terms & Conditions · Advertising · News Syndication RSS · Cookies · Privacy Policy ·

              © AllYouCanFind.ca for the Everyday Planet Earth & all living Creatures Global News ! 2026. All rights are reserved

                       

                      

              Start the new Vibrations with a Medbed Franchise today!  

              Protect your whole family with Quantum Orgo-Life® devices

                Advertising by Adpathway