Co-circulation of multiple influenza A variants in swine harboring genes from seasonal human and swine influenza viruses

Since the influenza pandemic in 2009, there has been an increased focus on swine influenza A virus (swIAV) surveillance. This paper describes the results of the surveillance of swIAV in Danish swine from 2011 to 2018. In total, 3800 submissions were received with a steady increase in swIAV positive submissions, reaching 56% in 2018. Ten different swIAV subtypes were detected. Full genome sequences were obtained from 129 swIAV positive samples. Altogether, 17 different circulating genotypes were identified including novel reassortants and subtypes harboring human seasonal IAV gene segments. The phylogenetic analysis revealed substantial genetic drift and also evidence of positive selection occurring mainly in antigenic sites of the hemagglutinin protein and confirmed the presence of a swine divergent cluster among the H1pdm09Nx viruses. The results provide essential data for the control of swIAV in pigs and for early detection of novel swIAV strains with zoonotic potential.


Introduction 39
Influenza A virus (swIAV) infection in swine causes respiratory disease, impairs the growth rate and 40 increases the risk of secondary infections 1-3 . SwIAV is enzootic globally and multiple subtypes and lineages 41 have been identified 4 . The influenza A virus genome consists of eight distinct gene segments and subtypes 42 are assigned by characterizing the two surface glycoproteins hemagglutinin (HA) and neuraminidase (NA) 5 . 43 Pigs are infected by the same subtypes as humans, including H1N1, H1N2 and H3N2 6 . The transmission of 44 H1N1 avian influenza A virus (IAV) to swine in the 1970s created the H1N1 Eurasian swine lineage also 45 called "avian-like swine H1N1" (1.C lineage 7 ) circulating in Europe and Asia 8 . An H3N2 influenza virus 46 related to a human strain from 1973 started to circulate in the European pig populations in 1984. In the mid-47 1980s, a reassortment between the avian-like swine H1N1 and H3N2 human virus resulted in a human-like 48 reassortant swine "H3N2sw" that became established in European swine 9,10 . In 1994, a H1N2 reassortant 49 (1.B lineage) comprising an HA gene from human seasonal H1N1, an NA gene from H3N2sw and internal 50 genes originating from avian-like swine H1N1 was first identified in the United Kingdom and subsequently 51 detected in many European countries 11 . This swIAV lineage is also known as European human-like 52 "H1huN2". However, this subtype has never been detected in Danish pigs. In the beginning of the 2000s, a 53 new "H1N2dk" reassortant virus was identified in Danish pigs 12 . This H1N2dk virus comprised an avian-like 54 swine HA gene and an NA from contemporary, circulating H3N2sw and has since been identified in several 55 European countries 13,14,15,16 . H1avN2hu subtype has been detected each year, with the exception of 2016 (Fig. 4). 140 In summary, six novel swIAV reassortant subtypes (H1pdmN2sw, H1pdmN2hu, H1pdmN1av, H1avN1pdm, 141 H3hu05N2sw and H1avN2hu) were discovered through the Danish surveillance of swIAV from 2011-2018 142 (Fig. 4). However, the diversity of circulating strains is even more complex, when all gene segments are 143 included in the analyses as described below. 144 EPI_ISL_247092) has previously been described 26 , but was also included in the analysis of the H3hu genes 150 of this study. 151

Hemagglutinin gene characterization 152
In total, 78 H1av, 48 H1pdm09 and three H3hu05 full-length HA sequences were obtained and analyzed 153 separately according to the lineage. 154 The H1av nucleotide sequences were fairly diverse, with an average pairwise sequence difference per site 155 (pi) of 0.099, SE: 0.0005. Phylogenetic trees constructed either with or without clock-models 156 (Supplementary figure 2, Figure 5), did not display the imbalanced, ladder-like structure typical for influenza 157 trees. The tree contained several clusters and one cluster (Cluster 6, Figure 5) that was dominated by 158 H1avNx strains carrying a complete internal gene cassette of H1N1pdm09 origin. There was low correlation 159 between sampling time and genetic divergence ( Table 2). 160 Analysis using CODEML indicated strong evidence for positive selection among the H1av sequences.  (Table 2). 164 The H1pdm09 nucleotide sequences had a lower nucleotide diversity: pi = 0.043, SE 0.0005. Both the clock 165 and non-clock trees for H1pdm09 sequences isolated from Danish pigs, showed that 30 of the sequences 166 were located in a well-defined cluster (Cluster 1; Fig. 6 and Supplementary figure 3), with the remaining 18 167 sequences branching out basally to this cluster ( Fig. 6 and Supplementary figure 3). The 30 H1pdm09 168 sequences of Cluster 1 were collected between 2015-2018, whereas the 18 H1pdm09 sequences outside this 169 cluster were collected between 2013-2017 (Fig. 6). The strict molecular clock tree and the TempEst analysis 170 of all the Danish H1pdm09 sequences suggested that the sequences evolved according to time with stable 171 substitution rate of 4.9 x 10 -3 per site per year ( Fig. 6 and Table 2). The diversion into the Cluster 1 appeared 172 to have occurred around 2011, however the most recent common ancestor for the sequences in Cluster 1 was 173 dated around 2014 (Fig. 6). In the phylogenetic tree that also included representative swine and human 174 seasonal H1pdm09 sequences, it was found that Cluster 1 only contained swine derived H1pdm09 175 sequences, while sequences outside of this cluster was a mix of swine and human seasonal H1pdm09 176 sequences. Cluster 1 was therefore termed the swine like "Sw-L cluster" ( The initial analysis of the Danish H1pdm09 aa sequences revealed a total of 20 aa positions that differed 180 between the Danish Sw-L and the Danish Hu-L sequences ( Table 3) and seven of these 20 aa differences 181 were specific, meaning that all the 30 Danish Sw-L aa sequences had a different aa compared to all of the 18 182 Danish Hu-L aa sequences (Bold positions in Table 3). Thirteen of the 20 aa residues defining the Sw-L 183 protein sequences were located either in previously defined antigenic sites (Ca and Sb) or the receptor 184 binding site (RBS). Six of these were among the seven "unique" Sw-L positions (Table 3). 185 Subsequently, the 20 aa residues were compared among all the sequences included in the phylogenetic tree 186 of Fig 7. These reference aa sequences were divided into three groups; one containing the foreign (non-187 Danish) swine H1pdm09 sequences (n=11) included in the "Sw-L cluster", one containing the European 188 swine H1pdm09 sequences located outside the Sw-L cluster (n=42) and one containing human seasonal 189 H1pdm09 sequences (n=59) (Table 3). Interestingly, all the 11 foreign swine H1pdm09 aa sequences 190 included in the Sw-L cluster, shared exactly the same aa residues as the Danish Sw-L sequences. Similarly, 191 the majority of the European swine-and human seasonal H1pdm09 aa sequences located outside the Sw-L 192 cluster carried residues similar to the Danish Hu-L aa sequences, and were different from the sequences 193 included in the Sw-L cluster (Table 3). Finally, no unique swine or human residues were revealed when all 194 H1pdm09 proteins derived from swine were compared to the H1pdm09 proteins derived from human 195 seasonal H1pdm09 viruses. Nonetheless, at position 273, significantly more swine H1pdm09 proteins (91 %) 196 carried an A compared to the human seasonal H1pdm09 proteins (27 %) (p = < 0.05). In summary, the 197 H1pdm09 proteins derived from Danish pigs were divided into two groups containing "Sw-L" and "Hu-L" 198 sequences, which were separated by 20 aa differences mainly located in antigenic sites or the RBS. The Sw-199 L cluster still clustered separately when swine-and human-derived H1pdm09 reference sequences were 200 included in the alignment, but 11 additional German and Italian H1pdm09 swine-derived sequences were 201 also part of this cluster. 202 The CODEML analysis for determining the best fitting substitution model revealed that the M2a model fitted sequences significantly better, suggesting that no positive selection occurred among these sequences. The 213 strict molecular clock and TempEst analysis were also repeated for the Danish Sw-L and Hu-L sequences 214 separately. Interestingly, the Hu-L sequences had a higher substitution rate and also showed a higher 215 correlation coefficient in the TempEst analysis compared to the Sw-L sequences (Table 3). In summary, 216 when analyzing all the H1pdm09 sequences as a whole, positive selection was evident among the sequences. 217 However, when dividing the H1pdm09 sequences into the Sw-L and Hu-L groups, positive selection was 218 only evident among the Hu-L H1pdm09 sequences and these sequences also showed a higher substitution 219 rate compared to the Sw-L sequences. 220 Additionally, differences in N-linked and O-linked glycosylation sites between the Sw-L and Hu-L 221 H1pdm09 proteins were examined. The results revealed that all proteins of both the Sw-L and Hu-L samples 222 were predicted to be N-glycosylated at position 28, 40, 304 and 557 (numbering from the first methionine). 223 In addition, 3/18 Hu-L H1pdm09 proteins were predicted to be N-glycosylated at position 136, which is in  Previously defined residues of the HA proteins regarded as important for host-adaptation, pathogenicity, 230 receptor binding and virulence were examined and compared between subtypes carrying an HA protein of 231 avian and H1N1pdm09 origin, respectively. The results can be visualized in Supplementary table 3. 232 The three H3 sequences obtained in this study, showed a low nucleotide diversity (pi) of 0.027, SE: 0.002. 233 The closest human IAV match in NCBI GenBank for all of the three sequences was 234 "A/Denmark/129/2005(H3N2)" with accession number EU103786. As only three sequences were obtained, 235 no further phylogenetic or evolutionary analysis were performed. 236

Neuraminidase characteristics 237
In total, 32 N1pdm, 14 N1av, 75 N2sw and 8 N2hu full-length NA sequences were obtained and analyzed 238 separately according to the lineage.  (Table 4). TempEST analysis showed a relative low correlation between the genetic divergence and time, and the Beast 248 analysis revealed a substitution rate of 5.9 x 10 -3 per site per year. No evidence of positive selection was 249 observed (Table 4). 250 The N2sw nucleotide diversity was 0.08, SE: 0.0005 and the Bayesian analysis revealed six main clusters. 251 Each cluster contained sequences dispersed over the majority of the surveillance period, suggesting no 252 temporal clustering. Interestingly, one major cluster only contained HxN2sw from strains having a full or 253 partial H1N1pdm09 internal gene cassette. Moreover, this cluster contained 28/30 of the same samples as 254 Cluster 3 of the H1av sequences, which also clustered according to the origin of the internal gene cassette 255  (Table 4). 258 The eighth N2hu sequences showed a sequence diversity of 0.085, SE: 0.006 and despite the limited number 259 of sequences, the Bayesian phylogenetic analysis revealed two main clusters; one containing sequences 260 derived from subtypes containing a full avian internal gene cassette and one only containing sequences with 261 a full or partial H1N1pdm09 internal cassette (Supplementary Figure 7). The TempEst analysis revealed a 262 low correlation between genetic divergence and time, and the low number of sequences resulted in an 263 overestimated substitution rate, which therefore was not included in the results. No evidence of positive 264 selection was observed (Table 4). 265 All of NA sequences across the different lineages (n=129) were examined for specific aa changes encoding 266 either neuraminidase resistance or increased virulence. However, none of the NA sequences had any of these 267 aa changes. 268 The internal gene cassette 269 In total, 17 different genotypes were identified in this study (Fig. 8). The subtypes H1N2dk, avian-like swine 270 Full genome sequencing of the swIAV isolates obtained over the eight years, revealed that since 2013, an 286 increasing number of the H1N2dk subtypes sequenced had acquired an internal gene cassette of H1N1pdm09 287 origin (Fig. 9). Similarly, though not as many samples were available, the H1avN2hu also seemed to gain 288 internal genes of H1N1pdm09 origin over time. In contrast, the avian-like swine H1N1 subtype, roughly 289 maintained an avian-like swine internal gene cassette, with an exception of three isolates, which contained an 290  (Supplementary table 2). 296 Previously defined important residues of the proteins encoded by the internal gene cassette was analyzed and 297 the results are summarized in Supplementary table 3. Furthermore, comparisons of the proteins encoded by 298 the internal gene cassette of the Sw-L and Hu-L H1pdm09Nx viruses were performed and some aa 299 differences between the two groups were identified. However, none of the aa differences were 100 % 300 specific to each group (Supplementary Table 4). considering swIAV to be a seasonal disease and are therefore not submitting samples for swIAV testing 315 during summer. In addition, no seasonality was documented for the prevalence of H1pdm09 positive 316 submissions, which is in accordance with a recent French study 41 . This could indicate that while 317 H1N1pdm09 reverse-zoonosis events occurs during the human influenza season, the high level of 318 H1N1pdm09 circulating in Danish pigs independent of the human influenza season hide the impact observed 319 on the H1N1pdm09 occurrence during the autumn and winter months. 320

Prevalent subtypes and reassortant swIAV 321
During the first three years of the surveillance program, the two most common influenza A virus subtypes in 322 Danish swine were avian-like swine H1N1 and H1N2dk, which harbor the same HA gene. However, soon 323 after the first introduction of H1N1pdm09 in January 2010, this subtype rapidly spread, and has since 2014 324 remained the second most prevalent subtype in Denmark. The swIAV subtype H3N2sw has almost 325 disappeared from Denmark, in line with surveillance data obtained in some other European countries such as 326 the UK and France 16,33 . Conversely, the H1N2dk has been steadily increasing in prevalence since 2012, and 327 is currently the most dominating subtype in Denmark. Concurrently, the H1N2dk has gradually gained an 328 internal gene cassette of H1N1pdm09 origin, suggesting that this gene constellation is beneficial for the 329 virus. In general, an increase in Danish swIAV subtypes carrying an internal gene cassette of H1N1pdm09 330 origin was observed, which indicates that an internal gene cassette of H1N1pdm09 origin is advantageous, 331 compared to an avian-like swine H1N1 derived internal gene cassette. The benefit of having a complete or 332 partial internal gene cassette of H1N1pdm09 origin, could be explained by the polymerase genes having a 333 better/increased replication efficiency 42  can influence each other, but it might be related to the specific reassortment event forming a common 340 ancestor for the cluster. Finally, the replacement of the avian-like swine internal gene cassette with an 341 H1N1pdm09 internal gene cassette, could enhance the zoonotic potential, as proposed for the American 342 H3N2v 44 and the British H1N2r 45 subtypes, which have resulted in several human infections. Therefore, the 343 pandemic potential of swIAV harboring gene segments of H1N1pdm09 origin should be a future research 344

focus. 345
Six novel reassortant swIAV subtypes and a total of 17 genotypes were identified during the eight-year 346 surveillance period. These findings underline the importance of having a national swIAV surveillance 347 program, which acts as an early warning system both for the swine industry and for the human health sector, 348 ensuring that novel subtypes and variants escaping current vaccines can be quickly identified. The 349 H3hu05N2sw subtype is a perfect example hereof, as it is a triple-reassortant swIAV including gene 350 segments from IAV of enzootic swIAV origin, H1N1pdm09 origin and human seasonal IAV origin 26 . 351 Surprisingly, this subtype has only been sporadically detected during the last five years. A possible 352 dissemination of this subtype among Danish swine herds would probably have devastating consequences, 353 because there is no population immunity towards the human seasonal H3hu05 26 . This indicates that other 354 factors than pre-existing immunity towards the HA protein are important for the spread of novel swIAV 355 subtypes and strains. Indeed, the most successful virus in Denmark during the last seven years has been the 356 H1N2dk, despite that there has been a high level of population immunity towards the HA protein of this 357 subtype since the 90's. Combined with the findings that the internal cassette of H1N1pdm09 origin seems to 358 benefit viral competitiveness, we might need to change our perception that pre-existing immunity to HA is 359 the main driver of evolution to focus also on the impact of the internal genes. Two other cases of human 360 seasonal IAV spillover into the swine population were observed during the surveillance, including the 361 H1avN2hu and H1pdmN2hu subtypes. Both subtypes contain the NA gene of a human seasonal IAV 362 circulating in the 90'ties 29 . The continued circulation of the H1pdmN2hu subtype in swine is worrying from 363 a zoonotic perspective, because all eight gene segments of this virus originates from viruses known to be 364 able to replicate in-and spread between humans. The circulation of the H1avN2hu subtype is even more 365 worrying, since there is no immunity against the HA protein of this subtype in the human population. The 366 H1avN2hu has gradually gained the internal cassette of H1N1pdm09 origin, meaning that some of these 367 viruses contain seven out of eight gene segments, which have been found in human IAV strains and thereby 368 may lead to increased zoonotic potential. Therefore, it is important to monitor the occurrence of these 369 subtypes in the futureboth in pigs and in humans. Another group of reassortant swIAV, that potentially 370 pose a problem for the swine herds, are those mixing the surface genes of enzootic swIAV and H1N1pdm09 371 subtypes. These novel reassortants includes H1pdmN2sw, H1pdmN1av and H1avN1pdm. Swine herds 372 experiencing infections with one of these three subtypes could potentially have a reduced effect of 373 vaccination, as no available vaccines currently include both the H1N1pdm09 subtype and the enzootic 374 swIAV subtypes. Thereby, these farms might need to apply two vaccines to reach an optimal immunity to the 375 circulating herd strain. 376

Genetic and antigenic drift 377
Another important aspect of swIAV evolution is the genetic drift, mainly affecting the two surface genes 378 (HA and NA) 46,47 . Especially the avian-like swine hemagglutinin protein (H1av) seem to have undergone 379 extensive genetic and antigenic drift, as a great sequence diversity was revealed. It was evident that the 380 evolution of the H1av gene did not evolve in one specific direction over time, but rather evolved in many 381 different directions, resulting in a vast number of different H1av clusters and variants. In a recent study 37 , we 382 found that the evolution of the H1av in a single herd followed a pectinate pattern mirroring the pattern seen 383 globally for the human seasonal influenza strains, which contradict the general perception that swine IAV is 384 not prone to selection driven by preexisting immunity like in humans. In the present study, we assessed the 385 H1av evolution over time in the Danish pig population as a whole and over several years and failed to 386 confirm this pectinate pattern. Thus, it seem that swIAV evolution at the single herd level is identical to the 387 pattern seen in the global human population, but when the swIAV evolution is evaluated on a national or 388 global scale, this pectinate pattern is disrupted. The reason for this difference is probably that the human 389 population, due to the extensive global interactions, can be regarded as a single "epidemiological unit", 390 whereas swine herds, due to a high level of external biosecurity and limited exchange of live animals 391 between herds, represents a vast variety of closed "epidemiological units", which each have a specific-and 392 probably pectinate-pattern of evolution. In other words, the global evolution of swIAV is characterized by a 393 vast number of different local clusters of viruses that on the herd level evolve similar to human seasonal 394 influenza viruses. This in turn results in very complex phylogenetic trees with a lot of clusters and 395 subclusters, which disrupt the pectinate structure. Still, despite the lack of a clear pectinate like evolution, the 396 H1av variants had clearly undergone positive selection on specific codons located in antigenic sites, which is 397 known to alter the binding of neutralizing antibodies 48-50 . This further confirm our previous findings, that the 398 herd immunity leads to significant antigenic drift in the globular head of the HA protein, as seen for human 399 seasonal IAV 51 . Furthermore, the finding of similar residues undergoing positive selection between different 400 herds, indicates that some residues in the HA protein are of particular importance for swIAV evolution. 401 Finally, the substitution rate estimated for H1av was similar to that documented in previous studies 52-55 , but 402 was lower than the substitution rate estimated for H1av in a single herd over time 37 . This emphasize that one 403 should differentiate when comparing evolutionary results based on data obtained from a single herd or data 404 obtained through extensive surveillance programs. 405 The H1pdm09 sequences analyzed in this study, revealed the existence of two groups of sequences. One 406 group of H1pdm09 sequences forming a well-defined cluster only containing sequences derived from swine 407 (Sw-L sequences) and another group of more diverse swine derived H1pdm09 sequences (Hu-L sequences) 408 that were scattered among human seasonal H1pdm09 sequences. In general, relatively long branches 409 separated the Danish Hu-L H1pdm09 swine sequences and the closest human sequences. However, a few of 410 the Hu-L sequences from Danish swine had a high level of identity to viruses isolated from humans during 411 the corresponding human influenza season, indicating a very recent "spill-over" from humans to pigs. 412 Indeed, all the Danish Hu-L viruses probably represents reverse zoonotic events, where the H1N1pdm09 413 virus was transmitted from humans to swine, and has started to evolve in pigs and by that has drifted away 414 from the human "seed" virus. The other group of H1pdmNx viruses found in Danish swine, formed a clearly 415 defined cluster (Sw-L cluster) that was different from the human seasonal H1N1pdm09 sequences. 416 Interestingly, this cluster also contained 11 viruses isolated from swine in Germany and one from Italy. This 417 is not surprising, since Denmark has an annual export of more than 10 million weaned pigs to mainly Eastern 418 Europe and Germany, which are not tested for swIAV prior to export. In contrast, all swine adapted 419 H1N1pdm09 viruses (SwD) found in France during recent years 41 formed another unique cluster that were 420 only distantly related to the "Danish-German" Sw-L cluster, confirming that the evolution of swIAV follows 421 different evolutionary traits in populations that are not epidemiologically connected. In summary, the presented data strongly indicate that the human seasonal H1N1pdm09 viruses still are 431 capable of infecting swine, despite more than ten years of adaption to humans, but it is unclear if the swine 432 adapted viruses of the Sw-L cluster also have retained its capability to infect humans. Studies to investigate 433 this in the ferret model are ongoing. 434 In comparison to the H1av sequences, the H1pdm09 sequences exhibited a lower level of sequence diversity, 435 probably because the H1pdmNx has circulated in Danish swine for significant shorter time than the H1avNx 436 strains. In contrast, the substitution rate and the positive selection on the RBS and antigenic sites were 437 comparable to that of the H1av sequences when the evolutionary analysis were performed on the Danish 438 antigenicity. This is highly important, as decreased cross protection between these two clusters would be 468 detrimental if the swine adapted virus jumps back into humans. Similarly having an IAV monitoring of 469 personal in affected herds should be considered. 470

Specific host and virulence markers 471
In summary, the HA proteins of the H1pdm09 viruses seem to be better adapted to elicit a strong receptor 472 binding to the α2.6-linked sialic acid receptor compared to the H1av HA proteins. This may reflect that the 473 H1pdm09 HA are descendants of the H1N1 "Spanish flu" strain 18 and by that have circulated in mammals 474 for at least 100 years, whereas the H1av HA protein was first detected in a mammal (pig) in the eighties' 56 . 475 In turn, these results could also explain why very few cases of zoonotic infection involving H1av have been 476 registered 57,58 . The fact that more Danish Hu-L sequences had "D" at position 225 support the assumption 477 that these H1pdmNx viruses are indeed more similar to human seasonal-like H1Npdm09 viruses compared 478 to the viruses of the Sw-L cluster and also indicate that the G225D transition may be more important in 479 humans than in swine. The comparison of swine H1pdm09 sequences and human seasonal H1pdm09 aa 480 sequences revealed that the residue at position 273 might be a potential marker important for distinguishing 481 between swine and human H1pdm09 sequences. However, this residue was not 100 % unique to one of the 482 two groups of sequences, and more studies should be performed to identify specific swine and human 483 markers of the H1pdmNx subtypes. In addition, the eight aa residues defined to differ between avian IAVs 484 and H1N1pdm09 origin viruses in the NP, PB1, PB2 and PA gene segments 59 were consistent with the 485 residues observed in the two clusters (avian-like swine and H1N1pdm09) of the NP, PB1, PB2 and PA genes 486 segments of this study. This suggests that these residues are indeed specific for H1pdmNx swIAV. 487 The recently identified residues 48Q, 98K and 99K of the Eurasian avian-like swIAV NP protein conferring 488 MxA resistance 60 , was documented in the majority (81%) of the Danish NP protein of avian-like swine 489 origin. MxA resistance is essential for zoonotic and pandemic potential of avian and swine IAV 60,61 , and 490 there is therefore a potential increased risk of zoonotic transmission in the Danish herds, where circulation of 491 swIAV strains carrying these three mutations is present. 492 As for the aa changes observed between the Sw-L and Hu-L sequences in the internal proteins the T76A 493 change in the PB2 protein has been linked an elevated interferon response 62 . Furthermore, the M283I aa 494 change in the PB2 protein has previously been linked to decreased virulence of avian H5 IAV 63,64 and the 495 N456S aa change has, on the other hand, been linked to human adaptations of the H3N2 subtype 65 . For the 496 PB1 protein, the M317I aa change has been identified in a H2N2 after multiple passaging in chicken eggs to 497 create a temperature sensitive IAV strain 66 . Finally, the C241Y in the PA protein has been linked to 498 mammalian adaptions of avian H5N1 viruses 67 . In summary, several of the aa changes observed between the 499 Sw-L and Hu-L internal proteins have previously been described to have an influence on the virulence, 500 replication efficiency or host-response/adaptation, thereby emphasizing that these changes could be 501 important in the adaption of H1pdm09Nx viruses to swine. However, this needs to be investigated further. 502

Importance of swIAV surveillance programs 503
The results generated in connection with the passive surveillance program of swIAV performed in Denmark 504 The total RNA from all sample types was eluted in 60µl RNase-free water and stored at -80 °C until further 538 analysis. Positive and negative controls were included in all extractions. 539

Detection of swIAV 540
The presence of swIAV was detected by an in-house modified version of a real time RT-PCR assay detecting 541 the M gene 69 . The assay was performed in a total reaction volume of 25 µl using the RNA Ultrasense One-542 Step Quantitative RT-PCR System (Invitrogen), 3 µl of extracted RNA, 300 nM forward primer (RimF), 600 543 Test for the HA gene of H1N1pdm09 origin by specific real time PCR 550 All swIAV positive samples were tested for the presence of the HA gene of H1N1pdm09 origin (H1pdm09) 551 by an in-house real time RT-PCR assay detecting specifically the HA gene of the pandemic virus (Table 1). 552 All reactions were analyzed in a Rotor-GeneQ machine (Qiagen) using the following PCR conditions: [ From 2011-2014, the swIAV positive samples were subtyped using Sanger sequencing of the HA and NA 560 genes according to a previously described PCR protocol 70 . The PCR products were purified using the High 561 Pure PCR product Purification Kit (Roche, Denmark). Subsequently the purified PCR products were sent for 562 sequencing at LGC Genomics (Berlin, Germany) with primers comprised of the "pQE" part of the PCR 563 primers (Table 1). 564 From 2015-2017, samples were subtyped using a multiplex real time RT-PCR assay strategy. Two multiplex 565 reactions including primers and probes for H3hu, N1pdm, H1av, N2hu or H3sw, H1pdm, N1sw, N2sw+hu, 566 respectively 34 (Table 1) were analyzed on the Rotor-GeneQ machine (Qiagen) using the following PCR Glutamin 2 mM, Non-essential amino acids 1x, 100 units/ml Penicillin, 100 µg/ml Streptomycin and TPCK-578 treated trypsin 2 µg/ml) using either 10 % lung tissue homogenate or 20 % nasal swab or oral fluid sample. 579 The inoculum was added to 70 % confluent MDCK cells for 45 minutes at 37 °C and 5 % CO2 followed by 580 the addition of fresh viral growth medium after wash of the inoculated cells. After 3 days, the cell culture 581 supernatant was harvested and tested for influenza A virus by real time RT-PCR. 582

Full genome sequencing 583
From 2013-2017 full genome sequencing was performed on cell culture-propagated influenza virus samples, 584 which had been subjected to full-length PCR amplification of all eight gene segments with in-house designed 585 primers (Table 1)  alignments, subsequent neighbor joining phylogenetic trees, and the function "BLAST against NCBI". 612 Moreover, sequence alignments of each lineage of the two surface gene segments (H1pdm09, H1av, N1pdm, 613 N1av, N2hu and N2sw) were analyzed for the average nucleotide diversity (pi) using author's own software. 614 In addition to the Bayesian phylogenetic analyses, strict molecular clock trees were constructed for the 625 surface gene segments of the lineages; H1pdm09, H1av, N1pdm, N1av, N2hu and N2sw to determine the 626 temporal evolution and the substitution rate. However, before the trees were constructed, all sequences were 627 investigated for the presence of a temporal signal (i.e., whether nucleotide changes accumulate roughly 628 proportionally to elapsed time) using the program TempEst 77 and evaluating the correlation coefficient. Subsequently, the alignments were manually examined to determine the presence of previously described aa 646 differences and residues. Specifically, for the HA proteins these included residues unique to the 647 H1pdmN2sw subtype 27 and residues linked to receptor binding 78,79 . Moreover, the five previously defined   Node labels represent posterior probabilities. A_California_2009 is the outgroup. The reference sequences 932 are named according to their given name in NCBI GenBank or GISAID and year of isolation. In addition, the 933 French swine derived sequences have the suffix "SwD" or "SeL". The red taxons with the suffix "Sw-L" 934 correspond to the Danish sequences of Cluster 1 - Fig. 6 and is now included in the "Sw-L" cluster. The 935 purple taxons correspond to the purple taxons of Fig. 6 and have been given the suffix "Hu-L".

Strain name
Accession # NCBI Genbank