key: cord-0267616-gqok8neh
authors: Guth, Sarah; Mollentze, Nardus; Renault, Katia; Streicker, Daniel G.; Visher, Elisa; Boots, Mike; Brook, Cara E.
title: Bats host the most virulent—but not the most dangerous—zoonotic viruses
date: 2021-07-26
journal: bioRxiv
DOI: 10.1101/2021.07.25.453574
sha: 5f8adac051deeef657ff548101a34a1016bfcab0
doc_id: 267616
cord_uid: gqok8neh

Identifying virus characteristics associated with the largest public health impacts on human populations is critical to informing zoonotic risk assessments and surveillance strategies. Efforts to assess “zoonotic risk” often use trait-based analyses to identify which viral and reservoir host groups are most likely to source zoonoses but have not fully addressed how and why the impacts of zoonotic viruses vary in terms of disease severity (‘virulence’), capacity to spread within human populations (‘transmissibility’), or total human mortality (‘death burden’). We analyzed trends in human case fatality rates, transmission capacities, and total death burdens across a comprehensive dataset of mammalian and avian zoonotic viruses. Bats harbor the most virulent zoonotic viruses even when compared to birds, which alongside bats, have been hypothesized to be “special” zoonotic reservoirs due to molecular adaptations that support the physiology of flight. Reservoir host groups more closely related to humans—in particular, Primates—harbor less virulent, but more highly transmissible viruses. Importantly, disproportionately high human death burden, arguably the most important metric of zoonotic risk, is not associated with any animal reservoir, including bats. Our data demonstrate that mechanisms driving death burdens are diverse and often contradict trait-based predictions. Ultimately, total human mortality is dependent on context-specific epidemiological dynamics, which are shaped by a combination of viral traits and conditions in the animal host population and across and beyond the human-animal interface. Understanding the conditions that predict high zoonotic burden in humans will require longitudinal studies of epidemiological dynamics in wildlife and human populations. Significance statement The clear need to mitigate zoonotic risk has fueled increased viral discovery in specific reservoir host taxa. We show that a combination of viral and reservoir traits can predict zoonotic virus virulence and transmissibility in humans, supporting the hypothesis that bats harbor exceptionally virulent zoonoses. However, pandemic prevention requires thinking beyond zoonotic capacity, virulence, and transmissibility to consider collective ‘burden’ on human health. For this, viral discovery targeting specific reservoirs may be inefficient as death burden correlates with viral, not reservoir, traits, and depends on context-specific epidemiological dynamics across and beyond the human-animal interface. These findings suggest that longitudinal studies of viral dynamics in reservoir and spillover host populations may offer the most effective strategy for mitigating zoonotic risk.

populations is critical to informing zoonotic risk assessments and surveillance strategies. Efforts 39 to assess "zoonotic risk" often use trait-based analyses to identify which viral and reservoir host 40 groups are most likely to source zoonoses but have not fully addressed how and why the impacts 41 of zoonotic viruses vary in terms of disease severity ('virulence'), capacity to spread within 42 human populations ('transmissibility'), or total human mortality ('death burden'). We analyzed 43 trends in human case fatality rates, transmission capacities, and total death burdens across a 44 comprehensive dataset of mammalian and avian zoonotic viruses. Bats harbor the most virulent 45 zoonotic viruses even when compared to birds, which alongside bats, have been hypothesized to 46 be "special" zoonotic reservoirs due to molecular adaptations that support the physiology of 47 flight. Reservoir host groups more closely related to humans-in particular, Primates-harbor 48 less virulent, but more highly transmissible viruses. Importantly, disproportionately high human 49 death burden, arguably the most important metric of zoonotic risk, is not associated with any 50 animal reservoir, including bats. Our data demonstrate that mechanisms driving death burdens 51 are diverse and often contradict trait-based predictions. Ultimately, total human mortality is 52 dependent on context-specific epidemiological dynamics, which are shaped by a combination of 53 viral traits and conditions in the animal host population and across and beyond the human-animal 54

interface. Understanding the conditions that predict high zoonotic burden in humans will require 55 longitudinal studies of epidemiological dynamics in wildlife and human populations. 56

Significance statement: 57 The clear need to mitigate zoonotic risk has fueled increased viral discovery in specific reservoir 58 host taxa. We show that a combination of viral and reservoir traits can predict zoonotic virus 59 virulence and transmissibility in humans, supporting the hypothesis that bats harbor 60 exceptionally virulent zoonoses. However, pandemic prevention requires thinking beyond 61 zoonotic capacity, virulence, and transmissibility to consider collective 'burden' on human 62

health. For this, viral discovery targeting specific reservoirs may be inefficient as death burden 63

correlates with viral, not reservoir, traits, and depends on context-specific epidemiological 64 dynamics across and beyond the human-animal interface. These findings suggest that

The vast majority of human pathogens are derived from animal populations (1) . In response to 69 increasingly frequent zoonotic spillovers and their substantial public health risks (2), there has 70 been a movement to identify the ecological systems and taxonomic groups of animals and 71

pathogens that are most likely to source the next emerging zoonosis in the human population (3-72 9). However, most of this work has centered on a binary definition of zoonotic risk-whether 73 particular pathogens are capable of infecting humans-without considering how pathogens vary 74

with respect to their impacts on humans after spillover. The ongoing SARS-CoV-2 pandemic has 75 re-emphasized the reality that not all zoonoses pose risks of equal magnitude-some are 76 exceptionally more dangerous than others due to the severity of disease they cause ('virulence') 77

or their capacity to spread within human populations ('transmissibility'), which combined, 78

influence the total number of human deaths ('death burden') (10). Given the extraordinary 79 diversity of both animal hosts and the viruses they harbor, understanding which animal and virus 80 groups are more likely to source dangerous zoonoses is an important public health aim. Many 81 high-profile zoonotic viruses-including Nipah and Hendra henipaviruses; Ebola filovirus; 82 SARS, MERS, and SARS-CoV-2 coronaviruses; pandemic avian influenzas; West Nile virus; 83 and Eastern Equine encephalitis virus-have emerged from Chiropteran (bat) or avian reservoirs 84 (11) . The high number of zoonotic viruses found in bats and birds has been attributed to their 85 large gregarious populations, mobility, ability to colonize anthropogenic environments, and sheer 86 species diversity (7, 11). Nonetheless, the question remains: are bat-and/or bird-borne viruses 87 disproportionately dangerous? 88 A recent meta-analysis (10) found that mammalian reservoir hosts most closely related to 89 humans harbor zoonoses of lower impact in terms of mortality relative to more phylogenetically 90 distant hosts. These results were consistent with phylogenetic trends in virulence that have been 91

reported in cross-species pathogen emergences in other systems (12, 13), and likely reflect 92 mismatches in host biology, physiology, and ecology. Notably, order Chiroptera (bats)-one of 93 the more distantly related host orders-had the highest positive effect size on case fatality rate in 94

humans. We apply generalized additive models (GAMs) to a comprehensive dataset of 120 mammalian and avian zoonotic viruses to identify reservoir host and viral traits predictive of the 121 (a) case fatality rate (CFR), (b) capacity for forward transmission, and (c) death burden induced 122 by infections in the human population-with the goal of characterizing sources of zoonotic 123 viruses that pose the greatest danger to global health. Our work builds on a small body of meta-124 analyses that have begun to explore variation in the virulence and between-human 125 transmissibility of zoonotic viruses (4, 19-21). We provide the most thorough analysis of  126  quantitative zoonotic virus data published to date, including the first analysis of burden and the  127  largest sample size-with trends examined across the majority of known zoonotic viruses. We  128  hypothesized that birds-given their capacity for flight and phylogenetic distance from  129  humans-might rival bats for the association with the most virulent zoonotic viruses. However,  130 we did not expect bats or birds to be responsible for the greatest burden on global health, instead 131

anticipating high burden to be largely a function of viral traits and associated with reservoir 132 orders that harbor less virulent, more transmissible viruses. 133

Drawing from existing databases (3, 7), we compiled a dataset of all mammalian and avian 135 zoonotic virus species that met a strict definition of zoonotic-requiring a record of natural 136 human infection confirmed by PCR or sequencing and animal-to-human directionality in 137

transmission. Virus species linked to multiple independent reservoir groups (e.g., canine and bat 138 rabies) or those which spillover to humans both directly from their reservoir and through bridge 139

hosts (e.g., Nipah virus) were subdivided into separate entries for each unique transmission chain 140 ending in spillover, creating a final dataset of 87 viruses with a total of 91 transmission chains 141

(SI Data and Results, Table S1 ). We then applied generalized additive models (GAMs) to assess 142 predictors (SI Data and Results, with previous work (10) and the hypothesis that bats are "special" zoonotic reservoirs, order 155

Chiroptera had the largest positive effect size on CFR in humans ( Figure 1b) . The top selected 156 model predicted a CFR of 65.4% for zoonotic viruses derived from order Chiroptera, 157

representing a more than 50% increase from the next highest predicted CFR ( Figure S2 ). predicted CFR within the same models (Figure 1a ), indicating that both reservoir and virus taxa 173

contributed to the observed variation in virulence. Chiroptera had the highest positive effect size 174 on CFR despite being associated with virus families that ranged from the most (Rhabdoviridae) 175

to least (Coronaviridae) virulent ( Figure 1c ). Removing the 100% fatal lyssaviruses (n=5) from 176 the dataset resulted in large reductions in the CFR predicted for bat-borne zoonoses ( Figure S4 ), 177 though order Chiroptera still had the highest and most significant positive effect size on CFR 178

( Figure S2 in SI Figures, Table S6a in SI Data and Results). 179

Previous work has demonstrated a positive correlation between reservoir host 180 phylogenetic distance from humans and the case fatality rates of zoonoses derived from those 181 reservoirs (10); in our analysis, however, reservoir host group phylogenetic distance from 182

Primates was not correlated with CFR, dropping entirely from the top ranked model and not 183 ranking significantly in any of the top 15 selected models (Figure 1a ). The combined effect of 184 reservoir host group and virus family as predictor variables in the same model likely 185

overwhelmed any correlation between host phylogeny and CFR, particularly given the lack of 186 granularity in our phylogenetic distance variable, based on a time-scaled phylogeny, which 187 produced only six unique distance values across nine host groups, with Chiroptera and four of 188 the other mammalian orders clustering at a single distance level (Materials and Methods). 189

Nevertheless, trends in effect size on CFR ( Figure 1b ) and predicted CFR ( Figure S2 ) across 190 reservoir host groups suggest that, in general, virulence increases with phylogenetic distance, but 191 this positive correlation may collapse at "extreme" distances. 192

To test whether these results held across a larger sample size, we ran a CFR analysis that To assess whether CFR trends might be influenced by health care differences among the 203 virus' differing geographic ranges, we tested whether Gross Domestic Product per capita 204 (GDPPC) significantly predicted country-specific CFR estimates-calculated from death and 205 case counts in countries that have reported the largest outbreaks of each given virus species, with 206 up to three country estimates for each species for a total of 119 estimates across the 86 unique 207 zoonotic transmission chains. First, we modeled all 119 country-specific CFR estimates 208 separately to test whether GDPPC predicts country-level variation in CFR ( Figure S6 in SI 209

Figures, Table S6c in SI Data and Results). Although significant, GDPPC explained a low 210 percentage of the deviance ( Figure S6a Table S7d in SI Data and Results). 215 GDPPC was not significant in any of the top models, often dropping entirely during model 216 selection ( Figure S7a in SI Figures), suggesting that health care differences among the virus' 217 geographic ranges most likely do not bias Figure 1 trends. Nevertheless, as with the 218 supplementary analysis presented in Figure S5 , both analyses of the country CFR estimates 219 echoed all key results presented in Figure 1 . 220

Predictors of transmissibility within human populations. We found that most zoonotic 221 viruses (72.1%) have not been reported to transmit within the human population following 222 spillover (i.e., transmissibility rank = 1, or R0 = 0) ( Figure S8 ). Only 15.1% of virus species had 223 demonstrated capacity for endemic transmission among humans, of which the majority (61.5%) 224

were sourced from Primates. The top selected GAM to predict the ordinal rank of 225 transmissibility within human populations-across the 86 unique zoonotic transmission chains 226

for which at least two human cases have been recorded-explained 56.7% of the deviance and 227

included virus family, the phylogenetic distance between each virus' reservoir host group and 228

Primates, vector-borne transmission, and the virus species publication count ( Figure 2 , Table S5b  229 in SI Data and Results). Transmissibility declined with phylogenetic distance from Primates, but 230 the estimated trend was highly uncertain (Figure 2c ). We therefore reran the analysis with 231 reservoir host group as the only host taxonomic predictor (excluding the phylogenetic distance 232 variable). This analysis identified Primates as the only host order significantly associated with 233 heightened transmissibility in humans, suggesting that this group is the primary driver of the 234 phylogenetic trend observed in the top selected model ( Figure S9a in SI Figures, Table S6c in SI 235

Data and Results). 236

Evolution of virulence theory typically assumes a tradeoff between virulence (death rate 237 due to infection) and transmission rate on the basis that while high within-host growth rates 238 increase infectiousness, they also increase damage to the host, increasing virulence and thus 239

shortening the infectious period and reducing opportunities for future transmission (23, 24). 240

Critically, CFR is not equivalent to virulence, but instead, a proxy that can be reliably quantified. 241

As defined by Day 2002 (25), CFR is a function of both pathogen virulence ( ) and clearance 242 rate ( ), in which = /( + ). Thus, virulent pathogens (high ) with high clearance 243 rates (high )-e.g., acute, short-lived infections such as Chikungunya virus (26)-could 244 produce low CFRs. In contrast, less virulent pathogens (low ) with low clearance rates (low 245

)-e.g., persistent infections such as HIV (27)-could produce high CFRs. Nevertheless, in our 246 data, we observed a relationship between CFR and transmissibility in humans that roughly 247 supports the fundamental theoretical tradeoff between virulence and transmission rate ( reliable death records were only available for a subset of the timeline, we standardized analyses 258 by including an offset for the number of years over which the death counts were recorded. The 259 raw death count distribution was highly left-skewed, with 39.5% of virus species linked to 0 260 deaths and more than half (62.7%) linked to fewer than 50 deaths ( Figure S11 Primates and species richness-as significant predictors. Reservoir groups most closely related 278

to Primates were associated with heightened death burdens relative to more distantly related 279 reservoirs, consistent with results from our transmissibility analyses that indicated that reservoirs 280 most closely related to Primates harbored more transmissible viruses (Figure 3c ). Reservoir 281 species richness positively correlated with death burden, as we would expect given that species 282 richness has been found to correlate with the number of viruses associated with a given reservoir 283 order (Figure 3d) (7). However, both reservoir predictors explained a small fraction of the 284 variation in death burden relative to virus family, confirming that death burden is largely a 285

function of viral traits (Figure 3a ). 286

While some reservoir groups-bats, primates, rodents, and birds-have sourced more 287 high burden viruses than others (Figure 4a ), both our model results and raw data suggested that 288 high burden viruses appeared to be function of viral traits, not the reservoirs themselves. No 289 single reservoir stood out as a consistent source of high burden viruses, with every reservoir that 290 harbors high burden viruses also harboring substantially more viruses that cluster at the lowest 291 death burdens (Figure 4a ). This was not the case for virus family (Figure 4b ) or primary 292 transmission route ( Figure 4c ); Coronaviridae and Orthomyxoviridae and a respiratory 293 transmission route were associated only with high burden zoonotic viruses. In general, the 294 viruses linked to the lowest death burdens were associated with the lowest transmission capacity. 295

As a deviation from this trend, Primates-which our models indicate harbor the most 296

transmissible, but generally less virulent zoonotic viruses-harbored several highly transmissible 297 viruses with low death burdens (Figure 4a ). 298

The highest death burdens were overall associated with zoonotic viruses that are less 299 virulent but highly transmissible in human populations (Figure 4d ). Respiratory pathogens with 300 capacity for human-to-human transmission have often incurred massive burdens over short 301

timeframes as a result of rare, but catastrophic spillover events that spark widespread 302 transmission in humans. Critically, while our dataset included only six viruses with respiratory 303 droplets as a primary transmission route-SARS-CoV-1, SARS-CoV-2, MERS CoV, Influenza 304

A, Nipah, and Monkeypox-these viruses accounted for more than 85.9% of the deaths recorded 305

for the 86 viruses in our death burden analysis, highlighting respiratory transmission as a high-306 risk zoonotic trait. However, these data were derived from a notably small sample size, as three 307 of the six respiratory viruses have caused only a single major epidemic. There was also 308 substantial variation among these respiratory viruses, with the death burdens associated with 309 SARS-CoV-1 and SARS-CoV-2 differing by more than 2.5 million. 310

Additionally, several outliers demonstrated that capacity for forward transmission in 311 human populations does not always predict death burden; it is critical to also consider 312 epidemiological dynamics across and beyond the human-animal interface. Less transmissible 313 viruses can accumulate large death burdens over many small, but frequent spillovers, particularly 314

in systems in which humans regularly interact with animal reservoirs. Rabies, Hantaan (HTNV), 315

and Japanese Encephalitis viruses have been associated with some of the highest death burdens 316

induced by viral zoonosis despite lacking forward transmission in human populations ( Figure  317 4d). This is likely because these viruses spill over to humans from animal host populations that 318 live amongst human communities-Rabies burden is largely driven by spillover from endemic 319 circulation in domestic dogs (28) an Ebola spillover event to spark a transnational epidemic that in just 2 years, caused more than 332 6.5 times the total number of deaths recorded from 1976-2013 (31, 33). These outliers suggest 333 that understanding epidemiological dynamics-within wildlife populations and across and 334

beyond the human-animal interface-in specific systems is a critical component of predicting 335

death burden and consequently, danger to human health. 336

A key insight from our work is that bats harbor the most virulent zoonotic viruses relative to 338 other mammalian and avian reservoirs ( Figure S1 in SI Figures) profile zoonotic viruses linked to significantly higher death burdens than we would expect based 354 on their capacity for forward transmission in the human population (Figure 4d ), suggesting that 355 death burden is highly dependent on both the contact rate at the human-animal interface and 356 epidemiological dynamics within the human population-factors which are not fully captured by 357 the broad explanatory variables considered in trait-based analyses. 358

The surprisingly low virulence of avian zoonotic viruses in contrast to bat-borne viruses 359 may reflect the extreme phylogenetic distance that separates birds from Primates. In our previous 360 analysis, we found that mammalian reservoir hosts most closely related to humans harbor less 361 virulent zoonotic viruses relative to more distantly related mammalian hosts such as bats (10) balance between virulence and transmission depends on how the reservoir host population 420 responds to the virus (the 'host selective pressure'), which is determined by the ecological, 421 physiological, and biological traits of the reservoir. While we identified "special" reservoirs of 422 virulent and transmissible zoonotic viruses, we found that the human death burden incurred by 423 viral zoonoses does not correlate with any one reservoir host order, including bats, and instead, is 424 a function of viral traits. Our data demonstrate that mechanisms driving high death burdens are 425 diverse and often contradict trait-based predictions. High death burdens have resulted from rare 426 spillover events of highly transmissible viruses that spread widely in the human population; 427

small, but frequent spillovers of the least transmissible viruses; and historically low-burden 428 pathogens that take off given the right ecological and evolutionary conditions. This suggests that 429 ultimately, death burden depends on epidemiological circumstances, which should be shaped, not 430 by reservoir host traits, but by a combination of viral traits and conditions in the animal host 431 population and across and beyond the human-animal interface. Notably, the pandemic spread of 432 SARS-CoV-2 can be attributed to its highly effective respiratory transmission between humans, 433 a trait linked to its identity within Coronaviridae, rather than its bat origins (indeed, CoVs 434 demonstrate gastrointestinal tropism in bat reservoirs) (51). 435

Over the course of the last decade, a significant amount of funding and research effort has 436 been dedicated to identifying correlates of zoonotic risk, often with a long-term aspiration of 437 identifying ways to anticipate and prevent emerging zoonoses in the future (52-54). This 438 research increasingly prioritizes viral discovery over longitudinal studies of epidemiological 439 dynamics and targets animal populations such as bats that have been identified as key zoonotic 440 reservoirs. While our analysis corroborates the hypothesis that bats are a 'special' reservoir for 441 virulent zoonotic viruses, we also demonstrate that viral traits-not bat reservoirs-pose the 442 greatest danger to human health. We argue that burden, which does not correlate with any animal 443 reservoir and instead appears to be a function of transmission conditions to and within the human 444 population, more correctly approximates "danger" to human health than does virus virulence. 445

While reservoir and viral traits can predict zoonotic capacity, virulence, and transmissibility, 446 death burden is dependent on system-specific epidemiological dynamics, which are shaped by a 447 combination of viral traits and conditions in the animal host population and across and beyond 448 the human-animal interface. Thus, understanding and controlling the mechanisms that drive high 449 death burdens in humans-high rates of human-animal contact and/or epidemiological dynamics 450

in the human population that allow discrete spillover events to trigger human epidemics-451 requires longitudinal surveillance of specific zoonotic or potentially zoonotic viruses in both 452 animal and human populations. reports would allow us to assess and account for potentially confounding effects of regional 485 differences in health care and overall infrastructure. For our death burden response variable, we 486 collected the total number of deaths recorded across the world since 1950. In many cases, our 487 death count began after 1950, either because a zoonosis first emerged in humans after 1950 or 488 reliable death records were only available for a subset of the timeline. To standardize, we added 489 a variable for the number of years over which death counts were recorded to use as an offset in 490 our models. Death and case counts were derived, when available, from the Global Infectious 491

Diseases and Epidemiology Network (GIDEON) (57)-which contains outbreak data from case 492 reports, government agencies, and published literature records-and supplemented with 493 literature searches. All variable descriptions are provided in Table S4 in SI Data and Results. 494

We additionally ranked each zoonosis' capacity for transmission within human 495

populations-a correlate of R0-on a four-point scale (10). We assigned a human 496 transmissibility level of "1" to viruses for which forward transmission in human populations 497 post-spillover had not been recorded; "2" to viruses for which forward transmission in humans 498 had been recorded but was described as atypical; "3" to viruses for which transmission within 499 human populations had occurred regularly but was restricted to self-limiting outbreaks; and "4" 500

to viruses for which endemic human transmission had been reported. 501

Recording death and case data from laboratory-confirmed outbreaks in the literature 502 required maintaining a strict definition of zoonotic, excluding some viruses that have been 503 included in previous meta-analyses (3, 7, 19) . We compiled excluded viruses that met looser 504 inclusion criteria-specifically, seven viruses that have only caused human infections in 505 laboratory settings and 25 viruses that lacked molecular confirmation of infection of humans, but 506 still had serological evidence of infection in humans-in a supplementary database (Table S2 in  507 SI Data and Results). Viruses included in previous meta-analyses that met neither our loose nor 508 strict inclusion criteria are outlined in Table S3 in SI Data and Results. 509

Drawing from previously published databases (3, 7, 10), we collected seven variables (SI 510

Data and Results, Table S7 ) that we hypothesized might predict observed variation in human 511 CFR, capacity for transmission within human populations, and death burden. Given published 512 correlations between phylogenetic distance and virulence in cross-species spillovers (10, 12, 13, 513 58, 59), we included the reservoir host group cophenetic distance from Primates. We calculated 514 this distance variable using a composite time-scaled phylogeny of the mean divergence dates for 515 all reservoir clades, as presented in the TimeTree database (7, 60). In our prior analysis (10), 516

phylogenetic distance values were derived from a phylogenetic tree of mammalian cytochrome b 517 sequences (3, 61, 62), which captured significantly more variation between host orders. The 518

time-scaled phylogeny used in this analysis produced only six unique distance values across all 519 reservoir groups in our database but represented the only available phylogeny that included both 520 mammals and birds. We considered both reservoir host and virus taxonomy, recording host order 521 and virus family. However, only ten avian zoonoses were distributed across several avian 522 reservoir host orders. To test our hypotheses regarding avian zoonoses, we addressed this small 523 sample size by aggregating avian reservoir orders into a single "Aves" group, while maintaining 524 separate host orders for the mammalian reservoirs. Given that the number of zoonoses harbored 525 by a reservoir group appears to correlate with species diversity within that group (7), we 526 hypothesized that species diversity might influence reservoir effect size on CFR in humans; thus, 527

we included reservoir species richness, which we derived from the Catalogue of Life using 528 version 0.9.6 of the taxize library in R (7, 63), taking the sum of values across bird orders for the 529 Aves reservoir group. If increasing a reservoir group's total number of zoonotic viruses also 530 increases their number of virulent zoonoses, reservoir species richness might inflate the mean 531 CFR of zoonotic viruses harbored by species rich reservoir groups-or alternatively, given that 532 most zoonotic viruses have low CFRs in humans, species richness might instead reduce the mean 533 CFR associated with these reservoirs. Nevertheless, we expected that higher numbers of zoonotic 534 virus species would inflate the total death burdens associated with species rich reservoir groups. 535 We defined a "spillover type" variable to account for the zoonotic transmission chain of each 536 virus, distinguishing between zoonoses that jump into humans directly from the reservoir 537 population and those that spillover to humans from bridge hosts (10). While the majority of 538 zoonoses were linked to single zoonotic transmission chains, there were a few exceptions with 539 both "direct" and "bridged" spillover. For example, zoonotic Influenza A virus and Nipah virus 540 (64, 65) have spilled over into the human population directly from their avian and bat reservoirs, 541 respectively, as well as from domestic pig bridge host populations. In such cases, each spillover 542 type (i.e., transmission chain) was entered separately in the database. We included an additional 543 binary variable that identified whether viruses were vector-borne, as both theory (23) and 544

previous meta-analyses (19, 20) have suggested a relationship between vector-borne 545 transmission and virulence. Finally, as has been done in other similar meta-analyses, we included 546 virus species publication count to account for any potential publication bias (3, 10, 59). 547

To pair with our country-specific CFR data, we collected an eighth predictor variable-548 gross domestic product per capita (GDPPC)-as a proxy for geographical differences in the 549 quality of health care and epidemiological control measures. 550

We additionally collected, for each virus species, the transmission route that contributes 551 the majority of human infections, extending data published by Brierley et al. (19) . We then 552 assessed trends in death burden across transmission types, hypothesizing that density-dependent 553 transmission, as characteristic of transmission via respiratory droplets, would be associated with 554 the highest death burdens in human populations. 555 Statistical analysis. Given the non-normal distribution of our data, expected nonlinear 556 relationships, and nested data structures within our predictor variables (66), we applied 557 generalized additive models (GAMs) in the mgcv package in R (67) to assess predictors of CFR, 558 transmissibility, and death burden in human populations. Rather than manually specifying higher 559 order polynomial functions, GAMs permit the use of smooth functions to capture nonlinear 560 relationships between response and predictor variables (66, 67). We fit continuous variables (i.e., 561 reservoir group species richness and phylogenetic distance from Primates, and virus species 562 publication count) as smoothed effects, and all binary (i.e., vector-borne status and spillover 563 type) and categorical (i.e., reservoir order and virus family) variables as random effects. For 564 variable selection, we ran all possible model combinations, ranked by AIC, and selected the 565 models with the lowest AIC values. 566

We first asked, which reservoir host and virus types are associated with elevated CFRs in 567 human populations following spillover? We constructed GAMs in the beta regression family to 568 query the predictive capacity of our predictor variables (SI Data and Results, GDPPC by the proportion of cases in the CFR calculation that were recorded in each country and 578 summing the weighted GDPPCs. We then modeled the global CFR estimates, which were not 579 tied to any specific system. For all CFR analyses, we modeled unique zoonotic transmission 580 chains-which we defined as unique reservoir orders and spillover type combinations per virus. 581

As a result, zoonoses with a single reservoir host order and spillover type were modeled as a 582

single CFR entry, while those with multiple reservoir orders and/or spillover types (e.g., 583

Influenza A and Nipah viruses) were modeled as multiple CFR entries. We excluded five viruses 584

for which only one human case has been recorded ( and points represent partial residuals. An effect is shaded in gray if the 95% CI crosses zero 833 across the entire range of the predictor variable; in contrast, an effect is shaded in purple and 834 considered "significant" if the 95% CI does not cross zero. Full model results are outlined in 835 Table S5c in SI Data and Results. 836 193 193 193 193 193 193 193 193 193 193 

Host Range and Emerging and Reemerging 617 Pathogens

Human viruses: 619 discovery and emergence

Host and viral traits predict zoonotic spillover from mammals

Spillover and pandemic properties of zoonotic viruses with high 623 host plasticity

Global Patterns of Zoonotic Disease in Mammals

Taxonomic patterns in the zoonotic potential of mammalian 627 viruses

Viral zoonotic risk is homogenous among taxonomic orders 629 of mammalian and avian reservoir hosts

Global shifts in mammalian population trends reveal key predictors of 631 virus spillover risk

Fast-lived Hosts and Zoonotic Risk

Host phylogenetic distance drives trends in virus 635 virulence and transmissibility across the animal-human interface

Bats and birds as viral reservoirs: A physiological and ecological 638 perspective

Determines Viral Persistence and Replication in Novel Hosts

Disease mortality in domesticated animals is predicted by host 643 evolutionary relationships

Accelerated viral dynamics in bat cell lines, with implications for 645 zoonotic emergence

Bats and birds: Exceptional longevity despite high 647 metabolic rates

Interspecies transmission and 649 emergence of novel viruses: lessons from bats and birds

WHO Coronavirus Disease (COVID-19) Dashboard

Tissue Tropism and Transmission 654 Ecology Predict Virulence of Human RNA Viruses

Virological factors that 657 increase the transmissibility of emerging human viruses

Transmissibility of emerging viral 660 zoonoses

Predicting reservoir hosts and arthropod vectors 662 from evolutionary signatures in RNA virus genomes

Host-Parasite Relations, Vectors, and the Evolution of Disease Severity

Coevolution of hosts and parasites

On the evolution of virulence and the relationship between various measures of 668 mortality

Replication cycle of chikungunya: A 670 re-emerging arbovirus

Dynamics and Genetics of Viral Populations 672 and Infected Cells

Estimating the Global Burden of Endemic Canine Rabies

The ecological dynamics of hantavirus diseases: From 676 environmental variability to disease prevention largely based on data from China

Review of Climate, Landscape, and 679 Viral Genetics as Drivers of the Japanese Encephalitis Virus Ecology

Predicting virus emergence amid evolutionary noise

Human Adaptation of Ebola Virus during the West African 684 Outbreak

Ebola virus disease. 686 The Lancet

The origin of specificity by means of natural selection: evolved 688 resistance in host-pathogen interactions

690 Molecular mechanisms of pathogenicity: how do pathogenic microorganisms develop 691 cross-kingdom host jumps?

Bats as "special" reservoirs for emerging zoonotic pathogens

Tacaribe Virus Causes Fatal Infection of An Ostensible 695 Reservoir Host, the Jamaican Fruit Bat

Re-emergence of Lloviu virus in Miniopterus schreibersii bats

Zwiesel bat banyangvirus, a potentially zoonotic Huaiyangshan 699 banyangvirus (Formerly known as SFTS)-like banyangvirus in Northern bats from 700

Immune responses of wild birds to emerging infectious diseases

The physiology and energetics of bat flight

Comparative Analysis of Bat Genomes Provides Insight into the Evolution 705 of Flight and Immunity

Dampened NLRP3-mediated inflammation in bats and implications for a 707 special viral reservoir host

Enhanced Autophagy Contributes to Reduced Viral Infection in Black 709 Flying Fox Cells

Dampened STING-Dependent Interferon Activation in Bats

Complementary regulation of caspase-1 and IL-1β reveals additional 713 mechanisms of dampened inflammation in bats

Adaptation of the master antioxidant response 716 connects metabolism, lifespan and feather development pathways in birds

Life without TTP: apparent absence of an important anti-inflammatory 719 protein in birds

Storage Are Associated with Weaker Complement-Mediated Bactericidal Ability in Birds 722 but Not Bats

The phylogenomics of evolving virus virulence

Global patterns in coronavirus diversity

How accurately can we assess zoonotic risk?

Predicting zoonoses

Prediction and prevention of the next pandemic zoonosis

The origin and diversity of the HIV-1 pandemic

Zoonotic origins of human coronaviruses

Global Infectious Diseases and Epidemiology Network (GIDEON): 736 A World Wide Web-Based Program for Diagnosis and Informatics in Infectious Diseases

The Causes and Consequences of Changes in Virulence following 739 Pathogen Host Shifts

Virulence mismatches 741 in index hosts shape the outcomes of cross-species transmission

General quantitative genetic methods for comparative 744 biology: phylogenies, taxonomies and multi-trait models for continuous and categorical 745 characters

Parasite specialization from a phylogenetic perspective: a new index 747 of host specificity

APE: Analyses of Phylogenetics and Evolution in R 749 language

Core Team, R: A Language and Environment for Statistical Computing (R Foundation 751 for Statistical Computing

Nipah virus: Impact, 753 origins, and causes of emergence

Foodborne Transmission of Nipah Virus

Mixed Effects Models and 757 Extensions in Ecology with R

Generalized additive mixed models using "mgcv" and "lme4

Beta Regression for Modelling Rates and Proportions

A better lemon squeezer? Maximum-likelihood regression with 763 beta-distributed dependent variables

Positions and QQ Plots