Highly pathogenic avian influenza virus H5N2 (clade 2.3.4.4) challenge of mallards age appropriate to the 2015 midwestern poultry outbreak

Abstract Background The 2015 highly pathogenic avian influenza virus (HPAIV) H5N2 clade 2.3.4.4 outbreak in upper midwestern U.S. poultry operations was not detected in wild birds to any great degree during the outbreak, despite wild waterfowl being implicated in the introduction, reassortment, and movement of the virus into North America from Asia. This outbreak led to the demise of over 50 million domestic birds and occurred mainly during the northward spring migration of adult avian populations. Objectives There have been no experimental examinations of the pathogenesis, transmission, and population impacts of this virus in adult wild waterfowl with varying exposure histories—the most relevant age class. Methods We captured, housed, and challenged adult wild mallards (Anas platyrhynchos) with HPAIV H5N2 clade 2.3.4.4 and measured viral infection, viral excretion, and transmission to other mallards. Results All inoculated birds became infected and excreted moderate amounts of virus, primarily orally, for up to 14 days. Cohoused, uninoculated birds also all became infected. Serological status had no effect on susceptibility. There were no obvious clinical signs of disease, and all birds survived to the end of the study (14 days). Conclusions Based on these results, adult mallards are viable hosts of HPAIV H5N2 regardless of prior exposure history and are capable of transporting the virus over short and long distances. These findings have implications for surveillance efforts. The capture and sampling of wild waterfowl in the spring, when most surveillance programs are not operating, are important to consider in the design of future HPAIV surveillance programs.


| INTRODUCTION
In November/December 2014, highly pathogenic avian influenza virus (HPAIV) H5N8 was discovered in British Columbia, Canada and the nearby state of Washington, USA. 1,2 This was from the Eurasian (EA) H5 clade 2.3.4.4 that originated in China and spread by waterfowl across Asia into Europe in 2014. 3 It was the first detection of this H5 clade 2.3.4.4 HPAIV in North America (NA) and it was likely transported into NA by migrating waterfowl from Asia through Alaska and the Pacific flyway. 4,5 The HPAIV quickly reassorted with low pathogenic avian influenza viruses (LPAIVs) that are essentially indigenous in NA waterfowl, into hybrid EA/NA HPAIV including H5N1 and H5N2 subtypes. 6,7 These viruses were subsequently detected in wild bird and poultry surveillance efforts throughout the Pacific flyway and sporadic locations elsewhere in the western United States. 8 In early March 2015, the HPAIV H5N2 clade 2.3.4.4A virus appeared in a Minnesota poultry operation and over the next 3-4 months led to the destruction of more than 50 million birds at 110 facilities in five states, in efforts to stem the outbreak. The economic cost of the outbreak and control measures was estimated to be nearly 5 billion dollars. 9 The roles of wild waterfowl in the transport and transmission of this virus during the outbreak are unknown. No HPAIV was detected in wild waterfowl in the poultry outbreak area, and the only wild birds found with the virus in that region were a black-capped chickadee (Poecile atricapillus) and a Cooper's hawk (Accipiter cooperii) in Minnesota and a Wisconsin snowy owl (Nyctea scandiaca). These detections all came from passive collection of moribund or dead birds and raise the issue of how efficient and effective surveillance in wild birds was during this agricultural emergency. 10 To better define how wild waterfowl were involved in the outbreak, we challenged adult wild mallards (Anas platyrhynchos), as well  Webster City, IA) and brooded and housed in a separate BSL-3 room from the adult birds to prevent naturally acquired infections being transmitted from the adults to the immunologically naïve young birds.
All birds had their primary flight feathers clipped, were provided food and water ad libitum, and allowed free movement within the room with access to tubs of fresh water for bathing and swimming. All husbandry and experimental procedures were performed according to methods approved by the NWHC Institutional Animal Care and Use Committee. Birds were housed in BSL-3 for $8 months until the commencement of the study in April 2019. Prior to start of the study, the birds were moved into high efficiency particulate air (HEPA)-filtered isolator cages (two birds/cage) with a mixture of seropositive adults, seronegative adults, and young birds ( Table 1). The birds acclimated in the cages for 3 days prior to the commencement of the study.

| Serological testing
Blood was collected from the wild-caught adult mallards by jugular venipuncture in November 2018 and from all birds in May 2019.
These were 6 months before and immediately prior to initiation of the study, respectively. Sera were separated from the cellular blood components by centrifugation and stored at À20 C until tested for influenza virus antibodies using the MultiS-Screen AI Virus Antibody Kit (IDEXX Laboratories Westbrook, ME) according to the manufacturer's instructions. This assay detects antibodies to influenza A virus (IAV) nucleoprotein (NP).
Hemagglutination inhibition (HI) assays were performed on a panel of reference viruses representing IAV HA subtypes frequently identified in wild waterfowl. 11 Sera were receptor-destroying enzyme (RDE) treated (Seiken) according to manufacturer's instructions. Sera were heat treated at 56 C for 30 min, and phosphate-buffered saline (PBS) was added to dilute the sera to 1:10. We titrated the sera using serial twofold dilutions in 96-well microtiter plates leaving 25-μl serum dilution in each well. Virus antigens were diluted to four hemagglutinating units, 25 μl added to the sera dilutions, and incubated 1 h at room temperature; 50-μl 0.5% chicken red blood cells were added to each well and incubated for 30 min, when HI reactions were read. In a separate assay, we used horse red blood cells instead of chicken red blood cells following the same procedure. Table S1 lists the IAV used in the HI assays.

| Virus
The Midwest United States and was passaged twice in embryonated chicken eggs.

| Inoculation and sampling
The virus inoculum was prepared by diluting the stock virus isolate in brain heart infusion (BHI). The inoculum virus titer was confirmed in embryonated chicken eggs according to the method of Reed and Muench. 12 One bird in each cage was inoculated intrachoanally with 10 5 50% egg infectious doses (EID 50 ) of virus in 1 ml of BHI using a 1-ml syringe tipped with a metal canula. The birds in cage #1, one adult and one young bird, were mock inoculated using a corresponding volume of BHI. The experimental design is outlined in

| Quantitative reverse transcriptionpolymerase chain reaction
Viral RNA was extracted from cloacal and oropharyngeal swabs using the MagMAXTM-96 AI/ND Viral RNA Isolation Kit (Applied Biosystems, Foster City, CA) following the manufacturer's procedures.
Real-time reverse transcription-polymerase chain reaction (RT-PCR) was performed using the published procedure of Spackman et al. 13 qRT-PCR assays used reagents provided in the Qiagen OneStep RT-PCR kit (Qiagen, Hilden, Germany) and performed on a Stratagene Mx3005P thermal cycler (San Diego, CA). C t values of duck swabs were compared with those from a standard curve of known viral concentrations to calculate the amounts of virus excreted by infected ducks and were reported as EID 50 RNA equivalents/ml. 14 On the basis of serial dilutions of the virus and the standard curve, we estimated the limit of detection of these methods to be between 0 and 10 EID 50 RNA equivalents/ml.

| Analysis of infection dynamics and transmission
We V p was the maximum EID 50 excretion at the infection peak for mallards that were seropositive for IAV exposure on the inoculation day with coefficients β v,1 À 2 estimated for effect sizes of seronegative adults (β ,1 ) and subadults (β 2 ); t p was the estimate of the time (days postinfection) of the maximum EID 50 , and β tp,1À2 were the estimated effect sizes of adult and subadult maximum EID 50 ; g was the exponential growth rate of viral shedding, d was the exponential rate of viral shedding decline, (β d,1 À 2 ) were the estimates of decline effect sizes of adults and subadults, and t was the time (DPI) of the observed EID 50 response. We modeled the EID 50 response as a lognormally distributed response variable with hierarchical components to incorporate individual variation using Bayesian implementation and Gibbs sampling to estimate the mean values for V p , t p , and d, as well as coefficients for effects of seronegative adults (β v,1 , β tp,1 , β d,1 ) and 8-month- . We did not estimate the exponential growth rate parameter, g, because it was nonidentifiable given the sampling frequency. Rather, we included this as an informative closed-form prior that was bounded by biologically relevant values. 17 We did not attempt to model the cloacal shedding dynamics because the detection through the time was much more variable than oral detection. Details on the model, selection of priors, and fit evaluation are presented in Appendix A.

| Preexposure serology
We tested adult mallards for the presence of avian influenza antibodies using a commercial enzyme-linked immunosorbent assay (ELISA) that detects antibodies to the viral NP. This assay is extensively used to measure the serological status of wild ducks. 18 In the November sampling, 16/28 adult mallards were positive for the presence of IAV antibodies (

| HPAIV infection, excretion, and transmission in mallards
Following intrachoanal inoculation with the HPAIV, all inoculated birds became infected and excreted detectable virus by DPI2 ( Figure 1A). Neither the serological status nor the age of the birds had detectable effects on intensity or timing of oral virus excretion.
Maximum oral viral shedding in adult mallards with previous expo- F I G U R E 1 Viral excretion measured in egg infectious doses (EID 50 ) RNA equivalents/ml, log scale via oral shedding (A) and cloacal shedding (B). Inoculated birds (purple lines) were intrachoanally inoculated with 10 5 EID 50 on Day 0 and placed in a cage with one noninoculated bird to measure transmission (green lines). Each panel represents one cage with one pair of ducks (see Table 1). The limit of detection was 0-10 EID 50 RNA equivalents  (Figure 1). Thus, we were unable to estimate an accurate endpoint of the infection process.

| Disease signs in HPAIV infected mallards
All mallards survived till the end of this study except one bird, Y252B imental studies on this topic are warranted. Regardless, any antibodies present in our subjects at the time of the study, including the bird with a weak LPAIV H5N2 HI titer, had no effect on infection, shedding dynamics, or transmission of HPAIV.
All exposed mallards in our study became infected with the HPAIV, either by direct inoculation or by transmission. There were no clinical signs or indications of disease in any of the infected mallards, similar to findings from other experimental challenge studies. [21][22][23][24][25] These studies are typically conducted with very young, immunologically naïve birds as test subjects. We were the first to capture, house, and challenge adult wild ducks that had a variety of IAV exposure his-  g i $ unif 1, 10 ð Þ: We did not estimate the exponential growth rate parameter, g, because it was nonidentifiable given the sampling frequency. Rather, we included as an informative closed-form prior that was bounded by biologically relevant values  The prior specification of the parameters for additive effects of seronegative status in adults and subadults (β v,1 À 2 ), β tp,1À2 , and (β d,1 À 2 ) was a noninformative Cauchy distribution, cauchy(10, 1).
Parameter values were estimated using Gibbs sampling implemented in JAGs (Plummer, 2003) and the R package jagsUI (Kellner, 2019) using three parallel MCMC chains of 250 000 iterations following a 50 000 iteration burn-in. Posterior distributions were generated after confirmed convergence of the MCMC chains (Table A1) and examination of parameter correlation ( Figure A1) for all parameters and sampled every 20th iteration.