Extension of pH-responsive trimeric building blocks

We used the WORMS helical fusion software29 to generate fusions between the base pH-trimer helical bundle and pairs of 82 helical repeat proteins, designed to cover a broad range of axial and radial displacement between repeat units55. Scaffold information was prepared in a json file, which is used as an input parameter in the WORMS software through the --dbfiles option. For each scaffold, the pdb file path ('file'), a geometric class specification ('class') and a connections dictionary ('connections') specifying the chain ('chain'), terminus direction ('direction') and residue positions that were available for fusion ('residues') were included in the json file. Geometric class specification for oligomeric scaffolds required either an N or C following the oligomeric architecture, to define the terminus available for subsequent fusions. A connections dictionary specifying each chain, the terminus direction and residue positions allowed for fusion was required for each chain and terminus direction. The oligomeric scaffold asymmetric unit required one connections dictionary, specifying fusion to one terminus and one chain. As monomers, helical repeat proteins require a list of two connections dictionaries, because they allow fusions to either the N or C terminus. An example json file example is shown below, where the connections dictionary specifies allowable fusions on the C terminus of the pH trimer from residue 50 and over, and anywhere along the helical repeat proteins:

 ] }, ,    ] }, ,    ] } }

To prepare the WORMS software for fusions, we also included a map for backbone connections (--bbconn), which specifies the order in which scaffolds of each class can be fused. We used the following backbone connection map, which specifies that the N terminus of the fusion of two helical repeat proteins in the monomer class is fused to the C terminus of an oligomer with C3 symmetry:

--bbconn C3 C3_C \  NC Monomer \  N_ Monomer \

Finally, we specified a --geometry NullCriteria() option, which allowed generation of arbitrary fusions of varying radius and length to the C terminus of the pH trimer.

Computational docking of plugged antibody nanoparticles

The generated fusions served as docking inputs into the antibody nanoparticle through the --inputs1 and --inputs2 options. To prepare the antibody nanoparticle for axle docking along its C3 axis, we generated a context structure PDB file by isolating the chains forming the C3 symmetric aperture. We fed this context structure and each trimer fusion separately into the RPXDock application31 using the axle docking protocol (--architecture AXLE_C3). To enable trimming, we set a maximum number of trimmed residues (--max_trim) and specified the chains representing the fusion to be trimmed (--trimmable_components). Finally, to enable greater sampling, we allowed the RPXDock application to sample the entire search space (--docking_method) at a cartesian and angular resolution of 0.25 angstroms and degrees, respectively (--grid_resolution_cart_angstroms and --grid_resolution_ori_degrees). An example axle docking executable is shown below.

PYTHONPATH = /path/to/rpxdock/site-packages /path/to/python/environment /path/to/rpxdock/dock.py \

 ––architecture AXLE_C3 \  ––inputs1 /path/to/plugfusion.pdb \  ––inputs2 /path/to/contextstructure.pdb \  ––max_trim 400 \  ––cart_bounds -300 300 \  ––docking_method grid \  ––grid_resolution_cart_angstroms 0.25 \  ––grid_resolution_ori_degrees 0.25 \  ––hscore_files ailv_h \  ––hscore_data_dir /path/to/rpxdock/hscore \  ––trimmable_components 'A' \  ––function stnd \

We evaluated the docked interfaces between each fusion and antibody nanoparticle using the RPX score and NContact metrics, which are predictive of geometric complementarity and designability. Empirically, we deemed docks with an RPX score greater than 50 and with more than one residue pair contact to be suitable for sequence optimization.

Computational design of plugged antibody nanoparticles

We used the Rosetta Macromolecular Modeling suite to optimize the sequence residues contributing to either the fusion domains or interface between plug and nanoparticle. With residue selectors, we identified clashing residue positions as a result of the WORMS fusion and residues contributing to the interface between the plug and antibody nanoparticle. The sequence of the selected residues was optimized on the basis of the local positions of each residue defined by secondary structure, solvent accessible surface area (SASA) of the backbone atom and the Cβ atom and the number of amino acid side chains in a cone extending along the Cα–Cβ vector side chain neighbors. An example of the fusion, docking and design of O432-17 is included at https://github.com/erincyang/plug_design.git.

Small-scale bicistronic bacterial protein expression

Synthetic genes encoding designed trimeric-plug and tetramer sequences bicistronically were purchased from Genscript in pET-29b+ vectors. They were connected by an intergenic region (5′-TAAAGAAGGAGATATCATATG-3′) and a C-terminal 6×histidine tag on the tetramer. Expression plasmids were transformed into BL21(DE3) Escherichia coli cells and were grown in LB medium supplemented with 50 mg l–1 kanamycin at 37 °C overnight. The overnight culture was diluted into autoinduction medium and incubated at 37 °C overnight. Cells were lysed chemically in BugBuster supplemented with 1 mM PMSF and 20 mM imidazole, and were cleared by centrifugation. Clarified lysates were purified using IMAC with Ni-NTA magnetic beads. The soluble fractions were washed with 25 mM Tris pH 8.0, 300 mM NaCl and 60 mM imidazole before elution with 25 mM Tris pH 8.0, 300 mM NaCl and 300 mM imidazole. Elution fractions from bicistronic expression plasmids were subsequently subjected to native PAGE and SDS–PAGE to identify slow migrating species that co-eluted two proteins of different molecular weights.

Small-scale bicistronic bacterial protein expression for trimeric-plug and tetrameric designs for sfGFP-Fc lysate assembly

Synthetic genes encoding designed trimeric plug and tetramer sequences bicistronically were purchased from Genscript in pET-29b+ vectors and connected by an intergenic region (5′-TAAAGAAGGAGATATCATATG-3′) and a C-terminal 6×Histidine tag on the tetramer. Expression plasmids were transformed into BL21(DE3) E. coli cells and grown in LB medium supplemented with 50 mg l–1 kanamycin at 37 °C overnight. The overnight culture was diluted into autoinduction medium and incubated at 37 °C overnight. Cells were lysed chemically in BugBuster supplemented with 1 mM PMSF and 20 mM imidazole, and were cleared by centrifugation. Clarified lysates were incubated with a 3 μM final concentration of sfGFP-Fc and purified by IMAC with Ni-NTA magnetic beads. The soluble fractions were washed with 25 mM Tris pH 8.0, 300 mM NaCl and 60 mM imidazole before elution with 25 mM Tris pH 8.0, 300 mM NaCl and 300 mM imidazole. Elution fractions from bicistronic expression plasmids were subsequently subjected to native PAGE and SDS–PAGE to identify slow migrating species that co-eluted three proteins of different molecular weights.

Production of Fc and Fc fusions

Fc, sfGFP-Fc, mRuby2-Fc and CTX IgG were purchased from Genscript in CMVR vectors and transfected into in ExpiHEK293F cells (Thermo Fisher Scientific) and purified by IMAC with 50 mM Tris pH 8.0, 300 mM NaCl and 500 mM imidazole elution buffer, and were then purified further through SEC over a Superdex 200 10/300 GL FPLC column (Cytiva) into 50 mM Tris pH 8.0, 300 mM NaCl and 0.75% CHAPS (Fc) or 50 mM Tris, pH 8.0, 300 mM NaCl and 0.05% glycerol (Cetuximab IgG, sfGFP-Fc, mRuby2-Fc). Stocks were frozen at −80 °C for subsequent analyses.

Large-scale expression and purification of O432-17 components

Designs that co-purified and yielded slowly migrating species, as determined by native PAGE, were subsequently subcloned into pET-29b+ vectors (Genscript). Each vector encoded either a trimeric plug or a tetramer variant—both with C-terminal hexahistidine tags—that was expressed at a larger scale (1–12 l of culture). Cells were lysed by microfluidization in 25 mM Tris pH 8.0, 300 mM NaCl, 1 mM DTT, 1 mM PMSF and 0.1 mg ml–1 DNAse and cleared by centrifugation. Clarified lysates were filtered through 0.7-μm filters and purified by IMAC using gravity columns with nickel-NTA resin or HisTrap HP columns (Cytiva) using 25 mM Tris pH 8.0, 300 mM NaCl and 60 mM imidazole wash buffer and 25 mM Tris pH 8.0, 300 mM NaCl and 300 mM imidazole elution buffer. Elution fractions containing pure proteins were concentrated using centrifugal filter devices (Millipore) and further purified on a Superdex 200 10/300 GL (for large-scale purification) or Superose 6 10/300 GL (for comparison to assembly) gel-filtration column (Cytiva) using 25 mM Tris pH 8.0, 150 mM NaCl and 0.75% CHAPS. Gel filtration fractions containing pure protein in the desired oligomeric state were pooled, concentrated and frozen in aliquots at −80 °C for subsequent analyses.

In vitro assembly of O432-17 nanoparticles

Purified tetramer and Fc or IgG from gel filtration fractions were assembled in a 1:1 molar ratio and with 1.1× excess trimeric plug. Molar ratios were calculated on the basis of the monomeric extinction coefficient. Assemblies were assembled in a total volume between 100 and 500 μl and a total concentration of each monomeric component between 5 and 50 μM, dialyzed overnight at 25 °C in 25 mM Tris pH 8.0 and 150 mM NaCl, and purified on a Superose 6 10/300 GL gel filtration column (Cytiva) using 25 mM Tris pH 8.0 and 150 mM NaCl as the running buffer.

Dynamic light scattering

DLS measurements were performed using the default sizing and polydispersity method on the UNcle (Unchained Labs). O432-17 variants (8.8 μl) were pipetted into glass cuvettes. DLS measurements were run in triplicate at 25 °C with an incubation time of 1 s; results were averaged across ten runs and plotted using Python3.7.

Negative-stain electron microscopy preparation and data collection of O432-17 nanoparticles and variants

Between 0.1 and 0.2 mg ml–1 of pre- or post-SEC O432-17 assemblies in 25 mM Tris pH 8.0 and 150 mM NaCl were applied onto 400- or 200-mesh carbon-coated copper grids and glow discharged for 20 s, followed by three applications of 3.04 μl 2% nano-W or Uranyless stain.

Micrographs were recorded using EPU software (Thermo Fisher Scientific) on a 120 kV Talos L120C transmission electron microscope (Thermo Fisher Scientific) at a nominal magnification of ×45,000 (pixel size, 3.156 Å per pixel) and a defocus range of 1.0–2.5 μm.

Negative-stain electron microscopy data analysis of O432-17 nanoparticles

Negative-stain electron microscopy datasets were processed using Relion3.0 software56. Micrographs were imported into the Relion3.0 software, and a contrast transfer function was estimated using GCTF57. Around 500 particles were manually picked and 2D classified, and selected classes were used as templates for particle picking in all images. Approximately 100,000 picked particles were 2D classified for 25 iterations into 50 classes.

Negative-stain electron microscopy data analysis of O432-17 nanoparticle variants

Negative-stain electron microscopy datasets were processed by CryoSPARC software v4.0.3 (ref. 58). Micrographs were imported into the CryoSparc software. Particles were picked using the blob picker in CryoSPARC and 2D classified, or 2D classes from blob picking were used as templates for particle picking in all images. All the picked particles were 2D classified for 20 iterations into 100 classes.

Cryo-electron microscopy preparation and data collection of O432-17 nanoparticles

Two microliters of pre-SEC-purified O432-17-Fc sample at 0.5 mg ml–1 in 25 mM Tris pH 8.0 and 150 mM NaCl were applied onto C-flat 1.2/1.3 holey carbon grids. Grids were then plunge-frozen into liquid ethane and cooled with liquid nitrogen using a Thermo Fisher Vitrobot Mk IV, with a 6.5-s blotting time and blot force of 0. The blotting process took place inside the vitrobot chamber at 22 °C and 100% humidity. Data acquisition was performed with Leginon on a Titan Krios electron microscope operating at 300 kV using a K3 summit direct electron detector equipped with an energy filter and operating in super-resolution mode. The nominal magnification for data collection was ×105,000 with a calculated pixel size of 0.42 Å per pixel. The final dose was 63.775 e– per A2 for 2,223 videos.

Cryo-electron microscopy data analysis of O432-17 nanoparticles

All data processing was done using CryoSPARC v3.0.0 (ref. 58). Alignment of video frames was performed using Patch Motion with an estimated B factor of 500 Å2, with the maximum alignment resolution set to 5. During alignment, all videos were Fourier cropped by one-half. Defocus and astigmatism values were estimated using patch CTF with default parameters. An initial 1,391 nanoparticle particles were manually picked and extracted with a box size of 576 pixels. This was followed by a round of 2D classification and subsequent template-picking using the best 2D class averages low-pass filtered to 20 Å. Particles were next picked with Template Picker and were manually inspected before extraction with a box size of 660 pixels, and were then further Fourier cropped to a final box size of 330 pixels, for a total of 66,904 particles. A round of reference-free 2D classification was performed in CryoSPARC, with a maximum alignment resolution of 6 Å. The best classes that revealed visibly assembled nanoparticles were used for 3D ab initio determination using the C1 symmetry operator. This was followed by a round of 3D heterogeneous refinement using C1 symmetry and sorting into four distinct classes, all of which revealed complete plugging of the octahedral three-component nanoparticle. Thus, all 50,017 of the best particles selected from 2D classification were subjected to non-uniform 3D refinement with octahedral symmetry applied, yielding a final map with an estimated global resolution of 7.06 Å, following per-particle defocus refinement. The final maps were deposited in the EMDB under accession code EMD-29602. Refined models were generated by rigid-body docking followed by refinement of the backbone of the design model into the final cryo-EM density map in Rosetta59.

Computational design of O432-17 electrostatically charged variants

A consensus design approach was used to first identify interior surface positions predicted to be the most robust to surface substitutions. These positions were divided into three tiers on the basis of the predicted enhancement to the design’s stability and/or solubility. Using the Rosetta modeling suite, the trimeric plug design model was redesigned, allowing optimization of the identities of interior surface residues that did not contribute to the interface between plug monomers or between the trimeric plug and antibody nanoparticle. We deployed three tiers of sequence-optimization strategies per charged variant. The first strategy enabled optimization of residue positions toward only charged amino acids (arginine or lysine for the positively charged variants, glutamate and aspartate for the negatively charged variants). The second strategy included the charged amino acids but also polar, non-charged amino acids, such as asparagine, glutamine and alanine, to maintain helical propensity. The third strategy included the residue identities in the first and second strategy, but also enabled reversion to the original identity in the trimeric plug design model. Substitutions that resulted in substantial losses of atomic packing interactions or side chain–side chain or side chain–backbone hydrogen bonds were discarded. The top-scoring designs for each design strategy and surface position tier were selected for inclusion as variant proteins.

Expression and purification of electrostatically charged trimeric plug variants

Plasmids encoding electrostatically charged trimeric plug variants and containing a C-terminal hexahistidine tag were cloned into pET-29b+ vectors (Genscript) and expressed overnight at 37 °C in autoinduction medium60. Cells were lysed by microfluidization or sonication in 25 mM Tris pH 8.0, 500 mM NaCl, 1 mM DTT, 1 mM PMSF and 0.1 mg ml–1 DNAse and cleared by centrifugation. Clarified lysates were filtered through 0.7-μm filters and purified by IMAC using gravity columns with nickel-NTA resin or HisTrap HP columns (Cytiva) using 25 mM Tris pH 8.0, 500 mM NaCl and 60 mM imidazole wash buffer and 25 mM Tris pH 8.0, 500 mM NaCl and 300 mM imidazole elution buffer. Elution fractions containing pure proteins were concentrated using centrifugal filter devices (Millipore) and further purified on a Superdex 200 10/300 GL (for large-scale purification) or Superose 6 10/300 GL (for comparison with assembly) gel-filtration column (Cytiva) using 25 mM Tris pH 8.0, 500 mM NaCl and 0.75% CHAPS. Gel filtration fractions containing pure protein in the desired oligomeric state were pooled, concentrated and frozen in aliquots at −80 °C for subsequent analyses.

Expression and purification of pos36GFP

pos36GFP plasmid (Addgene no. 62937) was transformed into BL21(DE3) bacteria cells in lysogeny broth and induced with isopropyl ß-d-1-thiogalactopyranoside (IPTG) at an optical density of 0.6 and grown for 16 h at 20 °C. Cells were lysed by microfluidization or sonication in 25 mM Tris pH 8.0, 500 mM NaCl, 1 mM DTT, 1 mM PMSF and 0.1 mg ml–1 DNAse and cleared by centrifugation. Clarified lysates were filtered through 0.7-μm filters and purified by IMAC using gravity columns with nickel-NTA resin or HisTrap HP columns (Cytiva) with 25 mM Tris pH 8.0, 500 mM NaCl and 60 mM imidazole wash buffer and 25 mM Tris pH 8.0, 500 mM NaCl and 300 mM imidazole elution buffer. Elution fractions containing pure proteins were concentrated using centrifugal filter devices (Millipore) and further purified on a Superdex 200 10/300 GL gel filtration columns (Cytiva) using 25 mM Tris pH 8.0, 500 mM NaCl and 0.75% CHAPS. Gel filtration fractions containing pure protein were pooled, concentrated and frozen in aliquots at −80 °C for subsequent analyses.

In vitro packaging of pos36GFP

O432-17(–) nanoparticles with pos36GFP cargoes were assembled in the same molar ratio of purified 1× tetramer, 1× Fc and 1.1× negatively charged trimeric plug from gel filtration fractions with 1× purified pos36GFP. Mixtures were dialyzed at 25 °C for 16 h into assembly buffer containing 25 mM Tris pH 8.0 and 200 mM NaCl, or 25 mM Tris pH 8.0 and 1 M NaCl. The ability of nanoparticles to package pos36GFP cargo was assessed using SEC on a Superose 6 10/300 GL column (Cytiva) in either buffer. Packaged GFP was quantified using the absorbance measurements at 280 and 488 nm obtained using a NanoDrop 8000 spectrophotometer. The absorbance of pure pos36GFP at 280 and 488 nm in 200 mM NaCl was used to calculate the absorbance at 280 nm due to pos36GFP in gel filtration fractions containing pos36GFP packaged in O432-17(–) nanoparticles. The relative absorbance due to pos36GFP and O432-17(–) nanoparticles was used to calculate the molar ratio of each protein using calculated extinction coefficients and was quantified against a standard curve generated using pure pos36GFP36.

In vitro packaging of pegRNA

O432-17(+) nucleocapsids with pegRNA (Integrated DNA Technologies) were assembled in a molar ratio of purified 1× tetramer, 1× CTX IgG and 1.1× positively charged trimeric plug from gel filtration fractions with 3× pegRNA per nucleocapsid. Mixtures were dialyzed at 25 °C for 16 h into 25 mM Tris pH 8.0 and 150 mM NaCl. Nucleocapsids were screened for in vitro encapsidation by native gel electrophoresis before and after a 1 h treatment with Benzonase or RNAse A.

Gel electrophoresis

Native agarose gels were prepared using 0.8% ultrapure agarose (Invitrogen) in TAE buffer (Thermo Fisher Scientific) containing SYBR Gold (Invitrogen). Nine microliters of O432-17(+) nucleocapsids were treated with 1 µl Benzonase (Invitrogen, diluted to 10 units per µl) or 1 µl RNAse A (Thermo Fisher Scientific, diluted to 10 units per µl) at 25 °C for 60 min, followed by mixing with 2 μl 6× loading dye (New England Biosciences, no SDS), and electrophoresed at 120 V for 30 min17,19. RNA in gels was imaged, and gels were subsequently stained with Gelcode Blue to analyze protein (Thermo Fisher Scientific).

Protein SDS–PAGE gels were performed using anyKD polyacrylamide gels (Bio-Rad) in Tris-glycine buffer and were electrophoresed at 180 V for 120 min.

In vitro assembly of O42.1 with pos36GFP

For each component, 1× O42.1 C4, 1× Fc and 1× pos36GFP per monomeric concentration were assembled and dialyzed for 16 h at 25 °C into 25 mM Tris pH 8.0 and 200 mM NaCl. The resulting assembly was screened using SEC on a Superose 6 10/300 GL column (Cytiva) in 25 mM Tris pH 8.0 and 200 mM NaCl.

Yeast display of Myc peptide

EBY100 S. cerevisiae cells were grown at 30 °C for 16 h in C-Trp-Ura medium supplemented with 2% (w/v) glucose, and were allowed to grow at 30 °C for 16 h in SGCAA expression medium to display myc-3×YPG peptide, followed by two washes in PBS + 3% BSA before incubation with O432-17 nanoparticle variants targeting the Myc peptide.

In vitro release of pos36GFP

O432-17(–) or O42.1 nanoparticles that had been in vitro assembled with pos36GFP were incubated with EBY100 S. cerevisiae cells expressing Myc-3×YPG peptide for 1 h at 25 °C. The cells were washed in 500 μl 25 mM Tris pH 8.0 and 500 mM NaCl until the supernatant fluorescence plateaued (after seven wash cycles) and were resuspended into citrate–phosphate buffer, pH 4.2, for 30 min at 25 °C. The supernatant was then collected and buffer exchanged using 25 mM Tris pH 8.0 and 500 mM NaCl before a fluorescence readout was generated using a Synergy Neo2 plate reader, with excitation wavelengths of 460 and 559 nm and emission wavelengths of 509 and 600 nm, and a gain of 120. Fluorescence was analyzed using GraphPad Prism version 9.3.1 for Windows, GraphPad Software (www.graphpad.com) and statistical significance was determined by two-way ANOVA with a 95% confidence interval and an alpha of 0.5.

Rational design of O432-17 pH-responsive variants

Bulky hydrophobic residues such as isoleucine and leucine amino acid positions within the pH-responsive trimeric interface were selected for optimization to either alanine or valine using the Rosetta software suite. The point substitution with the best scoring interface energy (ddG) at each position was selected as a trimeric plug variant. Combinatorial pairs of each point substitution variant were included as trimeric plug variants. We generated four O432-17 pH-responsive variants for both O432-17(–) and O432-17(+), two containing a third histidine hydrogen-bond network and either the p.I57V or p.L75A point substitution (3HIS_I57V and 3HIS_L75A), one containing a third histidine hydrogen-bond network and both p.I57V and p.L75A point substitutions (3HIS_ I57V_L75A) and one negative control containing zero histidine hydrogen-bond networks (0HIS), in which all histidines were mutated to asparagine.

AF647 conjugation to O432-17 trimeric plug variants

Trimeric plug variants containing a p.T359C substitution were generated by site-directed mutagenesis PCR. Alexa Fluor 647 C2 maleimide (Thermo Fisher Scientific) dissolved in DMSO and trimeric plug variants containing 10× TCEP were incubated in a 5:1 molar ratio with respect to the trimeric plug monomer for 16 h (overnight) at 4 °C in PBS + 0.75% CHAPS titrated to pH 7.2. The final reaction mixture contained 75 μM Alexa Fluor, 15 μM trimeric plug and 150 μM TCEP. The maleimide reaction was quenched with 1 mM DTT and buffer exchanged into 25 mM Tris pH 8.0 and 150 mM NaCl using PD-10 desalting columns with Sephadex-25 resin (Cytiva), according to the manufacturer’s protocol. It was then dialyzed for 3–4 days in 25 mM Tris pH 8.0, 150 mM NaCl and 0.75% CHAPS at 4 °C. Positively charged trimeric plug variants were buffer exchanged and dialyzed into 25 mM Tris pH 8.0, 500 mM NaCl and 0.75% CHAPS. The degree of labeling was estimated from a fluorophore extinction coefficient of 265,000 M−1 cm−1, and absorbance measurements were taken at 280 and 650 nm using a NanoDrop 8000 spectrophotometer.

Flow-cytometry-based pH titration

Linear Myc peptide with an N-terminal lysine side chain and 3× glycine linker (KGGGEQKLISEEDL) was produced through solid-phase peptide synthesis and biotinylated through amide formation. The resulting biotinylated Myc peptide was purified by reversed-phase high-performance liquid chromatography and quality checked for the proper molecular weight using liquid chromatography–mass spectrometry, lyophilized and dissolved in 100% DMSO for long-term storage. Then, 3.0–3.4 μm streptavidin-coated polystyrene particles (Spherotech) were incubated with biotinylated Myc peptide diluted in PBS to 5% DMSO 1 h at 25 °C. The antibody to Myc was purchased from Cell Signaling Technologies (cat. no. 9B11). The coated particles were washed in PBS + 3% BSA twice and split equally into pH-titrated citrate–phosphate buffer for 30 min at 25 °C. Coated particles were washed twice with PBS + 3% BSA and resuspended for flow cytometry. For O432-17(+) experiments, we prevented non-specific association with the polystyrene particles by using EBY100 S. cerevisiae cells.

All flow-cytometry experiments were performed on a LSR II Flow Cytometer (BD Biosciences). Lasers calibrated with coated particles were stained with either FITC-labeled anti-Myc-tag antibody (Abcam cat. no. 9E10), APC-labeled anti-Myc-tag antibody (Abcam cat. no. 9E10) or no antibody. Ten thousand events were collected per sample on three biological replicates. All flow-cytometry results were analyzed using FlowJo. Singlet beads were first isolated before analysis for AF647 or sfGFP signal. Normalization to the minimum and maximum fluorescence of each channel with each titration sample was performed in Python3.7, and the apparent pKa of AF647 and sfGFP fluorescence was estimated with a four-parameter non-linear logistic regression fit in GraphPad Prism version 9.3.1 for Windows (GraphPad Software).

Cells

Wild-type HeLa (ATCC CCL-2), EGFR-knockout HeLa (Abcam ab255385) and A431 cells (ATCC CRL-1555) were cultured at 37 °C with 5% CO2 in flasks with Dulbecco’s modified Eagle medium (DMEM) (Gibco) supplemented with 1 mM l-glutamine (Gibco), 4.5 g l–1d-glucose (Gibco), 10% fetal bovine serum (FBS) (Hyclone) and 1% penicillin–streptomycin (PenStrep) (Gibco). Wild-type HeLa and EGFR-knockout HeLa cells were also cultured with 1× nonessential amino acids (Gibco) supplemented in the medium. Cells were passaged twice per week. For passaging, cells were dissociated using 0.05% trypsin EDTA (Gibco) and split 1:5 or 1:10 into a new tissue-culture-treated T75 flask (Thermo Fisher Scientific cat. no. 156499). For serum starvation, cells were cultured in their respective cell medium without FBS for 16 h.

Immunostaining

For immunostaining, 35-mm glass-bottom dishes were seeded at a density of 20,000 cells per dish. A final monomeric concentration of 10 nM of O432-17-CTX or O432-17-Fc nanoparticles (417 pM final nanoparticle concentration) were incubated with cultured cells in DMEM or DMEM supplemented with 10% FBS. Cells were fixed in 4% paraformaldehyde, permeabilized with 100% methanol and blocked with PBS + 1% BSA. Cells were immunostained with anti-LAMP2A antibody (Abcam cat. no. ab18528) and/or anti-NaK ATPase (Abcam cat. no. ab76020, Abcam cat. no. ab283318), followed by Alexa Fluor 488-conjugated goat anti-rabbit-IgG secondary antibody (Thermo Fisher Scientific cat. no. A-11034) or Alexa Fluor 488-conjugated goat anti-mouse-IgG secondary antibody (Thermo Fisher Scientific cat. no. A-11029) and DAPI (Thermo Fisher Scientific cat. no. D1306), and were stored in the dark at 4 °C until imaging.

Nanoparticle uptake epifluorescence microscopy (A431 cells)

Cells were washed twice with FluoroBrite DMEM imaging medium and subsequently imaged in the same medium in the dark at room temperature. Epifluorescence imaging was performed on a Yokogawa CSU-X1 spinning dish confocal microscope with either a Lumencor Celesta light engine with seven laser lines (408, 445, 473, 518, 545, 635 and 750 nm) or a Nikon LUN-F XL laser launch with four solid state lasers (405, 488, 561 and 640 nm). A 40× objective with a numerical aperture (NA) of 0.95 and a Hamamatsu ORCA-Fusion scientific CMOS camera, both controlled by NIS Elements software (Nikon), were used. The following excitation (EX)/emission (EM) filter combinations (center/bandwidth in nm) were used: BFP, EX408 EM443/38; GFP, EX473 EM525/36; RFP, EX545 EM605/52; far red, EX635 EM705/72. The exposure times were 100 ms for the acceptor direct channel and 500 ms for all other channels, with no EM gain set and no neutral density filter added.

Nanoparticle uptake image acquisition in wild-type and EGFR-knockout HeLa cells

Four-color, 3D images were acquired with a commercial OMX-SR system (GE Healthcare). Toptica diode lasers with excitations of 405 nm, 488 nm and 640 nm were used. Emission was collected on three separate PCO.edge sCMOS cameras using an Olympus 60× 1.420-NA PlanApochromat oil immersion lens; 1024 × 1024 images (pixel size, 6.5 μm) were captured with no binning. Acquisition was controlled with AcquireSR Acquisition control software. Z-stacks were collected with a step size of 125 nm. Images were deconvolved in SoftWoRx 7.0.0 (GE Healthcare) using the enhanced ratio method and 200 nm noise filtering. Images from different color channels were registered in SoftWoRx using parameters generated from a gold-grid registration slide (GE Healthcare).

Nanoparticle uptake quantification

To quantify the percentage of cells that were bound to nanocages, the Multi-point tool in Fiji (ImageJ)61 was used to count nuclei in each image. The mRuby2 channel was used to identify and count cells, again using the multi-point tool, showing nanoparticle localization above the background signal using the identified nuclei as a guide for cell position. For cells positive for the LAMP2A nanoparticle signal, staining was used to aid in determining cell shape, and the freehand selection tool was used to draw manual regions of interest around the cell circumference. All regions of interest for positive cells were measured for cell area and integrated intensity in the mRuby2 channel for integrated intensity–cell area quantification. Cellular data were plotted and statistical tests were done in GraphPad Prism version 9.3.1 for Windows, GraphPad Software.

Statistics and reproducibility

Details on electron microscopy experiments can be found in Supplementary Table 4. All gel electrophoresis experiments were replicated in at least two biologically independent experiments.

Reporting summary

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