Structural and mechanistic insights into the bacterial amyloid secretion channel CsgG

Cloning and strains

Expression constructs for the production of outer membrane localized C-terminally StrepII-tagged CsgG (pPG1) and periplasmic C-terminally StrepII-tagged CsgG_C1S (pPG2) have been described in ref. 19. For selenomethionine labelling, StrepII-tagged CsgG_C1S was expressed in the cytoplasm because of increased yields. Therefore, pPG2 was altered to remove the N-terminal signal peptide using inverse PCR with primers 5′-TCT TTA AC CGC CCC GCC TAA AG-3′ (forward) and 5′-CAT TTT TTG CCC TCG TTA TC-3′(reverse) (pPG3). For phenotypic assays, a csgG deletion mutant of E. coli BW25141 (E. coli NVG2) was constructed by the method described in ref. 30 (with primers 5′-AAT AAC TCA ACC GAT TTT TAA GCC CCA GCT TCA TAA GGA AAA TAA TCG TGT AGG CTG GAG CTG CTT C-3′ and 5′-CGC TTA AAC AGT AAA ATG CCG GAT GAT AAT TCC GGC TTT TTT ATC TGC ATA TGA ATA TCC TCC TTA G-3′). The various CsgG substitution mutants used for Cys accessibility assays and for phenotypic probing of the channel constriction were constructed by site-directed mutagenesis (QuikChange protocol; Stratagene) starting from pMC2, a pTRC99a vector containing csgG under control of the trc promoter¹⁴.

Protein expression and purification

CsgG and CsgG_C1S were expressed and purified as described¹⁹. In brief, CsgG was recombinantly produced in E. coli BL 21 (DE3) transformed with pPG1 and extracted from isolated outer membranes with the use of 1% n-dodecyl-β-D-maltoside (DDM) in buffer A (50 mM Tris-HCl pH 8.0, 500 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol (DTT)). StrepII-tagged CsgG was loaded onto a 5 ml Strep-Tactin Sepharose column (Iba GmbH) and detergent-exchanged by washing with 20 column volumes of buffer A supplemented with 0.5% tetraethylene glycol monooctyl ether(C8E4; Affymetrix)and4 mM lauryldimethylamine-N-oxide (LDAO; Affymetrix). The protein was eluted by the addition of 2.5 mM D-desthiobiotin and concentrated to 5 mg ml⁻¹ for crystallization experiments. For selenomethionine labelling, CsgG_C1S was produced in the Met auxotrophic strain B834 (DE3) transformed with pPG3 and grown on M9 minimal medium supplemented with 40 mg l⁻¹ L-selenomethionine. Cell pellets were resuspended in 50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1 mM EDTA, 5 mM DTT, supplemented with cOmplete Protease Inhibitor Cocktail (Roche) and disrupted by passage through a TS series cell disruptor (Constant Systems Ltd) operated at 20 × 10³ lb in⁻². Labelled CsgG_C1S was purified as described19. DTT (5 mM) was added throughout the purification procedure to avoid oxidation of selenomethionine.

CsgE was produced in E. coli NEBC2566 cells harbouring pNH27 (ref. 16). Cell lysates in 25 mM Tris-HCl pH 8.0, 150 mM NaCl, 25 mM imidazole, 5% (v/v) glycerol were loaded on a HisTrap FF (GE Healthcare). CsgE-his was eluted with a linear gradient to 500 mM imidazole in 20 mM Tris-HCl pH 8.0, 150 mM NaCl, 5% (v/v) glycerol buffer. Fractions containing CsgE were supplemented with 250 mM (NH₄)₂SO₄ and applied to a 5 ml HiTrap Phenyl HP column (GE Healthcare) equilibrated with 20 mM Tris-HCl pH 8.0, 100 mM NaCl, 250 mM (NH₄)₂SO₄, 5% (v/v) glycerol. A linear gradient to 20 mM Tris-HCl pH 8.0, 10 mM NaCl, 5% (v/v) glycerol was applied for elution. CsgE containing fractions were loaded onto a Superose 6 Prep Grade 10/600 (GE Healthcare) column equilibrated in 20 mM Tris-HCl pH 8.0, 100 mM NaCl, 5% (v/v) glycerol.

In-solution oligomeric state assessment

About 0.5 mg each of detergent-solubilized CsgG (0.5% C8E4, 4 mM LDAO) and CsgG_C1S were applied to a Superdex 200 10/300 GL analytical gel filtration column (GE Healthcare) equilibrated with 25 mM Tris-HCl pH 8.0, 500 mM NaCl, 1 mM DTT, 4 mM LDAO and 0.5% C8E4 (CsgG) or with 25 mM Tris-HCl pH 8.0, 200 mM NaCl (CsgG_C1S), and run at 0.7 ml min⁻¹. The column elution volumes were calibrated with bovine thyroglobulin, bovine γ-globulin, chicken ovalbumin, horse myoglobulin and vitamin B₁₂ (Bio-Rad) (Extended Data Fig. 2). Membrane-extracted CsgG, 20 μg of the detergent-solubilized protein was also run on 3–10% blue native PAGE using the procedure described in ref. 31 (Extended Data Fig. 2). NativeMark (Life Technologies) unstained protein standard (7 μl) was used for molecular mass estimation.

Crystallization, data collection and structure determination

Selenomethionine-labelled CsgG_C1S was concentrated to 3.8 mg ml⁻¹ and crystallized by sitting-drop vapour diffusion against a solution containing 100 mM sodium acetate pH 4.2, 8% PEG 4000 and 100 mM sodium malonate pH 7.0. Crystals were incubated in crystallization buffer supplemented with 15% glycerol and flash-frozen in liquid nitrogen. Detergent-solubilized CsgG was concentrated to 5 mg ml⁻¹ and crystallized by hanging-drop vapour diffusion against a solution containing 100 mM Tris-HCl pH 8.0, 8% PEG 4000, 100 mM NaCl and 500 mM MgCl₂. Crystals were flash-frozen in liquid nitrogen and cryoprotected by the detergent present in the crystallization solution. For optimization of crystal conditions and screening for crystals with good diffraction quality, crystals were analysed on beamlines Proxima-1 and Proxima-2a (Soleil, France), PX-I (Swiss Light Source, Switzerland), I02, I03, I04 and I24 (Diamond Light Source, UK) and ID14eh2, ID23eh1 and ID23eh2 (ESRF, France). Final diffraction data used for structure determination of CsgG_C1S and CsgG were collected at beamlines I04 and I03, respectively (see Extended Data Fig. 10a for data collection and refinement statistics). Diffraction data for CsgG_C1S were processed using Xia2 and the XDS package^32,33. Crystals of CsgG_C1S belonged to space group P1 with unit cell dimensions of a = 101.3 Å, b = 103.6 Å, c = 141.7 Å, α = 111.3°, β = 90.5°, γ = 118.2°, containing 16 protein copies in the asymmetric unit. For structure determination and refinement, data collected at 0.9795 Å wavelength were truncated at 2.8 Å on the basis of an I/σI cutoff of 2 in the highest-resolution shell. The structure was solved using experimental phases calculated from a single anomalous dispersion (SAD) experiment. A total of 92 selenium sites were located in the asymmetric unit by using ShelxC and ShelxD³⁴, and were refined and used for phase calculation with Sharp³⁵ (phasing power 0.79, figure of merit (FOM) 0.25). Experimental phases were density modified and averaged by non-crystallographic symmetry (NCS)using Parrot³⁶ (Extended DataFig. 10; FOM 0.85). An initial model was built with Buccaneer³⁷ and refined by iterative rounds of maximum-likelihood refinement with Phenix refine³⁸ and manual inspection and model (re)building in Coot³⁹. The final structure contained 28,853 atoms in 3,700 residues belonging to 16 CsgG_C1S chains (Extended Data Fig. 2), with a molprobity40 score of 1.34; 98% of the residues lay in favoured regions of the Ramachandran plot (99.7% in allowed regions). Electron density maps showed no unambiguous density corresponding to possible solvent molecules, and no water molecules or ions were therefore built in. Sixteen fold NCS averaging was maintained throughout refinement, using strict and local NCS restraints in early and late stages of refinement, respectively.

Diffraction data for CsgG were collected from a single crystal at 0.9763 Å wavelength and were indexed and scaled, using the XDS package^32,33, in space group C2 with unit-cell dimensions a = 161.7 Å, b = 372.3 Å, c = 161.8 Å and β = 92.9°, encompassing 18 CsgG copies in the asymmetric unit and a 72% solvent content. Diffraction data for structure determination and refinement were elliptically truncated to resolution limits of 3.6 Å, 3.7 Å and 3.8 Å along reciprocal cell directions a*, b* and c* and scaled anisotropically with the Diffraction Anisotropy Server⁴¹. Molecular replacement using the CsgG_C1S monomer proved unsuccessful. Analysis of the self rotation function revealed D₉ symmetry in the asymmetric unit (not shown). On the basis of on the CsgG_C1S structure, a nonameric search model was generated in the assumption that after going from a C₈ to C₉ oligomer, the interprotomer arc at the particle circumference would stay approximately the same as the interprotomer angle changed from 45° to 40°, giving a calculated increase in radius of about 4 Å. Using the calculated nonamer as search model, a molecular replacement solution containing two copies was found with Phaser⁴². Inspection of density-modified and NCS-averaged electron density maps (Parrot³⁶; Extended Data Fig. 10) allowed manual building of theTM1 and TM2 and remodelling of adjacent residues in the protein core, as well as the building of residues 2–18, which were missing from the CsgG_C1S model and linked the α1 helix to the N-terminal lipid anchor. Refinement of the CsgG model was performed with Buster-TNT⁴³ and Refmac5 (ref. 44) for initial and final refinement rounds, respectively. Eighteen fold local NCS restraints were applied throughout refinement, and Refmac5 was run employing a jelly-body refinement with sigma 0.01 and hydrogen-bond restraints generated by Prosmart⁴⁵. The final structure contained 34,165 atoms in 4,451 residues belonging to 18 CsgG chains (Extended Data Fig. 2), with a molprobity score of 2.79; 93.0% of the residues lay in favoured regions of the Ramachandran plot (99.3% in allowed regions). No unambiguous electron density corresponding the N-terminal lipid anchor could be discerned.

Congo red assay

For analysis of Congo red binding, a bacterial overnight culture grown at 37 °C in Lysogeny Broth (LB) was diluted in LB medium until a D₆₀₀ of 0.5 was reached. A 5 μl sample was spotted on LB agar plates supplemented with ampicillin (100 mg l⁻¹), Congo red (100 mg l⁻¹) and 0.1% (w/v) isopropyl β-D-thiogalactoside (IPTG). Plates were incubated at room temperature (20–22 °C) for 48 h to induce curli expression. The development of the colony morphology and dye binding were observed at 48 h.

Cysteine accessibility assays

Cysteine mutants were generated in pMC2 using site-directed mutagenesis and expressed in E. coli LSR12 (ref. 7). Bacterial cultures grown overnight were spotted onto LB agar plates containing 1 mM IPTG and 100 mg l⁻¹ ampicillin. Plates were incubated at room temperature and cells were scraped after 48 h, resuspended in 1 ml of PBS and normalized using D₆₀₀. The cells were lysed by sonication and centrifuged for 20 s at 3,000g at 4 °C to remove unbroken cells from cell lysate and suspended membranes. Proteins in the supernatant were labelled with 15 mM methoxypolyethylene glycolmaleimide (MAL-PEG 5 kDa) for 1 h at room temperature. The reaction was stopped with 100 mM DTT and centrifuged at 40,000 r.p.m. (~100,000g) in a 50.4 Ti rotor for 20 min at 4 °C to pellet total membranes. The pellet was washed with 1% sodium lauroyl sarcosinate to solubilize cytoplasmic membranes and centrifuged again. The resulting outer membranes were resuspended and solubilized using PBS containing 1% DDM. Metal-affinity pull downs with nickel beads were used for SDS–PAGE and anti-His western blots. E. coli LSR12 cells with empty pMC2 vector were used as negative control.

ATR–FTIR spectroscopy

ATR–FTIR measurements were performed on an Equinox 55 infrared spectrophotometer (Bruker), continuously purged with dried air, equipped with a liquid-nitrogen-refrigerated mercury cadmium telluride detector and a Golden Gate reflectance accessory (Specac). The internal reflection element was a diamond crystal (2 mm × 2 mm) and the beam incidence angle was 45°. Each purified protein sample (1 μl) was spread at the surface of the crystal and dried under a gaseous nitrogen flow to form a film. Each spectrum, recorded at 2 cm⁻¹ resolution, was an average of 128 accumulations for improved signal-to-noise ratio. All the spectra were treated with water vapour contribution subtraction, smoothed at a final resolution of 4 cm⁻¹ by apodization and normalized on the area of the Amide I band (1,700–1,600 cm⁻¹) to allow their comparison⁴⁶.

Negative stain EM and symmetry determination

Negative stain EM was used to monitor in-solution oligomerization states of CsgG, CsgG_C1S and CsgE. CsgE, CsgG_C1S and amphipol-bound CsgG were adjusted to a concentration of 0.05 mg ml⁻¹ and applied to glow-discharged carbon-coated copper grids (CF-400; Electron Microscopy Sciences). After 1 min incubation, samples were blotted, then washed and stained in 2% uranyl acetate. Images were collected on a Tecnai T12 BioTWIN LaB6 microscope operating at a voltage of 120 kV, at a nominal magnification of ×49,000 and defocus between 800 and 2,000 nm. Contrast transfer function (CTF), phase flipping and particle selection were performed as described for cryo-EM. For membrane-extracted CsgG, octadecameric particles (1,780 in all) were analysed separately from nonamers and top views. For purified CsgE a total of 2,452 particles were analysed. In all cases, after normalization and centring, images were classified using IMAGIC-4D as described in the cryo-EM section. The best classes corresponding to characteristic views were selected for each set of particles. Symmetry determination of CsgG top views was performed using the best class averages with roughly 20 images per class. The rotational autocorrelation function was calculated using IMAGIC and plotted.

CsgG–CsgE complex formation

For CsgG–CsgE complex formation, the solubilizing detergents in purified CsgG were exchanged for Amphipols A8-35 (Anatrace) by adding 120 μl of CsgG (24 mg ml⁻¹ protein in 0.5% C8E4, 4 mM LDAO, 25 mM Tris-HCl pH 8.0, 500 mM NaCl, 1 mM DTT) to 300 μl of detergent-destabilized liposomes (1 mg ml⁻¹ 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) and 0.4% LDAO) and incubating for 5 min on ice before the addition of 90 μl of A8-35 amphipols at 100 mg ml⁻¹ stock. After an additional 15 min incubation on ice, the sample was loaded on a Superose 6 10/300 GL (GE Healthcare) column and gel filtration was performed in 200 mM NaCl, 2.5% xylitol, 25 mM Tris-HCl pH 8, 0.2 mM DTT. An equal volume of purified monomeric CsgE in 200 mM NaCl, 2.5% xylitol, 25 mM Tris-HCl pH 8, 0.2 mM DTT was added to the amphipol-solubilized CsgG at final protein concentrations of 15 and 5 μM for CsgE and CsgG, respectively, and the sample was run at 125 V at 18 °C on a 4.5% native PAGE in 0.5 × TBE buffer. For the second, denaturing dimension, the band corresponding to the CsgG–CsgE complex was cut out of unstained lanes run in parallel on the same gel, boiled for 5 min in Laemmli buffer (60 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 5% 2-mercaptoethanol, 0.01% bromophenol blue) and run on 4–20% SDS–PAGE. Purified CsgE and CsgG were run alongside the complex as control samples. Gels were stained with InstantBlue Coomassie for visual inspection or SYPRO orange for stoichiometry assessment of the CsgG–CsgE complex by fluorescence detection (Typhoon FLA 9000) of the CsgE and CsgG bands on SDS–PAGE, yielding a CsgG/CsgE ratio of 0.97.

CsgG–CsgE Cryo-EM

Cryo-electron microscopy was used to determine the in-solution structure of the C₉ CsgG–CsgE complex. CsgG–CsgE complex prepared as described above was bound and eluted with buffer supplemented with 100 mM imidazole from a TALON cobalt metal affinity resin to remove unbound CsgG, and on elution it was immediately applied to Quantifoil R2/2 carbon coated grids (Quantifoil Micro Tools GmbH) that had been glow-discharged at 20 mA for 30 s. Samples were plunge-frozen in liquid nitrogen using an automated system (Leica) and observed under a FEI F20 microscope operating at a voltage of 200 kV, a nominal magnification of ×50,000 under low-dose conditions and a defocus range of 1.4–3 μm. Image frames were recorded on a Falcon II detector. The pixel size at the specimen level was 1.9 Å per pixel. The CTF parameters were assessed using CTFFIND3 (ref. 47), and the phase flipping was done in SPIDER⁴⁸. Particles were automatically selected from CTF-corrected micrographs using BOXER (EMAN2; ref. 49). Images with an astigmatism of more than 10% were discarded. A total of 4,881 particles were selected from 75 micrographs and windowed into 128-pixel ×128-pixel boxes. Images were normalized to the same mean and standard deviation and high-pass filtered at a low-resolution cut-off of ~200 Å. They were centred and then subjected to a first round of MSA. An initial reference set was obtained using reference free classification in IMAGIC-4D (Image Science Software). The best classes corresponding to characteristic side views of the C₉ cylindrical particles were used as references for the MRA. For CsgG–CsgE complex, the first three-dimensional model was calculated from the best 125 characteristic views (with good contrast and well-defined features) encompassing 1,221 particles of the complex with orientations determined by angular reconstitution (Image Science Software). The three-dimensional map was refined by iterative rounds of MRA, MSA and anchor set refinement. The resolution was estimated to be 24 Å by Fourier shell correlation (FSC) according to the 0.5 criteria level (Extended Data Fig. 7). Visualization of the map and figures was performed in UCSF Chimera⁵⁰.

Bile salt toxicity assay

Outer-membrane permeability was investigated by decreased growth on agar plates containing bile salts. Tenfold serial dilutions of E. coli LSR12 (ref. 7) cells (5 μl) harbouring both pLR42 (ref. 16) and pMC2 (ref. 14) (or derived helix 2 mutants) were spotted on McConkey agar plates containing 100 μg l⁻¹ ampicillin, 25 μg l⁻¹ chloramphenicol, 1 mM IPTG with or without 0.2% (w/v) L-arabinose. After incubation overnight at 37 °C, colony growth was examined.

Single-channel current recordings

Single-channel current recordings were performed using parallel high-resolution electrical recording with the Orbit 16 device from Nanion. In brief, horizontal bilayers of 1,2-diphytanoyl-sn-glycero-3-phosphocholine (Avanti Polar Lipids) were formed over microcavities (of subpicolitre volume) in a 16-channel multielectrode cavity array (MECA) chip (Ionera)⁵¹. Both the cis and trans cavities above and below the bilayer contained 1.0 M KCl, 25 mM Tris-HCl pH 8.0. To insert channels into the membrane, CsgG dissolved in 25 mM Tris-HCl pH 8.0, 500 mM NaCl, 1 mM DTT, 0.5% C8E4, 5 mM LDAO was added to the cis compartment to a final concentration of 90–300 nM. To test the interaction of the CsgG channel with CsgE, a solution of the latter protein dissolved in 25 mM Tris-HCl pH 8.0, 150 mM NaCl was added to the cis compartment to final concentrations of 0.1, 1, 10 and 100 nM. Transmembrane currents were recorded at a holding potential of +50 mV and −50 mV (with the cis side grounded) using a Tecella Triton 16-channel amplifier at a low-pass filtering frequency of 3 kHz and a sampling frequency of 10 kHz. Current traces were analysed using the Clampfit of the pClamp suite (Molecular Devices). Plots were generated using Origin 8.6 (Microcal)⁵².

Measured currents were compared with those calculated based on the pore dimensions of the CsgG X-ray structure, modelled to be composed of three segments: the transmembrane section, the periplasmic vestibule, and the inner channel constriction connecting the two. The first two segments were modelled to be of conical shape; the constriction was represented as a cylinder. The corresponding resistances R₁, R₂ and R₃, respectively, were calculated as

where L₁, L₂ and L₃ are the axial lengths of the segments, measuring 3.5, 4.0 and 2.0 nm, respectively, and D₁, d₁, D₂ and d₂ are the maximum and minimum diameters of segments 1 and 2, measuring 4.0, 0.8, 3.5 and 0.8 nm, respectively. The conductivity κ has the macroscopic bulk value of 10.6 S m⁻¹ for the wider conical segments. The conductivity was half this value for the narrow central constriction, owing to the reduced ion mobility, in line with findings for the OmpF pore of similar dimensions⁵³. The current was calculated by inserting R₁, R₂ and R₃ and voltage V = 50 mV into

Access resistance also included in the calculations.