Substitutions affecting the HA–NA–receptor balance preceded the emergence of the pandemic 2009 H1N1 virus

Significance

Influenza A pandemics greatly affect the human population every few decades when a novel animal virus breaches the species barrier. The H1N1 pandemic of 2009 (H1N1pdm09) is the first pandemic in the genomics era, which allows us to identify and study in detail mutations in prepandemic swine viruses associated with the jump to humans. Here, we highlight adjustment of the balance between HA-mediated receptor binding and NA-mediated receptor cleavage (HA–NA balance) as a step that occurred in swine prior to the emergence of H1N1pdm09 virus in humans. These findings suggest that not only mutations in HA, but also in NA, may contribute to the pandemic potential of animal viruses.

Abstract

Swine are a crucial species in influenza A virus (IAV) ecology, capable of supporting replication and consequently evolution of human, avian, and swine strains. They are also the source of the 2009 pandemic H1N1 virus (H1N1pdm09). Understanding how swine IAV breach the species barrier is essential for detection of potentially pandemic viruses, but is often overlooked compared to avian IAV. We hypothesize that the HA–NA balance of (pre)pandemic swine viruses is a key feature that needs adjustment as swine and humans differ in their sialoglycome. We focused on a set of (pre)pandemic swine viruses collected in Mexico following the H1N1pdm09 emergence. Phylogenetic analysis showed that the emergence was preceded by substitutions in HA and NA predicted to affect their function. Thus, HA acquired substitution A227E, previously shown to reduce receptor binding in H1N1pdm09, while NA obtained substitution S369N in the 2nd sialic acid binding site (2SBS) predicted to reduce NA cleavage. The substitutions’ effect on the HA–NA balance, was confirmed with biolayer interferometry using recombinant proteins attached to nanoparticles. Engineering recombinant attenuated influenza viruses with swine HA and NA with/without the mentioned substitutions, showed that the 2SBS substitution negatively affected replication in primary human and swine cultures. Interestingly, replication was restored by the HA substitution. Overall, we show that the emergence of H1N1pdm09 in humans was preceded by substitutions in HA and NA that were acquired in swine and adjusted the HA–NA balance. We hypothesize that adjustment of the HA–NA balance was a critical step for crossing the host–species barrier.

Influenza A viruses (IAVs) are among the leading infectious agents responsible for viral infections in humans and animals worldwide (1, 2). In humans, IAV cause seasonal epidemics and occasional pandemics. Pandemics are driven by animal IAVs that manage to cross the host species barrier and to spread efficiently between humans. The emergence of pandemic IAV is infrequent, having occurred four times since the turn of the 20th century (1918, 1957, 1968, and 2009) (3), probably as these viruses not only need to be antigenically novel for the human population but also to adapt to the new host species. Incidentally, the 2009 pandemic was the first pandemic in the genomics era, thereby providing novel opportunities to study in detail the adaptive changes needed to enable the jump from an animal species to humans. In addition, the H1N1 pandemic virus from 2009 (H1N1pdm2009), was also the first pandemic IAV for which it was possible to unequivocally establish that after several reassortment events in swine the virus jumped to humans (4).

The host tropism of IAV is determined by a set of species-specific interactions between the viruses and their host. This has been most extensively studied by contrasting those of human and avian IAV, since aquatic birds are regarded as the natural host reservoir (reviewed in ref. 5). Adaptation of avian to humans or other mammalian hosts often includes substitutions in the polymerase subunit PB2 (6, 7) and the evasion of host restriction factors [e.g., Mx1; (8)]. Additionally, avian and human viruses differ in their receptor-binding preference. So while avian viruses commonly prefer binding with their hemagglutinin (HA) to sialic acid (Sia) bound to the penultimate sugar via an α2-3 linkage (α2-3Sia), human viruses generally display a preference for α2-6Sia as a result of substitutions in the receptor binding site (RBS) of their HA (9). The binding preferences reflect the abundance of these glycan receptors in the avian enteric and the human upper respiratory tract (10). Recently we have shown, however, that this binding preference may not be so absolute (11, 12). Besides the RBS, other regions of HA, such as the stem region or the trimer interface, can undergo changes that increase HA prefusion trimer stability of human compared to avian viruses, resulting in a lower pH fusion threshold (13).

In addition, the activity of the neuraminidase (NA), which cleaves Sia and is the functional counterpart of HA, appears to differ between human and avian viruses, with human viruses being among others relatively better at cleaving α2-6Sia (14–16). The NA protein is a type II transmembrane protein, which contains a short cytoplasmic tail, a transmembrane domain, a stalk region, and a head domain, forming homotetrameric protrusions (17). The NA head domain contains a highly conserved catalytic site as well as a second Sia binding site (2SBS) (originally known as hemadsorption site (reviewed in ref. 18). NA activity may be modified via several mechanisms besides substitutions in the catalytic site (19), including shortening of its stalk region (20), or substitutions in the 2SBS (16). A shortened NA stalk limits access of the catalytic site to substrates, while Sia-receptor binding by the 2SBS contributes to NA cleavage activity against multivalent substrates by bringing them closer to the catalytic site (21, 22). While the 2SBS is highly conserved in avian IAV, it is invariably lost in human viruses including H1N1pdm2009 (18).

Changes in HA and NA result in an adjustment of the balance between their activities, referred to as the HA–NA balance. The HA–NA balance can be defined as the synergy of HA-mediated receptor binding and NA-mediated receptor cleavage. The right equilibrium between the two activities allows for i) successful escape from the plentiful decoy receptors present in the mucus layer and the underlying respiratory epithelium, ii) virion motility on the cell surface needed for efficient viral entry and iii) efficient release of progeny virions at the end of the infection cycle (23–25). Clearly this balance needs to be readapted when IAV jumps from birds to humans (25, 26), but much less is known about changes required upon a jump from swine to humans. Nevertheless, we anticipate that for this species jump also, adaptation of the HA–NA balance is needed. Even though the general distribution of α2-6Sia and α2-3Sia is similar in the respiratory tract of humans and swine, with α2-6Sia and α2-3Sia particularly being found in the upper and lower respiratory tract, respectively (27, 28), the swine sialoglycan repertoire differs from that in humans. In contrast to humans, swine express besides the Neu5Ac Sia form also the Neu5Gc form, which is generally not bound by IAV, and the terminal α-Gal epitope, which cannot be sialylated (29). As a consequence, swine display less sialoglycans to which IAVs can attach, possibly resulting in an altered optimal HA–NA balance for swine compared to human viruses. Indeed, it has been suggested that human and swine viruses differ in their functional HA–NA balance (30), while the HA of H1N1pdm09 was shown to display reduced receptor binding compared to distantly related swine viruses, which was largely attributed to an A to E substitution at position 227 (30–32).

H1N1pdm2009 emerged in swine after multiple reassortment events between IAVs of different origins (33) resulting in a novel HA–NA combination, in which HA of the H1N1 classical swine lineage teamed up with NA from the H1N1 Eurasian avian-like swine lineage. We hypothesize that this novel HA–NA combination may have resulted in the selection of substitutions in HA and NA that altered the HA–NA balance as such that it allowed the jump from swine to humans. In this paper, we therefore delved deeper into the evolution of the HA–NA balance prior to and during the emergence of the H1N1pdm2009 virus. To this end, we performed extensive phylogenetic analysis using among others viruses collected at the site of emergence, descending from prepandemic strains (34), thereby identifying substitutions in RBS of HA and the 2SBS of NA associated with the jump to humans. Subsequently, we analyzed the functional effect of these substitutions on receptor binding and cleavage, and replication in differentiated air-to-liquid interface (ALI) cultures. Our results indicate that the emergence of H1N1pdm09 is associated with substitutions that reduced receptor binding by HA and cleavage by NA. While the substitution in HA did not affect replication in swine or human ALI cultures much, those in NA, located in the 2SBS, had a dramatic negative effect, which could be compensated, however, by the substitution in HA. We propose that adjustment of the HA–NA balance contributed to the jump of prepandemic H1N1 to the human host.

Results

Ancestral sequence reconstruction (ASR) by sequence alignment inspection and construction of phylogenetic trees enables prediction of substitutions in the HA and NA preceding the emergence of H1N1pdm2009. Phylogenetic analysis was performed for HA and NA coding regions of swine IAV viruses, circulating in Central America after the emergence of the H1N1pdm2009 (34). Trees were rooted by closely related sequences from the classical swine lineage (HA) or the Eurasian avian-like lineage (NA). A relevant part of the analysis is presented in Fig. 1, while the entire analysis is provided in Datasets S1–S5. Previously, swine viruses, referred to as genotype 1 (34), were identified with the same combination of segments derived from avian-like Eurasian H1N1 swine (N1), classical H1N1 swine (H1), and triple reassortant H3N2 swine virus lineages as H1N1pdm09. Genotype 1 viruses are closely related to H1N1pdm09, but are distinct from reverse zoonosis introductions of H1N1pdm09. Of these genotype 1 viruses, which are regarded as closely related descendants of prepandemic viruses (34), some HA and NA sequences cluster with H1N1pdm09 viruses that jumped back from humans to swine, while others form a separate cluster and are more distantly related. HA of this latter cluster differs from the other genotype 1 and the pandemic viruses by the following substitutions (H3 numbering): L74S, N87S, N145K, R149K A227E, K261E, S264A, and V301. Of these substitutions, only the A227E substitution was previously shown to (negatively) affect affinity to glycan receptors (31). The topology of the NA tree mostly follows that of the HA tree, with a similar division of the genotype 1 viruses. Again one group clusters with NA of H1N1pdm09 viruses, which differ from the other genotype 1 viruses by a set of substitutions, which include among others S369N. S369N is located in the 370-loop of the 2SBS and, as S369 is a Sia contact residue, is predicted to (indirectly) affect the catalytic activity of NA for multivalent substrates (18, 35). Pandemic viruses also acquired another substitution in this 2SBS loop (T365I). Altogether, the ASR indicates that descendants of genotype 1 prepandemic viruses fall apart in two groups both for their HA and NA proteins. One group clusters with H1N1pdm09 viruses, and can be regarded as descendants of viruses that jumped to humans, while the other group is a bit more distantly related. The phylogenetic analysis thus indicates that several substitutions in HA and NA were acquired in swine immediately prior to the emergence of the H1N1pdm09 virus in humans. Some of these substitutions are predicted to affect the HA–NA balance by modifying HA receptor-binding (A227E) or NA catalytic activity (S369N).

Fig. 1.

ASR: Phylogenetic analysis of HA and NA nucleotide sequences (coding regions) from all full-length swine IAV H1N1 (and H1N2 for HA) isolates originating from Central America were aligned with nucleotide sequences from swine viruses isolated in Mexico (indicated by AVX#) as described by Mena et al. (34). Genotype 1 viruses, which have the same combination of segments derived from avian-like Eurasian H1N1 swine (N1), classical H1N1 swine (H1), and triple reassortant H3N2 swine virus lineages as H1N1pdm09, but are distinct from reverse zoonosis introductions of H1N1pdm09, are underscored. The selected genotype 1 virus AVX3 is indicated in red and the selected swine isolate clustering with H1N1pdm2009, AVX56, is indicated in green. A few human H1N1pdm2009 viruses included in the analysis are colored in purple. Sequences of isolates carrying HA of the classical swine lineage or NA of the Eurasian avian-like lineage were added to root the trees, EPI_ISL_1120_A/swine/Wisconsin/1/61 and EPI_ISL_199471_A/swine/Gent/196/92, respectively. Bootstraps values above 80 are shown. Further information is provided in Materials and Methods and in Datasets S1–S5.

Replication and Receptor Interaction of Selected Swine Viruses.

A representative genotype 1 swine virus with a functional 2SBS in NA and with A227 in HA (A/swine/Mexico/AVX3/2011(H1N1), referred to as AVX3) and a swine virus clustering with H1N1pdm2009 with substitutions in the 2SBS of NA and with E227 in HA (A/swine/Mexico/AVX56/2013(H1N1), referred to as AVX56) were selected to assess their replicative capability in swine and human cell cultures (Fig. 2 and SI Appendix, Fig. SD1). To this end, primary tracheal swine and primary bronchial human cells from two different donors each were differentiated in ALI conditions and infected (5 × 10⁷ particles/transwell) with the selected viruses. Analysis of released virus at different time points post infection showed that the two viruses did not differ in their replication kinetics both on human and swine cells.

Fig. 2.

Replication and receptor interaction of selected swine viruses in differentiated primary airway cell cultures. (A) Pipeline of replication kinetics. Cells from two different donors of human or swine origins were differentiated in ALI conditions for 4 to 5 wk and then infected with the selected wt swine viruses: AVX3 and AVX56. Apical washes were collected every 24 h and analyzed by an end-point dilution assay (TCID50) on MDCK-II cells. Replication of AVX3 and AVX56 in human (B) and swine (C) differentiated primary airway cell cultures. (B) n = 2 with triplicates. (C) n = 3 with triplicates. Statistical analysis in SI Appendix, Fig. SD1.

Subsequently, biolayer interferometry (BLI) was used to study the receptor-binding properties of the HA proteins of AVX3 and AVX56. As stocks of these viruses displayed little if any binding to the sensors, regardless of the receptors loaded, we made use of a previously developed nanoparticle system, in which recombinant soluble hemagglutinin-neuraminidase (HN) of paramyxovirus was coupled to Nickel-NTA nanoparticles (NPs) (36). In the present study, NPs were first coated with recombinant soluble HA of AVX3 or AVX56, referred as AVX3-HA and AVX56-HA. These preparations displayed similar binding to sensors coated with synthetic α2,6-linked Sia (6′S[LN]3), while no binding was observed to α2,3-linked Sia (3′S[LN]3) (Fig. 3 A and B). To analyze the enzymatic activity of the corresponding NA proteins, we first performed an assay with the commercially available substrate MUNANA. The two proteins (AVX3-NA and AVX56-NA) displayed a similar ability to cleave this small compound, resulting in similar specific activities (Fig. 3 C and D). As a measure for the HA–NA balance, we subsequently analyzed the interaction of NPs displaying both HA and NA in a 3 to 1 ratio with receptor-coated surfaces in BLI. NPs displaying HA and NA that bound to the same extent to sensors loaded with α2,6-linked Sia in the presence of a NA inhibitor, Oseltamivir Carboxylate (OsC), differed in the absence of OsC. NPs displaying HA and NA of AVX3 were released faster [had a smaller area under the curve (AUC)] than those with HA and NA of AVX56 (Fig. 3 E–G). To confirm that this difference was due to the different NA rather than the HA proteins, which displayed similar receptor binding (Fig. 3 A and B), a similar experiment was performed in which NPs were loaded with the same HA protein (that of AVX3), but with different NA proteins. Analysis of the AUC of these NPs was similar in the presence of the NA inhibitor (Fig. 3 H–J), but clearly different when NA was not inhibited, with NPs containing NA of AVX3 being released faster than those with AVX56. From these results we conclude that the HA proteins of AVX3 and AVX56 have similar receptor-binding properties, but differ in the catalytic activity of their NA proteins, at least when analyzing the release of NPs from a receptor-coated surface.

Fig. 3.

Receptor-binding and cleavage properties of the selected viruses’ proteins. (A) Binding of NPs coupled to AVX3-HA and AVX56-HA proteins to Streptavidin sensors coated with 3′S[LN]3 (dashed lines) or 6′S[LN]3 (solid lines) synthetic glycans analyzed by BLI. (B) AUCs of graphs shown as in (A) for 6′S[LN]3 were graphed normalized to AVX3-HA 6′S[LN]3 (n = 3, data are mean ± SD). Significance was determined by with an Ordinary one-way ANOVA test followed by Tukey’s multiple comparison test. (C and D) Activity curves of soluble AVX3-NA and AVX56-NA using the commercial MUNANA substrate. (D) Specific enzymatic activity (1/EC50) normalized to AVX3-NA. Significance estimated with an unpaired two-tailed Welch’s t test (n = 2 with duplicates). (E) Binding to 6′S[LN]3 of NPs coupled to AVX3-HA and AVX3-NA or AVX56-HA with AVX56-NA in the presence (colored lines) or absence (gray lines) of OsC. Proteins were combined in a 3 (HA) to 1 (NA) ratio. Sample amount was adjusted to reach similar levels of binding between conditions in the presence of OsC. (F and G) Significance of differences in the AUC analyzed by unpaired one-tailed Welch’s t tests (n = 3, data are mean ± SD). Normalized to NPs carrying AVX3-HA + AVX3-NA. (H–J) NPs with AVX3-HA and AVX3-NA were compared to NPs with AVX3-HA and AVX56-NA similarly to (E–G).

Receptor-Binding and Cleavage Properties of Mutant HA and NA Proteins of AVX3.

Next we focused on those substitutions in HA and NA that preceded the emergence of H1N1pdm09 (and are found in AVX56 but not AVX3) and that are likely to have a (large) effect on either receptor binding or cleavage, being A227E in the RBS of HA (30, 31) and T365I and S369N in the 370 loop of the NA 2SBS (18). The effect of these substitutions on receptor binding and cleavage as well replication was studied after their introduction in HA and NA of AVX3 (resulting in HAmut and NAmut). First, we compared the binding of NPs displaying AVX3-HA or AVX3-HAmut (Fig. 4 A and B). Introduction of the A227E substitution in HA led to reduced binding to 2-6Sia compared to the wt HA. Substitutions in the 2SBS of NA negatively affected cleavage of MUNANA, resulting in a lower specific activity for the AVX3-NAmut protein (Fig. 4C and SI Appendix, Fig. SD2). Similar differences were observed when using linkage specific MUNANA assays (Fig. 4 D and E and SI Appendix, Fig. SD2). Mutations in the NA thus affect catalytic cleavage of small substrates, without affecting the cleavage specificity. We also evaluated the NA catalytic activity using NPs with AVX3-HA in combination with AVX3-NA or AVX3-NAmut. AVX3-NA resulted in faster release from the α2,6-linked SIA than AVX3NAmut (Fig. 4 F–H). Taken together, the BLI and cleavage assay data highlight the negative effect of the A227E substitution on HA-receptor binding and the T365I and S369N substitutions on NA substrate cleavage and release of NPs from a receptor-coated surface.

Fig. 4.

Receptor-binding and cleavage properties of mutant HA and NA proteins of AVX3. (A) Binding of NPs carrying AVX3-HA (solid line) and AVX3-HAmut (dashed line) proteins to Streptavidin sensors coated with 6′S[LN]3 synthetic glycan analyzed by BLI. (B) Significance of AUC analyzed by an unpaired one-tailed Welch’s t test (n = 4, data are mean ± SD) normalized to AVX3-HA. (C–E) Specific enzymatic activity (1/EC50) of soluble AVX3-NA and AVX3-NAmut using the commercial MUNANA substrate (C), an α2,3 specific (D) and an α2,6 specific (E) substrate. Data normalized to AVX3-NA were analyzed with unpaired one-tailed Welch’s t tests (n = 2 to 3 with triplicates). Activity curves provided in SI Appendix, Fig. SD2. (F) Binding to 6’S[LN]3 of NPs carrying AVX3-HA and AVX3-NA or AVX3-HA with AVX3-NAmut in the presence (colored lines) or absence (gray lines) of OsC. Proteins were combined in a 3 (HA) to 1 (NA) ratio. Sample amount was adjusted to reach similar levels of binding between conditions in the presence of OsC. (G and H) Significance of differences in the AUC analyzed by unpaired one-tailed Welch’s t tests (n = 4, data are mean ± SD). Normalized to NPs with AVX3-HA + AVX3-NA.

Infection of Differentiated Primary Airway Cell Cultures with Recombinant Viruses.

Subsequently, the individual and combined effects of the selected substitutions in the RBS of HA and 2SBS of NA on virus replication were studied in the background of the HA and the NA of AVX3. To this end, we engineered a set of four recombinant viruses in the background of the attenuated influenza A/Puerto Rico/8/1934(H1N1) virus that carried wt and/or mutant HA and NA genes of AVX3 (referred to as PR8-AVX3, PR8-AVX3-HAmut, PR8-AVX3-NAmut, PR8-AVX3-HA/NAmut) (Fig. 5A). Stocks of these viruses displayed similar particle numbers and infectious titers on MDCK-II cells (Fig. 5B). The viruses were next used to infect differentiated primary bronchial human and tracheal swine cells (Fig. 5C) from two different donors each under ALI conditions as described previously (Fig. 2). All viruses, apart from PR8-AVX3 NAmut, replicated to similar extents on the human and swine cultures. Notably, replication of PR8-AVX3-NAmut was much decreased compared to the other three recombinant viruses. However, when the substitutions in the 2SBS were combined with the A227E substitution in HA the negative effect of the NA substitutions was compensated (Fig. 5 D and E and SI Appendix, Fig. SD3). The double mutant virus even replicated to a higher extent than the parental virus at several time points both in the human and swine cultures. In agreement herewith, infection levels in primary tracheal swine cells analyzed at 14 hpi by immunofluorescence staining resulted in corresponding differences between the different viruses (Fig. 5 F and G). In conclusion, substitutions in the 2SBS of NA had a large negative impact on replication in swine and human ALI cultures, unless combined with a compensatory substitution in HA that negatively affected receptor binding.

Fig. 5.

Infection of differentiated primary airway cell cultures and HEK293 cells with different sialoglycan repertoire with recombinant viruses. (A) Schematic representation of the recombinant viruses (PR8-AVX3, PR8-AVX3-HAmut, PR8-AVX3-NAmut, PR8-AVX3-HA/NAmut) engineered to contain the HA and NA genes of AVX3 along with selected mutations in an A/Puerto Rico/8/1934(H1N1) background. (B) Table presenting the titers of the recombinant viruses’ stocks used in the following experiments as determined directly by Nanoparticle Tracking analysis (NanoSight NS300) or indirectly by end-point dilution assay (TCID₅₀) on MDCK-II cells. To determine the TCID₅₀/mL values, cells were fixed 3 d postinfection, stained with the HB65 (37) primary mouse antibody targeting the viral NP protein along with secondary anti-mouse Alexa Fluor 488 and observed under an EVOS microscope. (C) Pipeline of experiments performed with primary airway cultures. Cells from two different donors of human or swine origins were differentiated in ALI conditions for 4 to 5 wk and then infected with the recombinant viruses. Apical washes were collected every 24 h and analyzed by RT-qPCR by targeting the M1 gene. Swine cells were additionally fixed at 14 hpi to assess single round infection. (D and E) Replication kinetics of the recombinant viruses in human (n = 2 with triplicates) and swine (n = 4 with triplicates) ALI cultures respectively. Statistical analysis in SI Appendix, Fig. SD3. (F) Immunofluorescence assay of swine ALI cultures infected by the recombinant viruses for 14 h. Fixed and permeabilized samples were stained with a primary human HA stem targeting antibody (Fi6) recognized by Alexa-Fluor 488 anti-human secondary antibody. Samples from each condition were captured with an air UPLXAPO 4×/0,16 objective lens and HA-positive cells were quantified using ImageJ Fiji 1.54p. (G) Significance was determined with a one-way ANOVA test followed by Tukey’s multiple comparison test (n = 3, data are mean ± SD). (H) HEK^2,3Sia (solid bars) and HEK^2,6Sia cells (checkered bars) were transfected with the Gaussia luciferase-encoding vector pHH-Gluc for 24 h before infection with the recombinant viruses in single round conditions (absence of trypsin). In parallel, cells were also infected with Coxsackie virus B3 (CVB3) whose infection levels are independent of sialic acid. Statistical analysis by an unpaired two-tailed t test with Holm–Šídák correction for multiple comparisons, without assuming equal variances (n = 2, 5 to 10 replicates). Comparisons per cell line in SI Appendix, Fig. SD4.

Single-Round Infection with Recombinant Viruses on HEK293 Cells with Different Sialoglycan Repertoire.

Finally, we analyzed the preferred receptor usage of the four recombinant viruses using HEK293^2,3SIA and HEK293^2,6SIA cells (38) in combination with a Gaussian luciferase reporter construct. Cells were infected (100 particles/cell) with the recombinant viruses in the absence of trypsin, which results in a single-round of infection. We observed similar differences in replication efficiency as in the human and swine ALI experiments (Fig. 5H and SI Appendix, Fig. SD4). Specifically, the PR8-AVX3-NAmut virus gave low levels of infection unless the NA substitutions were combined with the HA substitution. Increased infection levels were observed for all viruses in the HEK293^2,6SIA compared to the HEK293^2,3SIA cells indicating that all viruses preferred infection via 2-6Sia, which is the preferred receptor of the AVX3 and AVX56 proteins as shown by the previous BLI experiments (Figs. 3 and 4).

Discussion

Exploring the events that led to the 2009 pandemic is necessary to comprehend the conditions that allow swine viruses to breach the swine-to-human species barrier. While swine-to-human infections are known to happen, most commonly due to contact with live animals (39, 40), subsequent human-to-human transmission is much less frequent. Previously, swine viruses were identified in Mexico with a similar genomic constellation as H1N1pdm09, that were distinct from reverse zoonosis introductions. These viruses are regarded as descendants of the most closely related ancestor viruses of H1N1pdm09 [referred to as genotype 1 (34)] and indicate that the swine virus responsible for the 2009 pandemic evolved in swine in Mexico. Performing detailed phylogenetic analyses, we identified substitutions in HA and NA that are observed in a subset of these viruses and preceded the emergence of H1N1pdm09 in humans. In addition, we showed that some substitutions (A227E in the HA RBS and T365I and S369N in the NA 2SBS) negatively affected receptor binding and cleavage. While the substitutions in NA had a large negative effect on virus replication in differentiated ALI cultures, their negative effect could be compensated by the substitution in HA. We conclude that adjustment of the HA–NA balance in prepandemic swine viruses preceded the emergence of H1N1pdm09 in the human host and propose that adjustment of this balance contributed to the jump of this virus to humans.

Of the substitutions in HA preceding the emergence of H1N1pdm09 in humans, we focused on residue 227 in HA, as the identity of this residue was previously shown to be the most important of all differences present for receptor binding when comparing H1N1pdm09 to more distantly related swine viruses (30, 31). In agreement herewith, we show here that substitution A227E in prepandemic AVX3 reduced receptor binding. Nevertheless, HA of AVX3 (with A227) displayed similar receptor binding as HA of AVX56 (with E227). Other differences between AVX3 and AVX56 HA, including R149K, which was previously reported to have a minor positive effect on receptor binding (30, 41), may compensate for the negative effect of A227E. Background-dependent effects of the reciprocal substitution (E227A) were also reported for A/California/04/2009(H1N1) with an important role for the residue at position 200 (31). Interestingly, since 2020 circulating H1N1pdm09 viruses contain A227 again (SI Appendix, Fig. SD5).

Several substitutions in NA preceded the emergence of H1N1pdm09. Two of these (T365I and S369N) are located in the 370 loop of the 2SBS, with position 369 being a Sia contact site. The 2SBS indirectly affects cleavage by NA, by bringing multivalent substrates closer to the catalytic site (18). S369N was previously shown to affect cleavage of multivalent substrates by the N1 of H5N1 virus (42). In agreement with these substitutions affecting the functionality of the 2SBS, AVX3 and AVX56 NAs differed in their ability to release NPs from a receptor-coated surface, but did not differ in their catalytic activity per se as assessed using the small molecule substrate MUNANA. Introduction of the two substitutions in AVX3 NA, negatively affected catalytic activity directly, which does not appear surprising in view of the close proximity of the 370 loop with the catalytic site (18). This makes it impossible, however, to discern the effect of these substitutions on the 2SBS from that of the catalytic site specifically as both will reduce the release of NPs from a receptor-coated surface. Regardless, we can conclude that substitutions located in the 2SBS that preceded the emergence of H1N1pdm09 in humans have a negative effect on cleavage of Sia receptors by NA.

The 2SBS is highly conserved in the NA of avian, but not in human viruses. Interestingly, the NA protein of swine viruses derived from human viruses consequently do not carry a functional 2SBS, indicating that a functional 2SBS is not required for efficient replication and transmission in swine. However, N1 proteins from the Eurasian avian-like swine lineage often contain a 2SBS with all SIA-contact residues conserved (approximately 50% prior to the emergence of H1N1pdm09). Swine viruses thus may carry, in contrast to human viruses, a functional 2SBS. While substitutions T365I and S369N are likely to affect the functionality of the 2SBS [(22) and as shown here], Sia binding by the 2SBS of early H1N1pdm09 viruses does not appear to be completely abrogated as some residual binding could still be observed (43). After the emergence of H1N1pdm09 in the human population, additional substitutions were observed in the 2SBS (N369K and K432E), of which N369K increased and K432E abrogated the functionality of the 2SBS as these substitutions specifically affected cleavage of multivalent substrates but not of MUNANA (44).

Substitutions in the 2SBS of AVX3 (that negatively affected NA cleavage) had a large negative effect on replication in swine and human ALI cultures. Combining these substitutions with the A227E substitution in HA (that negatively affected receptor binding), however, restored the replicative ability of the resulting virus, indicative of the importance of the HA–NA balance for replication in ALI cultures. We did not observe, however, appreciable differences between replication on the human and swine ALI cultures. To what extent swine and human viruses overlap and differ in their HA–NA balance is not yet known. Some overlap is expected as human viruses frequently jump to swine [reverse zoonosis (45)] and H1N1pdm09 emerged in swine prior to its jump to humans. However, swine viruses jump far less often to humans than the other way around (45). The swine and human sialoglycome are known to differ (as previously mentioned), with swine expressing relatively fewer functional sialoglycan receptors than humans, which is predicted to result in a different optimal HA–NA balance for replication in these species. This does not necessarily mean, however, that each virus will display an optimal balance as mutations may also be selected for other reasons, e.g., antigenic drift.

Although from our phylogenetic analyses we cannot deduce the exact order of the acquisition of the different mutations in HA and NA, the following scenario seems plausible: just prior to emergence of H1N1pdm09 in humans in 2009, triple-reassortant swine virus with H1 from the classical swine lineage obtained N1 from the Eurasian-avian lineage. This novel combination presumably required and/or allowed adjustment of the HA–NA balance, which opened a window of opportunity and (perhaps temporarily) resulted in a HA–NA balance that allowed the jump of the prepandemic swine virus to humans. More insight into the HA–NA balance of human and animal IAVs as well as into the associated genetic markers thereof would assist in identifying future potential prepandemic viruses.

Materials and Methods

ASR.

HA and NA nucleotide sequences (coding regions) from all full-length swine IAV H1N1 (and H1N2 for HA) isolates originating from Central America and present in the GISAID database (46–48) were aligned with nucleotide sequences from Genbank previously analyzed by Mena et al. (34) using MUSCLE (standard settings) from the Unipro UGENE software package (49). Sequences of isolates carrying HA of the classical swine lineage or NA of the Eurasian avian-like lineage were added to root the trees, while a few early human H1N1pdm2009 viruses were also added to the analysis. Gene trees were constructed by using PhyML 3.0 available from the ATGC website (https://atgc.lirmm.fr/phyml/) (50). Standard settings were applied and 100 times bootstrapping was performed. Complete trees are provided as Datasets S2 and S4 in Newick format. Full isolate names, date of isolation, and sequence identifiers are provided as Dataset S5 in excel format. Subtrees (with identical topology as the full trees), from which highly similar sequences were deleted, were subsequently generated and displayed in Fig. 1. The most relevant bootstrap values (above 80) are indicated. Amino acid substitutions occurring at the branches leading to the pdm09 virus and maintained in (nearly) all descendants were identified by inspection of protein alignments (provided in Datasets S1 and S3) and indicated in Fig. 1.

Cell Culture of Adherent Cells.

MDCK-II, HEK^2,3SIA (HEK^ΔSia (ΔST3GAL1/2/3/4/5/6 and ST6GAL1/2) stably transfected with ST3GAL4) and HEK^2,6SIA (HEK^ΔSia stably transfected with ST6GAL1) (51), were cultured in complete Dulbecco’s Modified Eagle’s Medium (DMEM) (Capricorn Scientific) media supplemented with 10% (v/v) fetal bovine serum (SIGMA), 100 U/mL penicillin, 100 µg/mL streptomycin and were maintained at 37 °C in a humidified incubator (5% CO₂).

Viruses.

A/swine/Mexico/AVX3/2011(H1N1) (referred to as AVX3) and A/swine/Mexico/AVX56/2013(H1N1) (referred to as AVX56) were collected from swine in Mexico as described previously (34). Recombinant viruses containing both the HA and NA genes of AVX3 were generated in the background of the attenuated A/Puerto Rico/8/34(H1N1) as previously described (52), using HA and NA genes amplified from wild-type viruses (A/swine/Mexico/AVX3/2011(H1N1) and A/swine/Mexico/AVX56/2013(H1N1)). Stocks were amplified in MDCK-II cells using DMEM (Gibco) containing 0.3% bovine serum albumin (BSA, SIGMA) supplemented with trypsin-TPCK (2 μg/mL; SIGMA). Particle number of the viral stocks used for infection assays was obtained by using the NanoSight NS300 instrument (Malvern Panalytical) as described previously (53). In short, viruses diluted in Phosphate Buffered Saline (PBS; Capricorn Scientifics) were injected in the NanoSight NS300 to generate five 1-min videos that were analyzed with the Nanoparticle Tracking analysis 3.0 software. The analysis resulted in quantitative data on the amount of viral particles. Sequences of HA and NA genes were confirmed for each stock used in the experiments.

Air–Liquid Interface Cultures of Swine and Human Airway Epithelial Cells.

Primary human airway bronchial epithelial cells from healthy donors were obtained from the University Medical Center (UMC) Utrecht Biobank, Utrecht with permission of the UMC Utrecht’s Biobank Research Ethics Committee (TCBio 21-265). Cells were cultured as previously described (54). In brief, airway epithelial cells (passage 2) were thawed and expanded in Collagen IV–coated (50 μg/mL; Sigma-Aldrich) plates using an expansion medium consisting of 50% (v/v) Bronchial epithelial cell medium-basal (Sciencell), 44% (v/v) Advanced DMEM/F12 (Thermo Fisher Scientific), 2% (v/v) B-27 (Thermo Fisher Scientific), 1% (v/v) GlutaMAX (Thermo Fisher Scientific), 10 mM HEPES (Thermo Fisher Scientific), 0.5 μg/mL (±)-epinephrine hydrochloride (Sigma-Aldrich), 0.5 μg/mL hydrocortisone (Sigma-Aldrich), 100 nM 3,3′,5-triiodo-L-thyronine (Sigma-Aldrich), 1.25 mM N-acetyl-L-cysteine (Sigma-Aldrich), 1 μM A83-01 (Tocris), 5 μM Y-27632 (Selleck Chemicals), 5 μM DAPT (Fisher Scientific), 100 ng/mL recombinant human FGF-10, 25 ng/mL recombinant human HGF, 50 nM recombinant human heregulin-β1 (PeproTech), supplemented with 2% RSPO3-Fc fusion protein conditioned medium (U-Protein Express), 1% (v/v) penicillin–streptomycin (Thermo Fisher Scientific), and 100 μg/mL primocin (InvivoGen). Subsequently, 0.2 × 10⁶ cells were seeded onto PureCol-coated (Advanced BioMatrix) 24-well Transwell inserts (0.4 μm pore size polyester membrane; Corning). Cultures were differentiated at the air–liquid interface (ALI) using Advanced DMEM/F12–based differentiation medium containing 0.5 μg/mL (±)-epinephrine hydrochloride, 0.5 μg/mL hydrocortisone, 100 nM 3,3′,5-triiodo-L-thyronine, 1% (v/v) penicillin–streptomycin, 50 nM A83-01, 100 nM TTNPB (Cayman Chemical), and 0.5 ng/mL recombinant human EGF. ALI cultures were differentiated for at least 21 d prior to use.

Primary swine airway epithelial cells were obtained from piglet trachea (surplus material), differentiated and maintained as described previously (55). In brief, piglet trachea epithelial cells were cultured in 24 well plates on collagen IV-coated Costar PET transwells (0.4 µm pore size, Corning) in bronchial epithelial growth medium (BEGM) containing primocin (1:2,000, InvivoGen). BEGM medium was refreshed every 2 to 3 d until confluency was reached. When transepithelial electrical resistance (TEER) measurements exceeded 500 Ω.cm², the apical and basolateral compartments were replaced with porcine ALI differentiation medium. After 2 d of submerged culturing, transwells were air-lifted and cultured in ALI differentiation medium for 28 to 33 d. Weekly, excess mucus was collected from the apical linings and every 2 wk cell surfaces were washed with Hanks’ Balanced Salt Solution (HBSS).

Virus Replication Kinetics.

The apical side of airway epithelial cultures was washed with HBSS prior to a 2 h incubation (37 °C, 5% CO₂) with the different viral stocks diluted in the same buffer. Each transwell was inoculated with 5 × 10⁷ viral particles. After the 2-h incubation, the inoculum was removed and the cultures were washed twice. HBSS buffer was supplied to the washed surfaces and eventually collected after a 15 min incubation (37 °C, 5% CO₂) to serve as the 0-h timepoint sample. This was repeated every 24 h until 3 d postinfection. The samples were stored at −80 °C and subsequently analyzed by RT-qPCR or titrated using the TCID₅₀ method.

RT-qPCR.

RNA was extracted using the QIAamp Viral RNA Mini Kit (QIAGEN) according to the manufacturer’s recommendations. Reverse transcription and quantitative PCR were performed in a single reaction strategy using the TaqMan Fast Virus 1-Step kit (ThermoFisher Scientific) and the LightCycler LC480 (Roche). Primers (forward primer: 5′-AAGACCAATYCTGTCACCTYTGA-3′; reverse primer: 5′-CAAAGCGTCTACGCTGCAGTCC-3′) and probe (5′-TTTGTKTTCACGCTCACCGTGCC3′ coupled to FAM) targeting the matrix genome segment were used. The Ct values were calculated using the Light Cycler 96 software (Roche) version 1.1.0.1320 and the Fit Points method.

Immunofluorescence.

Swine tracheal inserts were fixed at 14 h postinfection with 4% PFA for 20 min followed by 15 min of permeabilization with 0,1% Triton-100X (Thermo Fisher). Next, the samples were incubated with a human primary HA targeting antibody [Fi6; (56)] and then with an Alexa-Fluor 488 anti-human secondary antibody (Thermo Fisher) diluted in PBS-BSA 1%. Finally, samples were mounted in glass slides using the Prolong Diamond Antifade Mountant (Thermo Fisher). Images from three independent experiments were captured with an Olympus IX85 microscope using an air UPLXAPO 4×/0.16 objective lens and a Hamamatsu ORCA-Fusion camera. HA-positive cells were quantified using ImageJ Fiji 1.54p and further analyzed with GraphPad Prism 10.4.1.

Luciferase Assay.

Thirty thousand HEK^2,3SIA and HEK^2,6Sia adherent cells were seeded per well in 96-well plates coated with Poly-L-Lysine (Sigma). After 24 h, each well was transiently transfected with 100 ng of the Gaussia luciferase-encoding vector pHH-Gluc, described in ref. 37, using FuGENE6 (Promega Corporation) according to the manufacturer’s recommendations. The cells were infected with 5.5 × 10⁶ particles per well in Opti-MEM (Gibco) 24 h post transfection. Single round infection (in the absence of trypsin) was measured in the supernatants of the cultures the next day as described in ref. 37. Two independent experiments were performed with 5 to 10 replicates in total per condition.

Protein Expression.

Recombinant viral proteins were expressed in HEK^ΔSia (ΔST3GAL1/2/3/4/5/6 and ST6GAL1/2), adapted to suspension culture (130 RPM), at 37 °C in a humidified incubator (8% CO₂) in Freestyle 293-F expression medium (Gibco). Human codon-optimized cDNAs (Tsingke Biotech Co.) encoding the ectodomains of HA (Gene bank: KU976666.1 and KU976838.1) and NA (Gene bank: KU976552.1 and KU976728.1) of AVX3 and AVX56 were cloned into the pFRT expression plasmid (Thermo Fisher) in-frame with a N-terminal cleavable signal sequence, 6xHistidine tag, a Twin-Strep tag and a tetramerization domain (NA) or a C-terminal trimerization domain (HA) as described previously (36, 57). Site-directed mutagenesis, using the New England Biolabs (NEB) Q5 Site-directed Mutagenesis kit, was performed to introduce mutations in the AVX3-NA encoding gene. The plasmids of the viral glycoproteins were transiently expressed in the HEK^ΔSia cells with polyethyleneimine (PEI 1 mg/mL, Polysciences) for 4 d before eluting the proteins from the supernatant using Strep-Tactin beads (IBA Lifesciences) according to the manufacturer’s recommendations. The eluted proteins were quantified in a 10% SDS-PAGE gel stained with Coomassie Brilliant Blue (Westburg Lifesciences) along with a bovine serum albumin (BSA) standard curve through quantitative densitometry. Protein preparations to be compared in binding and cleavage assays were quantified on the same gel.

Coupling Purified HA and NA Ectodomains to Nanoparticles.

Purified HA ectodomains were coupled to dextran iron oxide composite Ni-NTA nanomag-D nanoparticles (130 nm diameter, Micromod) at 4 °C overnight as described previously (36, 58). Purified NA proteins were combined with HA proteins in a 3HA:1NA molar ratio. Each sample preparation reached 100 μL in volume consisting of 1.5 × 10⁸ nanoparticles and 3 μg of protein in total diluted in Phosphate-buffered saline with Ca²⁺ and Mg²⁺ (PBS+/+, Lonza). In each run 15% of the sample was used.

BLI.

“Dip and read” experiments were performed with the Octet RED384 (Fortebio) system while employing streptavidin biosensors (SA, Pall-ForteBio) as previously described (59). Biotinylated synthetic glycans 3′S(LN)3 and 6′S(LN)3, synthesized as described previously (SI Appendix), were loaded to saturation to the biosensors. Next, glycan-loaded SA sensors were dipped in wells containing HA-Ni-NTA or HA-NA-Ni-NTA nanoparticles in the presence or absence of Oseltamivir Carboxylate (OsC, Roche) at a 10 μM final concentration. PBS+/+ was used as the standard assay buffer. Experiments were carried out at 30 °C and were independently repeated three to four times.

NA Activity (MUNANA Assay).

Sialidase activity of purified NA proteins was determined in a fluorometric assay using three different monovalent substrates. First, 4-methylum belliferyl-N-acetyl-α-D-neuraminic acid (4-MUNANA, Sigma Aldrich) was used as previously described (42). Second, to elucidate specific NA mediated cleavage we used substrates with specific α2-3 or α2-6 linkage, 4-Methylumbelliferyl-O-(5-acetamido-3,5-dideoxy-d-glycero-α-d-galacto-non-2-ulopyranosylonate)-(2→3)-β-d-galactopyranoside (Sia2-3GalMU) and 4-Trifluoromethylumbelliferyl-O-(5-acetamido-3,5-dideoxy-d-glycero-α-d-galacto-non-2-ulopyranosylonate)-(2→6)-β-d-galactopyranoside (Sia2-6GalMU), that have been reported previously (60, 61). Two to three biological replicates were performed for each condition.

Software and Statistical Analysis.

BLI data acquisition and analysis were performed using the Octet System Data Acquisition and Analysis software and GraphPad Prism 10.4.1. All statistical analyses and graphical representations were generated using GraphPad Prism 10.4.1, with significance indicated as follows: ns; P > 0.05, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. Microscopy image analysis was assured by the ImageJ Fiji 1.54p software and schematic illustrations were prepared using BioRender (https://biorender.com).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (FAS)

Dataset S02 (TXT)

Dataset S03 (FAS)

Dataset S04 (TXT)

Dataset S05 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.