Binding properties of DA03E17 to antiviral-resistant variants and diverse influenza NAs

We previously isolated and characterized a human mAb DA03E17 from peripheral blood mononuclear cells (PBMCs) of an individual who was infected with A/H1N1pdm09 virus in the 2015–2016 influenza season³⁸. DA03E17 showed broad cross-reactivity against NAs from IAV group 1 (N1, N4, N5, N8), group 2 (N2, N3, N6, N7, N9), as well as both lineages of IBV (B/Yamagata and B/Victoria) (Supplementary Fig. 1a)³⁸. Furthermore, DA03E17 inhibited the sialidase activity of NA, neutralized both IAVs and IBVs in vitro, and provided in vivo protection against several subtypes of influenza virus³⁸. DA03E17 uses the IGHV4-31 and IGKV1-12 heavy and light chain V genes, respectively, and has a long CDR H3 (19 amino acids in the IMGT CDR definition scheme) (Supplementary Fig. 2). In addition to the previously reported broad reactivity of DA03E17, here we further characterized its ability to retain binding to NAs containing oseltamivir-resistant mutations. We generated recombinant N1 NA from H1N1 A/Brisbane/02/2018 (BB18), which contains previously reported stabilizing mutations derived from a computationally-designed NA (stabilized NA protein, sNAp) in the inter-protomeric interface to maintain a closed tetrameric state⁴¹. We produced BB18 N1 sNAp containing either the major oseltamivir-resistance mutation H274Y¹⁷ or other resistance substitutions, I222V or S246N (N2 numbering)¹⁸. In addition, we generated recombinant N2 NA from H3N2 A/Indiana/10/2011 (IN11) with either the E119V or I222L substitution⁴². We measured the binding of DA03E17 to these recombinant N1 and N2 NAs by ELISA. While DA03E17 retained its binding to BB18 N1 sNAp with either the H274Y, I222V, or S246N substitution, it exhibited reduced binding to IN11 N2 with either the E119V or I222L substitution, although some level of binding was still observed (Supplementary Fig. 1b). Compared to DA03E17, the previously described broadly neutralizing anti-NA mAb 1G01³⁶ showed a greater difference in binding to oseltamivir-resistant BB18 N1 NAs and was particularly affected by the H274Y substitution, consistent with a previous report³⁹, while its binding to oseltamivir-resistant IN11 N2 NAs remained relatively consistent (Supplementary Fig. 1b). We also assessed the NI activity of DA03E17 using ELLA with the same recombinant NA proteins containing oseltamivir-resistant mutations. DA03E17 retained inhibitory activity against BB18 N1 NAs carrying these mutations, with enhanced inhibition observed for H274Y, while activity was reduced against S246N and especially I222V variants. In contrast, 1G01 showed reduced inhibition against S246N and H274Y variants, consistent with its ELISA binding profile. Oseltamivir-resistant IN11 N2 mutants were excluded from ELLA due to a lack of detectable sialidase activity, but DA03E17 exhibited similar levels of inhibition to 1G01 against the wild-type IN11 N2 (Supplementary Fig. 1c). These results demonstrate that DA03E17 retains binding and inhibitory activity against several oseltamivir-resistant NA variants, while also revealing mutation-specific differences in sensitivity.

Biolayer interferometry (BLI) indicated that DA03E17 IgG bound to recombinant NAs from H1N1 A/California/07/2009 (CA09 N1 sNAp), H3N2 A/Perth/16/2009 (PT09 N2), and H3N2 A/Indiana/08/2011 (IN11 N2) with sub-picomolar apparent affinities, while it bound to NAs from B/Colorado/06/2017 (B/Victoria-lineage; CO17 B) and more recent H3N2 strains, including A/Kansas/14/2017 (KS17 N2) and A/Hong Kong/2671/2019 (HK19 N2), with sub-nanomolar to weaker nanomolar apparent affinities (Supplementary Fig. 3). Notably, DA03E17 IgG also bound to recombinant N1 sNAp derived from the HPAI H5N1 clade 2.3.4.4b virus (A/dairy cattle/Texas/24-008749-001/2024; TX24) with nanomolar apparent affinity, which is currently spreading across dairy herds and other mammals in 14 states in the United States^9,10,11 (Supplementary Fig. 3). DA03E17 Fab exhibited similar subtype-dependent binding patterns, with nanomolar to weaker nanomolar affinities for CA09 N1 sNAp, TX24 N1 sNAp, PT09 N2, and IN11 N2, and sub-micromolar affinities for KS17 N2, HK19 N2, and CO17 B. The binding of DA03E17 Fab to CO17 B NA was markedly reduced compared to IgG, indicating a greater contribution of avidity for this lineage (Supplementary Fig. 3). Together, these results confirm the broad cross-reactivity of DA03E17 to diverse influenza A and B virus NAs.

Cryo-EM structures of DA03E17 Fab in complex with N1, N2, and B NAs

To elucidate the epitope of DA03E17 and the structural basis for its broad cross-reactivity, we determined the cryo-EM structures of the DA03E17 Fab in complex with CA09 N1 sNAp, and KS17 N2 and CO17 B NAs at 2.67, 2.86, and 2.47 Å resolution, respectively (Supplementary Fig. 4 and Supplementary Table 1). In addition, a cryo-EM structure of KS17 N2 NA was determined in its apo-form at 2.75 Å resolution using the same data set of DA03E17-KS17 N2 NA complex (Supplementary Fig. 4). In all three complex structures, DA03E17 binds in a similar orientation to all three NAs, with each Fab interacting with just one protomer of the NA tetramer (Fig. 1a–c and Supplementary Fig. 5a–f). DA03E17 fully blocks the NA active site by protruding the CDR H3 into the active site pocket (Fig. 1c and Supplementary Fig. 5c, f), consistent with our previous study using the small molecule-based NA-Star assay, which demonstrated that the NA inhibition (NI) activity of DA03E17 is due to direct inhibition³⁸. No large conformational changes in the global structure of the NA protein are observed compared with corresponding wild-type structures, as indicated by an RMSD of 0.356, 0.402, and 0.314 Å across all pairs for the N1, N2, and B NAs, respectively (Supplementary Fig. 5g–i).

a, b Cryo-EM map at 2.67 Å resolution (a) and two orthogonal views of the atomic model from the top and side (b); only the Fab variable region is built into the density. DA03E17 heavy chain, dark blue; light chain, light blue; NA head domain tetramer; shades of gray; NA glycans, gold; Ca²⁺ ion, lime. c Ribbon diagram of the CA09 N1 sNAp protomer bound to one DA03E17 Fab. The NA active site targeted by CDR H3 is highlighted by a green oval. VH, heavy chain variable domain; Vk, kappa light chain variable domain. d, e DA03E17 epitope mapped on the NA protomer surface with CDR loops involved (d) or residue labels (e). The DA03E17 N1 epitope is highlighted in green. f–h, Detailed interactions between CA09 N1 sNAp and DA03E17 CDR H3 (f), CDR H1 and H2 (g), and CDR L1, L3, and FR L3 (h).

In the DA03E17-CA09 N1 sNAp complex structure, the buried surface area (BSA) of the DA03E17 and N1 NA interface is 1022.2 Ų, with the heavy chain accounting for 78% of the interaction. DA03E17 interacts with CA09 N1 sNAp using all CDR loops except CDR L2 (Fig. 1d), and the DA03E17 epitope on CA09 N1 sNAp consists of 32 residues around the active site pocket (Fig. 1e). The 19-residue CDR H3 contributes most of the Fab interactions with 19 N1 NA residues including seven catalytic site residues (R118, D151, R152, R224, R292, R371, and Y406) and six framework residues (E119, W178, I222, E227, E277, and N294) (Fig. 1f) which are strictly conserved across influenza A group 1, group 2, and influenza B NAs (Supplementary Table 2). Substitutions at D151 (D151G and D151N), which were previously identified as major escape mutations for DA03E17 in our earlier study³⁸, can be explained by the direct contact between the CDR H3 of DA03E17 and D151 observed in our structure (Fig. 1f). In contrast, although a substitution at T439 (T439A) was also identified as an escape mutation, no direct contact between DA03E17 and T439 was observed in our structures. This substitution may affect binding indirectly by altering the conformation of the NA active site. The CDRs H1, H2, L1, L3, and framework region (FR) L3 predominantly interact with residues located at the periphery of the active site (I149, K150, D198, N199, N247, W295, H296, S342, N344, G345, A346, N347, R430, P431, and K432), further strengthening binding (Fig. 1g, h). Overall, DA03E17 utilizes both heavy and light chain CDRs to bind the active site of NA, and a total of 24 hydrogen bonds were observed between DA03E17 and CA09 N1 sNAp. In the DA03E17-KS17 N2 and DA03E17-CO17 B NA complex structures, the BSAs at the interfaces are 1224.5 Ų and 977.6 Ų, respectively, indicating a slightly larger interaction surface with the KS17 N2 NA compared to the CA09 N1 and CO17 B NAs. Most of the interactions mediated by the CDR H3 in the DA03E17-KS17 N2 and DA03E17-CO17 B NA complexes are similar to those observed in the CA09 N1 sNAp complex (Supplementary Fig. 6a, e), with key epitope residues being conserved across IAV and IBV NAs (Supplementary Table 2), maintaining a consistent binding mode. These data establish that DA03E17 directly targets the highly conserved residues in the enzymatic active site of NA using its long CDR H3, thereby blocking the active site and inhibiting sialidase activity of NA.

DA03E17 targets highly conserved epitopes in N1, N2, and B NAs

To further understand the broad cross-reactivity of DA03E17, we examined the sequence conservation of DA03E17 epitopes on NAs from H1N1, H3N2, and IBV viruses that have been circulating for several decades. (Fig. 2a–c). The epitopes of DA03E17 are well conserved among human seasonal H1N1 (circulating from 1977 to 2023) and H3N2 (from 1968 to 2023) IAVs, as well as B/Victoria/2/87-like IBVs (from 1987 to 2023). The majority of epitope residues are highly conserved or conservatively substituted on H1N1 (25 out of 32, 78%), H3N2 (25 out of 35, 71%), and B/Vic (25 out of 30, 83%) NAs (Fig. 2d). Notably, although the B/Yamagata-like lineage has been considered extinct since 2020^2,3, the DA03E17 B/Vic NA epitope is also highly conserved in NAs of B/Yamagata/16/88-like IBVs (from 1988 to 2020; 25 out of 30, 83%). We also assessed epitope conservation in animal-origin IAVs, analyzing N1 and N2 NA sequences from HxN1 and HxN2 viruses that have circulated in avian and mammalian hosts between 1977 and 2023 (Supplementary Fig. 7). DA03E17 epitope residues on N1 were generally well conserved in avian and mammalian HxN1 NAs, similar to human seasonal H1N1 viruses. In contrast, conservation in HxN2 NAs from animal-origin IAVs was lower compared to human H3N2 NAs. Substitutions at positions 199 and 221—the two most variable residues within the N1 and N2 epitopes—from N199 (CA09 N1) to S199 (H1N1 A/Yokohama/94/2014) and N221 (CA09 N1) to K221 (H1N1 A/Puerto Rico/8/1934) in N1, as well as from K199/D221 (KS17 N2) to E199/K221 (H3N2 A/Fujian/411/2002; FJ02) in N2, did not compromise the NI activity of DA03E17, as shown in our previous study³⁸. Similarly, substitutions in the N2 epitope, from N147/R150/K344 (KS17 N2) to D147/H150/E344 (FJ02 N2), did not impair the NI activity of DA03E17, and substitutions in the influenza B NA epitope, from S244/R295/K343/E436 (CO17; B/Yamagata-lineage) to P244/S295/E343/T436 (B/Phuket/3073/2013; B/Victoria-lineage), had almost no effect on its NI activity³⁸. These results suggest that DA03E17 binding is resilient to these substitutions. The core of the DA03E17 epitope in the active site, which is targeted by the CDR H3, is strictly conserved across A/H1N1, A/H3N2, and B/Victoria-like viruses due to the functional constraints related to sialic acid receptor recognition (Fig. 2d, e). Therefore, the broad cross-reactivity of DA03E17 can be partially explained by the strict conservation or conservative substitution of key epitope residues, along with its resilience to substitutions around the NA active site. Notably, several conserved residues in the active site pocket are charged amino acids, forming distinct charged patches with R118, R292, and R371 contributing to positive charges, and E119, D151, E227, and E277 forming negative charges (Fig. 2e, f). D100b and R100c, which form the DR (Asp–Arg) motif at the tip of the DA03E17 CDR H3, have an electrostatic surface that is complementary to these charged patches, enabling extensive salt bridge formation for targeted interaction (Figs. 1f, 2g, and Supplementary Fig. 6a, e). Overall, we found that the DA03E17 epitope is largely conserved, especially the active site residues targeted by the CDR H3, which are strictly conserved across the NAs of H1N1, H3N2, and B/Victoria-like viruses that have been circulating for several decades, highlighting the critical role of the CDR H3 in the broad cross-reactivity of DA03E17.

a–c Sequence conservation of the DA03E17 epitope mapped onto surface representations of CA09 N1 (a), KS17 N2 (b), and CO17 B (c) NAs. NA residues conserved across A/H1N1, A/H3N2, and B/Victoria-like viruses are colored in dark blue, while residues conserved or conservatively substituted within each subtype or lineage are shown in sky blue. Semi-conserved or variable residues are colored in green. d Conservation of DA03E17 epitope residues based on NA sequences from human seasonal H1N1 (1977–2023) and H3N2 (1968–2023) IAVs, B/Victoria-like (1987–2023), and B/Yamagata-like (1988–2020) IBVs. Symbols mark NA residues forming hydrogen bonds with DA03E17: circled bullets for main chain, open circles with rays for side chain, and circled bullets with rays for both. e Superimposed protomers of CA09 N1 sNAp, and KS17 N2 and CO17 B NAs with conserved active site residues shown as sticks and key charged residues highlighted. f, g Surface representations of CA09 N1, KS17 N2, and CO17 B NA protomers (f) and DA03E17 Fv (g) colored by electrostatic potential from − 10 to + 10 k_BT/e_c. Dotted circles indicate charged patches in the NA active site pocket.

DA03E17 accommodates the N-glycans of the recently circulating human H3N2 viruses

The NA of circulating human A/H3N2 viruses has undergone substantial antigenic drift since 2014, resulting from mutations at positions 245 and 247 that introduced the N245 glycan near the active site⁶ (Fig. 3a and Supplementary Fig. 8a). By the 2016/17 season, nearly all circulating H3N2 viruses carried the N245 glycan, which has become fixed in recent strains and has been shown to shield the NA active site, reducing the efficacy of certain NA active site-targeting antibodies^6,7,8,39. Another N-linked glycan near the NA active site, the N146 glycan (Fig. 3a), is highly conserved in human and animal IAVs, including human H1N1 viruses circulating since 1918 and human H3N2 viruses circulating since 1968⁴³, and a corresponding N144 glycan is also conserved in IBVs. Our previous study showed that DA03E17 maintained inhibitory activity against the NAs of recently circulating H3N2 viruses, despite the presence of the N245 glycan³⁸. BLI analysis revealed that while DA03E17 Fab exhibited reduced affinities for drifted N2 NAs (KS17 and HK19) carrying the N245 glycan, as well as mutant IN11 N2 NA (IN11 N2 NAT) with an introduced N245 glycosylation site, compared to NAs without the N245 glycan, it still maintained sub-micromolar affinities (Fig. 3b and Supplementary Fig. 3). These results indicate that the N245 glycan may affect the binding of DA03E17 to the NA active site. To investigate in detail how the N245 glycan affects the binding of DA03E17, we compared the structure of the DA03E17-KS17 N2 NA complex with that of the apo-KS17 N2 NA, which was determined from the same data set of the DA03E17-KS17 N2 NA complex (Supplementary Fig. 4). In the apo-KS17 N2 structure, the N245 glycan would cause a steric clash with DA03E17 CDR H1, while the N146 glycan would clash with CDR L1. However, upon binding to KS17 N2 NA, DA03E17 induces conformational changes in both the N245 and N146 glycans, allowing it to maintain binding by repositioning these glycans and avoiding the clashes (Fig. 3c and Supplementary Fig. 8b).

a Cryo-EM structure of KS17 N2 NA in apo-form at 2.75 Å resolution. The inset shows a magnified view of a KS17 N2 NA protomer, with the active site indicated by a green circle, and the surrounding glycans are shown in yellow. b Binding affinity of DA03E17 Fab to recombinant NAs, measured by BLI with CA09 N1, TX24 N1, PT09 N2, IN11 N2, IN11 N2 NAT mutant, KS17 N2, and HK19 N2 NAs. NAs with an N-glycan at position 245 are highlighted in red. K_D was estimated using a 1:1 binding model. Source data are provided as a Source Data file. c–e Overlay of the apo-KS17 N2 NA protomer (gray) with Fab-bound KS17 N2 NA protomers (white): DA03E17 Fab (c), 1G01 Fab (d), and FNI9 Fab (e). The N146 and N245 glycans on the apo-KS17 N2 NA are shown in yellow, and those on the Fab-bound KS17 N2 NA are shown in green. Steric clash shown with red stars. Conformational changes in glycans and the 242–252 loop are indicated by arrows.

To further investigate the structural changes in the N245 and N146 glycans upon DA03E17 binding to KS17 N2 NA, we performed molecular dynamics (MD) simulations. In the apo-KS17 N2 NA structure, the MD simulations revealed that the N245 and N146 glycans formed clusters that obstructed access to the NA active site and potentially blocked the binding of DA03E17 (Supplementary Fig. 8c, d). In contrast, MD simulations of the DA03E17-KS17 N2 NA complex demonstrated that DA03E17 binding induced significant structural changes in both the N245 and N146 glycans. The glycans shifted to adopt specific conformations, moving away from the DA03E17 binding interface. These movements aligned with the cryo-EM structural data (Fig. 3a, c), further supporting that DA03E17 induces a rearrangement of the N245 and N146 glycans to facilitate stable binding to the NA active site. Such conformational changes in the glycans may require additional energy, potentially contributing to the reduced affinity of DA03E17 for N2 NAs carrying the N245 glycan. Considering that the N146 glycan is also present in other N1 and N2 NAs, such as CA09 N1, PT09 N2, and IN11 N2, where DA03E17 Fab binds with a range of nanomolar affinity (Supplementary Fig. 3), the N245 glycan, either alone or in the context of the N146 glycan, plays a prominent role in the observed reduction in apparent affinity for recent N2s.

Previous studies have also explored how NA active site-targeting antibodies maintain binding and protective efficacy against contemporary H3N2 viruses carrying the N245 glycan, offering insights into the mechanisms these antibodies employ to accommodate this glycan^8,39. The broadly cross-reactive antibody, 1G01, was shown to maintain protective efficacy in vivo despite reduced inhibition against viruses harboring the N245 glycan⁸. Although structural analysis using negative-stain electron microscopy indicated that 1G01 still binds the active site of N2 NA with the N245 glycan, the precise mechanism remained unclear. Another broadly cross-reactive antibody, FNI9, was shown to not only induce conformational changes in the N245 glycan but also alter the conformation of the 242–252 loop of HK19 N2 NA, which contains the N₂₄₅AT₂₄₇ glycosylation motif³⁹. To compare the N245 glycan accommodation mechanisms of 1G01 and FNI9 with that of DA03E17 directly, we determined the cryo-EM structures of the 1G01-KS17 N2 NA and FNI9-KS17 N2 NA complexes at 3.20 Å and 2.97 Å resolutions, respectively (Supplementary Fig. 9). Structural comparison revealed that CDR H3 of 1G01 would clash with the N245 glycan in the apo-KS17 N2 NA structure. However, similar to DA03E17, 1G01 induces conformational changes in the N245 glycan, shifting it away from the binding interface and maintaining binding to the NA (Fig. 3d). For FNI9, similar to previous findings with HK19 N2 NA³⁹, structural comparison revealed that it induces conformational changes not only in the N245 glycan but also in the 242–252 loop of KS17 N2 NA (Fig. 3e). This rearrangement allows FNI9 to maintain binding, suggesting that its glycan accommodation mechanism is not HK19 N2 NA-specific but rather a broader adaptation to drifted N2 NAs carrying the N245 glycan in general. To summarize, the three broadly cross-reactive antibodies, DA03E17, 1G01, and FNI9, accommodate the N245 glycan on drifted N2 NA through two primary mechanisms. DA03E17 and 1G01 induce conformational changes in the glycan, shifting it away from the binding interface, while FNI9 not only induces glycan rearrangement but also alters the conformation of the 242–252 loop, indicating a more extensive structural rearrangement. These results demonstrate two distinct strategies employed by NA active site-targeting antibodies to maintain effective active site binding.

DA03E17 mimics the interaction of sialic acid using a conserved motif in the CDR H3

The DA03E17 footprint on the NA spans the entire active site pocket, fully covering it and blocking access to critical residues involved in enzymatic activity (Fig. 1e), suggesting that DA03E17 may inhibit NA activity through a mechanism similar to known NA inhibitors, which block the active site by mimicking sialic acid interactions^44,45. We compared the interaction formed between the DA03E17 CDR H3 and the NA active site residues with that of the sialic acid receptor and oseltamivir. Remarkably, the carboxylate side chain of D100b in the DA03E17 CDR H3 forms the same salt bridge interaction network with NA residues R118, R292, and R371 as the carboxylate group of sialic acid and oseltamivir (Fig. 4a–c). The side chain of R100c in the CDR H3 also forms contacts with NA residues D151 and E227. While sialic acid and oseltamivir also contact these residues, R100c in the CDRH3 contacts through different interactions, forming a salt bridge that strengthens binding. Such receptor mimicry by antibody CDR H3 has been previously observed in other NA active site-targeting antibodies^37,39,46. In our structural comparison, we found that the DR (Asp–Arg) motif in DA03E17 functions similarly to those in the previously reported pan-influenza NA mAb FNI9³⁹ and the broadly cross-reactive anti-IBV NA mAb 1G05³⁷, mimicking sialic acid interactions and blocking the NA active site (Fig. 4d, e). Notably, the DR motif is reversed in order in FNI9, appearing as R–D instead of D–R, as seen in DA03E17 CDR H3 (Fig. 4i), yet both motifs mimic sialic acid interactions from nearly identical positions in the active site. In addition, we identified the same sialic acid-mimicking DR motif in the previously reported cross-group mAb Z2B3⁴⁷ (Fig. 4f), which binds to both N1 and N9 NAs. These findings highlight the conserved role of DR motifs in mediating receptor mimicry across multiple NA active site-targeting antibodies, supporting the idea of sequence and structural convergence in antibody responses to NA. Similarly, sialic acid mimicry has been previously reported in the N9 NA-specific mAb NA-45⁴⁶, where the negatively charged side chain of E (Glu), similar to D (Asp), is used to mimic the carboxylate group of sialic acid (Fig. 4g). This type of sialic acid mimicry, using D (Asp) and E (Glu) to mimic the carboxylate group of sialic acid, has also been observed in several HA receptor-binding site (RBS)-targeting antibodies^{48,49,50,51,52,53}, highlighting the critical role of negatively charged amino acids in mimicking the carboxylate group of sialic acid within both the HA RBS and the NA active site. We also compared the structure of another pan-influenza NA mAb, 1G01³⁶, which has an R (Arg) at the tip of its CDR H3, but its interactions were distinct from those mediated by the DR motif (Fig. 4h). In summary, these NA active site-targeting antibodies possess relatively long CDR H3 loops (Fig. 4i), and while the Fabs have varying angles of approach and the CDR H3 loops engage the active site pocket in diverse ways (Fig. 4j, k), the DR motifs remain structurally conserved among antibodies with receptor mimicry.

a Network of salt bridge interactions between D100b and R100c in the CDR H3 of DA03E17 and active site residues of CA09 N1 sNAp. b, c Salt bridge interactions between sialic acid (b) and oseltamivir (c) with active site residues of H3N2 A/Tanzania/205/2010 (TZ10) NA (PDB: 4GZQ, 4GZP). d Salt bridge interactions between D107 and R106 in the CDR H3 of FNI9 and TZ10 N2 NA (PDB: 8G3N). e Salt bridge interactions between D100a and R100b in the CDR H3 of 1G05 and B/Phuket/3073/2013 NA (PDB: 6V4N). f Salt bridge interactions between D107 and R108 in the CDR H3 of Z2B3 and H7N9 A/Anhui/1/2013 NA (PDB: 6LXJ). g Salt bridge interactions between E100a in the CDR H3 of NA45 and H7N9 A/Shanghai/2/2013 NA (PDB: 6PZE). h Hydrogen bonds and salt bridge interactions between R100c in the CDR H3 of 1G01 and CA09 N1 NA (PDB: 6Q23). i Alignment of the CDR H3 sequences of DA03E17, FNI9, 1G05, Z2B3, NA45, and 1G01, with DR motifs mimicking the interaction of sialic acid highlighted in red. j, k Overlay of low-pass filtered structures of NA active site-targeting antibodies bound to the tetrameric NA head domain (j), and a detailed view of their CDR H3 loops occupying the active site pocket of the NA protomer (k).

DR motif precursors are prevalent in human antibody repertoire

To assess the feasibility of eliciting NA active site-targeting antibodies that mimic sialic acid binding through vaccination, we employed a bioinformatic approach to identify the prevalence of antibodies with long CDR H3s containing DR motifs. Notably, in all three antibodies that had the DR motif and for which a D gene reading frame and position could be inferred (DA03E17, 1G05, and Z2B3), the DR motif was encoded at least partially by non-templated junction residues between the D and J genes (Fig. 5a, b), suggesting the DR motif could be limiting the precursor frequency. We employed two distinct search strategies using an ultradeep next-generation sequencing (NGS) dataset of 1.1 × 10⁹ antibody heavy chain sequences from 14 healthy human donors^54,55. First, we identified long CDR H3 loops (19–30 amino acids) with DR motif (DR or RD) positioned near the middle of the loop (flanked by at least 8 residues on either side) (Fig. 5c and Supplementary Fig. 10). CDR H3 loops bearing the DR motif were identified in all 14 donors, with median frequencies of 875 (DR) and 1348 (RD) per million sequences (Fig. 5d). No single D gene was dominant among these CDR H3 sequences. Given the potential for additional requirements, only a subset of precursors meeting this bioinformatic definition would be expected to have the potential to develop into NA active site-targeting antibodies. Therefore, this search represents an upper limit on the frequency of potential DR motif precursors that may be accessible to targeting by vaccination. The second search incorporated criteria that was specific for each of the three NA active site-targeting antibodies for which the D gene could be inferred (DA03E17, 1G05, and Z2B3). We defined precursors as having CDR H3 loops of equal length or longer than the template antibody (up to 30 amino acids) with the D gene in the same reading frame and flanked by at least as many amino acids as occurs in the template antibody (definition shown in Fig. 5e and in the “methods” section). Precursors for each antibody class were identified in all 14 donors at median frequencies of 1948, 228 and 6.9 per million for DA03E17, 1G05 and Z2B3, respectively (Fig. 5f). After establishing that all donors made precursors with these CDR H3 features, we then added a requirement for the DR motif to be present at the equivalent position as the template antibody relative to the D gene (Fig. 5e). Inclusion of the DR motif resulted in a large decrease in precursor frequencies for DA03E17 and Z2B3. DA03E17 precursors remained detectable in all 14 donors but at a 985-fold lower median frequency (~2 per million), whereas Z2B3 precursors were identified in only 2 of 14 donors, at a median frequency of ≤ 0.01 per million. In contrast, 1G05 precursors containing the DR motif occurred at a median frequency of 14 per million, representing only a 16-fold reduction likely due to D gene templating of the aspartic acid residue (Fig. 5f). These results suggest that targeting precursors with DA03E17- or 1G05-like CDR H3 loops with DR motifs in the naïve sequence may be feasible, as their precursor frequencies are comparable to other vaccine targets that have shown promising results^56,57,58. Elicitation of Z2B3-like responses may require initial targeting of sequences lacking the DR motif and acquiring the DR motif through somatic hypermutation (SHM).

a Genetic characteristics of DR motif antibodies targeting the NA active site through receptor mimicry. b The germline and amino acid sequences of DA03E17 CDR H3 are aligned, with non-templated junction residues in red and residues that have undergone SHM in light blue. Nucleotides removed by exonuclease trimming indicated with a line through the letters. Circled bullet points denote antibody main chain contacts only; open circles with rays denote antibody side chain contacts only; Circled bullet points with rays denote both main and side chain contacts. c Search criteria used to identify the minimal NA active site-targeting CDR H3 motifs in a NGS dataset of 1.1 billion human antibody heavy chain sequences. d Frequency of precursor BCRs with long CDR H3s containing DR (left) or RD (right) motifs near the middle of the CDR H3, displayed by total and the five D genes with the highest frequencies. Points represent the frequency observed in each of the 14 human donors. e Search criteria used to identify DA03E17-, 1G05- and Z2B3-like CDR H3s in the NGS dataset. f Frequency of precursors defined in (e) for three NA active site-targeting antibodies (DA03E17, 1G05, Z2B3). Points represent the frequency observed in each of the 14 human donors. Bars represent the median frequency across donors. Source data are provided as a Source Data file.

Additional DR motif antibodies converge on the NA active site through receptor mimicry

To explore whether similar DR motif NA antibodies have been independently observed and to get a broader qualitative sense of their prevalence, we surveyed antibody sequences from patents and the literature for NA antibodies with a central DR motif in the CDR H3, thereby complementing our repertoire analysis. Using the Patent and Literature Antibody Database (PLAbDab)⁵⁹, we searched for antibody sequences using “neuraminidase” as a keyword and identified 168 paired antibody sequences derived from 24 different sources. Through manual curation, we identified two antibodies (CR12042 and CR12044) with a centrally positioned DR motif in the CDR H3, both of which were reported in a single patent⁶⁰ (Fig. 6a). In addition, we identified three more antibodies (AF9C, Z1A11, and Z2C2) from the study that also reported Z2B3⁶¹, one of the DR motif antibodies included in our structural comparison (Fig. 4f). While CR12042 and CR12044 are described as human-derived antibodies, no details were provided regarding their source or method of isolation. CR12042 and CR12044 bound to multiple N1 NAs and showed weak binding to an N2 NA. Since they exhibited NI activity in a small molecule-based assay⁶⁰, we considered it likely that they target the NA active site. AF9C was isolated from an adult who received the 2014/15 Northern Hemisphere trivalent influenza vaccine (TIV), whereas Z1A11 and Z2C2 were isolated from a child who experienced a mild H7N9 infection in 2013, the same individual from whom Z2B3 had previously been isolated⁶¹. AF9C was N1-specific, while Z1A11 and Z2C2 were cross-reactive to N1 and N9 NAs, similar to Z2B3. Their NI activity was observed in ELLA, but was not assessed in a small molecule-based assay, leaving it unclear whether they directly target the NA active site. All five antibodies possessed a centrally located DR motif in the CDR H3 (Fig. 6a), and except for CR12042, which has an 18-residue CDR H3, they met the search criteria used for our precursor frequency analysis (Fig. 5c), which required a CDR H3 of 19–30 residues flanked by at least eight residues on either side of the DR motif. These five antibodies were expressed as IgGs, with AF9C, Z1A11, and Z2C2 produced in germline-reverted form (GL), as only gene usage and the CDR H3 and CDR L3 sequences were available. To evaluate their binding activity, we measured binding to CA09 N1 sNAp by BLI. Four of the five antibodies showed binding, with Z2C2-GL as the exception, possibly due to reduced affinity in its germline-reverted form lacking somatic mutations that may be required for stable interaction (Fig. 6b). To investigate whether the DR motifs in these antibodies mediate receptor mimicry in a structurally conserved manner, we determined cryo-EM structures of their complexes with CA09 N1 sNAp at resolutions ranging from 2.48 to 2.76 Å (Fig. 6c and Supplementary Fig. 11). All four antibodies targeted and blocked the NA active site, despite exhibiting distinct angles of approach (Fig. 6c). Remarkably, all DR motifs in these four antibodies were structurally conserved and engaged active site residues through receptor mimicry in a manner similar to the DR motif antibodies described earlier in this study (DA03E17, FNI9, 1G05, Z2B3) (Figs. 4 and 6d). These results further support the idea that receptor mimicry mediated by conserved DR motifs is a recurring strategy among NA active site-targeting antibodies.

a Genetic characteristics of additional DR motif antibodies. b Sensorgrams from BLI experiments showing binding of IgGs of CR12042, CR12044, and of germline-reverted (GL) AF9C, Z1A11, and Z2C2 to CA09 N1 sNAp immobilized on HIS1K biosensors. DA03E17 was included as a positive control. Source data are provided as a Source Data file. c Cryo-EM maps of additional DR motif antibodies in complex with CA09 N1 sNAp, with overall resolutions ranging from 2.48 to 2.76 Å. d Detailed interactions between the conserved DR motifs in the CDR H3 of additional DR motif antibodies and the NA active site residues. e Structural alignment of 8 DR motif antibodies by their bound NA protomer. The variable regions of the heavy and light chains (VH and VL) are colored in dark blue and light blue, respectively. f Cross-sectional view of the NA active site showing conserved DR motifs occupying the active site pocket.

We next investigated whether these eight DR motif antibodies share additional structural similarities beyond the DR motif. Although these antibodies exhibit diverse binding orientations to the NA active site (Fig. 6e), Z2B3 and AF9C-GL engage the active site with similar angles of approach (Supplementary Fig. 12). Both antibodies are derived from the IGHV1-69*01 germline gene, but their binding angles are slightly offset, and no notable structural similarities were observed at the level of detailed interactions, with their light chains also originating from different germline genes (Z2B3 from IGLV2-14*01 and AF9C from IGKV1-9*01)⁶¹. FNI9, CR12042, and CR12044 were also derived from, or predicted to originate from, the IGHV1-69 germline gene (Figs. 5a and 6a). Although FNI9 and CR12042 approached the NA active site with similar Fv angles, the orientations of their heavy and light chains were reversed (Supplementary Fig. 12). CR12044 also bound with a distinct orientation. Thus, aside from the relatively frequent use of IGHV1-69, no major structural similarities were observed among these antibodies. Nevertheless, one notable commonality was the relatively frequent usage of the IGHJ6 gene. DA03E17, 1G05, Z2B3, CR12042, AF9C, Z1A11, and even Z2C2, for which structural determination was not achieved due to loss of binding, all utilized IGHJ6 (Figs. 5a and 6a). The IGHJ6 gene is the longest among human IGHJ gene segments⁶², and its increased usage has been observed among antibodies with long CDR H3 loops⁶³. This may contribute to the maturation of antibodies capable of reaching into the deep NA active site pocket (Fig. 6f), thereby potentially facilitating the evolution of DR motif-bearing antibodies that mediate receptor mimicry. Together, these findings expand the set of known DR motif antibodies, providing additional evidence for the structural and mechanistic convergence of this antibody class and highlighting its potential as a target for NA-based universal influenza vaccine.