1 Host type governs influenza evolutionary strategy across 2 reservoir and spillover hosts 3 Maria A. Maltepes1, Alexey Markin2, Stephen Shank3, Jordan T. Ort4, Jared Sabre5, Lambodhar 4 Damodaran3,5, Grant Park1, Kathryn Kistler6,7, Tavis K. Anderson2, Louise H. Moncla3 5 1Department of Biology, UPenn, Philadelphia. 2National Animal Disease Center, Agricultural Research Service, 6 United States Department of Agriculture. 3Department of Pathobiology, UPenn, Philadelphia. 4Department of 7 Microbiology, Perelman School of Medicine, UPenn, Philadelphia. 5Office of Information Technology, UPenn, 8 Philadelphia. Department of Biology, Emory University, Atlanta. 6Fred Hutchinson Cancer Research Center, Seattle. 9 7 Howard Hughes Medical Institute, Seattle. 10 Abstract 11 Despite its high propensity for host switching, the evolutionary mechanisms underlying 12 influenza host adaptation remain unclear. H3Nx influenza viruses are uniquely 13 generalist, with long-term lineages that circulate in avian, human, swine, equine, and 14 canine hosts. Using 13,295 H3Nx sequences, we quantified host-specific adaptive 15 evolution, and developed a pipeline to map reassortment events onto trees with 16 measures of statistical uncertainty. We find that while H3Nx viruses in mammals 17 undergo adaptive evolution in HA and NA, viruses in birds experience very little 18 directional selection. Instead, avian lineages exhibit high rates of reassortment, 19 frequently generating novel reassortant lineages that persist transiently and turn over 20 rapidly. 29.8-47.4% of all avian reassortant lineages are purged within the first year of 21 circulation, and reassortment shows no fitness benefit in birds. In contrast, reassorted 22 lineages in swine are more likely to persist long-term, suggesting that reassortment in 23 swine may be broadly beneficial. Segment-specific reassortment patterns were also 24 distinct between avian and mammalian viruses, with NA reassorting more frequently 25 than expected in birds, but less frequently than expected in swine. Reassortment events 26 are enriched between mammalian, but not avian, host switches, suggesting that 27 reassortment may be most beneficial for mediating host switches among mammalian 28 species. Together, our data suggest that host differences drive fundamentally different 29 evolutionary outcomes for influenza viruses, transitioning from reassortment-dominant 30 evolution in their avian reservoir, to varying degrees of adaptation upon establishment in 31 mammals. 32 33 Introduction 34 Influenza A viruses (IAVs) have a high propensity to host-switch, posing persistent 35 challenges for human, animal, and wildlife health (Taubenberger and Kash 2010). 36 However, there are very few natural examples of viruses that have crossed species 37 barriers and been well-sampled in both the reservoir and novel hosts. H3Nx influenza 38 viruses are an unusually generalist subtype that circulate enzootically in global wild bird 39 populations (Yang et al. 2025; Yoon et al. 2014) and have established decades-long 40 circulating lineages in humans, swine, canines, and equines. From these established 41 lineages, additional spillovers into humans, swine, felines, camels, mink, seals, and 42 donkeys have occurred, providing a case study on how circulation in distinct hosts 43 impacts the fundamental forces that shape influenza virus evolution (Trovão et al. 2024; 44 Crawford et al. 2005; Song et al. 2008; Wasik et al. 2025; Venkatesh et al. 2020; 45 Yondon et al. 2014; Kuchinski et al. 2025; Yang et al. 2018; Song et al. 2011; Le Sage 46 et al. 2026; Webby et al. 2000). 47 Adaptation to a distinct host environment can be aided by two major evolutionary 48 forces acting on IAV’s eight-segmented genome: mutation and reassortment. 49 Reassortment is the only mechanism analogous to recombination in influenza viruses, 50 permitting the instantaneous acquisition of whole gene segments during co-infection. 51 While accumulation and selection of adaptive mutations is incremental, reassortment 52 can rapidly introduce advantageous or purge deleterious genotypes within a population 53 (Steel and Lowen 2014). Reassortment has been linked to the emergence of at least 3 54 naturally occurring human pandemics, and is thought to facilitate host switching by 55 bringing together novel gene segments to overcome host barriers (Ma et al. 2016; 56 Furuse et al. 2010; Nelson et al. 2008; Lindstrom et al. 2004). Reassortment within 57 avian hosts is thought to occur frequently and freely among co-infecting, homologous 58 strains (Dugan et al. 2008; Marshall et al. 2013). In contrast, reassorted progeny from 59 divergent parental strains can be constrained by segment mismatch in non-avian hosts 60 (Ganti et al. 2021; White and Lowen 2018). Reassortment is likely impacted by host- 61 specific differences in ecology, prevalence, and virologic constraint (Lowen 2017), but is 62 understudied due to methodologic limitations. 63 64 Reassortment inference frequently relies on comparing segment phylogenies to 65 identify incongruence. Common approaches include constructing tanglegrams, which 66 require subjective interpretations (de Vienne 2019; Ovadia et al. 2011), and classifying 67 sequences into genotypes as proxies for reassortment (Youk et al. 2023). Alternatively, 68 Bayesian models enable inference of the underlying reassortment network, but are 69 computationally intensive and are generally limited to small datasets (Müller et al. 70 2020). To address these limitations, TreeSort was developed and validated to 71 accurately detect reassortment on large influenza datasets (Markin et al. 2025). 72 TreeSort uses the molecular clock signal in the evolution of individual gene segment 73 trees to identify recent and ancestral reassortment events. The algorithm then maps 74 reassortment events onto a reference tree, and provides point estimates of 75 reassortment rates (i.e., the expected number of reassortment events per year) and 76 associated mappings, enabling large-scale reassortment mapping along with 77 information on host species. 78 79 Here, we leveraged the broad host range of H3Nx viruses to examine how 80 reassortment and selection jointly impact viral evolution and host specific adaptation. 81 We developed a pipeline to parallelize TreeSort measurements to quantify reassortment 82 along with measures of uncertainty, and measured reassortment and directional 83 selection in each host. We uncover patterns of adaptive evolution and reassortment that 84 vary substantially among species, and show that reassortment reflects host-dependent 85 patterns of segment mixing. Our data suggest differential fitness effects of reassortment 86 across hosts, with rapid lineage turnover in birds, but longer-term persistence and 87 fitness benefits in swine. Finally, we show that reassortment is enriched on branches 88 leading to host switches between mammals, but not between birds, suggesting that 89 reassortment is particularly beneficial for mediating mammal to mammal cross-species 90 transmission. These data establish reassortment as a ubiquitous generator of viral 91 diversity that differs systematically between avian and mammalian hosts, and suggest 92 that host differences drive fundamentally distinct outcomes for influenza virus evolution. 93 94 Results 95 96 H3Nx viruses have jumped hosts multiple times 97 H3Nx viruses have been studied extensively in humans and swine, with less 98 work focused on how these viruses evolve across their full host range (Wasik et al. 99 2025; Parrish et al. 2015; Webby et al. 2000). To reconstruct the complete evolutionary 100 history of these viruses, we curated a dataset of 13,295 non-human H3Nx sequences 101 sampled from horses, dogs, swine, birds, and other spillover hosts between 1963 and 102 2024. To mitigate sampling differences between species, we subsampled sequences by 103 year, host, country, and subtype to generate a dataset of 5,023 full genomes among all 104 non-human hosts. Human seasonal H3N2 viruses were included separately, 105 subsampled by year and country, generating a dataset of 1,078 full human influenza 106 genomes sampled from 1968-2023. Three human spillover H3N8 viruses of avian-origin 107 were also included. 108 Merging these subsampled datasets brought the final dataset size to 6,104 full 109 H3Nx genomes spanning 9 NA (N1-N9) subtypes. We inferred time-resolved 110 phylogenies for each of the eight genomic segments, revealing host jumps that have 111 established continuously circulating H3Nx viruses in North American avian, Eurasian 112 avian, equine, canine, human, and swine populations (Figure 1A, Supplementary Figure 113 1). These host switch events match known patterns of spillover that have led to 114 repeated spillbacks between humans and swine, establishment of H3N8 equine and 115 canine lineages, and the recent establishment of canine H3N2 viruses from birds 116 (Crawford et al. 2005; Song et al. 2008; Webby et al. 2000; Sharma et al. 2022; Zeller et 117 al. 2024; Rajão et al. 2015) (Supplementary Figure 2). We recapitulate a known 118 spillover of H3N8 viruses from birds to horses, which established the H3N8 equine 119 influenza lineage that still circulates as of 2026. Spillover of these viruses to canines 120 established a now-extinct canine H3N8 lineage that circulated from 1999 to 2016 121 (Wasik et al. 2023). In 2004, H3N2 influenza viruses spilled directly from birds to dogs, 122 establishing the extant canine H3N2 lineage (Wasik et al. 2025). Human H3N2 viruses 123 are descendants of the 1968 H3N2 pandemic strain. Notably, the continuously 124 circulating swine lineages across the HA tree descend from human introductions, 125 representing multiple independent spillover events that established distinct, endemic 126 swine lineages now globally distributed across multiple clades (Nelson and Vincent 127 2015; Nelson et al. 2014). For the purpose of comparing evolution across species, we 128 delineated host and geographic lineages (see Methods for details), and built 129 independent trees for each, providing the framework for subsequent analyses of 130 reassortment, adaptive substitution rates, and host-switching. 131 Adaptive substitution rates vary by host 132 Human seasonal influenza viruses evolve through yearly selective sweeps, in 133 which strong, directional selection fixes HA genes that mediate immune escape 134 (Rambaut et al. 2008). Differences in the strength of selection between species could 135 arise from differences in life span, vaccination patterns (e.g., in horses and swine), and 136 the frequency of repeat infections (Ferguson et al. 2003; Wille and Holmes 2020; 137 Petrova and Russell 2018). To determine the strength of selection across each host 138 lineage, we performed a modified McDonald-Krietman (MK) test on two surface 139 glycoproteins (the hemagglutinin (HA), neuraminidase (NA)), and a highly conserved 140 polymerase gene (the polymerase basic 1 (PB1)). While HA and NA are exposed to 141 host-specific receptor and immune factors and may undergo selection, PB1 is the most 142 conserved, and less likely to experience strong, host-specific selection. This test 143 calculates an adaptive substitution rate by measuring the rate of nonneutral substitution 144 accumulation over time that exceeds a neutral expectation, and was developed 145 specifically for temporally-sampled viral genomes (Kistler and Bedford 2023; Bhatt et al. 146 2011). Because this test assumes that viruses are co-circulating and competing with 147 each other to at least some degree, we subsetted viruses by host and geography for 148 analysis (see Methods). 149 150 Adaptive substitution rates varied substantially between gene segments, with the 151 highest adaptive rates in the HA and NA surface glycoprotein genes. Adaptive 152 substitution rates were largely similar across all host groups for PB1, in line with 153 previous findings suggesting limited direction selection on this polymerase protein 154 (Kistler and Bedford 2023). Human seasonal H3N2s and European swine viruses 155 exhibited the highest rates of positive selection in HA, followed by canine, equine, and 156 North American swine lineages. Differences in adaptive rates between North American 157 and European swine groups could arise due to regional variation in farming practices, 158 swine movement, and vaccination coverage, or from the distinct evolutionary histories of 159 these lineages. We found no detectable signal of adaptive evolution in HA among 160 Eurasian and North American avian lineages, with adaptive rates equivalent to those 161 measured in PB1. Adaptive rate estimates for NA were moderately elevated, but 162 overlapped across all host lineages, with no host lineage exhibiting significantly higher 163 rates relative to all other hosts (Figure 1B, Supplementary Figure 3). These data 164 indicate that surface protein genes experience distinct selection patterns across 165 species, with very little directional positive selection on HA in birds, and moderate 166 selection in swine, canine, and equine lineages. 167 168 169 170 Figure 1: H3Nx viruses have jumped hosts multiple times, where host clades 171 exhibit varying levels of selection. A) The HA phylogeny (N=6,104) is colored by host 172 type, where each leaf is a unique strain and each internal node is an inferred ancestor. 173 Notable host switches are highlighted. The 95% confidence interval for the TMRCA 174 estimates are between 1910-05-22 – 1917-09-14. B) Rates of adaptation (adaptive 175 substitutions per codon per year * 10-3) for each host are shown for the genes encoding 176 the surface proteins HA and NA, as well as the polymerase gene PB1. Significant 177 differences between host and genes are represented by estimates that fall outside of 178 another estimate’s error bars. 179 Reassortment rates vary by host and geography 180 181 In wild aquatic birds, low pathogenicity avian influenza viruses are thought to 182 evolve frequently via reassortment, a process in which genomic segments from co- 183 infecting virions mix within an infected host cell (Dugan et al. 2008; Wille et al. 2013). 184 Reassortment in wild birds is hypothesized to facilitate immune escape by mediating the 185 continual generation of novel subtypes, which may reduce the necessity of antigenic 186 selection (Roche et al. 2014). Influenza viruses also reassort frequently in swine, with 187 reassortment known to be associated with changes in phenotype that may mediate 188 interspecies transmission (Smith et al. 2009; Lycett et al. 2012; Khiabanian et al. 2009; 189 Thomas et al. 2024). We next aimed to determine whether reassortment rates and 190 patterns varied across hosts. 191 192 Reassortment inference is frequently limited by a lack of methods that can perform 193 robust inference across large datasets. TreeSort is an approach validated for identifying 194 reassortment among influenza viruses (Markin et al. 2025) that designates one segment 195 as the reference, and then systematically compares the evolutionary histories of the 196 remaining segments to identify signatures of incongruence. To account for statistical 197 uncertainty, we developed a pipeline in which TreeSort is run 1000 independent times, 198 and a summary reassortment tree is generated in which each reassortment event (and 199 the segments involved) is annotated with a support value. We then produce summary 200 trees retaining only high-support reassortment events that occur in at least 95% of 201 replicates (Figure 2B, Figure 2C, Supplementary Figure 4). 202 203 Reassortment rates were generally robust across TreeSort replicates. However, 204 the mapping of reassortment events onto specific branches varied across replicates, 205 particularly for high-reassortment hosts (avian and swine) (Supplementary Figure 8). 206 Overall, 54.4% of inferred reassortment events met the 95% support threshold, 207 supporting the utility of our replicate-based approach for clarifying high-support events. 208 Using these high-support events, we calculated reassortment rates for each host clade. 209 Host-specific reassortment rates were highest in North American avian viruses (mean ± 210 sd: 0.3472 ± 0.0131 reassortments per year per strain), and markedly lower in Eurasian 211 avian viruses (0.1248 ± 0.0018). While H3s are among the most prevalent subtype in 212 North American birds, H6 and H4 subtypes dominate in European wild migratory birds, 213 which may account for these differences (Diskin et al. 2020; Munster et al. 2007). Within 214 North America, reassortment was most common in the Mississippi flyway, suggesting 215 that geographic variation may also impact rates (Supplementary Figure 5), though 216 expanded datasets are necessary to confirm these results. Mammalian hosts exhibited 217 lower reassortment rates overall, with the highest mammal rates observed in swine 218 (0.1303 ± 0.0052), and lower rates in human (0.0952 ± 0.0063), canine H3N2 219 (0.0446 ± 0.0044), and equine (0.0096 ± 0.0001) (Figure 2A) lineages. Overall, patterns 220 of reassortment and adaptive substitutions were roughly inverse across species, with 221 the highest reassortment rates and lowest adaptive substitution rates in birds, and low 222 reassortment and higher adaptive substitutions in equines, canines, and humans. 223 Uniquely, swine influenza viruses exhibited both high reassortment rates and detectable 224 adaptive evolution in HA and NA. 225 226 To assess the sensitivity of reassortment inference to dataset size, we generated 227 serially subsampled datasets of North American avian viruses, ranging from N=400 to 228 N=1100 (increasing by 100 sequence increments, 5 trials each). Reassortment rates 229 increased with dataset size, suggesting that absolute rates may be sensitive to 230 sampling (Supplementary Figure 6). However, the relative ranking of reassortment rates 231 was consistent across datasets, with North American avian viruses exhibiting the 232 highest rates even when heavily downsampled (N=400). Similarly, North American 233 avian viruses exhibited higher rates than swine despite fewer available sequences. We 234 also found high reassortment support for a range of root-to-tip divergence values, 235 indicating that TreeSort does not preferentially detect recent reassortments 236 (Supplementary Figure 7A). Similarly, reassortment support showed minimal correlation 237 with branch length, suggesting that reassortment inference is not biased towards long, 238 divergent branches (Supplementary Figure 7B). Prior TreeSort validation has also 239 demonstrated a limited association between reassortment rates and genomic diversity, 240 suggesting that observed differences in reassortment frequencies are not solely driven 241 by differences in circulating diversity across these populations (Markin et al. 2025). 242 Together, these data suggest that the differences in reassortment rates among hosts 243 cannot be explained purely by differences in sampling or diversity. 244 245 Figure 2: Reassortment rates vary across hosts and regions. A) Reassortment 246 rates (reassortments per lineage per year) were calculated for each H3Nx host group. 247 Each point represents a reassortment rate calculated for an individual TreeSort 248 replicate. All 1000 replicates are shown, with the mean value indicated with a line. 249 Though more high-support reassortments were called in Eurasian avian viruses (415 250 reassortments) than in swine viruses (307 reassortments), the lower average rate in 251 Eurasian avian viruses reflects differences in the host-specific molecular clock rates 252 used to calculate the rate (see Methods). Summary TreeSort trees are shown for the B) 253 North American avian and C) swine host groups. Branches are colored to indicate 254 reassortment events supported at >= 95%. 255 256 257 Segments reassort nonrandomly 258 259 Influenza reassortment is known to be restricted by segment compatibility (White 260 and Lowen 2018; Lowen 2017). HA and NA must be functionally compatible, while 261 compatibility among segment packaging signals can further augment reassortment 262 viability (Mitnaul et al. 2000; White et al. 2017; Baker et al. 2014; Essere et al. 2013; Liu 263 et al. 2022). To determine whether segment-specific reassortment patterns differed 264 among hosts, we quantified segment specific reassortment frequencies. We then 265 compared these results to expectations under a null distribution generated by simulating 266 the same number of reassortment events as observed in our actual data, but randomly 267 sampling segments for each event with equal probability. This null thus represents the 268 expected frequency of reassortment of each segment under a model in which 269 reassortment is random. Given the relatively low rates of observed reassortment in 270 human, canine, and equine lineages, we focused subsequent reassortment analyses on 271 the avian and swine host groups for which we had sufficient statistical power. For all 272 subsequent analyses, we report reassortment events relative to HA’s evolution. 273 274 In both avian and swine lineages, many segments reassorted nonrandomly (p < 275 0.05 after Bonferroni correction). Among North American avian viruses, NA and PA 276 were significantly overrepresented in reassortment events (p = 0.007 for both), while NS 277 and MP were underrepresented (p = 0.02 and p = 0.007) (Figure 3A). In Eurasian avian 278 viruses, NA and NS reassortments were overrepresented (p = 0.007 and p = 0.02) and 279 PB1, PA, and MP were underrepresented (p = 0.02, p = 0.007, and p = 0.007, 280 respectively; Supplementary Figure 9A). Despite past evidence pointing to permissive 281 gene segment combinations in birds (Ganti et al. 2021; Dugan et al. 2008), these 282 patterns indicate that gene mixing is not entirely free or random among H3Nx viruses. 283 Instead, we find evidence in both Eurasian and North American avian populations for 284 increased NA reassortment events. Non-random reassortment could reflect functional 285 constraints on segment compatibility or non-random co-circulation of subtypes in the 286 avian reservoir, leading to unequal co-infection frequencies. In swine, reassortment 287 patterns were distinct, with reassortments involving NA and PB1 underrepresented (p = 288 0.007 and p = 0.03) (Figure 3B). The high overall rates of reassortment in swine confirm 289 that opportunities for co-infection and reassortment occur readily, while the strong 290 observed linkage between HA, NA, and PB1 (which rarely reassort away from each 291 other), could reflect that these gene pairs are selectively beneficial, or could arise from 292 constraints in gene segment packaging that conserves these pairings (Lowen 2017; 293 Nelson et al. 2014; Mitnaul et al. 2000; Ma et al. 2012). Because swine H3Nx viruses all 294 descend from spillovers from humans, it is also possible that these findings reflect 295 maintenance of these gene pairings in human-adapted H3N2 strains following spillover 296 (Yen et al. 2011; Xu et al. 2012). While our data cannot distinguish the reasons for 297 these patterns, they do show that reassortment patterns in both swine and avian 298 species are non-random, and differ substantially from each other. 299 300 Novel NA subtypes are introduced at a rate exceeding expectations based on 301 viral cocirculation alone 302 303 In both European and North American birds, we observed an excess of 304 reassortments involving NA, but did not distinguish between within and between 305 subtype events. Within-subtype reassortment events could be favored if only some 306 combinations of HA and NA are functionally compatible, allowing within-subtype 307 reassortments to better maintain HA:NA balance. Alternatively, if high subtype diversity 308 is preferentially maintained in wild birds (e.g., through selective or ecological 309 processes), then reassortment of new NA subtypes might occur more frequently. Using 310 the frequencies of NA subtypes within our dataset, we calculated the expected 311 frequency of within vs. between subtype reassortment events as the probability of 312 sampling two sequences of the same, or distinct, NA subtypes from the population 313 (Supplementary Figure 10). We then classified all observed NA reassortment events as 314 within- or between-subtype events using a divergence-based threshold (see Methods 315 for details). Analysis of pairwise divergence values showed clear delineation between 316 pairwise divergence rates for within vs. between-subtype NA sequences, supporting this 317 approach (Supplementary Figure 11, Supplementary Figure 12). We identified 35 318 within-subtype and 138 between-subtype reassortment events in the North American 319 avian lineage, and 39 within-subtype and 109 between-subtype events in the Eurasian 320 avian lineage. For avian viruses, the expected frequency of within- and between- 321 subtype reassortments were 0.382 and 0.618, respectively for North American viruses, 322 and 0.412 and 0.588 for Eurasian viruses. Comparing observed counts to these 323 expected probabilities under a binomial model revealed a significant excess of between- 324 subtype reassortments in both avian populations (Figure 3C, Supplementary Figure 9B). 325 326 Reassortment detection depends on identifying lineages that are more divergent than 327 expected, and could be biased towards detecting between-subtype reassortment 328 events. To estimate the robustness of our results to this bias, we conducted a sensitivity 329 analysis to estimate the robustness of this finding to varying underdetection rates for 330 within-subtype events. We first calculated the divergence value among all detected NA 331 reassortment events, and then determined the fraction of within-subtype NA sequence 332 pairs that were less divergent than the minimum detected reassortant value. Among 333 avian viruses, 1.6-2.4% of all NA sequences had pairwise divergences that fell below 334 the minimum divergence value of detected reassortments, suggesting the potential for 335 TreeSort to misclassify these events as false negatives. By sequentially assuming 336 varying rates of false negatives, we estimate that our finding of excess between-subtype 337 reassortants is robust to an underdetection rate of 7% (for Eurasian avian) to 11% (for 338 North American avian), substantially higher than our estimated underdetection rate (1.6- 339 2.4%) (Supplementary Figure 13, Supplementary Figure 14). These data suggest that in 340 both Eurasian and North American avian populations, novel NA subtypes are introduced 341 by reassortment more frequently than expected from cocirculation alone. Furthermore, 342 this finding is robust to a moderate degree of underdetection of within-ubtype 343 reassortment events. 344 345 Figure 3: Segments exhibit host-specific compatibility with HA. Segments reassort 346 nonrandomly relative to a null model in A) North American avian and B) swine viruses. 347 Data points represent simulated reassortment counts for each segment under random 348 sampling while observed counts are denoted with a black X. Asterisks indicate p-value < 349 0.05 after Bonferroni correction. C) Binomial testing revealed a significant excess of 350 between-subtype NA reassortments (p-value = 3x10-7). Success probabilities for the 351 binomial model were defined as the within-subtype (0.382) and between-subtype 352 (0.618) NA reassortment probabilities. Observed within-subtype (blue, 35 events) and 353 between-subtype (yellow, 138 events) NA reassortments are indicated by vertical lines. 354 355 Reassortment provides differential fitness effects across host groups 356 357 The high reassortment rate and non-random segment patterns observed in avian 358 and swine lineages supports reassortment as a critical component of the evolutionary 359 process in these species. To assess whether reassortment confers fitness advantages 360 within avian and swine lineages, we used reassortant lineage persistence as a proxy for 361 viral fitness. Selectively “fit” viruses should produce more offspring, which can be 362 quantified by assessing how long descendant lineages persist into the future (Müller et 363 al. 2020). Here, a reassorted lineage was defined as the longest path between a 364 reassortment event and either a subsequent reassortment event or terminal node, with 365 persistence measured in years of circulation. This analysis was restricted to the avian 366 and swine host groups due to limited statistical power in the remaining mammalian 367 hosts. While average reassortant persistence times did not differ significantly between 368 avian and swine (Supplementary Figure 15), we observed notable differences in 369 reassortant lineage turnover rates. North American avian reassortant lineages 370 experience rapid turnover, with approximately 47.4% of these lineages purged within 371 the first year of circulation. These findings align with previous, smaller-scale studies 372 demonstrating that circulating genotypes are continuously replaced by novel genotypes 373 created through reassortment (Macken et al. 2006). In contrast, only 29.8% and 34.7% 374 of reassortant lineages are purged within the first year in Eurasian avian and swine 375 viruses, respectively, indicating a much higher fraction of reassortment lineages that 376 persist into the future. 377 378 To determine whether differences in turnover demonstrate differential fitness 379 effects (vs. simply different reassortment rates), we compared reassortant lineage 380 persistences to those under a null model. We shuffled reassortment events across the 381 trees 1000 times and compared persistence times from our observed data to those 382 calculated from these null, shuffled datasets. Persistence times of reassortant lineages 383 showed contrasting patterns across host groups. North American avian reassortant 384 lineages generally persisted in accordance with null expectations (Figure 4A), while 385 Eurasian avian reassortant lineages showed decreased survival compared to the null 386 model at later persistence times (Figure 4B). In contrast, swine reassortant lineages 387 persisted marginally longer than expected under null expectations in the short-term 388 (Figure 4C), indicating that reassortment in swine may confer a fitness advantage. 389 These patterns were also recapitulated when comparing reassortant and 390 nonreassortant lineage persistence within each host group (Supplementary Figure 16), 391 indicating observed lineage turnover frequencies are not an artefact of tree structure. 392 393 We next quantified the probability that reassortant and non-reassortant lineages 394 sustain descendants at 0.5, 2, 6, and 10 years post-reassortment using Fisher's exact 395 tests. If reassortment is associated with a fitness benefit, we reasoned that reassortant 396 lineages should be more likely to sustain long-term descendants than non-reassortant 397 lineages. In North American avian viruses, reassortment had no significant effect on 398 long-term survival across most time points, with the exception of a marginally significant 399 decrease in survival at x = 10 years (OR: 0.63; p = 0.04). The odds ratios decreased 400 with increasing time intervals, suggesting a trend toward reduced long-term fitness 401 (Figure 4C, left panel). This trend was recapitulated in Eurasian avian viruses, where 402 significant decreases in survival emerged at x = 6 and x = 10 years (OR: 0.52 and 0.56; 403 p = 0.008 and 0.046, respectively). In contrast, swine viruses showed significantly 404 elevated survival probabilities at all time points except x = 10 years. Critically, at x = 0.5 405 and 2 years, survival effects of reassortment in swine viruses fell outside the distribution 406 of odds ratios under a null model of reassortment (Supplementary Figure 18), 407 suggesting that these results are unlikely caused by underlying tree topology. While we 408 cannot exclude the potential impacts of swine population structure and turnover and 409 immune heterogeneity on these findings, our data suggest an association between 410 reassortment and long-term lineage persistence in swine. These results are consistent 411 with surveillance data indicating that most novel swine genomes persist for ~1.8 years, 412 though dominant lineages can persist much longer (~10 years) (Janzen et al. 2025). 413 Together, these differences in long-term survival mirror the trends observed in the 414 reassortant persistence analysis, pointing to a potential association between 415 reassortment and fitness in swine, but not avian, viruses. 416 417 418 419 Figure 4: Reassortment exhibits differential fitness effects by host. Reassortant 420 persistence analyses for A) North American avian, B) Eurasian avian, and C) swine 421 host groups. Shaded regions indicate persistence values calculated under a null model 422 of random reassortment. Dashed lines represent observed reassortant persistence 423 values. Each time bin represents the proportion of reassortant lineages still in circulation 424 at that time. D–F) Long-term survival effects of reassortment were quantified using 425 Fisher's exact test, where fitness was defined as a reassortment event sustaining 426 descendants at x = 0.5, 2, 6, and 10 years into the future. Points represent odds ratios 427 (OR) with error bars denoting 95% confidence intervals. Filled points indicate 428 statistically significant results (p < 0.05). 429 430 Reassortment is enriched for switches between mammalian hosts 431 432 Reassortment is a key driver of pandemic virus formation, and is thought to be 433 critical for host switching (Ma et al. 2016; Ince et al. 2013; Mehle et al. 2012; 434 Taubenberger and Kash 2010). To determine whether reassortment was statistically 435 associated with host switches, we classified branches in the tree as “host switch” or “not 436 host switch” branches, and evaluated the frequency of reassortment events on each 437 branch type. We observed a significant enrichment of reassortment on host-switching 438 branches (OR= 5.49, p-value= 9 x 10-7, 95% CI=3.08, 9.81, Fisher’s exact tests) (Figure 439 5A), consistent with previous findings linking these processes. Among the 99 total host 440 switches identified in the global H3Nx tree, 35 (35%) involved reassortment. Of these 441 reassorted host switch branches, 21 occurred on internal nodes, which comprised of 442 mammal-to-mammal (18 events: 17 human-to-swine, 1 swine-to-human), avian-to- 443 mammal (2 events: 1 avian-to-seal, 1 avian-to-human) and mammal-to-avian (1 event: 444 swine-to-avian) transitions (Figure 5C). Details of these reassortment events, including 445 those occurring on terminal nodes, and the segments involved are provided in 446 Supplementary Table 1. 447 448 To determine whether this association varied significantly across the taxonomic 449 groups, we compared the observed frequency of reassorted host-switched internal 450 nodes to the expected frequency under random reassortment. This analysis revealed 451 that reassortment enrichment was taxonomically specific. Reassortment was 452 significantly overrepresented on nodes representing mammal-mammal host switches, 453 with the majority being human-to-swine transitions (Figure 5C). In contrast, no 454 significant enrichment was detected for host switches between avian orders (North 455 American avian: OR=0.49, p-value= 0.69, 95% CI= 0.058, 4.05; Eurasian avian: OR= 456 1.08, p-value= 0.84, 95% CI= 0.50, 2.33), which include switches between Anseriformes 457 (ducks, geese) and Galliformes (chickens, turkeys) (Figure 5A, Figure 5B). These 458 results suggest that reassortment may play a distinct role in mammalian adaptation 459 compared to avian host switching. However, this analysis cannot determine the 460 temporal relationship between reassortment and host-switching events, leaving open 461 whether reassortment precedes or follows successful host switches. 462 463 464 465 Figure 5: Reassortment is enriched for host-switching between mammalian hosts, 466 but not avian hosts. Fisher's exact test assessing whether reassorted branches are 467 more likely to be host-switched yielded a significant association for A) all host 468 transitions but no significant association when restricted to switches between avian 469 orders. B) For each transition type, observed reassortment events on host-switched 470 internal nodes (black X) are compared against expectations under a random model of 471 reassortment (colored points). Enrichment is significant only for mammal-to-mammal 472 transitions (p = 0.001; orange), with no significant enrichment for avian-to-mammal 473 (blue) or mammal-to-avian (green) transitions. C) Total host-switch counts, on both 474 internal and terminal nodes, are shown for each transition type. 475 476 Discussion 477 478 We established a robust phylogenetic approach for evaluating how host-specific 479 ecology and biology shape fundamental evolutionary processes in IAVs. Our data 480 reveal a dichotomy in how mutation and reassortment contribute to IAV diversification 481 across hosts. Across species, adaptive substitution rates fall along a spectrum, with the 482 highest rates in humans, followed by other mammals, and the lowest rates in birds. 483 Reassortment rates also vary substantially, but in the opposite direction, with the 484 highest rates in birds, and lowest rates in canines and equines. Among the H3Nx 485 viruses, mammalian lineages originally descend from spillovers from birds (and for 486 swine lineages, subsequent human to swine spillovers), suggesting that influenza 487 viruses can switch from a reassortment-dominant mode of evolution in birds to varying 488 degrees of adaptive evolution upon establishment in new species. The variation in these 489 evolutionary strategies across hosts may arise from a combination of ecological, 490 biological, and epidemiologic differences between species, including lifespan, 491 aggregation/housing, vaccination, and the degree of influenza diversity circulating in 492 that population. These features likely interact to determine whether mutation or 493 reassortment predominates in driving viral diversification across host species, 494 highlighting the flexibility of these viruses to modulate the strength of adaptive evolution 495 and reassortment. This flexibility may be one factor that enables efficient host switching 496 and adaptation, which is uniquely common among the H3Nx viruses. 497 Despite decades of study, surprisingly little is still known about how influenza 498 viruses evolve in their natural reservoir. We observe low adaptive rates in birds, which 499 may reflect the migration and shorter lifespans of wild bird species (particularly ducks, 500 which dominate our datasets), leading to more frequent population turnover when new 501 immunologically naïve susceptibles enter the population (Munster et al. 2007). As the 502 primary reservoir host, avian species also harbor extensive subtype diversity of low- 503 pathogenicity viruses, enabling mixing with a broad diversity of subtypes that co- 504 circulate. Under these conditions, reassortment may become the primary driver of viral 505 evolution in avian IAVs. We find that avian reassortant lineages are short-lived and 506 show no evidence of being selectively beneficial. The high reassortment rate coupled 507 with rapid reassortant lineage turnover in birds is consistent with a model in which short 508 avian lifespans and environmental durability of IAV promote a diverse reservoir that 509 repeatedly exposes naïve hosts to novel subtype combinations (Roche et al. 2014). 510 These data suggest that avian influenza viruses rely primarily on reassortment for their 511 evolution, with reassortment serving as a frequent, but largely stochastic phenomenon. 512 In contrast, the longer lifespans, biosecurity measures, and/or reduced environmental 513 persistence of IAV in other host species may limit opportunities for reassortment. 514 Finally, lower observed reassortment rates in other species with short life spans, like 515 swine, may be a consequence of reduced number of subtypes that co-circulate within 516 these species, providing fewer opportunities for reassortment to occur. 517 All H3Nx viruses circulating in swine reflect interspecies transmission events 518 from humans, with subsequent evolution occurring within established, enzootic swine 519 virus populations. However, these patterns vary substantially between lineages 520 circulating in North America and Europe. H3N2 is geographically restricted in Europe, 521 with widespread circulation primarily reported in Germany, Italy and the Netherlands 522 (Brown 2013; Simon et al. 2014). The European swine lineage (H3 1970.1) is most 523 similar to human-seasonal viruses detected in the 1970s and has circulated in 524 European swine since (Vincent et al. 2020). In contrast, North American swine lineages 525 are predominated by H3N2 viruses that have been seeded by recurrent human-to-swine 526 spillover events occurring multiple times since the late 1990s (Nelson and Vincent 2015; 527 Vincent et al. 2020), that then diversified. We find that swine influenza viruses exhibit 528 the unique combination of high rates of reassortment and directional positive selection 529 on HA and NA, a combination not observed in any other host group. We also find an 530 association between reassortment and viral persistence and lineage survival in swine, 531 pointing to a fitness advantage conferred by reassortment. Prior work has found that 532 influenza A viruses are detected in in farmed pigs year-round, and that multiple 533 cocirculating lineages can be detected within individual farms (Kyriakis et al. 2017; van 534 der Vries et al. 2025), potentially facilitating high co-infection and reassortment rates. 535 Swine housing density has been shown to correlate with H3 prevalence, which could 536 enhance opportunities for coinfection and reassortment (Maes et al. 2000; Poljak et al. 537 2008). Lifespans of swine vary across production types, with sows living up to 2-3 years 538 on average, and new susceptibles entering the population yearly, sustaining viral 539 prevalence (Pitzer et al. 2016; Brown 2000; EFSA Panel on Animal Health and Welfare 540 (AHAW) et al. 2022). Commercial swine are also moved frequently, enabling 541 transmission between herds, resulting in a heterogenous immune landscape at the 542 population level (Brown 2000). These patterns, in which susceptible hosts are 543 concentrated, moved frequently, and periodically turn over, may increase the probability 544 of reassortment, and of beneficial reassortants sustaining onward transmission 545 (Thomas et al. 2024; Neveau et al. 2022; Marshall et al. 2013). Alternatively, it is also 546 possible that the concentration of susceptible hosts could amplify the transmission of 547 reassortant lineages, even in the absence of a fitness benefit, as observed in co-housed 548 pigs where mutations persisted despite fitness costs (Murcia et al. 2012). Together, this 549 combination of swine farming practices and swine immunity may facilitate the high 550 reassortment and adaptive rates that we observe. 551 Our data recapitulate the frequent transmission of viral lineages between humans 552 and swine, a long-existing phenomenon. We find enrichment of reassortment on 553 human-to-swine host switches, which further supports the hypothesis that swine are 554 more often recipients of human IAV lineages rather than sources (Nelson and Worobey 555 2018; Nelson and Vincent 2015; Neumann et al. 2009). While we cannot determine 556 whether reassortment acts as the primary driver of the switch or a stabilizing force 557 following the jump, its enrichment on human-to-swine switches suggests that 558 reassortment is associated with host switches between these species, potentially 559 facilitated by year-round IAV circulation in swine that is permissive to reassorting with 560 human-adapted viruses. These findings suggest swine as unique progenitors of viral 561 diversity, and underscore the importance of continued surveillance and risk assessment 562 of swine IAV populations and at the human-swine interface. 563 This work reveals fundamental differences in how viral mixing is constrained 564 between avian and swine hosts. In swine, NA and PB1 reassortments were significantly 565 underrepresented relative to expectations under random reassortment, suggesting that 566 swine reassortant viruses face HA-NA-PB1 balance constraints. This pattern matches 567 the known evolutionary history of swine viruses, where North American swine H3N2 568 viruses received their HA, PB1, and NA segments from human H3N2 viruses in the late 569 1990s (Zhou et al. 1999). This human-derived constellation dates back to the 1968 570 pandemic strain which featured an HA-PB1 of avian origin paired with an NA from the 571 H2N2 human lineage (Lindstrom et al. 2004). While H3N2 remains less prevalent in 572 European swine, the H3N2 swine lineage possessed an HA-NA pairing from a human- 573 adapted virus with internal segments donated from Eurasian avian H1N1 viruses (Zell et 574 al. 2013; Castrucci et al. 1993). Phylogenetic analyses showing HA-NA diversity are 575 linked in swine viruses further confirm this pattern (Zeller et al. 2021). These pairings 576 may reflect an inherent functional constraint for circulation in mammalian hosts or 577 epistatic interactions that co-evolved in these viruses following their introduction into the 578 swine population (Nelson et al. 2014; Lowen 2017). In contrast, avian viruses showed 579 an excess of NA reassortments, with novel NA subtypes introduced at rates exceeding 580 those expected from observed viral circulation alone. Consistent with our findings that 581 reassortment is largely neutral in avian hosts, this pattern may reflect incomplete 582 surveillance of avian or environmental reservoirs. The NA subtype frequency 583 calculations underlying our null model depend on observed sequence data, and 584 substantial unsampled viral diversity could bias our estimates of expected reassortment 585 probabilities. Understanding this dynamic will require more comprehensive field 586 surveillance to better capture the true extent of viral diversity in natural avian reservoirs. 587 Interestingly, there seems to be a lack of NA cross-reactivity in mallards, indicating 588 complete sterilizing immunity against different NA subtypes is never established 589 (Latorre-Margalef et al. 2013). This may make the reservoir host highly permissive to 590 multi-subtype coinfection, which could directly facilitate the cocirculation of diverse NA 591 subtypes. 592 The past four human influenza pandemics have emerged through reassortment 593 events involving cross-species transmission, underscoring the importance of 594 understanding host-specific reassortment and evolutionary dynamics (Taubenberger 595 and Kash 2010). Here, we show that the H3Nx system provides a unique and 596 underappreciated opportunity to observe how influenza viruses operate under different 597 evolutionary pressures, highlighting its remarkable flexibility in evolutionary strategy 598 depending on host type. These findings reveal that influenza evolution is fundamentally 599 governed by host-specific ecology, epidemiology, and biology. Influenza relies primarily 600 on reassortment in its avian reservoir but shifts towards accumulating adaptive 601 substitutions upon endemic circulation in mammalian hosts, with the degree of this shift 602 varying by host species. Our findings support swine as key hosts to surveil, particularly 603 at the human-swine interface, where human-to-swine host switches frequently involve 604 reassortment. Reassorted viruses in swine have a tendency to persist, which provides 605 the opportunity to accrue mutations in the surface proteins that facilitate establishment 606 in the new pigs, and drift from the original seeding human viruses (Vijaykrishna et al. 607 2011; Rajao et al. 2022). In avian hosts, reassortment instead generates and sustains 608 extensive subtype diversity that co-circulates globally, where reassortant lineages are 609 largely short-lived with limited fitness effects. However, this large and dynamic reservoir 610 of circulating diversity provides repeated opportunities for mixing with strains with 611 pandemic potential, including highly pathogenic influenza viruses (Damodaran et al. 612 2026). Overall, this work emphasizes the need to prioritize surveillance and biosecurity 613 efforts, particularly in swine populations where reassortants may confer fitness benefits 614 for mammalian adaptation and in avian populations where reassortment generates viral 615 diversity capable of seeding future pandemic strains. 616 617 Methods 618 619 Data and phylogenetics 620 621 Initial H3Nx dataset curation and filtering 622 A comprehensive dataset of H3Nx genome sequences sampled from all non- 623 human hosts, regions, and across time was downloaded from NCBI virus and GISAID 624 (Khare et al. 2021) databases yielding 13,295 viral sequences sampled from birds, 625 horses, camels, donkeys, dogs, seals, mink, swine, and cats collected between 1963 626 and 2024. Three human spillover H3N8 viruses were also included from GISAID 627 (A/Guangdong/ZS-23SF005/2023, A/Henan/4-10/2022, A/Changsha/1000/2022). 628 629 To mitigate sampling biases across species and time periods, the following 630 workflow using Snakemake v9.9.0 and Nextstrain v8.5.4 was followed. Sequences were 631 filtered down to 5,023 full non-human genome based on these filtering parameters: (1) 632 minimum sequence lengths per segment (HA: 1,600 bp; NA: 1,270 bp; NP: 1,400 bp; 633 PA: 2,000 bp; PB1: 2,100 bp; PB2: 2,100 bp; MP: 900 bp; NS: 800 bp); (2) exclusion of 634 sequences with collection dates prior to 1960; (3) removal of sequences with unknown 635 country or region metadata; (4) removal of invalid characters. Sequences were grouped 636 by year, country, host species, and subtype, and within each group the data were 637 subsampled to 30 sequences per group. To ensure that all eight segments were 638 represented by the same set of strains, subsampling was performed first for the HA 639 segment, and then this filtering was applied to the other seven segments. Human 640 seasonal H3N2 viruses collected between 1968-2023 were downloaded separately from 641 NCBI (N=29,114) and filtered down to 1,078 sequences grouped by year and country. 642 These sequences were then added to the H3Nx dataset to make global phylogenies, 643 bringing the total filtered dataset size to 6,104 full genomes (5,023 non-human H3Nx 644 genomes, 3 spillover human H3N8 genomes, 1078 human seasonal H3N2 genomes). 645 646 Phylogenetic reconstruction of H3Nx evolution 647 Filtered sequences for each segment were aligned to an H3N8 mallard sequence 648 (A/blue-winged teal/Alberta/221/1978) using MAFFT v7.475. Maximum-likelihood 649 phylogenetic divergence trees were inferred for each segment using IQ-TREE v2.4.0 650 with a GTR substitution model. Time-resolved trees were then constructed using 651 TreeTime v0.11.4 using a coalescent time scale, where dates are inferred at internal 652 nodes. Temporal outliers were identified using a clock filter (interquartile distance 653 threshold: 4) and excluded. Nucleotide and amino acid sequences were reconstructed 654 at internal nodes using TreeTime. Ancestral state reconstruction was performed for host 655 type, bird order, country, region, and subtype via a discrete trait analysis. Segment trees 656 are shown in Figure 1A and Supplementary Figure 18. 657 658 659 Host subtree construction 660 Currently circulating host groups (North American avian, Eurasian avian, swine, 661 human H3N2 seasonal, equine, and canine H3N2 viruses) were parsed from the global 662 H3Nx phylogeny based on their clustering within the HA phylogenetic tree. Here, we 663 collapsed all geographic and phylogenetically distinct swine H3 lineages into a single 664 “swine” clade. This approach enables a comparative analysis of viral evolution across 665 mammalian and avian hosts at a global scale, but potentially obscures within-clade 666 variability among swine lineages. 667 668 Each host-specific dataset was realigned using MAFFT with host-specific 669 references for each segment. Maximum-likelihood phylogenetic divergence trees were 670 inferred for each segment using IQtree with a GTR substitution model. Clock rates were 671 estimated using TreeTime’s clock model function in python. These alignments and trees 672 were used for the subsequent mutation and reassortment analyses. 673 674 Adaptive substitution rates 675 676 Modified McDonald-Krietmand test (Bhatt method) 677 For each host group, we performed a modified McDonald-Krietman (MK) test on 678 the HA, NA, and PB1 genes(Kistler and Bedford 2023; Bhatt et al. 2011). Here, PB1 679 was chosen as a conserved reference gene, since as a core polymerase subunit under 680 strong purifying selection and not directly targeted by host immunity, it is not expected 681 to accumulate adaptive substitutions to the same degree as HA and NA (Kistler and 682 Bedford 2023; Bhatt et al. 2011). This test calculates adaptive substitution rates by 683 measuring the accumulation of nonneutral substitutions above a neutral expectation and 684 was specifically developed for temporally-sampled viral genomes. Aligned viral 685 sequences were binned into overlapping time windows of at least 3 years, with a 686 minimum of 3 sequences per time window. The test uses an updating outgroup, which 687 is first taken as the consensus of sequences in the earliest time window. This outgroup 688 was updated for future time windows when fixations at each site occur, to account for 689 recurring mutations. This outgroup is then compared to the aligned sequences in each 690 subsequent time window (the ingroup). Across each site, genetic differences between 691 the outgroup and ingroup are noted, and the number of adaptive substitutions is 692 calculated as the sum of nonsynonymous fixations and near fixations that exceed the 693 neutral expectation. Reflecting influenza’s high mutation rate, the neutral expectation is 694 measured by counting: 1) all mid-frequency (15-75%) polymorphisms, 2) synonymous 695 near fixations (>75%), and 3) synonymous fixations. Adaptive substitutions per codon 696 are then plotted as a point estimate for each time window, where the slope of the linear 697 regression fitting these estimates is calculated as the rate of adaptation (Supplementary 698 Figure 3, Figure 1B). Adaptive substitutions rates are measured as adaptive 699 substitutions per codon per year x 10–3. 700 701 The alignments for the host-specific subtrees were further filtered to include only 702 competing viruses, defined as viruses co-circulating within the same host and 703 geographic region. For avian viruses, competing viruses for the NA segment were 704 additionally required to belong to the same subtype. This filtering strategy accounts for 705 this test's sensitivity to population structure and ensures that adaptive substitution rates 706 reflect selection pressures within epidemiologically relevant viral populations. 707 Specifically, the swine host group was subdivided into separate European and North 708 American populations and North American and Eurasian avian NA lineages were 709 stratified by subtype (N2 and N8). These population-level groupings merge independent 710 introductions across these regions to enable larger scale comparisons of viral evolution 711 among these hosts. Reassortant strains identified through TreeSort were excluded to 712 prevent sequence differences arising from reassortment events from biasing selection 713 estimates. 714 715 Code for this test was adapted from github.com/blab/adaptive- 716 evolution/blob/master/adaptive-evolution-analysis. 717 718 TreeSort pipeline and reassortment rates 719 720 Overview of TreeSort and TreeSort Pipeline 721 TreeSort implements a phylogenetic incongruence approach to identify reassortment by 722 designating a single segment as a reference and systematically comparing the 723 evolutionary histories of the remaining segments against it. In the absence of 724 reassortment, these phylogenies are expected to be largely congruent where deviations 725 from this expectation are interpreted as evidence of reassortment. To do this, TreeSort 726 estimates segment-specific molecular clock rates and uses these to define probability 727 thresholds for detecting reassortment events along branches of the reference 728 phylogeny. This method produces a single annotated tree in which inferred 729 reassortment events and the reassorting segments are mapped. Here, a reassortment 730 event is defined as 1 or more segments reassorted relative to HA on a single branch. 731 732 Although TreeSort is highly informative, a single run yields only one annotated 733 phylogeny, thereby limiting inference to point estimates for different parameters (e.g. 734 reassortment rates). To address this limitation, we developed a pipeline to run TreeSort 735 in replicate (available at github/moncla-lab/treesort-pipeline). This approach allows for 736 more robust reassortment inference by incorporating statistical uncertainty in tree 737 building and TreeSort’s respective probability estimations. 738 739 The pipeline was generated using a Snakemake workflow and proceeds as follows: 740 TreeSort is run once to generate a binarized backbone tree annotated with 741 reassortment events, which serves as the fixed reference phylogeny for all subsequent 742 replicate runs. In each replicate, new divergence trees are generated for the challenge 743 (non-reference) segments while the --no-collapse flag is applied to ensure all TreeSort 744 outputs retain identical topology to the backbone tree. For these analyses, 1000 745 replicates were performed. Reassortment rates (reassortments per lineage per year) 746 are calculated for each replicate and compiled in a summary log file. 747 748 After running TreeSort in replicate, a summary JSON file is generated that 749 records the reassortment support value for each node, defined as the proportion of runs 750 in which a reassortment event was inferred. Support values are calculated 751 independently for each segment, such that a node involved in a multi-segment 752 reassortment event (e.g., PB2, PB1, and PA jointly) may have differing support values 753 across the reassorting segments. A summary reassortment tree is also generated, 754 annotated only with high-support reassortments (reassortments inferred in ≥95% of the 755 runs). The resulting summary reassortment tree is able to be visualized via the 756 Nextstrain Auspice interface using the cladeset-mapping tool (described in 757 github/moncla-lab/treesort-cladeset-mapping), enabling interactive exploration of 758 reassortment event inferences and their support across the reference phylogeny. 759 760 When TreeSort cannot confidently assign a reassortment event to a specific 761 branch, it labels both child branches with an uncertain tag (e.g., ?PB2 to indicate 762 ambiguity regarding the PB2 segment, see https://github.com/flu-crew/TreeSort for 763 details). In such cases, one of the two child branches was called at random to be 764 reassorted. 765 766 Custom Python scripts (v3.12.11) are used throughout the pipeline: (1) 767 converting TreeSort replicate outputs into Newick trees, (2) calculating reassortment 768 rates, (3) recording individual reassortment events and resolving uncertain calls across 769 replicates into individual JSON files, (4) aggregating results from individual JSONs into 770 a summary file, and (5) mapping consensus reassortments back to a nonbinarized tree 771 for Nextstrain visualization. 772 773 Reassortment rates and North American avian subsampling 774 775 As part of the TreeSort pipeline, the reassortment rate is calculated at each 776 replicate as the total number of reassortments divided by total tree length, and then 777 scaled by the evolutionary rate of the reference segment (see (Markin et al. 2025) for 778 details). For each host group, we calculated the mean rate and standard deviation 779 across replicates (Figure 2A). Reassortment rates are in units of reassortments per year 780 per lineage, enabling direct comparison across host groups. 781 782 To assess the sensitivity of reassortment inferences to dataset size, we 783 generated serially subsampled datasets of North American avian H3Nx viruses, ranging 784 from N=400 to N=1,100 sequences in 100-sequence increments. Sequences were 785 subsampled by year and region. Regions were defined as sequences sampled from 786 Africa, Europe, North America, China, South Asia, Japan/Korea, Oceania, South 787 America, and West Asia. For each dataset size, 5 independent subsampling trials were 788 performed, and reassortment rates were calculated using the full TreeSort replicate 789 pipeline described above (Supplementary Figure 6). 790 791 Null distribution - shuffled reassortment summary tree 792 To assess whether observed reassortment dynamics were ever consistent with 793 reassortment under a random model, we generated a null distribution of reassortment 794 for each host group and the global H3Nx datasets. For each host, we randomly 795 redistributed all high-support reassortment events uniformly across the branches of the 796 tree, such that each branch had an equal probability of being assigned a reassortment 797 event regardless of branch length or tree structure. This procedure was repeated 1000 798 times for each group, generating a distribution of 1000 trees in which the number of 799 reassortment events and tree topology were conserved, but placement was 800 randomized. 801 802 Flyway analysis for North American avian viruses 803 To determine whether reassortment events occurred more frequently in specific 804 flyways than expected by chance, we performed a discrete trait analysis on the North 805 American avian dataset to infer flyway annotations at internal nodes. Flyway 806 assignments for terminal taxa were based on sampling location according to US Fish 807 and Wildlife Service Administrative Flyway classifications (U.S. Fish & Wildlife Service 808 2023). We then calculated the frequency of each flyway among all sequences as a 809 proportion of the total, which served as the null expectation probability for each flyway. 810 811 For each flyway, we counted the number of reassorted nodes assigned to that 812 flyway as a proportion of the total number of reassorted nodes. We then used a two- 813 sided binomial test to determine whether the observed frequency of reassortment 814 events in each flyway differed significantly from the expected frequency based on the 815 overall strain distribution. Multiple hypothesis testing was corrected using the Bonferroni 816 method across all flyways, with significance threshold set at α = 0.05. We repeated this 817 flyway analysis on the null distribution of trees described earlier. 818 819 Segment reassortment dynamics 820 821 Null distribution for segment-specific reassortments 822 To assess whether gene segments reassort nonrandomly, we generated a null 823 model based on the observed reassortment events. The distribution of event sizes (the 824 number of segments involved per reassortment event) was first quantified from the 825 observed data. Simulations were then performed in which the observed distribution of 826 event sizes was maintained, while the identities of reassorting segments were randomly 827 assigned by sampling without replacement. This procedure was repeated 1,000 times to 828 generate a null distribution of randomly reassorting segments. Observed reassortment 829 frequencies were then compared to the simulated null distributions using empirical two- 830 sided p-values, accounting for small sample size using a Monte Carlo correction and 831 multiple testing using Bonferroni method (α = 0.05). P-values were calculated as p = (r + 832 1) / (n + 1), where r represents the count of simulated values with absolute deviation 833 from the mean as extreme or more extreme than the observed value in either direction, 834 and n equals the total number of simulations (n = 1,000). 835 836 Calculating within- and between-reassortment probabilities for avian host groups 837 To determine whether observed within- and between-subtype reassortment 838 counts deviated from expectations based on subtype circulation, expected reassortment 839 probabilities were calculated. For both North American and Eurasian avian datasets 840 separately, the frequency of each NA subtype was determined by counting the number 841 of strains belonging to each subtype and converting these counts into proportions of the 842 total datasets. Expected within-subtype reassortment probabilities were estimated by 843 summing the squared frequencies of each NA subtype, corresponding to the probability 844 of randomly selecting two viruses belonging to the same subtype. Expected between- 845 subtype probabilities were calculated as the complement of the within-subtype 846 probability (1-P(within)), representing the probability of selecting two viruses of different 847 subtypes. In North American avian viruses, the expected within-subtype reassortment 848 probability is 38%, and the expected for between-subtype is 62%. For Eurasian viruses, 849 the expected within-subtype reassortment probability is 41%, and the expected for 850 between-subtype is 59% (Supplementary Figure 10). These probabilities were 851 compared to observed counts under a binomial model. 852 853 Calculating within- and between-subtype divergence thresholds 854 To distinguish within- and between-subtype NA reassortment events, pairwise 855 nucleotide divergence was calculated among all NA sequences within the North 856 American and Eurasian avian datasets. Low-quality sequences (those containing fewer 857 than 80% valid nucleotide sites) were filtered out. Pairwise divergence was then 858 calculated for all possible sequence pairs within each dataset, labeled within-subtype 859 when both sequences belonged to the same NA subtype and between-subtype when 860 sequences belonged to different NA subtypes. Within- and between-subtype divergence 861 thresholds were determined from these empirical distributions of pairwise NA 862 sequences divergences. The within-subtype divergence threshold was defined as the 863 maximum pairwise divergence observed among sequences belonging to the same NA 864 subtype, whereas the between-subtype threshold was defined as the minimum pairwise 865 divergence observed among sequences belonging to different NA subtypes. The within- 866 subtype threshold was used as the upper bound for classifying NA reassortments as 867 within-subtype events. 868 869 Classification of NA reassortments as within- or between-subtype events 870 TreeSort calculates how diverged a reassorting segment is relative to the 871 parental segments, measured as the number of nucleotide differences. For example, an 872 annotation of PB2(136) denotes acquisition of a PB2 segment that differed by 136 873 nucleotides from the parental PB2 segment (see https://github.com/flu-crew/TreeSort for 874 details). Using only high-support NA reassortments, the inferred NA divergence values 875 inferred by TreeSort were compared to the divergence thresholds calculated above. 876 Reassortment events with TreeSort-inferred NA divergence values less than or equal to 877 the within-subtype upper bound were classified as within-subtype reassortments. 878 Reassortment events exceeding this bound were classified as between-subtype 879 reassortments. Among 173 NA reassortment events identified in North American avian 880 viruses, 35 involved reassortment of NAs from the same subtype, while 138 involved 881 introduction of NA segments from different subtypes. Among 148 NA reassortment 882 events identified in Eurasian avian viruses, 39 involved reassortment of NAs from the 883 same subtype, while 109 involved introduction of NA segments from different subtypes. 884 885 Testing if novel NAs are brought in more than expected by chance 886 To test whether within- and between-subtype NA reassortment events deviated 887 from expectations under a model of random reassortment, observed counts were tested 888 under a binomial model. The expected within- and between-subtype reassortment 889 probabilities derived from circulation frequencies were used as success probabilities. 890 The observed number of "successes" was taken as the number of within- or between- 891 subtype reassortments classified above. The binomial distributions were then evaluated 892 over all possible outcomes to generate the expected probability mass function for 893 within- and between-subtype reassortments under a null model. The p-value was 894 computed as the cumulative probability of observing the actual number of within- and 895 between-subtype events or fewer under the null model. The binom function from the 896 scipy.stats library was used for these tests. 897 898 Binomial sensitivity analysis 899 To evaluate whether potential detection bias against more highly divergent NA 900 reassortment events could affect inference of within- and between-subtype 901 reassortment, we performed a binomial model sensitivity analysis. First, we estimated 902 the proportion of within-subtype NA sequence pairs that fell below the minimum 903 divergence observed among NA reassortment events inferred by TreeSort. This 904 minimum divergence threshold was taken as an empirical lower bound for detectable 905 reassortment events, and the fraction of within-subtype sequences pairs below this 906 threshold was used as an estimate of potential underdetection. 907 908 To assess the sensitivity of our findings to this underdetection, we then 909 systematically increased fractions of observed within-subtype while keeping the total 910 number of events constant. Across a range of scenarios (1-12% additional within- 911 subtype events), binomial tests were recomputed using the same null probabilities 912 derived from cocirculation. For each scenario, we evaluated whether the observed 913 excess of between-subtype remained statistically significant as within-subtype events 914 increased. We identified the maximum proportion of within-subtype underdetection that 915 could be tolerated before statistical significance was lost (α = 0.05). Here, the 916 significance of our findings was robust up to 11.6% underdetection in North American 917 and 7.4% in Eurasian avian viruses (Supplementary Figure 13). 918 919 Fitness effects of reassortment 920 921 Reassorted persistence analysis 922 To assess whether reassortment confers a fitness advantage within avian and 923 swine host groups, we quantified reassortant lineage persistence as proxy for viral 924 fitness. For each high-support reassorted internal node inferred by the TreeSort 925 pipeline, we calculated the maximum distance to either the next downstream 926 reassortment event or a terminal leaf node. Branch lengths were scaled to time in years 927 by dividing by the host-specific molecular clock rate. 928 929 We compared persistence times of reassorted lineages to those for 930 nonreassortant lineages. Nonreassortant lineage persistence was measured as the 931 maximum distance from a non-reassorted node to the next terminal leaf or reassortment 932 event. Note that this calculation does not account for whether the nonreassortant 933 lineage descends from a recent reassortment event. Persistence times were binned into 934 annual intervals to create discrete persistence categories (0–1 years, 1–2 years, etc.). 935 For each host group and lineage type (reassorted vs. non-reassorted), we calculated 936 the proportion of lineages falling into each persistence bin relative to the total number of 937 lineages (Supplementary Figure 16). A similar analysis was performed to test whether 938 observed reassortant persistences differed from random expectation. Using the 939 distribution of null reassortment trees described earlier, we calculated mean null 940 reassortant lineage persistences and the 95% confidence intervals for each persistence 941 bin. 942 943 Long-term fitness effects of reassortment 944 To determine whether reassortment is associated with the long-term survival of 945 viral lineages, we assessed the likelihood that reassorted nodes produce descendants 946 surviving x years into the future (x = 0.5, 2, 6, and 10 years post-reassortment). We only 947 evaluated internal nodes with an estimated age at least equal to the specified time 948 interval. For each qualifying node, we calculated the maximum temporal distance from 949 the node to its most recent descendant leaf. Nodes were classified as "fit" if they 950 produced descendants reaching or extending beyond the specified time threshold, or 951 "unfit" if all descendants died out before the threshold. 952 953 For each time interval and host group (Swine, Eurasian, and North American 954 avian), a 2x2 contingency table was constructed crossing lineage type (reassorted vs 955 non-reassorted) and survival outcome (fit vs unfit). We performed a two-sided Fisher’s 956 exact test using the fishers_exact function from the scipy.stats library. Odds ratio and 957 95% confidence intervals were calculated from each contingency table confidence 958 intervals were computed using the log-transformed odds ratio function from the math 959 library. 960 961 We also performed these Fisher’s exact tests on the distribution of null, shuffled 962 reassortment trees described earlier. For each host group and time interval, we 963 identified how often null trees yielded statistically significant results (α = 0.05) and 964 compared the distribution of odds ratios across the null trees to the actual odds ratios 965 (Supplementary Figure 17). 966 967 Host-switching analysis 968 969 Reassortment enrichment on host-switched branches 970 We identified host switches in the global H3Nx phylogenetic tree by comparing 971 host assignments between parent and child nodes using inferences from the discrete 972 trait analysis. If a parent host assignment differed from its child node, then this was 973 considered a host-switched branch. For each internal node, we extracted its 974 reassortment status and classified all host switches into four categories: reassorted host 975 switches, non-reassorted host switches, reassorted non-switches, and non-reassorted 976 non-switches. To test whether reassortment events were statistically enriched on host- 977 switching branches, we constructed a 2×2 contingency table and performed a two-sided 978 Fisher's exact test. Odds ratios were computed using the fishers_exact function from 979 the scipy.stats library and 95% confidence intervals were calculated using the log- 980 transformed odds ratio function from the math library. Here, an odds ratio > 1 and p-val 981 < 0.05 indicates reassortment enrichment on host-switched branches. This analysis was 982 also conducted separately for the North American and Eurasian avian datasets but for 983 transitions between avian orders (e.g. galliform to anseriform) where no significance 984 was found. 985 986 To determine whether the observed association between reassortment and host- 987 switching varied by transition type, we compared observed reassorted transitions in the 988 global H3Nx tree to transitions in the distribution of null trees described earlier. For each 989 null tree, we identified all host switches occurring on reassorted internal nodes and 990 stratified them into 3 categories: avian-to-mammal, mammal-to-mammal, and mammal- 991 to-avian transitions. We aggregated counts across the 1,000 null trees to generate null 992 distribution for each host transition. Empirical p-values were calculated as p = (r + 1) / (n 993 + 1) where r represents the number of null trees with a host transition count as extreme 994 or more extreme than the observed count and n equals the total number of null trees 995 (n=1000) (Figure 5B). 996 997 Code availability and reproducibility 998 All code developed and used in this project are available in the 999 github.com/moncla-lab/h3nx-paper git repo. The TreeSort pipeline is available with 1000 documentation and example data at github.com/moncla-lab/treesort-pipeline. A public, 1001 interactive version of the H3Nx tree is available from the Moncla lab Nextstrain groups 1002 page at: nextstrain.org/groups/moncla-lab/h3nx/ha. All data that were used in this 1003 analysis were sourced from public databases. The acknowledgement table for GISAID 1004 isolates used in this analysis is provided in Supplementary Table 2, which can also be 1005 found at GitHub (https://github.com/moncla-lab/h3nx-paper). 1006 1007 Acknowledgments 1008 1009 We gratefully acknowledge all data contributors, i.e., the Authors and their 1010 Originating laboratories responsible for obtaining the specimens, and their Submitting 1011 laboratories for generating the genetic sequence and metadata and sharing via the 1012 GISAID Initiative, on which this research is based. 1013 1014 Support for this project was provided by the Pew Charitable Trusts, the Margaret 1015 Q. Landenberger Research Foundation, the National Institute of Allergy and Infectious 1016 Diseases, National Institutes of Health, Department of Health and Human Services 1017 (contract numbers 75N93021C00015 and 75N93021C00016), and the United States 1018 Department of Agriculture, Agricultural Research Service (ARS project number 5030- 1019 32000-231-000D). USDA is an equal opportunity provider and employer. 1020 1021 References 1022 Baker, Steven F., Aitor Nogales, Courtney Finch, et al. 2014. “Influenza A and B Virus Intertypic 1023 Reassortment through Compatible Viral Packaging Signals.” Journal of Virology 88 (18): 1024 10778–10791. 1025 Bhatt, Samir, Edward C. Holmes, and Oliver G. Pybus. 2011. “The Genomic Rate of Molecular 1026 Adaptation of the Human Influenza A Virus.” Molecular Biology and Evolution 28 (9): 2443– 1027 2451. 1028 Brown, Ian H. 2013. “History and Epidemiology of Swine Influenza in Europe.” Current Topics in 1029 Microbiology and Immunology (Berlin, Heidelberg), Current topics in microbiology and 1030 immunology, vol. 370: 133–146. 1031 Brown, I. H. 2000. “The Epidemiology and Evolution of Influenza Viruses in Pigs.” Veterinary 1032 Microbiology 74 (1-2): 29–46. 1033 Castrucci, M. R., I. Donatelli, L. Sidoli, G. Barigazzi, Y. Kawaoka, and R. G. Webster. 1993. 1034 “Genetic Reassortment between Avian and Human Influenza A Viruses in Italian Pigs.” 1035 Virology 193 (1): 503–506. 1036 Crawford, P. C., Edward J. Dubovi, William L. Castleman, et al. 2005. “Transmission of Equine 1037 Influenza Virus to Dogs.” Science (New York, N.Y.) 310 (5747): 482–485. 1038 Damodaran, Lambodhar, Joseph A. Lewnard, Gregg S. Davis, Sara Y. Tartof, Louise H. 1039 Moncla, and Nicola F. Müller. 2026. “Frequent Seasonal Reassortment between High and 1040 Low Path Viruses Drives the Diversification of Influenza A/H5N1.” In bioRxiv. BioRxiv, April 1041 18. https://doi.org/10.64898/2026.04.17.719307. 1042 Diskin, Elena R., Kimberly Friedman, Scott Krauss, et al. 2020. “Subtype Diversity of Influenza 1043 A Virus in North American Waterfowl: A Multidecade Study.” Journal of Virology 94 (11): 1044 e02022–19. 1045 Dugan, Vivien G., Rubing Chen, David J. Spiro, et al. 2008. “The Evolutionary Genetics and 1046 Emergence of Avian Influenza Viruses in Wild Birds.” PLoS Pathogens 4 (5): e1000076. 1047 EFSA Panel on Animal Health and Welfare (AHAW), Søren Saxmose Nielsen, Julio Alvarez, et 1048 al. 2022. “Welfare of Pigs on Farm.” EFSA Journal 20 (8): e07421. 1049 Essere, Boris, Matthieu Yver, Cyrille Gavazzi, et al. 2013. “Critical Role of Segment-Specific 1050 Packaging Signals in Genetic Reassortment of Influenza A Viruses.” Proceedings of the 1051 National Academy of Sciences of the United States of America 110 (40): E3840–8. 1052 Ferguson, Neil M., Alison P. Galvani, and Robin M. Bush. 2003. “Ecological and Immunological 1053 Determinants of Influenza Evolution.” Nature 422 (6930): 428–433. 1054 Furuse, Yuki, Akira Suzuki, and Hitoshi Oshitani. 2010. “Reassortment between Swine Influenza 1055 A Viruses Increased Their Adaptation to Humans in Pandemic H1N1/09.” Infection, 1056 Genetics and Evolution: Journal of Molecular Epidemiology and Evolutionary Genetics in 1057 Infectious Diseases 10 (4): 569–574. 1058 Ganti, Ketaki, Anish Bagga, Juliana DaSilva, et al. 2021. “Avian Influenza A Viruses Reassort 1059 and Diversify Differently in Mallards and Mammals.” Viruses 13 (3): 509. 1060 Ince, William L., Aissatou Gueye-Mbaye, Jack R. Bennink, and Jonathan W. Yewdell. 2013. 1061 “Reassortment Complements Spontaneous Mutation in Influenza A Virus NP and M1 1062 Genes to Accelerate Adaptation to a New Host.” Journal of Virology 87 (8): 4330–4338. 1063 Janzen, Garrett M., Blake T. Inderski, Jennifer Chang, et al. 2025. “Sources and Sinks of 1064 Influenza A Virus Genomic Diversity in Swine from 2009 to 2022 in the United States.” 1065 Journal of Virology 99 (9): e0054125. 1066 Khare, Shruti, Céline Gurry, Lucas Freitas, et al. 2021. “GISAID’s Role in Pandemic Response.” 1067 China CDC Weekly 3 (49): 1049–1051. 1068 Khiabanian, Hossein, Vladimir Trifonov, and Raul Rabadan. 2009. “Reassortment Patterns in 1069 Swine Influenza Viruses.” PloS One 4 (10): e7366. 1070 Kistler, Kathryn E., and Trevor Bedford. 2023. “An Atlas of Continuous Adaptive Evolution in 1071 Endemic Human Viruses.” Cell Host & Microbe 31 (11): 1898–1909.e3. 1072 Kuchinski, Kevin S., John Tyson, Tracy Lee, et al. 2025. “Detection of a Reassortant Swine- and 1073 Human-Origin H3N2 Influenza A Virus in Farmed Mink in British Columbia, Canada.” 1074 Zoonoses and Public Health 72 (3): 293–300. 1075 Kyriakis, Constantinos S., Ming Zhang, Stefan Wolf, et al. 2017. “Molecular Epidemiology of 1076 Swine Influenza A Viruses in the Southeastern United States, Highlights Regional 1077 Differences in Circulating Strains.” Veterinary Microbiology 211 (November): 174–179. 1078 Latorre-Margalef, Neus, Vladimir Grosbois, John Wahlgren, et al. 2013. “Heterosubtypic 1079 Immunity to Influenza A Virus Infections in Mallards May Explain Existence of Multiple Virus 1080 Subtypes.” PLoS Pathogens 9 (6): e1003443. 1081 Le Sage, Valerie, Michelle N. Vu, Maria A. Maltepes, et al. 2026. “Fatal Human H3N8 Influenza 1082 Virus Has a Moderate Pandemic Risk.” PLoS Pathogens 22 (3): e1013586. 1083 Lindstrom, Stephen E., Nancy J. Cox, and Alexander Klimov. 2004. “Genetic Analysis of Human 1084 H2N2 and Early H3N2 Influenza Viruses, 1957-1972: Evidence for Genetic Divergence and 1085 Multiple Reassortment Events.” Virology 328 (1): 101–119. 1086 Liu, Tongyu, Yiquan Wang, Timothy J. C. Tan, Nicholas C. Wu, and Christopher B. Brooke. 1087 2022. “The Evolutionary Potential of Influenza A Virus Hemagglutinin Is Highly Constrained 1088 by Epistatic Interactions with Neuraminidase.” Cell Host & Microbe 30 (10): 1363–1369.e4. 1089 Lowen, Anice C. 2017. “Constraints, Drivers, and Implications of Influenza A Virus 1090 Reassortment.” Annual Review of Virology 4 (1): 105–121. 1091 Lycett, S. J., G. Baillie, E. Coulter, et al. 2012. “Estimating Reassortment Rates in Co- 1092 Circulating Eurasian Swine Influenza Viruses.” The Journal of General Virology 93 (Pt 11): 1093 2326–2336. 1094 Macken, Catherine A., Richard J. Webby, and William J. Bruno. 2006. “Genotype Turnover by 1095 Reassortment of Replication Complex Genes from Avian Influenza A Virus.” The Journal of 1096 General Virology 87 (Pt 10): 2803–2815. 1097 Ma, Eric J., Nichola J. Hill, Justin Zabilansky, Kyle Yuan, and Jonathan A. Runstadler. 2016. 1098 “Reticulate Evolution Is Favored in Influenza Niche Switching.” Proceedings of the National 1099 Academy of Sciences of the United States of America 113 (19): 5335–5339. 1100 Maes, D., H. Deluyker, M. Verdonck, et al. 2000. “Herd Factors Associated with the 1101 Seroprevalences of Four Major Respiratory Pathogens in Slaughter Pigs from Farrow-to- 1102 Finish Pig Herds.” Veterinary Research 31 (3): 313–327. 1103 Markin, Alexey, Catherine A. Macken, Amy L. Baker, and Tavis K. Anderson. 2025. “Revealing 1104 Reassortment in Influenza A Viruses with TreeSort.” Molecular Biology and Evolution 42 1105 (8): msaf133. 1106 Marshall, Nicolle, Lalita Priyamvada, Zachary Ende, John Steel, and Anice C. Lowen. 2013. 1107 “Influenza Virus Reassortment Occurs with High Frequency in the Absence of Segment 1108 Mismatch.” PLoS Pathogens 9 (6): e1003421. 1109 Ma, Wenjun, Qinfang Liu, Bhupinder Bawa, et al. 2012. “The Neuraminidase and Matrix Genes 1110 of the 2009 Pandemic Influenza H1N1 Virus Cooperate Functionally to Facilitate Efficient 1111 Replication and Transmissibility in Pigs.” The Journal of General Virology 93 (Pt 6): 1261– 1112 1268. 1113 Mehle, Andrew, Vivien G. Dugan, Jeffery K. Taubenberger, and Jennifer A. Doudna. 2012. 1114 “Reassortment and Mutation of the Avian Influenza Virus Polymerase PA Subunit 1115 Overcome Species Barriers.” Journal of Virology 86 (3): 1750–1757. 1116 Mitnaul, L. J., M. N. Matrosovich, M. R. Castrucci, et al. 2000. “Balanced Hemagglutinin and 1117 Neuraminidase Activities Are Critical for Efficient Replication of Influenza A Virus.” Journal 1118 of Virology 74 (13): 6015–6020. 1119 Müller, Nicola F., Ugnė Stolz, Gytis Dudas, Tanja Stadler, and Timothy G. Vaughan. 2020. 1120 “Bayesian Inference of Reassortment Networks Reveals Fitness Benefits of Reassortment 1121 in Human Influenza Viruses.” Proceedings of the National Academy of Sciences of the 1122 United States of America 117 (29): 17104–17111. 1123 Munster, Vincent J., Chantal Baas, Pascal Lexmond, et al. 2007. “Spatial, Temporal, and 1124 Species Variation in Prevalence of Influenza A Viruses in Wild Migratory Birds.” PLoS 1125 Pathogens 3 (5): e61. 1126 Murcia, Pablo R., Joseph Hughes, Patrizia Battista, et al. 2012. “Evolution of an Eurasian Avian- 1127 like Influenza Virus in Naïve and Vaccinated Pigs.” PLoS Pathogens 8 (5): e1002730. 1128 Nelson, Martha I., Cécile Viboud, Lone Simonsen, et al. 2008. “Multiple Reassortment Events in 1129 the Evolutionary History of H1N1 Influenza A Virus since 1918.” PLoS Pathogens 4 (2): 1130 e1000012. 1131 Nelson, Martha I., and Amy L. Vincent. 2015. “Reverse Zoonosis of Influenza to Swine: New 1132 Perspectives on the Human-Animal Interface.” Trends in Microbiology 23 (3): 142–153. 1133 Nelson, Martha I., David E. Wentworth, Marie R. Culhane, et al. 2014. “Introductions and 1134 Evolution of Human-Origin Seasonal Influenza a Viruses in Multinational Swine 1135 Populations.” Journal of Virology 88 (17): 10110–10119. 1136 Nelson, Martha I., and Michael Worobey. 2018. “Origins of the 1918 Pandemic: Revisiting the 1137 Swine ‘Mixing Vessel’ Hypothesis.” American Journal of Epidemiology 187 (12): 2498– 1138 2502. 1139 Neumann, Gabriele, Takeshi Noda, and Yoshihiro Kawaoka. 2009. “Emergence and Pandemic 1140 Potential of Swine-Origin H1N1 Influenza Virus.” Nature 459 (7249): 931–939. 1141 Neveau, Megan N., Michael A. Zeller, Bryan S. Kaplan, et al. 2022. “Genetic and Antigenic 1142 Characterization of an Expanding H3 Influenza A Virus Clade in U.s. Swine Visualized by 1143 Nextstrain.” mSphere 7 (3): e0099421. 1144 Ovadia, Y., D. Fielder, C. Conow, and R. Libeskind-Hadas. 2011. “The Co Phylogeny 1145 Reconstruction Problem Is NP-Complete.” Journal of Computational Biology: A Journal of 1146 Computational Molecular Cell Biology 18 (1): 59–65. 1147 Parrish, Colin R., Pablo R. Murcia, and Edward C. Holmes. 2015. “Influenza Virus Reservoirs 1148 and Intermediate Hosts: Dogs, Horses, and New Possibilities for Influenza Virus Exposure 1149 of Humans.” Journal of Virology 89 (6): 2990–2994. 1150 Petrova, Velislava N., and Colin A. Russell. 2018. “The Evolution of Seasonal Influenza 1151 Viruses.” Nature Reviews. Microbiology 16 (1): 47–60. 1152 Pitzer, Virginia E., Ricardo Aguas, Steven Riley, Willie L. A. Loeffen, James L. N. Wood, and 1153 Bryan T. Grenfell. 2016. “High Turnover Drives Prolonged Persistence of Influenza in 1154 Managed Pig Herds.” Journal of the Royal Society, Interface 13 (119): 20160138. 1155 Poljak, Zvonimir, Catherine E. Dewey, S. Wayne Martin, Jette Christensen, Susy Carman, and 1156 Robert M. Friendship. 2008. “Prevalence of and Risk Factors for Influenza in Southern 1157 Ontario Swine Herds in 2001 and 2003.” Revue Canadienne de Recherche Veterinaire 1158 [Canadian Journal of Veterinary Research] 72 (1): 7–17. 1159 Rajao, Daniela S., Eugenio J. Abente, Joshua D. Powell, et al. 2022. “Changes in the 1160 Hemagglutinin and Internal Gene Segments Were Needed for Human Seasonal H3 1161 Influenza A Virus to Efficiently Infect and Replicate in Swine.” Pathogens 11 (9): 967. 1162 Rajão, Daniela S., Phillip C. Gauger, Tavis K. Anderson, et al. 2015. “Novel Reassortant 1163 Human-like H3N2 and H3N1 Influenza A Viruses Detected in Pigs Are Virulent and 1164 Antigenically Distinct from Swine Viruses Endemic to the United States.” Journal of Virology 1165 89 (22): 11213–11222. 1166 Rambaut, Andrew, Oliver G. Pybus, Martha I. Nelson, Cecile Viboud, Jeffery K. Taubenberger, 1167 and Edward C. Holmes. 2008. “The Genomic and Epidemiological Dynamics of Human 1168 Influenza A Virus.” Nature 453 (7195): 615–619. 1169 Roche, Benjamin, John M. Drake, Justin Brown, David E. Stallknecht, Trevor Bedford, and 1170 Pejman Rohani. 2014. “Adaptive Evolution and Environmental Durability Jointly Structure 1171 Phylodynamic Patterns in Avian Influenza Viruses.” PLoS Biology 12 (8): e1001931. 1172 Sharma, Aditi, Michael A. Zeller, Carine K. Souza, et al. 2022. “Characterization of a 2016-2017 1173 Human Seasonal H3 Influenza A Virus Spillover Now Endemic to U.s. Swine.” mSphere 7 1174 (1): e0080921. 1175 Simon, Gaëlle, Lars E. Larsen, Ralf Dürrwald, et al. 2014. “European Surveillance Network for 1176 Influenza in Pigs: Surveillance Programs, Diagnostic Tools and Swine Influenza Virus 1177 Subtypes Identified in 14 European Countries from 2010 to 2013.” PloS One 9 (12): 1178 e115815. 1179 Smith, Gavin J. D., Dhanasekaran Vijaykrishna, Justin Bahl, et al. 2009. “Origins and 1180 Evolutionary Genomics of the 2009 Swine-Origin H1N1 Influenza A Epidemic.” Nature 459 1181 (7250): 1122–1125. 1182 Song, Daesub, Bokyu Kang, Chulseung Lee, et al. 2008. “Transmission of Avian Influenza Virus 1183 (H3N2) to Dogs.” Emerging Infectious Diseases 14 (5): 741–746. 1184 Song, D. S., D. J. An, H. J. Moon, et al. 2011. “Interspecies Transmission of the Canine 1185 Influenza H3N2 Virus to Domestic Cats in South Korea, 2010.” The Journal of General 1186 Virology 92 (Pt 10): 2350–2355. 1187 Steel, John, and Anice C. Lowen. 2014. “Influenza A Virus Reassortment.” Current Topics in 1188 Microbiology and Immunology (Cham), Current topics in microbiology and immunology, vol. 1189 385: 377–401. 1190 Taubenberger, Jeffery K., and John C. Kash. 2010. “Influenza Virus Evolution, Host Adaptation, 1191 and Pandemic Formation.” Cell Host & Microbe 7 (6): 440–451. 1192 Thomas, Megan N., Giovana Ciacci Zanella, Brianna Cowan, et al. 2024. “Nucleoprotein 1193 Reassortment Enhanced Transmissibility of H3 1990.4.a Clade Influenza A Virus in Swine.” 1194 Journal of Virology 98 (3): e0170323. 1195 Trovão, Nidia S., Sairah M. Khan, Philippe Lemey, Martha I. Nelson, and Joshua L. Cherry. 1196 2024. “Comparative Evolution of Influenza A Virus H1 and H3 Head and Stalk Domains 1197 across Host Species.” mBio 15 (1): e0264923. 1198 U.S. Fish & Wildlife Service. 2023. “Migratory Bird Program Administrative Flyways.” FWS.gov. 1199 https://www.fws.gov/partner/migratory-bird-program-administrative-flyways. 1200 Venkatesh, Divya, Carlo Bianco, Alejandro Núñez, et al. 2020. “Detection of H3N8 Influenza A 1201 Virus with Multiple Mammalian-Adaptive Mutations in a Rescued Grey Seal (Halichoerus 1202 Grypus) Pup.” Virus Evolution 6 (1): veaa016. 1203 Vienne, Damien M. de. 2019. “Tanglegrams Are Misleading for Visual Evaluation of Tree 1204 Congruence.” Molecular Biology and Evolution 36 (1): 174–176. 1205 Vijaykrishna, Dhanasekaran, Gavin J. D. Smith, Oliver G. Pybus, et al. 2011. “Long-Term 1206 Evolution and Transmission Dynamics of Swine Influenza A Virus.” Nature 473 (7348): 1207 519–522. 1208 Vincent, Amy L., Tavis K. Anderson, and Kelly M. Lager. 2020. “A Brief Introduction to Influenza 1209 A Virus in Swine.” Methods in Molecular Biology (Clifton, N.J.) (New York, NY) 2123: 249– 1210 271. 1211 Vries, Erhard van der, Evelien A. Germeraad, Annelies Kroneman, et al. 2025. “Swine Influenza 1212 Virus Surveillance Programme Pilot to Assess the Risk for Animal and Public Health, the 1213 Netherlands, 2022 to 2023.” Euro Surveillance : Bulletin Europeen Sur Les Maladies 1214 Transmissibles [Euro Surveillance : European Communicable Disease Bulletin] 30 (22): 1215 2400664. 1216 Wasik, Brian R., Lambodhar Damodaran, Maria A. Maltepes, et al. 2025. “The Evolution and 1217 Epidemiology of H3N2 Canine Influenza Virus after 20 Years in Dogs.” Epidemiology and 1218 Infection 153 (e47): e47. 1219 Wasik, Brian R., Evin Rothschild, Ian E. H. Voorhees, et al. 2023. “Understanding the Divergent 1220 Evolution and Epidemiology of H3N8 Influenza Viruses in Dogs and Horses.” Virus 1221 Evolution 9 (2): vead052. 1222 Webby, R. J., S. L. Swenson, S. L. Krauss, P. J. Gerrish, S. M. Goyal, and R. G. Webster. 2000. 1223 “Evolution of Swine H3N2 Influenza Viruses in the United States.” Journal of Virology 74 1224 (18): 8243–8251. 1225 White, Maria C., and Anice C. Lowen. 2018. “Implications of Segment Mismatch for Influenza A 1226 Virus Evolution.” The Journal of General Virology 99 (1): 3–16. 1227 White, Maria C., John Steel, and Anice C. Lowen. 2017. “Heterologous Packaging Signals on 1228 Segment 4, but Not Segment 6 or Segment 8, Limit Influenza A Virus Reassortment.” 1229 Journal of Virology 91 (11): e00195–17. 1230 Wille, Michelle, and Edward C. Holmes. 2020. “The Ecology and Evolution of Influenza Viruses.” 1231 Cold Spring Harbor Perspectives in Medicine 10 (7): a038489. 1232 Wille, Michelle, Conny Tolf, Alexis Avril, et al. 2013. “Frequency and Patterns of Reassortment 1233 in Natural Influenza A Virus Infection in a Reservoir Host.” Virology 443 (1): 150–160. 1234 Xu, Rui, Xueyong Zhu, Ryan McBride, et al. 2012. “Functional Balance of the Hemagglutinin 1235 and Neuraminidase Activities Accompanies the Emergence of the 2009 H1N1 Influenza 1236 Pandemic.” Journal of Virology 86 (17): 9221–9232. 1237 Yang, Huanliang, Yihong Xiao, Fei Meng, et al. 2018. “Emergence of H3N8 Equine Influenza 1238 Virus in Donkeys in China in 2017.” Veterinary Microbiology 214 (February): 1–6. 1239 Yang, Jiaying, Xiaojing Chen, Xiyan Li, et al. 2025. “Global Spread of H3 Subtype Avian 1240 Influenza Viruses with an Accelerated Evolution after Interspecies Transmission.” The 1241 Journal of Infection 91 (2): 106542. 1242 Yen, Hui-Ling, Chi-Hui Liang, Chung-Yi Wu, et al. 2011. “Hemagglutinin-Neuraminidase 1243 Balance Confers Respiratory-Droplet Transmissibility of the Pandemic H1N1 Influenza 1244 Virus in Ferrets.” Proceedings of the National Academy of Sciences of the United States of 1245 America 108 (34): 14264–14269. 1246 Yondon, Myagmarsukh, Batsukh Zayat, Martha I. Nelson, et al. 2014. “Equine Influenza 1247 A(H3N8) Virus Isolated from Bactrian Camel, Mongolia.” Emerging Infectious Diseases 20 1248 (12): 2144–2147. 1249 Yoon, Sun-Woo, Richard J. Webby, and Robert G. Webster. 2014. “Evolution and Ecology of 1250 Influenza A Viruses.” Current Topics in Microbiology and Immunology (Cham), Current 1251 topics in microbiology and immunology, vol. 385: 359–375. 1252 Youk, Sungsu, Mia Kim Torchetti, Kristina Lantz, et al. 2023. “H5N1 Highly Pathogenic Avian 1253 Influenza Clade 2.3.4.4b in Wild and Domestic Birds: Introductions into the United States 1254 and Reassortments, December 2021-April 2022.” Virology 587 (109860): 109860. 1255 Zeller, Michael A., Daniel Carnevale de Almeida Moraes, Giovana Ciacci Zanella, et al. 2024. 1256 “Reverse Zoonosis of the 2022-2023 Human Seasonal H3N2 Detected in Swine.” Npj 1257 Viruses 2 (1): 27. 1258 Zeller, Michael A., Jennifer Chang, Amy L. Vincent, Phillip C. Gauger, and Tavis K. Anderson. 1259 2021. “Spatial and Temporal Coevolution of N2 Neuraminidase and H1 and H3 1260 Hemagglutinin Genes of Influenza A Virus in US Swine.” Virus Evolution 7 (2): veab090. 1261 Zell, Roland, Christoph Scholtissek, and Stephan Ludwig. 2013. “Genetics, Evolution, and the 1262 Zoonotic Capacity of European Swine Influenza Viruses.” Current Topics in Microbiology 1263 and Immunology (Berlin, Heidelberg), Current topics in microbiology and immunology, vol. 1264 370: 29–55. 1265 Zhou, N. N., D. A. Senne, J. S. Landgraf, et al. 1999. “Genetic Reassortment of Avian, Swine, 1266 and Human Influenza A Viruses in American Pigs.” Journal of Virology 73 (10): 8851–8856. 1267 Baker, Steven F., Aitor Nogales, Courtney Finch, et al. 2014. “Influenza A and B Virus Intertypic 1268 Reassortment through Compatible Viral Packaging Signals.” Journal of Virology 88 (18): 1269 10778–10791. 1270 Bhatt, Samir, Edward C. Holmes, and Oliver G. Pybus. 2011. “The Genomic Rate of Molecular 1271 Adaptation of the Human Influenza A Virus.” Molecular Biology and Evolution 28 (9): 2443– 1272 2451. 1273 Brockwell-Staats, Christy, Robert G. Webster, and Richard J. Webby. 2009. “Diversity of 1274 Influenza Viruses in Swine and the Emergence of a Novel Human Pandemic Influenza A 1275 (H1N1).” Influenza and Other Respiratory Viruses 3 (5): 207–213. 1276 Brown, Ian H. 2013. “History and Epidemiology of Swine Influenza in Europe.” Current Topics in 1277 Microbiology and Immunology (Berlin, Heidelberg), Current topics in microbiology and 1278 immunology, vol. 370: 133–146. 1279 Crawford, P. C., Edward J. Dubovi, William L. Castleman, et al. 2005. “Transmission of Equine 1280 Influenza Virus to Dogs.” Science (New York, N.Y.) 310 (5747): 482–485. 1281 Damodaran, Lambodhar, Joseph A. Lewnard, Gregg S. Davis, Sara Y. Tartof, Louise H. 1282 Moncla, and Nicola F. Müller. 2026. “Frequent Seasonal Reassortment between High and 1283 Low Path Viruses Drives the Diversification of Influenza A/H5N1.” In bioRxiv. BioRxiv, April 1284 18. https://doi.org/10.64898/2026.04.17.719307. 1285 Dugan, Vivien G., Rubing Chen, David J. Spiro, et al. 2008. “The Evolutionary Genetics and 1286 Emergence of Avian Influenza Viruses in Wild Birds.” PLoS Pathogens 4 (5): e1000076. 1287 Essere, Boris, Matthieu Yver, Cyrille Gavazzi, et al. 2013. “Critical Role of Segment-Specific 1288 Packaging Signals in Genetic Reassortment of Influenza A Viruses.” Proceedings of the 1289 National Academy of Sciences of the United States of America 110 (40): E3840–8. 1290 Ferguson, Neil M., Alison P. Galvani, and Robin M. Bush. 2003. “Ecological and Immunological 1291 Determinants of Influenza Evolution.” Nature 422 (6930): 428–433. 1292 Furuse, Yuki, Akira Suzuki, and Hitoshi Oshitani. 2010. “Reassortment between Swine Influenza 1293 A Viruses Increased Their Adaptation to Humans in Pandemic H1N1/09.” Infection, 1294 Genetics and Evolution: Journal of Molecular Epidemiology and Evolutionary Genetics in 1295 Infectious Diseases 10 (4): 569–574. 1296 Ganti, Ketaki, Anish Bagga, Juliana DaSilva, et al. 2021. “Avian Influenza A Viruses Reassort 1297 and Diversify Differently in Mallards and Mammals.” Viruses 13 (3): 509. 1298 Ince, William L., Aissatou Gueye-Mbaye, Jack R. Bennink, and Jonathan W. Yewdell. 2013. 1299 “Reassortment Complements Spontaneous Mutation in Influenza A Virus NP and M1 1300 Genes to Accelerate Adaptation to a New Host.” Journal of Virology 87 (8): 4330–4338. 1301 Khare, Shruti, Céline Gurry, Lucas Freitas, et al. 2021. “GISAID’s Role in Pandemic Response.” 1302 China CDC Weekly 3 (49): 1049–1051. 1303 Khiabanian, Hossein, Vladimir Trifonov, and Raul Rabadan. 2009. “Reassortment Patterns in 1304 Swine Influenza Viruses.” PloS One 4 (10): e7366. 1305 Kistler, Kathryn E., and Trevor Bedford. 2023. “An Atlas of Continuous Adaptive Evolution in 1306 Endemic Human Viruses.” Cell Host & Microbe 31 (11): 1898–1909.e3. 1307 Kuchinski, Kevin S., John Tyson, Tracy Lee, et al. 2025. “Detection of a Reassortant Swine- and 1308 Human-Origin H3N2 Influenza A Virus in Farmed Mink in British Columbia, Canada.” 1309 Zoonoses and Public Health 72 (3): 293–300. 1310 Latorre-Margalef, Neus, Vladimir Grosbois, John Wahlgren, et al. 2013. “Heterosubtypic 1311 Immunity to Influenza A Virus Infections in Mallards May Explain Existence of Multiple Virus 1312 Subtypes.” PLoS Pathogens 9 (6): e1003443. 1313 Le Sage, Valerie, Michelle N. Vu, Maria A. Maltepes, et al. 2026. “Fatal Human H3N8 Influenza 1314 Virus Has a Moderate Pandemic Risk.” PLoS Pathogens 22 (3): e1013586. 1315 Lindstrom, Stephen E., Nancy J. Cox, and Alexander Klimov. 2004. “Genetic Analysis of Human 1316 H2N2 and Early H3N2 Influenza Viruses, 1957-1972: Evidence for Genetic Divergence and 1317 Multiple Reassortment Events.” Virology 328 (1): 101–119. 1318 Liu, Tongyu, Yiquan Wang, Timothy J. C. Tan, Nicholas C. Wu, and Christopher B. Brooke. 1319 2022. “The Evolutionary Potential of Influenza A Virus Hemagglutinin Is Highly Constrained 1320 by Epistatic Interactions with Neuraminidase.” Cell Host & Microbe 30 (10): 1363–1369.e4. 1321 Lowen, Anice C. 2017. “Constraints, Drivers, and Implications of Influenza A Virus 1322 Reassortment.” Annual Review of Virology 4 (1): 105–121. 1323 Lycett, S. J., G. Baillie, E. Coulter, et al. 2012. “Estimating Reassortment Rates in Co- 1324 Circulating Eurasian Swine Influenza Viruses.” The Journal of General Virology 93 (Pt 11): 1325 2326–2336. 1326 Macken, Catherine A., Richard J. Webby, and William J. Bruno. 2006. “Genotype Turnover by 1327 Reassortment of Replication Complex Genes from Avian Influenza A Virus.” The Journal of 1328 General Virology 87 (Pt 10): 2803–2815. 1329 Ma, Eric J., Nichola J. Hill, Justin Zabilansky, Kyle Yuan, and Jonathan A. Runstadler. 2016. 1330 “Reticulate Evolution Is Favored in Influenza Niche Switching.” Proceedings of the National 1331 Academy of Sciences of the United States of America 113 (19): 5335–5339. 1332 Markin, Alexey, Catherine A. Macken, Amy L. Baker, and Tavis K. Anderson. 2025. “Revealing 1333 Reassortment in Influenza A Viruses with TreeSort.” Molecular Biology and Evolution 42 1334 (8): msaf133. 1335 Marshall, Nicolle, Lalita Priyamvada, Zachary Ende, John Steel, and Anice C. Lowen. 2013. 1336 “Influenza Virus Reassortment Occurs with High Frequency in the Absence of Segment 1337 Mismatch.” PLoS Pathogens 9 (6): e1003421. 1338 Mehle, Andrew, Vivien G. Dugan, Jeffery K. Taubenberger, and Jennifer A. Doudna. 2012. 1339 “Reassortment and Mutation of the Avian Influenza Virus Polymerase PA Subunit 1340 Overcome Species Barriers.” Journal of Virology 86 (3): 1750–1757. 1341 Mitnaul, L. J., M. N. Matrosovich, M. R. Castrucci, et al. 2000. “Balanced Hemagglutinin and 1342 Neuraminidase Activities Are Critical for Efficient Replication of Influenza A Virus.” Journal 1343 of Virology 74 (13): 6015–6020. 1344 Müller, Nicola F., Ugnė Stolz, Gytis Dudas, Tanja Stadler, and Timothy G. Vaughan. 2020. 1345 “Bayesian Inference of Reassortment Networks Reveals Fitness Benefits of Reassortment 1346 in Human Influenza Viruses.” Proceedings of the National Academy of Sciences of the 1347 United States of America 117 (29): 17104–17111. 1348 Munster, Vincent J., Chantal Baas, Pascal Lexmond, et al. 2007. “Spatial, Temporal, and 1349 Species Variation in Prevalence of Influenza A Viruses in Wild Migratory Birds.” PLoS 1350 Pathogens 3 (5): e61. 1351 Nelson, Martha I., Cécile Viboud, Lone Simonsen, et al. 2008. “Multiple Reassortment Events in 1352 the Evolutionary History of H1N1 Influenza A Virus since 1918.” PLoS Pathogens 4 (2): 1353 e1000012. 1354 Nelson, Martha I., and Amy L. Vincent. 2015. “Reverse Zoonosis of Influenza to Swine: New 1355 Perspectives on the Human-Animal Interface.” Trends in Microbiology 23 (3): 142–153. 1356 Nelson, Martha I., and Michael Worobey. 2018. “Origins of the 1918 Pandemic: Revisiting the 1357 Swine ‘Mixing Vessel’ Hypothesis.” American Journal of Epidemiology 187 (12): 2498– 1358 2502. 1359 Ovadia, Y., D. Fielder, C. Conow, and R. Libeskind-Hadas. 2011. “The Co Phylogeny 1360 Reconstruction Problem Is NP-Complete.” Journal of Computational Biology: A Journal of 1361 Computational Molecular Cell Biology 18 (1): 59–65. 1362 Parrish, Colin R., Pablo R. Murcia, and Edward C. Holmes. 2015. “Influenza Virus Reservoirs 1363 and Intermediate Hosts: Dogs, Horses, and New Possibilities for Influenza Virus Exposure 1364 of Humans.” Journal of Virology 89 (6): 2990–2994. 1365 Petrova, Velislava N., and Colin A. Russell. 2018. “The Evolution of Seasonal Influenza 1366 Viruses.” Nature Reviews. Microbiology 16 (1): 47–60. 1367 Rajão, Daniela S., Phillip C. Gauger, Tavis K. Anderson, et al. 2015. “Novel Reassortant 1368 Human-like H3N2 and H3N1 Influenza A Viruses Detected in Pigs Are Virulent and 1369 Antigenically Distinct from Swine Viruses Endemic to the United States.” Journal of Virology 1370 89 (22): 11213–11222. 1371 Rambaut, Andrew, Oliver G. Pybus, Martha I. Nelson, Cecile Viboud, Jeffery K. Taubenberger, 1372 and Edward C. Holmes. 2008. “The Genomic and Epidemiological Dynamics of Human 1373 Influenza A Virus.” Nature 453 (7195): 615–619. 1374 Roche, Benjamin, John M. Drake, Justin Brown, David E. Stallknecht, Trevor Bedford, and 1375 Pejman Rohani. 2014. “Adaptive Evolution and Environmental Durability Jointly Structure 1376 Phylodynamic Patterns in Avian Influenza Viruses.” PLoS Biology 12 (8): e1001931. 1377 Sharma, Aditi, Michael A. Zeller, Carine K. Souza, et al. 2022. “Characterization of a 2016-2017 1378 Human Seasonal H3 Influenza A Virus Spillover Now Endemic to U.s. Swine.” mSphere 7 1379 (1): e0080921. 1380 Simon, Gaëlle, Lars E. Larsen, Ralf Dürrwald, et al. 2014. “European Surveillance Network for 1381 Influenza in Pigs: Surveillance Programs, Diagnostic Tools and Swine Influenza Virus 1382 Subtypes Identified in 14 European Countries from 2010 to 2013.” PloS One 9 (12): 1383 e115815. 1384 Smith, Gavin J. D., Dhanasekaran Vijaykrishna, Justin Bahl, et al. 2009. “Origins and 1385 Evolutionary Genomics of the 2009 Swine-Origin H1N1 Influenza A Epidemic.” Nature 459 1386 (7250): 1122–1125. 1387 Song, Daesub, Bokyu Kang, Chulseung Lee, et al. 2008. “Transmission of Avian Influenza Virus 1388 (H3N2) to Dogs.” Emerging Infectious Diseases 14 (5): 741–746. 1389 Song, D. S., D. J. An, H. J. Moon, et al. 2011. “Interspecies Transmission of the Canine 1390 Influenza H3N2 Virus to Domestic Cats in South Korea, 2010.” The Journal of General 1391 Virology 92 (Pt 10): 2350–2355. 1392 Steel, John, and Anice C. Lowen. 2014. “Influenza A Virus Reassortment.” Current Topics in 1393 Microbiology and Immunology (Cham), Current topics in microbiology and immunology, vol. 1394 385: 377–401. 1395 Taubenberger, Jeffery K., and John C. Kash. 2010. “Influenza Virus Evolution, Host Adaptation, 1396 and Pandemic Formation.” Cell Host & Microbe 7 (6): 440–451. 1397 Thomas, Megan N., Giovana Ciacci Zanella, Brianna Cowan, et al. 2024. “Nucleoprotein 1398 Reassortment Enhanced Transmissibility of H3 1990.4.a Clade Influenza A Virus in Swine.” 1399 Journal of Virology 98 (3): e0170323. 1400 Trovão, Nidia S., Sairah M. Khan, Philippe Lemey, Martha I. Nelson, and Joshua L. Cherry. 1401 2024. “Comparative Evolution of Influenza A Virus H1 and H3 Head and Stalk Domains 1402 across Host Species.” mBio 15 (1): e0264923. 1403 U.S. Fish & Wildlife Service. 2023. “Migratory Bird Program Administrative Flyways.” FWS.gov. 1404 https://www.fws.gov/partner/migratory-bird-program-administrative-flyways. 1405 Venkatesh, Divya, Carlo Bianco, Alejandro Núñez, et al. 2020. “Detection of H3N8 Influenza A 1406 Virus with Multiple Mammalian-Adaptive Mutations in a Rescued Grey Seal (Halichoerus 1407 Grypus) Pup.” Virus Evolution 6 (1): veaa016. 1408 Vienne, Damien M. de. 2019. “Tanglegrams Are Misleading for Visual Evaluation of Tree 1409 Congruence.” Molecular Biology and Evolution 36 (1): 174–176. 1410 Wasik, Brian R., Lambodhar Damodaran, Maria A. Maltepes, et al. 2025. “The Evolution and 1411 Epidemiology of H3N2 Canine Influenza Virus after 20 Years in Dogs.” Epidemiology and 1412 Infection 153 (e47): e47. 1413 Webby, R. J., S. L. Swenson, S. L. Krauss, P. J. Gerrish, S. M. Goyal, and R. G. Webster. 2000. 1414 “Evolution of Swine H3N2 Influenza Viruses in the United States.” Journal of Virology 74 1415 (18): 8243–8251. 1416 White, Maria C., and Anice C. Lowen. 2018. “Implications of Segment Mismatch for Influenza A 1417 Virus Evolution.” The Journal of General Virology 99 (1): 3–16. 1418 White, Maria C., John Steel, and Anice C. Lowen. 2017. “Heterologous Packaging Signals on 1419 Segment 4, but Not Segment 6 or Segment 8, Limit Influenza A Virus Reassortment.” 1420 Journal of Virology 91 (11): e00195–17. 1421 Wille, Michelle, and Edward C. Holmes. 2020. “The Ecology and Evolution of Influenza Viruses.” 1422 Cold Spring Harbor Perspectives in Medicine 10 (7): a038489. 1423 Wille, Michelle, Conny Tolf, Alexis Avril, et al. 2013. “Frequency and Patterns of Reassortment 1424 in Natural Influenza A Virus Infection in a Reservoir Host.” Virology 443 (1): 150–160. 1425 Yang, Huanliang, Yihong Xiao, Fei Meng, et al. 2018. “Emergence of H3N8 Equine Influenza 1426 Virus in Donkeys in China in 2017.” Veterinary Microbiology 214 (February): 1–6. 1427 Yang, Jiaying, Xiaojing Chen, Xiyan Li, et al. 2025. “Global Spread of H3 Subtype Avian 1428 Influenza Viruses with an Accelerated Evolution after Interspecies Transmission.” The 1429 Journal of Infection 91 (2): 106542. 1430 Yondon, Myagmarsukh, Batsukh Zayat, Martha I. Nelson, et al. 2014. “Equine Influenza 1431 A(H3N8) Virus Isolated from Bactrian Camel, Mongolia.” Emerging Infectious Diseases 20 1432 (12): 2144–2147. 1433 Yoon, Sun-Woo, Richard J. Webby, and Robert G. Webster. 2014. “Evolution and Ecology of 1434 Influenza A Viruses.” Current Topics in Microbiology and Immunology (Cham), Current 1435 topics in microbiology and immunology, vol. 385: 359–375. 1436 Youk, Sungsu, Mia Kim Torchetti, Kristina Lantz, et al. 2023. “H5N1 Highly Pathogenic Avian 1437 Influenza Clade 2.3.4.4b in Wild and Domestic Birds: Introductions into the United States 1438 and Reassortments, December 2021-April 2022.” Virology 587 (109860): 109860. 1439 Zeller, Michael A., Daniel Carnevale de Almeida Moraes, Giovana Ciacci Zanella, et al. 2024. 1440 “Reverse Zoonosis of the 2022-2023 Human Seasonal H3N2 Detected in Swine.” Npj 1441 Viruses 2 (1): 27. 1442 Zeller, Michael A., Jennifer Chang, Amy L. Vincent, Phillip C. Gauger, and Tavis K. Anderson. 1443 2021. “Spatial and Temporal Coevolution of N2 Neuraminidase and H1 and H3 1444 Hemagglutinin Genes of Influenza A Virus in US Swine.” Virus Evolution 7 (2): veab090. 1445 Zell, Roland, Christoph Scholtissek, and Stephan Ludwig. 2013. “Genetics, Evolution, and the 1446 Zoonotic Capacity of European Swine Influenza Viruses.” Current Topics in Microbiology 1447 and Immunology (Berlin, Heidelberg), Current topics in microbiology and immunology, vol. 1448 370: 29–55. 1449 Zhou, N. N., D. A. Senne, J. S. Landgraf, et al. 1999. “Genetic Reassortment of Avian, Swine, 1450 and Human Influenza A Viruses in American Pigs.” Journal of Virology 73 (10): 8851–8856. 1451