Probing Clostridium difficile infection in innovative human gut cellular models

Interactions of anaerobic gut bacteria, such as Clostridium difficile, with the intestinal mucosa have been poorly studied due to challenges in culturing anaerobes with the oxygen-requiring gut epithelium. Although gut colonization by C. difficile is a key determinant of disease outcome, precise mechanisms of mucosal attachment and spread remain unclear. Here, using human gut epithelial monolayers co-cultured within dual environment chambers, we demonstrate that C. difficile adhesion to gut epithelial cells is accompanied by a gradual increase in bacterial numbers. Prolonged infection causes redistribution of actin and loss of epithelial integrity, accompanied by production of C. difficile spores, toxins and bacterial filaments. This 2-D dual chamber system was used to examine C. difficile interactions with the commensal Bacteroides dorei, and interestingly, C. difficile growth is significantly reduced in presence of B. dorei. Furthermore, in novel multilayer and 3-D gut models containing a myofibroblast layer, C. difficile adheres more efficiently to epithelial cells, as compared to the 2-D model, leading to a quicker destruction of the epithelium. Our study describes new controlled environment human gut models that enable host-anaerobe and pathogen-commensal interaction studies in vitro.


23
Words: 6165  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46 Although animal models of CDI have been used to understand C. difficile pathogenesis and 92 investigate grown as cysts in a matrigel monitored the adhesion and translocation of C. difficile for 1h 130 . While both studies investigated the role of C. difficile toxins in these 131 models, neither was able to follow the infection for a longer period of time due to the lack of an 132 optimal environment for bacterial growth. 133 134 In this study, we have determined the infection dynamics of C. difficile over an extended time 135 frame using novel human gut epithelium models. We demonstrate an increase in the numbers of 136 adherent C. difficile accompanied by production of spores, toxins and bacterial filamentous forms, 137 along with host chemokine production, over 48h in a IEC monolayer VDC model. We demonstrate 138 that this system can be used to study interactions of obligate anaerobes such as the gut commensal, in the control IECs incubated within the VDC, although a slight decline was observed over 24h 155 ( Figure 2A). However, no significant disruption to the IECs was seen after 24h or 48h by 156 microscopy; actin staining showed that the cytoskeleton of the control cells was intact at 24h and 157 48h ( Figure 2B). Immunofluorescent staining of the Snapwell inserts for MUC2, a major mucus 158 protein produced by goblet cells, showed that a small amount of mucus was produced ( Figure S1) 159 in this cell layer which contained 10% goblet cells (9 Caco-2:1 HT29-MTX).

161
Furthermore, to ensure that anaerobic conditions were maintained in the apical chamber, growth 162 of C. difficile in the apical compartment of the VDC was compared to growth in the anaerobic 163 cabinet for 3h and 24h. No negative impact on growth was observed at 3h and 24h compared to 164 bacteria grown in the anaerobic cabinet; instead, a slight increase in C. difficile growth in the VDC 165 was noted at 24h ( Figure S2). 166 167 C. difficile colonization leads to disruption of the intestinal epithelium 168 To determine how C. difficile interacts with the human host in the short and long term, Caco-169 2/HT29-MTX layers were infected with C. difficile R20291 at an MOI of 100:1 for different 170 periods of time in the anaerobic chamber of the VDC (Figure 1). In order to study bacteria that 171 adhere to the IECs and their replication, for all experiments, at 3h p.i., the apical supernatant 172 containing the C. difficile was removed, the IECs washed in PBS, fresh prereduced media added, 173 followed by incubation for the required time. Uninfected controls shown in Fig 2B were run in  174 parallel. The number of adherent C. difficile was determined by counting colony forming units 175 (CFUs) from the cell lysates, after washing off non-adherent bacteria. A significant increase in the 176 number of cell-associated C. difficile was observed from 3h to 24h p.i. ( Figure 2C). This increase 177 in cell-associated bacteria corresponded to a decrease in TEER measurements indicating disruption 178 of the intestinal epithelial barrier (Figure 2A, Figure S3A). Confocal microscopy showed C. 179 difficile present as small micro-communities on the IECs at 24 and 48h ( Fig 2D). An additional 180 image of microcommunities formed at 24h is shown in Fig S3B At early time points (3h and 6h 181 p.i.), there was little disruption of the actin filaments but at 24h and 48h p.i., destruction of the 182 cytoskeleton was evident ( Figure 2D). Interestingly, at 48h p.i., immunofluorescent staining of 183 bacteria showed the presence of filamenting C. difficile ( Figure 2D). A Coloc 2 ImageJ 184 colocalisation analysis revealed partial colocalisation of C. difficile with actin ( Figure 2D) as 185 indicated by the Manders' M2 value (channel 2 for C. difficile) which shows 30% and 20% C. 186 difficile colocalised with the actin at 24h and 48h p.i respectively (Table 1) Positive Li's ICQ  187  values and Costes signifance test values (Table 1) further confirmed colocalisation at both time  188 points.

190
Prolonged infection is associated with spore and toxin production, and host responses 191 In order to fully understand bacterial factors necessary for CDI persistence, we studied the 192 production of spores, toxins and a host chemokine in the 2-D monolayer epithelial model. We 193 measured spores and total bacteria from the cell-associated fraction at different times after 194 infection. Although there were ~0.1% spores in the inoculum, there was a mild but significant 195 increase in spore numbers from 3h to 48h p.i. (p < 0.05, Figure 3A). After washing off the non-196 adherent bacteria at 3h, we also tracked the bacterial spore numbers from 3h-48h in the 197 supernatants ( Figure S4). By 48h p.i., there were equal numbers of spores and total cells in the 198 supernatants ( Figure S4). revealed that IL-8 levels were low at 6h p.i. but increased at 24h and 48h p.i. in this infection 205 model compared to uninfected controls ( Figure 3C). 206 207 Co-culturing C. difficile with Bacteroides dorei results in reduced C. difficile growth within 208 an epithelial gut model.

209
To determine if our VDC-based system could be extended for use with other anaerobic bacteria, 210 we studied a strict commensal anaerobe, B. dorei within the epithelial monolayer model using 211 similar conditions as for C. difficile. Adherence to the IECs was determined at 3h and 24h by CFU 212 counts, as described for C. difficile. B. dorei adhered to the epithelial cells at 3h and multiplied 213 over 24h, as observed for C. difficile (Figure 4 A, B). A mixed culture of B. dorei and C. difficile 214 (1:1), prepared as described in Experimental Procedures was then cultured with the monolayer in 215 the VDC. Cell-associated bacteria were quantitated by plating on a medium used to selectively 216 isolate C. difficile colonies, which also allowed the growth of B. dorei. Bacteroides colonies were 217 distinguished by colony size, color and morphology (small colonies). A significant decrease in the 218 number of C. difficile was observed when co-cultured with B. dorei in the presence of IECs when 219 compared to mono-cultures of C. difficile at 24h p.i., but not at 3h p.i. (Figure 4A, B, p < 0.001). 220 We also observed higher colony counts when B. dorei was grown in co-culture with C. difficile at 221 3h and 24h, compared to mono-cultures of B. dorei ( Figure 4A, B).

223
Increased cell-associated bacteria in multilayer and 3-D models 224 Typically, a myofibroblast layer underlies the basement membrane in the human gut. To develop 225 this system further by increasing its cell complexity, we incorporated myofibroblasts into our 226 model. Human CCD-18co myofibroblasts were first grown on the basolateral side of the polyester 227 Snapwell insert before seeding the IECs on the apical side (multilayer model, Figure 5A). 228 Furthermore, to recreate the highly porous architecture of the ECM in the basement membrane, 229 epithelial and fibroblasts cells were grown apically and basolaterally respectively on inserts 230 containing electrospun nanofiber scaffolds generated from polyethylene terephthalate to develop 231 a 3-D gut epithelium as reported previously (Morris et al., 2014). Scanning electron microscopy 232 revealed that these scaffolds exhibited a uniform nanofibrous matrix ( Figure 5B) that supported 233 the attachment and proliferation of Caco-2 cells to form a confluent monolayer ( Figure 5C) and 234 enabled the proliferation of the CCD-18co cells ( Figure 5D). The average fiber diameter was 457 235 ± 170 nm ( Figure S5).

237
Interestingly, in the multilayer model, we find that after infection with the same C. difficile strain, 238 same MOI's (100:1) and conditions as the monolayer model, for 3h and 24h (Figure 2), higher 239 numbers of bacteria adhered to the IECs at both time points ( Figure 6A) compared to the 240 monolayer model without fibroblasts (Figure 2). Anti-fibronectin staining of the basolateral side 241 of the membrane containing the fibroblast cells indicated likely degradation of fibronectin, as 242 indicated by the destabilsed fibronectin network, and damage to the fibroblast layer at 24h p.i. 243 ( Figure 6B). As before, immunofluorescent staining showed the localisation of C. difficile on the 244 epithelial cells at 3h and 24h p.i. ( Figure 6C). Infections were not followed for longer times (48h) 245 as the cell layer was badly damaged by 24h p.i.

247
Similar to the multilayer model, we observed higher adhesion of C. difficile to the IECs at 3h and 248 24h p.i. in the 3-D gut epithelium ( Figure 6D) compared to the 2-D model data in Figure 2.

249
Similarly, anti-fibronectin staining showed a destabilisation of the discrete fibronectin network 250 and likely cellular damage of the fibroblast layer in the 3-D model, as indicated by the lack of 251 discrete nuclear staining ( Figure 6E). Confocal microscopy revealed the presence of numerous C. 252 difficile on the IECs at 24h p.i. ( Figure 6F). Interestingly, at 24h p.i., immunofluorescent staining 253 showed the presence of filamenting C. difficile ( Figure 6F), much earlier than seen in the 254 monolayer infection model. Bacterial staining did not always correlate to the CFU counts ( Figure  255 6C and 6F), which we attribute to the loss of attached bacteria during the staining procedure.

257
Spore, toxin and host chemokine production in response to C. difficile in 3-D and multilayer 258 models 259 In both the 3-D and multilayer models of infection, spores were found to adhere to the epithelium 260 at numbers comparable to the 2-D monolayer model ( Figure 7A, Figure 3A, Figure S6A), although 261 there was no increase in spore numbers over time. A higher increase in total cell numbers was 262 observed over time in the 3-D and multilayer model (Figure 7, Figure S6A), unlike the monolayer 263 model ( Figure 3A). While there was higher variability, levels of toxin A in the apical compartment 264 supernatants from both models were comparable to that produced in the 2-D models ( Figure 7B, 265 Figure S6B). Levels of toxin B detected in these models were also low ( Figure 7B, Figure S6B). 266 Levels of IL-8 , increased over time (3-24h p.i.) in the 3-D ( Figure 7C), and multilayer models 267 ( Figure S6C), although these were not significantly different to the uninfected controls incubated 268 for 3h or 24h within the VDCs. human gut models for studying interactions of obligate anaerobes with the host gut epithelium. 275 We have tracked C. difficile infection events over a prolonged time frame in an in vitro dual 276 environment human gut model. Along with demonstrating C. difficile adhesion, micro-277 communities, toxin and spore production, we report formation of C. difficile filaments, a potential 278 adaptation mechanism during infection.

280
Previous in vitro infection models have shown the attachment of C. difficile on IECs either using 281 conditions that are not appropriate for the bacterium (such as growing in aerobic conditions) or 282 that are limited to short term infection due to the anaerobic growth requirements of C. difficile 283 ( and sporulate, our findings clearly support previous studies indicating spore adhesion to gut cells 296 during infection. C. difficile spores were reported to adhere to undifferentiated Caco-2 cells after 297 1h of infection as determined by viable spore counts and fluorescence microscopy (Paredes-Sabja 298 and Sarker, 2012), and recently, a spore surface protein CotE was shown to be essential for spore 299 binding to mucus producing epithelial cell layers (Hong et al., 2017). 300 301 Toxin production by C. difficile is known to play a role in pathogenesis by disrupting the barrier 302 integrity of the intestinal epithelium leading to increased permeability and re-organization of actin 303 (Aktories et al., 2017). Surprisingly, although there is a decrease in TEER and actin reorganization, 304 we detect very low levels of toxins at early time points of infection (6-24h). It is possible that low 305 toxin levels, particularly of toxin B, are sufficient to cause loss of the membrane integrity, as they 306 are augmented by other secreted enzymes produced by C. difficile. Although a partial bacterial 307 colocalisation with actin was observed later during infection (24 and 48h) we do not at present 308 understand if C. difficile mediates any direct interactions with the acin cytoskeleton. The formation 309 of C. difficile filaments seen at 24 and 48 p.i. may be associated with host cell contact and infection, 310 as our studies indicate that incubation in similar conditions (conditioned medium from infected 311 cells or growth in DMEM-10 for 48h) in the absence of cells does not induce this phenotype 312 ( Figure S8A and S8B provides structural and biochemical support to cells, fibroblasts are also known to produce 358 chemokines when activated by bacteria (Smith et al., 1997). The increased C. difficile adhesion 359 observed in the multilayer and the 3-D models, which presumably leads to quicker progressing 360 infection, could be the result of indirect modulation of the epithelial barrier by the myofibroblast 361 cells, though at present we do not understand the underlying mechanisms. The increased amount 362 of the chemokine IL-8 observed in the uninfected 3-D and multilayer models could be attributed 363 to the myofibroblasts, although this chemokine response is not specific to C. difficile infection. It 364 is interesting to note that inspite of having higher numbers of bacteria attaching in these models, 365 there was no significant increase in the toxin levels. While we do not observe an increase in spore 366 numbers or distinct toxin production profiles in the 3-D and multilayer models, this may be due to 367 the quick progression of infection due to higher bacterial adhesion. We are currently optimising 368 these models using lower MOI's to enable longer infection experiments.

370
Both the multilayer and 3-D gut models have huge potential in studying C. difficile pathogenesis, 371 particularly for investigating host invasion and interactions with basolateral surface of the host 372 epithelium and are being utilized for studying functions of secreted C. difficile factors in the 373 laboratory. Overall, we have developed highly useful tools for studying C. difficile host-374 interactions which we expect to have broad applications in studying anaerobic gut commensals 375 and their interactions with pathogens and the host. UK), and 1% penicillin-streptomycin (10,000 units/mL penicillin, 10 mg/mL streptomycin, 390 Sigma-Aldrich, UK). HT29-MTX was grown in DMEM and CCD-18co in Eagle's Minimum 391 Essential Medium media. Both media were supplemented with 10% FBS, 1% penicillin-392 streptomycin, 2 mM glutamine and 1% non-essential amino acids (Sigma-Aldrich, UK). All cell 393 lines were maintained in 5% CO2 in a humidified incubator at 37°C and free from mycoplasma 394 contamination as determined by the EZ-PCR Mycoplasma kit (Biological Industries, USA).

396
For the epithelial 2-D models, Caco-2 and HT29-MTX were mixed in a 9:1 ratio and 2 x 10 5 397 cells/ml were seeded on a 12 mm Snapwell inserts (tissue culture treated polyester membrane, 398 Corning, New York, USA) supported by a detachable ring for 2 weeks to form a polarized 399 monolayer. For the multilayer and 3-D models, CCD-18co (1 X10 4 cells/ml) were first seeded on 400 the basolateral layer of the polyester Snapwell insert or electrospun nanofiber scaffold for 5 days 401 after which Caco-2 and HT29-MTX were seeded on the apical side as in the 2-D models for 14 402 days. Prior to the infection experiments, the cell culture medium in the Snapwell inserts was 403 replaced with antibiotic-free medium. The Snapwell inserts containing the polarized cell layer were placed between the two half 407 chambers of the vertical diffusion chamber (Harvard Apparatus, Cambridge, UK) and sealed with 408 the provided clamps. 2.7 ml DMEM with 10% FBS (DMEM-10) was placed on both sides of the 409 chamber. Transepithelial electrical resistance measurements was performed using 410 Harvard/Navicyte electrodes on the EC-800 Epithelial Voltage Clamp (Harvard Apparatus, 411 Cambridge, UK) over 24h. 412 413

Infection of intestinal epithelial cells (IECs) in 2-D, 3-D and multilayer models 414
A single bacterial colony was inoculated in pre-reduced BHI broth (Oxoid, UK) supplemented 415 with 1 g/L cysteine (Sigma-Aldrich, UK) and 5 g/L yeast extract and incubated at 37 o C overnight. 416 The culture was centrifuged at 10,000g for 5 mins (Eppendorf 5810R) and bacterial pellet was 417 resuspended in DMEM-10 and incubated at 37 o C for at least an hour. Bacterial counts determined 418 from this culture were confirmed to be 2 x 10 7 -3 x 10 7 CFU/ml for every experiment. This culture 419 was used to infect the IECs at a predetermined MOI of 100:1 in the apical side of the VDC 420 containing the IECs. The apical chamber was diffused with anaerobic gas mixture (10% CO2, 421 10% H2, 80% N2, BOC, UK) and the basolateral compartment with 5% CO2 and 95% air (BOC, 422 UK). At 3h p.i., the apical media containing the C. difficile was removed, the IECs washed in PBS 423 and 2.7 ml prereduced DMEM with 10% FBS added. It was incubated for a further 3h or up to 424 48h. The apical and basolateral media was then removed and stored at -80 o C. The IECs were 425 washed thrice in pre-reduced PBS before lysing in 1 ml sterile water. Serial dilutions prepared 426 from the IEC lysates were performed and plated on appropriate plates to determine the number of 427 cell-associated bacteria. 428 429

Spore and total cell counts 430
To determine the number of spores, the lysed cells and apical supernatants were heat treated at 431 65 o C for 20 mins as previously described (Fimlaid et al., 2015). Untreated and heat treated samples 432 were serial diluted and plated on BHI and BHI-T agar (supplemented with 0.1% sodium 433 taurocholate, Sigma-Aldrich, UK the apical supernatants were centrifuged at 2500 g for 5 min and 100 µl of the supernatant in 453 duplicates was used for the assay. Duplicates (100 µl) of the standards (toxin A and B) provided 454 with the kit was run in the same ELISA assay as the apical supernatants from which the amount 455 of toxin produced was calculated. IL-8 production was also determined by analysis of basolateral 456 supernatants from VDC's using a human IL-8 ELISA kit (R&D systems, Minneapolis, USA) 457 following the manufacturer's instruction. Duplicates (100 µl) of the standards (IL-8) provided with 458 the kit was run in the same ELISA assay as the basolateral supernatants from which the amount of 459 IL-8 produced was calculated. 460 461 4.8 Electrospinning protocol 462 Electrospinning procedure was performed as described previously ( and M2 values showed that 100% of the actin colocalised with C. difficile. After applying an 522 automatic threshold determined by the software, the levels of colocalisation was lower (Table S1). 523 One hundred randomizations were used to determine the Costes significant P value. The Costes P 524 value if > 95% or 0.95 was deemed to be significant. For all analyses except the Li ICQ, 1 525 represents perfect colocalisation with lesser values indicating the various degree of colocalisation. 526 For the Li ICQ, the values range from maximum 0.5 to -0.5, i.e. with random (or mixed) staining 527 ICQ= ~0; dependent staining 0 < ICQ ≤ +0.5, and for segregated staining 0 > ICQ ≥ -0.5 (Li et  528 al., 2004). 529 530 4.12 Statistical analysis 531 One-way or two-way anova was used to compare two or more groups when there was one or more 532 independent variables respectively with Tukey's test for multiple comparison using Graphpad 533 Prism. Student t test (two tailed) was used to determine the significance between two groups. 534 Significance is represented as *p < 0.05, **p < 0.01, ***p < 0.001 and ****p< 0.0001. Except 535 stated otherwise, all the results presented are the average of 3 independent experiments performed 536 in duplicates or triplicates. The error bars indicate the standard deviation (SD). induce filament formation. Staining with anti C. difficile for 1h was performed, followed by Alexa 894 Fluor 488 goat anti-rabbit secondary antibody (green). Images were taken with a Leica DMi 895 microscope at 100x magnification. Scalebar =10 µM 896 897 898