How a paramyxovirus fusion/entry complex adapts to escape a neutralizing antibody

Abstract

Paramyxoviruses including measles, Nipah, and parainfluenza viruses are public health threats with pandemic potential. Human parainfluenza virus type 3 (HPIV3) is a leading cause of illness in pediatric, older, and immunocompromised populations. There are no approved vaccines or therapeutics for HPIV3. Neutralizing monoclonal antibodies (mAbs) that target viral fusion are a potential strategy for mitigating paramyxovirus infection, however their utility may be curtailed by viral evolution that leads to resistance. Paramyxoviruses enter cells by fusing with the cell membrane in a process mediated by a complex consisting of a receptor binding protein (HN) and a fusion protein (F). Existing atomic resolution structures fail to reveal physiologically relevant interactions during viral entry. We present cryo-ET structures of pre-fusion HN-F complexes in situ on surfaces of virions that evolved resistance to an anti-HPIV3 F neutralizing mAb. Single mutations in F abolish mAb binding and neutralization. In these complexes, the HN protein that normally restrains F triggering has shifted to uncap the F apex. These complexes are more readily triggered to fuse. These structures shed light on the adaptability of the pre-fusion HN-F complex and mechanisms of paramyxoviral resistance to mAbs, and help define potential barriers to resistance for the design of mAbs.

Monoclonal antibodies hold promise for combating serious respiratory virus infections but viruses may evolve to evade them. Here, using structural analysis, the authors show how human parainfluenza virus adapts to escape a powerful antibody by modulating its cell entry mechanism.

Introduction

The respirovirus human parainfluenza virus type 3 (HPIV3) is a leading cause of illness in pediatric¹, elderly, and immunocompromised populations^2,3, yet there are no vaccines or therapies available. For example, HPIV3 accounts for the vast majority of HPIV infections following transplantation, causing pneumonia with up to a 50% mortality rate⁴. The development of effective neutralizing mAbs (nAbs) against several respiratory viruses has led to an important new therapeutic strategy, with evidence that nAbs generated using several different approaches can be clinically effective against a range of respiratory viruses^5–12. Advancing prophylactic or therapeutic mAbs against HPIV3 could directly ameliorate child health¹ and the health of patients with immune compromise^2,3 while advancing a general strategy for respiroviruses. However, the utility of such mAbs may be severely limited by viral evolution that leads to resistance, a lesson emphasized by the experience with SARS-CoV-2¹³ where each newly effective nAb was inevitably evaded by viral evolution. To elucidate the mechanisms of HPIV3 nAb resistance and address the challenges to developing this approach, we investigate the structural basis for evasion of mAbs against HPIV3 by deriving structural information that directly reflects authentic states of the HPIV3 glycoprotein complex—the target of immunity—on the surface of virions.

HPIV3, like other paramyxoviruses, enters cells by fusing directly with the cell membrane in a process mediated by a receptor binding protein and a fusion protein (^14–18, reviewed in ref. ¹⁹). For HPIV, the fusion/entry complex consists of a receptor binding hemagglutinin-neuraminidase (HN) in close interaction with the fusion protein (F). Upon engagement of a cellular receptor by HN, the complex undergoes a series of structural transitions that may provide optimal targets for antibody-mediated inhibition. We have shown that prior to receptor engagement, HN stabilizes F to prevent F’s premature activation^17,20,21. Upon receptor engagement, HN switches to its F-triggering activity, and this initiates the process whereby F undergoes its large conformational shift that ultimately mediates membrane fusion and viral entry^{14,15,19,22–24}. The interface between HN’s globular heads in the HN dimer modulates the activation of F by HN and regulates fusion¹⁸.

Cryo-electron tomography (cryo-ET) structures of the HN-F fusion complex in its prefusion state on the surface of an authentic clinical virus at sub-nanometer resolution revealed the architecture of a paramyxovirus entry/fusion complex^25,26, showing possible mechanisms for stabilization of the pre-fusion complex. The structural organization of the glycoproteins in relation to each other in situ, where one of the globular heads of the HN dimer caps the apex of the pre-fusion F trimer, suggests how the pre-fusion HN-F complex is maintained in a ready but quiescent state prior to receptor engagement. This cap contains an HN loop structure that appears to interact with the apex of the F trimer and is highly conserved across paramyxoviruses, potentially pointing to a general mechanism for maintaining the fusion/entry complex’s stability and ensuring that activation of fusion occurs only at the right time and location²⁶.

Two antigenic sites have been characterized on HPIV3 F: one is located in the apex of the trimeric F²⁷ and a second is at the side of the globular portion of F²⁸. We showed that an anti-F neutralizing mAb, PIA174, that targets the apex of pre-fusion HPIV3 F blocks the structural transition of F to its activated intermediate and perturbs the HN/F complex^26,29. The nAb PIA174 stabilizes the prefusion state of F and thereby inhibits F activation, mirroring the proposed function of the HN domain that caps this site on F²⁶. The pocket in the F trimer apex that receives the HN loop overlaps the epitope of PIA174. Under the selective pressure of nAb PIA174 in viral evolution experiments, resistant variant viruses rapidly emerged²⁶ including two PIA174Ab escape variant (EV) viruses. One that we previously characterized functionally (EV-1) bears a mutation at the precise site on the F apex that we predict to be the site of HN/F interaction, A194T, along with a previously characterized mutation in the HN dimer interface²⁶. The mutation in F alters both PIA174Ab binding and HN’s activation of F, confirming experimentally that this HN/F interaction site is an important interface between HN and F and functionally overlaps with the site of PIA174 neutralizing antibody binding. In contrast, here we describe a second PIA174 escape mutant (EV-2) that bears two mutations in the HN dimer interface²⁶ and a unique mutation adjacent to the central helix of the F trimer, distant from the apex, that is responsible for its resistance to PIA174Ab.

In this work, the structures of the pre-receptor engaged HN/F complexes of these two EV viruses reveal that in both the HN protomer head is shifted away from the apex of F in comparison to the position of HN in the parental HN-F complex, “uncapping” the F apex. This shift produces a significant change in the relationship between the HN and F. The apex of F is exposed in the structure of the HN-F complexes of the variants, and in both cases, the nAb no longer binds the mutated F. The pre-fusion F proteins of both EVs are less stable and transition more readily from the pre-fusion state. The EVs differ in the functional impact of the mutations, their mechanisms of nAb escape, and the fitness costs to the viruses in human airway. The structural alterations in the nAb EV HN-F complexes highlight the notion that the relationship of the HN and F glycoproteins to each other in the complex correlates with function. Alterations at the apex—but also distal to the F apex—can affect F structure and alter antibody binding at the apex, affect HN-F interaction, thereby altering the HN-F complex, impact the stability of the F protein trimer, and modulate fusion. The emergence of these viruses with structurally and functionally altered HN-F complexes highlights the capacity of the paramyxovirus fusion/entry complex to adapt to external stimuli while supporting multiple essential functions. The HN/F complex has an inherent adaptability that will allow it to evade antibodies that target many exposed epitopes²⁶. The structures of these variants reveal distinct mechanisms of viral evolution to evade nAbs and help define potential barriers to resistance.

Results

Resistance properties of antibody escape variants

Viral evolution experiments were conducted by using a previously characterized HPIV3 neutralizing monoclonal antibody PIA174Ab²⁹ to select for variants that are infectious in the presence of the antibody. Two EV viruses emerged (Fig. 1A, B). One that we previously characterized functionally (EV-1) bears a mutation at precisely the site on the F apex that the complex’s structure predicts to be the site of HN/F interaction, A194T²⁶. This mutation alters both PIA174Ab binding and HN’s activation of F. The second PIA174 escape mutant bears a mutation L234F at the side of F, distant from the trimeric F apex but adjacent to the central helix of the F trimer (Fig. 1B). Both viruses also bear a mutation at the HN dimer interface (H552Q) that we have previously shown is a mutation that confers increased HN-F interaction and F activation in persistent HPIV3 variants^18,22,30,31 (for full sequences see NCBI BioProject PRJNA1083633). This PIA174 Fab-escape variant, called “escape variant (EV)−2” to distinguish it from the F trimer apex/HN dimer interface variant (EV-1), like EV-1, spread throughout a cell culture monolayer in the presence of high concentrations of Ab PIA174 (Fig. 1C). However, the initial viral entry, as measured by plaque formation (Fig. 1D) is inhibited for EV-2 in distinction to EV-1 which is completely resistant to inhibition of entry.

Fig. 1. EV1 and EV2 resistance to inhibition by PIA174.

Model of the full-length HN/F complex with mutations for (A) EV1 and (B) EV2 indicated by red atomic spheres. C Vero cells were infected with mCherry-bearing parental, EV1, or EV2 HPIV3 and overlaid with serial dilutions of PIA174 FAb. Total fluorescence was measured 2 days post infection and y-axis shows the percent inhibition of total read fluorescence for each virus. D Inhibition of viral entry for EV1 and EV2 was quantitated by plaque counting normalized to no treatment. E Cryo-EM model of prefusion F (PDBID: 6MJZ) with both PIA174 and 3 × 1 structures showing location of L234 (green) and A194 (red) residues in relation to 3x and PIA174 mAbs. F, G A194T and L234F recognition by PIA174. Cells transfected with A194T, L234F, or parental Fs stabilized by Q162C-L168C, I213C-G230C, A463V, I474Y mutations were treated with 1ug/mL PIA174 Fab or 3 × 1 mAb and fluorescent secondary anti-human antibody. Fluorescence was quantitated. Data are means ± SE from three separate biological replicates for (C, D) and (F, G).

Escape variant Fs evade nAb binding

Both the A194T and the L234F mutations (Fig. 1A, B) decrease PIA174 antibody binding to equally expressed F (Fig. 1F and Supplementary Fig. 1), a surprising finding in light of the distance of L234F from the F trimer apex binding site of PIA174 Fab. To assess whether the L234F mutation destabilizes the F central helix and thereby alters the PIA174 antibody binding epitope, we compared Ab binding to F-A194T, F-L234F, and parental F with Ab binding to disulfide-stabilized versions of these proteins (Fig. 1E). We introduced A194T and L234F mutations in a stabilized prefusion F protein that had been used as an antigen to generate both PIA174 and another known anti-HPIV3 F-specific neutralizing antibody, 3 × 1²⁸. The engineered stabilized F bears 4 cysteine mutations; one pair that connect the central helix (and A194) to the domain III helix-turn-helix that contains the L234F mutation, and another pair in the N-terminal heptad repeat domain containing the 3 × 1 Ab epitope. The cysteines substituted for residues I213 and G230 staple the central helix domain of the structure and reduce flexibility conferred by L234F, and cysteines substituted for Q162 and L168 staple the N-terminal heptad repeat domain of the structure. We asked whether the stabilizing mutations could compensate for the EV1 and EV2 mutations and restore PIA174 Fab binding. Figure 1G shows that while PIA174 Fab binds the stabilized parental F, it fails to bind the cysteine-stabilized F-A194T. However, PIA174 Fab binding is fully restored to the stabilized F-L234F (Fig. 1G). As expected, Ab 3 × 1 binds equally to all three stabilized Fs and is unaffected by A194T. L234F partially reduces Ab 3 × 1 binding, restored in the stabilized F (Fig. 1F, G). These results suggest that the escape mutation L234F, distant from the F trimer apex, nonetheless destabilizes the F trimer interface thereby perturbing the trimer apex PIA174 antibody epitope.

Resistance mutations alter HN-F complex function so that F readily activates but is defective in completing the fusion process

The PIA174 Fab neutralizes by stabilizing the pre-fusion state of F²⁶. Both escape mutations reverse this effect by conferring increased ‘triggerability’ of F from the pre-fusion state, i.e., facilitating F’s exit from the prefusion state to initiate its conformational change. Cells expressing parental F, A194T F, or L234F F were incubated at 55 °C for different times, and the % of F in pre-fusion state was detected using a conformation specific pre-fusion F antibody that binds a site distal to the apex of F (Fig. 2A). Compared to the parental F, both variant F-proteins exited the pre-fusion state much more quickly, with the pre-fusion F signal gone within 15 min for the variants and 1 h for parental F. While parental F is retained in the prefusion state by the presence of the PIA174 Fab in the face of exposure to high temperature (Fig. 2B), neither A194T F nor L234F F are stabilized by the antibody; they exit the pre-fusion state at the same rate in the absence (Fig. 2A) or presence (Fig. 2B) of Fab.

Fig. 2. Key PIA174 escape mutations alter the activatability of F.

Percent of F in pre-fusion state at different time points at 55 °C without (A) or with (B) PIA174 Ab fragment. C Rate of conformational change of uncleaved F with A194T or L234F measured by rate of lipopeptide capture at 37 °C with red blood cells. The HN bears a mutation (D216R) that makes it sialidase-deficient to maximize the HN-receptor contact and HN’s fusion promotion in order to compare the properties of the F proteins. D, E Triggering of F by heat (left) assessed at a range of temperatures with parental, A194T, or L234F Fs. F activation is quantitated by % of red blood cells (RBCs) released, bound, or fused with F-expressing cells. Activation of F by HN (right) was assessed at a range of temperatures for 60 min. with parental, A194T, L234F F. Data are means ± SE from at least three separate biological replicates for (A–E).

To measure the rate of conformational change of the various F proteins, we took advantage of the ability of HPIV3 HRC-lipopeptides to capture the transient extended intermediate of F^32,33. Cells co-expressing F bearing L234F, A194T, or no mutations and influenza HA or HN were allowed to bind to their sialic acid receptors on red blood cells (RBCs) at 4 °C, then incubated at 37 °C to permit F activation (Fig. 2C). The use of unprocessed F₀ proteins for this experiment allows specifically for measurement of the rate of conformational change of unfolding the pre-fusion F, without progress to F-insertion or fusion. The HRC-lipopeptides insert into the RBCs and capture extended F proteins, thereby tethering the RBCs to the F-expressing cells, without requiring membrane piercing or fusion mediated by F. Under these conditions, the F bearing L234F undergoes conformational change, extends, and is captured by lipopeptide on the RBC at a somewhat higher rate than parental F—although not as rapidly as the F bearing A194T (Fig. 2C).

To evaluate the impact of the L234F F mutation on the ability of F to be activated by heat or by HN and to progress towards insertion in a target membrane and complete fusion, compared to the impact of the F apex mutation A194T, we used an assay that distinguishes between different states of F activation and quantitates each state^22,34. The readout for F activation, insertion, and progress to fusion is fusion of RBCs with F-expressing cells. Cells co-expressing cleaved F bearing L234F, A194T, or no mutations and influenza HA or HN were allowed to bind to their sialic acid receptors on RBCs at 4 °C and transferred to a range of temperatures to permit F activation (Fig. 2D, E and Supplementary Fig. 2). Influenza HA serves only to tether the target RBCs to the F-expressing cells and does not actively trigger F¹⁷; this condition measures activation of F by heat (Fig. 2D). At each temperature, we measured the percentage of target RBCs that were either released into the medium (blue circles), bound by either HN/HA (red square) but had not fused, or had undergone fusion (green triangle). The F bearing A194T is more readily triggered by heat compared to the parental F in conditions where HA is used to tether but no HN is present. However, the F bearing L234F, despite more readily transitioning from the pre-fusion state as shown in Fig. 2A, does not progress to insertion in the target and to complete fusion. Fusion mediated by this F co-expressed with HA is less than for the parental F; even at the highest temperatures used in the assay, 50% fusion was not attained for the L234F F when paired with the parental HN. In order to achieve 50% fusion, this F must be paired with a low-neuraminidase or high-avidity variant HN³⁵ (Supplementary Fig. 2). Thus, while both F proteins are more readily triggered to exit the pre-fusion state by heat, the L234F mutation appears to interfere with execution of fusion.

In the presence of co-expressed parental HN to activate F (Fig. 2E), A194T F is more activated and executes more fusion than parental F, but L234F F is much less readily activated than parental F. For parental F with HN, 50% fusion occurs at about 24 °C, shown by the blue dotted line. A194T F with HN shows 50% fusion at about 20 °C, shown by the red dotted line, indicating more efficient activation of F. L234F F shows less HN mediated fusion and fails to attain 50% fusion with HN even at 37 °C, but is not inactive, simply less responsive to activation by their HN, indicative of the perturbation of a critical HN-F interaction site.

To understand how the L234F F can function in the context of an HN-F complex from an infectious virus, we examined the effect of the HN present in the EV2 variant virus—HN bearing H552Q/I243V—on HN-promoted fusion mediated by the L234F F. We have previously shown that the H552Q mutation in HN confers enhanced fusion promotion. This residue comprises HN’s sialic acid binding site II, that we have shown to be critical for activating HPIV3 F; alteration of Site II affects fusion, infectivity, and viral fitness in vivo^{18,21,26,30,31,36,37}. The I243V mutation is in the globular head of HN at the dimer interface underneath H552Q (Fig. 1B), and its function has not been previously characterized. Using a cell-cell fusion assay where fusion of cells co-expressing HN/F or HA/F pairs with receptor-bearing cells is quantitated by beta-galactosidase complementation, we determined the effect of this mutated HN on fusion mediated by L234F F (Fig. 3A). The H552Q/I243V is highly fusion-promoting and restores fusion by L234F F. The I243V mutation in HN has an enhanced fusion promotion effect and is additive in its fusion promotion effect with the H552Q dimer interface mutation (Fig. 3B). The HN with H552Q and I243V mutations has an enhanced fusion promotion effect on all three F proteins (Fig. 3A).

Fig. 3. HN escape mutations enhance fusion promotion of F.

A Cell-to-cell fusion measured by beta-galactosidase complementation with cells expressing HN or HA and parental F, A194T F, or L234F. Fusion values are normalized to max value in each experiment. B Cell-to-cell fusion measured by beta-galactosidase complementation with cells expressing parental F and parental HN, I243V, H552Q, or I243V/H552Q. Data are means ± SE from three separate biological replicates for (A, B).

Cryo-ET structure of escape variant HN-F complexes reveal altered relationships between HN and F

To examine the structural consequences of the Ab escape mutations on the HN/F complex on the viral surface, we compared the cryo-ET structures of parental HPIV3, EV1, and EV2 HN/F complexes. Recombinant viruses were generated and confirmed by deep sequencing prior to these experiments. We used antibody affinity capture to purify each viral sample directly on the grid without high speed centrifugation, using a procedure that avoids the disruption of surface glycoproteins associated with other purification methods²⁵. We used our previous processing pipeline for subtomogram averaging of viral surfaces to obtain medium resolution of the HN/F complex for each virus (Supplementary Table 1 and Supplementary Fig. 3)²⁶. The cryo-EM model of the soluble F (Protein Data Bank (PDB) ID: 6MJZ) was fit into the final density for each variant final subtomogram average (Fig. 4A, B, D, E and Supplementary Movie 1). The models of individual HN protomer heads, derived from the crystal structure of the soluble portion of the HN dimer (PDB ID: 4MZA), were also fit into the final density of each subtomogram average (Fig. 4A, B, D, E). The Fourier shell correlation (FSC) was between 12 and 13 Å for the final subtomogram averages (Supplementary Fig. 5). At this level of resolution, it can be observed that remarkably, the HN-F interaction at the apex of F has been altered in the fusion complexes of both variants. The globular head of HN that—in the parental complex—sits directly atop the apex of F, capping it, is shifted off the F apices. The HN dimer heads extend 175 Å above the viral membrane surface as the heads rest above F in the parental HN/F complex (Fig. 4A, D and Supplementary Fig. 5), but only 160 Å above the viral membrane surface for both EV1 and EV2 (Fig. 4B, E), where the HN head is not capping the F apex. The F apex is exposed in both viruses.

Fig. 4. HN/F fusion complex subtomogram averages for EV1 and EV2 variants.

A Side views of the field strain viral HN-F fusion complex with prefusion F [Protein Data Bank (PDB) ID: 6MJZ] and HN (PDB ID: 4MZA) models fit into the density. B Side views of the EV1 HN- F reconstruction with prefusion F and HN models fit into the density. Insets show the interaction between the HN and F models for field strain and EV1 viral fusion complexes. C Overlay of the final subtomogram averages for field strain (blue) and EV1 (orange) fusion complexes. D 90 degrees rotated view (with respect to A) of the field strain viral fusion complex with prefusion F and HN models fit into the densities. E Side views of the EV2 HN-F reconstruction with prefusion F and HN models fit into the density. Insets show the interaction between the HN and F models for field strain and EV2 viral fusion complexes. F Overlay of the subtomogram averages for field strain (blue) and EV2 (green) fusion complexes. G Overlay of EV2 HN/F complex (green) with L234F HN/F complex (blue). H The buried surface area (residues of interaction with HN) on F for the field strain (blue), EV1 (orange), and EV2 (green). I Buried surface area (residues of interaction with HN) on F, overlaid with the model of PIA174 Fab to show overlap between Fab and HN interacting residues. Scale bars (A–F): 5 nm.

For both EV1 and EV2, the 386-392 loop in HN interacts with F differently to this interaction on the parental virus, as suggested by the fitted domains, although at this resolution the loop cannot be resolved directly. The domains were fitted with a map to model FSC between 17 and 25 Å for the final subtomogram averages (Supplementary Fig. 5). The structure of the HN-F complex in EV1 shows that, when viewed from the top, the HN protomer proximal to F is shifted forward off the apex compared to the parental HN/F complex, resulting in an interaction between HN and F over a smaller surface area, suggesting that it is a less stable interaction (Fig. 4B inset). The only interaction between the HN protomer and F is the 386-392 loop of HN that loosely interacts with the 56-64 loop in the F2 region of F, near the disulfide bond that connects F1 and F2 (Fig. 4H, I and Supplementary Fig. 6B). This likely instability permitted minor conformations of the HN and F complex resulting in a final density in Dynamo with a resolution of 15.8 Å (Supplementary Fig. 4A, B). To further investigate these minor conformations, subtomograms from Dynamo were further refined and classified in Relion to separate out the conformations (Supplementary Fig. 4C). We then proceeded to further refine two separately grouped classes (Supplementary Fig. 4E). Separately, 3D classification in Relion for EV2 did not reveal any differences in the orientation of HN at the interaction site between HN and F (Supplementary Fig. 4D). In contrast, flexibility analysis of EV1’s HN-F complex revealed changes in the orientation of HN with respect to F (Supplementary Fig. 4F).

The EV2 HN-F structure is different from that of the EV1 complex, in that, when viewed from the top, the HN protomer proximal to F rotates and shifts laterally off the apex making extensive interactions with the F-HRC region adjacent to the central helical core (Fig. 4E inset). Based on our model fitting, the HN 386-392 loop is no longer in contact with the F protein (Supplementary Fig. 6C) but the buried surface area between the HN and F increases from 490 Å in the parental complex to 608 Å in the EV2 complex (Fig. 4H, I). While these buried surface area differences reveal distinctions between the interfaces in the wt and variants, more accurate surface area calculations will require higher resolution data and models. We by necessity interpret the reconstructions at this resolution by docking available high-resolution structures of the F pre-fusion trimer and HN dimer heads despite the fact that the cryo-EM structure of the stabilized pre-fusion F²⁹ differs from our authentic virus surface F, and this difference allows for some inaccuracy. What we derive from the docking is comparison of the interaction between HN and F for the parental, EV1, and EV2 complexes. While not identifying specific residues of interaction at contact sites, the buried surface areas show the different relationships between HN and F for the parental, EV1, and EV2 complexes. The structure of the fusion complex for a recombinant virus with F bearing only the L234F change was solved to examine whether the HN mutations in EV2 contribute to the displacement of HN from the apex of F, or whether this displacement is attributable to the L234F mutation in F alone. This complex that differed from the parental virus only in the L234F mutation in F showed the same displacement of the HN head from the F apex as EV2 (Fig. 4G).

Escape mutations exact a fitness cost for the virus in human airway

To identify fitness consequences of the mutations in the viruses that emerged during selection for escape from PIA174, we evaluated infection in human airway epithelia (HAE) (Fig. 5). We have shown that HAE cultures represent an authentic model of the human lung, reflecting the cell environment and selective pressure of the natural tissue^37,38. These airway cultures have been validated for evaluating features of infection and HPIV3 fitness^38,39. We compared the growth and evolution of the variants to the parental strain. HAE cultures at an air-liquid interface were infected with EV1 or EV2 viruses or the parental virus and viral titer measured daily for 7 days (Fig. 5A). Compared with the parental virus, EV1 showed significantly reduced production of infectious viral particles (red line) which only reached the parental level (blue line) by day 7. For EV2 (green line) there was an initial delay in production of infectious virus but beginning at day 3 the growth curve dramatically changed and by day 4 overtook the titer of the parental virus.

Fig. 5. Fitness cost of escape mutations.

A Human Airway Epithelial (HAE) cells were infected with EV1, EV2, or Parental HPIV3 and apical washes were titered daily up to 7 days post infection. Allele frequencies (points) and titers measured in plaque forming units (PFU; gray bars) of (B) EV1 at 0, 7 days post infection and (C) EV2 at 0, 2, 4, 5, and 7 days post infection. D Location of H552Q on the field strain HN model, which remained at 100% allele frequency in all wells of EV1, and compensatory mutations in EV2 wells that increased in allele frequency during the 7 days of infection for HN. Data are means ± SE from three separate biological replicates for (A).

Alterations in HN or F that confer a disadvantage in the human airway will lead to evolution. We have shown that the HN-F fusion complexes of HPIV3 field strains from humans are not under selective pressure to evolve during growth in HAE^36–39. By growing the variant viruses in HAE, we determine whether these mutations place the virus at a disadvantage and under selective pressure to evolve in the HAE. Remarkably, deep sequencing of the day 7 viruses revealed that while EV1 retained its escape mutations—which resulted in slowed growth that caught up—EV2 had entirely reverted to the parental sequence under the selection pressure of growth in human airway (Fig. 5B, C). This rapid response to airway conditions revealed the severe fitness cost of the L234F change in the background of EV2. EV1, despite having a fitness cost, is not significantly impaired in essential functions as we had shown in Figs. 1 and 2 and did not exhibit a disadvantage significant enough to spur evolution. Figure 5C shows the daily accumulation of changes in EV2 until by day 7 the reversion is complete. In each well, a single mutation arose in HN (R242K for well 1 or T230A for well 2) and rose in frequency over the days, as the EV2 mutations decreased to complete loss. In contrast, Fig. 5B shows that EV1 retained all its mutations including the F-A194T apex mutation. The I243V mutation (dark blue spheres) that arose in EV2 during viral evolution is adjacent to both R242K (brown spheres) and T230A (cyan spheres) (Fig. 5D) and is located in the dimer interface distal from the H552 residue (Fig. 5D, yellow spheres).

Discussion

Respiratory viruses affect millions of children globally each year, yet for most we have no vaccine or drug treatment. Acute respiratory infection accounts for nearly 20% of childhood deaths worldwide, or 2–3 million children each year. Most cases of childhood croup, bronchiolitis, and pneumonia, the most common manifestations of acute respiratory infections in infants, are caused by two viruses—HPIV3 and respiratory syncytial virus (RSV)^1,2. HPIV3 is responsible for 30–40% of all acute respiratory tract infections in infants and children³. The toll on immune compromised patients is highly significant with paramyxoviruses including HPIV3 responsible, for example, for a large percentage of the infectious complications after transplantation³. This vulnerable group is growing yearly as treatments for autoimmune and other serious illnesses that have a side effect of immune suppression become more readily available.

Our recent data offer the first detailed view of the HPIV3 HN-F fusion complex—not individual isolated proteins, but the complex as it actually exists and works—on the surface of an intact virus^25,40. Our analysis of the HN-F fusion complex leads us to propose a paradigm for how the complex, viewed in its entirety as a fusion machine, promotes viral entry. The receptor binding protein is complexed with F and activates the F-mediated fusion process (^14–18, reviewed in ref. ¹⁹), a finding that has been confirmed across a broad swath of paramyxoviruses with possible exceptions (e.g., the pneumoviruses respiratory syncytial virus-RSV^41–43 and human metapneumovirus-hMPV^44–46)^47,48. Each paramyxovirus solves the problem of entry using variations on this theme. By imaging interactions between viral glycoproteins at high resolution in situ as the interacting proteins progress through their roles in the cell it is now possible to directly investigate entry mechanisms^25,26. Here we build on the structural understanding of the authentic virus by characterizing viruses that have adapted to an external stimulus (PIA174 nAb) and correlating these structural adaptations with function.

Neutralizing mAbs against respiratory viruses have shown significant promise for prophylaxis as well as therapeutics. mAbs against RSV, Ebola virus, influenza virus, HIV-1, and SARS-CoV-2 have been evaluated clinically, and several are authorized for administration to patients for prophylaxis or therapy^5–12. However, resistance presents a critical challenge for respiratory nAbs—and specifically for HPIV3—given the evidence of antigenic drift in the field⁴⁹, the viral evolution within immunocompromised human hosts³¹, and the finding that a neutralizing antibody directed at stabilized F readily elicits resistant variant viruses²⁶. In this paper we show that the HN/F complex in its pre-fusion state on the viral surface has an inherent adaptability that will allow it to evade antibodies that target exposed epitopes. The opportunity to not only characterize progression through steps of entry of the field strain virus, but also to view structural changes as the virus evolves to evade immunity or inhibitors, will allow us to understand this dynamic and adaptable viral fusion mechanism.

Previous structural or computational analyses of the individual glycoproteins did not predict the actual configuration of the glycoproteins in the biologically relevant complex on the virus surface. Examining intact viral surface HN/F complexes was crucial since the structures of the complexes shown here differ from their soluble counterparts. We cannot rule out the possibility that the viral surface may contain HN/F complex heterogeneity including instances of uncoupled HN and F. In HN/F complexes, we observe that HN can maintain contact with F by binding to F just adjacent to the apex. These observations may provide new targets for inhibitors. Crystal or single particle cryo EM structures of HN and F isolated from their biological context previously suggested hypotheses about how the HN-F complex might mediate viral entry as a unit^18,27,29, but determining the biological validity of these mechanistic insights required structural data in an authentic virus-host dynamic. This information about the in situ complex is also key to identifying therapeutic targets to reduce the problem of resistance.

An anti-HPIV3 F neutralizing mAb, PIA174, that targets the apex of pre-fusion HPIV3 F blocks the structural transition of F to its activated intermediate and perturbs the HN/F complex^6,9, stabilizing the prefusion state of F and thereby inhibiting F activation⁶. Viral evolution experiments under the selective pressure of PIA174 readily led to two EV viruses. One of these viruses bears a mutation at the precise site on the F apex where PIA174 binds and that we predict to be an important site of HN/F interaction, A194T⁶. This mutation alters both PIA174Ab binding and HN’s activation of F, leading F to be more readily activated. In contrast, a second PIA174 escape mutant bears a mutation adjacent to the central helix of the F trimer, distant from the trimer apex, L234F. This mutation destabilizes the F trimer interface thereby perturbing the apex antibody epitope and rendering the viruses resistant to nAb inhibition.

Remarkably, the structures of the pre-receptor engaged HN/F complexes of both escape mutants show that in both variant complexes the HN protomer head is shifted away from the apex of F in comparison to the position of HN in the parental HN/F complex, uncapping the apex of F. The relationship between HN and F in the complex changes significantly in both EVs. In both variants the F apex is affected and HN-F interaction is altered, and the altered structure correlates with reduced stability of F in its pre-fusion state, suggesting that the F-apex may be critical to maintaining the pre-fusion state and supporting the proposed mechanism where the HN loop interacts with the apex of F to stabilize it, mirroring the function of PIA174.

While the EVs show a fitness cost and reduced growth in human airway tissue, in both cases they are infectious and viable viruses. We previously had the opportunity to study within-host evolution of HPIV3 in two immune compromised patients with long term persistent HPIV3 infection³¹. The A194T mutation at the apex of EV’s F emerged in these patients, confirming the viability of viruses bearing this mutation in F and highlighting the availability of this structural solution that allows viruses to adapt to stressed conditions.

The two-protein entry complex of paramyxoviruses provides a wealth of options to these viruses for adjusting their entry processes and adapting to conditions of new environments. We have shown that during adaptation to monolayer culture, the HPIV3 HN-F complex rapidly evolves to more vigorous fusion, more avid receptor binding, and more readily activated F³⁷. During acute lung infection, in contrast, the complex rapidly evolves to stabilize the pre-fusion complex and decrease receptor avidity and fusion^21,36,38. During long-term infection, the HN-F complex modulates to favor fusogenicity and cell-to-cell spread³¹. Here we show that in adapting to treatment with a nAb that suppresses F activation, the complex evolves to become more readily triggered while evading nAb binding.

In order to design nAbs with a higher barrier to resistance, that would be therefore more useful clinically, it is necessary to incorporate information about nAb mechanisms, the potential routes of escape, and use this information to inform which antigen designs are least likely to elicit resistance. We propose that the intermediates of entry during virus-host target membrane interaction may provide vulnerable and genetically constrained antigens for nAbs. Future work will assess the potential for adaptability of the HN-F complex when structural intermediates are targeted. We anticipate that mechanisms we uncover for HPIV3 will be shared across multiple paramyxoviruses and suggest targets for nAbs.

These results uncover an entirely unanticipated HPIV3 nAb resistance mechanism where the HN-F fusion complex altered such that the HN no longer interacts with F’s apex and is shifted from its position in the complex. The result of eliminating the nAb binding site can be accomplished in at least two distinct ways: one with a mutation directly in the site (A194T) and the other with a distant destabilizing mutation (L234F), both accomplishing the same end by shifting HN from its position atop F and uncapping the apex. It is entirely unexpected that this complex on viral surfaces would exhibit this degree of adaptability while retaining critical functions on the virus. This information will inform the search for nAb targets that circumvent the remarkable adaptability of this complex. This comprehensive structure-function study provides the only view of the adaptability of the complex formed by a paramyxovirus receptor binding and fusion protein on the virus surface, and of its wealth of strategies for evading nAbs. These mechanistic findings have broad implications for the design of nAbs that target paramyxoviruses. Understanding features of the HN-F complex that lead to resistance also enables the design of a new generation of paramyxovirus nAbs less likely to elicit resistant viruses.

Methods

Virus growth and purification

Recombinant viruses were generated by reverse genetics using an HPIV3 virus sequence (modified from^50,51, a gift from Ursula Buchholz and Peter Collins, NIAID) containing an mCherry cassette between genes P and M. Resulting viruses were propagated using Vero cells (ATCC). Viruses were titered by limiting dilution infection of Vero, and infected cells were quantified using a Cytation 5 imager (Biotek). All recombinant viruses were sequenced using mNGS⁵² prior to experimental use.

Constructs

Plasmids encoding HPIV3 HN and F were generated through site-directed mutagenesis of a previously constructed pCAGGS mammalian expression vector⁵³ and sequenced via Sanger sequencing prior to experimental use. Transfections with plasmids were performed in HEK293T cells (ATCC) using Lipofectamine 2000 per the manufacturer’s specifications (Invitrogen). Consensus sequence of the laboratory-adapted strain of HPIV3 (Wash/47885/57) used throughout the study was obtained from the NIH (HA-1, NIH no. 47885, catalog no. V323-002-020⁵⁴).

Cells

HEK293T cells (ATCC; CRL-3216) for transfections were grown in DMEM (Gibco) supplemented with 10% FBS and 1% penicillin-streptomycin (Gibco) at 37 °C and 5% CO₂. Vero cells (ATCC; CCL-81) were grown in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin (Gibco) at 37 °C and 5% CO₂. All cells tested negative for mycoplasma presence using MycoAlert Mycoplasma Detection kit per the manufacturer’s specifications (Lonza).

Chemicals and antibodies

Zanamivir (Acme Bioscience) was dissolved in Opti-MEM at a concentration of 50 mM and stored at −80 °C. Monoclonal anti-HPIV3 HN antibodies were custom elicited in rats (Aldevron) using eGFP-HN complementary DNA, diluted in Dulbecco’s PBS (DPBS) to 100 μg/ml, and kept at 4 °C. PIA174 antibody fragment (Fab) and full antibody (Ab) were purchased from Creative Biolabs. The 3 × 1 antibody was obtained from Jim Boonyaratanakornkit (Fred Huchinson Cancer Center). VG-PEG4-chol fusion inhibitory peptide was produced by standard Fmoc-solid phase methods, and the cholesterol moiety was attached using displacement of an α-bromoamide³⁸. Bromoacetyl–PEG4-cholesterol was custom synthesized by Charnwood Molecular (UK). VG-PEG4-chol (5 mM in DMSO) was kept at −20 °C.

Viral evolution strategy

Vero cells (ATCC) in a 96-well plate were infected with 200 plaque-forming units (PFUs) per well of HPIV3 expressing mCherry (modified from^50,51, a gift from Ursula Buchholz and Peter Collins, NIAID) in Opti-MEM (Gibco, 31985070) supplemented with 1% penicillin-streptomycin. After 2 h, the infection medium was replaced with decreasing concentrations of PIA174 antibody (10 μg/ml; 1:2 dilutions down to 0.04 μg/ml). Because significant viral spread was observed in all wells after 2 days, this virus was passaged onto new cells with four times higher concentration of antibody (40 μg/ml; 1:2 dilutions to 0.16 μg/ml). Infection was monitored, and antibody was replaced every 2 days. Virus was collected and frozen at −80 °C on days 1, 3, 4, 6, and 7 after infection. After day 4 (EV2) and day 7 (EV1), putative EVs (observed exponential growth in wells at the highest concentration of antibody) were propagated in Vero cells in the presence of antibody (4 μg/ml). These viruses, and control viruses that were passaged alongside variant viruses without antibody, were sequenced.

Viral entry inhibition assay

Vero cells (ATCC) in a 96-well plate were infected with 500 PFUs of parental, EV1, or EV2 virus in the presence of serial dilutions of PIA174 FAb (200 µg/mL; 1:2 dilutions to 0.2 µg/mL) and incubated for 2 h at 37 °C. Then, the antibody was removed, and the cells overlaid with 0.5% carboxymethyl cellulose²⁶. After 18 h, cells were imaged on a Cytation 5 imager (Biotek), and fluorescently labeled cells were counted. Inhibition of entry was quantified by comparing the number of infected cells at different concentrations of antibody to the number of infected cells in the absence of antibody for each virus.

Antibody binding assay

HEK 293 T cells were transiently transfected with HPIV3 F, HPIV3 F bearing A194T, or HPIV3 F bearing L234F F with or without additional stabilizing mutations (Q162C, L168C, I213C, G230C, A463V, I474Y) based on²⁹ for 18 h at 32 °C. The cells were then washed and treated with PIA174 FAb or 3 × 1 mAb at a series of concentrations (1 µg/mL; 1:2 dilutions to 0.016 µg/mL) for 1 h at 4 °C. The cells were then washed and treated with secondary anti-human antibody conjugated with DyLight 594 (0.5 µg/mL) (Abcam) for 30 min at 4 °C. The cells were fixed with 4% PFA and treated with 1 µg/mL DAPI (Thermo). Cells were imaged on a Cytation 5 imager (Biotek) and intensity per well was calculated. Concurrently, cell surface expression of the F proteins was determined.

F stability assay

HEK 293T cells were transiently transfected with HPIV3 F, HPIV3 F bearing A194T, or HPIV3 F bearing L234F for 18 h at 32 °C. 1 µg/mL of PIA174 FAb was added to half of each plate and the cells were equilibrated to 4 °C for 15 min. Cells were either kept at 4 °C or transferred to 55 °C for 5, 15, 30, or 60 min. After incubation, the cells were washed, and 1 µg/mL of 3 × 1 mAb was added to all wells for 60 min on ice to detect loss of the pre-fusion state. After washing, cells were stained with anti-human antibody conjugated with DyLight 594 (0.5 µg/mL)(Abcam) for 60 min at 4 °C. Cells were imaged on a Cytation 5 (BioTek) and fluorescent intensity was calculated. The data were normalized to the fluorescence at 4 °C for each condition.

β-Gal complementation-based fusion assay to assess fusion promotion by HN

We previously adapted a fusion assay based on α complementation of β-galactosidase^22,55. In this assay, receptor-bearing cells expressing the omega peptide of β-Gal were mixed with cells coexpressing designated envelope glycoproteins and the α peptide of β-Gal, and cell fusion leads to complementation. Fusion was stopped by lysing the cells. Substrate was added (Galacton-Star substrate; Applied Biosystems, T1012), and luminescence was read after 1 h at 500 ms on a Tecan M1000 Pro. Relative fusion was calculated by normalizing all luminescence values to the maximum value in the experiment.

Conformational change of unprocessed F (F₀)

HEK 293T cells were transiently transfected with HPIV3 HN bearing D216R and uncleaved HPIV3 F, uncleaved HPIV3 F bearing A194T, uncleaved HPIV3 F bearing L234F, or empty vector (pCAGGs) for 18 h at 32 °C. Cells were treated overnight with 25 mU/well of exogenous neuraminidase (Sigma Aldrich) to deplete sialic acid receptors and then washed and incubated with 1% RBC suspensions (pH 7.5) for 30 min at 4 °C. After the samples were washed to remove unbound RBCs, they were treated with 2 µM of lipid conjugated HRC derived peptide³³ in CO₂ independent medium (Gibco) and brought to 37 °C for 0, 5, 15, 30, or 60 min. The plates were rocked, and the liquid phase collected in V-bottom plates for measurement of released RBCs. The cells were incubated with 10 mM zanamivir in CO₂ independent medium to release RBCs that were attached via HN receptor engagement. The liquid phase collected in V-bottom plates for measurement of reversibly bound RBCs. Plates were spun down and pelleted RBCs were lysed in milli-Q water and transferred to a flat bottom 96-well plate for quantitation. The cells were then incubated with RBC lysis solution (ammonium-chloride-potassium lysis buffer; Thermo Fisher Scientific, A1049201), where the lysis of unfused RBCs removes cells that were attached only via F (bound to inhibitory lipopeptide). The liquid phase collected in flat-bottom plates for measurement of irreversibly bound RBCs. The cells were then lysed in 200 μl of dodecyl maltoside HEPES (DH) buffer [5 mM Hepes, 10 mM NaCl, and dodecyl maltoside (0.5 mg/ml)] 1:10 in PBS and transferred to flat-bottom 96-well plates for quantification of fused RBCs. Hemoglobin absorbance for the above four compartments was determined by measuring absorbance at 405 nm on a Tecan M1000 Pro. Percentage of irreversibly bound RBCs was calculated by dividing irreversibly bound absorbance by the sum of released, reversibly bound, irreversibly bound, and fused RBCs. Initial slopes were calculated by taking the first derivatives of the percent irreversibly bound RBCs over time at the halfway point between time points 0 and 1 hour.

Measurement of F-activation and fusion between RBCs and envelope glycoprotein-expressing cells

HEK 293T cells were transiently transfected with HPIV3 HN or uncleaved influenza HA and HPIV3 F, HPIV3 F bearing A194T, and HPIV3 F bearing L234F, for 18 h at 32 °C. Cells were treated overnight with 25 mU/well of exogenous neuraminidase (Sigma Aldrich) to deplete sialic acid receptors and then washed and incubated with 1% RBC suspensions (pH 7.5) for 30 min at 4 °C. After the samples were washed to remove unbound RBCs, they were kept at 4 °C or brought to 17, 27, 32, or 37 °C for 60 min. The plates were rocked, and the liquid phase collected in V-bottom plates for measurement of released RBCs. Plates were spun down and pelleted RBCs were lysed in milli-Q water and transferred to a flat bottom 96-well plate for quantitation. The cells were then incubated with RBC lysis solution (ammonium-chloride-potassium lysis buffer; Thermo Fisher Scientific, A1049201), where the lysis of unfused RBCs removes unfused cells. The liquid phase collected in flat-bottom plates for measurement of irreversibly bound RBCs. The cells were then lysed in 200 μl of dodecyl maltoside HEPES (DH) buffer [5 mM Hepes, 10 mM NaCl, and dodecyl maltoside (0.5 mg/ml)] 1:10 in PBS and transferred to flat-bottom 96-well plates for quantification of fused RBCs. Hemoglobin absorbance for the above three compartments was determined by measuring absorbance at 405 nm on a Tecan M1000 Pro.

Viral evolution in HAE

The HAE EpiAirway AIR-100 system (MatTek Corporation) contains cultured human-derived tracheo/bronchial epithelium that forms pseudostratified, differentiated mucociliary epithelium recapitulating in vivo human tissue³⁹. Upon receipt from the manufacturer, HAE cultures were transferred to provided 6-well plates containing 2 mL of AIR-100-ASY assay medium (MatTek Corporation) per well with the apical surface remaining exposed to air and incubated at 37 °C in 5% CO₂ overnight prior to infection. HAE cultures were infected with 500 PFU per well of EV1, EV2, and parental virus at the apical surface for 3 h at 37 °C, followed by inoculum removal and incubation at 37 °C for the remainder of the experiment. Maintenance medium (MatTek Corporation) was changed every other day with 2 mL medium via the basolateral surface. Viruses were harvested by adding 200 μL of provided 1× PBS containing magnesium and calcium (MatTek Corporation) per well via the apical surface and incubated for 30 min at 37 °C. Supernatant was subsequently collected and viral titers were determined by limiting dilution infection of Vero cells (ATCC), with infected cells quantified using Cytation 5 (BioTek). The RNA genomes of these viruses, and control viruses that were passaged without antibody alongside variant viruses, were sequenced.

Sequencing library generation and analysis

Shotgun RNA-sequencing libraries were made by using Turbo DNAse (Thermo AM1907) to deplete host genomic DNA, Superscript IV (Thermo 18090200) primed with random hexamers and Sequenase 2.0 (Thermo 70775Z1000UN) to create double-stranded cDNA, and the Illumina DNA Prep (S) kit to fragment DNA and add adapters with 14 cycles of PCR (Illumina 20025520)⁵². The libraries were sequenced 2 × 150 bp on the NextSeq 2000. Reads were adapter- and Q20 quality-trimmed using fastp⁵⁶. Variants for all samples were called using the reference-based options in LAVA. Briefly, shotgun RNA sequencing reads for the viral genome from the virus that escaped during the viral evolution experiment were aligned to the reference sequence for the HPIV3 expressing mCherry (GenBank accession OP821798) using bwa-mem v0.7.17-r1188 (https://arxiv.org/abs/1303.3997), and variant allele frequencies were extracted using samtools v1.7⁵⁷ and annotated via VarScan v2.3⁵⁸. Sequencing reads are available in NCBI BioProject PRJNA1083633.

Cell surface biotinylation

HEK 293T cells were transiently transfected with viral glycoprotein variants. Cells were then incubated at 4 °C with 3.3 mM NHS-S-S-dPEG₄-biotin (Quanta Biodesign 10194) in DPBS (Gibco 4287080) for 1 h before lysing with DH Buffer [50 mM HEPES (Gibco Cat#15630080), 100 mM NaCl, 5 mg/mL dodecyl maltoside (Thermo Scientific 89903) in Milli-Q Water] supplemented with complete Protease Inhibitor Cocktail (1 tablet/50 mL; Roche 11836145001). Biotinylated proteins were pulled down with Streptavidin Sepharose (Invitrogen 434341) for 16 h at 4 °C. Protein was eluted from the beads in reducing Laemmli SDS Sample Buffer (Boston BioProducts BP-110R), boiled for 10 min, and run on a 4–20% Novex Tris-Glycine Protein Gel (Invitrogen WXP42026BOX). The gel was transferred to nitrocellulose using iBlot quick transfer method (Invitrogen IB23001). The blot was blocked (Invitrogen WB7050), treated with Anti-HPIV3 F HRC (GenScript; Rabbit) and alkaline phosphatase-conjugated anti-rabbit secondary antibody (Invitrogen WB7105). The blot was then developed using NBT/BCIP Substrate (Invitrogen WP20001). Analysis of the blot was performed in imageJ.

Cryo-electron tomography preparation of HPIV3 HN-F complex

Lacey carbon gold grids, containing a continuous layer of thin carbon (Ted Pella), were plasma cleaned with Fischione M1070 Nanoclean on 70% power for 20 s with a 25% Oxygen, 75% Argon gas mixture. 8 μl drops containing 100 μg/ml of the anti-HPIV3 HN antibody were incubated on the grids for 10 min, and the grids were washed with DPBS to remove unadsorbed antibodies. Next, grids were blotted and placed face-down in the cell supernatant fluid, containing in a 6-well plate. Plates were incubated for 30 min at 4 °C with rocking. Grids were then washed in a DPBS solution (Gibco) containing 5 nm gold nanoparticles (Sigma Aldrich) added at a 10% concentration of the final solution. Grids were then plunge frozen in liquid ethane, using a Vitrobot (Mark IV; Thermo Fisher Scientific Co.). For a negative control, we applied an antibody specific for measles H to the grids, and we did not observe viral particles on these grids.

Cryo-electron tomography data collection and image processing

Vitrified grids were imaged using a Titan Halo 300 kV transmission electron microscope (Thermo Fisher Scientific Co.), equipped with a Gatan K3 direct detector and no energy filter. Images were captured at a magnification of 18000x, giving a pixel size of 1.71 Å at the specimen level. Images were acquired with SerialEM software⁵⁹ with a 3.5–5.0 μm defocus and a bidirectional tilt-series of 3° steps starting from −9° to 51° and then −12° to −51°. Each tilt had a dose of 2.8 e-/Ų resulting in a total dose of ~100 e-/Ų for the tilt series. All micrograph movies were aligned using WarpEM⁶⁰. Tilt series were aligned with Etomo⁶¹ and then the aligned parameters were used to reconstruct the CTF corrected tomograms in WarpEM. 12 tomograms for the EV1 HN-F complex, 12 tomograms for the EV2 HN-F complex, and 16 tomograms for the L234F HN-F complex were selected for subtomogram averaging.

Subtomogram averaging of the HN-F complex

The subtomogram averaging process was performed using the Dynamo software package^62,63. Sub-volumes of the viral particle surfaces (438) ų were extracted from 4X binned tomograms. The first round of reference free alignment was followed by centering and re-cropping to (438) ų on a density corresponding to an HN and F complex. A second round of 6 iterations with a 360⁰ azimuth range applied to these centered particles was performed with an initial reference of an HN (PDBID:4MZA) and F (PDBID:6MJZ) model prefiltered to 50 Å and a cylindrical mask including the membrane. In addition, an initial model missing one HN and one F protomer was prefiltered to 50 Å and used as a reference in a separate run. This separate run converged to the same density as the initial reference of HN and F (Sup. Figure 4). Iterations in rounds 2–5 had an initial reference that was low pass filtered to 30 Å and came from the previous rounds average. Sub-volumes in rounds 3-5 (18 iterations) were aligned in a process that involved decreasing the azimuth range from 120° to 15° in the presence of cylindrical mask encompassing only the HN and F density. Overlapping particles separated by less than 60 Å were removed in the 5th round and particles with the highest cross correlation were selected for re-extraction of subtomograms. L234F and EV2 HN-F complexes were unbinned using Warp with a box size of (410) ų. Subvolumes were then subjected to 2 more rounds of refinement with azimuth ranges of 15° and 5° with a tight mask surrounding HN and F generated with the Relion mask creation tool^64,65. For the EV1 HN-F complexes, 2X binned subtomograms were re-extracted using Warp with a box size of (430) ų. The initial reference in Relion was low pass filtered to 20 Å and the subtomograms were subjected to 2 more rounds of refinement and classification in Relion with a tight mask surrounding HN and F. Prior to classification, multi-body refinement in Relion was used to determine dominant motions of the HN dimer. Resolution of the resulting maps of EV1, L234F, or EV2 HN-F complexes was estimated by Fourier shell correlation with a 0.143 or 0.5 threshold using 3DFSC⁶⁶ either unmasked or with a mask that extended from the HN/F density.

Model fitting and image analysis

All cryo-EM movie images were visualized using ImageJ, IMOD⁶⁷, ChimeraX^68,69, and Chimera⁷⁰. Fourier Shell Correlations performed by 3DFSC⁶⁶, Mtriage⁷¹, and Resmap⁷² were used to validate the final resolution. The number of particles included in the analysis and the strategy utilized are summarized in Supplementary Table 1. Model fitting and model-to-map cross correlation fit were performed in Chimera⁷⁰ and ChimeraX^68,69 using the protomers from the crystal structure of HN (PDBID:4MZA) and the cryo-EM structure of F solved with pre-fusion F Fab antibody fragment (PDBID:6MJZ). Buried surface areas between HN and F were calculated in ChimeraX. Backplots of particles were generated using ArtiaX in ChimeraX⁷³

Material availability and licenses

Materials are available by MTA with the Trustees of Columbia University, NYC. Reagents are available from the corresponding authors under a material agreement with Columbia University. This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/. This license does not apply to figures/photos/artwork or other content included in the article that is credited to a third party; obtain authorization from the rights holder before using such material.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization T.C.M., M.P., A.M.; Formal analysis T.C.M., M.P., A.M.; Funding acquisition T.C.M., M.P., A.M.; Investigation T.C.M., G.Z., E.B.S., K.G., E.H., K.S. Resources M.P., A.M. Supervision T.C.M., A.L.G., M.P., A.M. writing—original draft T.C.M., A.M.; final version: all co-authors provided feedback to the final draft.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

Final subtomogram averages have been deposited in the Electron Microscopy Data Bank (EMDB) with the accession codes: EMD-43914 for the EV1 HN and F complex; EMD-43913 for the EV2 HN and F complex; EMD-43912 for the L234F HN and F complex. The subtomogram average for the field strain can be found in the EMDB with the accession code EMD-27550. Pre-aligned tilts for all tomograms have been deposited in the Electron Microscopy Public Image Archive (EMPIAR) database with the accession code: EMPIAR-11936 for the EV1 HN and F complex, EMPIAR-11937 for the EV2 HN and F complex, and EMPIAR-11938 for the L234F HN and F complex. Raw data for Figs. 1,2,3, and 5, uncropped western blots, and plasmid sequences encoding HN, HA, and F have been deposited in the Dryad repository (10.5061/dryad.g1jwstqz6). Sequencing reads are available in NCBI BioProject PRJNA1083633. For domain fitting into our density maps, we used the cryo-EM model of the soluble F (Protein Data Bank (PDB) ID: 6MJZ; https://www.rcsb.org/structure/6MJZ) and the X-ray crystal structure of the HN dimer (PDB ID: 4MZA; https://www.rcsb.org/structure/4mza). All other relevant data are within the paper and its Supporting Information files.

Competing interests

T.C.M., M.P., and A.M. are listed as inventors on a provisional patent related to this work, “Subnanometer Structure of an Enveloped Virus Fusion Complex on Viral Surface Reveals New Entry Mechanisms.” The other authors do not have competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-024-53082-y.