Structure and assembly of the S-layer determine virulence in C. difficile


 Many bacteria and archaea possess a cell surface layer – S-layer – made of a 2D protein array that covers the entire cell. As the outermost component of the cell envelope, S-layers play crucial roles in many aspects of cell physiology. Importantly, many clinically relevant bacterial pathogens possess a distinct S-layer that forms an initial interface with the host, making it a potential target for development of species-specific antimicrobials. Targeted therapeutics are particularly important for antibiotic resistant pathogens such as Clostridioides difficile, the most frequent cause of hospital acquired diarrhea, which relies on disruption of normal microbiota through antibiotic usage.
Despite the ubiquity of S-layers, only partial structural information from a very limited number of species is available and their function and organization remains poorly understood.
Here we report the first complete atomic level structure and in situ assembly model of an S-layer from a bacterial pathogen and reveal its role in disease severity. SlpA, the main C. difficile S-layer protein, assembles through tiling of triangular prisms abutting the cell wall, interlocked by distinct ridges facing the environment. This forms a tightly packed array, unlike the more porous S-layer models previously described. We report that removing one of the SlpA ridge features dramatically reduces disease severity, despite being dispensable for overall SlpA structure and S-layer assembly. Remarkably, the effect on disease severity is independent of toxin production and bacterial colonization within the mouse model of disease. 
Our work combines X-ray and electron crystallography to reveal a novel S-layer organization in atomic detail, highlighting the need for multiple technical approaches to obtain structural information on these paracrystalline arrays. These data also establish a direct link between specific structural elements of S-layer and virulence for the first time, in a crucial paradigm shift in our understanding of C. difficile disease, currently largely attributed to the action of potent toxins.
This work highlights the crucial role of S-layers in pathogenicity and the importance of detailed structural information for providing new therapeutic avenues, targeting the S-layer. Understanding the interplay between S-layer and other virulence factors will further enhance our ability to tackle pathogens carrying an S-layer. We anticipate that this work provides a solid basis for development of new, C. difficile-specific therapeutics, targeting SlpA structure and S-layer assembly to reduce the healthcare burden of these infections. 

of SLPH (Fig. 1e); N-terminal sequencing revealed truncation of the HID, indicating that this 135 region is unstable in the absence of the LID/HID interaction (Fig. 1e). 136 Our structural model shows that SLPL protrudes from the interacting motif, with D1 closest 138 to the SLPH plane and D2 extending outwards at an angle of ~120°, away from the long axis 139 of D1. Whilst D1 is well ordered, formed by a 5-strand -sheet packed against two -helices, 140 D2 is predominantly composed of long, flexible loops, particularly at the externally-exposed

Crystal lattice reflects in situ S-layer assembly 159
Due to the natural tendency of SlpA to form 2D crystal arrays, we hypothesized that the 160 packing of our crystal structures might reflect the in situ S-layer arrangement. Two H/L 161 complexes, related by pseudo-twofold symmetry, are present in the P1 asymmetric unit, 162 packed in a 2D planar array. The 2D lattices are then stacked to extend the crystal into the 163 third dimension (Extended Data Fig. 3a). The 2D lattice is achieved by tiling of SLPH, with 164 interlocked ridges of SLPL molecules covering gaps between the tiles, creating a tightly 165 packed layer (Fig. 2). Lattice contacts between CWB2 motifs of neighbouring SLPH molecules 166 involve helix-helix interactions between the symmetry-related copies of helix 12 (see 167 topology in Extended Data Fig. 1), as well as electrostatic interactions generating a tightly 168 bonded network ( Fig. 2a and Extended Data Fig. 4). The charge distribution generated by 169 the trimeric arrangement of the CWB2s provides complementary charges across the lateral 170 faces of the SLPH triangular prism tile (Extended Data Fig. 3b and 5a), allowing these 171 interactions to be established (Extended Data Fig. 4). The pseudo-threefold organization of 172 the CWB2s that define the CWP family is also seen in other minor constituents of the S-layer 173 whose structures have been determined 18 . Analysis of the charge distribution in Cwp6 and 174 Cwp8 CWB2s (Extended Data Fig. 5) indicates that charge complementarity could play a role 175 in interaction between lateral faces of CWB2s triangular prisms from different CWPs and 176 SLPH within a mature S-layer. The environment-and cell-facing sides of SLPH exhibit 177 considerable charge differences, with a mostly negatively charged external surface and a 178 largely non-polar cell wall-facing base, decorated by positive patches (Extended Data Fig.  179 5a). The positive patches at the cell-wall base could provide the mechanism for anchoring 180 SlpA to the cell wall via interactions with the anionic secondary cell wall polymer PSII 19 . 181 is similar to wild type, with equivalent interactions between SLPH tiles and D1 domains 229 (Extended Data Fig. 7). Importantly, the absence of D2 exposes pore 1 between SLPH tiles 230 ( Fig. 3d and Extended Data Fig. 7c), which is occluded by interlocking D2 domains in the full-231 length structure. This creates two openings in the array of about 16 Å, with potential 232 functional implications as it could indicate a more permeable S-layer than in the wild type 233 structure, with twice as many pores, of slightly increased size. The retention of S-layer integrity, despite loss of an exposed structural domain, provided an 249 unprecedented opportunity to directly assess the role of SlpA in C. difficile infection. We 250 employed the mouse model of acute disease, which allows a nuanced analysis of 251 colonization and pathology, typified by weight loss and caecal and colon inflammation 21 . 252 Animals infected with the wild type strain R20291 (producing SlpAWT) lost on average 8% of 253 their body weight within 24 h of infection and 12% by the peak of infection at 48 h, before 254 returning to pre-infection weights approximately 4 days after infection. In contrast, 255 infection with R D (producing SlpAR D ) resulted in little apparent disease; animals 256 displayed only a 1% weight loss after 24 h and 5% after 48 h, before a gradual return to pre-257 infection weight (Fig. 4a). As R D was derived from our previously characterized slpA 258 mutant FM2.5 15 , a control for any potential background genetic variation was needed. A 259 previously characterized strain FM2.5RW 15 contains a watermarked copy of the wild type 260 slpA gene (encoding SlpARW) and R D was constructed in a similar way but contains a 261 truncated version of slpA (see Methods for details). Animals infected with FM2.5RW strain 262 showed similar patterns of disease as those infected with wild type. 263 To determine if the surprising loss of virulence seen for R D could be attributed to 264 impaired colonization, we quantified C. difficile present in faeces collected each day post-265 infection (Fig. 4b) and in intestinal contents post-mortem (Extended Data Fig. 8). No 266 statistically significant differences in bacterial recovery were observed, demonstrating that, 267 although D2 is surface-exposed in the S-layer, it is not required for efficient colonization. 268 R D sporulated normally in vitro (Extended Data Fig. 9a), but displayed an increased 269 sensitivity to lysozyme in comparison to the wild type strain R20291 (Extended Data Fig. 9b). 270 As the perceived dogma is that the intestinal pathology and symptoms associated with CDI 271 are largely toxin-driven, we next examined toxin expression and activity in vitro (Extended 272 Data Fig. 9c, 9d) and in intestinal contents ( Fig. 4c and Extended Data Fig. 8f). Surprisingly, 273 no toxin production defect was observed, suggesting the decrease in disease severity seen 274 with R D is a direct consequence of the loss of the surface-exposed D2 domain. Strikingly, 275 although toxin activity was equivalent in all strains, a reduced level of epithelial damage was 276 observed in tissue from R D2-infected mice. Indeed, assessment of other markers including 277 the extent of inflammatory infiltrate, tissue edema and crypt length measurements 278 indicated that damage was reduced in R D infected mice when compared with animals 279 infected with WT or control strain at the acute point of infection (48 h post-infection; Fig. 4d  280 and Extended Data Fig. 8c). 281 Together, these observations demonstrate that the S-layer contributes directly to C. difficile 282 disease severity, in a toxin-independent manner. Importantly, our data also reveal that the 283 most surface-exposed domain of SlpA is dispensable for effective colonization, suggesting 284 that other regions of the protein or other CWPs are likely to be involved in direct host-285 pathogen interactions required for colonization. 286 287

Discussion 288
Here we report the first experimentally determined structure of a complete S-layer protein 289 from a medically important human pathogen, which allows us to observe the organization 290 of the paracrystalline array at an atomic level. The tight packing of H/L complexes in the 291 crystal replicates assembly into the functional S-layer which we observe in situ by electron 292 microscopy. The repeating crystallographic array is created by tiling of one of the subunits 293 (SLP H ), which also anchors the S-layer to the cell wall. Most other S-layers characterized to 294 date are composed of distinct domains that contribute either to lattice formation 295 (crystallization domain) or cell surface attachment (anchoring domain) 2 . In the C. difficile S-296 layer, the crystallization and anchoring functions are combined in the SLPH, with assembly 297 relying on contacts between adjacent tiles. Moreover, contacts between SLPL and 298 neighbouring H/L complexes further expand the S-layer assembly network. The SLP L ridges 299 are also important for generating a structure impermeable to the majority of folded 300 proteins such as lysozyme and other large molecules as it covers the pores present within 301 the packing of the triangular prism SLPH tiles. This tight packing raises the question of how 302 large molecules such as the C. difficile toxins 22 are exported to the environment. 303 Furthermore, S-layers must be able to accommodate cell growth and division and this array 304 needs to adapt to the curvature of the cell poles. Having tightly packed core subunits or 305 domains, maintained by interchangeable electrostatic interactions, that are then decorated 306 with more flexible regions is a simple, yet seemingly effective, way to achieve both 307 requirements. Points of mismatched symmetry as observed by tomography (Extended Data 308  Our investigation of the functional role of the S-layer revealed that toxins are necessary but 314 not sufficient for full disease severity, a paradigm shift in our understanding of C. difficile 315 infections. Despite being dispensable for protein fold or even S-layer assembly, D2 seems to 316 confer a functional role to SLPL and its absence leads to reduced disease severity. 317 Surprisingly, this is not due to changes in colonization of the gut, suggesting other domains 318 or S-layer proteins are involved in this type of interaction with the host. Instead, the 319 presence of D2 is associated with increased levels of inflammation when compared to the 320 full-length protein (Fig. 4). In C. difficile infections, tissue inflammation has been associated 321 with activation of additional immune pathways that results in enhanced disease 23 . The high 322 sequence variability and structural flexibility of the D2 domain, in contrast to the conserved 323 and relatively rigid SLPH, could therefore confer an immune-evasion mechanism as a result 324 of the evolutionary pressure of the dynamic environment of the gut. It is therefore tempting 325 to speculate that D2 is directly involved in the activation of the host immune response, 326 however, the molecular mechanisms involved remain to be elucidated (see SI discussion). 327 Importantly, we have established a direct link between the S-layer and disease severity and 328 our characterization of S-layer assembly in C. difficile reveals new potential therapeutic 329 avenues. The interacting SLPH subunits and the flexible D2 domains present key targets for 330 disruption of the S-layer, and molecules that affect S-layer assembly are attractive 331 therapeutic agents. 332           Table 2. Plasmid pRPF233, containing a copy of the 14 complete slpA gene from C. difficile strain R20291 was modified by inverse PCR using 15 oligonucleotides RF102 and RF103 to delete the coding sequence of SlpA residues 115-259 16 and replace with GGAGGT, encoding two glycine residues. The resulting plasmid, pOB001, 17 was transferred to the C. difficile S-layer mutant strain FM2.5 3 by conjugation 4 . FM2.5 18 displays an aberrant colony morphology that is easily distinguished from wild type C. 19 difficile. Recombination between the plasmid-borne slpA gene and the mutated copy on the 20 chromosome was detected by reversion to normal colony morphology. Plasmid curing was 21 confirmed by loss of thiamphenicol resistance, the chromosomal location of the engineered 22 slpA gene was confirmed by PCR and the resulting protein profile was determined by SDS-23 PAGE of S-layer proteins isolated using standard methods (see below). 24

Plasmid construction 25
For crystallization studies, fragments of R7404 slpA, encoding mature HID (residues 1-41) 26 and SLPL (residues 1-316) or LID (residues 240-316) were amplified from genomic DNA and 27 cloned into pACYC-Duet1 yielding plasmids pJAK149 and pJAK147, respectively. C-terminally 28 6x His-tagged HID was amplified using RF1396 and RF1397 and cloned into pACYC-Duet 29 (MCS1) linearized using RF1398 and RF1400 by Gibson assembly, and SLPL or LID were 30 amplified using RF1394 and RF1395 or RF1395 and RF1396, respectively, and cloned into 31 MCS2 using NdeI-KpnI restriction cloning. 32 To study protein-protein interactions in vitro, DNA encoding mature SLPL or SLPH of CD630 33 and R7404 was amplified using Q5 (NEB) PCR and cloned into pET28a using NcoI-XhoI 34 restriction cloning, in frame with a C-terminal 6x His-tag. Deletion variants (see Table 1 for 35 construct details) lacking HID or LID, or point mutants within HID and LID were constructed 36 by inverse PCR, using primers listed in Table 2, as previously described 5 . 37 To study the impact of individual LID and HID point mutations on H/L complex assembly in C. 38 difficile, codons for SLPL F274 or SLPH Y27 in pRPF170 were mutated to GCA (Ala) by inverse 39 PCR cloning, yielding plasmids pRPF209 and pJAK186, respectively. Tris-HCl pH 7.5, 500 mM NaCl, 0.1% Triton X-100, followed by 50 ml of 50 mM Tris-HCl pH 74 7.5, 250 mM NaCl, 5 mM β-cyclodextrin. Affinity purification was performed in 50 mM Tris 75 pH 8.0, 250 mM NaCl, 10 mM imidazole, with protein elution by a linear gradient of 76 imidazole (10-500 mM). Each chromatographic step was followed by analysis of the eluate 77 by 12% SDS-PAGE. induced with anhydrotetracycline (20 ng ml -1 ). Surface-localized H/L subunits were extracted 85 using low pH glycine as described above and normalized to an equivalent OD600nm of 25. 86 Culture supernatants were filtered, concentrated to an equivalent OD600nm of 50 using a 87 Vivaspin column with a 10 kDa MWCO. Samples were then subjected to SDS-PAGE and 88 western immunoblotting using polyclonal antibodies specific for the CD630 SLPH or SLPL 5 . 89 90

Analysis of protein-protein interactions by enzyme-linked immunosorbent assay (ELISA) 91
The assays were performed as previously described 5 . Briefly, Maxisorp microtiter plates 92 (Nunc) were coated with 10 μg ml -1 of SLPL or SLPH and their variants (Table 1) Protein Crystallography Beamline (ISPyB), processed with XDS 7 , iMosflm 8 or DIALS 9 and 113 scaled with Aimless 10 within CCP4i 11 or CCP4i2 12 software suites. When needed, density 114 modification was performed with PARROT 13 . 115 The structure of LID/HID-6x His was solved de novo using Arcimboldo_lite 14 within CCP4i, 116 starting from several 10-14 residues-long polyalanine models of α-helices. Automatic model 117 building was performed with Buccaneer 15 , followed by manual building with Coot 16 and 118 refinement with Phenix_refine 17 . 119 The structure of SLPL/HID-6x His-tag was determined by sequential molecular replacement 120 in Phaser 18 searching first for SLPL D1-D2 domains model (PDB ID: 3CVZ 5 ), followed by the 121 search with LID/HID structure into a fixed SLPL solution, and subsequent manual building 122 (COOT) and refinement (Phenix_refine). Ramachandran outlier rotamers, respectively. Full data collection and refinement statistics 141 are summarized in SI Table 1. Validation of final models was performed using COOT and 142 Phenix internal tools, as well as MOLPROBITY 24 web server. All structural models were 143 validated using wwPDB validation server prior to deposition of files. Data collection and 144 refinement details are summarized in SI Table 1

Electron crystallography data processing 178
Images were initially processed using the 2dx suite 28-30 . Most micrographs of S-layer ghosts 179 showed two rotationally separated lattices in Fourier transforms and these were indexed 180 independently. Images were masked based on crystal size and good crystalline order and 181 subjected to two cycles of unbending using the programs QUADSEARCH and CCUNBEND. 182 The symmetry was determined from images of untilted crystals using ALLSPACE 31 . Phase 183 origins for individual images were refined against each other using ORIGTILTK, sequentially 184 adding images of higher tilt to the refinement. Crystal tilt angles were estimated from lattice 185 distortion. LATLINE 32 was used to determine interpolated amplitudes and phases on a 186 regular lattice of 1/160 Å −1 in the z* direction. A Gaussian tapered real-space envelope of 187 width slightly larger than that of the H/L complex estimated from the X-ray crystal structure 188 (70 Å for wild type and 60 Å for RΔD2) was applied. The phase origin and tilt parameters 189 were further refined using the output interpolated lattice lines as reference. The variation of 190 amplitude and phase along 0,0,l was estimated by examining a plot of maximum contrast on 191 each Z-section in real space 33 . The final structure factors were sampled from the 192 interpolated lattice lines 32 and a 3D map generated within the CCP4 suite of programs 11,28 . 193 Cryo-EM micrographs of untilted R20291 and RΔD2 samples were processed similarly to 194 generate 2D projection maps. B-factors were calculated using SCALIMAMP3D with 195 bacteriorhodopsin diffraction amplitudes as reference 34 . Data collection, processing and 196 analysis details are summarized in SI Tables 2 and 3. 197 198 Fitting X-ray structures to EM density 199 The coordinates of R20291 and RΔD2 H/L complex X-ray structural models were fitted using 200 For cryo-electron tomography (cryo-ET), the homogenized S-layer ghost sample used in 209 electron crystallography was mixed with an equal volume of 10 nm BSA-treated nanogold 210 beads, and 3 μl of this mixture was applied to a glow discharged lacey carbon with ultra-thin 211 carbon 300 mesh grid, blotted for 3 s and plunged into liquid ethane, using a Leica EM GP. 212 The frozen grids were stored in liquid nitrogen temperature for later observation. To assess resistance to lysozyme, overnight C. difficile cultures were grown in TY broth, 231 subcultured to an OD600nm of 0.05 in 1 ml fresh TY in a 1.5 ml cuvette and then grown for 8 h 232 with hourly OD600nm measurements. Where appropriate, lysozyme (500 μg ml -1 ) was added 233 after 2.5 h growth. Experiments were performed in triplicate on biological duplicates and 234 data expressed as the mean and standard deviation.  Our structural models and functional analysis of the S-layer of C. difficile provide the first 3 detailed insights of this important layer in a human pathogenic bacterium. This work allows 4 us to explore this array in unprecedented detail, both elucidating key features and raising 5 new questions requiring further investigation. 6 7 S-layer assembly: how can a 2D crystal array remain flexible? 8 A degree of conformational flexibility is required to accommodate the wrapping of the 2D S-9 layer lattice around the curved surface of the C. difficile cell. Indeed, dynamic flexibility 10 between S-layer protein domains has been shown to promote efficient crystal nucleation on 11 the curved cellular surface in Caulobacter crescentus 1 . Our recent work showed formation of 12 C. difficile S-layer at specific sites coinciding with cell wall synthesis 2 , suggesting discrete S-13 layer assembly points. Furthermore, Fourier analysis of S-layer ghosts and tomographic 14 imaging (Extended Data Fig. 10) indicates a highly mosaic surface, with many crystal defects, 15 particularly at the cell poles, where the paracrystalline array must be disrupted to allow for 16 cell curvature. The pattern of crystalline patches with grain boundaries observed is 17 consistent with the random secretion of S-layer protein monomers and self-assembly of 2D 18 crystals occurring at gaps and grain boundaries within the curved S-layer, as proposed for 19 other organisms 3-5 . 20 While increasing numbers of S-layer structural models are available 6-11 , to our knowledge, 22 this is the first report of a complete X-ray structure of a major S-layer protein where the 23 crystal lattice mimics S-layer assembly in the cell. This indicates that S-layer assembly in C. 24 difficile does not require an underlying ordered polysaccharide array, unlike the apparent 25 organization observed in LPS-mediated S-layer anchoring in the Gram-negative C. 26 crescentus 10 . As the C. difficile S-layer is anchored via interactions of the CWB2 motifs with 27 PSII 12 , a much simpler glycan that is unlikely to be ordered at the cell surface, it is not 28 surprising that the protein can assemble independently. In order to elucidate the anchoring 29 mechanisms of C. difficile S-layer, we are investigating the interactions of SPLH and the H/L 30 complex with PSII using a combination of biochemical, biophysical and structural methods. 31 SlpA is the main component of the S-layer in C. difficile but additional proteins, which 33 together correspond to an estimated 10% of the protein molecules forming the array, must 34 be accommodated within the layer. Our assembly model suggests that tiling of the CWB2 35 triangular prism present in SLPH and in all minor cell wall proteins (CWPs) 13 , is a mechanism 36 that allows insertion of these proteins while maintaining the crystalline arrangement. In our 37 crystal structure, SLPH tiling is maintained by charge complementarity of interacting 38 triangular prism surfaces (Extended Data Fig. 3b). Our structural analysis of homology 39 models of other SLCTs suggests that most of the interactions between neighbouring H/L 40 complexes which define those interfaces are conserved across different SLCTs (Extended 41 Data Fig. 4), indicating that they are likely to be important for C. difficile S-layer assembly. It 42 is worth noting that the most conserved interactions are at the interface of neighbouring 43 CWB23-CWB21 motifs and around pore 2, and involve residues from both SLPH and SLPL 44 (Extended Data Fig. 4b, top 6 rows). This conservation across SLCTs suggests that SLPL is also 45 important to maintain the S-layer assembly and that these interactions could be potential 46 targets for disrupting the array. 47 Our analysis of the charge distribution of the CWB2 motifs in Cwp6 and Cwp8, the only 48 other CWB2-containing proteins whose structures have been determined to date 14 , 49 indicates that the charge complementarity between H/L complex and these minor S-layer 50 components would also be possible, supporting our proposed global assembly model. The The pores observed in our crystal lattice (Fig. 3) are highly hydrophilic (Extended Data Fig.  86 3e), suggesting most small charged molecules could diffuse into the cell. Pore 1, mostly 87 occluded by D2 in the full H/L complex, has a mixed charged distribution, with patches of 88 both positive and negative charges throughout (Extended Data Fig. 3f), suggesting 89 important metabolites such as ATP (negatively charged) or metal ions (positively charged) 90 could have access. In contrast, the fully exposed pore 2, is mostly negatively charged, 91 indicating that positively charged small metabolites could preferably be diffused via this 92 pore. The presence of a positive patch formed by two pseudo symmetry-related lysines 93 covering the outermost opening of this pore (Extended Data Fig. 3f) could provide a gating 94 mechanism for these metabolites. 95 It is worth noting that interacting D1 domains from neighbouring SLPL molecules completely 96 cover the widest cavity in the SLPH CWB2s tiling. This interface, defined by neighbouring 97 CWB21-CWB22 motifs, at around 20 Å wide but spanning over 100 Å across the triangular 98 prism tiles, is also hydrophilic, with complimentary charges (Extended Data Fig. 3b). The 99 SLPH CWB2 motifs tiling also creates a cavity of approximately 70 Å between symmetry-100 related molecules which is partly occluded by the HID and LID domains, with the 101 interlocking D1 domain ridges covering this gap (Extended Data Fig. 3b). If the interacting D1 102 domains are flexible and can at least partially expose these cavities, it could potentially 103 allow diffusion of larger molecules through the S-layer. 104 Absence of D2 creates a more permeable S-layer as it exposes pore 1 in the SLPH tiling ( Fig.  105 3, Extended Data Fig. 3). Moreover, many of the residues lining the two exposed pores in 106 this lattice are not resolved in the electron density of the SlpARΔD2 and could not be 107 modelled, suggesting weaker interactions. A scenario where D1-LID/HID interactions with 108 the CWB2s are less stable and could allow access to the wider openings in the CWB2s tiles 109 would possibly explain the susceptibility to lysozyme seen in the RΔD2 strain (Extended Data 110 Our work with RΔD2 strain, producing a modified S-layer is, as far as we are aware, the first 115 example of a C. difficile strain exhibiting both similar toxin and colonization levels but 116 reduced disease severity in a mouse model. The reduction in weight loss upon infection with 117 a strain of C. difficile lacking the surface-exposed D2 domain of SLPL suggests that the S-layer 118 and toxins act synergistically to mediate epithelial damage, highlighting the possible 119 Another possibility is that modifications in S-layer-mediated signaling influences the 144 downstream generation of specific proinflammatory cytokines such as IL-23. Infection of IL-145 23 -/-mice with C. difficile resulted in limited tissue edema, reduced inflammatory influx 146 and less epithelial damage 24 ; mirroring the observations in animals infected with the 147 modified RΔD2 strain in this study. Further, in vitro, a combination of cell filtrates and toxin 148 but not toxin alone, was required to stimulate the expression of IL-23 in both mouse and 149 human bone marrow-derived macrophages 25 . As these filtrates are likely to contain high 150 levels of SlpA, it could be speculated that IL-23 expression relies on two independent 151 signals, the first provided by the toxin (through inflammasome activation) and the second, a 152 MyD88-dependent event implicating TLR signaling, possibly involving SlpA. Therefore, if the 153 D2 domain is essential for TLR signaling, reduction of disease could be linked to prevention 154 of downstream IL-23 mediated events such as enhancement of inflammation and sustained 155 barrier damage. 156 While the specific mechanism by which S-layer is able to modify the severity of toxin 157 mediated-disease is currently unclear, we now have the tools and structural knowledge to 158 allow us to dissect and determine the relevance of the S-layer structure in C. difficile 159 disease. 160  1 Two crystals were used to determine the structure of H/L complex R7404. All others required one crystal only. 3.5 4.1 3.6 -* Mean value phase error against symmetry-imposed phase of 0° or 180° (45° is expected for random phases 1 ). *Represents most likely plane group 1 a and b represent the respective symmetry axis for the plane group 2 Target residual indicates the expected phase residual of each symmetry group based on the signal-tonoise ratio of the respective reflections 2 .