Archaic chaperone-usher pilus self-secretes into a superelastic zigzag spring architecture

Anton Zavialov (  antzav@utu. ) University of Turku Natalia Pakharukova University of Turku https://orcid.org/0000-0002-8363-6105 Henri Malmi University of Turku Minna Tuittila University of Turku Sari Paavilainen University of Turku Tobias Dahlberg Umeå University Magnus Andersson Umeå University https://orcid.org/0000-0002-9835-3263 Si Lhyam Myint Umeå University Bernt Eric Uhlin Umeå University https://orcid.org/0000-0002-2991-8072 Debnath Ghosal California Institute of Technology Yi-Wei Chang University of Pennsylvania https://orcid.org/0000-0003-2391-473X Grant Jensen California Institute of Technology https://orcid.org/0000-0003-1556-4864 Stefan Knight Uppsala University https://orcid.org/0000-0002-7180-8758 Urpo Lamminmäki University of Turku

positioned at the usher C-terminal domains CTD1 and CTD2 23-26 (steps 1-2, Fig. 4e and 23 Extended Data Fig.7). The Gd of chaperone-bound Sub4 replaces the G1 strand of the base 24 chaperone through donor strand exchange (DSE), linking Sub4 to Sub3 14,15 . DSE results in the 25 complete folding of Sub3 and formation of the A¢-A¢¢ and B-B¢ hairpins 27 . However, Sub3 26 cannot form a clinch contact with its neighbouring subunits, as Sub2 has entered the narrow 1 usher channel and the chaperone-bound Sub4 is only partially folded and lacks its own twin-2 hairpin 22 . The clinch contact can only be formed between subunits Sub1 and Sub2 on the cell 3 surface (Fig. 4e, step 3). Therefore, the formation of the twin hairpin not prior to, but after DSE 4 serves three purposes. First, it provides an additional folding potential to drive assembly 27 . 5 Second, it prevents premature subunit clinching, thereby keeping the fibre in the elongated 6 conformation required for secretion. Finally, it enables the formation of the quaternary 7 structure immediately after subunit translocation. 8 The secretion step involves handover of the base from NTD to CTDs. The handover 9 cannot be driven by the binding of the base to CTDs, as neither significant hydrophobic 10 interactions nor affinity between the base and CTDs have been observed (Extended Data 11 Fig.8), questioning the origin of forces and energy driving secretion in CUPs. The FimD usher 12 NTD was recently shown to escort the base until it reaches CTDs and form interactions with 13 CTD2 that could potentially facilitate the release of the base from NTD 28 . Our findings 14 demonstrate that secretion of the pilus rod is greatly facilitated by quaternary structure 15 formation, representing an alternative driving force. Since formation of a single clinch reduces 16 the fibre length exactly by the length of one pilin, clinch formation may actively pull subunits 17 to the cell surface without introducing shifts in their positions at each cycle (Extended Data 18 Fig.7). In addition, clinch formation may prevent backtracking of the secretion step potentially 19 leading to the base slipping away from the usher after its release from NTD, permanently 20 jamming assembly (Extended Data Fig.9). Future structural studies on the assembly-secretion 21 mechanism of the CsuD usher will help validate these hypotheses. 22

23
The clinch-DSC-based zigzag filament is probably the earliest and the most widely used 24 architecture of pili assembled via the CUP. This economical design gives the pilus a 25 surprisingly high mechanical stability, rapid dynamic properties and superelasticity. The pilus 1 secretion process involves an elegant mechanism that allows clinch formation only at the cell 2 surface (Fig. 4e). Hence, similar to the chaperone that preserves folding energy of the subunit 3 to drive pilus assembly, the usher inhibits the formation of the quaternary structure, preserving 4 energy of inter-subunit contacts to drive pilus secretion through the membrane. Interestingly, 5 polymers of classical CUP subunits can easily adopt the zigzag filament architecture of the 6 archaic pili (Extended Data Fig.10), suggesting that both types of pili might follow a similar 7 conserved secretion pathway before they reach the stage of forming the final quaternary 8 structure. Hence, it is not excluded that secretion of helical pili might be partially driven by 9 intermediate rather than the final inter-subunit contacts, inviting further studies on this topic. 10 The elucidation of the structure and assembly-secretion process of ubiquitous archaic pili 11 should pave the way for the development of clinch-formation inhibitors against persistent 12 bacterial infections. 13 formation on abiotic surfaces by Acinetobacter baumannii: involvement of a novel 24 chaperone-usher pili assembly system. Microbiology 149, 3473-3484, (2003). 25 5 Giraud, C., Bernard,C. S.,Calderon,V.,Yang,L.,Filloux,A.,Molin,S.,Fichant,G.,1 Bordi, C. & de Bentzmann, S. The PprA-PprB two-component system activates CupE, the 2 first non-archetypal Pseudomonas aeruginosa chaperone-usher pathway system 3 assembling fimbriae. Environmental microbiology 13, 666-683, (2011). Acta. Crystallogr. D Biol. Crystallogr. 58, 1948-1954, (2002. were generated using reverse PCR. The oligonucleotides are listed in Extended Data Table 2. 18

Protein expression and purification 19
Wild type (WT) and mutant CsuA/B were co-expressed with the CsuC chaperone, carrying a 20 C-terminal His6-tag, in the periplasm of E. coli harbouring the pET101-CsuC6H-CsuA/B-## 21 plasmid series and were co-purified by Ni-chelate chromatography essentially as described in 22 22 . WT and mutant CsuA/Bsc were present in the periplasm of E. coli harbouring the pET101-23 6HCsuA/Bdsc-## plasmid series and were purified by Ni-chelate chromatography as described 24 earlier 27 . Depending on the remaining impurities, proteins were dialyzed against 20 mM bis-25 TRIS propane, pH 9.0 and were purified further by anion exchange chromatography on a Mono 1 Q 5/50 GL column (GE Healthcare). For the circular dichroism measurements, the buffer for 2 the proteins was exchanged to 12.5 mM potassium phosphate, pH 7.0 using a PD-10 desalting 3 column (GE Healthcare). Protein concentrations were measured on a NanoDrop™ 2000 4 Spectrophotometer (Thermo Scientific). 5 To express wild-type and mutant variants of Csu fimbriae, E.coli BL-21 AI cells were 6 transformed with ampicillin-resistant pBAD-Csu and its derivatives. Selected clones were 7 cultivated in Luria-Bertani (LB) medium supplemented with 100 µg ml −1 ampicillin overnight 8 at 37°C and refreshed by 1:400 dilution of LB medium containing 80-100 µg ml −1 ampicillin. 9 The cells were grown at 37°C to OD of 0.8-1.0 at 600 nm, then were induced with 0.2% L(+)-10 arabinose for protein expression and were grown for a further 2.5 h. The cells were harvested 11 by two rounds of centrifugation at 5000×g for 30 min and 7000×g for 10 min. The bacterial 12 pellet was resuspended in 0.5 mM Tris-HCl, pH 7.4, 75 mM NaCl and incubated at 65°C for 13 1h. After incubation, the bacteria were pelleted by two rounds of centrifugation at 9500×g for 14 10 min. Supernatant containing detached Csu fimbriae was carefully collected and stored at 15 4°C before analysis. Prior to cryo-EM, the quality of the preparation was assessed by negative 16 stained samples in transmission electron microscopy. 17

Electron microscopy of negative stained pili 18
Purified Csu pili were applied on Formvar-coated glow-discharged gold grids (Agar Scientific) 19 and incubated for 1 min. After blotting the excessive sample, the grid was washed with two 20 drops of water, blotted again and then stained with 2% uranyl acetate. Images were acquired 21 on a JEM-1400 Plus transmission electron microscope (JEOL Ltd.) operated at 80 kV. 22

Cryo-electron microscopy 23
Supernatant containing detached Csu fimbriae was concentrated to approximately 10 g l −1 24 using a Vivaspin device (Sartorius Stedim) with a molecular mass cut-off of 100 kDa. 4 µl of 25 sample was applied to glow-discharged Quantifoil R2/2 300 mesh copper grids coated with 1 ultrathin carbon (Electron Microscopy Sciences). The grids were blotted and plunge-frozen in 2 liquid ethane using Vitrobot Mark IV (ThermoFisher Scientific) at 4°C and 100% humidity. 3 The data were collected on a 300 kV Titan Krios electron microscope (Thermo Scientific) 4 equipped with a Gatan K3 direct electron detector operated in super-resolution mode with a 5 pixel size of 0.433 Å and a defocus range of -1.0 to -3.0 µm. A total dose of 60 electrons/Å 6 was applied and equally divided among 40 frames to allow for dose weighting. Details on cryo-7 EM data collection are summarized in Extended Data Table 1. 8

Cryo-EM image processing and reconstruction 9
Dose-fractionated movie frames were subjected to beam-induced motion correction using 10 MotionCor2 29 . Image processing and helical reconstruction were performed in RELION 3.0 11 30 . Filaments manually picked from 602 selected micrographs using e2helixboxer program 12 within EMAN 2 31 were subjected to 2D classification to generate auto-picking templates. After 13 autopicking of helical filaments, a total of 480,064 segments were extracted with a box size of 14 400 pixels. After 2D and 3D classification steps 255,833 segments were used for 3D 15 refinement. The segments were rescaled to a pixel size of 1.35 Å. A starting model for 16 reconstruction was generated de novo from the 2D particles using Stochastic Gradient Descent 17 algorithm in RELION 3.0. The helical symmetry parameters were estimated using conventional 18 Fourier-Bessel analysis and segclassreconstruct and seggridexplore modules in SPRING 32 . 19 The initial estimates of helical parameters (-157° helical twist, 26.3 Å helical rise) were tested 20 using a search range of -150° to -165° for the twist and 26 Å to 30 Å for the rise. After 3D 21 classification 255,833 particles were used for high resolution 3D refinement. The helical 22 symmetry (-153° helical twist, 28 Å helical rise) was applied and refined during high resolution 23 3D refinement producing a map with a resolution of 6.18 Å. Applying a soft mask with a raised 24 cosine edge of 14 pixels and B-factor sharpening yielded a map with a global resolution of 4.8 25 Å assessed by the gold standard Fourier-shell correlation procedure between independently 1 refined half reconstructions (FSC 0.143) (Scheres and Chen, 2012). The resolution was further 2 improved to 3.42 Å after two iterations of Bayesian polishing followed by 3D refinement and 3 post-processing. The final map showed clear β-strand separation and density for bulky side 4 chains consistent with the reported resolution. The pixel size of the cryo-EM maps from 5 RELION was slightly off and was adjusted by calculating the cross-correlation coefficient of 6 the map to the refined model using "Fit in Map" tool in UCSF Chimera package (Pettersen et 7 al., 2004). 8

Model building and refinement 9
The initial model of the Csu pilus was built manually by fitting the crystal structure of 10 CsuA/Bsc (PDB: 6FM5) (Pakharukova et al. 2018) into the experimental electron density using 11 the UCSF Chimera. The angle between two subunits was adjusted using the Chimera "Fit in 12 Map" tool in several iterations of first docking three subunit dimers into adjacent regions in the 13 map with one subunit overlap and averaging the orientations of the overlapping subunits, then 14 overlapping the three dimers fully and averaging the subunit orientations of all three dimers. A 15 short linker connecting the donor strand with strand A was modelled with Coot 33 . The structure 16 was refined by combining manual adjustments in Coot and real space refinement in PHENIX 17 34 . The initial four-subunit model was reduced to a model with three donor strand 18 complemented subunits (four chains) that occupy the highest resolution positions in the map. 19 The model was validated with MolProbity 35 . The refinement statistics are given in Extended 20 Data Table 1. 21

Atomic force microscopy 22
The bacteria were grown on the LB agar plate supplemented with carbenicillin and induced 23 with 0.02% arabinose to produce pili. The bacterial cells with pili were imaged by atomic force 24 microscopy (AFM) as described earlier 36 with some modifications. Briefly, bacterial cells were 25 suspended in 100 µL of Milli-Q water and 10 µL of which was placed onto a freshly cleaved 1 mica surface (Goodfellow Cambridge Ltd., Cambridge). Samples were incubated for 5 min at 2 room temperature and blotted dry before they were placed into a desiccator for a minimum of 3 2 h in order to dry. Images were collected with a Nanoscope V Multimode 8 AFM equipment 4 (Bruker) using Bruker ScanAsyst mode with Bruker ScanAsyst-air probe oscillated at resonant 5 frequency of 50-90 kHz, selected by the Nanoscope software. Images were collected in air at 6 a scan rate of 0.8-1.5 Hz, depending on the size of the scan and the number of samples (256 or 7 512 samples/image). The final images were plane fitted in both axes and presented in amplitude 8 (error) mode. 9

E. coli strain BL21 harboring pBAD-Csu or its derivatives was cultured overnight in Luria 11
Broth (LB) medium in the presence of 100 mg l -1 ampicillin. 5 ml of the fresh medium in a 50 12 ml polypropylene tube was inoculated with 100 µl of the overnight culture and then grown at 13 37 o C with vigorous shaking for 2 h. Dilutions of the anti-CsuEN (aEN) polyclonal antibody 14 (Innovagen AB) in 50 µl LB were divided on microtiter plates. Bacterial cultures were induced 15 with 0.2 % arabinose, and 150 µl triplicates were mixed with the serum dilutions on microtiter 16 plates. The plate was incubated at 37 o C for 2 h with gentle shaking. Wells were then emptied 17 and washed two times with 300 µl of phosphate buffered saline. Any remaining biofilm was 18 stained with 1 % crystal violet for 15 min, rinsed with water, allowed dry, and dissolved in 250 19 µl of 0.2 % Triton X-100. Optical density at 595 nm was determined with a 96-well plate 20 spectrometer reader. 21

Western blotting 22
Periplasmic fractions were mixed with Laemmli buffer, and the samples were boiled. The 23 proteins were separated by electrophoresis in 18 % SDS polyacrylamide gels and transferred 24 onto an Immuno-blot polyvinylidene difluoride membrane (Bio-Rad Laboratories, USA) in 25 Bio-Rad A-buffer (25 mM Tris, pH 8.3, 192 mM glycine, with 20% methanol and 0.1% SDS) 1 at 100 V or 350 mA for 1 h. Membrane was blocked with 5% skim milk in phosphate-buffered 2 saline/Tween, incubated with primary anti-CsuA/B rabbit polyclonal antibody (Innovagen AB) 3 followed by incubation with secondary IRDye 680RD-conjugated anti-rabbit goat antibody 4 (Li-Cor Biosciences). Protein bands were detected with the Odyssey system (Li-Cor 5 Biosciences) and quantified with ImageJ. 6

Force measuring optical tweezers 7
To measure the biomechanical properties of Csu pili we used an custom made force measuring 8 optical tweezers setup constructed around an inverted microscope Olympus IX71 (Olympus, 9 Japan) equipped with a water immersion objective (model: UPlanSApo60XWIR 60X N.A. = 10 1.2; Olympus, Japan) and a 1920 x 1440 pixel CMOS camera (model: C11440-10C, 11 Hamamatsu) 37 . To sample force data with high signal-to-noise ratio with minimal amount of 12 drifts we used the Allan variance method to identify noise 38 . We used the Power Spectrum 13 method to calibrate the trap by sampling the microspheres position at 131,072 Hz and 14 averaging 32 consecutive data sets acquired for 0.25 s each. To extend a pilus, we moved the 15 piezo stage at a constant speed of 50 nm/s and sampled the force and position at 50 Hz. To 16 assess the mean contour length of the pilus quaternary structure, we "buckled" pili by reversing 17 the piezo-stage until the bead touched the bacterial cell wall. We measured the distance from 18 our starting position (rise of region I) to the cell. To study ultrashort pili, we had to "rub" the 19 bacterial surface with the laser-controlled bead in our OT system. 20

Temperature-depended folding transition analysis 21
Circular dichroism was measured using Chirascan™ CD Spectrometer (Applied Photophysics) 22 and a Macro-cuvette 110-QS with 1 mm layer thickness (Hellma). Background for the spectra 23 was first measured four times from the buffer 12.5 mM potassium phosphate at pH 7.0 before 24 inserting the target protein at 0.150 mg/ml concentration. CD spectra at 20 o C were measured 25 four times with a 195-260 nm wavelength range and using 1 nm intervals between each 3 s 1 measurements. For the melting spectra, proteins were heated using 4 o C temperature ramping 2 from 19 to 99 o C. Each spectrum was measured once after a 30 s temperature stabilization time 3 using wavelength range 260-195 nm and 1 nm intervals between each 2 s measurements. The 4 measurement of all melting spectra took 1 h 28 min. Each spectrum was smoothed by a factor 5 of 4. Melting curves were recorded at 225 nm wavelength by heating samples from 20 to 99 o C 6 at the rate of 1 o C/min. Circular dichroism was measured for 12 s at first every 1.0 o C and later 7 every 0.5 o C with 0.15 o C error margin. Each recording took 1 h 19 min. The cuvette was 8 purified of residual protein using 2 M potassium hydroxide between samples. The "Curve 9 Fitting function" in the Chirascan user interface was used to fit melting data to the "Sigmoid 10 curve + slope" equation. 11

Modeling of assembly-secretion process 12
The Csu pilus models were constructed based on the cryo-EM structure of the Csu pilus rod 13 Surface diagram of a 12-subunt fragment of the Csu pilus rod. Subunits are numbered in the 4 direction of pilus growth, from the pilus tip to its assembly base at the outer membrane (OM). 5 c, Cartoon diagram of the rod focusing on the donor strands. d, Cartoon diagrams of 13-subunit 6 fragments of archaic Csu and classical P pilus rods. The zigzag filament is ~3 times as long as 7 the helical tube rod. The handedness is indicated by a black curved arrow. 8  Table 3. e, Csu pilus assembly-secretion cycle. Modelling is described in Extended Data Fig.7. 11 The DGF folding energy and DGQ free energy of quaternary structure formation preserved by 12 the chaperone and usher drive assembly and secretion, respectively. See also Supplementary 13 Video 3. 14 Extended Data  CsuA/Bsc. Periplasmic extracts were obtained from equal number of cells. CsuA/Bsc was 3 detected with rabbit polyclonal anti-CsuA/Bsc antibody and secondary IRDye 68RD-4 conjugated anti-rabbit goat antibody as described in Procedures. b, CD spectra of WT 5 CsuA/Bsc at different temperatures. c, Temperature dependence of molar ellipticity of WT and 6 substituted variants of CsuA/Bsc at 225 nm. Melting temperatures generated from the data are 7 listed in Extended Data Table 3. 8 2. The Gd donor strand of the subunit in the preassembly complex (Sub4) replaces the G1 donor 1 strand of the chaperone capping the base of the pilus in the zip-in-zip-out donor strand 2 exchange (DSE) process 14,16 , linking Sub4 to the pilus. This also results in the complete folding 3 of Sub3 and formation of the A¢-A¢¢ and B-B¢ hairpins 27 . The former pilus-capping chaperone 4 is released. 5 6 3-4. In a reversible process, the pilus translocates up the usher channel, rotating clockwise. Due 7 to the presence of the A¢A¢¢-BB¢ twin hairpin, rotation within the usher is restricted to a 8 relatively narrow path. To bring Sub2 A¢-A¢¢ hairpin closer to the acceptor pocket on Sub1, 9 secreted pilus or (most probably) the usher has to rotate as well (see Supplementary video 3). 10 11 5. Still in a reversible process, Sub2, emerging from the usher secretion channel, leans to the 12 edge of the usher bringing Sub2 A¢-A¢¢ hairpin closer to Sub1 DD¢ loop. The globular domain 13 of NTD dissociates from CsuC as the angle between these proteins becomes suboptimal for the 14 interaction, while the flexible N-tail remains bound as evident from the structure of FimD 15 conformer 1 28 . 16 17 6. A clinch contact between Sub1 and Sub2 forms as the A¢-A¢¢ hairpin of Sub2 and the Gd 18 donor strand N-terminus of Sub3 bind to Sub1 from two sides, while the DD' loop covers the 19 A¢-A¢¢ hairpin from the side, locking its conformation with a second layer of interactions. The 20 formation of the clinch is likely facilitated by narrowing down the number of pathways towards 21 its formation: when subunits come close to one another, the A¢-A¢¢ hairpin and Gd N-terminus 22 prevent sideways rotations of Sub1 as shown on Fig. 2c. Once the clinch forms, Sub2 becomes 23 part of the rigid pilus stalk and is unable to pull back into the usher channel. This mechanism 24 prevents pilus backtracking, driving unidirectional secretion. 25 26 7. Upon the completion of the secretion cycle, the N-tail of the usher NTD reaches CTD2 and 27 loses its affinity to the chaperone as suggested by Du et al. 28 . NTD is released to accept a new 28 preassembly complex. 29