Design of functionalised circular tandem repeat proteins with longer repeat topologies and enhanced subunit contact surfaces

Constructs and nomenclature

The construct names, sequences, and figures for all constructs described in this paper are provided in Supplementary Table S1. The constructs described in this article (and listed in that Table) are referred to as ‘tcTRPs’ (‘thick circular Tandem Repeat Proteins’). Following the nomenclature for previous cTRP constructs described and used in a prior study⁵, the exact size and assemblages are further annotated as ‘tcTRP9’ and ‘tcTRP24’, respectively, where the underlying tcTRP9 and tcTRP24 scaffolds contain a total of 9 or 24 repeats. For scaffolds that are assembled from smaller identical protein subunits, the constructs are annotated as tcTRP9_x and tcTRP24_x, where each subunit contains ‘x’ repeats. For example, ‘tcTRP9₃’ (read as ‘tcTRP9 sub3’) refers to a particle containing a total of nine repeats, assembled from the trimerization of protein subunits containing 3 repeats each. Similarly, ‘tcTRP24₈’ corresponds to a particle containing a total of 24 repeats, assembled from the trimerization of protein subunits containing eight repeats each_. Some tcTRP24 constructs contain disulfide staples between protein subunits and are named ‘tcTRP24_xSS’.

Finally, constructs harboring additional protein cargo fused at defined positions around the tcTRP periphery are further denoted in the form ‘tcTRP24SS-Cargo_Y’, where ‘y’ is the total number of cargo domains per construct. One example in this paper (‘tcTRP24₈-SH2₆’) corresponds to a trimeric tcTRP containing a total of 24 repeats (‘tcTRP24…’), assembled from three identical subunits that contain 8 repeats each (‘tcTRP24₈…’), and with each subunit further displaying two SH2 peptide-binding domains, for a total of six such domains displayed (‘tcTRP24₈-SH2₆’).

Computational protein design

Protein design simulations were conducted exactly as described previously⁵. That approach corresponds to a geometry-guided repeat computational strategy implemented in the Rosetta package¹⁸ with additional de novo design elements¹⁷. Key features include the application of parametric symmetrization of backbone and side-chain conformations applied across all repeats (such that computational complexity scales only with repeat length); a pseudo-energy term that optimizes the inter-repeat geometry; clustering and resampling protocols that allow intensified exploration of promising topologies; and an in silico validation step that assesses sequence-structure compatibility by attempting to re-predict the designed structure given only the designed sequence. Applying this design procedure produces a diverse array of toroidal structures.

In this work, two additional modifications of the previously described approach were implemented: the ‘Ref2015’ energy function³¹ was used for all protein design and structure recapitulation calculations, and the range of allowed helix lengths was increased to 20−45 residues. Initial simulations explored helical linkers of length 1−5 residues with unconstrained backbone torsion angles. Clustering analysis of low-energy designs from these simulations revealed convergence on a 2-residue, antiparallel connection with backbone conformation ‘GB’ (one residue in a left-handed alpha-helical conformation and one residue in an extended conformation). A subsequent round of designs focused on ‘GB’ linkers was conducted to enhance sampling in this low-energy region of conformational space.

The identification of residue positions for the incorporation of disulfide staples into the tcTRP24₈ trimer was performed by utilizing the Rosetta ‘Disulfidize Mover’ routine³². Each edge helix involved in the trimerization was selected and corresponding residues scanned. The distance between adjoining beta-carbons was used to determine potential residues; once identified they were mutated to cystine residues and tested through rotamer optimization and energy minimization.

Protein expression and purification

All constructs encoding tcTRPs described in this study were designed and ligated into an in-house pET15HE expression vector³³ or a commercially available pET28b expression vector and sequence verified. The coding sequence and the corresponding translated protein sequences, including the N-terminal poly-histidine affinity tag and thrombin cleavage site preceding the first tcTRP repeat, are provided in Supplementary Table S1. The free SH2 domain was subcloned and purified as previously described⁵.

Plasmids were transfected into BL21(DE3)-RIL Escherichia coli cells (Agilent Technologies) and plated on LB medium augmented with 100 μg mL⁻¹ ampicillin. Protein was expressed via a previously described autoinduction protocol³⁴. Briefly, 1 L of ZYP-5052 media containing 100 μg mL⁻¹ ampicillin was inoculated with individual transformants, shaken at 37 °C for 8 h followed by 16 °C for 24 h. Expression cultures were pelleted by centrifugation and stored at −20 °C until purification.

Frozen cell pellets were thawed at room temperature and resuspended in 100 mL of 1× phosphate-buffered saline (‘PBS; 137 mM NaCl, 10 mM Na₂HPO₄, 2.7 mM KCl, pH 7.4.). PMSF was added to a final concentration of 0.5 μM. Cells were lysed via sonication and centrifuged in an SS34 rotor at 16,000 rpm for 20 min at 4 °C to remove cell debris. The supernatant was passed through a 5 μm filter, added to 2 mL of nickel-NTA metal affinity resin (Invitrogen) equilibrated with 1× PBS, and then incubated on a rocker platform at 4 °C for 1 h. After loading onto a gravity-fed column, the resin was washed twice with 25 mL of PBS containing 25 mM Imidazole. The protein was then eluted from the column by three additions of 5 mL 1× PBS containing 300 mM Imidazole. Fractions containing the eluted protein were pooled, concentrated, and buffer exchanged into 1× PBS. The sample was then filtered through a 0.2 μm filter and run over a size exclusion column (Cytiva HiLoad 16/60 Superdex 200) equilibrated in either 1× PBS or 20 mM Tris pH 7.5 + 150 mM NaCl.

Those constructs encoding tcTRPs trimers functionalized with three identical copies of VHH 72D61R (an anti-SARS-CoV-2 camelid nanobody targeting the viral receptor-binding domain or ‘RBD’)²³ were subcloned into a modified pET28 b(+) vector where the Kanamycin bacterial resistance gene is replaced with Ampicillin bacterial resistance gene. The constructs were sequence verified by Sanger sequencing. Protein expression was carried out using the method described above for tcTRPs expression, with the additional step of buffer exchange via dialysis from Ni-NTA elution buffer to 1×PBS, prior to use in assays.

Circular dichroism

Purified proteins were dialyzed overnight into 10 mM potassium phosphate buffer at pH 7.0, then diluted to 20.8 μM, as determined using the trimeric molecular weight of 166077 Daltons and extinction coefficient on a NanoDrop spectrophotometer (Thermo Fisher). Thermal denaturation experiments were performed on a JASCO J-815 spectrometer with a Peltier temperature controller. Wavelength scans from 190–250 nm were performed at 25 °C, 95 °C, and cooling back to 25 °C.

Crystallization and structure determination of tcTRP9 and tcTRP9₃

Both purified proteins were crystallized at 22 mg/mL with 100 mM sodium acetate pH 4.5 and 25% polyethylene glycol 400 in a 24-well hanging drop tray. Crystals were cryocooled in the same buffer via direct plunge into liquid nitrogen. Data was collected under cryocooled conditions (−150 °C) on a Saturn 944+ CCD area detector (Rigaku Inc.) using X-rays produced at 1.54 Å wavelength by a Rigaku HF-007 rotating anode generator. Data were processed using program HKL2000³⁵. Molecular replacement was performed using the de novo designed model of tcTRP9 using PHASER³⁶ in the PHENIX program suite³⁷. Refinement was done utilizing programs COOT³⁸ and REFMAC³⁹. Figures were generated using program PYMOL⁴⁰. The final Ramachandran distribution for backbone angles (favored, allowed, outliers) were 98.6, 1.4, 0% for tcTRP9 and 96.3, 2.5, 1.2% for tcTRP9₃ (Table 1).

Cryogenic electron microscopy (CryoEM) visualization of tcTRP24₈ and tcTRP24₈SS

Both purified proteins were screened with negative-stained transmission electron microscopy (TEM) using a 120 KV JOEL1400 electron microscope equipped with a 16 megapixel (4k × 4k) GATAN RIOL CMOS detector. The samples were prepared by depositing 4 μL of purified proteins at approximately 40 nM to the surface of a glow-discharged uniform carbon-coated grid. The particles were allowed to adsorb to the carbon film for ~ 1 min and washed three times with 20 ul of water and once with a drop of 0.7% uranyl formate followed by staining for 25 s with a 40 μL droplet of uranyl formate solution. Excess stains were wicked away with filter paper and the grids were air-dried overnight prior to analysis.

The tcTRP24₈SS particles were further analyzed by CryoEM. Samples were prepared by applying an aliquot of 3 μL protein sample of tcTRP24₈SS to a glow-discharged Quantifoil1.2/1.3 holey carbon grid, blotted with filter paper for 5 s and plunge-cooled in liquid ethane using an FEI Vitrobot Mark IV. Cryo-EM micrographs were collected on a 200 kV Glacios microscope (FEI) equipped with a Gatan K2 Summit direct detection camera. The microscope was operated at a calibrated magnification of 37,000×, yielding a pixel size of 1.16 Å on micrographs with an accumulated dosage of 60 e−/A²S. In total, 627 movies were collected from two screening sessions, including 82 at a tilt angle of 45°.

All data preprocessing, 2D classification, and 3D model generation and refinement, as well as post refinement polishing, were performed using the software package CryoSPARC2⁴¹. For each movie stack, the frames were aligned for beam-induced motion correction using Patch-motion-correction. Patch-CTF was used to estimate the contrast transfer function (CTF) parameter. A new ring-shape algorithm with inner/outer diameters of 100/120 was used for automated blob picking. After inspection and local motion correction, 627763 particles were accepted for reference-free 2D classification. Two consecutive runs of 2D classification/selection were used to root out false positive and bad (overlapping) particles. A total of 121426 particles in 20 classes were used for ab initio 3D reconstruction.

It is obvious from the selected classes that there were at least two populations of particles with different diameters. Three models were requested for ab initio 3D reconstruction. Results from 3D reconstruction showed multiple circular-disk particles with different diameters. The proportion of the three 3D classes varied with the number of consecutive 2D classification/select and images selected. Multiple trials were performed with different particle picking protocols and particle diameters. All approaches yielded similar results.

Peptide binding assays in solution via fluorescence polarization

A 10-residue peptide, Tir10, containing a phosphorylated tyrosine (‘pY’), was chemically synthesized with a FITC tag at the 5′-end linked to the peptide with a 7 atom aminohexanoyl space, Ahx (GenScript).

Tir10 stock was re-suspended to 5.7 mM in DMSO, then diluted to 0.5 μM in fluorescence polarization (‘FP’) Buffer (20 mM HEPES, 150 mM KCl, pH 7.4). Proteins were exchanged into FP Buffer then two-fold serially diluted from 23 to 0.01 μM (tcTRP24₈SS and free SH2) or 34 to 0.02 μM (tcTRP24₈SS-SH2₂). Diluted proteins were mixed with Tir10 at a ratio of 9:1 for final concentrations of 21–0.01 μM or 31–0.015 μM protein, respectively, and 0.05 μM Tir10, then incubated, shielded from light, at room temperature for 20 min. FP values were read at excitation of 485 nm and emission of 525 nm (SpectraMax M5). After subtracting FP buffer only background from the raw perpendicular (S) and parallel (P) measurements, polarization (mP) and anisotropy (r) were calculated with the following equations:

Peptide binding assays on a surface via surface plasmon resonance (SPR)

SPR experiments were performed at 25 °C on a Biacore T100 instrument (Cytiva) with a Series S SA chip using a running buffer of 10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.05% surfactant P20 with 0.1 mg/mL bovine serum albumin. Biotinylated Tir-10_v2 (Biotin-Ahx-EHI-pY-DEVAAD) at 10 ng/mL was injected over one flow cell at a flow rate of 10 μL/minute to capture ~15 RUs of peptide. A flow cell with streptavidin alone was used as a reference surface. Analytes were repurified by SEC just prior to use. Buffer blanks and analytes (10 nM tcTRP24₈SS, 60 nM free SH2, and 10 nM tcTRP24₈SS-SH2₂) were injected at a flow rate 50 μL/minute with 7 min of association and 10 min of dissociation. An overlay plot of double-referenced data was generated, then normalized for off-rate comparison by dividing each curve by its maximum response in Scrubber 2.0b software (BioLogic Software). Maximum binding responses observed were 125 and 123 RUs for free SH2 and tcTRP24₈SS-SH2₂, respectively. The tcTRP24₈SS control did not bind. The figure was made in Prism 7 (GraphPad) for Mac OS X version 7.0d.

SARS-CoV-2 pseudoviral neutralization assays

The SARS-CoV-2 pseudotyped lentiviral particle generation, tittering and neutralization assays were performed as previously described protocol⁴² with minor modifications, as described here. Poly-L-lysine coated clear bottom 96-well black plates (Thermo Scientific, 12-566-70) were seeded with 293-ACE2 cells (provided by the Jessie Bloom lab at the Fred Hutchinson Cancer Research Center) at a density of 1.25 × 10⁴ cells per well in 50 µL volume. Twelve hours after seeding, five-fold antibody dilutions, starting at 50 µg/mL, were prepared. Control, virus only and cell only samples were prepared as previously reported. 60 µL of the titered virus was added and mixed with antibody dilutions and virus only wells. The mix was incubated at 37 °C for 1 h. 100 µL of the mix from each well was added to the corresponding well on 293-ACE2 seeded plates. Polybrene (Sigma Aldrich, P4707) was added to each well at a final concentration of 5 µg/mL. Plates were incubated at 37 °C for 60 h post infection. Virus neutralization was assessed by measuring luminescence. While incubating, Bright-Glo Luciferase reagents (Promega, E2610) were thawed, equilibrated at room temperature, and prepared following the manufacturer’s recommendation. 100 µL of growth media was removed from each well and 30 µL per well of luciferase reagent added. Plates were incubated for 2 min at room temperature in the dark and luminescence was measured using an M2 plate reader (Molecular Devices). Luminescence RLUs from virus only wells were normalized as 100% infectivity and RLUs from cells only were normalized as 0% infectivity. Infectivity and IC50 were calculated using Four Parameter Logistic Regression on GraphPad Prism (GraphPad Software).

ELISA based trimer VHH binding to SARS-CoV-2 RBD

To assay trimeric VHH binding to RBD, ELISA-based binding assays were performed using the following protocol. High-affinity ELISA plates (Greiner Bio-one, Catalog Number 655084) were coated with SARS-CoV-2 RBD (Roland Strong, Fred Hutchinson Cancer Research Center) at 1 µg/mL in bicarbonate buffer (Sigma, Catalog Number C3041-100CAP) overnight at 4 °C. Coated plates were washed three times with ELISA wash buffer (1XPBS, (Fisher BioReagents, Catalog Number B399-4), supplemented with 0.05% Tween-20 (Thermo Fisher Scientific, Catalog Number B2337-500)). ELISA plates were washed on AquaMax 2000 Microplate Washer (Molecular Devices). Washed plates were blocked with ELISA blocking buffer (ELISA wash buffer supplemented with 5% Non-fat dried milk) for 2 h at room temperature. While blocking, VHH dilutions were prepared as follows. Trimeric VHH 72D61R were diluted with ELISA blocking buffer in 4-fold serial dilution starting at 10 µg/mL. ELISA blocking solutions were aspirated, and 100 µL of antibody dilutions were added into each corresponding well. Plates were incubated at room temperature for 60 min and washed three times with ELISA wash buffer.

The detection antibody, HRP conjugated MonoRab^TM Rabbit Anti-Camelid VHH Cocktail (GenScript, A02016) was diluted 1:10,000 in ELISA blocking buffer. The detection antibody was added at 100 µL/well and plates were incubated for 30 min at room temperature then washed two times with ELISA wash buffer followed by a wash with 1×PBS. Peroxidase activity was measured using chemiluminescence by adding 100 µL Sera Care KPL TMB Microwell Peroxidase Substrate (Sera Care Life Sciences Inc., 5120-0047) following the manufacturer’s recommendation. Peroxidase activity was quenched after 5 min incubation at room temperature by adding 50 µL per well 1 M Hydrochloric acid. Chemiluminescence was measured on M2 plate reader (Molecular Devices) at 450 nm wavelength using SoftMax software (Molecular Devises). Binding EC50 was calculated using averages of replicates and Four Parameter Logistic Regression on GraphPad Prism (GraphPad Software).

Bio-Layer Interferometry (BLI) based Trimer VHH binding to SARS-CoV-2 RBD

BLI measurements were performed on the Octet RED96 system (ForteBio) using High Precision Streptavidin (SAX) Biosensors (ForteBio). Biosensors were hydrated with phosphate-buffered saline (PBS) at pH 7.4 at room temperature for 10 min in 96-well flat-bottom microplate (Greiner, 655209). All kinetics experiments were performed at 30 °C with 1000 rpm agitation in the kinetics module. Biosensors were dipped into PBS containing wells for 60 s prior to antigen loading. Biosensors were loaded with enzymatically biotinylated RBD (provided by the Roland Strong Lab at the Fred Hutchinson Cancer Research Center) at 1 µg/mL in phosphate buffer, pH 7.4 for 300 s to achieve ~0.6–1 nm response. Loading was quenched by incubating biosensors in 50 µM Biocytin (Sigma Aldrich, 576-19-2) for 60 s. Baseline were established by incubating antigen-loaded biosensors in kinetics buffer (PBS + 0.02% Tween 20, 0.1% BSA, 0.05% Sodium azide) for 120 s. Following baseline measurements to determine the rate of association, antigen-loaded biosensor tips were dipped for 50 s into three-fold dilution series of trimeric tcTRP9₃-VHH₃ fusions starting at a protein concentration corresponding to approximately 100 nM. Analyte bound biosensors were dipped into kinetics buffer for 120–300 s to measure the rate of dissociation. Kinetic analyses were performed using the HT 11.1.1.39 Data Analysis module (ForteBio). Results were double referenced. The association and dissociation steps were both used in a 1:1 binding model with global fitting.

Statistics and reproducibility

Biochemical experiments reporting binding interactions (flourescence polarization (FP) assays for the SH2-tcTRP fusions in Fig. 7; BLI and ELISA assays for VHH-tcTRP constructs in Fig. 8 and viral pseudoneutralization assays in Fig. 8) were conducted in triplicate using independent aliquots of each protein. Data are shown as mean and standard for n = 3 measurements.

Reporting summary

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