key: cord-0307716-zg8zm5gy
authors: Belser, Jessica A; Lau, Eric HY; Barclay, Wendy; Barr, Ian G.; Chen, Hualan; Fouchier, Ron AM; Hatta, Masato; Herfst, Sander; Kawaoka, Yoshihiro; Lakdawala, Seema S; Lee, Leo Yi Yang; Neumann, Gabriele; Peiris, Malik; Perez, Daniel R; Russell, Charles; Subbarao, Kanta; Sutton, Troy C; Webby, Richard J; Yang, Huanliang; Yen, Hui-Ling
title: Robustness of the ferret model for influenza risk assessment studies: a cross-laboratory exercise
date: 2022-04-02
journal: bioRxiv
DOI: 10.1101/2022.04.02.486825
sha: e8fb9443728687140e6eb072aedaf44f882e1357
doc_id: 307716
cord_uid: zg8zm5gy

Ferrets represent the preferred animal model for assessing the transmission potential of newly emerged zoonotic influenza viruses. However, heterogeneity among established experimental protocols and facilities across different laboratories may lead to variable results, complicating interpretation of transmission experimental data. Between 2018-2020, a global exercise was conducted by 11 participating laboratories to assess the range of variation in ferret transmission experiments using two common stock H1N1 influenza viruses that possess different transmission characteristics in ferrets. Inoculation route, dose, and volume were standardized, and all participating laboratories followed the same experimental conditions for respiratory droplet transmission, including a strict 1:1 donor:contact ratio. Additional host and environmental parameters likely to affect influenza transmission kinetics were monitored throughout. Overall transmission outcomes for both viruses across 11 laboratories were concordant, suggesting the robustness of the ferret model for zoonotic influenza risk assessment. To attain high confidence in identifying zoonotic influenza viruses with moderate-to-high or low transmissibility, our analyses support that as few as three but as many as five laboratories, respectively, would need to independently perform viral transmission experiments with concordant results. This exercise facilitates the development of a more homogenous protocol for ferret transmission experiments that are employed for the purposes of risk assessment.

Pandemic influenza viruses may arise through interspecies transmission events of animal 20 influenza viruses. Assessing the human-to-human transmission potential of animal influenza 21 viruses that cause spillover infections in humans is essential for pandemic risk assessment. Ferrets 22 have been used as a surrogate model for investigating the transmission potential of influenza 23 viruses in humans, as they are naturally susceptible to infection with human and zoonotic influenza 24 viruses, exhibit clinical signs during infection which closely resemble those of humans, and 25 support influenza virus transmission via similar modes as humans. In particular, the respiratory 26 droplet transmissibility of a specific influenza strain among ferrets often correlates with its 27 transmission potential in humans (1). As such, ferrets are commonly used for assessing the 28 pandemic potential of newly emerged zoonotic influenza viruses, and data from these experiments 29 inform formal risk assessment rubrics established by the WHO and CDC (2, 3). 30

The transmission potential of influenza viruses is determined by multiple viral, host, and 31 environmental parameters. As the ferret model becomes commonly employed in laboratories 32 worldwide, there is an underappreciated heterogeneity among established experimental protocols 33 and facility setups across different laboratories, which may lead to variable results between 34 transmission experiments performed (4). Some of these variables, such as the dose, volume, and 35 route of inoculation and animal age, have been confirmed to affect the kinetics of virus infection, 36 replication, and transmission in the ferret model (5-7). However, the impact of other parameters, 37 such as virus propagation procedures, caging designs, airflow directionality and number of air 38 exchanges, and environmental conditions such as relative humidity, is largely unknown. 39

Consequently, interpretation of results from ferret transmission experiments can represent a 40 challenge when comparing data generated from multiple laboratories, even when the same virus 41 strain or subtype is being investigated (8). Considering the statistical limitations on small sample 42 sizes in ferret experiments, and high potential for strain-specific variability, investigators often 43 assess the pandemic potential of emerging virus subtypes as an aggregate of multiple viruses tested 44 (9-11). As many public health efforts require cross-laboratory risk assessment studies for newly 45 emerged zoonotic influenza viruses (12) and antiviral efficacy studies aiming to block influenza 46 transmission between ferrets (13), a greater understanding of variability in transmission results 47 obtained between independent groups is critical. 48

To assess the variability of ferret transmission results across laboratories under established 49 protocols, we performed a global exercise using two common stock influenza viruses that possess 50 different transmission characteristics in ferrets. Eleven independent laboratories inoculated ferrets 51 with these stock viruses under uniform conditions; parameters known to affect influenza 52 transmission kinetics were controlled in the experimental protocols while other potential 53 parameters were carefully monitored and recorded, both prior to and during the transmission 54 experiments. All aggregated data from these experiments were blinded and analyzed by an 55 independent statistician. To inform future risk assessment activities, the confidence of drawing 56 conclusions on virus transmissibility with concordant or discordant outcomes from multiple 57 laboratories was also investigated. By assessing the range of variation present among ferret 58 transmission experiments performed under established experimental protocols, this global exercise 59 provides helpful guidance for data interpretation when cross-laboratory results are to be compared. 60

The relatively concordant transmission results across 11 laboratories suggest that the ferret model 61 is highly robust for influenza pandemic risk assessment studies under the semi-standardized 62 conditions employed here. Furthermore, analyses investigating the role of host and environmental 63 parameters as they contribute to virus transmission kinetics and outcomes is valuable for both 64 (14) (15) (16) (17) . Transmissibility was evaluated with 4 donor:contact pairs at a 1:1 ratio in each laboratory. 73 Transmission to exposed respiratory droplet contact ferrets was defined by detection of infectious 74

virus or seroconversion to the homologous virus in post-exposure sera. Following establishment 75 of contact with donor ferrets 24 hours post-inoculation, detection of infectious virus and 76 seroconversion in contacts was observed in 10/11 and 11/11 laboratories, respectively, with the 77 reported transmission frequency ranging from 50-100% (Table 1) 1-2) were normalized to TCID50 units (Figure 1 ), employing strain-specific conversions prior to 97 analyses (Supplemental Table 1 ). From the inoculated donor ferrets, the peak viral titers detected 98 in the nasal washes or throat swabs were at 5.72 ± 0.95, mean ± SD log10 TCID50/mL after 99 normalization, with the peak titers detected from 95.5% (42/44) of donors on the first sampling 100 time point on 1 or 2 days post-inoculation (dpi), followed by a decline of infectious titer over time 101

( Figure 1A) . Area under the curve (AUC) after normalization was calculated to approximate total 102 viral load shed by the Cal/09-inoculated donors, with a mean ± SD log10 AUC of 5.84 ± 0.89.

donors (left bars) and aerosol contact ferrets (right bars) after inoculation or exposure to 105 A(H1N1)pdm09 virus Cal/09. B, normalized viral loads of donors (left bars) and aerosol contact 106 ferrets (right bars) after inoculation or exposure to avian H1N1 virus ruddy turnstone/09. Nasal 107 washes (all groups except Group F) or throat swabs (Group F) were sampled to determine 108 infectious viral loads which were normalized to log10 TCID50/mL. Each bar represents individual 109 ferrets. Limit of detection is indicated with a dashed line. 110

Next, to evaluate the transmission efficiency, the serial interval (first detection of viral 112

shedding in contacts post-exposure from specimens collected every-other-day) was calculated for 113 each infected contact ferret. The serial interval was 1 day for 3.1% (1/32) of the Cal/09 infected 114 contact ferrets, followed by 3 days for 68.8% (22/32), 5 days for 21.9% (7/32), and 11 days for 115 6.3% (2/32), with a median serial interval of 3 days post-contact. Peak viral titers detected in the 116 contact nasal washes or throat swabs were at 5.41 ± 1.06 mean ± SD log10 TCID50/mL after 117 normalization, with peak titers detected from 50% (16/32) and 34.4% (11/32) infected contacts on 118 3 dpi and 5 dpi, respectively. Altogether, the AUC for Cal/09 infected contact ferrets was 5. 

A/ruddy turnstone/Delaware/300/2009 (ruddy turnstone/09) (18, 19) , which has been reported to 131 transmit in ferrets via respiratory droplets under the experimental setting of donor: direct contact: 132 respiratory droplet contact at a 1:1:1 ratio, but not at a 1:1 donor:respiratory droplet contact ratio 133 (R Fouchier, unpublished data) (18, 19) . Here, the experimental setup and conditions were 134 identical to those assessing Cal/09 virus transmissibility including a donor: respiratory droplet 135 contact 1:1 ratio with no direct contact ferret. Transmission of an egg-derived isolate of ruddy 136 turnstone/09 virus to exposed respiratory droplet contacts was only observed in 4 out of the 11 137 laboratories, with the transmission frequencies ranging from 25-75% across these four laboratories 138 (Table 1) . Compared to Cal/09 virus, there were greater differences in the ruddy turnstone/09 virus 139 transmission outcomes across 11 laboratories, but the difference did not reach statistical 140 significance by Fisher's exact test of homogeneity (p=0.068). Viral shedding and seroconversion 141 to ruddy turnstone/09 virus were detected from 6/43 exposed contact ferrets across all laboratories, 142 resulting in a transmission efficiency of 14.0%, which was significantly lower compared to that of 143 Cal/09 virus (72.3%, paired t-test, p < 0.001). 144

From the inoculated donor ferrets, the peak viral titers detected in the nasal washes or throat 145 swabs were at 4.85 ± 0.94, mean ± SD log10 TCID50/mL after normalization, which was 146 significantly lower than those detected in the Cal/09 inoculated donors (Mann-Whitney test, p 147 <0.0001). Peak titers were detected from 88.6% (39/44) donors on the first sampling time point (1 148 or 2 dpi) followed by a decline of infectious titer over time ( Figure 1B) . Compared with Cal/09 149 virus inoculated donors, the mean ± SD log10 AUC of ruddy turnstone/09 virus-inoculated ferrets 150 was 5.06 ± 1.86, significantly lower than those inoculated with the Cal/09 virus (Mann-Whitney 151 test, p <0.0001) ( Figure 2 ). Overall, ruddy turnstone/09 virus-inoculated donor ferrets shed lower 152 titers of infectious virus than the Cal/09 virus-inoculated donors. 153

In contrast to the transmission efficiency of Cal/09 virus with a median serial interval of 3 154 days, for the ruddy turnstone/09 transmission experiments, the serial interval was 3 days, 5 days, 155 or 7 days for 33.3% (2/6), 33.3% (2/6), and 33.3% (2/6) of the infected contact ferrets, respectively, 156 with the median serial interval at 5 days. Peak viral loads (3.94 ± 0.94 mean ± SD log10 157 TCID50/mL) detected from the six infected contact ferrets were lower when compared to the Cal/09 158 infected contact ferrets (Mann-Whitney test, p=0.0022). Peak titers were detected from 16.7% 159

(1/6), 33.3% (2/6), and 50% (3/6) infected contacts on 3 dpi, 5 dpi, and 7 dpi, respectively. 160

Furthermore, ruddy turnstone/09 virus-infected contact ferrets shed significantly less infectious 161 virus (4.31 ± 0.98, mean ± SD log10 AUC) when compared to those animals directly inoculated 162

with Cal/09 virus (Mann-Whitney test, p=0.0033) (Figure 2 ). Taken together, there was a longer 163 serial interval and lower infectious virus shed by ruddy turnstone/09 virus-exposed contact ferrets 164 when compared to those exposed to Table 3 ). Groups employing 193 caging with airflow directionality from inoculated to contact cages more frequently reported 194 moderate to high transmissibility (≥50%) of both viruses compared with groups lacking this 195 airflow directionality (6/6 vs 3/5 groups for Cal/09 virus, 3/6 vs 1/5 groups for ruddy turnstone/09 196 virus), however these findings did not reach statistical significance (both p>0.3, Supplemental 197 Table 5 ). Other specific features of cage setups, including distance between inoculated and contact 198 cages and air changes per hour (ACH) were also not statistically linked to the ruddy turnstone/09 199 transmission outcomes (both p>0.4, Supplemental Table 5 ). Taken together, despite substantial 200 heterogeneity in numerous non-standardized parameters in experimental setups employed between 201 groups, no one feature was identified as modulating transmission outcomes to a significant degree. 202 203 Contributing factors to virus pathogenicity. All ferrets inoculated with either Cal/09 or ruddy 204 turnstone/09 were productively infected, however measurements of morbidity varied between 205 groups for both viruses. Among Cal/09 virus-inoculated ferrets, mean maximum weight loss and 206 peak rise in body temperature between groups ranged from <1.0-15.6% and 0.6-2.1°C, 207 respectively (Supplemental Table 6 weight loss reported between groups was generally similar (56% and 52% for Cal/09 and ruddy 213 turnstone/09 viruses, respectively). No commonality with increased morbidity and ferret vendor, 214 gender, or pre-inoculation body weight was identified. Furthermore, no association was found 215 between morbidity and viral load (peak titer or AUC) or other environmental parameters, with the 216 exception of room temperature (with higher mean room temperatures associated with greater mean 217 weight loss) (Supplemental Table 8 ). 218 219

the results from this exercise demonstrate a capacity for groups possessing differences in facilities 221 designs and experimental protocols to report varying levels of relative transmissibility and 222 pathogenicity following inoculation of ferrets with the same virus. To illustrate how confidence in 223 risk assessments of virus transmissibility can increase as results from multiple groups are 224 combined, we evaluated the hypothetical risk of a virus capable of moderate to high transmission 225 (defined as p ≥50% transmission events per total pairs of ferrets as defined in Table 2 ) or non-226 transmissible (defined as p≤25% transmission events). In these analyses, concordant results are 227 defined as multiple groups identifying a virus exhibiting the same transmission capacity, and 228 discordant results are defined as multiple groups identifying a virus with different transmission 229 capacities, as defined above. By assuming concordant results across laboratories which permits 230 pooling of all transmission outcomes, as few as three groups (12 pairs of ferrets) will yield a 231 probability of over 80% to conclude moderate to high transmissibility when transmission was 232 observed in at least half of all experiments, and a probability of over 85% to conclude low 233 transmissibility when at most one transmission event was observed over all experiments. 234 235 Despite generally consistent results between all groups in this exercise, discordant results 268 are possible (Table 1) , highlighting the need to better understand how to responsibly interpret and 269 account for these findings. As such, we also considered the scenario when discordant results 270 between laboratories are recorded. To demonstrate moderate to high transmissibility, we found 271 that 6 laboratories with 1 discordant result could still provide 80% confidence in the conclusion, 272 while any discordant result significantly reduced confidence for concluding low transmissibility 273 (Table 3) . In both scenarios, if the results from different laboratories were more heterogeneous, 274 the uncertainty around the conclusion from each lab increases and the overall confidence would 275 decrease. This exercise is an illustration of the possible scenarios and confidence in drawing 276 conclusions on transmissibility but would be affected by how moderate to high or low 277 transmissibility were defined. 278

Discussion 280

The importance of the ferret model for influenza virus risk assessment studies cannot be 281 understated (4, 22). Recent advances in molecular biology, aerobiology, genomics, and other areas 282 highlight the ways the ferret model in general, and studies evaluating virus transmissibility by the 283 airborne route specifically, continue to contribute towards our understanding of influenza viruses 284 and the threat they pose to human health (23-25). However, as this model becomes more 285 commonly employed in laboratories worldwide, there is a pressing need to capture the level of 286 variability and heterogeneity intrinsic to this research. Cross-laboratory exercises have been 287 employed in the past to evaluate the reproducibility of assays employed for influenza virus public 288 health efforts (26), but no such exercise has been performed to date evaluating influenza virus 289 transmissibility in the ferret. In this exercise, 11 laboratories across different continents 290 independently evaluated the transmission potential of Cal/09 and ruddy turnstone/09 viruses with 291 distinct transmission potential. With only a few experimental parameters (common virus stock, 292 standardized inoculation dose, route, volume, and the 1:1 donor:contact ratio) being controlled 293 across the participating laboratories, we observed homogenous transmission outcomes (that is, 294 outcomes did not differ statistically) across laboratories. Our results demonstrate the robustness of 295 the ferret model in influenza risk assessment studies. 296

Risk assessment rubrics have thoroughly evaluated a wide scope of influenza A viruses, 297 from viruses associated with poultry outbreaks in the absence of confirmed human infections, to 298 viruses such as A(H5N1) and A(H7N9) influenza viruses that have caused substantial human 299 disease and death (3, 27). As such, there is a need to evaluate heterogeneity of ferret transmission 300 models employing viruses possessing a similar scope of transmissibility phenotypes. While the 301 variability in transmission results for either the Cal/09 or ruddy turnstone/09 viruses tested in this 302 study were not statistically significant, the range of results obtained, especially with the ruddy 303 turnstone/09 virus, nonetheless illustrates a level of variability that can be present in transmission 304 readouts of viruses exhibiting both low to high transmission efficiency (Table 1) . This variability 305 was present despite a high degree of standardization of virus stock, inoculation procedures, and 306 uniformity of donor:contact ratio. 307

As shown in the Supplemental Methods and Supplemental Tables 1-6, this exercise 308 captured the extensive heterogeneity in laboratory protocols and setups present between different 309 groups. Documented variation was present in every parameter examined, inclusive of ferrets, cage 310 setups, titration methods, and environmental conditions, among other features. Caging and airflow 311 considerations were especially variable (Supplemental Table 2 ). It is impossible to standardize all 312 contributing variables to these experiments, as institutional, animal welfare, and governmental 313 guidelines and requirements vary worldwide, as do cost implications. That said, this exercise 314 supports the capacity to harmonize results generated between disparate groups when a small 315 number of procedural parameters are fixed. Interestingly, the four groups that detected infectious 316 virus in contact nasal wash specimens in ruddy turnstone/09 transmission experiments all found 317 4/4 virus transmission in the Cal/09 experiment; transmission percentages between the two viruses 318 were highly correlated between laboratories (Spearman correlation = 0.86, p < 0.001). 319

Furthermore, while directional airflow (OR=4) did not reach statistical significance, it is 320 nonetheless of note that 3/4 laboratories for which ruddy turnstone/09 virus transmission was 321 detected possessed directional airflow, versus 3/7 of the laboratories for which transmission with 322 this virus was not detected; directional airflow from inoculated to contact animals was a feature in 323 6/11 laboratories in this exercise (Supplemental Table 3 ). While our results did not conclusively 324 identify any one experimental parameter statistically associated with enhanced transmissibility 325 outcomes, it is possible that a confluence of parameters is nonetheless capable of creating a more 326 permissive environment for virus transmission to occur. 327

To improve interpretation of results from this standardization exercise, we concurrently 328 investigated the hypothetical confidence in concluding low transmissibility (≤25% or ≤1 ferret 329 infected out of 4 ferrets) or moderate to high transmissibility (≥50% or ≥2 ferrets infected out of 4 330 ferrets) from multiple contributing laboratories. These analyses assumed both a uniform prior 331 distribution for the transmission probability for a novel pathogen, and independent transmission 332 outcomes from the laboratories. We considered two scenarios: one scenario where strong 333 homogeneity across laboratories could be assumed so the samples were pooled from multiple 334 laboratories, and another scenario where each laboratory drew their own conclusion on 335 transmissibility such that an overall conclusion was drawn as a voting system. As influenza viruses 336 of notable public health importance are frequently assessed across multiple independent 337 laboratories, these analyses provide a framework to rigorously interpret independently generated 338 findings, especially when discordant results between laboratories are reported. This is most critical 339 in the event of a novel virus believed to possess moderate-to-high transmissibility; our analyses 340 support that 4 independent laboratories with concordant results supporting an enhanced 341 transmissibility phenotype yields a 95% probability of this finding, with additional independent 342 groups or a greater number of total ferret donor:contact pairs necessary when discordant results 343 are present. 344

Collectively, the findings of this exercise support the potential benefit of increased 345 uniformity, or standardization, of some parameters when conducting risk assessment-specific 346 activities on the same viruses. Specifically, the donor:contact ratio represents such a parameter. 347

For a virus with moderate to high transmissibility, such as Cal/09 virus, modulation of this ratio 348 (e.g., conducting experiments with a 2:1 donor:contact ratio, as is the case when transmission 349 evaluations in a direct contact setting and via respiratory droplets employ a common donor) would 350 not substantially alter conclusions drawn. However, for a virus with reduced transmissibility at a 351 1:1 ratio, such as the ruddy turnstone/09 virus evaluated here, it is likely that an increased 352 donor:contact ratio (eg., 2:1) may enhance transmissibility by increasing virus-laden aerosols 353 exhaled from infected ferrets. Previous studies on ruddy turnstone/09 virus demonstrated airborne 354 transmission potential when employing a donor: direct contact: aerosol contact at 1:1:1 ratio; 355 efficient transmission by direct contact will subsequently affect the quantity and kinetics of virus-356 laden aerosols that mediate transmission by air (18, 19) . There is a need to better understand how 357 modulation of this ratio contributes to assessments of virus transmissibility. However, this does 358 underscore the potential complications posed by harmonizing data generated for risk assessment 359 purposes for which the donor:contact ratio diverges. With increased heterogeneity in results 360 between labs, uncertainty around the conclusions increases, and there is a corresponding decrease 361 in confidence in the results (Table 3) , showing the utility in increasing homogeneity across findings 362 from different labs in order to reduce the total number of labs required to yield statistically 363 meaningful results in this sort of analysis. 364

The emergence of SARS-CoV-2 further corroborates the pandemic potential of viruses of 365 zoonotic origin. Early identification and risk assessments of novel viruses are essential for 366 preventing the next pandemic. Continued optimization and refinement of risk assessment protocols 367 will facilitate data interpretation in response to emerging pandemic threats. Collectively, a greater 368 appreciation of this heterogeneity, and understanding of the scope of variability present in risk 369 assessment settings, will permit more robust conclusions to be drawn from these efforts in the 370 future. 371 372

Viruses. The A(H1N1)pdm09 virus A/California/07/2009 (Cal/09) was propagated in MDCK 374 cells (passage C3) at the US CDC as described previously (28). The low pathogenic avian influenza 375 A(H1N1) virus A/ruddy turnstone/Delaware/300/2009 (ruddy turnstone/09) was propagated in 376 eggs (passage E3) by St. Jude Children's Research Hospital as described previously (19). Stocks 377 were fully sequenced and tested for exclusivity to rule out the presence of other influenza virus 378 subtypes prior to distribution. 379

Animal and experimental variability. Groups obtained ferrets from multiple vendors and 380 independent breeders from North America, Europe and Asia, and animals varied in their age, 381 gender, health status, and other parameters (Supplemental Table 1 ). There was substantial 382 differences between laboratories in the specific caging employed for transmission experiments, 383 distance between cages, airflow directionality between cages, and air changes per hour 384 (Supplemental Table 2 ). Anesthesia protocols, sample collection methods, and decontamination 385 procedures to prevent cross-contamination between contact and donor animals varied between 386 groups and are reported in Supplemental Methods. All experiments were performed under country-387 specific legal guidelines and approved institutional-specific animal protocols as specified in the 388 Supplemental Methods. 389

Standardized procedures. All laboratories received common stock viruses prepared by CDC and 390

St. Jude Children's Research Hospital with the shipping temperature recorded. Stock viruses were 391 diluted to 10 6 plaque forming units (PFU) in 500µl PBS based on predetermined viral titers, and 392 donor ferrets were inoculated intranasally under in-house protocols for anesthesia (Supplemental 393 Methods). On day 1 post-inoculation, one respiratory droplet contact ferret was introduced and 394 exposed to each donor by housing in an adjacent cage, employing a strict 1:1 donor:contact ratio, 395 with 4 transmission pairs tested for each virus. Ferret temperatures, weights, and nasal 396 washes/swabs were collected every 24-48 hours. Daily room temperature and relative humidity 397 readings were collected and are reported in Supplemental Table 3 employing pre-validated  398 thermohygrometers with comparable readings (Testo Inc., 608-H1). Sera was collected at the end 399 of each experiment for determination of seroconversion to homologous virus by hemagglutinin 400 inhibition assay using established in-house serology protocols. 401

Sample titration and normalization. Infectious virus titers were determined by plaque assay, 402 50% tissue culture infectious dose (TCID50) assay, or 50% egg infectious doses (EID50) assay at 403 each laboratory with varying limits of detection (Supplemental Table 4 ). To facilitate subsequent 404 statistical assessments across laboratories, reported titers from each laboratory were normalized to 405 TCID50/mL for each virus based on PFU, TCID50, and EID50 values pre-determined by a single 406 laboratory to minimize titration methodology-specific variation. 407

Mapping influenza transmission in the ferret model to transmission in humans

Pandemic preparedness and the Influenza Risk 435 Assessment Tool (IRAT)

Ferrets as Models for Influenza Virus 439

441 Severity of clinical disease and pathology in ferrets experimentally infected with 442 influenza viruses is influenced by inoculum volume

Influenza virus aerosol exposure and analytical system for ferrets

Pathogenesis of Influenza A/H5N1 virus infection in ferrets differs 448 between intranasal and intratracheal routes of inoculation

Complexities in Ferret Influenza 450

Considerations regarding 452 appropriate sample size for conducting ferret transmission experiments

Sample size considerations for one-to-one 455 animal transmission studies of the influenza A viruses

Identification, characterization, and natural selection 459 of mutations driving airborne transmission of A/H5N1 virus

Characterizing Emerging Canine H3 Influenza Viruses

Baloxavir treatment of ferrets 472 infected with influenza A(H1N1)pdm09 virus reduces onward transmission

Antigenically Diverse Swine Origin H1N1 Variant 477 Influenza Viruses Exhibit Differential Ferret Pathogenesis and Transmission Phenotypes

Nonreplicating influenza A virus vaccines confer broad protection against lethal 481 challenge

Eurasian-origin gene segments 484 contribute to the transmissibility, aerosol release, and morphology of the 2009 pandemic 485 H1N1 influenza virus

Pathogenesis and transmission of swine-origin 2009 A(H1N1) 489 influenza virus in ferrets

The potential of avian 491 H1N1 influenza A viruses to replicate and cause disease in mammalian models

Molecular basis of 495 mammalian transmissibility of avian H1N1 influenza viruses and their pandemic 496 potential

Mechanistic insights into the 498 effect of humidity on airborne influenza virus survival, transmission and incidence

Environmental Conditions Affect Exhalation of H3N2 Seasonal and Variant Influenza 502 Viruses and Respiratory Droplet Transmission in Ferrets

Moving Forward: Recent Developments for the 505

Visualizing real-time influenza virus infection, transmission and protection in 508 ferrets

Hemagglutinin-512 neuraminidase balance confers respiratory-droplet transmissibility of the pandemic H1N1 513 influenza virus in ferrets

Influenza A virus transmission bottlenecks are defined by infection 516 route and recipient host

Reproducibility of serologic assays for influenza virus A 519 (H5N1)

Transmission and pathogenesis of swine-origin 2009 A(H1N1) 526 influenza viruses in ferrets and mice

Data blinding and analyses. Data blinding, aggregation and all statistical analyses were 408 performed by an independent statistician. Transmission outcomes were compared across 409 laboratories by each virus, using Fisher's exact test of homogeneity. Viral load between viruses 410 were compared by testing difference in area under the curve (AUC) using t-test. Factors associated 411 with transmissibility and morbidity were assessed by using logistic regression and linear regression 412 models. We also investigated the confidence in concluding low transmissibility (≤25%, or ≤1 ferret 413 infected out of 4 ferrets) or moderate to high transmissibility (≥50% or ≥2 ferrets infected out of 4 414 ferrets) from multiple contributing laboratories. We assumed a uniform prior distribution for the 415 transmission probability for a novel pathogen was assumed, and independent transmission 416 outcomes from the laboratories. The confidence of drawing conclusion on transmissibility with 417 concordant or discordant outcomes from the laboratories is presented. We considered a scenario 418 where strong homogeneity across laboratory can be assumed so the samples were pooled from 419 multiple laboratories, and another scenario that each laboratory draw their own conclusion on 420 transmissibility and the overall conclusion was drawn as voting system. All analyses were 421 conducted in R version 4.0.4 (R Development Core Team