Genetic and Antigenic Evolution of European Swine Influenza A Viruses of HA-1C (Avian-Like) and HA-1B (Human-Like) Lineages in France from 2000 to 2018

This study evaluated the genetic and antigenic evolution of swine influenza A viruses (swIAV) of the two main enzootic H1 lineages, i.e., HA-1C (H1av) and -1B (H1hu), circulating in France between 2000 and 2018. SwIAV RNAs extracted from 1220 swine nasal swabs were hemagglutinin/neuraminidase (HA/NA) subtyped by RT-qPCRs, and 293 virus isolates were sequenced. In addition, 146 H1avNy and 105 H1huNy strains were submitted to hemagglutination inhibition tests. H1avN1 (66.5%) and H1huN2 (25.4%) subtypes were predominant. Most H1 strains belonged to HA-1C.2.1 or -1B.1.2.3 clades, but HA-1C.2, -1C.2.2, -1C.2.3, -1B.1.1, and -1B.1.2.1 clades were also detected sporadically. Within HA-1B.1.2.3 clade, a group of strains named “Δ146-147” harbored several amino acid mutations and a double deletion in HA, that led to a marked antigenic drift. Phylogenetic analyses revealed that internal segments belonged mainly to the “Eurasian avian-like lineage”, with two distinct genogroups for the M segment. In total, 17 distinct genotypes were identified within the study period. Reassortments of H1av/H1hu strains with H1N1pdm virus were rarely evidenced until 2018. Analysis of amino acid sequences predicted a variability in length of PB1-F2 and PA-X proteins and identified the appearance of several mutations in PB1, PB1-F2, PA, NP and NS1 proteins that could be linked to virulence, while markers for antiviral resistance were identified in N1 and N2. Altogether, diversity and evolution of swIAV recall the importance of disrupting the spreading of swIAV within and between pig herds, as well as IAV inter-species transmissions.


Introduction
Influenza A viruses (IAVs) are members of the Orthomyxoviridae family and are composed of eight negative single-stranded RNA segments. The lack of RNA polymerase proofreading activity during virus replication produces a continuous diversity of variants and the segmented genome allows gene reassortment between virus strains in case of co-infection [1]. Inter-species transmissions also play an important role in IAV evolution. They provide new sources of diversity, either following adaptation to a

Virus Isolation
All M-gene positive samples with Ct < 30 were selected for virus isolation (only one sample per farm at a given sampling date), except for supernumerary H1 av N1 viruses from the Brittany region for which we selected a representative panel based on location and period of detection. Selected M-gene positive samples were subjected to virus propagation by one passage onto 9-day-old specific pathogen-free (SPF) embryonated chicken eggs or Madin-Darby Canine Kidney (MDCK) cells, according to standard procedures [28]. For swine antisera production and in some cases for antigenic characterization, reference virus strains and/or swine H1Ny isolates first isolated on MDCK cells were further propagated by a second passage on eggs or MDCK cells. Harvested cell culture supernatants and allantoic fluids were clarified by centrifugation and stored at −70 • C until sequencing, pig inoculation, or hemagglutination inhibition assays.
Additional virus strains propagated on embryonated chicken eggs from biological samples, all taken in 2000-2004, were kindly provided by LABOCEA 22, Ploufragan. These allantoic fluids were submitted to molecular subtyping, as described above, before sequencing.

Sequencing
Viral RNA was extracted from viral cultures using the NucleoSpin ® RNA Kit (Macherey-Nagel, Düren, Germany). The nucleotide sequences were obtained either by the Sanger method on an Applied Biosystems ® Sanger Sequencing 3500 Series Genetic Analyzer (Hitachi High Technologies, Naka, Japan) or by next-generation sequencing (NGS) on Thermo Fisher Scientific's Ion Proton instrument (Thermo Fisher, Carlsbad, CA, USA). The reads obtained by NGS were cleaned with Trimmomatic 0.36 software [29], then assembled by mapping on reference genomes using Burrows-Wheeler Aligner software (version 0.7.15-r1140 [30]) and by de novo using the SPAdes and MIRA programs (versions v3. 10.0 [31] and 4.0.2 [32]). A single consensus sequence per viral segment was generated by combining the contigs produced using both methods with Vector NTI Advance 11.0 software (Invitrogen Corp., Carlsbad, CA, USA). Sequences obtained by NGS for this study were available in GenBank: Bioproject # PRJNA623701. Genbank identification numbers of the sequences used in this study are reported in Supplementary Table S1.

Phylogenetic Analyses of HA-, NA-and M-Encoding Genes
The sequences obtained in this study, and sequences of other swIAVs isolated in France and previously deposited in public databases, were compared to swIAV sequences retrieved from the Influenza Research Database (IRD) (available at https://www.fludb.org/). Phylogenetic trees were generated by a Bayesian approach using BEAST v1.8.4 [33], with the substitution models 'HKY + Gamma + Codon Position (CP)' for HA-and NA-genes or 'GTR + Gamma + CP' for M gene, improved by the sample collection years [34]. A Markov chain Monte Carlo (MCMC) method of 10,000,000 samplings for every 1000 was performed. The quality of the posterior distribution of each setting was checked through the Effective Sample Size (ESS) with Tracer v1.6 [33], with 10% of burning. A minimum ESS value of 150 was accepted for prior when the values were over 300 for posterior, likelihood and clock-rate settings. Trees were summarized applying 10% of burning with TreeAnnotator v1.8.4 [33] and visualized with FigTree v1.4.3 [35]. The HA clades were defined according to the global swine H1 clade scheme [7], using the 'Swine H1 Clade Classification Tool' accessible in IRD.

Large-Scale Phylogeny of M-Encoding Segment
A large-scale phylogenetic analysis was performed to evaluate the origin of the divergent phylogenetic clusters identified among M gene sequences from French swIAVs. After checking that the M gene sequences obtained in this study formed a distinct cluster from M genes of avian IAV lineages as previously described [36], we focused the large-scale analysis on all sequences of M segments from swIAV isolated in the Northern hemisphere, and human IAV isolated in Europe when associated with a known HA/NA subtype (sequences available in IRD at the 16th of November 2017). In total 12,386 complete M gene sequences were selected after cleaning, representing 7154 M-gene sequences of swIAVs from the Northern hemisphere and 5232 European human IAV strains. The sequences were aligned with MAFFT v7.311 [37], then a cluster analysis was performed with R 3.4.0 using the packages 'phyclust' and 'ape'. The phylogenetic clustering was defined by the main function 'phyclust', which uses an evolutionary Continuous Time Markov Chain model-based approach to identify population structure from DNA data.

Amino Acid Sequence Analyses
DNA sequences were translated into amino acid sequences with Seaview 4.6.1 [38]. For all segments, amino acid positions were numbered from the first methionine. Thus, the amino acid numbering included the signal peptide for HA. The variability of peptides in protein sequence alignments was measured by the entropy values with the Antigen Variability ANAlyzer 0.31 (AVANA) software with default parameters [39]. Genogroup-specific amino acid mutations were identified using two methods: (i) the genogroup pairwise comparison in AVANA to identify all genogroup-specific mutations without a priori; (ii) the R 3.4.0 analysis only of residues identified to be involved in change of pathogenicity, virulence or fitness of IAVs, according to the literature (Table S2). Antigenic and receptor-binding sites in HA of H1 viruses were determined according to the literature [40,41], as well as those of NA which were deduced from descriptions of N2 viruses [42,43].

Antigenic Characterization
Swine antisera against swIAV strains representative of European enzootic lineages of H1 subtype and/or H1 swIAVs isolated in France during the study period were produced in SPF pigs at ANSES facilities (see [25] for details) ( Table 1). Strain A/Sw/England/117316/86, a representative of the formal "classical swine H1N1" (clswH1N1) lineage which was transiently isolated in Europe in the 1980s, was also included, as it belongs to the HA-1A clade as H1N1pdm strain. Hemagglutination inhibition (HI) assays were performed according to standard procedures [28]. Briefly, four hemagglutinating units (HAU) of egg-propagated virus were incubated with two-fold dilutions (starting at dilution 1:10) of swine antiserum and tested against 0.5% chicken red blood cells. HI antibody titers were expressed as the reciprocal of the highest dilution inhibiting 4 HAU of virus. The means of HI antibody titers obtained by groups of strains were compared with R 3.4.0 using the function 'pairwise.wilcox.test' on log-transformed HI titers without adjustment. Difference of means was significant for p < 0.05. The H1 antibody titers <10 were replaced by 1 within the test to give no weight in the mean of log-transformed value.

Relative Proportions of H1 av Ny and H1 hu Ny swIAVs in France between 2000 and 2018
In this study, 1220 swIAVs from as many independent respiratory outbreaks that occurred from 2000 to 2018 in pigs reared in mainland France were HA/NA subtyped molecularly. From 2000 to 2009, less than 25 virus strains were subtyped each year (N = 133). The HA-and NA-specific RT-PCRs allowed discriminating 67 H1 av N1 strains (50.4%), 59 H1 hu N2 strains (44.4%), five reassortant H1 hu N1 strains (3.8%) and two reassortant H1 av N2 strains (1.5%), but no H3N2 strain. From 2010, the yearly number of subtyped viruses increased more than five-fold thanks to the establishment of the national swIAV surveillance network. In total, 1087 swIAVs were subtyped between 2010 and 2018: 744 H1 av N1 strains (68.4%), 251 H1 hu N2 strains (23.1%), 54 H1N1pdm strains (5.0%), 25 reassortant H1 av N2 strains (2.3%), seven reassortant H1 hu N1 strains (0.6%), five H3N2 strains (0.5%) and one reassortant H1pdmN1 (0.1%). No reassortant of H3N1 subtype, or reassortant exhibiting the H1 av -or H1 hu -encoding gene with the N1pdm segment were detected. Thus, viruses with an HA belonging either to the H1 av or the H1 hu lineages represented more than 95% of the swIAVs identified from 2000 to 2018. Despite H1N1pdm introduction in 2010 [25], H1 av N1 swIAVs have remained the predominant viruses since 2004, with an annual frequency varying from 40% to 77%, while H1 hu N2 viruses counted for 15% to 48% ( Figure 1). Hemagglutination inhibition (HI) assays were performed according to standard procedures [28]. Briefly, four hemagglutinating units (HAU) of egg-propagated virus were incubated with two-fold dilutions (starting at dilution 1:10) of swine antiserum and tested against 0.5% chicken red blood cells. HI antibody titers were expressed as the reciprocal of the highest dilution inhibiting 4 HAU of virus. The means of HI antibody titers obtained by groups of strains were compared with R 3.4.0 using the function 'pairwise.wilcox.test' on log-transformed HI titers without adjustment. Difference of means was significant for p < 0.05. The H1 antibody titers <10 were replaced by 1 within the test to give no weight in the mean of log-transformed value.
Finally, three strains with H1 from clade 1C.2.2 were identified in the Eastern part of France ( Figure 3A). They harbored 22 mutations in HA (data not shown), including S179I in an antigenic site.

Figure 2.
Bayesian inference tree of HA-1C (H1avNy) swIAV strains. Strains isolated in pigs in France are indicated in black type whereas those isolated in pigs in other countries are in grey type. Nodes supported by more than 50% of the sampled trees are indicated by a blue bar displaying the 95% credibility intervals of highest posterior density for the node heights. HA-1C clades are defined according to the classification of Anderson et al., 2016 [7]. The three boxes on the right grid are colored according to the genogroups defined on genomic segments 4 (HA), 6 (NA) and 7 (M) phylogenies, respectively. For segment 6, the boxes were colored in orange for N1 'Eurasian avian-like' lineage, in blue for N2 'Scotland-like' lineage, in yellow for N2 of 'H3N2 human-like' lineage, and in green for N2 of 'Gent-like' lineage (see Section 3.4.1). For segment 7, the boxes were colored in brown for 'M ancestral Eurasian avian-like swine' group, in red for 'M recent Eurasian avian-like swine' group, in green for 'M European swine' group, in yellow for 'M Danish-like' group, and in purple for M segment from the H1N1pdm lineage (see Section 3.4.2).
Within the HA-1C.2.1 clade, three genogroups named A, B and C, respectively, can be pointed out, based on large numbers of substitutions on non-silent sites compared with A/Sw/Cotes d'Armor/0388/09 (CA0388/09), a representative strain for clade 1C.2.1, and/or geographical locations. Nodes supported by more than 50% of the sampled trees are indicated by a blue bar displaying the 95% credibility intervals of highest posterior density for the node heights. HA-1C clades are defined according to the classification of Anderson et al., 2016 [7]. The three boxes on the right grid are colored according to the genogroups defined on genomic segments 4 (HA), 6 (NA) and 7 (M) phylogenies, respectively. For segment 6, the boxes were colored in orange for N1 'Eurasian avian-like' lineage, in blue for N2 'Scotland-like' lineage, in yellow for N2 of 'H3N2 human-like' lineage, and in green for N2 of 'Gent-like' lineage (see Section 3.4.1). For segment 7, the boxes were colored in brown for 'M ancestral Eurasian avian-like swine' group, in red for 'M recent Eurasian avian-like swine' group, in green for 'M European swine' group, in yellow for 'M Danish-like' group, and in purple for M segment from the H1N1pdm lineage (see Section 3.4.2). strains also presented many mutations in antigenic sites: L88I, PNH->LSY at positions 141-143, H155N, G172D, N173S, K180Q, T183K, K186R, D202Y and G219V for the Southwestern strains; P141E, V151G, A152S, H155R, A158T, N159K, L178I, N185D, K186R, D202Y, I207A, N211S, G219V and S220T for the Northwestern strain.
Finally, three strains with H1 from clade 1C.2.2 were identified in the Eastern part of France ( Figure  3A). They harbored 22 mutations in HA (data not shown), including S179I in an antigenic site.   Table 2). It could be noted that the serum produced against a H1 av N2 reassortant virus of the same HA clade, i.e., HIS-CA0186/10, showed HI titers significantly weaker against strains of clade HA-1C.2.1 than HIS-CA0388/09 and HIS-Mo0070/05 ( Table 2). The relationships between the HI antibody titers obtained by HIS-CA0388/09 against each strain of clade HA-1C.2.1 and the observed mutations in antigenic sites or RBS of the tested strains were analyzed with a generalized linear model. This analysis revealed a statistical link between the decrease in HI antibody titers and the appearance of mutation N142D or K180N in HA-1C.2.1 strains. Their prevalence increased from 2014, with an individual estimated effect of -1.8 log for HA-N142D and -1.5 log for HA-K180N on HI antibody titer of HIS-CA0388/09. Interestingly, HIS-CA0388/09 exhibited higher mean HI titers against strains from groups A and C (427.15 and 470.32, respectively) as compared to strains from group B or other strains belonging to clade HA-1C.2.1 and isolated between 2015 and 2018 (246.75 and 265.52, respectively) ( Table 2). Although strains from group B harbored mutations in antigenic sites, the HI antibody titers of HIS-CA0388/09 remained similar to those against HA-1C.2.1 parental strains named 'others' in Table 2. HIS-65-150242 (H1 av N2, clade HA-1C.2), but not HIS-CA0186/10 (H1 av N2, clade HA-1C.2.1), inhibited the hemagglutination properties of three H1 av N2 strains from clade HA-1C.2 (Table 2). Finally, HIS containing antibodies directed against H1N1pdm (clade HA-1A.3.3.2) and H1 hu N2 (clade HA-1B.1) virus strains showed no or low HI antibody titers against H1 av Ny strains, whatever the HA-1C clade they belong to (Table S3A). Table 2. Cross-hemagglutination inhibition (HI) data obtained for H1 av Ny swIAVs isolated in France in 2000-2018 against hyperimmune swine antisera directed towards HA-1C representative strains. The table reports the HI titers for reference antigens (homologous HI titers are underlined), then the geometric mean HI titers +/− the standard deviation (in bold) obtained by strains grouping in the different HA-1C (H1 av ) clades over time. Results of the Wilcoxon pairwise comparison test are reported by a letter in superscript, a different letter indicating a significant difference between groups in the row and column (p < 0.05). HI titers of subgroups within 1C.2.1 were tested independently of the other rows. HI titers for HA-1C.2 and 1C.2.2 clades were not included in the statistical test because the sample sizes were below five. The range of HI antibody titers obtained within each group of tested strains is reported in square brackets with the number of tested strains in italics. Raw data for individual studied strains are available upon request.

Genetic and Antigenic Evolution of HA-1B (H1 hu Ny) swIAVs Isolated in France
Between 2000 and 2018, 111 out 324 H1 hu Ny strains were sequenced. Eighty-seven of them (78 H1 hu N2 and 9 H1 hu N1), for which at least HA-and NA-encoding genes were completely sequenced, were included in the phylogenetic analyses.
The first strains belonging to the European H1 hu N2 lineage that were isolated in France and sequenced dated back to 1997 [24].  Figures 3B and 4). Since 2016, some H1 hu Ny virus strains were also identified in the north of France but they were found to belong to clade 1B.1.2.1, which groups swIAV strains mostly isolated in other European countries ( Figures 3B and 4).
All N1 segments of HxN1 swIAVs isolated in France in 2000-2018 derived from the so-called "Eurasian avian-like" (EA) N1 lineage, except those from the H1N1pdm lineage that has been described previously [25] (Figure 5). Most EA-N1 segments were associated with HA-1C genes in H1 av N1 strains (Figure 2), and only a few with HA-1B gene in H1 hu N1 reassortants detected from 2005 to 2012 (Figure 4). Congruently with H1 av phylogeny, the strains from HA-1C.2.3, 1C.2.1-B and 1C.2.1-C phylogenetic groups formed three clusters within the EA-N1 lineage, mentioned as N1-α, -β and -δ, respectively ( Figure 5). Comparison of deduced amino acid sequences revealed that NA genes in group N1-α were distinguished from other N1-encoding sequences by 25 group-specific mutations of which four were located in sites predicted to be antigenic determinants for NA, i.e., S339F, H341N, V394I and I396M. In N1-β group, the encoded NA were characterized by 16 amino acid mutations including three in sites described as antigenic determinant, i.e., R331G, H341N/D and T397N/S, and one mutation creating an additional glycosylation site, NA-S88N. Strains belonging to the N1-δ group were characterized by seven mutations including N/S385K in a predicted antigenic site. No compensatory mutation was identified in N1 amino acid sequences to optimize HA/NA balance when they were associated with a HA-1B in reassortant H1 hu N1 strains.
The N2 segments sequenced in this study were associated either with H1 hu (HA-1B) or H1 av (HA-1C) genes in H1 hu N2 or H1 av N2 strains, respectively (Figures 2 and 4). Phylogeny of N2 segment exhibited more genetic diversity than those for N1 segments as they classified into three distinct lineages ( Figure 6). The majority of H1N2 strains grouped in the "A/Sw/Scotland/410440/1994-like lineage" which derived from the original H1 hu N2 strain in Europe ( Figure 6). Within this N2 "Scotland-like" lineage, strains harboring H1 hu ∆146-147 genes formed a cluster and were characterized by two specific amino acid mutations in N2, i.e., I50M and V317I. A second N2 lineage derived from human-like reassortant swine H3N2 strains belonging to the "A/Gent/1/1984-like lineage". This N2 "Gent-lineage" comprised H1 hu N2 strains isolated since 2016 in the northern part of France, as well as reassortant H1 av N2 strains with a HA-1C.2. Within this lineage, French H1N2 strains had N2 genes closer to those described in Danish enzootic H1 av N2 viruses than to those of Eurasian swine H3N2 viruses ( Figure 6). Finally, three reassortant H1 av N2 strains shared a N2 gene derived from a human seasonal H3N2 lineage, as detailed in a previous study [44] (Figure 6).
Analysis of residues involved in IAV resistance to antiviral drugs identified only two H1 av N1 strains that each naturally acquired a mutation, assessed to impact the efficacy of NA inhibitors, NA-Y155H classified as highly reduced inhibition (HRI) to oseltamivir and zanamivir, and NA-S247N classified as reduced inhibition (RI) to oseltamivir for H5N1 viruses, according to the list of the WHO [45]. By contrast, 80% of H1 hu N2 strains harbored mutation NA-S331R, also classified as RI to oseltamivir and zanamivir [45]. The H1 av N2 strains with N2 of human seasonal origin were the only ones harboring mutation NA-N329K classified RI to oseltamivir and zanamivir [45]. Finally, mutations Y347K and V149I in HxN1 strains and mutation K249R in HxN2 strains were located in amino acid sites described as impacting the efficiency of NA inhibitors, but their resistance phenotype has never been tested.
Viruses 2020, 12, x FOR PEER REVIEW 17 of 31 Figure 5. Bayesian inference tree of NA-N1 segment of HxN1 swIAV strains. Strains isolated in pigs in France are indicated in black type, those isolated in pigs from other countries in grey type. Nodes supported by more than 50% of the sampled trees are indicated by a blue bar displaying the 95% credibility intervals of highest posterior density for the node heights. Vertical colored bars in right correspond to the color reported in cases of column #6 in the grids of HA trees (Figures 2 and 4). Nodes supported by more than 50% of the sampled trees are indicated by a blue bar displaying the 95% credibility intervals of highest posterior density for the node heights. Vertical colored bars in right correspond to the color reported in cases of column #6 in the grids of HA trees (Figures 2 and 4). Nodes supported by more than 50% of the sampled trees are indicated by blue bar displaying the 95% Figure 6. Bayesian inference tree of NA-N2 segment of HxN2 swIAV strains. Strains isolated in pigs in France are indicated in black type, those isolated in humans or pigs in other countries in grey type. Nodes supported by more than 50% of the sampled trees are indicated by blue bar displaying the 95% credibility intervals of highest posterior density for the node heights. Vertical colored bars at the right correspond to colors reported in cases of column #6 in the grids of HA trees (Figures 2 and 4).

Evolution of M-Encoding Segment
Complete sequences of M-encoding genes were obtained for 170 H1Ny strains (100 H1 av N1, 12 H1 av N2, 54 H1 hu N2 and 4 H1 hu N1) isolated from 2000 to 2018 and included in the analysis. The M-encoding genes of French swine H1Ny strains segregated into several genogroups (Figure 7), that were defined from a large scale phylogenetic analysis ( Figure S1). Two H1 av N2 strains with a Viruses 2020, 12, 1304 18 of 27 HA-1C.2 harbored a M segment closer to those of Danish enzootic H1 av N2 strains (M 'Danish-like' group) than of other H1Ny strains isolated in France, and congruent with the NA segment. Three reassortant H1 av N2 strains harbored a M segment from the H1N1pdm lineage (Mpdm). Two of these, with a N2 from human seasonal IAV, acquired only the M segment from the H1N1pdm, whereas the third one harbored all internal segments from the H1N1pdm lineage, as detailed in previous studies [44,46]. Finally, the majority of French swine H1Ny strains classified into two main genogroups we called the "M recent Eurasian avian-like swine" group (MswEA-'recent') and the "M European swine" group (MswEU), respectively, which showed no specific associations with HA or NA clades (Figures 2 and 4). France are indicated in black type, those isolated in pigs from other countries in grey type. Nodes supported by more than 50% of the sampled trees are indicated by blue bar displaying the 95% credibility intervals of highest posterior density for the node heights. Vertical colored bars at the right correspond to M genogroups with the colors reported in cases of column #7 in the grids of HA trees (Figures 2 and 4).

Figure 7.
Bayesian inference tree of M-gene segment of H1Ny viruses. Strains isolated in pigs in France are indicated in black type, those isolated in pigs from other countries in grey type. Nodes supported by more than 50% of the sampled trees are indicated by blue bar displaying the 95% credibility intervals of highest posterior density for the node heights. Vertical colored bars at the right correspond to M genogroups with the colors reported in cases of column #7 in the grids of HA trees (Figures 2 and 4).

Evolution of Other Internal Protein-Encoding Genomic Segments
Phylogenetic analyses of polymerase (PB2, PB1 and PA), nucleoprotein (NP) and non-structural (NS) protein-encoding segments of H1 av and H1 hu swIAVs identified in France until the end of 2018 showed that they all derived from the "Eurasian avian-like swine" lineage whatever the H1 lineage or clade (data not shown), excluding the reassortant strain with internal segments from H1N1pdm.
The deduced amino acid sequences of the internal proteins of the H1 av and H1 hu swIAVs were overall conserved, with a mean entropy value of 0.077, considering that a residue entropy exceeding 1.0 is highly variable and the maximum value is 4.322 (log 2 20). In any case, the PB1-F2, PA-X, M2 and NS1 proteins showed the greatest residue variability with less than 50% of conserved residues and mean entropy values that were 1.5-to 5-fold higher than the overall mean (Table 4). Added to their weak residue conservation (8.9%) (Table 4), the PB1-F2 proteins presented a wide variation in length. Almost 40% of the 162 amino acid sequences deduced for strains isolated from 2000 to 2018 were truncated before the 90th amino acid, and harbored a stop codon at positions 9 (0.6%), 12 (3%), 26 (5.6%), 58 (18.5%), 80 (5.6%) or 88 (6.2%) after the methionine. PB1-F2 proteins of strains with an H1 av belonging to HA-1C.2.3 clade were all truncated at position 12, those with HA-1C.2.2 and HA-1B.1.2.1 at position 80, and those with HA-1C.2.1-B at position 88. Three H1 av N1 strains from HA-1C.2.1 clade (1.8%) probably did not even encode PB1-F2 protein since they had a threonine instead of a methionine at the first position.
Almost all French swIAVs presented the PA-X full length pattern that was associated with highly virulent strains [49]. Only three strains (1.8%) of H1 hu N1, H1 hu N2 and H1 av N2 lineages, respectively, had a PA truncated at residue 42 in the X-domain, as in PA from H1N1pdm strains. Moreover, seven strains were found to have additional amino acids as compared to those encoded by the commonly described full-length 61-codon X-ORF. Five of them, including the three reassortant H1 av N2 strains with a N2 from seasonal human IAV, harbored a base substitution that transformed the stop codon into a cysteine codon at position 253 followed by seven additional residues (MPGLNHS). Two other strains, of H1 hu N2-∆146-147 and H1 av N1 lineages, had a leucine codon at position 253 and encoded three or seven additional residues (MPG or MGPLNHS), respectively.
Comparison of amino acid sequences, as well as research of molecular markers for virulence, allowed the identification of mutations in internal segments that were statistically linked to HA gene phylogenetic clusters. In addition to PB1-F2 truncation, all the H1 av strains from HA-1C.2.1-B clade and the two H1 av N2 strains from HA-1C.2 clade isolated in 2015 harbored the mutation N66S in PB1-F2, described as a virulence marker in several IAV subtypes [50]. H1 av strains from HA-1C.2.1-B clade were also characterized by 40 other co-occurring amino acid changes in internal proteins (Table S4), including mutation D92E in NS1 that was suspected of conferring virulence and resistance to IFN-α Viruses 2020, 12, 1304 20 of 27 of H5N1 virus strains in pigs [51]. H1 av strains from HA-1C.2.1-A clade were characterized by ten co-occurring mutations (Table S4), including PB1-F2-E87G described to decrease lethality of H5N1 strains in ducks [50]. H1 av strains from HA-1C.2.1-C clade were characterized by 29 co-occurring mutations (Table S4), including PB1-S654N and PA-T343A that were associated with a poor fitness of A/Puerto Rico/8/34(H1N1) and H1N1pdm virus strains, respectively [52,53]. H1 av strains from 1C.2.2 clade were characterized by PB1-K353R and PA-S409N mutations reported to increase virulence and transmission of several IAV subtypes [54,55]. Finally, within the H1 hu lineage, strains from clade 1B.1.2.1, isolated in the northern France, were characterized by two mutations in NP that were described as increasing virulence and resistance to MxA. The first, mutation NP-Y289H, was also observed in reassortant H1 av N2 strains with a N2 from seasonal human IAV. The second mutation, NP-Q357K, was also observed in H1 av strains isolated in the same geographic area [56,57].

Overview of H1 av and H1 hu Genotypes Identified in France from 2000 to 2018
Taking into account the clade and/or lineage identified for each genomic segment, 17 distinct genotypes were identified from the 161 H1 av and H1 hu swIAV strains fully sequenced in this study (Table 5). Among them, 16 genotypes covered eight genotypes previously described by Watson et al. ([13], Table 5).

Discussion
SwIAV strains of European enzootic H1 av N1 and H1 hu N2 lineages were quickly introduced into the pig population in France after their emergence in 1979 and 1994, respectively [8][9][10]. Between 2000 and 2010, the epidemiological surveillance of swIAVs was poorly organized and few samples were collected. The IAV pandemic in 2009 led to the implementation of a national surveillance network which, combined with improvement of molecular subtyping and sequencing techniques, has permitted the production of a sharper picture of the evolution, diversity and distribution of swIAV strains in France.
H1 av N1 was predominant in France between 2000 and 2018 (66.5%), as in most European countries except the United Kingdom and Denmark, according to the European studies conducted between 2010 and 2017 [12,13,21]. The H1 av Ny strains identified in pigs in France clustered mainly in the HA clade 1C.2.1, described in Europe and Russia [7]. Only a few strains isolated in the East of the country carried an HA from clade 1C.2.2, which was detected in other European countries such as Germany, Italy, Luxembourg, Netherlands, Poland and Spain [7]. Three H1 av N2 strains with an HA belonging to the 1C.2 clade, similar to contemporary strains circulating in Denmark, were detected in France. The first two were observed in Southwestern France in spring 2015 and were close to the "Danish-like lineage" in all other segments, whereas the third was detected in Northwestern France in 2018 and had internal segments of H1N1pdm [46,58]. These sporadic detections during the study period sustain the hypothesis that these strains were introduced thanks to live pig importations from abroad. H1 hu N2 was the second most frequent subtype (25.4%) detected in pigs in France in 2000-2018. Its proportion was similar to those observed between 2010 and 2013 in Italy, Spain and Netherlands, but twice than that reported in the United Kingdom and Germany at that time [12,18]. However, European H1 hu Ny subtypes classified in different HA clades according to their geographical location [7]. In France, between 1997 and 2016, viruses from the H1 hu N2 subtype were only detected in the north-west. They belonged mainly to the HA-1B.1.2.3 clade, which was found only in France, in association with a NA segment from the A/Sw/Scotland/410440/1994-like lineage. From 2017, new H1 hu N2 strains were detected in the north of France. They harbored an H1 from 1B.1.2.1 clade and a NA segment from A/Sw/Gent/1984-like lineage, similarly to strains described in Belgium, Germany, Netherlands and Denmark [7,13]. Commercial exchanges and human movements between these border countries could explain the presence of this lineage in this part of France.
Within the HA-1B.1.2.3 clade, several groups of strains diverged from A/Sw/Cotes d'Armor/790/97 but co-circulated in the same geographical area. The more recent strain belonging to group D was detected in 2013, which suggests that this group is rare or no longer present in the territory, maybe due to a lack of competitiveness with other strains from HA-1B.1.2.3 clade. Interestingly, most strains since 2012 harboring the double 146-147 deletion and associated mutations in HA also formed a phylogenetic cluster. A specific RT-qPCR was developed by the French NRL for to discriminate H1 hu ∆146-147 within H1 hu strains [26]. This tool revealed that H1 hu ∆146-147 strains quickly spread until reaching approximately 54% of H1 hu viruses detected in pigs in France in 2013-2014, a proportion that further decreased to 18% in 2018 (data not shown). Interestingly, the same deletion into the RBS has repeatedly been described in swIAV isolates and observed within various HA clades: HA-1B.1.2.2 in Italy [59], HA-1B-Other-human clade in Russia [60] and the "classical swine clade" HA-1A. 1.3 (H1α3) in Canada [61]. One of the two residues, either 146 or 147, was also deleted in the three HA-1C.2 strains identified in France. Fixation of the same amino acid deletions into RBS of HA from the three 1A, 1B and 1C lineages is not yet explained but has been investigated elsewhere to understand the impact on virus fitness, pathogenicity and immune escape [62].
Interestingly, the regionalization of HA clades, either of H1 av or H1 hu lineages, could be explained by the limited number of pig movements through the French territory as compared to other countries. The majority of farms perform farrow-to-finish production, and the leading destination of movements is slaughterhouses/rendering plants (75.2%), followed by exchanges between farms (22.8%) and trade operators (2.0%) [63]. Movements of live pigs between European countries in both import and export represent only 1.4% of transport rounds [63].
Antigenic drift for swIAVs was previously estimated to be of 0.15 and 0.17 antigenic units (AUs) per year for H1 av and H1 hu lineages, respectively [64], which is weak compared to human IAV strains (0.6 to 1.2 AU/year [65,66]). Assuming that an antigenic distance above 2 AU, or a four-fold difference in HI assay titers, is significant and considered as an antigenic drift [67], it was expected that genetic swIAV evolution would generate a significant antigenic drift every 12-13 years. In the present study, such an antigenic distance was observed between the newly introduced H1 av strains HA-1C.2 and the predominant strains HA-1C.2.1, as well as between the H1 hu ∆146-147 strains and other H1 hu strains. However, a slight stepwise antigenic drift could be noted within H1 av 1C.2.1 lineage every four years, which could be associated with specific mutations in RBS and antigenic sites, e.g., HA-N142D and K180N, rather than gradual accumulation of mutations. Such punctual mutations in HA as those described in this study could lead swIAV to escape vaccine protection quicker than expected, suggesting that vaccine antigens should be regularly and/or locally updated. In France, vaccination is applied in a little less than half of the breeding herds [22]. During the study period, the adjuvanted inactivated trivalent vaccine Respiporc FLU3 ® (Ceva, Libourne, France, formerly IDT-Biologika, Dessau-Roßlau, Germany) was the only vaccine that has been issued [22]. Respiporc FLU3 ® contains antigens representative of European H1 av N1, H1 hu N2, and H3N2 lineages, which have not been changed since the vaccine has been launched in 2010.
Exchange of HA-and NA-encoding segments were regularly observed during the studied period. However, H1 hu N1 viruses (n = 7) have seemed rarer than H1 av N2 viruses (n = 25) since 2010. The co-evolution of HA and NA segments, linked to the maintenance of the HA-NA balance [68], may explain in part the limited number of viruses that exchanged their glycoproteins. Among H1 av N2, N2 segment was generally provided by enzootic H1 hu N2 viruses (N2-Scotland lineage) but not swine H3N2 viruses (N2-Gent lineage), in line with their frequency and area of co-circulation with H1 av N1 viruses. Indeed, swine H3N2 subtype is rare in France, with less than one isolate detected each year on average, and furthermore restricted to the north (data not shown), which limits reassortment opportunities. However, three H1 av N2 strains were shown to contain a N2 gene from a seasonal human H3N2 lineage, recalling the importance of implementing biosecurity measures in holdings to avoid inter-species transmissions that may lead to the emergence of new viruses in pigs [44].
In contrast to other European countries in which swIAV diversity has been strongly impacted by the transmission of the H1N1 pandemic virus to pigs in 2009, the genotype diversity of French swIAVs has not changed greatly due to this new introduction between 2009 and 2018 [13,[15][16][17][18][19]. The H1 av and H1 hu viruses were widely established for years in northwestern France and might have hampered the H1N1pdm spread in French herds located in this high pig density area [22,25]. Thus, co-infections of pigs with H1 av or H1 hu strains and H1N1pdm virus would have remained a rare event, which would explain the limited number of reassortant strains between these lineages. Only three H1 av N2 strains harboring internal segments from the H1N1pdm lineage have been identified in our study. Given their detailed genotypes, two of them, containing only the Mpdm, would have been generated on the territory [44]. However, the third, with a HA from clade 1C.2, a N2-Gent and all the internal genes from H1N1pdm, would have been introduced in Brittany in toto from abroad in 2018 [46]. Four other reassortant strains of H1 av N1 subtype, containing one or several internal segments from H1N1pdm, have been recently reported using SYBR-Green RT-qPCRs [21]. Thus, at least three additional genotypes would complete the list of the seventeen we have discriminated here. Whereas these reassortant detections have remained sporadic, the risk of reassortment events could increase in the near future, as the number of H1N1pdm virus strains detected in France has increased since 2016 [25,44].
Except for three reassortant strains, internal segments of H1Ny viruses belonged to the Eurasian avian-like (EA) lineage and showed a weak genetic evolution. However, we highlighted the co-occurrence of two M clades within the EA lineage in France and elsewhere in Europe. Before the emergence of the H1N1pdm virus, Furuse and coll. identified only one M gene swine lineage in Europe [36]. This lineage was maintained in European and Asian swIAVs, but it appears that a new lineage derived in Europe and spread in all swIAV subtypes from 2002. The two M clades, swEA and swEU, were identified in French swIAVs regardless of region or lineage. The M swEU clade presented specific mutations in the ectodomain (M2e) and transmembrane domain of the matrix protein M2. However, the origin and the consequences of the emergence of this new M clade are poorly understood up to now. The M2e domain has heretofore been considered to be host-specific and invariable between IAV subtypes, suggesting that antibodies against M2e might confer cross-protection which is currently being explored for a universal IAV vaccine development [69][70][71]. However, the variability in the M2e domain, such as here observed in swIAVs could help virus strains to evade the immune system, raising questions about the accuracy of such M2e-based vaccines.
Finally, the analysis of the internal protein-encoding genes showed a variability in length for PB1-F2 and PA-X, and detected several mutations suspected of altering the virulence of strains. Thus, the H1 av N1 strains of clade HA-1C.2.1 group B harbored a truncated PB1-F2 protein and mutations PB1-F2-N66S and NS1-D92E were described as increasing virulence [50,[72][73][74]. However, these strains remained located in the Eastern part of France, in an area with a low pig density, which may have limited their spread.

Conclusions
To conclude, the genetic evolution of swIAVs in France in the last two decades is characterized by geographical divergence in H1 av Ny viruses and apparition of H1 hu ∆146-147 strains, which entailed the adaptation of RT-PCR primers for HA subtyping. Besides marked antigenic drift in H1 hu ∆146-147 strains, some amino acid mutations in HA induced slight antigenic drifts every four years within H1 av and H1 hu lineages, which would require the reference antigens to be used in HI tests for serological investigations to be re-evaluated regularly. The low number of samples taken in some areas of France (the center and the East) do not preclude the presence of additional minor H1 genotypes in the French territory. Thus the diversity could be even greater than observed. The genotype constellation was widely dominated by H1 av N1 and H1 hu N2 strains, with internal segments of the Eurasian avian-like lineage, between 2000 and 2018. However, the situation might evolve with further adaptation of new variant or reassortant viruses, in line with increasing frequency of H1N1pdm infections and/or new virus spreading following imports. Finally, the identification of molecular markers for virulence and antiviral resistance also recalls the importance of implementing strategies to disrupt swIAV spreading within and between pig herds, as well as IAV transmissions between humans and pigs.
Supplementary Materials: The following are available online at http://www.mdpi.com/1999-4915/12/11/1304/s1, Table S1: Genbank accession numbers for swIAV sequences isolated in France and used in this study, Table S2: List of residues identified as possibly involved in changing the pathogenicity, virulence or fitness of IAVs, Table S3: HI titers of hyperimmune antisera directed against reference H1 av , H1 hu , H1pdm and classical swine H1 viruses when tested to swIAV of (A) H1 av Ny or (B) H1 hu Ny lineage isolated in France between 2000-2018, Table S4: Specific mutation patterns identified with AVANA software which characterize deduced amino acid sequences of internal proteins from strains of the three genogroups A, B and C within HA-1C.2.1 clade, Figure S1