Hierarchical design of pseudosymmetric protein nanocages

Pseudosymmetric trimer design

To identify mutations for altering trimer assembly specificity, we first identified all pairs of interacting residues in the trimer interface. Contacts were defined as any residue with a heavy (that is, non-hydrogen) atom within 4 Å of a heavy atom in a residue across the interface. We then used Rosetta to calculate the total score of poses containing all possible pairs of mutations, as well as the difference in score between the trimeric and monomeric states using the ddG filter. Example scripts are provided as supplementary files. Individual mutations were evaluated by comparing their ddG and total scores to those of the wild-type (WT) interface according to equation (1). The total scores and ddG values of the paired mutations were similarly normalized according to equation (2).

Ideal mutant pairs were those where one or both single mutations increased the energy of the trimer relative to the wild type (that is, normalized scores > 0) while the double mutation had no effect or stabilized the trimer (that is, normalized scores ≤ 0). We also identified likely positions for design using coevolutionary analysis^48,49. Strongly co-evolving residues at the protein–protein interface were identified using GREMLIN. We then identified mutations that were negatively correlated with the wild-type pair for testing experimentally.

Mutant protein expression

Mutant I53-50A trimers were expressed at three scales. Small-scale expression was performed at 1 ml culture volume in 96-well plates with 2 ml well volume. Medium-scale expressions were performed at 50 ml culture volume in 250 ml baffled shake flasks. Large-scale expressions were performed at 500 ml culture volumes in 2 l baffled shake flasks. All proteins were expressed in T7 competent E. coli in TB medium, with IPTG induction for 3 h at 37 °C. Cells were pelleted and frozen at −20 °C until lysis. Prior to lysis cells were defrosted on ice in lysis buffer (50 mM Tris pH 8.0, 250 mM NaCl, 20 mM imidazole, 1 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol (DTT), 0.1 mg ml⁻¹ DNase, and 0.1 μM RNase, unless otherwise noted). Small-scale expressions were lysed with a plate sonicator (QSonica), medium-scale expressions were lysed with a probe sonicator, and large-scale expressions were lysed by microfluidization (18,000 psi, one pass). Lysates from small-scale expressions were clarified by centrifugation in a swinging bucket rotor at 4,000g. Lysates from medium- and large-scale expression lysates were clarified by centrifugation at 12,000g in a fixed-angle rotor.

I53-50B expression and purification

Pentameric I53-50B was produced recombinantly in E. coli. A pET29b expression plasmid encoding I53-50B.4PT1¹¹ was synthesized by GenScript using the NdeI and XhoI restriction sites with a double stop codon just before the C-terminal polyhistidine tag. Tagless protein was expressed in Lemo21(DE3) cells (NEB) in LB (10 g Tryptone, 5 g Yeast Extract, 10 g NaCl) grown in a 10 l BioFlo 320 Fermenter (Eppendorf). At inoculation, impeller speed was set to 225 rpm, gas flow rate was set to 5 standard litres per minute with O₂ supplementation as part of the dissolved-oxygen aeration cascade, and the temperature set to 37 °C. At the onset of a dissolved oxygen spike (OD ~ 12), the culture was fed with a bolus addition of 100 ml of 100% glycerol and induced with 1 mM IPTG. During this time, the culture temperature was reduced to 18 °C and O₂ supplementation was ceased, with expression continuing until OD reached ~20. The culture was collected by centrifugation and the protein was purified from inclusion bodies. First, pellets were resuspended in PBS, homogenized, and then lysed by microfluidization using a Microfluidics M110P at 18,000 psi. Following sample clarification by centrifugation (24,000g for 30 min), the supernatant was discarded and protein was extracted from the pellet using a series of three washes. The first wash consisted of PBS, 0.1% Triton X-100, pH 8.0. The second wash consisted of PBS, 1 M NaCl, pH 8.0, and the final wash (extraction) consisted of PBS, 2 M urea, 0.75% CHAPS (3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate), pH 8.0. Following extraction, the sample was applied to a DEAE Sepharose FF column (Cytiva) on an AKTA Avant150 FPLC system (Cytiva). After sample binding, the column was washed with 5 column volumes of PBS at pH 8.0 with 0.1% Triton X-100, followed by a wash with 5 column volumes of PBS at pH 8.0 with 0.75% CHAPS. The protein was eluted with 3 column volumes of PBS at pH 8.0 with 500 mM NaCl. After purification, fractions were pooled and concentrated in 10K MWCO centrifugal filters (Millipore), sterile filtered (0.22 μm), aliquoted and flash-frozen in liquid nitrogen, and stored at −80 °C until use.

Assembly competency analysis

Single mutations were introduced into the I53-50A trimer¹¹ by QuikChange site-directed mutagenesis. Sequence-verified mutants were expressed at small scale. Clarified lysates were separated from pellets and a 5 µl aliquot was set aside for characterization by SDS–PAGE. The pellet was resuspended in lysis buffer and a 5 µl aliquot was set aside for characterization by SDS–PAGE. Clarified lysate was immediately mixed with purified I53-50B.4PT1 pentamer. Because trimer expression levels varied from mutant to mutant, pentamer was added at three different concentrations. To 10 µl lysate, 7.5, 2.5 or 0 µl lysis buffer was added, followed by 2.5, 7.5 or 10 µl I53-50B.4PT1 pentamer at 1.8 mg ml⁻¹. The assembly reaction was allowed to proceed for 30 min at room temperature. Purified I53-50B.4PT1 pentamer was included on all native PAGE gels. A 10 µl aliquot of each assembly reaction was mixed 1:1 with Native Sample Buffer (Bio-Rad Laboratories), loaded into precast 4–15% polyacrylamide gels (Bio-Rad Laboratories), and run with 1× Tris-Glycine Native PAGE buffer for 3 h at 200 V. The gel was stained with GelCode Blue (Thermo Fisher Scientific) and destained in water. The lack of an I53-50 nanocage band on the native gel indicated single mutations that disrupted either trimer formation or trimer geometry such that the mutant trimer was no longer assembly-competent.

Screening of mutant combinations

Single mutants that disrupted I53-50A trimer—and therefore I53-50 nanocage—formation were combined with ‘rescue’ mutations intended to generate pseudosymmetric I53-50A trimers. Synthetic DNA encoding potential combinations were ordered as heterotrimeric operons cloned into pCDB179 from IDT. To facilitate detection of the distinct components of the heterotrimer, a 6×His-SUMO domain was added to one subunit and sfGFP and an avi-tag added to a second subunit via genetic fusion. The third subunit bore a Strep-tag via genetic fusion. Variants were tested for I53-50 nanocage formation using trimer-containing E. coli lysates and purified I53-50B pentamer as described above. Combinations that formed I53-50 nanocages were expressed at large scale and purified by Ni²⁺ affinity chromatography on a HisTrap FF column (Cytiva). In brief, clarified lysate was passed through a pre-equilibrated 5 ml HisTrap FF column, washed with 3–5 column volumes of wash buffer (50 mM Tris pH 8.0, 250 mM NaCl, 20 mM imidazole, 1 mM DTT), and heterotrimer was eluted with either a step elution or a gradient over 40 min at 3 ml min⁻¹ flow rate into 100% elution buffer (50 mM Tris pH 8.0, 250 mM NaCl, 500 mM imidazole, 1 mM DTT). Major fractions corresponding to the two observed peaks in the elution profile were pooled separately, concentrated in a 30-kDa cut-off Amicon concentrator (Millipore), and injected onto a pre-equilibrated Superdex 200 Increase 10/300 column (Cytiva). The SEC buffer was 25 mM Tris pH 8.0, 150 mM NaCl, 1 mM DTT. Fractions corresponding to the trimer peak from each chromatogram were collected for analysis by native mass spectrometry. Alternatively, the IMAC eluate was pooled and loaded onto a StrepTrap HP column (Cytiva) pre-equilibrated in binding buffer (100 mM Tris pH 8.0, 150 mM NaCl, 1 mM EDTA, and 1 mM DTT). The column was then washed with 10 column volumes of binding buffer, or until the A280 absorbance leveled off at baseline and eluted with a step elution in binding buffer plus 2.5 mM desthiobiotin. Major fractions were analysed by reducing SDS–PAGE.

Native mass spectrometry

Trimer purity, identity, and oligomeric state were analysed by on-line buffer-exchange mass spectrometry⁶¹ in 200 mM ammonium acetate using a Vanish ultra-performance liquid chromatography coupled to a Q Exactive ultra-high mass range Orbitrap mass spectrometer (Thermo Fisher Scientific). The recorded mass spectra were deconvolved with UniDec version 4.2+ (ref. ⁶²).

Assembly of I53-50 nanocages using pseudosymmetric I53-50A heterotrimers

The native mass spectrometry-verified pseudosymmetric I53-50A heterotrimer was expressed and purified at medium scale as described above and mixed at a 1:1 molar ratio with purified I53-50B.4PT1 pentamer and allowed to assemble at room temperature for 30 min. Assembled nanocages were characterized by DLS and negative-stain electron microscopy as described below.

Computational design of T = 4 nanocages

We created a model of the pentasymmetron by extracting five trimers surrounding the icosahedral five-fold from I3-01¹⁴. We reverted the interface residues on the unpaired subunit back to the original 1WA3 sequence, mutated 12 residues to negatively charged amino acids to enhance expression and facilitate purification, then combined each trimer into a single chain so that the pentasymmetron could be treated computationally as a simple homopentamer. We used previously described protocols¹¹ to dock and design T = 4 nanocages, with some modifications to the design script. Example design scripts are provided as supplementary files. Docked configurations were manually screened to ensure interfaces were between the unpaired pentasymmetron subunit and the homotrimer. Designs were visually inspected and any overly exposed hydrophobic residues introduced during design were reverted to their wild-type identities.

Screening of T = 4 nanocages by co-purification

Three tricistronic genes were ordered from IDT. An N-terminal GFP was included on the A subunit of the pentasymmetron heterotrimer as a mass tag. A C-terminal 6×His tag was added to the C subunit. Genes were expressed at medium scale. Clarified lysate was loaded onto 1 ml of Ni-NTA resin (Thermo Fisher Scientific) pre-equilibrated in wash buffer. After washing with three column volumes of wash buffer, the protein was eluted with two column volumes of elution buffer. Eluate was screened for the presence of all three gene products by SDS–PAGE.

Purification of co-expressed GI₄-F7

GI₄-F7 nanocages expressed tricistronically at large scale were purified by loading on a 5 ml HisTrap FF column (Cytiva) equilibrated in wash buffer (50 mM Tris pH 8.0, 250 mM NaCl, 20 mM imidazole, 1 mM DTT). After loading, the column was washed with 3–5 column volumes of wash buffer and protein was eluted with a gradient into 100% elution buffer (50 mM Tris pH 8.0, 250 mM NaCl, 500 mM imidazole, 1 mM DTT) over 40 min at 3 ml min⁻¹. The major fractions from elution were pooled, concentrated to ~1 ml, and loaded onto an equilibrated Sephacryl S-500 HR 10/300 GL. SEC buffer was 25 mM Tris pH 8.0, 150 mM NaCl, 1 mM DTT.

Purification of GI₄-F7 heterotrimeric and homotrimeric components

For in vitro assembly, the heterotrimeric component of GI₄-F7, comprising only the A and B chains, was expressed bicistronically. The A chain was modified with an N-terminal 6×His tag. When expressed this way, some AAA nanocages and BBB homotrimers probably assemble in addition to AAB and ABB heterotrimers. To purify AAB from ABB heterotrimers, the bicistronic gene was expressed at large scale with the modification that 0.75% CHAPS was added to the lysis buffer and DTT was omitted. Clarified lysate was purified with a 5 ml HisTrap FF column as described above. Elution chromatograms contained three peaks. The first peak was predominantly the ABB heterotrimer, the second peak was predominantly AAB heterotrimer, and the third peak was predominantly AAA homotrimers assembled into an I3-01-like particle. Any BBB homotrimer would be in the flow-through. The first and second peaks were pooled separately and concentrated to ~1 ml. To remove any residual I3-01-like nanocage, we further purified the concentrated fractions on a Superose 6 Increase 10/300 column. The SEC buffer was 25 mM Tris pH 8.0, 150 mM NaCl, 0.75% w/v CHAPS. Glycerol was added to purified heterotrimer to a final concentration of 5%, the concentration was determined by A280, and 1 ml aliquots were flash-frozen in liquid nitrogen. Aliquots were stored at −80 °C until use. The homotrimer components were expressed at large scale and purified by IMAC in the same way as the co-expressed GI₄-F7 nanocages except that 1% CHAPS was added to all buffers. It was further purified by SEC on a HiLoad 26/600 Superdex 200 PG column in 25 mM Tris pH 8.0, 150 mM NaCl, 5% glycerol, 1.0% w/v CHAPS, 1 mM DTT. The total trimer protein concentration was measured by A280, flash-frozen in liquid nitrogen in 1 ml aliquots, and stored at −80 °C until use.

In vitro assembly of GI_T-F7 nanocages

To assemble GI_T-F7 nanocages, components were mixed at various stoichiometries depending on the target assembly state in the presence of 3% CHAPS, a condition that prevented premature assembly. This was necessary to prevent the assembly of off-target species during addition of the multiple components required to generate the target assemblies. For example, mixing BBB–CCC heterotrimers prior to the addition of AAB and ABB components under assembly-permissive conditions would result in 2D arrays instead of GI_T-F7 nanocages (see Extended Data Fig. 6). Once all components were added, the mixtures were dialysed into 0% CHAPS overnight at room temperature in a 30-kDa cut-off dialysis cassette. As an extra precaution, the AAB and ABB heterotrimers were mixed first since they do not directly interact with the BBB homotrimer, followed by addition of the CCC homotrimer. Nanocages were prepared fresh for each experiment, or stored at 4 °C for up to three days. To assemble BBB–CCC 2D arrays, the components were first individually dialysed to remove CHAPS, and then mixed at a 1:1 stoichiometric ratio and allowed to assemble overnight at room temperature.

Characterization of assemblies

Assemblies were characterized in solution by DLS. Samples were measured in triplicate, technical replicates, using an UNcle (UNchained Labs) according to the manufacturer’s directions. In brief, 8.8 µl of sample was loaded in triplicate into the capillary cassette. For each replicate, 10 acquisitions 10 s in length were collected. Assemblies were further characterized by negative-stain electron microscopy. Samples were diluted to between 0.1 and 0.5 mg ml⁻¹ total protein depending on the assembly stoichiometry, applied to a glow-discharged thick carbon film 400 mesh copper grid (Electron Microscopy Sciences), and stained with 2% uranyl formate. Care was taken to ensure the stain thickness was sufficient to support the larger assemblies. Micrographs were collected on a Talos L120C (FEI) at up to 48,000× magnification. Individual micrographs were processed with ImageJ.

Conjugation of RBD antigens to GI_T-F7 nanocages and characterization by negative-stain electron microscopy

To enable conjugation of antigens to assembled nanocages, CCC trimers were fused to a SpyCatcher002 motif at their C terminus, expressed in E. coli, and purified via IMAC and SEC, as described above. Nanocages were then assembled at the appropriate stoichiometries for T = 4 and T = 9 assemblies by dialysing into a 0% CHAPS solution overnight. Assembled particles were then mixed with an excess of RBD-SpyTag002 and mixed at 4 °C for 3 h. Conjugation was confirmed by SDS–PAGE, wherein the mass of RBD showed a ~30 kDa increase, consistent with conjugation to CCC proteins in nanocage assemblies. After conjugation, particles were also visualized by negative-stain electron microscopy to confirm intact assemblies. Samples were prepared by applying 3 µl of a 5 µM nanocage solution to glow-discharge carbon-coated grids, followed by staining with uranyl formate 3 times prior to imaging. EPU software (Thermo Fisher) was used to collect at least 100 micrographs of each sample. Images were imported to CryoSparc and particles were averaged to obtain initial 2D classes. Selected classes were then used to generate templates for a second round of particle picking, and new particles were averaged multiple times to obtain the 2D classes shown in Fig. 3i.

B cell activation assay

The COVA2-15 IgG RAMOS cell line was generously provided by the van Gils laboratory⁶⁰ and not authenticated further. For Ca²⁺ flux experiments, cells were loaded with FuraRed cell-permeable dye (Thermo Fisher) for 30 min in RPMI1640 supplemented with 10% fetal clone II, 1% l-glutamax, and 1% penicillin–streptomycin (complete medium) at a cell concentration of 1 × 10⁷ per ml. Cells were then washed with 10× volume complete medium, resuspended at 2 × 10⁶ cells per ml in complete medium, and aliquoted at 0.25 ml into individual FACS tubes. Samples were kept at room temperature and then warmed in a 37 °C bath for 3 min immediately before use. Acquisition was performed on an Attune CytPix flow cytometer (Thermo Fisher) with baselines recorded for 30 s for each sample before addition of antigen and measurement of BCR-specific activation. For gating strategy, see Supplementary Fig. 6. A polyclonal goat anti-human IgG F(ab′)2 (Southern Biotech) was used as a positive control for signalling resulting from IgG BCR cross-linking by addition of 2.5 µg to cells. The FuraRed ratio of bound (fluorescence in VL3) and unbound (fluorescence in BL1) Ca²⁺ was used for analysis using FlowJo v10 (BD Biosciences). Cell lines were not tested for mycoplasma.

Cryo-EM sample preparation, data collection and data processing

Three microlitres of 3 mg ml⁻¹ GI₄-F7, GI₉-F7, and GI₁₆-F16 were loaded onto freshly glow-discharged R 2/2 UltrAuFoil grids, prior to plunge freezing using a Vitrobot Mark IV (Thermo Fisher Scientific) with a blot force of 0 and 6 sec blot time at 100% humidity and 22 °C. Data were acquired using an FEI Titan Krios transmission electron microscope operated at 300 kV and equipped with a Gatan K3 direct detector and Gatan Quantum GIF energy filter, operated in zero-loss mode with a slit width of 20 eV. For GI₄-F7 and GI₉-F7, automated data collection was carried out using Leginon⁶³ at a nominal magnification of 105,000× with a pixel size of 0.843 Å. 7,249 and 2,558 micrographs were collected with a defocus range comprised between −0.5 and −2.5 μm, respectively. The dose rate was adjusted to 15 counts per pixel per s, and each movie was acquired in super-resolution mode fractionated in 75 frames of 40 ms. For the GI₁₆-F7 data set, automated data collection was carried out using Leginon⁶³ at a nominal magnification of 64,000× with a pixel size of 1.42 Å. In total, 2,268 micrographs were collected with a defocus range between −0.5 and −3.5 μm. The dose rate was adjusted to 15 counts per pixel per s, and each movie was acquired in super-resolution mode fractionated in 50 frames of 100 ms. Movie frame alignment, estimation of the microscope contrast-transfer function parameters, particle picking and extraction were carried out using Warp⁶⁴.

Two rounds of reference-free 2D classification were performed using CryoSPARC⁶⁵ to select well-defined particle images. These selected particles were subjected to two rounds of 3D classification with 50 iterations each (angular sampling 7.5° for 25 iterations and 1.8° with local search for 25 iterations) using Relion⁶⁶ with an initial model generated with ab initio reconstruction in cryoSPARC. 3D refinements were carried out using non-uniform refinement along with per-particle defocus refinement in CryoSPARC. Selected particle images were subjected to the Bayesian polishing procedure⁶⁷ implemented in Relion 3.1 before performing another round of non-uniform refinement in cryoSPARC followed by per-particle defocus refinement and again non-uniform refinement. To further improve the density of the asu, the particles were symmetry-expanded and subjected to focus 3D classification without refining angles and shifts. Particles belonging to classes with the best resolved asu density were selected and then subjected to local refinement using CryoSPARC. Local resolution estimation,and sharpening were carried out using CryoSPARC. Reported resolutions are based on the gold-standard Fourier shell correlation of 0.143 criterion and Fourier shell correlation curves were corrected for the effects of soft masking by high-resolution noise substitution^68,69.

Model building and refinement

UCSF Chimera⁷⁰ and Coot⁷¹ were used to fit atomic models into the cryo-EM maps. GI₄-F7 and GI₉-F7 asu models were refined and relaxed using Rosetta using sharpened and unsharpened maps^72,73. For GI₄-F7 or GI₉-F7 icosahedral model, all of the side chains of GI₄-F7 or GI₉-F7 asu model are truncated except Gly, Cys, and Pro residues and the symmetry-related copies were generated in ChimeraX with cryo-EM maps.

Alignments and images

To align the cryo-EM models to the design model, both models were centred at the origin and their icosahedral symmetry axes aligned in PyMOL⁷⁴. The Cα r.m.s.d. was calculated using the rms_cur function in PyMOL. To measure deviations in the rigid-body degrees of freedom, copies of the pentasymmetron, disymmetron, and trimer (or trimers for GI₉-F7) from the cryo-EM model were aligned to the design model using the ‘super’ function in PyMOL. We then calculated the rotations and translations from the transformation matrix between the corresponding component of the original cryo-EM model and the aligned cryo-EM model. We applied the same approach to the heterotrimer (and homotrimer for GI₉-F7) components to obtain rotations and translations within the pentasymmetron, disymmetron, and homotrimer components, respectively. We found that the ‘super’ function in PyMOL was very sensitive to chain and residue numbering, as well as some of the minor differences between the design model and cryo-EM model. Therefore, for all alignments using PyMOL, we made sure to harmonize residue numbering, chain IDs, and remove any residues present in only one model or the other. For that reason, aligned images were generated using the mm command in ChimeraX⁷⁵ and verified to ensure that the alignments closely matched those generated on the trimmed models created with the super function in PyMOL.

Scripts and plots

All data were processed and plotted using Python 3.8.8, matplotlib 3.3.4 and seaborn 0.11.1.

Reporting summary

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