Changing resource landscapes and spillover of henipaviruses

Bat distribution

Infections are harbored within reservoir hosts, therefore the reservoir host distribution broadly delineates the areas of spillover risk, tempered by the spatiotemporal dynamics of pathogens within the populations, survival or dispersal of pathogens outside of reservoir hosts, overlap with the distribution of potential recipient hosts, and behaviors that increase recipient host risk of exposure⁵⁵. Anthropogenic pressure from habitat loss, fragmentation, and agricultural and urban development interact to redistribute resources; critical habitat patches that sustain animal populations or movements are removed, alternative resources are introduced outside the species’ native range, and the resource structure in small remnant habitat patches is modified by edge effects^52,56,57. Wildlife responses to changes in the availability and quality of resources across the landscape are species-specific, ranging from population-level decline and local extinction to successful adaptation or exploitation of urban environments^58–61; thus, the effects on wildlife distributions and the subsequent effects on spillover risk must be considered separately for each zoonotic disease system.

The distribution of flying foxes and other pteropodids that harbor HNVs is strongly modulated by the distribution and phenology of their diet foods. Many species move long distances across the landscape, driven by ephemeral floral and fruit resources in native forests^62–66. For example, populations of the straw-colored fruit bat (Eidolon helvum) annually migrate 1000 km from the Democratic Republic of Congo to feed on fruit in central Zambia⁶². However, few species, if any, are conventionally migratory. Instead, many pteropodids are nomadic, exhibiting irregular patterns of movement associated with complex fluctuations in the abundance of food resources across large landscapes⁶⁴. In Australia, local abundance of flying foxes can exponentially increase during highly productive flowering events^67,68. Both the quality and abundance of resources seem to determine the attractiveness of a habitat patch. Small stands or even isolated trees, if highly productive and nutritious, can attract large numbers of bats^43,69–71. The longevity, productivity, and diversity of diet species influences the time bats spend in one location^68,72,73, and nomadic searching continues once the resource pulse ends^72,74,75. Their dependence on concentrated, ephemeral sources of high-yield diet plants^70,75,76 makes nomadic bats especially vulnerable to small changes in their habitat. Habitat loss reduces not only the quantity of available food but also the connectivity of foraging patches if seasonally important resources are removed. By contrast, resource provisioning, which is biased towards monoculture (e.g., fruit orchards) and introduced species, changes the composition and seasonality of available food and the overall nutritional landscape.

Research on the effects of habitat fragmentation on frugivorous and nectarivorous bats has disproportionately focused on Neotropical species; surprisingly little is understood about the direct impacts of land-use change on pteropodid bats, or to what degree habitat fragmentation and increased distance between resource patches is tolerated (see Ref. 77). However, abundant evidence demonstrates that altered resource landscapes are changing the movement behaviors of pteropodids. Migratory E. helvum and Epomophorus wahlbergii have formed colonies in African cities that now persist year round^76,78–80, mirroring the behavior of several flying fox species in Australia and Asia^81,82. In Australia, periods of resource scarcity driven by habitat loss may drive flying foxes to seek alternative food sources in urbanized areas, where they form sedentary subpopulations^69,83,84. A similar attraction to anthropogenic food operates within African pteropodid systems. For example, Madagascan fruit bats (E. dupreanum and P. rufus) increasingly forage within agricultural landscapes in response to habitat change^85,86, and E. helvum on the island of Príncipe favor year-round fruit trees provided by agricultural plantations^87,88. Population numbers of urban E. helvum in Accra, Ghana increase during the dry season, when bats preferentially forage on native floral resources. Although much of the population departs during periods of resource concentration, the remaining bats rely on cultivated trees in urban habitats⁷⁸. Similarly, stopover duration during migration of E. helvum is driven by food supply⁶². In Bangladesh, roosting ecology is associated with forest fragmentation and food supplementation, with greater numbers of P. medius roosts reported in areas with higher human population density, more fragmented forest, and a diversity of planted food resources^46,82. Similar patterns are being observed in numerous species elsewhere around the world, with additional studies from Australia, Thailand (P. lylei), Fiji (P. tonganus), India (Cynopterus sphinx), Pakistan (P. medius), Japan (P. dasymallus inopinautus), Malaysia (P. vampyrus and P. hypomelanus) suggesting that food resource limitation following habitat loss drives fruit bats to feed and roost in urban and agricultural landscapes^{42–47,89–91}.

Such changes in flying fox distribution in response to urban and agricultural habitats could change the spatial dynamics of henipaviruses and the risk of spillover. Urban habituation increases the overlap of bats and spillover recipient hosts of disease⁴⁰. Many of the affected species are known reservoirs of HeV and NiV in Australia and Southeast Asia or are suspected reservoirs of African henipaviruses (e.g., E. helvum)^83,90,92,93. For example, attraction of P. medius to date palm sap collection pots facilitates foodborne transmission of NiV to humans in Bangladesh^94–96. Villages in central and northwestern Bangladesh where most NiV spillovers occur (i.e., the Nipah Belt) are also characterized by fragmented forests that support more P. medius roosts, likely by providing more consistent food via household gardens and agroforestry⁸². Similarly, attraction of P. vampyrus to mango plantations facilitated spillover of NiV from bats to pigs in Malaysia^41,97. However, theoretical models provide a range of predictions for the consequences of urban habituation on the distribution of pathogen across bat populations. Resource shifts from habitat loss or resource provisioning that changes flying fox site fidelity and migratory behavior or stopover patterns could also influence HNV prevalence in bat populations^62,98–100 and spillover risks¹⁰¹. For pathogens with long infectious periods relative to bat movement rates, urban colonization could spread pathogens across new areas near humans and domestic animals. Alternatively, if these pathogens have short infectious periods and long recovery periods, the disease will infect quickly and fade out of the population in the absence of reintroduction^39,102. Over time, births and other processes replenish the number of susceptible hosts. Thus, a decrease in connectivity, which potentially occurs in sedentary urban populations, could lead to extinction–colonization dynamics of virus, resulting in larger outbreaks when virus is reintroduced to susceptible populations⁴⁰.

Bat density

The mode of pathogen transmission plays a major role in determining how host density changes pathogen dynamics and persistence within populations and spillover from reservoirs to new hosts. Transmission tends towards either density-dependence, where infectious contacts between hosts increase with population density (e.g., most acute, directly transmitted infections), or frequency-dependence, where infectious contacts remain constant regardless of host population density (e.g., pathogens transmitted through sexual partners or insect vectors)¹⁰³. Many pathogens, particularly those indirectly transmitted via contaminated objects (i.e., fomites) and latent infections, often fall somewhere between the extremes of density- and frequency-dependent transmission^39,104. This direction may shift seasonally; spatiotemporal resource patterns influence births, deaths and dispersal that subsequently affect the density of reservoir populations. High-quality habitat and resource abundance can increase population density by improving birth rates, juvenile survival, and immigration into the population, while food shortages and poor-quality habitat can have the opposite effect. Resources may also change contact patterns among reservoir hosts and between reservoir and recipient hosts. Resources can also decouple the density–habitat quality relationship¹⁰⁵. For example, concentration of resources in shrinking wildlife habitat can locally increase wildlife population densities, while human-mediated or environmental events can disperse or relocate wildlife populations to inferior habitats.

In pteropodids, change in local population density in response to resource availability and quality may be species-specific. Cyclones and fire in the 1990s removed much of the traditional food resources for flying foxes on the islands of Samoa, but the behavioral response of native pteropodids (P. tonganus and P. samoensis) affected overall population density in different ways. After the primary diet resource (Syzgium nectar) for P. tonganus was removed, bats travelled greater distances to feed on fallen cultivated fruits in villages¹⁰⁶. Hunting and predation of bats in villages combined with juvenile mortality driven by severe food shortages led to a 90% population decline. By contrast, P. samoensis responded to the food shortage by shifting their diet to incorporate unusual native foods and foraging close to their day roosts, in turn only suffering a 10% local population reduction¹⁰⁶. The density of P. alecto and P. poliocephalus also changes in response to food scarcity, as these species tend to aggregate near patches of productive floral habitat^68,107. Urban resource availability has been proposed as the driver of the exponential increases in the number of urban roosts of Australian Pteropodids¹⁰⁸. However, few studies determine whether population sizes and demographic rates change in new resource landscapes (e.g., fecundity and survival)⁴⁹. This is in part owing to difficulties in quantifying abundance of wide-ranging, mobile species¹⁰⁹.

Changes in flying fox density driven by altered resource landscapes could dramatically affect the local and spatial dynamics of HNVs. Pteropodid bats live in dense arboreal colonies, and may experience near continuous exposure to low level viral excretion in urine and feces³⁵. Three-dimensional roost structures may facilitate indirect transmission of virus and lead to higher exposure within roosts. If henipavirus dynamics in bat populations are driven by density-dependent transmission, then increasing local population density would increase virus excretion and spillover risk⁴⁹. However, it is unclear how bat density scales with roost size and little empirical support exists for HNV dynamics being solely driven by density-dependent processes^35,39. In addition, viral dynamics in bat populations that are driven by transmission versus within-host processes (e.g., reactivation of latent infections) will exhibit different responses to resource restriction or bat density³⁹. Thus, mechanisms of maintenance in reservoir populations play a role in determining how HNV spillover risk scales with host density, although little ecological or laboratory evidence exists to support particular scenarios.

In an ecological context, resource provisioning changes patch occupancy times and potentially colony size, which theory predicts would facilitate landscape-level spread of viruses^52,110,111. In Australia, conflicting associations between flying fox density and spillover have been reported^101,112,113. Smith et al.¹¹² found that Australian flying fox density correlated positively with clusters of equine HeV cases. However, their study used a database of P. alecto occurrence observations collated into presence/absence data, rather than population counts, suggesting that the data are perhaps more representative of the density of roost locations¹¹⁴. As such, the results indicate that the density of flying fox roosts, rather than the population density of flying foxes, is associated with HeV spillover. This agrees with the findings of Giles et al.¹⁰¹, who found that the presence of more smaller camps was associated with spillover. Similarly, a study on the roost characteristics of P. medius in Bangladesh noted more roosts, but not more bats, in villages reporting NiV cases, and that increased numbers of smaller roosts and human disease were associated with forest fragmentation⁸². Recent work by Paez et al.³⁴ on HeV suggested that bat abundance may play a role in peak HeV prevalence detected at roost sites³⁴, so the association between bat density, roost density, HeV prevalence, and spillover remains unclear.

Prevalence in bat populations

Disease prevalence, or the proportion of infected individuals in a population, is influenced by resource abundance and quality through behavioral and population-level factors, like reservoir host density and distribution, and through within-host factors, such as nutritional effects on immunity. Habitat loss and resource provisioning change the types and availability of diet foods, with potential downstream effects on wildlife nutritional status and immunity^115,116. A number of studies have addressed correlations between nutrition and HNV infection dynamics. In Australia, chronic reduction in flying fox winter food availability¹¹⁷, as well as acute food shortages associated with El Niño/La Niña climate cycles^101,118, may drive seasonal and annual patterns of elevated HeV prevalence and seroprevalence in flying foxes^35,36. Decreased body condition following an acute food shortage event was associated with elevated HeV seroprevalence in P. scapulatus³⁶, whereas poor body condition in P. giganteus in India was associated with lower seroprevalence of NiV²⁶. However, evidence of interactions between age and body condition on HNV seroprevalence^36,119 hinders clear interpretation of these seemingly contradictory data. Moreover, seroprevalence may not be a good proxy for infection prevalence^17,120. Seroprevalence measures the proportion of animals that have been exposed to the virus, and seroconverted (produced antibodies to the pathogen), and thus reflects cumulative rather than current infection history. Early studies relied on seroprevalence, but later research transitioned to detection of viral RNA after sensitive PCR techniques were developed (e.g., for HeV see Ref.121).

There have been behavioral observations consistent with nutritional stress spatiotemporally concomitant with HeV spillovers (Eby unpublished data; Kessler unpublished data)³⁵. Although McMichael et al.¹²² found that flying foxes with poor body condition display biomarkers consistent with nutritional stress (low plasma total protein, albumin, and globulin levels)¹²², nutritional biomarkers (aside from low plasma triglycerides) were not associated with detection of HeV RNA in individual bat urine or blood¹²³. Beyond these nutritional markers, HeV positivity was associated with other health markers such as greater lymphocyte counts, plasma alkaline phosphatase levels, and urinary protein levels, alongside lower neutrophil counts¹²³. Two separate studies also report a statistically significant relationship between detection of HeV RNA in pooled under-roost urine and increased concentration of urinary cortisol, which can be elevated during acute or chronic stress and malnutrition^124,125. The current state of knowledge is inconclusive as few studies include extensive spatial and temporal sampling of HNVs with direct consideration of bat ecology. Therefore, several other hypotheses for the above associations have been proposed. Several studies have found elevated cortisol in flying foxes during winter months^124,126 and have suggested that the relationship between HeV and cortisol is driven by cold temperatures and thermoregulatory stress, although the authors acknowledge that the study was not designed to explicitly test cold temperature as a driver of excretion¹²⁴. Confounding factors, such as food restriction during winter, limit our ability to establish causal links from the study design. In other systems, the influence of nutrition and health on HNV prevalence is less well-characterized. Where seasonality in prevalence has been observed outside Australia, namely NiV in P. lylei in Thailand, patterns have not been studied in the context of bat diet or health³¹. While few studies have explicitly evaluated the links between shifting resource availability and quality and HNV infection dynamics in pteropodids, when considered together, the studies above suggest that health-related effects play a role in HNV dynamics, and provide a valuable starting point for future work on the effects of resources, nutrition, and immunity on HNV prevalence.

Intensity of infection in bat hosts

Bat immune responses to viral infection play a critical role in determining HNV prevalence in populations, infection intensity within hosts, and shedding (i.e., excretion) of virus. The intensity of infection, which we define as viral load (or the quantity of any pathogen) in an individual host, is a critical driver of spillover because it probably correlates with the amount of pathogen excreted and available for infecting susceptible hosts⁵⁵. Infection intensity is primarily controlled by innate and adaptive immunity, which utilize an array of cells and compounds to limit viral replication and infection of host cells and tissues. Immune responses are energetically costly and resource-intensive to develop and activate, and the required allocation of host resources makes immunity dependent on nutrition for immune system development, maintenance, and function (see Refs. 127–129). Any dietary shift that restricts intake of sufficient energy or nutrients needed to sustain immune function has the potential to disrupt control of viral replication, leading to higher infection intensity.

The immune systems of bats and other mammal species are fundamentally similar and likely regulated by nutrition in similar ways. However, several differences seem to make bats uniquely able to host an array of pathogens without suffering apparent pathology. In humans and other mammals, induction of interferon alpha (IFNα), an important protein of the innate immune system, is important for the initiation of antiviral immune responses. IFNα signaling activates numerous immune genes called IFN-stimulated genes (ISGs) that activate the immune system and restrict viral replication and infection. Unlike other species, pteropodids continuously express IFNα and subsequently also have persistent induction of ISGs even when no infection is present⁵³. Continuously heightened innate immune responses may serve to control viral replication in bats. There are also key differences in bat adaptive immune systems. The adaptive immune system uses specialized cells and antibody responses that target specific pathogens, and it allows a host to mount a more effective immune response upon subsequent exposures to the same pathogen. Compared to other mammals, bats have genes that produce a wider array of antibodies¹³⁰. This allows bats to more specifically and effectively target pathogens, but likely at the expense of immune processes that generate high antibody levels (titers)¹³¹. In addition, in wild-caught P. alecto, immune cells called T-cells that are classically involved in antiviral responses in other mammals seem to co-occur with other T-cell subsets that regulate immune response¹³². This suggests that that cell-mediated immunity in bats may be biased towards regulatory and anti-inflammatory processes. Combined with reduced viral burden from innate immune activity, the low titers and slow antibody responses of bats^133,134 and patterns of cellular immunity could prevent viral clearance, leading to the hypothesized patterns of persistent henipavirus infection and reactivation in bats¹³⁵.

These and other critical immune responses are dependent on complex interrelationships between levels of certain dietary components and immunity that may be affected by changing the nutritional landscape. The innate and adaptive immune defenses in bats and other species are affected differently by nutrition because each branch has different resource costs. In general, constituents of the innate immune system have low resource costs during development but can be costly to produce and use during infection. Expression of IFNα is tightly controlled in most species because over-activation of the immune system can cause damaging inflammatory responses¹³⁶. Continuously elevated IFNα and immune gene expression in bats likely imposes high maintenance costs that are sensitive to resource restriction and nutrition¹³⁷. Bats may reduce some of this cost by inducing ISGs that help prevent damage from acute inflammatory responses typically associated with innate immunity⁵³, but IFN production itself is sensitive to vitamin A deficiencies and nutritional stress^138,139. By contrast, it takes longer and is more energetically expensive to develop the specialized cells and antibody repertoires of the adaptive immune system, but it has lower maintenance and activation costs during infection¹³⁶. Like innate immunity, it is also affected by certain nutritional conditions. As another example, starvation lowers levels of leptin, a fat-derived hormone that regulates energy balance and the activation and proliferation of different T-cell types, resulting in an overall decrease in antibody secretion (T-cells help activate B-cells, the immune cells that produce antibodies) and most other adaptive immune processes¹⁴⁰. Numerous other examples demonstrate that deficiencies in certain vitamins, minerals, and amino acids, free fatty acids, glucose, and other nutritional compounds differentially act on specific parts of the immune system in ways that range from stimulating complex signaling pathways to highly localized immune responses (e.g., Refs. 127,128).

Undernutrition, including starvation and vitamin and mineral deficiencies, constrains early-life investment in development of adaptive immunity and suppresses innate and adaptive immune processes during infection. Adult bats experiencing marginal nutrition or resource restriction are expected to rely on immune defenses that are less costly to mount. Although this should manifest as a shift towards adaptive rather than innate immune response to infections, individuals that experienced marginal nutrition during development (or otherwise preferentially invested in innate immunity, as bats might) continue to show greater reliance on innate rather than adaptive immune responses as adults^136,137. This general shift in immune functioning and impaired maintenance of costly IFN-mediated innate immune response could increase susceptibility to and replication of viral infections¹⁴¹. The studies of resource allocation and bat immunity are rare, although Christe et al.¹⁴² demonstrated that pregnant greater mouse-eared bats (Myotis myotis) showed weaker cell-mediated responses during early gestation than non-pregnant bats in the same roost¹⁴². Reproduction is resource intensive, and this response is consistent with immunological responses to resource shifts observed in other mammals. Increased susceptibility to infection during mammalian pregnancy is associated with suppression of inflammatory responses, including T-cell mediated immune responses and interferon production¹⁴³. HeV has been isolated from placental fluid of an aborted Pteropus poliocephalus fetus²², and elevated seroprevalence has been observed during pregnancy and lactation^36,37. Elevated seroprevalence in nutritionally stressed and pregnant flying foxes could be explained by this mechanism, where seroconversion is the result of compromised immunological states that increase susceptibility or reactivation of latent virus³⁶. Resource-driven changes in immune function could be particularly important in bats, where cellular, innate antiviral mechanisms are believed to be key in controlling viral replication early in the immune response, and are hypothesized to allow individuals to resist the pathological effects of viral infection and tolerate latent infections¹⁴⁴. Thus, these changes in immune function could alter within-host processes necessary for controlling viral infection in individuals, and increase viral intensity, prevalence, and excretion, particularly in pregnant and nutritionally compromised animals.

Shedding from infected bats

Spillover depends on excretion of a sufficient infectious dose from reservoir hosts, and contact of reservoir hosts with excreted pathogens. The amount of virus or other pathogens shed from reservoir hosts depends on infection intensity and all the factors involved in immunological control that were discussed in the previous section. As outlined above, clearing infection is energetically costly and allocation trade-offs exist between different immunological processes. Trade-offs also exist between immunity and life-history traits^145–147. In particular, several proposed mechanisms link flight to bat immune defense. For example, the hypothesis that flight acts as a fever and leads to selection of avirulent pathogens with low levels of replication (Ref. 148, though see Ref. 131), or the hypothesis that adaptations that allow bats to avoid oxidative cell damage when they increase metabolic rates for flight also allow them to mitigate inflammatory damage and host viruses without clinical disease^148,149. Changed movement behavior in altered resource landscapes could affect such adaptations¹⁵⁰. Quantifying the costs of bat flight and long-distance movement on immune system function may prove to be important in the future as bats move either more or less in response to habitat loss⁷². In addition, abortion rates among pregnant flying foxes increase during food shortages¹⁵¹, suggesting that trade-offs also occur during reproduction. Considering trade-offs between investing in immunity or other traits provides a useful framework for understanding the eco-evolutionary interactions and population dynamics of hosts and pathogens^72,152 and may also prove useful in identifying environmental conditions associated with pathogen shedding from bats or transmission within bat populations.

Understanding the environmental drivers of HNV dynamics in bat populations is difficult without knowing the time-course of viral infection in individuals³⁹. Intense episodic shedding of HeV and NiV in bat urine seems to occur in discrete spatiotemporal pulses that are distinct from low levels of background HNV shedding^35,153. Recent work suggests that HNV shedding is correlated with environmental conditions that are presumably stressful³⁴. In Australia, the roosts closest to HeV spillover clusters in 2011 and 2013 were occupied by actively shedding bats experiencing conditions consistent with food shortage³⁵. In addition, intensity of HeV excretion from P. alecto is associated with low rainfall levels in preceding months, which is known to negatively affect the flowering and nectar intensity of primary eucalypt diet plants³⁴. When preferred foods are unavailable, P. alecto and other pteropodids feed on alternative food sources, including native and exotic fruits that may have low nutritional or energetic value, poor digestibility, and harmful secondary metabolites^{83,115,154,155}. However, any inference from these studies relies on urine collected under bat colonies, which only measures shedding of virus. Few studies have simultaneously tested both bat excrements and tissues^38,123, making it impossible to determine whether observed patterns are driven by acute or persistent infections. Resource-related pressure or stress could drive increased shedding in a variety of ways; impaired immune control could increase susceptibility to HNV infections, recrudescence of latent infections, or opportunities for transmission if bats aggregate in response to resource concentration^34,39,46. If HNV infection in bats is acute, observed episodic patterns could be driven by extinction-recolonization dynamics (HNVs quickly spread through the population and then go locally extinct, followed by re-introduction after the colony immunity wanes)⁴⁰. By contrast, if bats are persistently infected, loss of the immune control may drive reactivation and periodic shedding long after individuals are initially infected³⁵. This is important to consider in light of a potential environmental driver, as the environmental conditions that enable infection and reactivation may be very different. Finally, abortion of infected fetuses represents another potential mechanism of HNV excretion from bats²², and represents another likely spillover risk that is sensitive to resource trade-offs.

Virus survival outside bat hosts

While direct transmission of HNV from bats to domestic animals and humans may occur under rare circumstances¹⁵⁶, indirect transmission is likely through contamination of foods or fomites or virus in aerosolized urine and feces^{35,96,156,157}. Therefore, environmental survival emerges as an important feature that determines zoonotic potential of pathogens. Although HNVs can survive outside their hosts in urine and on fruit, they are sensitive to increasing temperatures and desiccation and have an average half-life of only a few hours¹⁵⁸. Consequently, spillover requires that contact between recipient hosts and HNVs must occur shortly after excretion¹⁵⁶. Models show that HeV survival increases with latitude but decreases with temperature¹⁵⁹. Not surprisingly then, HeV was found to survive longer in winter than in summer, although virus survival at any location varied annually¹⁵⁹.

Whether changing resource landscapes decrease or increase virus survival will depend on the microclimatic conditions where transmission takes place. For example, landscape changes that increase temperature, such as deforestation¹⁶⁰, may decrease virus survival by increasing desiccation, while increases in precipitation or humidity may increase survival by decreasing desiccation. Modeling studies of HeV survival in the environment suggest that dense overhead tree canopy structure and long grass increase HeV survival¹⁵⁷, so local climate and management are critical factors that influence the likelihood of horse exposure to live virus. By contrast, it is less clear how environmental conditions affect NiV survival. Ongoing spillover and human-to-human transmission of NiV in Bangladesh is primarily linked to consumption of date palm sap after bats contaminate collection buckets⁹⁶. Sap harvesting is concentrated during seasonal periods, so virus survival in collection buckets is likely to be less variable, and the liquid media reduces the effects of temperature and desiccation. Thus, human behavior, not environmental conditions may be the key barrier to spillover events at the human–virus interface¹⁶¹.

Exposure of spillover hosts in a changing landscape

The cumulative effects of habitat loss and resource provisioning (e.g., changes to the abundance, quality, and composition of resources) on each of the processes described in the previous sections change the probability that susceptible recipient hosts will become exposed to an infectious dose of excreted pathogens (Fig. 2)⁵⁵. Within-host factors of recipient hosts (beyond the scope of this review) also determine the course of spillover infections. However, henipaviruses target cell receptors that are highly conserved across humans and animals, explaining why HNVs are able to target a broad range of recipient host species^29,162. Numerous domestic and peridomestic animals, including pigs, horses, cats, rats, dogs, and laboratory animals are susceptible to HNVs (see Ref. 163), as are humans and many pteropodid and insectivorous bat species^{17,30,164–166}.

Unlike some of the indirect effects of changing resource landscapes on nutrition and animal movements, the act of recipient host exposure to HNVs is almost exclusively tied to bat resources. In Bangladesh, date palm sap is only available as a food for pteropids because of harvest by humans^161,167. Infection in Bangladesh has also been documented following consumption of bat-dropped fruits from native and cultivated trees, such as mangos, star fruit, olives, and banana¹⁶⁷. Empirical evidence points to land-use change as a primary driver of contact between P. medius and recipient hosts of NiV. Case-control studies of villages found anthropogenic changes to landscape composition to be a risk factor for NiV spillover events. Hahn et al.⁴⁶ found that villages with NiV spillover events had nearby roosts associated with Indian mast and silk cotton trees. These trees are planted decoratively, but they also provide nectar during winter, a season of resource limitation for bats in Bangladesh, and attract large numbers of P. medius⁴⁶. Date palm sap, another provisioned resource for flying foxes, is also harvested in winter, thus potentially increasing bat-human contact^46,96. The seasonal importance of cultivated trees as bat food sources may be further exacerbated by forest fragmentation, which is another key predictor of NiV spillover events in Bangladeshi villages^46,82.

Like spillover in Bangladesh, shifting resource landscapes were implicated in the early Malaysian outbreak of NiV. A retrospective epidemiological survey documented overlap between flying fox food trees (mango orchards) and pig enclosures at the NiV index farm, suggesting that spillover occurred after pigs consumed contaminated fruit dropped by feeding bats⁹³. Tripling of mango and swine production in Malaysia from the 1970s and 1990s and mango–swine polycropping on Malaysian farms suggests that agricultural intensification changed the availability of bat foods, thereby attracting flying foxes into farms and promoting spillover⁴¹. Forest degradation in Malaysia¹ may have increased the importance of these cultivated orchards as food sources for bat populations.

Similarly, native habitat loss and resource provisioning play a role in exposure of spillover hosts to HeV. As a result of urban and rural development, critical flying fox habitat has been removed or replaced with alternative foods^68,168, resulting in increased overlap between human, horse, and flying fox populations. For instance, 64% of HeV spillover events between 1994 and 2010 (9/14 spillovers) occurred within the foraging radius of continuously occupied urban camps⁴⁰, and the trend has been more pronounced in recent spillover events (Eby et al. unpublished data). Flying foxes have also increased their use of degraded remnant forest patches and planted trees in peri-urban pastureland^{69–71,107,169}. Whether this is the result of resource restriction^34,35,40,71 or dietary preference for cultivated and weedy species⁶⁹ remains unresolved, but the frequent occurrence of diet plants in horse paddocks presents a spillover risk.

Unfortunately, much less is known about how changing resources affect contact between bats and spillover hosts for African HNVs or what risk these viruses pose for humans or domestic animals. Interestingly, the distribution of E. helvum in urban areas depends on hunting pressure on the local bat population. Populations in regions without hunting show a greater tendency to roost in urban areas, while hunted populations generally roost in more remote areas⁸⁰. Thus, in addition to land-use change, other factors that increase bat-human interaction increase spillover risk. A cross-sectional study associated the land-use change with HNV-neutralizing antibodies amongst humans in Cameroon¹⁷⁰, and a separate analysis in the same study identified an association with butchering bats¹⁷⁰. Because deforestation is likely correlated both with HNV seroprevalence and hunting bats, carefully designed field studies and statistical adjustments are needed to elucidate how deforestation affects HNV spillover in Africa and more broadly. However, this is extremely challenging due to the panmictic continental distribution of E. helvum fruit bats¹⁷¹. HNV antibodies detected in domestic animals indicate prior exposure to HNV, although details surrounding spillover events are unavailable^163,172. Nonetheless, these findings add to a consensus that the human-driven resource shifts are altering the bat-human interface and risk of HNV spillover.

Unanswered questions

Although our review presents parsimonious relationships between flying fox resources and cross-species transmission, we identify many remaining questions about how shifting resource landscapes drive particular components of the spillover process³⁵, and avenues for future research (Box 2). These break down into two major knowledge gaps that hinder causal inference about changing resource landscapes and HNV cross-species transmission. First, our limited understanding of within-host bat–HNV interactions and population-level persistence mechanisms constrains our ability to consider the consequences of stress and undernutrition. Second, and perhaps most critically, we emphasize that it is particularly important to disentangle the effects of native habitat loss from resource provisioning in urban and agricultural landscapes.

Within-host infection dynamics

Contact between wildlife and recipient hosts is required for spillover, but the underlying disease dynamics in reservoir hosts play a role in whether a pathogen will be transmitted to recipient hosts. There is a recognized need for data that will distinguish between different viral infection scenarios in bats (e.g., acute immunizing infection with or without waning immunity, versus persistent chronic or latent infections³⁹), which we strongly reiterate. Although mounting evidence from wild-caught bats points to maintenance and recrudescence of latent viral infection as a plausible contributor to HNV persistence in flying fox populations^32,173, acceptance of the wrong hypothesis may negatively affect our inference and ability to successfully mitigate spillover risks. Viral dynamics driven by transmission of acute infections or reactivation of latent infections may be influenced differently by the availability and quality of resources, as well as by the nutritional and physiological stressors associated with changing resource landscapes. This can scale to effects on HNV transmission between bats, between bat populations, and to other species. For example, poor nutritional condition is perhaps more likely to drive reactivation of latent infections, either by reduction in immunocompetence or immunity trade-offs during pregnancy, both of which are suspected drivers of viral recrudescence in bats^40,135. By contrast, transmission between hosts will likely be more affected by the distribution of resources and its effects on bat population density and connectivity³⁹.

Key advances in bat immunology have been derived from the studies of flying foxes^53,174, but few of those studies have specifically addressed the immunological consequences of stress and nutrition. Most work on these topics has been conducted in other bat families. This hinders our understanding of the general ways that the poor resource landscapes might affect the immune status of pteropodids, as well as more complex allocation trade-offs between immunity and other life-history traits that are thought to occur during resource restriction. For example, the general links between migration and disease^98,100, urbanization and disease^50,175, and anthropogenic resources and disease⁴⁹ have been well reviewed in the literature elsewhere. Across these contexts, pathogen transmission and spillover risk seems to be particularly heightened if migratory capacity is reduced. Migration is an energetically costly activity, with many physiological changes occurring at different stages during long-distance animal movements⁹⁸. The metabolic pathways linking nutrition to immune system functioning remain poorly understood for flying foxes and many other migratory species. Empirical field studies comparing HNV prevalence, nutrition, and immune function between migratory flying fox populations and urbanized colonies are also lacking. These data are key to understanding how future reductions in migratory behavior in response to the resource landscape may affect innate and adaptive immune responses in flying foxes.

Habitat loss and resource provisioning

The confirmed or putative primary reservoir hosts involved in ongoing spillover of NiV (P. medius), HeV (P. alecto), and African henipaviruses (E. helvum) all show a tendency to habituate to urban and agricultural landscapes for roosting and foraging. Evidence gives rise to two hypotheses that are widely discussed in the bat community but remain largely untested in these systems. First, that resource provisioning attracts bats to peri-urban areas, or second, that land-use change is pushing bats into peri-urban areas or changing the resource landscape in a way that results in the use of suboptimal food sources. Cultivated plants and landscaping provide bats with an abundance of novel foods that attract bats from native forests into towns⁶⁸. Given the choice between ephemeral native foods that require long-distance travel, bats choose to roost and forage on crops and other planted foods, and increased contact between bats and recipient hosts allows for more spillover opportunities. In other words, the spillover process is driven primarily by the availability of potential resources and attraction of bats to new food subsidies. The major Australian metropolitan cities of Melbourne, Sydney, and Brisbane all report a spatiotemporal distribution of diet plants that support local resident populations of flying foxes^83,84,169, and the foraging radius of flying foxes is large enough that cultivated fruit trees can also supplement their diet. Similarly, plants in urban habitats in Pakistan and Bangladesh seem to fulfill the diet requirements of P. medius^45,82.

However, these studies fail to consider the fitness of urban populations relative to their rural or migratory counterparts. In the Australian HeV system, flying foxes are experiencing population-level declines at the same time that the proportion of urban bats is increasing⁴⁰. The ecological trap literature and numerous other examples point towards scenarios where wildlife select or otherwise occupy inferior habitat, often in response to rapid or extensive habitat change that results in a mismatch between the available resources and environmental cues that indicate habitat quality^176,177. For example, if bats have evolved preference for heterogenous native forest because it provided good foraging opportunities, land-use change and unintentional resource provisioning with weeds and planted vegetation might result in use of forest patch edges or urban environments that subsequently reduce fitness¹⁷⁸. The assumption that high population densities exist in high quality habitat breaks for highly mobile species that rely on unpredictable resources. First, considering habitat use during particular seasons is critical, particularly if seasons are associated with food shortages or animal movements to different ranges. There also may be high multi-annual variation in population density that reflects short-term or current resource conditions rather than the underlying habitat quality¹⁰⁵. This leads to the second hypothesis, which argues that land-use change may increase stress and undernutrition (or other health factors) in bats, with downstream consequences on their ability to control viral infections.

Land-use changes that unintentionally provision wildlife with novel resources are usually concomitant with loss of native habitat. Much of the current research critically ignores the joint effect of natural resource restriction outside the provisioned environment, and the potential changes in the quality of available resources following land-use change (Fig. 2)⁸⁴. Flying foxes are highly mobile and the effects of habitat change on urban habituation must be evaluated at the landscape scale, rather than the immediate vicinity of new roosts^46,81. Furthermore, the temporal distribution of resources on the landscape must also be considered in both urban and natural environments. The fruiting and flowering phenology of tropical and subtropical plants is highly irregular with very few species providing reliable annual resources, and predicting which habitats are critical for maintaining nomadism is exceeding difficult. Extensive deforestation in Bangladesh similarly seems to have concentrated available seasonal resources, changing the roosting behavior of P. medius in Bangladeshi villages^46,82. In Australia, agricultural and urban development has preferentially cleared forest on rich soils in coastal lowlands, removing much of the only habitat that reliably provides floral resources for bats in winter when resources are generally scarce^68,117,118. Temporal bottlenecks in resource availability following habitat loss, particularly removal of intensely productive patches that drive long-distance movement^70,75, are overlooked as potential drivers behind the formation of continuously occupied colonies.

Whether habitat loss forces bats out of their native habitats or an abundance of planted resources attracts them to developed landscapes, the effects of composition and quality of available foods on health and foraging strategies of bats must also be considered¹⁷⁹. A preference of flying foxes for native diet resources over cultivated plants has been demonstrated, as well as a higher nutritional value in native fruits in some regions^{43,85,115,180}. Similarly, native floral resources for flying foxes may provide greater volumes and more concentrated nectar than introduced species^7,117, and native vegetation provides superior habitat for other pollinators^7,181. Research also demonstrates that food shortages force bats to utilize alternative and unusual food sources that may have low nutritional value^{83,115,154,155}. Use of exotic species outside their natural diet is often cited as evidence of diet preference or high nutritional value of those species (e.g., Refs. 69,84); however, this is impossible to determine without direct measurement of forage quality and consideration of possible alternative food choices, which is rarely performed. It is critically important to evaluate dietary preferences and available food resources, particularly in the context of spillover.