Tailored design of protein nanoparticle scaffolds for multivalent presentation of viral glycoprotein antigens

1Multivalent antigen presentation, in which antigens are presented to the immune system in a repetitive array, has been demonstrated to increase the potency of humoral immune responses. This has been attributed to increased cross-linking of antigen-specific B cell receptors at the cell surface and modulation of immunogen trafficking to and within lymph nodes. An ongoing challenge has been to develop multimerization scaffolds capable of presenting complex oligomeric or engineered antigens, as these can be difficult to stably incorporate into non-protein-based nanomaterials (e.g. liposomes, polymers, transition metals and their oxides). Epitope accessibility, proper folding of the antigen, and stability are also important considerations in any strategy for antigen presentation. Several reports have utilized non-viral, naturally occurring protein scaffolds, such as self-assembling ferritin, lumazine synthase, or encapsulin nanoparticles, to present a variety of complex oligomeric or engineered antigens. These studies have illustrated the advantages of using self-assembling proteins as scaffolds for antigen presentation, including enhanced immunogenicity and seamless integration of antigen and scaffold through genetic fusion. More recently, computationally designed one- and two-component protein nanoparticles have been used to present complex oligomeric antigens, conferring additional advantages such as high stability, robust assembly, ease of production and purification, and increased potency upon immunization.

2The ability to predictively explore new structural space makes designed proteins attractive scaffolds for multivalent antigen presentation. In our previous work with computationally designed nanoparticle immunogens, the nanoparticles were generated from naturally occurring oligomeric proteins without initial consideration of geometric compatibility for antigen presentation. A more comprehensive solution would be to de novo design nanoparticles which present complex antigens of interest. For homo-oligomeric class I viral fusion proteins, a large group that includes many important vaccine antigens, a close geometric match between the C termini of the antigen and the N termini of a designed nanoparticle component would enable multivalent presentation without structural distortion near the glycoprotein base, and potentially allow for better retention of antigenic epitopes relevant to protection. More generally, precise control of antigen presentation geometry through de novo nanoparticle design would enable systematic investigation of the structural determinants of immunogenicity.

3We sought to develop a general computational method for de novo designing protein nanoparticles with geometries tailored to present antigens of interest, focusing specifically on the prefusion conformations of the trimeric viral glycoproteins HIV-1 Env (BG505 SOSIP), influenza hemagglutinin (H1 HA), and respiratory syncytial virus (RSV) F (DS-Cav1). To make the antigen-tailored nanoparticle design problem computationally tractable, we employed a two-step design approach (Figure 1). In the first step, we de novo designed antigen-tailored trimers, featuring N termini geometrically matched to the C termini of the viral glycoproteins. In the second step, we generated tetrahedral, octahedral, and icosahedral two-component nanoparticles by designing secondary interfaces between a designed trimer (fusion component) and a de novo homo-oligomer (assembly component). This design approach yielded nanoparticles tailored to present 4, 8, or 20 copies of the viral glycoproteins in defined geometries (Figure 1d). Sequences for all designed trimers and homo-oligomers, two-component nanoparticles, and antigen-fused components in this study can be found in Supplementary file 1A, B, and C, respectively. Details on each step of the design approach are described in the following sections.

4We chose to design our antigen-tailored trimers from monomeric repeat proteins composed of rigidly packed 20- to 50-residue tandem repeat units, as their high stability and tunable length (through variation of repeat number) are desirable properties for the design of protein-based nanomaterials. These structurally diverse alpha-helical repeat proteins featured three to six repeat modules and total lengths between 119 and 279 residues. They were docked into C3-symmetric trimers using our RPX docking method, which identifies configurations likely to accommodate favorable side chain packing at the de novo designed interface. To identify trimeric configurations with N termini compatible for fusion to the C termini of the three viral glycoproteins, docks with an RPX score above 5.0 were screened using the sic_axle protocol. Geometrically compatible docks (non-clashing termini separation distances of 15 Å or less) were subjected to full Rosetta C3-symmetric interface design and filtering (see Materials and Methods), and twenty-three designs were selected for experimental characterization (Figure 1—figure supplement 1).

5Synthetic genes encoding each of the designed trimers were expressed in E. coli and purified from lysates by Ni2+ immobilized metal affinity chromatography (Ni2+ IMAC) followed by size-exclusion chromatography (SEC). Twenty-two designs were found to express in the soluble fraction, and nine formed the intended trimeric oligomerization state as assessed by SEC in tandem with multi-angle light scattering (SEC-MALS; examples in Figure 2 top panel, second row; SEC-MALS chromatograms for the remaining designs are in Figure 2—figure supplement 2 and data in Figure 2—figure supplement 1—source data 1; SEC chromatograms for remaining designs with off-target retention volumes are in Figure 2—figure supplement 2). Four of the designs that were trimeric and expressed in high yield, 1na0C3_2, 3ltjC3_1v2, 3ltjC3_11, and HR04C3_5v2, were selected for solution small angle X-ray scattering (SAXS) experiments. The proteins exhibited scattering profiles very similar to those computed from the corresponding design models, suggesting similar supramolecular configuration (Figure 2 top panel, third row; metrics in Table 1 and Figure 2—source data 1). These four trimers were derived from three distinct designed helical repeat proteins from TPR, HEAT, or de novo topological families (1na0, 3ltj, and HR04, see Materials and Methods). Crystals were obtained for the two designs 1na0C3_2 and 3ltjC3_1v2. Structures were determined at resolutions of 2.6 and 2.3 Å, revealing a backbone root mean square deviation (r.m.s.d.) between the design model and structure of 1.4 and 0.8 Å, respectively (Figure 2—figure supplement 3, and Figure 2—figure supplement 3—source data 1, crystallization conditions, structure metrics, and structure-to-model comparisons are described in Materials and Methods). The structures confirmed in both cases that the designed proteins adopt the intended trimeric configurations, and that most of the atomic details at the de novo designed interfaces are recapitulated.

6As secondary assembly components were required to design our antigen-tailored nanoparticles, validated trimers were docked pairwise with de novo designed symmetric homo-oligomers to generate tetrahedral, octahedral, and icosahedral nanoparticle configurations using the TCdock program. To increase the probability of generating icosahedra which confer the highest valency among the targeted symmetries, three naturally occurring homopentamers were also included in the docking calculations (PDB IDs 2JFB, 2OBX, and 2B98). Analogously to the designed trimers, nanoparticle docks were scored and ranked using the RPX method to identify configurations likely to accommodate favorable side chain packing at a secondary de novo designed interface. High-ranking and non-redundant nanoparticle configurations featuring outward-facing N termini for antigen presentation were selected for Rosetta interface design. Fifty-three nanoparticle designs across all three targeted symmetries that exhibited the best interface metrics were selected for experimental characterization (see Materials and Methods). The nomenclature for the eleven tetrahedra, twenty-one octahedra, and twenty-one icosahedra indicate the symmetry of the nanoparticle (T, O, or I), the oligomeric state of the first component (A) and second component (B) used in each design, the letters “dn” reflecting the de novo nature of the input oligomers, and the rank by RPX score from the docking stage (e.g., “I53_dn5” indicates an icosahedral nanoparticle constructed from a pentameric and trimeric component, ranked 5th in RPX-scoring for the two input oligomers).

7Synthetic genes encoding each of the two-component nanoparticles were obtained with one of the components fused to a His6-tag, and the designs were purified using Ni2+ IMAC (see Materials and methods). Pairs of proteins at the expected molecular weights were found to co-elute by SDS-PAGE for twenty-four of the designs, consistent with spontaneous assembly of the nanoparticles followed by pulldown His6-tagged component (featured co-eluting designs are presented in Figure 2—figure supplement 4). SEC chromatograms revealed that nineteen designs did not form assemblies of the expected size or that the resulting assemblies were heterogeneous (Figure 2—figure supplement 5). Five designs comprising a panel of unique geometric configurations, T33_dn2, T33_dn5, T33_dn10, O43_dn18, and I53_dn5, ran as monodisperse particles of the predicted molecular mass by SEC-MALS and were further investigated by SAXS. The experimental solution scattering curves closely matched the scattering curves computed from the design models for all five designs (Figure 2, bottom panel and Figure 2—figure supplement 6; metrics in Table 1 and Figure 2—source data 1, bottom five designs).

8Due to its high valency and production yield, we selected the I53_dn5 nanoparticle to investigate the capacity of its two components to be separately produced and assembled in vitro. The two components of I53_dn5 were re-cloned, expressed, and separately purified (pentameric "I53_dn5A" with His6-tag and trimeric “I53_dn5B”). Nanoparticle assembly appeared to be complete within minutes after equimolar mixing (Figure 2—figure supplement 7). This capability is noteworthy as it enables production of each component independently, even from different host systems, which provides more flexibility in nanoparticle manufacturing. In vitro assembly also confers more control over nanoparticle assembly and composition, for example by assembling with a mixture of components fused to different antigens.

9The five SAXS-validated nanoparticles were structurally characterized using negative stain electron microscopy (NS-EM). 2,000–5000 particles were manually picked from the electron micrographs acquired for each designed nanoparticle and classified in 2D using the Iterative MSA/MRA algorithm (see Materials and Methods). 3D classification and refinement steps were performed in Relion/3.0. Analysis of the NS-EM data confirmed high sample homogeneity for all five nanoparticle designs as evident from the micrographs and 2D class-averages (Figure 3). While some free nanoparticle components were detected in the T33_dn5 sample, suggesting a certain propensity towards disassembly, analysis of the reconstructed 3D maps revealed that all five nanoparticles assemble as predicted by the design models, at least to the resolution limits of NS-EM.

10In order to obtain higher-resolution information, three designs, T33_dn10, O43_dn18, and I53_dn5, representing one nanoparticle from each targeted symmetry (T, O, I), were subjected to cryo-electron microscopy (cryo-EM). Cryo-EM data acquisition was performed as described in the Materials and Methods section and data acquisition statistics are displayed in Figure 4—source data 1. The data processing workflow is presented in Figure 4—figure supplement 1. Appropriate symmetry (T, O, and I for T33_dn10, O43_dn18, and I53_dn5, respectively) was applied during 3D classification and refinement and maps were post-processed in Relion/3.0. The final resolutions of the reconstructed maps for the T33_dn10, O43_dn18, and I53_dn5 nanoparticles were 3.9, 4.5, and 5.3 Å, respectively. Some structural heterogeneity was observed in the cryo-EM data, particularly in the case of I53_dn5. In 2D classification results we generated particle projection averages that range from spherical to ellipsoid shape (Figure 4—figure supplement 1c), indicating some degree of flexibility. There is less evidence of flexibility in T33_dn10 and O43_dn18, in agreement with the higher final map resolution for these nanoparticles.

11Nanoparticle design models were relaxed into the corresponding EM maps by applying multiple rounds of Rosetta relaxed refinement and manual refinement in Coot to generate the final structures. Refined model statistics are shown in Figure 4—source data 2. Reconstructed cryo-EM maps for T33_dn10, O43_dn18, and I53_dn5 and refined models are superimposed in Figure 4. Overall, the refined structures show excellent agreement with the corresponding Rosetta design models. Backbone r.m.s.d. values estimated for the asymmetric unit (consisting of a single subunit of component A and component B) were 0.65, 0.98, and 1.3 Å for T33_dn10, O43_dn18, and I53_dn5, respectively (Table 1).

12To explore the capability of the designed nanoparticles to present viral glycoproteins, we produced their trimeric fusion components genetically linked to a stabilized version of the BG505 SOSIP trimer. Synthetic genes for BG505 SOSIP fused to the N termini of T33_dn2A, T33_dn10A, and I53_dn5B (BG505 SOSIP–T33_dn2A, BG505 SOSIP–T33_dn10A, and BG505 SOSIP–I53_dn5B) were transfected into HEK293F cells. The secreted fusion proteins were then purified using a combination of immuno-affinity chromatography and SEC. The corresponding assembly component for each nanoparticle was produced recombinantly in E. coli, and in vitro assembly reactions were performed as equimolar mixtures of the two components overnight.

13Assembled nanoparticles were purified by SEC and analyzed by NS-EM to assess particle assembly and homogeneity. ~ 1000 particles were manually picked and used to perform 2D classification and 3D classification/refinement in Relion. Models for the BG505 SOSIP-displaying nanoparticles fit into their reconstructed 3D maps are displayed in Figure 5 (left). BG505 SOSIP trimers are clearly discernible in 2D class-averages and reconstructed 3D maps. However, the trimers appear less well-resolved than the corresponding nanoparticle core in the three reconstructions, likely due to the short flexible linkers between the BG505 SOSIP trimer and the fusion component. The self-assembling cores of the antigen-fused T33_dn2, T33_dn10, and I53_dn5 nanoparticles were very similar to the NS-EM maps of the unmodified nanoparticles (at least to the resolution limits of NS-EM), demonstrating that fusion of the BG505 SOSIP trimer did not induce any major structural changes to the underlying nanoparticle scaffolds. Free components were detected in raw EM micrographs of BG505 SOSIP–I53_dn5, indicating some degree of disassembly. This finding is supported by stability data reported in a parallel study, where BG505 SOSIP–I53_dn5 demonstrated sensitivity to various physical and chemical stressors.

14To further characterize the capability of the designed nanoparticles to present viral glycoproteins, we characterized the structures and antigenic profiles of I53_dn5 fused to the prefusion influenza HA and RSV F glycoproteins (HA–I53_dn5 and DS-Cav1–I53_dn5). Constructs were generated with each glycoprotein genetically linked to the N terminus of the I53_dn5B trimeric fusion component, and the proteins were secreted from HEK293F cells and purified by Ni2+ IMAC. The fusion proteins were mixed with equimolar pentameric I53_dn5A for HA–I53_dn5 or I53_dn5A.1 (a stabilized and redox-insensitive variant of I53_dn5A lacking cysteines, see Materials and Methods) for DS-Cav1–I53_dn5, and the assembly reactions purified by SEC. For both assemblies, the majority of the material migrated in the peak expected for assembled nanoparticles, and NS-EM analysis showed formation of I53_dn5 nanoparticles with spikes emanating from the surface (Figure 5—figure supplement 1 and 2). In both cases, there was considerable variation in the spike geometry, again suggesting some flexibility between the glycoproteins and the underlying scaffold. The GG linker connecting DS-Cav1 to I53_dn5 likely accounts for the observed flexibility and suboptimal definition of the glycoprotein trimer in two-dimensional class averages (Figure 5—figure supplement 1, bottom right). There was no engineered linker between the glycoprotein and fusion component in the case of HA–I53_dn5, and more clearly defined spike density was observed in the class averages (Figure 5—figure supplement 2, bottom right).

15To determine if the presented glycoproteins were properly folded, we examined their reactivity with conformation-specific monoclonal antibodies (mAbs). The DS-Cav1–I53_dn5 nanoparticle was found by an enzyme-linked immunosorbent assay (ELISA) to bind the RSV F-specific mAbs D25, Motavizumab, and AM14 similarly to soluble DS-Cav1 trimer with foldon, indicating that the F protein is presented in the desired prefusion conformation on the nanoparticle (Figure 4—figure supplement 1, top). Biolayer interferometry binding experiments with anti-HA head - and stem-specific mAbs analogously showed that both the HA–I53_dn5 nanoparticle and the HA–I53_dn5B trimer presented the head and stem regions with wild-type-like antigenicity (Figure 5—figure supplement 2, top).

16Previous work involving icosahedral nanoparticle scaffolds presenting HIV-1 Env trimers has shown that antigen crowding can modulate the accessibility of specific epitopes and thereby influence the humoral immune response. The nanoparticle scaffolds developed in this work were specifically designed to exhibit varying geometries and valencies, providing a unique way to manipulate and understand epitope accessibility in the context of nanoparticle vaccines. We selected the BG505 SOSIP–T33_dn2 tetrahedral nanoparticle (assembled in vitro using BG505 SOSIP–T33_dn2A and T33_dn2B) to compare against a previously published SOSIP-displaying icosahedral nanoparticle (BG505 SOSIP–I53-50) through surface plasmon resonance (SPR) experiments. BG505 SOSIP–T33_dn2 presents four copies of the BG505 SOSIP trimer with much greater spacing than BG505 SOSIP–I53-50 with twenty copies. This difference derives from the angles between neighboring three-fold rotational symmetry axes—where the displayed BG505 trimers are located on the nanoparticle surfaces—in icosahedral and tetrahedral point group symmetries (41.8° and 109.5°, respectively). To first validate mAb binding to BG505 SOSIP–T33_dn2A, NS-EM class averages and a 3D reconstruction were generated in complex with the VRC01 Fab, confirming the expected binding mode (Figure 6a). Next, in part to simulate surface B cell receptors, a panel of anti-Env mAbs targeting epitopes ranging from the apex to the base of the BG505 SOSIP trimer were immobilized on SPR sensor chips (Figure 6b). BG505 SOSIP–T33_dn2A trimer or BG505 SOSIP–T33-dn2 nanoparticle was flowed over the mAbs and the ratio of macromolecule bound was calculated from the binding signal as previously described. For mAbs that target apical, V3-base, and CD4-binding site epitopes (PGT145, PGT122, 2G12, and VRC01), the number of molecules of trimer or nanoparticle bound was relatively similar (ratio ~ 1). However, for mAbs that target more base-proximate epitopes in the gp120-gp41 interface (ACS202, VRC34, and PGT151), an inter-protomeric gp41 epitope (3BC315), and the main autologous neutralizing antibody epitope in the glycan hole centered on residues 241 and 289 (11B), the accessibility was reduced in the nanoparticle format. Binding to a mAb directed to the trimer base (12N) was not observed for nanoparticle BG505 SOSIP–T33_dn2 (Figure 6b). We compared epitope accessibility of BG505 SOSIP–T33_dn2 to that of BG505 SOSIP–I53-50 for six different mAbs. As for BG505 SOSIP–T33_dn2, mAbs to the apex, V3-base, and CD4-binding site (PGT145, PGT122, and VRC01) gave molar ratios of ~ 1 compared to BG505 SOSIP–I53-50. However, for mAbs that target the more base-proximate epitopes in the gp120-gp41 interface (VRC34 and PGT151), there was nearly three-fold higher epitope accessibility on T33_dn2 compared to I53-50 (Figure 6c). Further down the trimer, no accessibility difference was again observed for a mAb that targets the gp41 inter-protomeric epitope (3BC315), which was relatively inaccessible on both nanoparticles, likely due to steric hindrance by neighboring trimers. These findings demonstrate that antigen epitope accessibility can be finely tuned through presentation geometry, which could be used as a strategy to target the immune response against specific epitopes of interest.

17Strong BCR signaling is required for eliciting robust humoral immune responses, but the molecular mechanisms by which this can be accomplished are not fully understood. Historically, live-attenuated or inactivated viruses and engineered virus-like particles (VLPs) have been able to confer protective immunity without pathogenicity, but the empirical discovery and compositional complexity of such vaccines has hampered understanding of possible mechanisms for obtaining sufficient levels of protective antibodies. De novo designed protein nanoparticles provide a modular way to present antigens to the immune system in defined geometries and of known composition. Multivalent antigen presentation can enhance antigen-specific antibody titers by orders of magnitude, but presentation of complex antigens is challenging due to the required geometric compatibility between antigen and scaffold. The design approach described here, in which nanoparticles incorporate de novo designed homo-oligomers tailored to present antigens of interest, is a general solution to this problem. More broadly, the ability to build protein-based nanomaterials with geometric specifications from scratch represents an important step forward in computational protein design, and provides a systematic way to investigate the influence of antigen presentation geometry on immune response.

18The ability to fully tailor structures of nanoparticle scaffolds could be particularly useful for HIV-1 Env-based immunogens. While previous studies of HIV-1 Env trimers presented on nanoparticle scaffolds have demonstrated enhanced immunogenicity, the effects are often modest compared to those observed for other antigens. While there may exist intrinsic peculiarities to HIV-1 Env that limit increases in antibody responses upon multivalent presentation, limitations associated with epitope inaccessibility caused by crowding of the trimers on nanoparticle surfaces have also been identified. This observation strongly motivates the exploration of antigen presentation across a range of scaffolds to identify geometries that most effectively elicit the desired immune response, particularly when it is of interest to direct the humoral immune response to specific epitopes. Indeed, the SPR experiments presented here demonstrate that epitopes proximate to the BG505 SOSIP base were markedly more accessible to immobilized mAbs on BG505 SOSIP–T33_dn2 than BG505 SOSIP–I53-50, directly implicating steric crowding on the nanoparticle surface as a determinant of antigenicity. Furthermore, the availability of multiple antigen-displaying nanoparticles makes possible the usage of different scaffolds during prime and boost immunizations, which could limit immune responses directed toward the scaffolds while boosting antigen-specific antibody responses. Finally, the ability to tune antigen presentation geometry through de novo nanoparticle design provides a route to investigate the effects of this parameter on B cell activation, as well as the potency and breadth of the ensuing humoral response. This design approach could help overcome the intrinsically low affinity of germline BCRs for viral glycoproteins, and enable development of broadly neutralizing antibodies.

19Listed below are the RCSB Protein Data Bank entries for monomeric repeat proteins used for trimer docking and design in this study. An additional set of monomeric repeat proteins is provided in which experimental SAXS data agreed with the computational model.

20The monomeric repeat proteins were used as input to C3-symmetric trimer docking and design against the three viral antigens of interest: HIV-1 BG505 SOSIP, influenza H1 HA, and RSV F protein (PDB IDs 5VJ6 res. 518–664, 5KAQ res. 11–501, 5TPN res. 27–509). A C3-symmetric docking search was first performed, and output was assessed by the previously described RPX scoring method which discerns docks with more potential favorable pair-wise interactions than others. Up to the top-scoring 100 docks for each repeat protein monomer were aligned against each of the three antigens along the shared C3 axis of symmetry and sampled translationally along and rotationally about the axis in 1 Å and 1° increments, respectively. For each sample, the distance between the C-terminal residue of the antigen and N-terminal residue of the docked trimer was measured until a minimum, non-clashing distance was determined (Figure 1—figure supplement 1). Solutions for docks that were less than or equal to 15 Å for one or more of the three antigens were selected for full Rosetta symmetric interface design as described in previously published methods. Individual design trajectories were filtered by the following criteria: difference between Rosetta energy of bound (oligomeric) and unbound (monomeric) states less than −30.0 Rosetta energy units, interface surface area greater than 700 Å2, Rosetta shape complementarity (sc) greater than 0.65, and less than 50 mutations made from the respective native monomer. Designs that passed these criteria were manually inspected and refined by single point reversions, and one design per unique docked configuration was added to the set of trimers selected for experimental validation.

21Two-component nanoparticle docks were scored and ranked using the RPX method as opposed to prior methods involving only interface residue contact count. High-scoring and non-redundant nanoparticle configurations were selected for Rosetta interface design with an added caveat that they include trimers with outward-facing N termini for antigen fusion. The design protocol took a single-chain input. pdb of each component, and a symmetry definition file containing information for a specified cubic point group symmetry. The oligomers were then aligned to the corresponding axes of the symmetry using the Rosetta SymDofMover, taking into account the rigid body translations and rotations retrieved from the. dok file output from TCdock. A symmetric interface design protocol was employed which included pair-wise interaction motifs found from the RPX method within each Rosetta symmetric interface design trajectory. Individual design trajectories were filtered by the following criteria: difference between Rosetta energy of bound and unbound states less than −30.0 Rosetta energy units, interface surface area greater than 700 Å2, sc greater than 0.6, and less than 50 mutations made from each native oligomer. Designs that passed these criteria were manually inspected and refined by single point reversions for mutations that did not appear to contribute to stabilizing the bound state of the interface. The sequence with the best overall metrics for each unique docked configuration was selected for experimental characterization.

22Synthetic genes for designed proteins were optimized for E. coli expression and assembled from genes (purchased through Genscript or Gen9) ligated into the pET21b(+) (designed trimers) or pET28b(+) (designed nanoparticles) vector at restriction sites NdeI and XhoI or NcoI and XhoI, respectively. A second ribosome-binding site was inserted between the open-reading frames of individual components of nanoparticle designs (‘AGAAGGAGATATCAT’), such that the two proteins would be co-expressed and screened for co-elution by SDS-PAGE. Plasmids were cloned into BL21 or Lemo21 (DE3) E. coli competent cells (New England Biolabs). Transformants were inoculated and grown in 5 mL of either LB or TB medium with either 100 mg/L carbenicillin or 100 µg/L kanamycin at 37°C overnight. Subsequently, liquid cultures were inoculated 1:100 (v:v) and grown at 37°C until an OD600 of 0.5–0.8. Isopropyl-thio-β-D-galactopyranoside (IPTG) was then added at a concentration of 0.5–1 mM and growth temperature was reduced to 18°C to induce protein expression, or cultures were left at 37°C and auto-induced by media-included galactose according to the Studier protocols. Expression proceeded for 20 hr until the cell cultures were harvested by centrifugation. Cell pellets were resuspended in 25 mM Tris, 150 mM NaCl, 5 mM imidazole, DNase, EDTA-free protease inhibitors (Pierce), pH 8.0. and lysed by sonication or microfluidization. Each protein was then purified from lysate by Ni2+ IMAC with Ni-NTA Superflow resin (Qiagen or GE). Resin with bound cell lysate was washed with 15 column volumes of 25 mM Tris, 150 mM NaCl, 40 mM imidazole, pH 8.0. Proteins were eluted with five column volumes of 25 mM Tris, 150 mM NaCl, 400 mM imidazole, pH 8.0 for further purification by SEC.

23Elution samples for each designed protein were concentrated down using a 10,000 MWCO protein concentrator (Novagen) and fractionated by size on an AKTA pure chromatography system using a Superdex 200 (for designed trimers) or Superose 6 10/300 GL column (for designed nanoparticles) in 25 mM Tris, 150 mM NaCl, pH 8.0 (TBS). Sizing profiles were noted based on absorption at 220 nm and 280 nm wavelength light for each fraction. Molecular weights for predominant species in each protein trace were estimated by comparison to the corresponding monomeric profile.

24Fractions containing single predominant species from an initial SEC purification were concentrated down with 10,000 MWCO protein concentrators (Novagen) to a concentration of 1.0–2.0 mg/mL and run through a high-performance liquid chromatography system (Agilent) using a Superdex 200 or Superose 6 10/300 GL column (GE Life Sciences) in TBS buffer. These fractionation runs were coupled to a multi-angle light scattering detector (Wyatt) to determine the absolute molecular weights for each designed protein complex.

25Designed proteins that predominantly formed the target oligomeric species were re-expressed and purified for low-resolution solution structure determination by SAXS at the SIBYLS High Throughput SAXS Advanced Light Source in Berkeley, California. A beam exposure time of between 0.3 and 10 s was used to obtain averaged diffraction data (SAXS FrameSlice Application), which are represented in plots of log intensity (I) vs. q. A 11kEV/1.125A X-ray beam was used with a 2 m beamstop.

26Design 1na0C3_2 was found to crystallize in 1 M LiCl, 100 mM citrate, 20% w/v PEG 6000, pH 4, and was frozen using 25% glycerol as cryoprotectant. Design 3ltjC3_1 crystallized in 1 mM DL-glutamic acid monohydrate, 100 mM DL-alanine, 100 mM glycine, 100 mM DL-lysine monohydrochloride, 100 mM DL-serine; 100 mM Tris, 100 mM BICINE, 20% v/v ethylene glycol, 10% w/v PEG 8000, pH 8.5. Diffraction data for each of these designs were collected at the Advanced Light Source (Beamline 8.2.1) at Lawrence Berkeley National Laboratory in Berkeley, California. Both designed trimers contained uncleaved C-terminal His6-tags in crystallized conditions.

27Diffraction data for 3ltjC3_1 was collected on beamline 5.0.1 at the Advanced Light Source (Berkeley, CA) and 1na0C3_2 on beamline 8.2.1, both using an ADSC Q315R CCD area detector. Both datasets were scaled and merged in HKL2000. The structures were phased by molecular replacement, with the computational design as the search model, using the program PHASER in the PHENIX software suite. Iterative rounds of manual model building and refinement were conducted in Coot and Phenix.refine, respectively for both structures. Hydrogens were added for all refinement runs. The geometric quality of the final model was assessed using the Molprobity server. Resolution cutoff was determined by monitoring the refinement statistics in the context of the reflection data completeness and the CC1/2 and I/σI values.

28Genes for the individual nanoparticle components were cloned into expression vectors and expressed independently in E. coli. The His6-tagged proteins were purified following the purification protocol described for the designed trimers. Initial SEC chromatograms for the components were obtained on a Superdex 200 10/300 GL column, and predominant peak species were stoichiometrically mixed in TBS buffer for 20 min at 25°C. A secondary SEC step was performed on a Superose 6 10/300 GL column to assess assembly of the intended particle based on expected retention volume.

29Multiple rounds of designs were performed to remove native unpaired cysteines from I53_dn5A. In the first round of design, cysteines were mutated to alanines (C94A, C119A), which caused the protein to bind and retain through purification a bright yellow metabolite. Further mutations were introduced to knock out metabolite binding in the native enzymatic active site (W18G), which led to protein precipitation during purification. Additional mutations were made (K84R, M88P, E91D, L117I, L120D) to re-stabilize the protein, named I53_dn5A.1.

30Synthetic genes were optimized for mammalian expression and subcloned into the pPPI4 vector. BamHI and NheI restriction sites were used for insertion of different nanoparticle components to the C terminus of BG505 SOSIP. Quick Ligation kit, BamHI-HF, and NheI-HF restriction enzymes were purchased from New England Biolabs. BG505 SOSIP variant used for all early optimizations steps was engineered with a combination of v5.2 (mutations: E64K, A73C, A316W, A561C) and MD3D (mutations: M271I, A318Y, R585H, L568D, V570H, R304V, F519S) stabilizing mutations, and had glycosylation sites introduced at positions 241 and 289 (mutations: P240T, S241N, F288L, T290E, P291S). This construct was termed BG505 SOSIP.v5.2 (7S). For epitope-accessibility experiments (by surface plasmon resonance), a version of this construct was designed without the 241 and 289 glycans. HEK 293F (RRID:CVCL_6642) cells were grown in suspension using FreeStyle 293 Expression Medium (Thermo Fisher Scientific) at 135 RPM, 8% CO2, 80% humidity, 37°C. At confluency of ~ 1 × 106 cells/ml, the cells were co-transfected with pPPI4 DNA vectors encoding the appropriate fusion component (250 µg per 1 L of cells) and furin protease (80 µg per 1 L of cells). Polyethylenimine (Polysciences Inc) was used as a transfection reagent (1 mg per 1 L of cells). Cells were incubated for 6 days, after which they were spun down by centrifugation (7,000 RPM, 1 hr, 4°C) and the protein-containing supernatant was further clarified by vacuum-filtration (0.45 µm, Millipore Sigma). For immuno-affinity chromatography steps, Sepharose 4B columns with immobilized PGT145 IgG were used (RRID:AB_2491054). Fusion components were eluted with 3 M magnesium chloride, 250 mM L-Arginine buffer, pH 7.2 into an equal volume of SEC buffer (25 mM Tris, 250 mM L-Arginine, 500 mM NaCl, 5% glycerol, pH 7.4). Eluates were concentrated and buffer exchanged into SEC buffer. A Sephacryl S200 16/600 column was used for subsequent SEC purification.

31Synthetic genes were optimized for mammalian expression and subcloned into the CMV/R vector (VRC 8400). XbaI and AvrII restriction sites were used for insertion of I53_dn5B component to the C terminus of the H1 HA ectodomain (residues 1–676 from A/Michigan/45/2015), which also contained a Y98F mutation to prevent sialic-acid binding and self-aggregation during expression. Gene synthesis and cloning was performed by Genscript. HEK 293 F cells were grown in suspension using Expi293 Expression Medium (Thermo Fisher Scientific) at 150 RPM, 8% CO2, 70% humidity, 37°C. At confluency of ~ 2.5 × 106 cells/mL, the cells were co-transfected with the vector encoding HA–I53_dn5B (1000 µg per 1 L of cells). Expifectamine was used as a transfection reagent according to the manufacturer’s protocol. Cells were incubated for 96 hr, after which they were spun down by centrifugation (4,000 RPM, 20 min, 4°C) and the protein-containing supernatant was further clarified by vacuum-filtration (0.45 µm, Millipore Sigma). For nickel-affinity chromatography steps, a background of 50 mM Tris, 350 mM NaCl, pH 8.0 was added to clarified supernatant. For each liter of supernatant, 4 mL of Ni Sepharose excel resin (GE) was rinsed into phosphate-buffered saline (PBS) using a gravity column and then added to the supernatant, followed by overnight shaking at 4°C. After 16–24 hr, resin was collected and separated from the mixture and washed twice with 50 mM Tris, 500 mM NaCl, 30 mM imidazole, pH 8.0 prior to elution of desired protein with 50 mM Tris, 500 mM NaCl, 300 mM imidazole, pH 8.0. Eluates were concentrated and applied to a HiLoad 16/600 Superdex 200 pg column pre-equilibrated with PBS for purification by SEC.

32Gene synthesis and cloning was performed by Genscript. HEK 293 F cells (RRID:CVCL_6642) were grown in suspension using Expi293 expression medium (Thermo Fisher Scientific) at 150 RPM, 8% CO2, 70% humidity, 37°C. At confluency of ~ 2.5 to 3 × 106 cells/ml, the cells were transiently transfected with the vector encoding DS-Cav1–I53_dn5B (1 mg per 1 L of cells). Expifectamine was used as a transfection reagent according to the manufacturer’s protocol. Cells were incubated for 96 hr and spun down by centrifugation (4,000 RPM for 20 min at 4°C). Supernatant was vacuum-filtered (0.45 µm, Millipore Sigma) and 50 mM Tris, 350 mM NaCl, pH 8.0 was added for nickel-affinity chromatography. Ni Sepharose resin (GE) was washed three times with PBS by centrifugation (2,000 RPM for 5 min at 4°C) and added to the supernatant. Nickel-supernatant was incubated either overnight at 4°C or for 2 hr at room temperature. Resin was collected and separated from the mixture and washed twice with 50 mM Tris, 500 mM NaCl, 30 mM imidazole, pH 8.0 prior to elution of desired protein with 50 mM Tris, 500 mM NaCl, 300 mM imidazole, pH 8.0. Eluates were concentrated and applied to a HiLoad 10/300 Superdex 200 Increase GL column pre-equilibrated with PBS for purification by SEC.

33Several reactions containing 5–10 µg of the purified fusion component and an equimolar amount of the corresponding assembly component were prepared and incubated under different conditions (varying temperature and assembly buffer) for 24 hr. Native PAGE Bis-Tris gels (Thermo Fisher Scientific) and NS-EM was used for detection of assembly. Following the identification of optimal assembly conditions, milligram quantities of particles were assembled and purified by SEC (Superose six or Sepharose 500 column) with TBS as the running buffer. Fractions corresponding to the fusion component were pooled and concentrated (Amicon Ultra Centrifugal Filter Units, Millipore Sigma).

34NS-EM experiments were performed as described previously. Fusion components and assembled nanoparticle samples (with and without antigen) were diluted to 20–50 µg/ml and loaded onto the carbon-coated 400-mesh Cu grid that had previously been glow- discharged at 15 mA for 25 s. VRC01 (RRID:AB_2491019) complexes with BG505 SOSIP–T33_dn2A were formed by combining the trimer with a six-fold molar excess of the VRC01 Fab and subsequent incubation for 1 hr at room temperature. Complex sample was diluted to 20 µg/ml and loaded onto the glow discharged Cu grids. Grids were negatively stained with 2% (w/v) uranyl-formate for 60 s. Data collection was performed on a Tecnai Spirit electron microscope operating at 120 keV. The magnification was 52,000 × with a pixel size of 2.05 Å at the specimen plane. The electron dose was set to 25 e-/Å. All imaging was performed with a defocus value of −1.50 µm. The micrographs were recorded on a Tietz 4k × 4 k TemCam-F416 CMOS camera using Leginon automated imaging interface. Data processing was performed in Appion data processing suite. For BG505 SOSIP-fused nanoparticle samples (v5.2 (7S)), approximately 500–1000 particles were manually picked from the micrographs and 2D-classified using the Iterative MSA/MRA algorithm. For non-fused nanoparticle samples, 2,000–5000 particles were manually picked and processed. For BG505 SOSIP-fused trimer samples and Fab complexes, 10,000–40,000 particles were auto-picked and 2D-classified using the iterative MSA/MRA algorithm. 3D classification and refinement steps were done in Relion/2.1 (RRID:SCR_015701). The resulting EM maps have been deposited to EMDB with IDs: 21162 (T33_dn2), 21163 (T33_dn5), 21164 (T33_dn10), 21165 (O43_dn18), 21166 (I53_dn5), 21167 (BG505 SOSIP–T33_dn2A), 21168 (BG505 SOSIP–T33_dn2A + VRC01 Fab), 21169 (BG505 SOSIP–T33_dn2 nanoparticle), 21170 (BG505 SOSIP–T33_dn10 nanoparticle), 21171 (BG505 SOSIP–I53_dn5 nanoparticle).

35The HA–I53_dn5 complex was adsorbed onto glow-discharged carbon-coated copper mesh grids for 60 s, stained with 2% uranyl formate for 30 s, and allowed to air dry. Grids were imaged using the FEI Tecnai Spirit 120 kV electron microscope equipped with a Gatan Ultrascan 4000 CCD Camera. The pixel size at the specimen level was 1.60 Å. Data collection was performed using Leginon with the majority of the data processing carried out in Appion. The parameters of the contrast transfer function (CTF) were estimated using CTFFIND4. All particles were picked in a reference-free manner using DoG Picker. The HA–I53_dn5 particle stack from the micrographs collected was pre-processed in Relion (RRID:SCR_015701). Reference-free two-dimensional (2D) classification with cryoSPARC was used to select a subset of particles, which were used to generate an initial model using the Ab-Initio reconstruction function in CryoSPARC. The particles from the best class were used for non-uniform refinement in CryoSPARC to obtain the final 3D reconstruction.

36The sample was diluted with a buffer containing 10 mM HEPES pH 7.0 and 150 mM NaCl to a concentration of 0.025 mg/ml and adsorbed for 15 s to a glow-discharged carbon-coated copper grid. The grid was washed with the same buffer and stained with 0.7% uranyl formate. Images were collected at a nominal magnification of 57,000 × using EPU software on a ThermoFisher Talos F200C electron microscope equipped with a 4k × 4 k Ceta camera and operated at 200 kV. The pixel size was 0.253 nm. Particles were picked automatically using in-house written software (unpublished) and extracted into 200 × 200 pixel boxes. Reference-free 2D classifications were performed using Relion 1.4 and SPIDER.

37Grids were prepared on Vitrobot mark IV (Thermo Fisher Scientific). Temperature was set to 10°C, humidity at 100%, wait time at 10 s, while the blotting time was varied in the 4–7 s range with the blotting force at 0. The concentrations of T33_dn10, O43_dn18, and I53_dn5 nanoparticle samples were 4.2, 3.0, and 1.9 mg/ml, respectively. n-Dodecyl β-D-maltoside (DDM) at a final concentration of 0.06 mM was used for sample freezing. Quantifoil R 2/1 holey carbon copper grids (Cu 400 mesh) were pre-treated with Ar/O2 plasma (Solarus plasma cleaner, Gatan) for 10 s prior to sample application. Concentrated nanoparticle samples were mixed with appropriate volumes of stock DDM solution and 3 µl applied onto the grid. Excess sample and buffer was blotted off and the grids were plunge-frozen into nitrogen-cooled liquid ethane. Cryo-grids were loaded into a Titan Krios (FEI) operating at 300 kV, equipped with the K2 direct electron detector camera (Gatan). Exposure magnification of 29,000 was set with the resulting pixel size of 1.03 Å at the specimen plane. Total dose was set to ~ 50 e-/Å with 250 ms frames. Nominal defocus range was set to −0.6 to −1.6 µm for all three nanoparticle samples. Automated data collection was performed using Leginon software. Data collection information for acquired datasets is shown in Figure 4—source data 1.

38MotionCor2 was run to align and dose-weight the movie micrographs. GCTF v1.06 was applied to estimate the CTF parameters. Initial processing was performed in cryoSPARC 2.9.0. Template-picked particles were extracted and subjected to 2D classification. Multiple rounds of heterogeneous refinement were performed to further clean-up particle stacks in three acquired datasets. Selected subsets of particles were then transferred to Relion 3.0 (RRID:SCR_015701) for further processing. Reference models were generated using Ab-Initio reconstruction in cryoSPARC v2.9.0 with the application of appropriate symmetry (tetrahedral, octahedral, and icosahedral for T33_dn10, O43_dn18, and I53_dn5, respectively). Several rounds of 3D classification and refinement were used to sort out a subpopulation of particles that went into the final 3D reconstructions. Tetrahedral, octahedral, and icosahedral symmetry restraints were applied for all 3D refinement and classification steps during the processing of T33_dn10, O43_dn18, and I53_dn5 datasets, respectively. Soft solvent mask around the nanoparticle core was introduced during the final 3D classification, refinement, and post-processing. The resolutions of the final reconstructed maps were 3.86 Å for T33_dn10, 4.54 Å for O43_dn18, and 5.35 Å for I53_dn5. The resulting EM maps have been deposited to EMDB with IDs: 21172 (T33_dn10), 21173 (O43_dn18) and 21174 (I53_dn5). A graphical summary of the data processing approach and relevant information for each dataset are displayed in Figure 4—figure supplement 1.

39The antigenicity of BG505 SOSIP–T33_dn2A trimer and BG505 SOSIP–T33_dn2 nanoparticle was analyzed on a BIAcore 3000 instrument at 25 °C and with HBS-EP (GE healthcare Life sciences) as running buffer, as described. Affinity-purified goat anti-human IgG Fc (Bethyl Laboratories, Inc) and goat anti-rabbit IgG Fc (Abcam) were amine-coupled to CM3 chips and the anti-HIV-1 Env human and rabbit mAbs were captured to an average density of 320 ± 1.5 RU (s.e.m). BG505 SOSIP–T33_dn2A or BG505 SOSIP–T33_dn2 (both v5.2(7S) without N241/N289) was allowed to associate for 300 s and then dissociate for 600 s at a concentration of 10 nM assembled macromolecule (trimer or nanoparticle). The low background binding in parallel flow cells with only anti-Fc was subtracted. The lack of binding of nanoparticles lacking Env was ascertained for each mAb. To illustrate how epitope accessibility affects the relative binding of the trimers and nanoparticles, we converted the signals, which are proportional to mass bound, to moles bound and calculated the ratio for nanoparticles/trimers. For this comparison historic data on icosahedral nanoparticles were included. The number of moles binding to the immobilized IgG at the end of the association phase was calculated: where n is the number of moles of macromolecules, R the response at 300 s (RU), m the mass bound per area and RU (g/(mm RU)), A the interactive area of the chip (mm), and M the molar mass of the macromolecule (g/mol). This analysis corrects for the greater mass (and thereby greater signal) for each bound nanoparticle such that the number of binding events by differing macromolecules can be directly compared.

40To produce biotin-labeled antibodies specific to the H1 HA head, mAb 5J8 in PBS was mixed with a 20 × molar excess (relative to complete antibodies) of EZ-Link NHS-LC-Biotin (Thermo Fisher Scientific) and allowed to sit for 2 hr at 4°C, followed by two rounds of overnight dialysis against PBS at 4°C to remove excess biotinylation reagent. All biosensors were hydrated in assay buffer (25 mM Tris, 150 mM NaCl, 0.5% bovine serum albumin, 0.01% TWEEN 20, pH 8.0) before use. Biotinylated 5J8 (20 μg/mL in assay buffer) was immobilized on SA biosensors, then briefly dipped in assay buffer prior to exposure to designed H1 HA fusions (500 nM per asymmetric unit, in assay buffer). The biosensor was again dipped in assay buffer and then exposed to the stem-specific mAb CR6261 (20 μg/mL in assay buffer).

41Purified HA-displaying nanoparticles or trimers were applied to a Sephacryl S-500 column pre-equilibrated with 25 mM Tris, 2 M NaCl, 5% glycerol, pH 8.0. Sizing profiles were recorded based on absorption at 280 nm wavelength light.

42ELISA was used to measure binding kinetics of DS-Cav1–I53_dn5 to RSV F-specific mAbs D25, Motavizumab, and AM14. D25 is a prefusion specific mAb that binds site Ø. Motavizumab binds site II of the pre and post-fusion conformations. AM14 is trimer-specific binding across protomers of the prefusion conformation. 96-well ELISA plates were coated with 2 μg/mL DS-Cav1 nanoparticles. Plates were incubated at 4°C overnight and blocked with PBS containing 5% skim milk at 37°C for 30 min. mAbs listed were serially diluted in fourfold steps, and then added to the plates and incubated at 37°C for 45 min. Horseradish peroxidase (HRP)-conjugated anti-human IgG (Southern Biotech., Birmingham, AL) was added and incubated at 37°C for 30 min, followed by 3,3′,5′,5- Tetramethylbenzidine (TMB; Sigma-Aldrich, St. Louis, MO) HRP substrate, and yellow color that developed after the addition of 1 M H2SO4 was measured by absorbance at 450 nm.