NA is one of the three membrane proteins expressed on IAV and IBV particles, next to HA and matrix protein 2 (M2). Label-free protein quantification of purified influenza A and B virions revealed that the NA : HA ratio ranges from 0.1 to 0.2 (6). NA is a homotetrameric type II membrane protein with a mushroom-like shape. Each protomer comprises approximately 470 amino acid residues and consists of a cytoplasmic tail, a transmembrane domain (TMD), a stalk and a head domain ( Figure 1 ).

The cytoplasmic tail of NA consists of 7 highly conserved amino acid residues. Alanine-scanning mutagenesis of the cytoplasmic tail of NA of A/WSN/33 virus indicated a role in virus budding. Notably bud release rather than bud formation was affected in the tested NA cytoplasmic tail mutants (7). The cytoplasmic tail is also important for the association of NA with lipid rafts. Lipid rafts on the apical site of polarized cells are sites where newly formed influenza virions initiate assembly and budding (8). Together with the cytoplasmic tail, the TMD plays a role in the transport of newly expressed NA to the apical plasma membrane (9). The sequence of the TMD is moderately conserved across IAV subtypes and predicted to form an alpha helix. The TMD ensures the membrane anchoring of NA and serves as a translocation signal. In addition, this domain is reported to be an important stabilizing factor for the tetrameric NA formation (10, 11). Substitution of hydrophilic residues in the TMD by alanine reduced or even abolished the interaction between the four TMDs of the NA tetramer (10).

The stalk domain connects the TMD with the catalytic head domain. There is no crystal structure available of the stalk domain. The NA stalk varies in length within and across NA subtypes, carries multiple predicted N-glycosylation sites and, with few exceptions, contains at least one cysteine residue that can form an intermolecular disulphide bond with a neighbouring NA molecule (12). Glycosylation of the stalk region may contribute to NA stability, whereas inter-stalk disulphide bond formation is important for the tetramer formation. Whereas the length of the NA stalk of human influenza viruses seems to be relatively constant, the NA stalk of avian NA subtypes, in particular N1, N2, N3, N5, N6, and N7, tolerate deletions (13). The stalk length affects NA enzymatic activity, presumably by modulating the accessibility to the sialic acid-containing substrates (14).

Crystal structures of the catalytic head domain of at least one representative NA from N1 to N9 and from influenza B NA have been resolved (15–22). The NA head domain is characterized by a six-bladed propeller that is folded around the catalytic site and which is typical for all known sialidases (23). Each blade is made up of four antiparallel β-sheets that are stabilized by disulphide bonds and connected by loops of variable length (12). Each monomer harbours a catalytic site, oriented towards the lateral side of the NA head, that is highly conserved over the different IAV subtypes ( Figure 1 ). Remarkably, unlike the tetramer, influenza NA monomers and dimers show very little if any enzymatic activity (24). It is not known why tetramer formation is essential for NA to be active. Possibly, this is linked to calcium binding by tetrameric NA, which contributes to NA activity and stability. NA can bind up to nine Ca2⁺ ions in the case of the 2009 H1N1 pandemic (H1N1pdm09) virus derived NA (16, 25–28).

Despite the significant primary sequence variation between IAV subtype NAs, the catalytic site residues in N1–N9 NAs are highly conserved. Among these, residues R118, D151, R152, R224, E276, R292, R371, and Y406 (N2 numbering) directly contact the substrate, while residues E119, R156, W178, S179, D198, I222, E227, H274, E277, N294, and E425 play a key role in stabilizing the catalytic site residues (19) ( Figure 1 ). Phylogenetic analysis indicates that IAV NAs fall into two distinct groups. In group 1 NAs, a cavity adjacent to the catalytic site is observed which is absent in group 2 NAs (29). This cavity is created by the so-called 150-loop that consists of residue 147-152 and is flexible in group 1 NAs such that it can adopt an open and a closed conformation. Group 2 NAs on the other hand lack this second cavity as the formation of a salt bridge between D147 and H150 stabilizes the 150-loop, which is absent in group 1 NAs due to the presence of a G147 there (29, 30). The catalytic site and its adjacent 150-cavity are further explored as targets for NA inhibitors (31). Additionally, most avian influenza NAs, but not NAs of human viruses, have a functional second sialic acid binding site (2SBS) or hemadsorption site next to the catalytic site (32). The 2SBS consists of three loops with residues that facilitate binding to sialic acid in its so-called chair conformation. The catalytic site, in contrast, binds sialic acid in its twisted boat conformation (16). Studies comparing human and avian NA catalytic properties show that the presence of a functional 2SBS in avian NA increases NA activity against multivalent substrates (33, 34).

HA and NA exert different functions in the influenza virus life cycle. HA is vital in the entry process, by mediating binding to sialic acids on host cell glycoproteins or -lipids, which results in virion uptake into endocytic vesicles, and the subsequent fusion of the host cell and virus membrane through a pH-induced conformational change (35). NA, on the other hand, catalyses the removal of the terminal sialic acids and thus functions as a receptor-destroying enzyme. NA activity is involved in multiple steps of the virus life cycle ( Figure 2 ). During viral entry NA cleaves decoy receptors present in the mucus that lines the epithelial cells of the respiratory tract, allowing the infection of underlying epithelial cells (36–38). In line herewith, inhibition of NA activity was shown to result in severely decreased infection of differentiated primary human airway epithelium cells (39). NA activity was also reported to stimulate HA-mediated membrane fusion (40). The best-known function of NA in the influenza virus replication cycle is its critical role in the release of newly formed virions from the infected cell and in prevention of HA-mediated virion aggregation by removing sialic acid from the viral and host cell membrane (41).

Given these partly opposing functions, a functional balance between HA and NA is critical for viral replication. If HA binding is much stronger than NA activity, for example, virions may be trapped by decoy receptors present in mucus and eventually be removed by mucociliary clearance before reaching the underlying epithelial cells. On the contrary, if NA activity dominates, the virion will quickly move through the mucus, but will more likely fail to stably bind to an epithelial cell and enter it. HA and NA can be found on the virion surface in a patch-wise distribution, which contributes together with the HA-NA balance to virus motility and entry. In filamentous influenza virions, NA is mainly localized at one pole of the filamentous virus, and the few NA molecules present along the side of the virus tend to cluster in patches as well (42). Such polarized viruses seem to move by Brownian ratchet-like diffusion in a mucus-rich extracellular environment, in which filamentous particles exhibit directed mobility away from their NA-rich pole and toward the HA crowded part of the virus (42). HA binding to sialic acid is reversible due to the low affinity. This reversible binding likely allows NA patches to catalyse the removal of the sialic acid when the HA is no longer bound thereby initiating movement of the virion through mucus and over the cell surface. This NA-driven virion motility presumably allows the virus to find and to dock to specific spots on the cell surface that trigger entry into the host cell (38, 43, 44). Although detailed mechanistic insights into the importance of the HA-NA balance are still lacking, it is clear that besides HA also NA contributes to IAV pathogenicity. For example, truncation of the NA stalk, which has been associated with adaptation of IAV from aquatic birds to poultry, resulted in increased virulence in poultry and mice (13, 45–49).

Several studies addressed the genetic evolution of HA and NA using IAV and IBV strains, which revealed that their evolution differs and is often asynchronous (74–78). Although the general topology of the NA and HA phylogenetic trees is similar with the typical ‘ladder-like’ gradual evolution and rapid replacement of old strains by newer ones, NA evolved slower and more gradually at the nucleotide level than HA. Looking at 40 years of evolution of human H3N2 viruses, Westgeest et al. showed that NA had fewer nucleotide substitutions over this time span compared with the HA head HA1, the most variable part of the HA gene (78). Nevertheless, it was concluded by Bhatt et al. (74) that with almost all adaptive evolution in NA being concentrated in residues on the surface of NA, the adaptation rate is higher for surface NA residues than for HA1. The observation that adaptive evolution in NA occurs almost exclusively in solvent-accessible surface residues indicates an important role for antibody-mediated immune responses in NA evolution. In Figure 3 , the IAV N1, IAV N2 and IBV NA structural conservation for the viruses that were used in seasonal influenza vaccines from 1970 to 2021 are depicted. Overall, a relatively high conservation of the head domain residues is observed for each NA (sub)type. As expected, the catalytic site is for all NA subtypes highly conserved. However, the surface residues do show variation, especially for IAV N2, which is in agreement with a higher rate of adaptation for NA in H3N2 than H1N1 viruses (74).

The analysis of NA antigenic drift usually relies on the determination of NAI titers of polyclonal ferret antisera raised against a particular influenza virus strain (e.g. a vaccine strain) in an enzyme-linked lectin assay (ELLA) (79, 80). The antigenic relatedness of the NAs from two viruses can be assessed by comparing NAI titers, calculating percent relatedness (81) or by performing antigenic cartography (76) to generate an antigenic map (80). On such a map, based on the quantification of the raw data of NAI assays, the antigenic distances between influenza viruses or sera are visualized. Importantly, the reactivity of antisera of ferrets that have experienced a single experimental virus infection is quite different from human sera which have an antibody repertoire that has typically been shaped by multiple prior infections and vaccinations (80, 82). The antigenic changes identified this way based on sera from ferret that were infected once with a particular influenza virus strain may therefore not always accurately reflect the response of an (adult) human individual (83) although such discrepancies have so far only been demonstrated for HA-specific responses (84–86).

Only a few studies have investigated the antigenic drift of NA using large NAI data sets and by constructing antigenic maps. In 2011, Sandbulte et al. reported on the characterization of the antigenic drift of NA in human H1N1 and H3N2 viruses that were recommended for influenza vaccine implementation over a period of 15 years in the USA (76). Similarly, Gao et al. examined the antigenic drift of NA sequences of H1N1pdm09 viruses (80). Overall, the antigenic differences between NAs of human H1N1 viruses occurred in one direction, meaning that antisera raised against older strains reacted weakly with more recent NAs whereas reactivity between antisera against more recent strains and older NAs was high (76, 80, 87). Both raw NAI data and the antigenic map obtained by Sandbulte et al. show that the NA of pre-2009 seasonal human H1N1 was in antigenic stasis for over a decade despite its genetic evolution during that time. For human H3N2 viruses the antigenic drift of NA was also not proportional to the number of amino acid changes induced. Collectively, these data show that there is discordance between the antigenic and genetic evolution of NA. Indeed, For N2 it was found that a single point mutation (E329K) was responsible for the abrupt NA antigenic drift between 2006 and 2007 (76). Similarly, for H1N1pdm09 few substitutions appeared to be largely responsible for the observed antigenic changes (80). The presumed important role in antigenic drift of K432E (80) was later confirmed using recombinant NA proteins that only differed at this position (87). Epistatic interactions as well as by biophysical constraints may also play a role in NA antigenic drift as was shown for the evolution of an antigenic site in H3N2 NA (88, 89). Furthermore, changes in HA receptor binding, e.g. as a result of antigenic drift, may in turn select for substitutions in NA that affect enzymatic activity to restore a functional balance in HA and NA, while additionally affecting NA antigenicity (90).

Bidirectional IAV transmission between humans and swine represents a serious public health challenge. In the last decade, an increasing number of zoonotic infections with IAV from swine has been reported. For example, after two distinct human-to-swine H3N2 spill overs in 1998 and 2002, N2-98 and N2-02 lineage viruses have circulated in swine in the USA and antigenically evolved over the next 20 years. After IAV infection, pigs typically produce broadly cross-reactive NAI antibodies to the N2 protein, but no NAI cross-reactivity between the N2-98 and N2-02 lineages was observed (91). The antigenic distance between swine and human N2 antigens increased over time to the extent that there is by now very little antigenic similarity with the human seasonal H3N2 NA that these viruses were derived from, nor with the NA of currently circulating human H3N2 viruses. This suggest that there will be little to no cross-reactive NA mediated immunity in both the swine and human population which may impact the occurrence of future spill overs (91).

Early antibody characterization and vaccination studies have shown that immunity to NA can be very broad within, but usually not across, subtypes (105). Nevertheless, mAbs that directly bind into the highly conserved catalytic site of NA (115) and inhibit enzymatic activity in small molecule assays are expected to exert a degree of heterosubtypic NAI. Recently, Stadlbauer et al. reported on a human mAb, 1G01, which binds to and inhibits the activity of several group 1 and group 2 NAs as well as NAs from both influenza B virus antigenic lineages in vitro. MAb 1G01 was shown to inhibit NA activity in a NA-Star assay and to occupy the catalytic site via a long CDR H3 loop by co-crystal structure analysis ( Figure 4 ) (105). A similar mechanism of action was described for mAb NA-45 where the CDR H3 loop adopts a protruding conformation with a tip that inserts into the NA catalytic site (98). This type of substrate mimicry is unique among all structurally characterized NA antibodies so far but has also been reported for anti-HA antibodies that target the receptor-binding site (116, 117). Interestingly, the number of residues in the antibody footprint might be important for the cross-reactivity of mAbs. For example, the binding footprint of 1G01 (105) and Z2B3 (106, 107) significantly overlaps. However, the 1G01 footprint includes more catalytic and framework residues which explains the broader cross-reactivity of 1G01 compared to Z2B3 (107). It is important to note that the footprint of mAbs that bind into the highly conserved catalytic site of NA often overlaps with the rim of the catalytic site. The rim is less conserved and can tolerate amino acid substitution without NA losing enzymatic activity.

Next to conventional mAbs, other antibody moieties that target the NA catalytic site have been described. The first NA-specific single domain antibodies or VHHs were described by Harmsen et al. with several candidate cross-NA binders and some VHHs that could affect NA activity (118). The isolation and characterization of a set of alpaca-derived H5N1 NA-specific VHHs (N1-VHH) with NA-inhibitory activity was also described. Two monovalent candidates N1-3-VHHm and N1-5-VHHm could inhibit NA activity with N1-3-VHHm also inhibiting oseltamivir resistant H5N1 virus NA. Bivalency of these constructs enhanced their NA inhibitory capacities and resulted in VHH constructs that could protect mice against H5N1 challenge (119).

A large proportion of the reported NAI mAbs only display NAI activity in the ELLA, but not in assays with small molecule substrates, indicating they do not contact the catalytic site directly (99). Binding of these mAbs likely results in steric hindrance and restricts access of large glycoconjugate substrates to the catalytic site (52). The NAI activity in ELLA assays has been associated with the capacity to block virus egress from infected cells (98, 99). An example of such a mAb is CD6, which effectively inhibited H1N1pdm09 viruses both in vivo and in vitro (103). Interestingly, CD6 more efficiently inhibited NA as an intact IgG compared to Fab and Fab₂ molecules of CD6, demonstrating that CD6 inhibits NA activity through steric hindrance and not structural distortion. Crystallization studies revealed a quaternary epitope that spans the lateral faces of two neighbouring N1 monomers ( Figure 4 ). Such a quaternary epitope present in intact NA dimers had been proposed earlier by Saito et al. for the N8-4 mAb but until then had never been structurally defined (114). For HA, quaternary epitope-specific mAbs that span two HA monomers that are efficient in neutralizing virus in vitro have also been described (120, 121). Some studies showed a correlation between the relative distance of epitopes to the NA catalytic site and the in vitro properties of mAbs. For example, mAb HF5 which has contact residues that surround the catalytic site pocket, showed more efficient NAI activity compared to other mAbs with an epitope that is located more laterally on the NA head (92, 107). Additionally, murine anti-N9 mAbs that inhibit NA enzymatic activity with a small molecule substrate in vitro, provided superior in vivo protection compared with mAbs that inhibited NA activity only in the ELLA assay (93).

It is becoming increasingly clear that Fc-mediated effector functions have an important role in providing in vivo protection against influenza virus challenge (122). The fragment crystallisable (Fc) region of an antibody can interact with Fc receptors and thereby mediate indirect effector functions such as antibody-dependent cellular cytotoxicity (ADCC), antibody-dependent cellular phagocytosis (ADCP) and complement-dependent cytotoxicity (CDC) (123). Using Fc receptor knockout mice and DA265 mutant mAbs that are unable to bind Fc receptors, it was demonstrated that broadly reactive HA-stalk antibodies depend on Fc-Fcγ receptor interactions in vivo (122, 124, 125). The contribution of Fcγ receptor engagement for in vivo protection by some NA-specific mAbs was demonstrated by Job et al. (126). The mouse monoclonal antibody N1-7D3 binds to a conserved linear epitope near the carboxy-terminus of N1 NA and shows no NAI activity ( Figure 4 ) (113). However, a recombinant mouse-human chimeric version comprised of the variable domains of mouse N1-7D3 and the constant regions of human IgG1 could engage activating Fcγ receptors and protected FcγR-humanized mice against challenge with H1N1pdm09 virus (126). In addition, mAbs with no detectable NAI activity reported by Yasuhara et al., could protect mice against a lethal influenza virus infection. Interestingly, a N297Q mutant version of these mAbs that lacks Fcγ receptor-binding activity failed to protect, thereby supporting the crucial role for Fcγ effector functions in protection by non-NAI mAbs (127). In line with this, grafting of the non-NAI N1-7-VHH on a mouse IgG2a Fc could protect mice against an otherwise lethal H5N1 challenge (119).

Several studies have demonstrated that the Fc-mediated effector functions might also be crucial for in vivo protection by some anti-NA mAbs that possess weak NAI activity (99, 127). Thus, while for potent N9 NAI mAbs Fc effector functions were not needed, the ability to interact with Fc receptors was required for weakly NAI mAbs to protect against challenge (99). Likewise, a broadly cross-reactive mAb with low NAI activity required FcγR interactions to mediate protection, while a strain-specific mAbs with high NAI did not (124). Likely, the sum of the NAI activity and the FcγR-mediated effector functions determines the potency of an anti-NA antibody in vivo. An anti-NA mAb with high NAI activity can reduce virus spread in vivo mainly through its NAI activity, whereas an anti-NA mAb without NAI activity can suppress virus replication mainly through FcγR-mediated effector cell activation (127).

Despite the recognized importance of NA-reactive antibodies in protection against influenza, the current commercially available influenza vaccines do not consistently elicit these antibodies (50, 128). NA immunogenicity of these vaccines varies widely between manufacturers (65, 128–130) and is poor compared to natural infection. Whereas following natural infection the number of NA-reactive B cells were equal to or exceeded HA-reactive B cells, current vaccines rarely induced NA-reactive B cells (50). Furthermore, broadly cross-reactive NAI antibodies elicited by natural infection were unable to bind multiple commercially available inactivated vaccines, indicating that the vaccines lack the NA epitopes targeted by these antibodies (50).

The poor NA immunogenicity of the licensed influenza vaccines can be attributed to a few factors. As mentioned, current vaccines are optimized and standardized specifically for inducing high HAI antibody titers. As a result, the immunogenicity of the NA is not guaranteed. Quantity and quality of the NA antigen varies between manufacturers and vaccine batches (131, 132). Virus inactivation procedures may also affect NA immunogenicity. While treatment with EDTA or formalin did not compromise the immunogenicity of recombinant NA proteins (133), native NA in a viral particle may respond differently to these conditions. The stability and immunogenicity of NA also varies between strains (132). Nevertheless, in case vaccination does elicit NA antibodies, these can be long lived in healthy human subjects (134, 135). Better consideration of the NA component of current vaccines could therefore mean an important step forward in effectivity of the current vaccines. This could be achieved by also considering NA in selection of the vaccine strains, optimizing the manufacturing process to keep NA antigenically intact, and also standardizing the amounts of immunogenic NA in vaccine preparations.

Next-generation seasonal and universal influenza vaccines call for an improved NA-directed immune response. Various options for rational design of the NA antigen and mode of presentation to the immune system are in development aiming to boost the breadth and magnitude of the NA-specific response ( Figure 5 ). The following section summarizes promising ideas and the rationale behind them.