Adult Puzzle Books

DNA manipulation

Plasmid design for PmCPS genes

All PmCPS genes were codon optimized for E. coli expression, synthesized by Bio Basic and cloned into pETDuet and pACYCDuet plasmid backbones, giving rise to expression plasmids 5 and 6, pETDuet Hex Secretion System and pACYCDuet Biosynthesis components, respectively (see ‘Plasmids’ in the Supplementary Information). Plasmid 6 encodes KpsC (with an N-terminal Strep-Tag II), KpsS (N-terminally 8×His-tagged), HyaE (N-terminally Myc-tagged), HyaB (N-terminally Flag-tagged) and HyaD (N-terminally 8×His tagged), each under a separate T7 promoter and lac operator. Plasmid 5 encodes KpsD (with a C-terminal Strep-Tag II), KpsE (N-terminal S-tag), KpsM (untagged) and KpsT (C-terminally 8×His tagged), each under a separate T7 promoter and lac operator. In both plasmids, each CPS gene is flanked by a unique restriction site. Single-gene deletions for in vivo encapsulation assays were achieved by restriction enzyme (NEB) digest and religation using T4 ligase (NEB). The restriction enzymes used in single-gene deletions from plasmid 6 were NcoI (KpsC), BamHI (KpsS), SacI (HyaE), EcoRI (HyaB) and XhoI (HyaD); and for plasmid 5 they were NdeI (KpsD) and BamHI (KpsE).

Two-plasmid genome-editing system

Plasmid design was based on a previous study⁴², in which efficient E. coli genome editing was achieved using a two-plasmid approach: plasmid 7 (pCasJZ) and plasmid 8 (pUC57_region1_tet). Plasmid 7 encodes SpCas9 under an arabinose promoter, the red λ phage proteins Gam, Beta and Exo under a T7 promoter, the SacB cassette and the chloramphenicol resistance cassette. Plasmid 8 encodes two 500-bp-long homology regions (1 and 2), flanked by protospacer adjacent motifs (PAMs) (1 and 2) and target sequences (1 and 2), two guiding RNA sequences with target sequences (1 and 2) and the tetracycline resistance locus. PAMs 1 and 2 and the target sequences 1 and 2 were chosen to be upstream and downstream, respectively, of CPS region 1 in the E. coli C43 genome. Homology regions 1 and 2 are 500 bp sequences upstream and downstream of target sequences 1 and 2.

Plasmid 7 was generated using the Gam, Beta, Exo, λ tL3 terminator, SacB, SacB promoter, araBAD promoter and SpCas9 elements from plasmid pCasPA (Addgene 113347), which were cloned by Gibson Assembly (NEB) into MCS-1 of the pACYCDuet plasmid. Next, araBAD and araC were cloned in place of the deleted MCS-II of that plasmid. The homology regions, guiding RNAs, PAMs and target sequences were synthesized by Gene Universal into the pUC57 vector. Then, using Gibson assembly, the ampicillin resistance cassette was replaced with the tetracycline resistance cassette from pBR322 plasmid, giving rise to plasmid 8.

Genome editing

The procedure has been described in detail elsewhere⁴². In brief, plasmids 7 and 8 were co-transformed into E. coli C43 (DE3). The working concentrations of isopropyl-β-d-thiogalactopyranoside (IPTG), arabinose, chloramphenicol, tetracycline, glucose and sucrose were 1 mM, 20 mM, 25 mg l⁻¹, 10 mg l⁻¹, 1% and 2%, respectively. Cells were plated on LB agar plates, supplemented with chloramphenicol, tetracycline and glucose. Single colonies were used to inoculate 2 ml of LB medium supplemented with chloramphenicol, tetracycline and glucose. After 2 h of growth at 37 °C, IPTG was added, and the expression of red λ phage proteins was induced. After an another 1 h, arabinose was added and the expression of SpCas9 and transcription of sgRNAs was induced. After 3 h, cells were plated on LB agar plates containing chloramphenicol, tetracycline and arabinose and incubated overnight at 37 °C. Positive colonies were verified by colony PCR using primers flanking the 500-bp-long homology regions and sequencing (Supplementary Figs. 2 and 3). Then, the positive clone was grown in LB containing sucrose and tetracycline to remove plasmid 7. No chloramphenicol-resistant colonies were detected. Genome-edited C43 cells lacking the CPS1 region, termed C43ΔCPS1, were made electrocompetent and used in the in vivo encapsulation assay.

Plasmid design for CBM70 and SNAP–CBM70

The codon-optimized gene encoding Streptococcus pneumoniae carbohydrate-binding module 70 (CBM70)²⁰, was synthesized by Bio Basic and cloned into the pET30 vector with a C-terminal 10×His tag, generating plasmid 9. Then, using Gibson Assembly, the SNAP tag from the pSNAP-tag vector (Addgene 101135) was N-terminally fused to CBM70 with a GSSMGS linker, creating plasmid 10.

Plasmid design for StKpsMT-E-D

The codon-optimized KpsD, KpsE, KpsM and KpsT genes from S. thermodepolymerans were synthesized and cloned into expression vectors by Gene Universal, generating plasmids 1a, 3a and 4. Plasmid 1a contains KpsM in MCS-I and KpsT (C-terminal Flag) in MCS-II in a pETDuet backbone. Plasmid 3a contains KpsE (C-terminal 10×His tagged) in MCS-I in a pCDFDuet backbone. Plasmid 4 contains KpsD in a pET30 backbone. To increase the expression yields of KpsM and KpsE, we introduced three amino acids as the third to the fifth residues of the polypeptide chains (K₃-I₄-H₅) (using polymerase incomplete primer extension (PIPE) cloning) that were shown to increase translation initiation⁴³. KpsE was further modified by introducing two cysteines at positions 77 and 138, giving rise to plasmids 1b and 3b. Next, using Gibson assembly, KpsD was inserted into MCS-II of plasmid 3b, creating plasmid 3c, used for in vivo encapsulation assays.

Mutagenesis

Mutagenesis was performed by PIPE cloning with overlapping primers on plasmids 1b and 3b, using Phusion HF DNA polymerase (NEB), resulting in plasmids 17–24.

Protein and CPS expression

Bacterial growth

All bacterial cultures described in this work were grown at 37 °C and with shaking at 220 rpm, unless noted otherwise. Working concentrations: ampicillin 100 mg l⁻¹, kanamycin 50 mg l⁻¹, streptomycin 50 mg l⁻¹ and chloramphenicol 25 mg l⁻¹. Appropriate plasmids were transformed into C43 cells (for protein purification) or C43ΔCPS1 cells (for in vivo encapsulation, MINFLUX and CPS purification) for overnight growth in the presence of suitable antibiotics. All collected cell pellets in this study were flash-frozen in liquid nitrogen and stored at −80 °C for further use.

For western blotting: For expression testing, an overnight starter culture of cells expressing all PmCPS components was used to inoculate 1 l of LB medium supplemented with ampicillin and chloramphenicol. At an optical density at 600 nm (OD₆₀₀) of 0.6, protein expression was induced with 100 mg l⁻¹ of IPTG. Growth was continued for another 3–4 h, after which cells were collected (4,500 rpm for 20 min) and flash-frozen in liquid nitrogen. This cell pellet was used to prepare inverted membrane vesicles, as described previously⁴⁴. The inverted membrane vesicles were then run on a 12.5% polyacrylamide gel and analysed using the western blot technique detecting the engineered affinity tags, as described previously⁴⁵.

For in vivo encapsulation assays and spheroplasting: 20 ml LB culture was inoculated with a single stab of the appropriate transformants and grown in the presence of appropriate antibiotics and 100 mg l⁻¹ IPTG. Growth was carried out for 6–8 h, after which cells were collected. Spheroplasts were prepared as described previously⁴⁶. After removal of the outer membrane, spheroplasts were resuspended in PBS supplemented with 200 mM sucrose, and used for in vivo encapsulation assays.

For purification of CPS: 8× 1 l of 2× LB supplemented with appropriate antibiotics were inoculated from an overnight starter culture. At an OD₆₀₀ of 0.6, the medium was cooled to 20 °C, and then cells were induced using 100 mg l⁻¹ of IPTG, grown overnight and collected.

For purification of the CBM70, SNAP–CBM70 and KpsMT–KpsE proteins: 6× 1 l of 2× LB supplemented with appropriate antibiotics were inoculated from an overnight starter culture. At an OD₆₀₀ of 0.6, protein expression was induced using 200 mg l⁻¹ of IPTG, and cells were grown for another 3–4 h at 37 °C, after which they were collected.

In vivo encapsulation assay

A total quantity of 200 μl of cells at OD₆₀₀ of 4 was washed with ice-cold PBS three times and then incubated on ice with 10 μl of 2 mg ml⁻¹ of ^Alexa647SNAP–CBM70 or ^Flux680CBM70 (see below) for a total of 2 h. For CPS digestion, samples were treated with 1 mg ml⁻¹ of bovine testicular hyaluronidase (MP Biomedicals) for 2 h on ice. For two-colour MINFLUX nanoscopy, 12.5 μl of ^Alexa680Streptavidin (Thermo Fisher Scientific) was added after 1 h of incubation, followed by incubation on ice for 30 min. Then 2 µl of 100× Cellbrite Fix 488 (Biotum) was added for another 30 min, after which cells were washed three times in 1,000 µl of ice-cold PBS, fixed with 4% PFA (Electron Microscopy Sciences) in a total of 1,000 µl PBS for 20 min, blocked with 50 mM NH₄Cl in PBS for 30 min, resuspended in 200 µl ice-cold PBS and imaged.

Confocal microscopy

Imaging was performed on a Zeiss LSM880 confocal microscope with an Airyscan detector, at 40× using a water-immersion objective. Membrane and CPS channels were recorded sequentially using 488-nm and 633-nm excitation lasers, respectively, and suitable filter sets. The pixel size was set to 52 nm. Images were processed in ImageJ–Fiji⁴⁷.

Metabolic labelling of LPS with AZ6470 DBCO

Metabolic labelling of the LPS was achieved following the protocol described previously²³. In brief, cells expressing all necessary components for CPS biosynthesis were grown for 8 h at 37 °C in the presence of 1 mM Kdo azide (Click Chemistry Tools) and 100 mg l⁻¹ IPTG. After that, cells were washed three times and resuspended in M9 medium. Then AZDye647 DBCO (Click Chemistry Tools) was added to a final concentration of 0.5 µM for the copper-free click chemistry reaction. Cells were incubated in the dark for another 1 h at 37 °C with shaking. Finally, the cells were washed three times with PBS, aliquoted, flash-frozen in liquid nitrogen and stored at −80 °C.

Sample mounting, imaging buffer and nanoscopy for MINFLUX

Fixed cells were applied on glass slides precoated with poly-l-lysine. Samples were mounted in the imaging buffer as described⁴⁸. In brief, gold nanoparticles (Nanopartz, A11-200-CIT-DIH-1-10) were used as fiducials. GLOX buffer supplemented with 10–14 mM MEA (Cysteamine) was used as imaging buffer. Samples were sealed with EliteDouble22 (Zhermack).

MINFLUX nanoscopy and corresponding confocal microscopy were performed using an in-house MINFLUX set-up⁴⁸. In two-colour MINFLUX, a single event originates from one of two different red fluorophores (Alexa 647/AZdye647 or Alexa 680/Flux680) and is split into Cy5-near and Cy5-far detectors. The number of photons from this single event reaching both detectors is represented as a detector channel ratio (DCR) and is characteristic for each emitter. Experimental DCR values were separately acquired for each fluorophore, and then used to assign the colour of fluorophores to localizations in two-colour MINFLUX experiments. Pixel size in rendered images was based on localization precisions of raw burst.

Protein purification

When possible, all purification steps were performed at 4 °C.

CBM70 and SNAP–CBM70 purification

Cell pellets were thawed and resuspended in 10 % glycerol, 100 mM NaCl, and 20 mM Tris pH 7.5, then incubated for 1 h with 1 mg ml⁻¹ lysozyme. PMSF (1 mM) was added and cell suspensions were lysed by three passes through a microfluidizer. Intact cells were removed by low-speed centrifugation for 25 min at 12,500 rpm, in a JA-20 rotor (Beckman). The supernatant was centrifuged for 1 h at 200,000g in a Ti45 rotor (Beckman) and the insoluble material was discarded. The supernatant was spiked with 20 mM imidazole and incubated with Ni-NTA resin for 1 h with agitation. The resin was washed with (1) 1 M NaCl, PBS (pH 7.4) and 40 mM imidazole and (2) PBS and 60 mM imidazole. Protein was eluted after a 30-min incubation in PBS containing 320 mM imidazole, and was concentrated using a 10-kDa filter (Amicon) to 1 ml. Then, in the case of CBM70, the sample was run over an S200 16/60 gel filtration column equilibrated in PBS and the peak fractions were collected and used for αHA column preparation, or aliquoted and flash-frozen for storage; or, in the case of SNAP–CBM70, dialysed against PBS overnight, aliquoted, flash-frozen in liquid nitrogen and stored at −80 °C.

Purification of KpsMT-KpsE

Cell pellets were processed as described above. After low-speed centrifugation, membranes were isolated from the lysate by centrifugation for 2 h at 200,000g in a Ti45 rotor, then collected, flash-frozen in liquid nitrogen and stored at −80 °C.

Membranes were thawed and resuspended in 300 mM NaCl, 20 mM Tris pH 7.5, 10 % glycerol, 40 mM imidazole, 1 % n-dodecyl-β-maltoside (DDM) and 0.1 % cholesterol hemisuccinate and incubated for 1 h with agitation. Aggregated material was removed by centrifugation at 200,000g for 30 min, and the supernatant was incubated with Ni-NTA resin for 1 h. The resin was washed with (1) 1.5 M NaCl, 20 mM Tris pH 7.5, 10 % glycerol, 40 mM imidazole and 0.1% LMNG and (2) 300 mM NaCl, 20 mM Tris pH 7.5, 10% glycerol, 80 mM imidazole and 0.1% LMNG. Protein was eluted after a 30-min incubation in 300 mM NaCl, 20 mM Tris pH 7.5, 5% glycerol, 400 mM imidazole and 0.05% LMNG, and was concentrated to 500 µl using a 100-kDa filter (Amicon), followed by overnight incubation on ice. The next day, the sample was run over a S6-increase 10/300 gel filtration column equilibrated in 100 mM NaCl, 50 mM Tris pH 7.5 and 0.025% LMNG. This buffer was supplemented with 5 mM MgCl₂ for KpsMT(E151Q)-KpsE preparations. The peak fractions were collected and concentrated to 2–3 mg ml⁻¹ using a 100-kDa filter, and were used for grid preparation or in vitro ATPase activity assays.

KpsMT-KpsE in complex with ADP–AlF₄⁻ was purified similarly. The concentrated Ni-NTA elution sample was dialysed overnight against buffer containing 100 mM NaCl, 50 mM Tris pH 7.5, 0.05% LMNG, 5% glycerol, 10 mM NaF, 2 mM AlCl₃, 5 mM ADP and 5 mM MgCl₂. The same buffer containing 0.025% LMNG and lacking glycerol was used to equilibrate the S6-increase gel filtration column. Peak fractions were collected on the basis of elution times, concentrated to around 2–3 mg ml⁻¹ and used for cryo grid preparation.

Preparation of Alexa647-tagged SNAP–CBM70 and Flux680-tagged CBM70

All steps were performed in the dark at 4 °C. SNAP–CBM70 aliquots were thawed and mixed with dimethyl sulfoxide (DMSO)-solubilized Alexa647 (NEB) in a 1:1 molar ratio, in the presence of 1 mM DTT. The sample was incubated with agitation for 6–8 h and run over an S200 10/30 gel filtration column equilibrated in PBS to separate ^Alexa647SNAP–CBM70 from the free dye. Peak fractions with string absorbances at 280 and 671 nm were aliquoted at 2 mg ml⁻¹, flash-frozen in liquid nitrogen and stored at −80 °C.

^Flux680CBM70 (lacking the SNAP domain) was prepared in a similar manner. CBM70 aliquots were thawed and mixed with DMSO-solubilized Flux680-maleimide (Abberior) in a 1:3 molar ratio, in the presence of 1 mM DTT. The sample was incubated with agitation for 16 h and run over an S200 10/30 gel filtration column equilibrated in PBS to separate ^Flux680CBM70 from the free dye. Peak fractions with string absorbances at 280 and 695 nm were aliquoted at 2 mg ml⁻¹, flash-frozen in liquid nitrogen and stored at −80 °C.

ATPase activity assays

The ATPase activity of KpsMT-KpsE was quantified using an enzyme-coupled assay as previously described³². The peak fraction of the complex eluting from a S6-increase column was concentrated to 0.5–1 mg ml⁻¹ and used for activity assays. ATPase activity was initiated by adding ATP, and the depletion of NADH was monitored at 340 nm for 1 h at 27 °C in a SpectraMax M5 plate reader. The rate of NADH depletion was converted to nmol ATP hydrolysed using an ADP standardized plot. The data were processed in Microsoft Excel and GraphPad Prism. All experiments were performed at least in triplicate and error bars represent deviations from the means.

Anti-hyaluronan affinity column preparation

Purified CBM70 was coupled to NHS-activated Sepharose 4 Fast Flow beads (Cytiva) following the manufacturer’s protocol. In brief, the resin was washed with (1) MQ water, (2) 1 mM HCl and (3) PBS. Then, 20 ml of 5 mg ml⁻¹ CBM70 in PBS was mixed with 25 ml of the washed resin, and left agitating for two days at 4 °C. After that, the liquid was drained from the beads, and the resin was washed with PBS, followed by blocking buffer (PBS containing 200 mM ethanolamine) for 24 h with agitation at 4 °C. The beads were washed and stored in 20% ethanol at 4 °C.

CPS purification

Cell pellets and membranes were prepared as described above. Membranes were resuspended in PBS containing 1% LMNG and incubated for 1 h at room temperature with agitation. Aggregated material was removed by centrifugation at 200,000g for 30 min, and the supernatant was incubated with anti-HA resin (see above) for 1 h. Next, the beads were washed three times with PBS containing 0.1 % LMNG. The CPS was eluted from the column after a 30-min incubation with 2 M NaCl, 100 mM sodium citrate pH 3.0 and 0.01 % LMNG, concentrated to 250 µl using a 3-kDa filter (Amicon) and dialysed overnight against PBS in a 3.5-kDa dialysis membrane. The next day, the sample was run on a 1.5% agarose gel (Ultra-pure agarose, Invitrogen) or a 4–20% gradient polyacrylamide gel (Bio-Rad), and stained with Stains-All dye (Sigma) as described⁴⁹. The obtained CPS sample was also used for cryo-EM analyses.

HA ELISA assay

The total amount of cell-surface-exposed HA was quantified using an ELISA-based kit (Echelon Biosciences) following the manufacturer’s protocol. Cell densities were adjusted on the basis of OD₆₀₀. HA was quantified on chemiluminescence using a Promega GloMax plate reader. The data were processed in Excel and GraphPad Prism. All experiments were performed at least in triplicate.

Grid preparation and data collection

To obtain the ATP- and glycolipid-bound states, KpsMT(E151Q)-KpsE and wild-type KpsMT-KpsE were supplemented with 2 mM ATP or 30 µl of lipid-linked HA, respectively, before grid preparation. Quantifoil holey carbon grids (Cu 1.2/1.3, 300 mesh) were glow-discharged in the presence of two drops (about 200 µl) of amylamine. Four microlitres of sample was applied, blotted for 4–10 s with a blot force of 4–7 at 4 °C and 100% humidity, then plunge-frozen in liquid ethane using a Vitrobot Mark IV (FEI).

Cryo-EM data were collected at the University of Virginia Molecular Electron Microscopy Core (MEMC) on a Titan Krios (FEI) 300-kV electron microscope using a Gatan Imaging Filter (GIF) and a K3 direct electron detection camera. Movies were collected in EPU (Thermo Fisher Scientific) at a magnification of 81,000× with an energy filter width of 10 eV, using counting mode with a total dose of 51 e⁻ per Å over 40 frames, and with a target defocus of −1.0 to −2.0 μm.

Cryo-EM data processing

All datasets were processed in cryoSPARC (versions 3.3 and 4.0)⁵⁰. Raw movies were subjected to patch motion correction and patch contrast transfer function (CTF) estimation. For all four datasets, particles were automatically selected by blob picker to generate initial templates, followed by template picker. After several rounds of 2D classification, selected particles were used for ab initio reconstructions in C1, followed by heterogenous refinement.

For dataset 1 yielding the Apo 1 state, C2 symmetry was applied in both non-uniform and local refinements. To improve the KpsT density, focused three-dimensional (3D) classification followed by non-uniform and local refinement was applied. Using the Phenix Combine Focused Maps job⁵¹, a composite map 1 was created from maps A and B, on the basis of the model and half maps from the focused refinements.

For dataset 2 yielding the ATP-bound state, C2 symmetry was applied during non-uniform refinement, followed by local refinement in C1. To improve the density of the crown region of KpsE, focused 3D classification followed by non-uniform and local refinement was applied. Using Phenix Combine Focused Maps job, a composite map 2 was created from maps A and B, on the basis of the model and half maps from the focused jobs.

For dataset 3 yielding the ADP–AlF₄⁻-bound state, C2 symmetry was applied in non-uniform refinement.

For dataset 4 yielding the glycolipid 1 and 2 and Apo 2 states, C1 symmetry was applied in both non-uniform and local refinements. Next, the particles were 3D classified into seven classes using seven identical Apo 1 volumes as input. Three of the resulting classes (class 0, 1 and 4) revealed distinct states and were subjected to non-uniform refinement. Classes 0 and 1 had noticeable extra density in the polysaccharide canyon. To improve the KpsT density in classes 0 and 1, a separate 3D classification focused on KpsT was performed, resulting in improved KpsT density for the glycolipid 1 and 2 states (respective map A for both classes), enabling rigid-body docking of an AlphaFold2-predicted KpsT model. Particles from classes 0 and 1 were also subjected to 3D classification focused on the KpsM subunits, which resulted in improved glycolipid density maps (respective map B for both classes). Then, the maps focused on KpsT and KpsM (respective maps A and B for both classes) were combined, resulting in composite maps 4 and 5 for glycolipids 2 and 1, respectively.

For this dataset, class 4 from the original 3D classification job revealed a novel arrangement of the KpsM transmembrane helices. As for classes 0 and 1, the KpsT map quality was improved by 3D focused classification, followed by non-uniform refinement. The improved map allowed rigid-body docking of a KpsT model. The focused maps A and B were combined using Phenix Combine Focused Maps job, resulting in map 6. Maps were sharpened on the basis of models using either the autosharpen (maps 1, 2, 3 and 6) or the local anisotropic sharpening (maps 4 and 5) jobs in Phenix:refine. Half maps were used to generate global-resolution estimates using EMBL’s Fourier shell correlation (FSC) server, and local-resolution estimates using cryoSPARC’s local resolution estimation job.

In all datasets, we observed a minor particle population with an incomplete KpsE cage. In these cases, the disordered eighth KpsE subunit is proximal to the interface of the KpsM subunits.

Model building and refinement

To generate the initial model of the Apo 1 state of KpsMT-KpE, the Alphafold2 models²⁵ of the individual subunits were rigid-body-docked into the EM map using Chimera⁵², and the model was iteratively real-space refined in Coot and Phenix:refine^51,53. The obtained structure was used to build all other states. Chain completeness for all states is reported in ‘Chain completeness’ in the Supplementary Information. For Apo 2 and the glycolipid 1 and 2 states, two sets of real-space refine jobs were run, with and without the KpsT subunits rigid-body-docked into the model (Extended Data Table 1). For glycolipid-bound states 1 and 2, the putative substrate model was drawn using Coot Ligand Builder, on the basis of 1,2-dipalmitoyl-phospatidylglycerol and four Kdo sugars linked by β-2-7 and β-2-4 glycosidic linkages. The SMILES output from this job was used in Phenix eLBOW to generate coordinate and constraint files. To preserve the correct linkage geometry connecting the first two Kdo units during refinement, the generated poly-Kdo-phosphatidylglycerol lipid was trimmed to two Kdo sugar units, docked into the substrate density in Coot and real-space-refined in Phenix.refine. After this refinement, the second Kdo sugar was removed from the substrate and the whole model was refined again in Phenix.refine. After the glycolipid 1 and 2 models were refined, the corresponding maps were locally anisosharpened in Phenix⁵⁴, resulting in maps 4 and 5. All structural figures were prepared using ChimeraX⁵⁵ and Inkscape (https://inkscape.org).

AlphaFold2 predictions

The full-length KpsE octamer was predicted at servers of the University of Virginia Molecular Electron Microscopy Core. PmKpsD and StKpsD were predicted on AlphaFold2 Collab servers using truncated protein sequences (no outer membrane signal sequence, N-terminal truncation), to limit the octameric protein sequence to fewer than 3,300 amino acids. Omitted regions were then backfitted on the predicted octameric backbone from a monomer model.

Reporting summary

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