Protein-lipid interaction at low pH induces oligomerisation of the MakA cytotoxin from Vibrio cholerae

Many pathogenic bacteria produce protein toxins that target and perturb host cell membranes. The secreted α-pore-forming toxins (α-PFTs) cause membrane damage via pore formation. This study demonstrates a remarkable, hitherto unknown mechanism by an α-PFT protein from Vibrio cholerae. As part of the MakA/B/E tripartite toxin, MakA is involved in membrane pore formation similar to other α-PFTs. In contrast, MakA protein alone induces tube-like structures in the acidic lysosomal host cell compartment. In vitro studies unravel the dynamics of tubular growth, which occur in a pH-, lipid- and concentration-dependent manner. A 3.7-Å cryo-electron microscopy structure of MakA filaments reveals a unique protein-lipid superstructure. In its active α-PFT conformation, MakA embeds its transmembrane helices into a thin annular lipid bilayer and spirals around a central cavity. Our study provides molecular insights into a novel tubulation mechanism of an α-PFT protein, revealing a new mode of action by a secreted bacterial toxin.


Introduction
observed co-localization of the Alexa568-labeled MakA (Alexa568-MakA) with GFP-132 LAMP1 or lysotracker in tubular structures ( Fig. 1a and Supplementary Fig. 1a). The 133 tubulation in the lysosomes was further confirmed by transmission electron microscopy 134 (TEM) of lysosomes isolated from MakA-treated HCT8 cells (Fig. 1b). To investigate 135 if MakA can also induce tubulation of lysosomes outside the intracellular environment, 136 we purified lysosomes from untreated HCT8 cells and exposed them to native MakA 137 or to Alexa568-MakA. Both confocal microscopy and TEM analysis revealed 138 aggregation and tubulation of lysosomes at pH 5.0 ( Fig. 1c-d). In addition, a majority 139 of the lysosomes showed well-organized tubulation when exposed to MakA at pH 6.5 140 (Fig. 1d). In contrast, we did not observe any MakA-induced tubular structures in 141 lysosomes at pH 7.0 (Fig. 1d). Western blot analysis confirmed pH-dependent binding 142 of MakA to lysosomes (Fig. 1e). Alexa568-MakA was subsequently shown to bind 143 epithelial cells in a pH-dependent manner ( Fig. 1f-g). To determine the kinetics of 144 MakA binding to the target cells, HCT8 cells were exposed to Alexa568-MakA at pH 145 5.0, and live-cell imaging was performed using spinning disc confocal microscopic 146 analysis. Consistent with our earlier findings 21 , we observed accumulation of 147 Alexa568-MakA on the plasma membrane, including filipodia-rich tubular structures, in 148 a time-dependent manner ( Fig. 1h and Supplementary Fig. 1b-c). The time scale of 149 MakA binding to individual HCT8 cells ranged from ∼40 minutes to 4 hours after 150 Alexa568-MakA treatment (Fig. 1f-h and Supplementary Fig. 1b-c). Ultimately, 151 Alexa568-MakA was detected on the plasma membrane of the entire cell population, 152 with most cells positive for tubular structures protruding out from the plasma membrane 153 ( Fig. 1f and 1h). Taken together, these results suggest that MakA can cause tubulation 154 of both endolysosomal membranes and plasma membranes in a pH-dependent 155 manner. 156 toxicity was quantified by a propidium iodide uptake assay using flow cytometry and 166 fluorescence microscopy ( Fig. 2b and Supplementary Fig. 2a). In addition to causing 167 pH-dependent toxicity of HCT8 cells, MakA was similarly toxic to other colon cancer 168 cells, Caco-2 and HCT116 cells, as assessed by the MTS cell viability assay 169 (Supplementary Fig. 2b). 170 171 Erythrocytes are widely used as a cell model to investigate the cytolytic activity of the 172 toxins that belong to the ClyA pore-forming toxins family 14,17,18,24,25 . To test if pH-173 dependent binding and oligomerization of MakA may cause hemolysis of erythrocytes, 174 human erythrocytes were exposed to increasing concentrations of MakA at different 175 pH conditions for either 90 min or 5 h (Fig. 2c). When erythrocytes were exposed to 176 MakA at pH 5.0, hemolysis was observed in a concentration dependent manner within 177 90 min (Fig. 2c). In contrast, MakA failed to induce hemolysis of erythrocytes at pH 6.5 178 or pH 7.4 during the 90 min treatment. A detectable, but low level of hemolysis was 179 observed after 5 h with erythrocytes exposed to MakA at pH 6.5 (Fig. 2c). With 180 confocal microscopy, we observed pH-dependent binding of Alexa568-MakA to 181 erythrocytes. A majority of erythrocytes at pH 5.0 and 6.5 were covered by Alexa568-182 MakA whereas there was virtually no MakA binding observed at pH 7.4 ( Fig. 2d and 183 Supplementary Fig. 2c). Notably, we detected the presence of tubular structures in 184 association with MakA at the red blood cell surface, as shown by a maximum 3D 185 projection of the z-stack images of erythrocytes (Fig. 2e). The presence of MakA-186 induced tubular structures on the surface of erythrocytes was further observed by TEM 187 and scanning electron microscopy (SEM) (Fig. 2f-g). Together, these results suggest 188 that MakA could accumulate in a pH-dependent manner at the surface of both epithelial 189 cells and erythrocytes, thereby inducing formation of tubular structures that ultimately 190 might lead to cell lysis.  Fig. 3a). The results indicated that more MakA was associated with 199 the liposomes at pH 5.0 and 6.5 than at pH 7.4. The interaction between MakA and 200 ECLE liposomes at pH 6.5 was quantified by surface plasmon resonance (SPR) 201 analysis (Fig. 3a). MakA displayed significant interaction with ECLE liposomes, with 202 an estimated KD of 49.2 nM. Importantly, MakA at the highest tested concentration 203 (200 nM) failed to interact with the liposomes prepared from zwitterionic 1-palmitoyl-2-204 oleoyl-sn-glycero-3-phosphocholine (POPC), used as a negative control (Fig. 3a). 205

206
To determine the effect of pH on MakA's conformational properties, the protein was 207 subjected to circular dichroism (CD) spectroscopy analysis at different pH in the 208 absence or presence of ECLE liposomes (Fig. 3b-c). In the absence of liposomes, the 209 CD spectra indicated a decrease in the α-helical content of the protein when in the 210 acidic environment. This decrease in α-helical content of MakA was restored when 211 ECLE liposomes were present, suggesting that liposomes somehow stabilized the 212 structure of MakA (Fig. 3b- Fig. 3b).  Fig. 3c). Concomitant with the 221 assembly of tubular structures emanating from the liposomes, the size of the liposomes 222 appeared to shrink as the tubules grew up to several micrometers in length. Ultimately, 223 the entire liposome seemed to be transformed into long tubules ( Fig. 3h and  224 Supplementary Fig. 3c-e). The tubulation of ECLE liposomes was also observed by 225 confocal microscopy upon treatment with Alexa568-labeled MakA (Supplementary 226 Fig. 3d-e). In the same population of small liposome particles, we also detected 227 Alexa568-MakA-positive large lipid vesicles (5-10 μm in size, less than 1% of the entire 228 liposome fraction). The z-stack projection suggested that the whole lipid vesicle was 229 decorated with a bundle of fluctuating tubules (Supplementary Fig. 3e). To further 230 assess whether or not any protein or glycolipid receptor mediated the observed 231 membrane tubulation by MakA, liposomes were prepared from a well-defined synthetic 232 lipid mixture (SLM); whose composition was inspired by the distribution of lipids found 233 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in was observed by TEM (Fig. 3d). In addition to the tubular structures, we observed a 235 large number of well-organized, star-shaped oligomeric particles of MakA among the 236 ECLE liposomes (Supplementary Fig. 3c). Furthermore, the presence of MakA 237 protein in the tubular structures was evidenced by immunogold staining using MakA-238 specific antibodies (Fig. 3e). By fluorescence microscopy we were able to visualize 239 tube growth originating from a supported lipid bilayer (SLB) prepared from SLM 240 liposomes containing the fluorescent lipid Texas Red-DHPE, demonstrating that the 241 tubes also contain lipids from the SLB (Fig. 3f and Movie 1). 242

243
We next investigated the kinetics of MakA protein-lipid tubulation. Using fluorescence 244 microscopy, we found that administering Alexa568-MakA to a SLB prepared from SLM 245 liposomes prompted a significant and highly dynamic membrane remodeling ( Fig. 3g  246 and Movie 2). Within 10 minutes, Alexa568-MakA binding to the SLBs resulted in 247 formation of MakA-associated tubules of various sizes (Fig. 3g). Based on these 248 findings, we propose a schematic model for how the MakA-liposome interaction can 249 result in the formation of the observed protein-lipid tubular structures (Fig. 3h). At pH 250 6.5 or lower, MakA may adopt a conformation that allows the protein to insert into the 251 lipid membrane in the form of an oligomer assembly that can start to spiral around the 252 lipids of the membrane leading to formation of a growing tube structure. Concomitantly 253 the size of the vesicle appear to shrink, and the tube may grow up to several 254 micrometers in length. Our results suggest that the MakA-lipid tubulation can occur 255 without the involvement of other proteins or some specific protein receptor under the 256 pH conditions tested. 257 258

Structure of the MakA filament 259
We used helical reconstruction to solve the cryo-EM structure of a MakA filament 260 assembled in vitro in the presence of ECLE liposomes at pH 6.5 and high protein 261 concentrations ( Fig. 4a and Supplementary Fig. 4). An initial 2D classification 262 allowed us to identify repetitive elements and measure a helical repeat distance of 263 ~216 Å ( Fig. 4b and Supplementary Fig. 5a). A subsequent investigation of the layer 264 line distances in a collapsed power spectrum of selected well-resolved class averages 265 confirmed this distance (Supplementary Fig. 5b). Next, we performed a preliminary 266 3D refinement of filament segments from well-defined 2D class averages without 267 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 12, 2021. ; https://doi.org/10.1101/2021.09.10.459761 doi: bioRxiv preprint imposing symmetry (Supplementary Fig. 4). This volume was visually inspected to 268 deduce the helical symmetry parameters (Supplementary Fig. 5c-d). One repeating 269 element of the right-handed spiral consists of 37 tetramers that complete five turns 270 around the helical axis, spanning a length of 216.5 Å and a diameter of 322 Å. This 271 arrangement results in an axial rise of 5.85 Å per subunit and a helical twist of 48.65˚ 272 ( Fig. 4b and Supplementary Fig. 5d). Application of these initially calculated values 273 with local searches in a 3D refinement further optimized symmetry parameters, 274 resulted in a cryo-EM map at an overall resolution of 3.7 Å (Supplementary Table 1  275 and Supplementary Fig. 6). The obtained cryo-EM map features a well-resolved 276 central transmembrane helix (TMH) region and a less well-resolved peripheral region 277 (Fig. 4c). We isolated two MakA tetramers from the segments using signal subtraction 278 and subjected the resulting particles to 3D classification and refinement to improve the 279 peripheral density and connectivity. The clear connectivity of the obtained density map 280 ( Fig. 4d and Supplementary Fig. 6e) allowed for reliable secondary structure 281 placement using the MakA soluble state crystal structure (PDB-6EZV 11 ). High-282 resolution features in the helical reconstruction ( Fig. 4c and Supplementary Fig. 6f) 283 allowed for de novo model building of structural elements in the central region. 284 However, due to continuous rotation along the filament axis, flexibility (Supplementary 285 Fig. 4), and the conformational difference with respect to the crystal structure, the 286 MakA tail domain structure is less reliable and modeled with poly-alanine secondary 287 structure elements. 288 289 MakA oligomerizes into a filamentous structure growing from or ending in membranous 290 vesicles (Fig. 4a, Supplementary Fig. 4). The building blocks of this filament are 291 formed by two MakA dimers ( Fig. 4b and 4d) that organize into a pinecone-like 292 architecture, spiraling around a central cavity (Fig. 4c). From the top view along the 293 filament axis, the helix features a propeller-like structure with a weak, annular density 294 embedded in between the blades formed by MakA (Fig. 4c). This density resembles 295 lipid tails and contains some spherical features, which could be associated with 296 phospholipid heads, suggesting the presence of a thin phospholipid bilayer that spirals 297 around the central cavity of the filament (Fig. 4c). Interestingly, the annular density is 298 located between the transmembrane helices of MakA (Fig. 4d), indicating that the 299 active toxin form interacts with lipid vesicles and starts to oligomerize by internalizing 300 membrane lipids. 301 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The basic building block of the observed protein-lipid filament is formed by four MakA 305 subunits in a membrane-bound active conformation (Fig. 5a,b). This conformation of 306 MakA is significantly different from the previously reported soluble state structures 307 resembling the inactive form (PDB-6DFP and PDB-6EZV 11 ). In the soluble form, a C-308 terminal tail (res. 351-365, Fig. 5c, purple) inactivates the predicted transmembrane 309 domains by forming a β-tongue, consisting of three β-sheets 11 , that shields the 310 hydrophobic residues from the surrounding solvent. This shielding characteristic is represents the handle, the transition from the tail to the neck region forms two hinges, 316 and the β-tongue together with 4 and 5 resembles the blade that folds out (Fig. 5c, α4 317 & α5; light & dark green). Additionally, the β-tongue changes its secondary structure 318 and, together with α4 and α5/α6, forms two extended helices (Fig. 5c). This significant 319 conformational change leads to the formation of two TMHs and a short loop region 320 CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in we show that MakA is able to produce a novel protein-lipid polymeric superstructure at 344 low pH (6.5 or below) that perturbs host cell membranes. forms of mammalian pore-forming toxin, mammalian perforin 2 (mPFN2), provide 359 interesting insights into the pore generation process. It was demonstrated that pre-360 pore-to-pore transformation occurs at an acidic pH, which is accomplished by a 180° 361 rotation of the membrane attacking domain and β-hairpin P2 domains with respect to 362 one another, allowing membrane insertion to take place 34 . Similar to these other toxins, tested. Consistent with these results, we recently found that MakA co-localizes with β-368 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in the rapid growth of tubules towards the extracellular space as shown with red blood 393 cells (Fig. 2g). We propose that the appearance of tubular structures in response to 394 MakA was a direct consequence of MakA insertion into the membrane, which 395 appeared to create the conditions for generating a positive curvature, as described 396 previously for protein-lipid complexes and multi-anchoring polymers 43,44 . 397

398
Our findings allowed us to propose a model for MakA oligomer assembly and its 399 implications for pore formation. Within the oligomerized MakA filament, the TMHs line 400 the inner cavity, interacting with lipids, which are intercalated both within the dimer 401 interface and between the dimers in the helical spiral ( Fig. 4c and Fig. 5b). In the 402 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in with a monomer with a pronounced elbow-like kink from α3 to α4 in the transition from 417 the tail to the neck region (Fig. 5c). Subsequently, the two dimers form a tetramer with 418 the stretched and kinked MakA states associating with each other, respectively. The were dissolved in a mixture of chloroform: methanol (2:1) to a concentration of 1 442 mg/mL. Next, they were protonated by addition of 0.5 μL of 1 M HCl to 100 μg of PIP2, 443 kept at room temperature for 15 min and dried with nitrogen gas. The dried lipid was 444 redissolved in chloroform: methanol (3:1) mixture to 1 mg/mL followed by drying again. 445 Finally, the 100 μg of PIP2 was redissolved in 100% chloroform to 1 mg/mL and stored 446 at -20°C until used for liposome production. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in For the lysosome pull-down assay, intact lysosomes freshly isolated from HCT8 cells 474 were diluted in three times their volume of freshly prepared binding buffer (120 mM 475 sodium citrate, pH 5.0, pH 6.5 or pH 7.0), followed by incubation with MakA (20 µg/mL) 476 at 37°C (60 min). These MakA-lysosome complexes were centrifuged at 21,000 x g 477 Live cell experiments were conducted in phenol-red-free IMDM media adjusted to pH 498 5.0 supplemented with 10% FBS and 1 mM sodium pyruvate (Thermo Fisher Scientific) 499 at 37°C in 5% CO2. Alexa568-MakA (500 nM) was added to HCT8 cells, and images 500 were recorded every 1 min during a period of 120 min using a 63X lens and Zeiss 501 Spinning Disk Confocal controlled by ZEN interface (RRID:SCR_013672) with an Axio 502 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in For confocal microscopy, HCT8 cells were loaded with the nuclear staining marker 513 Hoechst 33342 (2 μM, 30 min) and exposed to Alexa568-MakA (500 nM) in different 514 pH-adjusted IMDM complete media for 4 h at 37°C in 5% CO2. Cells were visualized 515 live using a Leica SP8 inverted confocal system (Leica Microsystems) equipped with 516 an HC PL APO 63x/1.40 oil immersion lens. Images were captured using the LasX 517 (Leica Microsystems) and processed using ImageJ -FIJI distribution 45 . 518 519 For the propidium iodide uptake experiment, HCT8 cells were treated with MakA (500 520 nM, 4 h) in IMDM complete media (pH 5.0 or pH 7.4), followed by adding propidium 521 iodide (0.5 μg/mL, 30 min). Fluorescence and bright-field images were captured with 522 a fluorescence microscope (Nikon, Eclipse Ti). Images were processed using the NIS-523 Elements (Nikon) and ImageJ -FIJI distribution 45 . 524 525 For Alexa568-MakA binding to erythrocytes, freshly prepared human erythrocytes 526 (0.25% in PBS) were loaded into an 8-well chamber slide (µ-Slide, ibidi), cells were 527 allowed to adhere to the glass surface for 10 h, followed by buffer exchange to citrate 528 buffer (pH 5.0, pH 6.5 or pH 7.4). The erythrocytes were exposed to Alexa568-MakA 529 (500 nM) for 3 h at 37°C in 5% CO2. Cells were visualized using a Leica SP8 inverted The liposome pull-down assay was performed as previously described 21,47 . Briefly, the 584 liposome suspension was diluted in five times its volume of freshly prepared binding 585 buffer (120 mM sodium citrate, pH 5.0, pH 6.5 or pH 7.4), followed by centrifugation at 586 21,000 x g for 30 min at room temperature. The liposome pellet was resuspended in 587 binding buffer followed by incubation with MakA (20 µg/mL). The liposome-protein 588 mixtures were incubated at 37°C (60 min), followed by centrifugation at 21,000 x g at 589 room temperature (30 min). To reduce the background, pellets were washed in the 590 respective binding buffer two to three times. The resulting sample was loaded onto the 591 SDS-PAGE, transferred to a nitrocellulose membrane and subjected to Western blot 592 analysis using anti-MakA antiserum (1:10,000 dilution, overnight at 4°C). The MakA 593 antibodies were detected with HRP-conjugated goat anti-rabbit secondary antibodies. 594 The membranes were developed with a chemiluminescence reagent (Bio-Rad). 595 Images were acquired using ImageQuant LAS 4000 instrument and processed using 596 ImageJ -FIJI distribution 45 .  To analyze the interaction of MakA with ECLE or POPC liposomes, L1 sensor chips 639 and a Biacore 3000 instrument were used as previously described 21 . In brief, the ECLE 640 or POPC liposomes were immobilized onto the cells of the L1 sensor chip surfaces at 641 low flow rates of 2 μL/min for 40 min, stabilized with 50 mM NaOH, and the successful 642 Immun-Star™ AP Chemiluminescence (Bio-Rad). Images were acquired using 669 ImageQuant LAS 4000 instrument and processed using ImageJ -FIJI distribution 45 .

Cryo-EM data processing and helical reconstruction 693
The MotionCorr implementation of RELION-3.1 was used for drift correction and dose 694 weighting of the micrographs 48 . The contrast transfer function (CTF) was determined 695 using CTFFIND-4.1.14 (ref. 49 ), and empty or micrographs with poor CTF fits or low ice 696 quality were removed after manual inspection, which reduced the total to 1,351 697 micrographs (Supplementary Fig. 4). Helical reconstruction tools in RELION-3.1 (ref.  Supplementary Fig. 4). In parallel, a 2D classification was 707 performed by extracting 65'241 segments from 2x binned data, applying a box size of 708 646 Å (310 px, 2.084 Å/px), an IBD of 62 Å, and a spherical mask of 580 Å. The volume 709 refined without symmetry and the 2D class averages obtained from large helical 710 segments were used to determine the helical symmetry parameters. Diameter and 711 repeat distance were visually analyzed and measured in a representative 2D class 712 average in RELION-3.1 (Fig. 4b and Supplementary Fig. 5a). Additionally, the repeat 713 distance was calculated from the corresponding collapsed power spectrum (layer-line 714 distance-1) in SPRING-0.68 (ref. 51 ) (Supplementary Fig. 5b). Next, to determine the 715 helical twist and rise, the number of turns and subunits per repeat were counted from 716 the initial reconstructed model (Supplementary Fig. 5c-d), and the handedness of the 717 reconstruction was, after the subsequent high-resolution refinement described below, 718  Fig. 6). As the peripheral region was less well-resolved, a subsection 729 of the structure was isolated via signal subtraction, centered in a 260-pixel box and 730 subjected to 3D classification without symmetry and local searches with increasing 731 sampling rate from 3.7˚, 1.8˚, and 0.9˚ angles. From the resulting three classes, further 732 refinement of class 2 (56.8%) yielded a map of the two tetramers in isolation at an 733 overall resolution of 4.1 Å with improved peripheral density (Supplementary Fig. 4,  734 blue branch, and Supplementary Fig. 6d,e). As the resulting three classes showed 735 different conformations of the MakA head region, we examined whether these 736 conformations exist across the volume. Two subsequent 3D classifications into five 737 classes, without image alignment but with local symmetry searches, first with all 738 classes, then with the top class from the first run, showed a normal distribution of 739 angles, ranging from 48.48˚ to 48.68˚ suggesting continuous motion/rotation along the 740 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 12, 2021. ; https://doi.org/10.1101/2021.09.10.459761 doi: bioRxiv preprint filament axis, which is most pronounced in the tail region (Supplementary Fig. 4, blue  741 branch, lower right). 742 743

Model building, refinement, and validation 744
To obtain an initial model of the tail domain, the MakA crystal structure (PDB-6EZV 11 ) 745 was rigid-body docked into the density map using Chimera 52 and Coot 53 . Regions 746 where the density/model fit was poor (no density, difference in conformation) were 747 trimmed. This included the C-terminal tail (res. 351-365) and the central region (res. 748 ~160-260). Elements in the tail domain with poor density fit were rigid-body docked. 749 The central region of the protein, which includes the neck and head, was built de novo 750 in Coot 53 . The model was first refined against the asymmetric map of the two tetramers 751 in isolation (4.1 Å) using phenix.real_space_refine (version 1.14-3260) 54 . Next, this 752 model was rigid-body docked into the 3.7-Å helical map, two neighboring placeholder 753 molecules were symmetry expanded to provide interaction interfaces, and refined with 754 secondary structure restraints. The final model contains four MakA subunits with 755 trimmed sidechains in the tail domain (N-terminus to 159, 281 to C-terminus, 756 Supplementary Fig. 6e,f). To validate the final model, all atomic coordinates were 757 displaced randomly by 0.5 Å, refined against half map 1, followed by calculating the 758

Statistical analysis 766
The result from replicates is presented as mean ± s.e.m. or mean ± s.d. The statistical 767 significance of different groups was determined by Student's t-tests (two-tailed, 768 unpaired) or one-way ANOVA using Microsoft Excel or GraphPad Prism. *p < 0.05, **p 769 < 0.01, ns = not significant. . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in

1002
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in

1008
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in

1037
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in

1097
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in

34
Nuclei were counterstained with Hoechst 33342. Scale bars, 10 μm. Images in (a) were acquired for a 35 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in

37
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in

47
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 12, 2021.    CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 12, 2021. ; https://doi.org/10.1101/2021.09.10.459761 doi: bioRxiv preprint Two 90-degree related views of the helical reconstruction are shown from the side and the top, next to 111 two slab views. d Two 180-degree related views of the two tetramers refined in isolation are visualized.

112
The cryo-EM densities in (c,d) are colored according to local resolution, which was estimated using

117
Selected bulky residues are indicated and labeled with amino acid code and residue number in red.

119
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted September 12, 2021. ; https://doi.org/10.1101/2021.09.10.459761 doi: bioRxiv preprint