Characterization of the glycoproteins of fish and amphibian influenza B–like viruses

Abstract

Influenza-like virus sequences previously identified in fish and amphibians cluster as a sister clade of influenza B viruses but remain largely uncharacterized. We demonstrate that salamander influenza-like virus (SILV) hemagglutinin (HA) is functionally divergent from influenza B virus HA and does not bind to α2,3- and α2,6-linked sialic acids. However, the HAs of Siamese algae-eater influenza-like virus (SAEILV) and chum salmon influenza-like virus (CSILV) bind to α2,3-linked sialic acid. Furthermore, SAEILV HA binds to sialyated Lewis X, is activated by human airway enzymes, and is fusogenic over a broad pH range. SAEILV neuraminidase (NA) has a highly conserved active site and a similar structure to other known NAs. We also determined the cryo–electron microscopy structure of the HA of a previously described virus from the same sister clade, the Wuhan spiny eel influenza virus (WSEIV). No cross-reactive antibodies against these HAs or NAs were found in human serum, suggesting that humans are immunologically naïve to these viruses.

Fish and amphibian influenza B–like viruses show functional divergence and lack immunity in humans.

INTRODUCTION

Influenza viruses are highly contagious respiratory pathogens responsible for seasonal epidemics and, occasionally, global pandemics, posing a major threat to public health and the economy worldwide. Influenza viruses are classified into four distinct types: A, B, C, and D (1, 2). Waterbirds and shorebirds serve as natural reservoirs for influenza A viruses, although zoonotic spillover events have resulted in influenza A virus transmission to a wide variety of avian and mammalian species, including humans, terrestrial birds, pigs, poultry, horses, cats, cows, bats, marine mammals, etc. (3, 4). Influenza A H1N1 and H3N2 virus strains are currently circulating in humans, but 19 HA subtypes and 11 NA subtypes have been detected in the animal reservoir. The introduction of other subtypes—to which humans have little to no immunity—into the population carries the potential to cause devastating pandemics (1). In contrast, the influenza B virus is relatively stable due to its limited antigenic drift and narrow host range (5–7). Influenza B viruses are primarily restricted to human hosts and are known to cause seasonal and epidemic outbreaks of influenza (8). However, sporadic reports of infections detected in harbor and gray seals as well as bamboo rats have raised questions about the possibility of nonhuman influenza B virus reservoirs (5, 9–11). Phylogenetic analysis of viruses isolated from seals suggests that the influenza B virus detected there was of human origin (12). Thus far, there has been limited evidence showing sustained animal reservoirs of influenza B virus strains and there have been no documented cases of influenza B virus transmission from animals to humans.

Despite extensive research on the characterization of influenza viruses, our overall understanding of the true diversity and evolutionary history of vertebrate RNA viruses, particularly concerning those found outside of mammalian and avian hosts, remains limited. A 2018 study identified 214 previously unidentified vertebrate-associated RNA viruses across more than 186 host species by using a large-scale metatranscriptomic approach. Influenza-like viruses discovered in fish and amphibians included Wuhan Asiatic toad, Wuhan spiny eel, and Wenling hagfish influenza-like viruses (13) of which the Wuhan spiny eel virus (WSEIV) appeared to be influenza B virus–like (14).

Recently, Parry et al. (15) identified the complete coding segments of five divergent vertebrate influenza-like viruses. Two amphibian influenza-like viruses exhibit relatively high pairwise amino acid identity to segments from the influenza D virus. The other three viruses were found to be closely related to influenza B virus: Salamander influenza-like virus (SILV), Siamese algae-eater influenza-like virus (SAEILV), and chum salmon influenza-like virus (CSILV), detected in Mexican walking fish (Ambystoma mexicanum) and plateau tiger salamander (Ambystoma velasci), Siamese algae-eater fish (Gyrinocheilus aymonieri), and chum salmon (Oncorhynchus keta), respectively. Furthermore, the genome arrangements of these three viruses are similar to those of influenza A and B viruses, containing eight segments (PB2, PB1, PA, HA, NP, NA, M, and NS). Sequence alignment and phylogenetic analysis of the coding regions indicate that these three viral genomes display the closest similarities to the influenza B virus species. The amino acid percentage identity between SILV and influenza B viruses (B/Victoria/593/2011) ranges from 75.57% (PB1) to 25.69% (M2); between SAEILV and influenza B viruses (B/North Carolina/11/2019), it ranges from 69.46% (PB1) to 24.56% (NEP); and between CSILV and influenza B viruses (B/North Carolina/11/2019), it ranges from 69.50% (PB1) to 34.53% (M1). The surface glycoproteins, hemagglutinin (HA), and the neuraminidase (NA) of SILV exhibit 43 to 46% identity, whereas those of SAEILV show 34 to 43% identity, and CSILV HA and NA display 32 to 38% amino acid identity to the respective HA and NA of influenza B viruses. These findings suggest the existence of a vast number of uncharacterized influenza B–like viruses and highlight our limited understanding of the origin, evolution, and cross-species transmission of these viruses.

The viral surface glycoproteins HA and NA play a critical role in the life cycle of the influenza A and B viruses. HA mediates viral attachment and entry into host cells through interactions with sialic acid containing glycans on the cell surface (16, 17). NA cleaves the glycosidic linkage between the terminal sialic acids and galactose, facilitating the release of viral progeny from the cell surface (18). Influenza B virus HA can interact with both α2,3- and α2,6-linked N-acetylneuraminic acid (Neu5Ac) (19). In contrast, WSEIV, an influenza B–like virus, lacks the ability to bind to the canonical influenza B virus receptors. Instead, WSEIV uses a sialylated ganglioside, GM2, as its receptor (14). On the basis of NA sequence comparisons, the nine subtypes of the influenza A virus NAs can be categorized into two groups: group 1 (N1, N4, N5, and N8) and group 2 (N2, N3, N6, N7, and N9). The NA-like proteins from bat influenza viruses H17N10 and H18N11 do not exhibit NA activity despite their structural similarity with authentic NAs (20–22). However, the NA protein of WSEIV has NA activity similar to the influenza B virus NA, despite low sequence identity (14, 23). Collectively, these findings underscore the need for studies characterizing influenza B–like viruses in undersampled hosts, as well as for understanding the functionalities and antigenic properties of these viruses. To better understand the complexity and divergence of HAs and NAs from genetically distinct influenza B–like viruses, we report the functional, antigenic, and structural characterization of the SILV, SAEILV, and CSILV glycoproteins and structural characterization of the HA of the previously described WSEIV.

RESULTS

Phylogenetic and comparative sequence analysis

The phylogenetic analysis of representative influenza A and B virus HA and NA genetic sequences with the HA and NA of SILV, SAEILV, CSILV, and WSEIV indicated that these influenza-like viruses were more closely related to influenza B viruses than influenza A viruses (Fig. 1, A and B). The position of the sequences strongly suggests that these viruses have a long evolutionary history in vertebrates. Identical amino acid residues of each glycoprotein revealed by paired sequence analysis with the reference strain B/Brisbane/60/2008 (B/Bris/08) were highlighted in blue on top of the structures of B/Bris/08 HA and NA to indicate conserved patches. Compared to the HA and NA of B/Bris/08, SILV showed ~42% (HA) and ~44% (NA) amino acid identity (Fig. 1, C and F), SAEILV was ~30% (HA) and ~40% (NA) identical (Fig. 1, D and G), and CSILV ~28% (HA) and ~34% (NA) identical (Fig. 1, E and H). Multiple sequence alignment of the influenza-like viruses with the influenza B/Malaysia/2506/2004 virus HA and NA showed distinct glycosylation patterns and receptor binding site (RBS) residues for each strain (Fig. 1I). In contrast, comparison of the NA sequences demonstrated that residues within the enzymatic active site are highly conserved among influenza B–like viruses, along with the regions in the immediate vicinity of the enzymatic site (Fig. 1, F to H and J). In the context of the HA, we noticed a lack of conservation in the residues that form the sialic acid–interacting RBS (24, 25), suggesting a potentially altered receptor binding profile for these influenza B–like HAs. Furthermore, these HA sequences appeared to have a reduced number of putative N-linked glycosylation sites compared to the B/Malaysia/2506/2004 HA, whereas the N336 glycan remained conserved in all the sequences analyzed. To analyze the extend of glycosylation, the HAs were expressed as recombinant protein trimers in insect cells using the baculovirus expression system as described in a prior study (26) and analyzed with and without PNGaseF treatment via sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) in which an increased electrophoretic mobility indicates a loss of molecular weight due to the removal of N-linked glycans. As a control, we used recombinant B/Malaysia/2506/2004 HA, which is part of the B/Victoria/2/1987-like lineage, also expressed using the baculovirus expression system as described previously (27). Compared to B/Malaysia/2506/2004 HA, SILV, SAEILV, and CSILV HAs showed only a modest change in molecular weight after deglycosylation (Fig. 1K). This confirms the sequence-based observation that fewer glycosylation sites are present on these HAs. From an antigenic perspective, the target epitope of the pan-influenza virus HA monoclonal antibody (mAb) CR9114 also exhibited some mismatches (25). However, a substantial number of matched residues were found in the stalk domain of the HAs, consistent with previous findings of higher levels of conservation in this region across all influenza virus HAs (28). Mismatches were observed in the region immediately upstream of the fusion peptide, which is largely conserved and encompasses the proteolytic cleavage site essential for HA activation.

Fig. 1. Phylogenetic and comparative sequence analysis of SILV, SAEILV, and CSILV glycoproteins.

Phylogenetic analysis of the SILV, SAEILV, and CSILV HA and NA amino acid sequences and representative sequences from influenza A and B virus HAs (A) and NAs (B) were performed by the maximum likelihood method. The scale bar shows estimated amino acid substitutions per site. Conserved amino acid residues of the SILV HA and NA (C and F), SAEILV HA and NA (D and G), CSILV HA and NA (E and H), respectively, relative to influenza B virus HA and NA are highlighted in blue on top of the publicly available structure of influenza B/Brisbane/60/2008 HA (PDB 4FQM) and NA (PDB 4CPL), visualized by ChimeraX. The HA RBS, central stem (CS) epitope, and NA active site are circled in red, orange, and purple. Amino acid sequence alignment of the SILV HA and NA, SAEILV HA and NA, CSILV HA and NA, and WSEIV HA and NA are displayed against influenza B/Malaysia/2506/2004 HA and NA. (I and J) Key features are shown in different colors. Asterisks indicate identical amino acids. SDS-PAGE analysis of recombinant expressed SILV, SAEILV, and CSILV HAs (K) and SAEILV NAs (L) and influenza B/Malaysia/2506/2004 HA and NA under nondeglycosylated and deglycosylated conditions.

It is worth noting that we were only able to produce SAEILV NA protein recombinantly in our baculovirus expression system. An N-terminal vasodilator stimulating phosphoprotein (VASP) tetramerization domain was used to maintain the tetramerized structure of the head domain of NA. The PNGaseF treatment of the B/Malaysia/2506/2004 NA and SAEILV NA did not reveal any overt differences in glycosylation patterns across the glycoproteins with comparable shifts in molecular weights pre- and postdeglycosylation (Fig. 1L).

Receptor specificity

To study the receptor binding specificity of these influenza B–like viruses, we performed a hemagglutination assay that exploits the ability of the HA to bind sialic acid receptors on red blood cells (RBCs), thereby cross-linking them and preventing gravity-induced pelleting (29). We expressed and tested the recombinant HA proteins (SILV, SAEILV, and CSILV) as soluble trimers in the baculovirus system. Recombinant SILV failed to induce hemagglutination, whereas both SAEILV and CSILV caused hemagglutination with both chicken and turkey RBCs (Fig. 2A), suggesting that they can recognize canonical influenza virus receptors. The absence of hemagglutination was also noted for the recently identified influenza B–like WSEIV, which binds to GM2 and for H17, H18, and H19, which use the major histocompatibility complex as a receptor, raising questions about the receptor usage of the SILV HA (14, 21, 22). To investigate the receptor usage of these influenza B–like viruses more comprehensively, we used glycan microarrays containing synthetic glycans with α2,3- and α2,6-linked sialic acid (Neu5Ac) presented as linear structures and bi- and triantennary glycans including sialyl Lewis X (SLe^x) structures (17, 30). The same approach had been previously used to identify a sialylated ganglioside receptor for WSEIV (14). As expected, the B/Netherlands/2914/2015 control bound to both glycans containing α2,6-linked Neu5Ac or α2,3-linked Neu5Ac (Fig. 2E). However, we did not detect binding of the recombinant SILV HA to any of the glycans present on the microarray (Fig. 2B), which agrees with the lack of hemagglutination activity. We also did not detect any binding to glycolipids as we did for the WSEIV (fig. S1A) (14). SAEILV HA showed robust binding to α2,3-linked and to SLe^x structures and even appeared to bind some α2,6 Neu5Ac (Fig. 2C). CSILV HA had a more restricted binding repertoire and predominantly bound to α2,3-linked Neu5Ac (Fig. 2D). CSILV apparently needs at least two α2,3-linked Neu5Ac within an N-glycan as only compounds 10 and 21 are bound and not compounds 6 and 17; furthermore, SAEILV was able to bind linear structures whereas CSILV was not. In contrast to SAEILV, CSILV was not able to accommodate SLe^x, also not when presented as complex N-glycans as structures 11 and 22. Conclusively, all three HAs have a distinct receptor binding specificity with that of the SILV being unknown as we detected no binding on the array, whereas SAEILV and CSILV both can bind canonical avian-type receptors.

Fig. 2. Functional profile of SILV, SAEILV, and CSILV HAs.

(A) Hemagglutination assays were performed using recombinant HAs from SILV, SAEILV, and CSILV with turkey and chicken RBCs. HA proteins were added at 10 μg/ml and serially diluted twofold. (B to E) Glycan microarray analysis was used to assess binding of recombinant SILV, SAEILV, and CSILV HAs to synthetic glycans with α2,3- or α2,6-linked sialic acids (Neu5Ac), including linear, biantennary, triantennary, and SLe^x structures. B/Netherlands/2914/2015 HA was used as a positive control. Glycans A to M and O to W represent linear and biantennary structures, whereas glycans 1 to 22 represent triantennary N-glycans with LacNAc extensions. Glycans are color coded by terminal motif: yellow (no NeuAc), pink (α2,6-NeuAc), white (α2,3-NeuAc), blue (Lex), and black (SLex). Bars with dual colors indicate distinct epitopes on different arms. Data are shown as mean RFUs ± SD. (F) HA cleavage was assessed in HEK293T cells cotransfected with plasmids expressing full-length HAs and human airway proteases. After 48 hours, cleavage of HA0 to HA1 was evaluated by Western blot. Cells transfected with pCAGGS-HA, treated or untreated with TPCK-treated trypsin, served as controls. (G) Cell-cell fusion was assessed in HeLa cells expressing the respective HAs. Following trypsin treatment and low-pH exposure, polykaryon formation was evaluated. Representative images show fusion induced by SILV, SAEILV, CSILV, and B/Malaysia/2506/2004 HAs. White arrows indicate polykaryons, with insets showing magnified views of the fused cells. UT, untransfected.

Cleavage profile of HAs by human airways proteases

The activation of viral glycoproteins by host proteases is often necessary for viral fusion, infection, and pathogenicity. HA usually binds to sialyated glycans on the host cell surface from where the virus can then be transported into the cell via endocytosis. To gain membrane fusion competence, HA relies on host cell proteases for the cleavage of the HA0 precursor into the HA1 and HA2 polypeptides (31). A 2020 study investigated the differences in preferential proteolytic activation of influenza A and B virus HAs by human respiratory epithelium enzymes (32). Influenza B virus HA0 has been found to be activated by a broader panel of type II transmembrane serine proteases (TTSPs) and kallikrein (KLK) enzymes with much higher efficiency than influenza A virus HA0. Given the established involvement of TTSPs and KLKs in the cleavage and activation of influenza B viruses (32), we aimed to investigate the extent to which of these proteases were able to activate the HAs of these influenza B–like viruses. Human embryonic kidney (HEK) 293T cells were cotransfected with pcDNA3.1 plasmids encoding human airway enzymes and pCAGGS expression plasmids encoding the corresponding full-length HA of SILV, SAEILV, CSILV, or B/Malaysia/2506/2004 HA (Fig. 2F). Cells transfected with only the pCAGGS expression plasmid encoding the corresponding full-length HA served as an untreated control. For trypsin control, pCAGGS HA transfected cells were incubated briefly with N-tosyl-l-phenylalanine chloromethyl ketone (TPCK)–treated trypsin before harvesting the cells. At 48 hours posttransfection, the transfected cells were harvested for detection of HA cleavage by Western blot. Cleavage was detected using polyclonal antisera generated in mice immunized with each corresponding recombinant HA. As evidenced by the presence of both an HA1 band (~50 kDa) and an HA0 band (~75 kDa), B/Malaysia/2506/2004 HA was cleaved by various human airways enzymes. No cleavage was observed in cells that were transfected with only the pCAGGS HA plasmid. Notably, none of the selected proteases or trypsin was able to cleave and activate the SILV and CSILV HA. On the other hand, SAEILV HA0 was cleaved by TMPRSS4, TMPRSS13, KLK5, KLK14, and trypsin. The percent cleaved HA was calculated relative to that of the trypsin control, consisting of cells transfected with corresponding HA and briefly exposed to exogenous trypsin (fig. S1B).

Activation pH of HAs from influenza B–like viruses

We next investigated whether HA0 cleavage translated into HA activation and subsequent membrane fusion using a polykaryon assay. HeLa cells transfected with a pCAGGS plasmid expressing full-length HA were treated with trypsin to activate the HA on the cell surface. Cells were then exposed to different pH conditions (4.5 to 5.9) for 15 min to promote membrane fusion. Cell-cell fusion was detected via the presence of polykaryons. Only cells transfected with SAEILV HA induced polykaryon formation at a variety of different pH values. Cells treated at a pH of 7.4, and untransfected cells did not form any syncytia, showing that the polykaryon formation was HA specific and pH dependent (Fig. 2G). Because SILV HA was not cleaved by trypsin and CSILV only to a low degree, it is not unexpected that they did not show fusion activity in this assay. Their pH of fusion remains to be determined once appropriate proteases for their cleavage have been identified.

The structural basis of influenza B–like HA receptor diversity

We were also interested in determining and comparing the structures of HAs of influenza B–like viruses with those of bona fide influenza B viruses. We first determined the cryo–electron microscopy (cryo-EM) structure of the previously described influenza B–like WSEIV HA at a 2.67-Å resolution and observed HA1 and HA2 subunits, glycans, RBS, and fusion peptide. Of the three predicted glycosylation sites in the WSEIV HA sequence, clear glycan densities were observed on residues N36 and N336 but not on the HA2 C-terminal proximal residue N503 (Fig. 3A). The overall WSEIV HA trimer resembles the influenza B/Bris/08 virus HA [Protein Data Bank (PDB) 4FQM] (25), despite only sharing 44% amino acid sequence homology. There was a slight shift of angle in the central stem helix between WSEIV HA and B/Bris/08 HA, which led to a more drastic shift of the HA head (Fig. 3B). Surface charge analysis revealed distinct charge patterns on the HA head: WSEIV HA had negatively charged patches within and around the RBS, whereas B/Bris08 HA had positively charged patches within and around the RBS (Fig. 3C). In addition, the key residues within the RBS pocket of WSEIV HA are different from B/Bris/08 HA (Fig. 3D). These discrepancies in the RBS also support different receptor preferences for these influenza B–like HAs.

Fig. 3. WSEIV HA cryo-EM structure resembles influenza B virus HA.

(A) Cryo-EM map and model of WSEIV HA showing the HA1/HA2 subunits, fusion peptide, RBS pocket, and N-linked glycans. (B) Models of WSEIV HA and B/Brisbane/60/2008 HA (PDB 4FQM) are aligned by the central helix of HA2. (C) The surface of WSEIV HA and B/Brisbane/60/2008 HA is colored by electrostatic potential using ChimeraX. (D) Key residues within the RBS of WSEIV HA and B/Brisbane/60/2008 HA. Ab initio reconstruction maps and selected 2D classes of SAEILV HA (E) and SILV HA (F) from nsEM.

Our attempts to obtain high-resolution cryo-EM structures of SAEILV/CSILV/SILV HA were unsuccessful. However, low-resolution negative stain EM (nsEM) maps were acquired for SAEILV and SILV HA (Fig. 3, E and F). The three-dimensional (3D) reconstruction maps show that SILV HA resembles the canonical shape of influenza virus HA, whereas SAEILV HA has an open and flexible head domain. The unstable head domain in the recombinant HA protein makes structural determination difficult. Next, we predicted the models of the four influenza B–like virus HAs using the AlphaFold3 (AF3) server (33). The predicted models showed overall high confidence of accuracy on the protein backbones of the HA ectodomains, especially on the HA head regions, but the model quality of the cleavage sites and fusion peptides was quite poor (fig. S2A). The AF3 predicted model of WSEIV HA highly resembles the cryo-EM structure with a root mean square deviation (RMSD) of 1.298 Å (fig. S2B). The predicted models also revealed divergent key residues located within the RBS pockets (fig. S2C and table S1), which provide structural basis for their distinct receptor binding profiles (Fig. 2, B to E).

NA activity and sensitivity to multiple NA inhibitors

NA is a fascinating protein that plays multiple essential roles in the influenza virus life cycle. Our efforts to express recombinant protein from SILV and CSILV NAs were unsuccessful, thereby hindering further functional and structural characterization of these previously unidentified influenza B–like virus NAs. Given that the primary active-site residues that are crucial for the sialidase activity of the NA are conserved, we examined the functional activity of the SAEILV NA by performing an enzyme-linked lectin assay (ELLA) using fetuin as a substrate. Fetuin contains mono-, bi-, and triantennary glycans with α2,3- and α2,6-linked sialic acids in a 2:1 ratio (34). Fetuin-coated 96-well plates were incubated with serially diluted proteins overnight at four different temperatures (4°, 20°, 33°, and 37°C) to determine the temperature-dependent profile of the NAs. These temperatures were partially chosen to cover the range of temperatures that Siamese algae eater fish might experience in their freshwater habitat. NA enzymatic activity was quantified using peanut agglutinin (PNA) lectin, which binds to exposed galactose residues that remain once terminal sialic acid is cleaved off. B/Malaysia/2506/2004 was used as a comparative control. We found that SAEILV NA has NA activity, and enzymatic activity is also slightly higher compared to B/Malaysia/2506/2004 NA at the tested temperatures (Fig. 4A). The inverse of the half-maximum lectin binding was used to calculate the specific activity of the NA to better understand the temperature dependency (Fig. 4B). As expected, we found that each NA’s activity gradually decreased at lower temperatures. In addition, the pattern of specific activity profiles for the B/Malaysia/2506/2004 NA and the SAEILV NA were nearly the same.

Fig. 4. NA specific activities and cleavage specificities of the SAEILV NA.

(A) NA enzymatic activity of recombinant SAEILV and B/Malaysia/2506/2004 NA proteins was examined via ELLA using fetuin, at four different temperatures: 4°, 20°, 33°, and 37°C. The curves indicate absorbance measured at 490 nm with error bars indicating the SD. OD, optical density. (B) Specific enzymatic activity (inverse of half-maximum lectin binding) is shown for each individual NA at each temperature. (C) Inhibitory susceptibility to three NA inhibitors (oseltamivir, peramivir, and zanamivir) was evaluated in an ELLA-based NA inhibition assay. (D) Specific activity of SAEILV and B/Malaysia/2506/2004 NA as determined by ELLA using different glycoprotein-lectin combinations normalized to the fetuin-ECA.

To explore the susceptibility of the SAEILV NA to clinically approved NA inhibitors, the median inhibitory concentration (IC₅₀) of oseltamivir, peramivir, and zanamivir was determined using an ELLA-based NA inhibition assay. In concordance with previous findings for WSEIV NA (14, 23), SAEILV NA was also more sensitive to peramivir and zanamivir than oseltamivir (Fig. 4C). These results support our finding that residues within the enzymatic active site are highly conserved between influenza B virus NA and SAEILV NA, enabling NA inhibitor binding.

Enzymatic activity of the SAEILV NA for differently linked sialic acids

Next, we examined the activities of the SAEILV NA for differently linked sialic acids using lectins with different binding specificities in a previously described ELLA protocol (35). Lectins with different binding specificities were used to determine whether the NAs preferentially cleaved α2,3- and α2,6-linked sialic acid. Erythrina crista-galli agglutinin (ECA) and PNA bind to nonsiaylated N- and O-linked sugars, respectively, and their binding would be, in general, higher in the presence of active NA (36, 37). Maackia amurensis lectin I (MAL I) and Sambucus nigra lectin (SNA) specifically bind α2,3- or α2,6-linked Neu5Ac, respectively, and their binding decreases with NA cleavage of the target substrates (25, 38). Specific activities (activity per amount of protein) of the SAEILV and B/Malaysia/2506/2004 NA proteins were determined for each glycoprotein-lectin combinations and plotted relative to the specific activity determined for fetuin-ECA. In line with the enzymatic activity, both proteins have equal substrate specificity and comparable cleavage profiles (Fig. 4D).

The structural basis of NA active sites

To structurally compare the NA active site, we obtained a 2.75-Å cryo-EM 3D reconstruction of the SAEILV NA head (Fig. 5A). We observed cryo-EM density for one glycosylation site at position N281 located on the underside of NA (Fig. 5A). This glycan is also conserved in CSILV NA, previously reported WSEIV NA (PDB 7XVU) (23) as well as influenza B virus NA (Figs. 1J and 5B). Despite the ~40% sequence identity, the catalytic residues within the NA active site are highly conserved among SAEILV, WSEIV, and influenza B virus NA, with only one exception of a W407G mutation in SAEILV NA (Fig. 5B and table S1). The conserved residues in the NA active-site pocket provide the structural basis of the susceptibility of SAEILV NA to NA inhibitors. On the contrary, the residues around the NA active site are less conserved as indicated by different surface charge patterns of SAEILV, WSEIV, and influenza B virus NAs (Fig. 5C), which also suggests that their preference for substrates could be different. In addition, we observed the potential density of a calcium ion in the cryo-EM reconstruction map (Fig. 5D), indicating that the activity of SAEILV NA might be Ca²⁺ dependent, like NAs for seasonal influenza A and B viruses. We superimposed the structure of 1G01+ A/California/04/2009 (H1N1) NA (PDB 6Q23) (38) with SAEILV NA and observed a clash between SAEILV NA Y239 and the 1G01 HCDR3 loop. In addition, A/California/04/2009 N1 K150/K432 form additional salt bridges with 1G01, whereas SAEILV NA lacks such interactions (Fig. 5E). These structural discrepancies likely explain why the broadly reactive NA antibody 1G01 does not bind SAEILV NA (Fig. 6C). Like HA, we used AF3 to generate models from the sequences of CSILV and SILV NAs. Both models were predicted with >90% overall accuracy on the tetrameric NA head (fig. S2, D and E). The AF3 predicted SAEILV NA structure highly resembles our cryo-EM structure with an RMSD of 0.752 Å on the NA protomer (fig. S2F). Again, the key residues within the NA active site are highly conserved among these three NAs (fig. S2G and table S1).

Fig. 5. SAEILV NA cryo-EM structure resembles influenza B virus NA.

(A) Cryo-EM map of SAEILV NA showing the tetrameric NA head, active site, and N-linked glycan. (B) Aligned models of SAILV NA, WSEIV NA (PDB 7XVU), and B/Brisbane/60/2008 NA (PDB 4CPL) showing conserved secondary structures, key active-site residues (B NA numbering), and underside glycans among three NAs. (C) The surface and the active-site pockets of SAEILV NA, WSEIV NA, and B/Brisbane/60/2008 NA are colored by electrostatic potential using ChimeraX. (D) Cryo-EM map and model showing the conserved calcium binding pocket in SAEILV NA. (E) Aligned models of SAEILV NA and 1G01 + A/California/04/2009 (H1N1) showing that 1G01 HCDR3 clashes with SAEILV NA Y239.

Fig. 6. Binding profiles of the anti–influenza B virus HA and NA mAbs.

(A and C) Broadly cross-reactive anti–influenza B virus HA and NA human and mouse mAb binding to recombinant protein (SILV, SAEILV, CSILV, and B/Malaysia/2506/2004 HA and SAEILV and B/Malaysia/2506/2004 NA) in ELISA. (B) Anti–influenza B virus human and mouse mAb binding to cell surface expressed NAs of SILV, SAEILV, CSILV, and B/Malaysia/2506/2004 by flow cytometry. (D) Cross-reactive antibody profile of sera from postseasonal influenza vaccination recipients. Antibodies against SILV, SAEILV, CSILV HA, and SAEILV NA were determined via ELISA. Mopeia virus glycoprotein was used as the negative control for baseline establishment.

Characterizing the binding profile of mAbs using enzyme-linked immunosorbent assay and immunofluorescence

Next, we focused on characterizing the binding of broadly protective monoclonal antibodies to these influenza B–like glycoproteins (Fig. 6, A to C). We used a panel of broadly cross-reactive human and mouse monoclonal antibodies, which had been earlier characterized to bind to diverging lineages of influenza B virus (25, 38–40). To do so, enzyme-linked immunosorbent assay (ELISA)– and flow cytometry–based cellular assays were performed and robust binding for most of the tested human and mouse monoclonal antibodies was observed for B/Malaysia/2506/2004 HA and NA, which served as our experimental control. We did not detect any binding to SILV, SAEILV, and CSILV HAs by any of the human and mouse monoclonal antibodies. Only a positive control antibody targeting the His-tag on the recombinant proteins produced a signal, showing that the proteins were actually coated on the plates. As aforementioned, our efforts to express recombinant protein from SILV and CSILV NAs were unsuccessful, and we took two different approaches to characterize the binding profiles of these NAs with a panel of human and mouse mAbs. Of the panel of human and mouse mAbs used to probe SAEILV NA (which we could produce as recombinant protein), only one antibody, 2E01, showed low but detectable binding in ELISA– and flow cytometry–based assay and in an immunofluorescence (IF) assay (Fig. 6, B and C, and fig. S3C). Our group earlier showed that anti–influenza B NA-mAb 2E01 demonstrated remarkable NA inhibition and binding activities against viruses belonging to the B/Yamagata/16/88-like and B/Victoria/2/87-like lineages and the ancestral B/Lee/1940 strain, which cumulatively span more than 70 years of antigenic drift (41). A positive control in IF assay, mouse antisera generated against SAEILV NA raised by vaccinating mice with recombinant SAEILV NA protein, was used. This serum detected SAEILV NA expression on transfected cells. We did not detect any binding in flow cytometry or IF for SILV or CSILV NAs, but because we did not have specific sera as positive controls, we can also not prove that the transfected cells actually expressed these NAs. 4F11 (anti–influenza B virus NA) and 4C2 (anti–influenza B virus HA) mouse mAbs were used as the negative control for the assay against the respective HAs and NAs and showed no signal as expected.

Last, to determine whether humans have immunity to these HAs and NAs, we performed ELISAs with a panel of human sera (Fig. 6D). A glycoprotein from an arenavirus, Mopeia virus, to which humans are naïve, was used as the negative control for baseline establishment. No reactivity above baseline was found against the HAs and NA of influenza B–like viruses (whereas reactivity to bona fide HA and NA of influenza B virus was high as expected), further supporting the finding that there is limited conservation of cross-reactive epitopes as also shown by our panel of mAbs.

DISCUSSION

Influenza viruses are well known to infect a broad spectrum of host species; however, there have been few studies of influenza-like viruses from fish and amphibians. In a recent study, HA and NA sequences were identified in gills and respiratory tissue samples from salamander as well as Siamese algae and chum salmon fishes (15). Here, we report the functional and structural characterization of HA and NA of influenza B–like viruses from fish and amphibians. The identification of influenza viruses in fish and amphibians that are genotypically and phenotypically distinct from other influenza A viruses and clusters close to influenza B viruses suggest the occurrence of prolonged virus-host codivergence with several host-switching events over time (13, 15). We found that, like the previously described WSEIV HA (14), SILV HA does not bind to canonical human or avian receptors. This lack of canonical receptor binding is likely due to specific structural features of the putative RBS of SILV and WSEIV HA. The WSEIV HA structure contains a smaller RBS pocket with negatively charge patches within and around the RBS; this may result in steric or charge constraints that limit canonical α2,3- and α2,6-linked Neu5Ac binding. CSILV and SAEILV, on the other hand, do interact with canonical avian-type receptors and are different from each other in their ability to interact with SLe^x.

Notably, SILV, SAEILV, CSILV, and WSEIV NA show the highest similarity to influenza B virus NA (B/Brisbane/60/2008) with amino acid sequence identities of 44, 40, 34, and 48%, respectively. Although this is a relatively low sequence identity, these NAs display similar overall structures to other typical influenza virus NAs with a conserved active site. Our data clearly showed that SAEILV NA has similar NA activity as typical influenza B virus NAs and is sensitive to multiple NAIs like zanamivir, oseltamivir, and peramivir, which is in concordance with the previous studies on WSEIV NA (14, 23). In addition, we found limited antigenic conservation between these influenza B–like viruses and bona fide influenza B viruses. No antibody-based immunity was detected against the HA and NA of these influenza B–like viruses in humans.

In this study, we resolved the high-resolution structures of WSEIV HA and SAEILV NA by cryo-EM and performed nsEM 3D reconstructions of SAEILV HA and SILV HA. However, we were unable to observe the trimeric CSILV HA by nsEM, suggesting that the recombinant CSILV HA is unstable and might exist in a monomeric form. AF3 predicted models were highly consistent with our WSEIV HA and SAEILV NA cryo-EM structures, suggesting that the AF3 models of the remaining HAs are reliable preliminary models that can be used to facilitate future attempts to further stabilize these proteins for recombinant expression.

Seasonal epidemics caused by influenza B viruses were responsible for up to 52% of influenza-associated pediatric mortality during the past 15 years (42, 43). Thus, influenza B virus surveillance and diagnostics in nonhuman hosts are critically important and improve our understanding of influenza B virus ecology and diversity. Looking at influenza B–like viruses from other animal lineages and vertebrate classes will add to our understanding of the origins and evolution of this important group of viruses.

In summary, the data from our structural and functional characterization of the influenza B–like viruses from fish reveal that the SILV HA is functionally divergent from influenza B virus HA. The SILV HA does not induce hemagglutination with chicken or turkey RBCs nor does it bind to canonical human or avian receptors. Conversely, CSILV and SAEILV, on the other hand, do interact with canonical avian-type receptors. In addition, SAEILV was activated by human airway enzymes and is fusogenic at a wide range of pH conditions. No cross-reactive antibodies were found in the human serum, suggesting that humans are immunologically naïve to these glycoproteins. We demonstrated that SAEILV NA protein has a canonical overall structure and function compared to other typical influenza virus NAs and is sensitive to all currently used NAIs. Our detailed structural studies of WSEIV HA and SAEILV NA provide insights into understanding the previously unidentified influenza B–like virus HAs and NAs and understanding the mechanism of lack of binding to pan-NA mAb 1G01.

MATERIALS AND METHODS

Cell and recombinant proteins

Sf9 cells [CRL-1711, American Type Culture Collection (ATCC)] used for baculovirus rescue were cultured in Trichoplusia ni medium-formulation Hink insect cell medium (TNM-FH, Gemini Bioproducts), supplemented with 10% fetal bovine serum (FBS; Gibco) and a penicillin (100 U/ml)–streptomycin (100 μg/ml) solution (Gibco). High Five cells (BTI-TN-5B1-4 subclone; Vienna Institute of Biotechnology) (44), used for protein expression, were grown in serum-free Sf-900 medium (Gibco) supplemented with penicillin (100 U/ml)–streptomycin (100 μg/ml) solution (Gibco). HEK293T (CRL-3216, ATCC) cells were propagated using Dulbecco’s modified Eagle’s medium (DMEM; Gibco) supplemented with 10% FBS, penicillin (100 U/ml)–streptomycin (100 μg/ml) solution and 1% 1 M (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (Hepes). HeLa cells (CCL-2, ATCC) were grown in Eagle’s minimum essential medium (EMEM) supplemented with 10% FBS and penicillin (100 U/ml)–streptomycin (100 μg/ml) solution.

Nucleotide sequences coding for the HA and NA proteins of SILV (HA: MT926392; NA: MT926391), SAEILV (HA: MT926384; NA: MT926383), and CSILV (HA: MT926407; NA: MT926406) were synthesized commercially (Genewiz). Recombinant HA and NA proteins were expressed in the baculovirus expression system, as described in a previous report (26). Ectodomains from SILV HA (Met¹-Gly⁵²⁹), SAEILV HA (Met¹-Leu⁵¹⁵), CSILV HA (Met¹-Val⁵²¹), and WSEIV HA (Met¹-Thr⁵²²) were cloned into modified pFastBacDual transfer plasmids containing a T4 trimerization domain with a thrombin cleavage site and a hexahistidine purification tag (His-tag). Coding sequences from SILV NA (Ser⁷⁹- Ala⁴⁶⁷), SAEILV NA (Ser⁸²-Ala⁴⁶⁸), and CSILV NA (Ser⁸¹-Leu⁴⁶⁴) were cloned also into modified pFastBacDual transfer plasmids, which included a signal peptide, a hexahistidine purification tag (His tag), a VASP tetramerization domain, and a thrombin cleavage site. The recombinant pFastBac constructs were introduced into DH10Bac bacteria (Invitrogen). Transformed colonies were selected, screened, and cultured for the extraction of bacmid DNA. This bacmid DNA was used to transfect Sf9 cells to rescue recombinant baculovirus and then used to infect High Five cells for protein expression. Supernatants were subsequently collected by low-speed centrifugation ~72 hours postinfection, and the protein was purified by Ni²⁺–nitrilotriacetic acid resin chromatography (Qiagen).

Phylogenetic and sequences analyses

Representative HA and NA amino acid sequences of influenza A and B viruses for the phylogenetic trees were obtained from the Global Initiative on Sharing All Influenza Data (GISAID) and GenBank (H1 to H18: EPI1349891, EPI899625, EPI673678, EPI1007628, EPI942074, EPI1154383, EPI1090164, EPI1154159, EPI1103524, EPI953583, EPI774886, EPI1007631, EPI967018, EPI750076, EPI965435, EPI939704, EPI356309, and EPI486922; B/Lee/1940 HA: EPI243230; B/Phuket/3073/2013 HA: EPI1799824; B/Colorado/06/2017 HA: EPI969380; SILV HA: QOE76814.1; SAEILV HA: QOE76806.1; CSILV HA: QOE76830.1; WSEIV HA: AVM87624.1; N1 to N11: EPI1381203, EPI899627, EPI939823, EPI1154448, EPI1007658, EPI939830, EPI750078, EPI941550, EPI965439, EPI356311, and EPI356298; B/Lee/1940 NA: EPI366432; B/Phuket/3073/2013 NA: EPI1799823; B/Colorado/06/2017 NA: EPI969379; SILV NA: QOE76813.1; SAEILV NA: QOE76805.1; CSILV NA: QEO76829.1; and WSEIV NA: AVM87625.1). Amino acid sequences were aligned using multiple sequence comparison by log-expectation (MUSCLE), and a phylogenetic tree was constructed by the maximum likelihood method with the MEGA11 software (45) using default parameters and 1000 bootstrap replications. The tree was cleaned and edited using FigTree v1.4 (http://tree.bio.ed.ac.uk/software/figtree/), and labels along with highlights were added in Microsoft PowerPoint. Pairwise alignment of the SILV HA and NA, SAEILV HA and NA, CSILV HA and NA, WSEIV HA and NA and B/Malaysia/2506/2004 HA (ACR15732.1), and B/Malaysia/2506/2004 NA (ACR15736.1) was carried out using Clustal Omega V1.2.4. Potential N-glycosylation sites of the HA and NA were predicted using the NetNGlyc 1.0 server (https://services.healthtech.dtu.dk/services/NetNGlyc-1.0) on the basis of the sequon motif N-X-S/T, where X can be any amino acid except proline. Identical residues were indicated by asterisks. Functionally and antigenically relevant features have been annotated.

Deglycosidase treatment and SDS-PAGE

Deglycosylation was carried out using the PNGaseF kit (NEB) according to the manufacturer’s instructions. Five micrograms of recombinant protein was mixed with 1 μl of deglycosylation mix buffer 2 (NEB), and sufficient phosphate-buffered saline (PBS) was added to achieve a total volume of 8 μl. The mixture was heated at 75°C for 10 min to denature the glycoproteins and then chilled on ice. Next, 1 μl of protein deglycosylation mix II was added to the noncontrol tube and mixed gently, whereas 1 μl of PBS was added to the control tube. The reaction mixtures were incubated at room temperature (RT) for 30 min, followed by incubation at 37°C for 16 hours. Subsequently, the samples were treated with 2X Laemmli loading buffer (Bio-Rad), supplemented with 10% β-mercaptoethanol. The samples were then heated for 10 min at 98°C before loading them onto an SDS–polyacrylamide gel (4 to 20% Mini-PROTEAN TGX Precast Protein Gels, Bio-Rad) and were subsequently stained with Coomassie blue fast stain solution for 1 hour, shaking at RT. The staining solution was removed, and the gel was destained following incubations with distilled water.

Hemagglutination assay

Ten micrograms of recombinant HA proteins were serially diluted twofold in PBS in V-bottom 96-well plates and incubated with 0.5% chicken or turkey RBCs in a total volume of 100 μl per well for 45 min at 4°C. Hemagglutination of RBCs was visually assessed.

Glycan array

Glycan array binding analysis of the recombinant HAs was conducted as previously described (17, 30, 46). Briefly, recombinant hexahistidine-tagged HA was precomplexed with a mouse anti-His Alexa Fluor 647 (Thermo Fisher Scientific, catalog no. MA1-21315-A647) and goat anti-mouse Alexa Fluor 647 antibodies (Invitrogen, catalog no. A21235). This preparation was performed in 50 μl of PBS-T (PBS with 0.1% Tween 20) at a 4:2:1 molar ratio, incubated for 15 min on ice, and applied on the array for 90 min. After multiple washes with PBS-T, PBS, and deionized water, the arrays were scanned to detect HA binding using an Innoscan 710 (Innopsys). The B/Netherlands/2914/2015 strain was a kind gift from R. Fouchier (Erasmus Medical Center, Rotterdam, The Netherlands) and was detected using CR9114 and goat anti-human Alexa Fluor 647 antibodies (Invitrogen, catalog A-21445). Mean relative fluorescence units (RFU) and SD values were imported into Prism 7.0, and the corresponding graph was generated.

Protease cleavage assay

To assess HA0 cleavage in protease-expressing cells, HEK293T cells were seeded onto a poly-d-lysine–coated 12-well plate at a density of 300,000 cells per well. The following day, cells were cotransfected with 0.5 μg of the pCAGGS expression plasmid encoding the corresponding full-length HA and 0.5 μg of pCDNA3.1 plasmids encoding human airway proteases (GeneScript) using the TransIT-LT1 transfection system (Mirus Bio). At 48 hours posttransfection, the trypsin control well was exposed to TPCK-treated trypsin (5 μg/ml) for 15 min at 33°C. Cells were lysed with radioimmunoprecipitation assay (RIPA) buffer (Thermo Fisher Scientific, USA) in the presence of phosphatase/protease inhibitors cocktail (Thermo Fisher Scientific, USA). The lysate was centrifuged (15 min at 4°C and 17,005g), and the supernatant was collected. Western blot was performed according to standard protocols. Briefly, proteins were separated on a 4 to 20% Mini-PROTEAN TGX Precast Protein Gel (Bio-Rad) and transferred on a nitrocellulose membrane using the iBlot transfer and stack device (Invitrogen) at 20 V for 7 min. The membranes were blocked in 5% nonfat dry milk (Bio-Rad) in PBS-T for 1 hour at RT with shaking. The blocking solution was removed, and the corresponding anti-HA antisera (generated in-house by immunizing female BALB/c mice intramuscularly with SILV HA, SAEILV HA, and CSILV HA recombinant protein) diluted 1:2000 in PBS supplemented with 1% nonfat dry milk were added. The membranes were incubated overnight and washed three times with PBS-T. The following day, anti-mouse immunoglobulin G (IgG)–horseradish peroxidase (HRP) linked (1:5000; catalog no. 7076; Cell Signaling Technology Inc.) was added for 1 hour at RT and the membranes were washed. The membranes were developed by adding ECL prime and incubated for 5 min. Imaging was performed with a ChemiDoc MP Imaging System (Bio-Rad).

Cell-cell fusion assay

A polykaryon formation assay to determine the fusogenicity of the HAs was performed as previously described (32). HeLa cells were seeded in a 96-well tissue culture plate at a density of 20,000 cells per well. The next day, cells were transfected with a pCAGGS plasmid expressing the full-length HA using the TransIT-LT1 transfection system (Mirus Bio). After 24 hours, the growth medium was replaced with 0.2% FBS, and cells were further incubated at 33°C. The following day, cells were washed with Dulbecco’s PBS (DPBS) and treated with TPCK-treated trypsin (5 μg/ml) for 10 min at 33°C before being exposed to pH-adjusted DPBS, ranging from pH 4.5 to 5.9, adjusted with citric acid. After a 15-min acid pulse, the cells were washed twice with DPBS and subsequently incubated with DMEM supplemented with 10% FBS and penicillin-streptomycin (Pen-Strep) for 4 hours at 33°C. The cells were fixed with 3.7% (v/v) paraformaldehyde (PFA) and permeabilized with 0.1% Triton X-100 for 15 min each. Staining was performed using the HCS CellMask Green (Life Technologies) as per the manufacturer’s instructions for 30 min at 5 μg/ml. Images were acquired using an EVOS M5000 (Thermo Fisher Scientific) fluorescence microscope.

Enzyme-linked immunosorbent assays

Ninety-six-well flat-bottom Immulon 4 HBX plates (catalog no. 3855; Thermo Fisher Scientific) were coated overnight at 4°C with purified recombinant protein (2 μg/ml) at 50 μl per well in PBS. The next day, the plates were washed using PBS-T, 225 μl per well, using a BioTek 405 Microplate washer. Plates were blocked using 3% milk protein in PBS-T for 1 hour at RT. The antibodies were diluted to a starting concentration of 30 μg/ml, serially diluted 1:3 in blocking solution, and incubated for 2 hours at RT. For human sera ELISAs, the sera were diluted with initial dilution (1:100) with twofold serial dilutions and incubated at RT for 2 hours. The microtiter plates were washed three times with PBS-T, and 50 μl of goat anti-human IgG (Fab specific) HRP antibody (Sigma-Aldrich, #A0293 RRID: AB_257875) or goat anti-mouse HRP antibody (Rockland; 610-1302 RRID: AB_219656) diluted 1:3000 in blocking solution was added to all wells and incubated for 1 hour at RT. The plates were washed three times, and 100 μl of SigmaFast o-phenylenediamine dihydrochloride (OPD; Sigma-Aldrich) was added to all wells. After 10 min, the reaction was stopped with 50 μl of 3 M hydrochloric acid (Thermo Fisher Scientific) and the plates were read at a wavelength of 490 nm with a plate reader (BioTek). The data were analyzed in Microsoft Excel and GraphPad Prism 7. The data were visualized as binding curves by applying a nonlinear fit.

IF microscopy

HEK293T cells were plated at 3 × 10⁵ cells/ml in a sterile, 96-well plate and incubated overnight at 37°C with 5% CO₂ as described previously (47). The following day, cells were checked for >99% confluence and washed with 1X PBS. Cells were transfected with a pCAGGS plasmid expressing the full-length NA using the TransIT-LT1 transfection system (Mirus Bio). After 24 hours, the growth medium was replaced with DMEM supplemented with 0.2% FBS, and cells were further incubated at 33°C for another day. The following day, cells were fixed with 3.7% PFA in PBS for 1 hour at RT. Next, PFA was removed, and the cells were then washed two times with PBS and blocked with 3% milk in PBS for 1 hour at RT. The blocking solution was then removed, and primary mAbs were added to their respective wells at a concentration of 10 μg/ml in 1% milk in PBS. Primary antibodies were incubated with shaking for 1 hour at RT. Plates were then washed three times with 1X PBS and goat anti-mouse Alexa Fluor 488–conjugated antibody (Invitrogen; A-11001 RRID: AB_2534069) or goat anti-human Alexa Fluor 488 (Invitrogen; A-11013; RRID: AB_2534080) and 4′,6-diamidino-2-phenylindole (DAPI) both at 1:500 in 1% milk added to the cells for secondary staining for 1 hour. The plates were washed three times with PBS and then imaged using the EVOS M5000 (Thermo Fisher Scientific) fluorescence microscope. Overlays between red and blue channels were made in ImageJ.

Flow cytometry for cell surface staining

In a 96-well plate, HEK293T cells were transfected with a pCAGGS plasmid expressing the full-length NAs as explained above in IF assay. Cells were collected after 48-hour transfection and were washed once with 2% FBS (HyClone) in PBS [Corning, DPBS 1X (without calcium and magnesium)]. Next, cells were incubated with primary mAbs (10 μg/ml) in V-bottom 96-well plates (Greiner) at 4°C for 30 min. After washing with 2% FBS in PBS, cells were incubated with and goat anti-mouse Alexa Fluor 488–conjugated antibody (Invitrogen; A-11001 RRID: AB_2534069) or goat anti-human Alexa Fluor 488 (Invitrogen; A-11013; RRID: AB_2534080) secondary antibody at 4°C for 30 min in the dark. The samples were washed and resuspended in PBS. Flow cytometry data were acquired on a FACScanto II flow cytometer. At least 50,000 events for each sample were recorded, and data were analyzed with the FlowJo 10 software.

Enzyme-linked lectin assays

ELLAs, used to determine enzymatic activity of the NAs and their sensitivity to inhibition, were performed as described in detail in previous reports (14, 35). Flat-bottom Immulon 4 HBX microtiter plates (Catalog no. 3855; Thermo Fisher Scientific) were coated with fetuin (25 μg/ml; Sigma-Aldrich) diluted in 1x PBS, at 100 μl per well and incubated overnight at 4°C. The next day, the fetuin-coated plates were incubated with twofold dilutions of recombinant NAs starting at 10 μg/ml in sample diluent buffer [1x DPBS supplemented with 1% bovine serum albumin (BSA) and 0.5% Tween 20] overnight. To determine the temperature-dependent profile of the NAs, the plates were incubated overnight at either 4°, 20°, 33°, or 37°C. The following day, the plates were washed six times with PBS-T. PNA conjugated with HRP (PNA-HRP; Sigma-Aldrich) was diluted to 5 μg/ml in a conjugate diluent buffer (1x DPBS supplemented with 0.1% BSA) and added to the washed plates. The plates were incubated for 2 hours in the dark at RT. The plates were washed three times with PBST and developed with 100 μl of SigmaFast OPD.

To evaluate the potency of NA inhibitors on SAEILV NA, NA inhibition assays using oseltamivir, peramivir, or zanamivir were performed. Plates were coated overnight with fetuin as described above. In a separate U-bottom plate, inhibitor solutions with a starting concentration of 156.28 μM were serially diluted twofold in 1x PBS and were preincubated with 15 μg of recombinant NAs for 1 hour on a shaker at RT. The fetuin-coated plates were washed three times with PBS-T as described above. The inhibitor-protein dilutions were transferred to the fetuin plates and incubated at 37°C for 18 hours (overnight). The following day, the ELLA procedure was performed as described above. Substrate specificity characterization and specific enzyme activity determination were performed using previously described ELLAs (24). In brief, fetuin-coated plates (25 μg/ml) were incubated with serial dilutions of recombinant NA protein sample diluent buffer. After the overnight incubation at 37°C, the plates were washed and incubated with either biotinylated ECA (1.25 μg/ml; Vector Laboratories), biotinylated PNA (2.5 μg/ml; Vector Laboratories), biotinylated SNA (1.25 μg/ml; Vector Laboratories), or biotinylated MAL I (2.5 μg/ml; Vector Laboratories). The binding of ECA, PNA, SNA, and MAL I was detected using HRP-conjugated streptavidin (Thermo Fisher Scientific). Absorbance measurements were carried out at 490 nm following the development with SigmaFast OPD, and specific enzyme activity (inverse of half-maximum lectin binding) was determined following a nonlinear regression analysis in GraphPad Prism 7 and was plotted normalized to the fetuin-ECA to determine activity per amount of protein in the context of cleavage of α2,3- or α2,6-linked sialic acids.

Human serum samples

The serum samples were obtained from an observational study approved by the Mount Sinai Hospital Institutional Review Board, IRB-16-00772 (PI: V. Simon). Sera collected from 15 (nine female and six male) study participants were used. The participants were healthy adults aged 18 years or older. All study participants provided written consent before the biospecimens, and data were collected. All specimens were coded before processing and analysis. Sera were stored at −80°C until analysis.

Negative stain EM

Purified recombinant HAs or NAs were deposited on glow-discharged (PELCO easiGlow, Ted Pella Inc.) carbon-coated 400-mesh copper grids (Electron Microscopy Sciences) at a concentration of ~20 μg/ml. Excess sample was blotted with filter paper. The grids were stained two times with 2% (w/v) uranyl formate for 10 and 60 s, respectively. Excess stains were blotted with filter paper. Grids were imaged on either a 200-kV Talos (Thermo Fisher Scientific) or a 120-kV Tecnai Spirit T12 (FEI) with an Eagle charge-coupled device (CCD) 4k camera (FEI). Images were collected at a ×72,000 or ×52,000 magnification with pixel sizes of 2 and 2.06 Å, respectively. Micrographs were acquired using the EPU (Thermo Fisher Scientific) or Leginon (48), and data were processed with the Appion (49) software package or CryoSPARC 4.0 (50).

Cryo-EM grid preparation and imaging

Three microliters of purified recombinant HA or NA (0.4 to 0.8 mg/mL) was mixed with 0.5 ml of 0.7% (w/v) octyl-β-glucoside (OBG) before applying to glow-discharged grids (Electron Microscopy Services Cu 1.2/1.3 300 mesh). The grids were vitrified with Mark IV Vitrobot (Thermo Fisher Scientific) with the following settings: 4°C, 100% humidity, 0-s wait time, 4.5- to 5.5-s blot time, and blot force of 1. After freezing, cryogrids were loaded into the 300-kV FEI Titan Krios (for WSEIV HA) or the 200-kV Glacios (Thermo Fisher Scientific) (for SAEILV NA), which were equipped with K3 Summit direct electron detector cameras (Gatan) and TFS Falcon IV (Thermo Fisher Scientific), respectively. Data were collected with approximate exposures of 50 e⁻/Ų at magnifications of ×105,000 or ×190,000 for the Krios or Glacios, respectively. Data collection was automated using Leginon (48) or EPU. Further details are described in table S2.

Cryo-EM data processing

Image preprocessing was performed with the Appion software (49) package or CryoSPARC Live. Micrograph movie frames collected from Krios were aligned and dose-weighted using the UCSF MotionCor2 software, and GCTF (51) was estimated. All micrographs were processed by CryoSPARC v4.0 for particle picking and reference-free 2D classification. Initial 2D classes of high quality were used as templates for template picking of datasets followed by 2D classifications to remove bad particles. The final maps of HA and NA were generated by nonuniform refinement with C₃/C₄ symmetry, followed by Global CTF refinement and another round of nonuniform refinement. More details are described in fig. S4 and table S2.

Atomic model building and refinement

AF3 predicted structures were used as the initial WSEIV HA and SAEILV NA models for refinement. Both models were manually fit into density using ChimeraX (52) followed by iterative manual model building in Coot (53) followed by refinements with Phenix (54) and Rosetta (55). All the figures of maps and models were generated using ChimeraX.

Data analysis

Statistical analyses were performed using Prism 9 software (GraphPad). IC₅₀ and area under the curve (AUC) values were calculated using nonlinear regression (four parameters) on the basis of log₁₀-transformed protein concentrations.

Supplementary Materials

This PDF file includes:

Figs. S1 to S4

Tables S1 and S2