Mechanistic insights into the small-molecule inhibition of influenza A virus entry

Significance

Influenza A virus (IAV) poses a persistent global health threat due to its rapid evolution and immune evasion, limiting the effectiveness of existing vaccines and treatments. Targeting hemagglutinin (HA), the key viral entry protein, offers a promising antiviral approach, yet developing broad-spectrum inhibitors is hindered by structural differences between HA subtypes and their susceptibility to mutations. Using cryo-EM, we uncover how two small molecules inhibit group 2 IAV by stabilizing HA in its prefusion state, blocking viral entry. Our study highlights the challenges of universal inhibition and introduces a “divide and conquer” strategy—designing group-specific inhibitors and combination therapies to achieve broader protection. These insights advance antiviral drug development against IAV and emerging viral threats.

Abstract

Influenza A virus (IAV) is a zoonotic pathogen responsible for seasonal and pandemic flu. The extensive genetic and antigenic diversity within and between IAV phylogenetic groups presents major challenges for developing universal vaccines and broad-spectrum antiviral therapies. Current interventions provide limited protection due to the virus’s high mutation rate and capacity for immune evasion. Recent advancements in viral hemagglutinin (HA)-targeting small-molecule entry inhibitors offer a promising avenue to overcome these limitations. Here, we present structural and functional analyses of two group 2 HA-specific small-molecule inhibitors recently identified by our team. Cryogenic electron microscopy (cryo-EM) structures revealed that these inhibitors bind a conserved pocket within the HA stalk, likely interfering with the conformational rearrangements necessary for membrane fusion and viral entry. Structure-guided mutagenesis confirmed the critical roles of key interacting residues and uncovered distinct resistance profiles between the two compounds, as well as in comparison to Arbidol, a previously reported HA inhibitor. Notably, our structural analysis highlights intrinsic barriers to achieving cross-group inhibition with current small-molecule designs. To address this, we propose an alternative strategy for broadening antiviral coverage. Together, these findings provide mechanistic insights into IAV entry inhibition and a foundation for the rational design of next-generation anti-influenza therapeutics.

Influenza A virus (IAV) is a major pathogen responsible for causing seasonal epidemics and occasional pandemics, thereby posing significant global public health challenges (1). Seasonal influenza infections result in considerable morbidity and mortality, with the World Health Organization (WHO) estimating millions of severe cases and hundreds of thousands of deaths annually (2). Influenza pandemics lead to even more devastating consequences, as seen with the 1918 H1N1 pandemic, which caused an estimated 50 million deaths worldwide (3). The more recent pandemic, the H1N1 “swine flu” in 2009, and ongoing outbreaks of H5N1 and H7N9 avian influenza strains underscore the need for effective preventive and therapeutic measures (4–7). Lately, there has been heightened vigilance regarding the cross-species transmission of highly pathogenic avian influenza (HPAI) viruses to humans (8, 9).

The IAV genome exhibits high rates of mutation and reassortment, leading to substantial genetic diversity (10, 11). This diversity is particularly evident in the hemagglutinin (HA) surface protein, which is the primary target for neutralizing antibodies and a key determinant of viral infectivity and host specificity (12, 13). HA mediates IAV entry into host cells through a series of conformational changes from a prefusion to a postfusion state that occurs in the low pH environment of the host cell endosome (14). IAVs are phylogenetically classified into two major groups based on the HA sequences and there are a total of 19 lineages: group 1, which includes subtypes such as H1, H2, and H5, and group 2, which includes subtypes such as H3, H7, and H10 (15, 16). Additionally, H19, a recently discovered subtype in the common pochard, has not yet been antigenically classified (17). Significant antigenic and sequence variations exist not only between the two major HA groups but also within each group, posing a significant challenge in the development of universal vaccines and broad-spectrum antiviral therapies (18).

Current influenza vaccines are designed to protect against specific seasonal virus strains and require annual immunological revisions to match the most prevalent circulating variants (19). The effectiveness of these vaccines varies every season, primarily depending on the antigenic match between the vaccine strains and the circulating viruses; a close match typically results in higher efficacy (20). However, since influenza viruses mutate rapidly through antigenic drift, there is a reduction in the effectiveness of the vaccine if the circulating strains differ from those predicted (21, 22). These challenges underscore the need for alternate and complementary approaches to influenza prevention and treatment, such as small-molecule inhibitors targeting the conserved regions of IAV proteins.

Small-molecule drugs have been developed as direct-acting antivirals (DAAs) for combating seasonal and pandemic flu (23). The US FDA has approved adamantanes (M2 ion channel inhibitors), neuraminidase (NA) inhibitors (NAIs) oseltamivir (Tamiflu®), zanamivir (Relenza®), peramivir (Rapivab®), and laninamivir (Inavir®) and a cap-dependent endonuclease inhibitor baloxavir marboxil (Xofluza) as anti-influenza therapeutics (24). However, their efficacy is diminished or limited by the rapid mutations of IAV, which lead to drug resistance. Therefore, there is an urgent need to develop antivirals with novel mechanisms which can be used alone or in combination clinically to treat IAV infection.

Many small molecules inhibit IAV by blocking the low pH-induced conformational changes in HA required for viral entry. These inhibitors often show group-specific activity. For example, tert-butyl hydroquinone (TBHQ) (25, 26), and MBX2546 (27) target group 2 HAs, while F0045(S) (28), JNJ4796 (29), CBS1117 (30, 31) are specific for group 1 HAs. Some compounds, like umifenovir (Arbidol) (32), N-cyclohexyltaurine (33), and IY7640 (34), show cross-group activity but with reduced potency.

In this study, we employed cryogenic electron microscopy (cryo-EM) to determine the structures of two group 2 HA-targeting small-molecule inhibitors, ING-16-36 and SA-67, in complex with H7 HA. Both compounds demonstrated strong inhibitory activity against H7-pseudotyped influenza viruses. Together with structure-guided mutagenesis and comparisons to previously reported HA-Arbidol structures, our results provide mechanistic insights into small-molecule inhibition of HA-mediated entry and establish a structural framework for rational lead optimization and influenza antiviral development.

Results

Structural Analysis of the Complexes of Small-Molecule Inhibitor and HA by Cryo-EM.

We previously developed two influenza entry inhibitors through structure–activity-relationship (SAR)-guided optimization of a selective imidazo[1,2-a]pyrimidine-based initial hit (CBS1194). These efforts led to the identification of an improved inhibitor ING-16-36 [2-(2-chloropyridin-4-yl)-N-cyclohexyl-5,7-dimethylimidazo(1,2-a)pyrimidin-3-amine], previously reported as “compound 43” (35) (H7 pseudovirus IC₅₀ = 0.31 ± 0.04 µM; H5 pseudovirus IC₅₀ > 30 µM), and subsequently an ultrapotent inhibitor SA-67, a desmethyl analog, N-cyclohexyl-2-(4-fluorophenyl)imidazo(1,2-a)pyrimidin-3-amine], previously reported as compound 4 h (36) (H7 pseudovirus IC₅₀ = 0.04 ± 0.01 µM; H5 pseudovirus IC₅₀ > 30 µM) (SI Appendix, Fig. S1 and SI Appendix, Table S1). To elucidate the molecular mechanism underlying the inhibitory activity of these compounds against HA-mediated viral entry, we conducted cryo-EM structural studies of ING-16-36 and SA-67 in complex with the HA protein of a group 2 IAV (H7 subtype, H7N9-SH13). The trimeric recombinant HA protein, stabilized with a T4 fibritin trimerization motif, was expressed and purified and used for the cryo-EM studies. The H7 SH13 HA trimer was incubated with a threefold molar excess of each compound at 4 °C for 1 h and vitrified on Quantifoil grids. Cryo-EM data were collected on a Titan Krios equipped with a Gatan K3 camera operating in movie mode (2 s exposure, 50 e/Ų dose). Final 3D refinements with C3 symmetry were performed using CryoSPARC, yielding resolutions of 2.77 Å for SA-67 and 2.76 Å for ING-16-36 (Table 1 and SI Appendix, Figs. S2 and S3). The resulting structures revealed that both inhibitors bind to the same hydrophobic pocket located in the stem region of the HA trimer (Fig. 1A), stabilizing interprotomer interactions. Each HA trimer binds three symmetric copies of the small-molecule inhibitors, which are tightly embedded in the hydrophobic pockets (Fig. 1B). This binding pocket is formed by residues from both HA1 and HA2 (indicated in italics) domains of adjacent HA protomers. For SA-67, the hydrophilic pocket entrance helps anchor the compound via several charged or polar residues, including a hydrogen bond network formed by R54 and E57 of HA2 from one protomer, and E97 of HA2 from an adjacent protomer. Additionally, Q302 from the neighboring HA1 and E97 form hydrogen bonds with the fluorine atom of SA-67. The interior of the pocket is predominantly hydrophobic, with the cyclohexyl group of SA-67 engaging in nonpolar interactions with L19 (HA1), L98 (HA2), and L99 (HA2 from an adjacent protomer). The imidazopyrimidine core is further stabilized by hydrophobic contacts with Y94 and L99 (HA2) and P284 and F285 (HA1). At the top of the pocket, the nitrogen atoms at positions 1 and 8 of the imidazopyrimidine ring form hydrogen bonds with R298 (HA1), which also contributes to pocket stabilization through hydrogen bonding with T59. T59, in turn, interacts hydrophobically with the phenyl ring of SA-67.

Table 1.

Data collection andrefinement statistics for cryo-EM

	H7 HA: SA-67	H7 HA: ING-16-36
	EMD-70657	EMD-70658
	PDB-9ONZ	PDB-9OO1
Data Collection
Microscope	FEI Titan Krios	FEI Titan Krios
Voltage (kV)	300	300
Electron dose (e-/Å2)	50	50
Detector	Gatan K3	Gatan K3
Pixel Size (Å)	0.899	0.899
Defocus Range (µm)	−1.0 to −2.5	−1.0 to −2.5
Magnification	81,000	81,000
Reconstruction
Software	cryoSPARC	cryoSPARC
Particles	155,278	195,885
Symmetry	C3	C3
Box size (pix)	360	360
Resolution (Å) (FSC0.143)*	2.77	2.76
Refinement (Phenix) †
Protein residues	1470	1470
Chimera CC	0.9	0.9
Resolution (Å) (FSC0.5)	3	2.9
EMRinger Score	3.62	3.94
R.m.s. deviations
Bond lengths (Å)	0.003	0.003
Bond angles (°)	0.554	0.599
Validation
MolProbity score	1.88	1.87
Clash score	4.52	4.3
Rotamers outliers (%)	0	0
Ramachandran
Favored regions (%)	96.3	97.19
Disallowed regions (%)	0	0

^*Resolutions are reported according to the FSC 0.143 gold-standard criterion.

^†Statistics are reported for the protein residues within the complex.

Fig. 1.

Cryo-EM structure of HA in complex with SA-67 and ING-16-36. (A) Cryo-EM density map of HA (H7N9-SH13) complexed with the influenza entry inhibitor SA-67. HA domains are individually colored (HA1: cyan, blue, and gray; HA2: purple, pink, and light gray). SA-67 is shown in green. (B) Ribbon representation of the HA:SA-67 complex, with the compound shown as spheres. Right: Zoomed-in views. Top, SA-67 (sticks) occupies a binding pocket (surface) formed by the HA2 domains of two adjacent protomers. Bottom, three SA-67 molecules symmetrically occupy the trimer interface (Top–Down transection view). (C) Cryo-EM density map of H7N9-SH13 HA in complex with ING-16-36. HA domains are colored as in (A); ING-16-36 is shown in yellow. (D) Ribbon representation of the HA:ING-16-36 complex with the compound shown as spheres. Right: Zoomed-in views. Top, ING-16-36 (sticks) is embedded within a similar HA2 pocket (surface) as in SA-67 binding. Bottom, three copies of ING-16-36 symmetrically occupy the HA trimer interface (Top–Down view).

ING-16-36 binds to HA in a manner similar to SA-67 (Fig. 1 C and D). However, the addition of two methyl groups at the 5- and 7- positions of its imidazopyrimidine ring induces conformational adjustment in the binding pocket to accommodate the bulkier structure. These protruding groups displace key pocket-forming residues, including the side chains of R298 and K301 (from two protomers), which are pushed upward at the top of the pocket. Additionally, the loop region at the pocket entrance, comprising residues R54-T59 from the same protomer, shifts outward slightly to accommodate the chlorine atom on the pyridine ring (SI Appendix, Fig. S4). These structural rearrangements require the local remodeling of the HA stalk pocket, which likely incurs an energetic penalty and reduces the thermodynamic favorability of binding. As a result, despite forming similar contacts as SA-67, ING-16-36 exhibits suboptimal binding affinity and reduced inhibition potency. This suggests that the compound’s potency could be further improved by optimizing its steric and electrostatic complementarity to the preformed pocket conformation of HA.

Structure-Based Mutagenesis.

To validate the importance of pocket-forming residues in HA for small-molecule binding, we performed structure-based mutagenesis in H7 HA and evaluated the impact of residue substitutions on inhibitor activity using an HIV-Luc-H7 pseudovirus system. Targeted substitutions were introduced into key residues in the HA1 subunit (L19, P284, F285, and R298) and the HA2 subunit (L55, Y94, L98, and L99). For each position, we selected one amino acid with similar physicochemical properties (charge or hydrophobicity) and one that differed more drastically. Only substitutions generating >10⁴ RLU luciferase signals were included in IC₅₀ analysis. We defined a ≥5-fold change in IC₅₀ compared to wild-type (WT) H7 as significant, reflecting an important role of the residue in drug binding. We observed that all selected substitutions resulted in increased IC₅₀ for SA-67 (WT IC₅₀ = 0.04 μM), and fold changes ranging from 10-fold (P284A) to 229-fold (Y94H) (Fig. 2A and SI Appendix, Fig. S5). The positive charge of R298 appears to mediate ionic interactions of SA-67 with HA, as its substitution with lysine had less impact on the IC₅₀ (sevenfold) compared to a substitution with alanine (13-fold). For ING-16-36 (WT IC₅₀ = 0.31 μM), substitution at L19, P284, L55, L98, and L99 reduced its inhibitory potency (Fig. 2B and SI Appendix, Fig. S6). In contrast to SA-67, residue R298 played a less crucial role for ING-16-36, likely a result of the steric effects of the 5- and 7-methyl groups in the imidazopyrimidine scaffold. Substitutions Y94H and L99A caused the most dramatic reductions in potency for both compounds (220 and 198-fold for SA-67, and over 300-fold for ING-16-36), strongly supporting the critical role of these two residues in compound binding. For both inhibitors, residues L19, L98, and L99 contribute to the overall hydrophobicity of the binding pocket. Y94 is essential for π-π stacking with the imidazopyrimidine scaffold. L19 is particularly important for accommodating the cyclohexane ring of the inhibitors in this snug hydrophobic pocket. F285 plays an important role through aromatic stacking interaction for SA-67, as evidenced by a 43-fold and 20-fold increase in IC₅₀ upon leucine and tryptophan substitutions, respectively. In contrast, this residue is less critical for ING-16-36 binding, as substitution with tryptophan caused a negligible IC₅₀ change, although replacing it with leucine reduced inhibition by fivefold, likely due to a loss of aromaticity. Overall, our mutagenesis studies highlight the key residues involved in small-molecule binding and inhibition, providing valuable insights for further structure-based optimization of these small molecules.

Fig. 2.

Mutagenesis analysis of HA interactions with SA-67 and ING-16-36. (A) Structural visualization of the HA–SA-67 complex. SA-67 and key interacting HA residues are shown as sticks; hydrogen bonds are indicated with yellow dashed lines. The accompanying table lists IC₅₀ values for SA-67 inhibition across various HA mutants, along with the fold-change relative to wild-type virus. (B) Structural visualization of the HA–ING-16-36 complex. ING-16-36 and interacting HA residues are shown as sticks, with hydrogen bonds represented by yellow dashed lines. The table summarizes IC₅₀ values for ING-16-36 against each mutant and the corresponding fold-change compared to wild-type.

Comparison with Arbidol.

Arbidol is a small-molecule inhibitor with antiviral activity against various respiratory viruses, including IAV (32), and demonstrates greater entry inhibition of group 2 versus group 1 IAVs. In pseudovirus entry assays using wild-type H7, SA-67 (IC₅₀ = 0.04 μM) and ING-16-36 (IC₅₀ = 0.31 μM) were approximately 100-fold and 10-fold more potent, respectively, than Arbidol (IC₅₀ = 4.75 μM; SI Appendix, Table S1). High-resolution X-ray crystallography structures of Arbidol bound to H3 and H7 HA have been reported (37), enabling direct comparison of its binding modes with those of SA-67 and ING-16-36. Structural analysis revealed that all three compounds target the same hydrophobic pocket in the HA stem region (Fig. 3A) but differ in the extent and specificity of their interactions with pocket-forming residues. These differences in binding rationalize the mutagenesis results obtained using the same panel of pocket residue substitutions as our compounds: While SA-67 and ING-16-36 rely on a dense network of specific contacts, especially involving residues like Y94, L99, and F285, Arbidol forms fewer and more flexible interactions, making it less sensitive to these mutations. For example, the Y94H substitution led to a 229-fold decrease in SA-67 potency but only a fourfold reduction in Arbidol activity, consistent with the more adaptable fit of Arbidol within the pocket (Fig. 3 B and C and SI Appendix, Fig. S7). Likewise, substitution at L19, P284, L55, L98, and L99 resulted in less than a fivefold change in Arbidol’s IC₅₀ relative to wildtype H7 pseudovirus, indicating these residues play a lesser role in its binding. Notably, the F285L and R298A replacements unexpectedly increased Arbidol’s inhibitory potency, reducing its IC₅₀ by 10-fold and fivefold, respectively. This potency enhancement is likely attributable to the longer, more flexible ethyl-ester side group of Arbidol, in contrast to the rigid imidazopyrimidine scaffold of SA-67 and ING-16-36 (Fig. 3 B and C). The flexibility of Arbidol allows it to tolerate conformational variations within the binding pocket and adapt to residue changes that might otherwise hinder tighter binders. For example, the F285L replacement introduces a smaller side chain, creating additional space that permits deeper insertion of Arbidol’s methyl group into the pocket, thereby improving its binding affinity and antiviral potency.

Fig. 3.

Structural and mutational analyses reveal distinct HA binding modes and resistance profiles for SA-67, ING-16-36, and Arbidol. (A) Structural visualization of the HA–Arbidol complex (PDB: 5T6S). Arbidol and its interacting HA residues are shown as sticks. The accompanying table lists IC₅₀ values for Arbidol inhibition across HA mutants, along with fold-change relative to wild-type virus. (B) Structural overlay of SA-67 and Arbidol bound to H7 HA. Both compounds are shown as sticks, with interacting HA residues depicted as lines. (C) Structural overlay of ING-16-36 and Arbidol bound to H7 HA. ING-16-36 and Arbidol are shown as sticks, with interacting HA residues depicted as lines.

Structural Basis for Group-Specific IAV Entry Inhibition.

Structural analyses revealed that the HA pocket targeted by SA-67 and ING-16-36 is conformationally conserved across group 2 IAV but structurally altered in group 1 IAVs (Fig. 4A). In group 1 HA proteins, residue M59 (equivalent to T59 in the H7 HA used for this work) adopts an inward-pointing conformation, forming hydrophobic interactions that occlude the pocket and prevent inhibitor binding (Fig. 4B). This structural divergence likely accounts for the lack of strong inhibition by the two compounds observed against group 1 IAVs (SI Appendix, Table S1) (35). To further investigate the basis of this group-specificity, we compared the binding mode of our group 2 inhibitors with that of CBS1117 (31), a previously reported group 1-specific inhibitor (Fig. 4C). Notably, the binding site of SA-67 and ING-16-36 is approximately 20 Å away from that of CBS1117. Additionally, an N-linked glycosylation site near the CBS1117 binding region in group 2 HA likely interferes with its binding, contributing to its lack of activity against group 2 viruses (Fig. 4D). These structural and functional differences clarify the lack of cross-group inhibition and underscore the distinct inhibitory mechanisms between group 1 and group 2 HA inhibitors. Broader comparisons of available cocrystal structures of small-molecule inhibitors bound to group 1 or group 2 HA (28, 29, 31, 33, 36, 37) reveal a consistent pattern: Group 1–specific inhibitors target a pocket located near the fusion peptide in the lower stalk region (Fig. 4E). In contrast, group 2 inhibitors, including SA-67 and ING-16-36, engage a pocket located further from the fusion peptide, in a upper region that remains structurally accessible to entry inhibitors across diverse group 2 IAV HAs.

Fig. 4.

Comparison of group 1 and group 2 influenza HA entry inhibitors. (A) Structural comparison of the HA stem regions from group 1 and group 2 viruses, illustrating distinct conformations in the SA-67 binding pocket between the two groups. (B) The M59 residue in group 1 HA inserts into the SA-67 binding pocket, blocking inhibitor binding, whereas the corresponding residues in group 2 (T or P) do not interfere with SA-67 inhibition. (C) Spatial comparison of the binding pockets for the group 2 inhibitor SA-67 and the group 1 inhibitor CBS1117. (D) The N-linked glycosylation site at N38 in group 2 HA contributes to the incompatibility of CBS1117 with group 2 HAs. (E) Side-by-side display of seven small-molecule inhibitors: F0045(S) (red spheres, PDB: 6WCR), JNJ4796 (purple spheres, PDB: 6CFG), CBS1117 (yellow spheres, PDB: 6VMZ), TBHQ (blue spheres, PDB: 3EYM), N-cyclohexyltaurine (cyan spheres), Arbidol (orange spheres), and SA-67 (green spheres, this work). HAs are shown as surfaces in gray, light purple, and pink for the three protomers. A reference dashed line highlights the distinct locations of inhibitor binding pockets between group 1 and group 2 HAs.

Discussion

In this study, we present cryo-EM structures of IAV HA in complex with two potent small-molecule inhibitors, SA-67 and ING-16-36. Structural comparison with previously reported HA-inhibitor complexes determined by X-ray crystallography provides insights into the mechanism by which small molecules block HA-mediated membrane fusion and viral entry. These insights offer a foundation for rational, structure-guided design and lead optimization of next-generation antivirals against diverse IAVs.

Both SA-67 and ING-16-36 exhibit group 2-specificity, effectively inhibiting H3 and H7 IAV subtypes but not group 1 IAV subtypes, such as H1 and H5 (35) (SI Appendix, Table S1). These compounds bind a conserved pocket in the group 2 IAV HA stalk region (characterized here using H7 HA), engaging both the HA1 and HA2 subunits and enhancing interprotomer interactions. This binding mode stabilizes the prefusion conformation of HA, thereby preventing the conformational rearrangements required for membrane fusion and viral entry. This mechanism is supported by our prior observation that the original hit compound from this series of IAV inhibitors, 2-(3-bromophenyl)-N-[2,3-dihydrobenzo(b)(1, 4)dioxin-6-yl]-5,7-dimethylimidazo[1,2-a]pyrimidin-3-amine, CBS1194, inhibits limited trypsin-mediated cleavage of H7 HA protein (38). Conservation of this binding pocket across group 2 subtypes, such as H3 and H7, supports the broad reactivity of these compounds (SI Appendix, Fig. S8). Notably, ING-16-36 is an early intermediate in our SAR campaign, while SA-67 is a more advanced compound that has been further optimized for potency and fit. Although ING-16-36 binds to HA in a similar way to SA-67, its inhibitory potency is somewhat lower. Structural analysis shows that the two extra methyl groups at the 5- and 7-positions of its imidazopyrimidine ring push key pocket-forming residues upward and cause the loop at the pocket entrance to shift outward. This localized remodeling of the HA stalk pocket indicates that ING-16-36 binding involves conformational adjustments that incur an energetic cost. Such induced fit likely reduces the thermodynamic favorability of binding, resulting in suboptimal affinity and inhibitory activity. These findings highlight the importance of maintaining steric and electrostatic complementarity with the preformed pocket conformation to achieve maximal inhibitor efficacy. Future optimization efforts should aim to strengthen protein–ligand interactions while minimizing the structural rearrangements required for compound accommodation.

Several other small-molecule inhibitors, TBHQ (26), Arbidol (37), and N-cyclohexyltaurine (33), also target the same region in group 2 HA. Although SA-67, ING-16-36, and Arbidol all bind the same HA stalk pocket and interact with a similar set of residues, SA-67 exhibits two orders of magnitude greater inhibitory potency than Arbidol, while ING-16-36 is approximately one order of magnitude more potent (Fig. 3). This enhanced activity can be attributed to optimized protein–ligand interactions that increase HA engagement and entry inhibition. However, this enhanced binding also renders SA-67 and ING-16-36 more sensitive to sequence variation in the pocket. Indeed, substitutions of key pocket residues had a significantly greater impact on their potency compared to Arbidol (Figs. 2 and 3). Thus, while improved interactions confer higher potency, they also reduce tolerance to natural mutation or targeted residue substitution, resulting in reduced activity against some HA variants, approaching the potency of Arbidol in some cases.

Notably, residue 298 is an Arginine in H7 and H15 HAs, but a Lysine in H3, H4, H10, and H14 HAs. Our mutagenesis analysis revealed that the R298K substitution in H7 HA caused only a moderate reduction in inhibitor potency, a sevenfold decrease for SA-67 and a threefold decrease for ING-16-36, while the R298A substitution had a more pronounced effect. These findings suggest that the side chain at position 298 contributes to compound binding and that charge retention (as in R298K) partially preserves activity. Many of the substitutions that confer resistance to our inhibitors, such as L99A and Y94H, are located at highly conserved and structurally buried positions within the HA stem. These residues are likely to play important roles in HA folding or in mediating the conformational rearrangements required for membrane fusion. Given their location in the hydrophobic core, mutations at these sites may carry a fitness cost, though this remains to be confirmed experimentally. These features suggest that the HA stem pocket may be subject to functional constraints that could limit the emergence of resistance. Future work is required to profile drug resistance-conferring mutations in the HA stalk to better understand the genetic barrier to resistance and to define sequence constraints that preserve the druggable conformation of this pocket.

Compared to group 2-targeting small-molecule inhibitors, the three reported group 1-specific inhibitors, JNJ4796 (29), CBS1117 (31), and F0045(S) (28), bind a distinct pocket in the HA stalk lower region roughly 20 Å from that of group 2 inhibitors (Fig. 4 C−E). This spatial separation may explain the persistent challenges in identifying small-molecule inhibitors with cross-group activity. Although Arbidol has been reported to inhibit both group 1 and group 2 IAVs, it shows ~4-fold greater potency against group 2 IAVs, and only the group 2 HA/Arbidol complex has been structurally characterized (37). Whether Arbidol binds the upper HA stalk pocket (as it does in group 2) or instead occupies the lower pocket targeted by group 1 inhibitors remains unresolved. Functional data, including protection of HA from trypsin digestion, support Arbidol’s role in stabilizing HA and preventing fusion (37). Efforts to optimize Arbidol have shown promise. Wright et al. reported that Arbidol derivatives displayed comparable binding affinities to HA proteins from both groups (39). While the antiviral efficacy of these derivatives across both IAV groups remains to be fully characterized, these findings suggest that pan-IAV entry inhibitors may be achievable.

Alternatively, we propose a “Divide and Conquer” strategy for developing HA-targeted small-molecule entry inhibitors: to develop separate compounds with HA group-specific activities. For example, ING-1466, a group 1-specific inhibitor developed in our lab, has demonstrated in vivo antiviral efficacy and is a promising candidate for treating infections caused by H1N1 and H5N1 IAVs (40). Similarly, SA-67 and ING-16-36, which potently inhibit group 2 subtypes, such as H3N2 and H7N9, represent strong leads for group 2-specific antiviral development. Therefore, future efforts could focus on the structure-guided optimization of these inhibitors, including chemical modifications to improve binding affinity, broaden subtype coverage within each IAV group, and enhance in vivo antiviral efficacy. The successful development of an mRNA vaccine encoding all 20 HA subtypes (41) provides further rationale for pursuing subtype-specific inhibitors as part of a broader antiviral toolkit. The integration of cryo-EM structural information, structure-based mutagenesis, and computational modeling will be collectively instrumental in guiding this next phase of drug design. Finally, compared to X-ray crystallography, our cryo-EM protocol offers greater generalizability and faster determination of HA-compound complex structures with more reproducible outcomes, despite slightly lower resolution. With ongoing advances in cryo-EM detector technology and sample preparation methods, achieving atomic-level resolution (<2 Å) is increasingly feasible and will further empower structure-based antiviral drug design.

Materials and Methods

Cell Culture.

Human A549 lung epithelial cells (ATCC# CCL185) and 293 T embryonic kidney cells (ATCC# CRL-1573) were cultured in the growth medium Dulbecco’s Modified Eagle Medium (DMEM) that was supplemented with 10% (v/v) fetal bovine serum (FBS) and 1% penicillin-streptomycin (P/S). Expi293F cells, acquired from Thermo Fisher Scientific, were maintained in Expi293 Expression Medium (Thermo Fisher) at 37 °C with 8% CO₂ with orbital shaking at 120 rpm.

HA Protein Preparation.

H7 HA trimer from the H7N9 A/Shanghai/02/2013 (SH13) virus was prepared by fusing the HA construct to a thrombin cleavage sequence, T4 fibritin trimerization motif (foldon), and a hexa-histidine (6X His) affinity tag at its C-terminus. The DNA sequence was codon optimized for mammalian expression and synthesized by GenScript and cloned into a pVRC8400 expression plasmid as previously described (42). Expi293 cells were diluted to 1.2 × 10⁶ cells/mL and transfected with 1 mg/L of HA expression plasmid using Turbo293 transfection reagent. After six days, the medium was clarified by centrifugation at 10,000×g and filtered, then concentrated and loaded onto Complete His-Tag Resin (Roche) by gravity flow. The resin was washed with three column volumes of PBS containing 50 mM imidazole, and the target protein was eluted with PBS containing 300 mM imidazole. The eluted protein was concentrated and further purified by size exclusion chromatography using a HiLoad 16/600 Superdex 200 pg column (Cytiva), with fractions corresponding to the SH13 HA trimer being collected, pooled, and concentrated.

Cryoelectron Microscopy Structural Analyses.

The H7 SH13 HA trimer was incubated with a threefold molar excess of compound on ice for 1 h to form a complex. 3 µL of the complex at a concentration of 0.5 mg/mL was deposited on a copper mesh 300 1.2/1.3 carbon grid (QUANTIFOIL). Vitrification was performed at 4 °C using an FEI Vitrobot Mark IV, with a 30-s wait time, 4-s blot time, blot force of 2, and 100% humidity. Automated data collection was carried out on a Titan Krios microscope using a Gatan K3 direct detection device via Thermo Fisher EPU software. Exposures were taken in movie mode for 2 s with a total dose of 50 e/Ų. Data processing was performed in CryoSPARC 3.3 (43). C3 symmetry was applied for the final refinement. Model building was performed in Coot (44), followed by simulated annealing and real-space refinement in Phenix (45), and iterative manual fitting in Coot (44). Geometry and map fitting were assessed using MolProbity (46) and EMRinger (47). Figures were generated using PyMOL (www.pymol.org) and ChimeraX (48).

Mutagenesis Studies.

As previously described (36), H7 HA pseudotyped viruses with targeted residue substitution were generated using the Agilent Technologies QuickChange Lightning Site-Directed Mutagenesis kit (Cat# 210518). Site-specific nucleotide substitutions were introduced by PCR amplification using primers containing the desired mutations, designed via Agilent’s online QuickChange Primer Design Program. Mutated plasmids were verified by Sanger sequencing to confirm successful incorporation of the intended changes. For downstream analysis, only mutant pseudoviruses that produced luciferase signals exceeding 1 × 10⁴ relative light units (RLU) in viral entry assays were selected for further characterization.

Viral Entry Assays.

Antiviral activities of the small molecules were determined using a lentiviral pseudotyped virus system as described previously (36). Briefly, plasmids encoding monobasic H7 HA [A/Anhui/1/2013 (H7N9)], influenza neuraminidase 1 (NA1), and replication-defective HIV-1 vector pNL4-3.Luc.R − E− (from the NIH AIDS Research and Reference Reagent Program), containing a luciferase reporter gene, was used to create pseudovirions. Pseudovirus production utilized a polyethylenimine (PEI)-based transient cotransfection system on 293 T cells. Five hours posttransfection, the cells were washed with phosphate-buffered saline (PBS) and replaced with fresh reduced serum media Opti-MEM (Gibco). Supernatants containing HIV-luc-H7N1 pseudovirus were collected 16 h later and filtered through a 0.45 μm pore size filter (Corning). The pseudovirions carrying monobasic H7 HA were then treated with 1 μg/mL of N-p-tosyl-L-phenylalanine chloromethyl ketone (TPCK)-treated trypsin (Sigma) and incubated at 37 °C in a water bath for 1 h to cleave the H7 glycoprotein prior to infection. To obtain IC₅₀ values of the inhibitors, A549 cells were seeded at 2 × 10⁴ cells/well into white 96-well plates. Cells were infected with the pseudovirions with or without compounds, all maintained at 1% DMSO concentration. For cytotoxicity, cells were treated with or without compounds in DMEM. Compounds were serially diluted at two- or threefold ratios starting at 100 μM to obtain the CC₅₀ and IC₅₀ values, respectively. After incubating the plates for 48 h, the level of viral entry was determined by measuring the luciferase activity using the Neolite Reporter Gene Assay System (PerkinElmer); cytotoxicity was measured using CellTiter-Glo Luminescent Cell Viability Assay (Promega). H7 pseudovirus inhibition and compound cytotoxicity were normalized using the 1% DMSO control. IC₅₀ and CC₅₀ values were determined by evaluating the dose−response data using four-parameter logistic regression (4PL) analysis in GraphPad Prism (version 9.3.0). The 4PL regression is commonly used for modeling dose–response curves in pharmacology because it effectively captures the sigmoidal (S-shaped) behavior typically observed in biological responses to increasing concentrations of a drug or compound.

Sequence Conservation Analysis and Logo Plot Generation.

To assess the conservation of residues within the compound binding pocket among group 2 IAV HA, we collected representative HA sequences from multiple subtypes within group 2 (including H3, H4, H7, H10, H14, and H15) from the NCBI Influenza Virus Resource. Sequences were aligned using Clustal Omega. Residues lining the binding pocket were identified based on structural analysis of the HA-compound complex. The aligned residues corresponding to the pocket positions were extracted and visualized using WebLogo (49, 50) v.2.8.2 (https://weblogo.berkeley.edu/). The resulting sequence logo plot illustrates the degree of conservation and variability of each pocket residue across group 2 HAs.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Structural data have been deposited in PDB and EMDB [PDB-9ONZ (51), PDB-9OO1 (52), EMD-70657 (53), EMD-70658 (54)]. All study data are included in the article and/or supporting information.