A urine-dependent human urothelial organoid offers a potential alternative to rodent models of infection

Murine models describe a defined host/pathogen interaction for urinary tract infection, but human cell studies are scant. Although recent human urothelial organoid models are promising, none demonstrate long-term tolerance to urine, the natural substrate of the tissue and of the uropathogens that live there. We developed a novel human organoid from progenitor cells which demonstrates key structural hallmarks and biomarkers of the urothelium. After three weeks of transwell culture with 100% urine at the apical interface, the organoid stratified into multiple layers. The apical surface differentiated into enlarged and flattened umbrella-like cells bearing characteristic tight junctions, structures resembling asymmetric unit membrane plaques, and a glycosaminoglycan layer. The apical cells also expressed cytokeratin-20, a spatial feature of the mammalian urothelium. Urine itself was necessary for full development, and undifferentiated cells were urine-tolerant despite the lack of membrane plaques and a glycosaminoglycan layer. Infection with Enterococcus faecalis revealed the expected invasive outcome, including urothelial sloughing and the formation of intracellular colonies similar to those previously observed in patient cells. This new biomimetic model could help illuminate invasive behaviours of uropathogens, and serve as a reproducible test bed for disease formation, treatment and resolution in patients.


Results
HBLAK and HBEP can form three-dimensional urothelial organoids. HBEP cells were derived from normal human bladder biopsies by CellNTec and provided commercially in cryovials. We also grew HBLAK cells, which are spontaneously immortalised but not transformed version of HBEP available from the same company. These latter cells retain the ability to differentiate but have increased longevity without senescing. On thawing, both cell types were seeded on plastic in fully defined, serum-free, BPE containing CNT-Prime media, which favours the proliferative phenotype. Both HBEP and HBLAK shared a 'spindle-like' morphology, a hallmark of multipotent epithelial progenitors. When 70% confluent, the cells were transferred to Millicell transwells (Millipore) with CNT-Prime media in the apical and basal chambers and grown to confluency. At this point, the media was shifted to high-calcium, differentiation media (CNT-Prime-3D) in both chambers. After overnight incubation, we replaced filter-sterilized human urine in the apical chamber and left the cultures to develop for 14 (HBLAK) to 25 (HBEP) days depending on the experiment, with periodic media and urine changes. At endpoint, the organoid-coated filters were retrieved, fixed, and stained for various biomarkers and inspected microscopically.
3D confocal analyses showed that both cultures were viable despite the prolonged presence of urine. The HBEP tissue contained multiple layers (approximately 3) with tightly-packed spheroid basal cells, intermediate cells and enlarged and flattened umbrella-like cells at the apical surface (Fig. 1a,b). The HBLAK tissue was morphologically similar, but in contrast, likely due to increased rate of growth, these produced approximately 5-7 cell layers with multiple layers of intermediate cells (Fig. 1c,d). Optical slices at the basal region of the HBEP and HBLAK organoids respectively showed the typically small, tightly-packed and spheroid morphology of urothelial basal cells (Fig. 1e,g). Basal cells in the HBLAK organoid (~ø10 µm) appeared to be slightly smaller than those in the HBEP culture (~ø20 µm) (Fig. 1e,g). Similarly, single optical slices at the apical regions of the HBEP and HBLAK organoids showed the formation of a large, flat and often hexagonal cellular morphology typical of well-differentiated umbrella cells ( Fig. 1f and i for HBEP, Fig. 1h and j for HBLAK). As shown in Fig. 1j, lower-magnification views of HBLAK show a more heterogeneous differentiation pattern, with cells elaborating distinct multi-layered zones of organoid formation with large umbrella-like cells at their surface, flanked by regions of very much smaller undifferentiated basal cell-like monolayers and occasional areas of hypertrophy.
In summary, both models formed large zones of three-dimensional epithelia in a manner reminiscent of a urothelium, with umbrella-like cells at their apical surface. Moreover, both models were exposed to sterile human urine for several weeks (14-25 days) without exhibiting any signs of toxicity.
HBLAK and HBEP organoids exhibit key biomarkers of the human urothelium. We characterised the HBEP and HBLAK human urothelial organoids further by targeting urinary tract-specific antigen expression using indirect IF in conjunction with high-resolution laser scanning confocal microscopy. The resulting organoids had the correct spatial expression of several key biomarkers. Studies exploring cytokeratin (CK) expression in normal human bladders have shown a relationship between the level of cytodifferentiation and sub-type CK expression 17 . Both HBEP-and HBLAK organoids exhibited correct spatial expression of CK8 and CK20; specifically, CK8 was expressed throughout the strata of the in vitro tissues whereas CK20 was expressed preferentially by the umbrella-like cells at the apical surface ( Fig. 2a,b,c,d). In contrast to the rodent bladder, muscarinic receptors are expressed throughout the human urothelium 46,47 . This finding was echoed in our model with evidence of muscarinic receptors M2 and M3 expressed in all cell layers of both organoids ( Fig. 2e and g). Uroplakin-III (UP3), an indispensible part of the asymmetric unit membrane 48 , was present in the apical cells of the HBEP (Fig. 2f,i) model, although the orthogonal view (bottom, Fig. 2f) shows that this protein was not heavily enriched on the surface. In contrast, in the HBLAK (Fig. 2h) organoids, UP3 was strongly surface-expressed (see orthogonal view, bottom Fig. 2j). In both models, from the apical view, the UP3 staining appeared in a speckled pattern similar to that seen in urothelial cells shed in the urine of a chronic UTI patient (Fig. 2j,l). Multinucleation of the umbrella-like cells was a relatively rare occurrence (e.g. examples can be seen in Fig. 2gi and k).
HBLAK organoids possess correct topographical and ultrastructural features. Given these promising results, we went forward with the HBLAK model, as it was more tractable, thicker, and presented better surface-staining of UP3. We used scanning electron microscopy (SEM) and transmission electron microscopy (TEM) to analyse its ultrastructure, organisation and overall topography. Low-power SEM echoed what was seen in Fig. 1j, namely that the HBLAK cells elaborated distinct multi-layered zones of organoid formation with umbrella-like cells at their surface (approximately two-thirds of the tissue, 64.8% +/− 10.4) along with zones of undifferentiated basal cell-like monolayers (Fig. 3a) and areas of disorganised hyperplasia (Fig. 3b). The surface of each differentiated zone exhibited very large (up to ~80 µm), often hexagonal umbrella-like cells (Fig. 3c,d,e). A comparison with the undifferentiated basal cells (Fig. 3e vs. f) showed that the umbrella-like cells were up to approximately 50 times larger in terms of surface area than their undifferentiated counterparts. SEM micrographs also showed evidence of potential tight junction formation (Fig. 3c) 49 and structures resembling, in size and spacing, characteristic microplicae or 'hinges' at the apical surface of each umbrella-like cell ( Fig. 3g) 49,50 . Specifically, the protrusions were approximately 200 nm long, 100 nm in width and spaced approximately 500 nm apart, which is consistent with the reported size of microplicae 50,51 ; this length is far smaller than the usual length of epithelial microvilli, which is usually approximately 1 µm 52 . Orthogonal sections of fully differentiated organoid zones were analysed using TEM. As with the laser scanning confocal imaging, TEM elucidated distinct layers of basal, intermediate and umbrella cells (Fig. 3h). Structures consistent in size and spacing with rigid plaques could be seen residing between the 'hinges' at the apical surface of the umbrella cells (Fig. 3i,j) 49,50 . Structures were also seen that may possibly correspond to the specialised fusiform vesicles (Fig. 3j) necessary for trafficking uroplakins 53 -although the use of specific markers would be necessary to confirm these. In contrast, TEM analysis of the small, undifferentiated cells making up the monolayers flanking the organoid zones did not exhibit the putative hinges, plaques or the structures than might be fusiform vesicles (Fig. 3k,l). Taken together, these observations suggest that the fully differentiated regions of the organoids possess some key features expected of a human urothelium.

Urine strongly influences HBLAK differentiation, organoid development and GAG formation.
As shown above, the HBLAK organoid expresses key urothelial markers and is morphologically reminiscent of a urothelium. Of key importance, however, is the tolerance exhibited by these cultures to urine. In this group of experiments we investigated whether urine is merely tolerated or indeed necessary for differentiation in 2D, organoid formation and the elaboration of a GAG layer. To achieve this, we exposed the cells, in 2D and 3D, to various dilutions of sterile human urine before examining microscopically.
To decouple the influence of urine from any effects that might be exerted by developing intermediate cells below, we grew HBLAK cells as 2D monolayers on chamber slides and exposed them to medium containing urine. As shown in Fig. 4, HBLAK cells exhibited a marked dose response to urine after 72 hours. Cells grown in high calcium medium alone took on a small, tightly packed basal cell-like appearance with visually well-formed intercellular junctions as assessed by actin staining (Fig. 4a) and the elaboration of actin-containing microvilli (Fig. 4a inset, red arrow). With the addition of 25% human urine, however, a small proportion of the cells began to exhibit an enlarged morphology, but intercellular junctions were partially disrupted (Fig. 4b). After 72 hours in 50% sterile human urine, extensive colonies of HBLAK cells took on a large, flat umbrella cell-like morphology (Fig. 4c) devoid of microvilli ( Fig. 4c inset, red arrow). However, intercellular junctions seemed almost entirely compromised, possibly because increased urine came at the expense of calcium concentration. Strikingly, we found human urine to be necessary for organoid formation and the elaboration of a GAG layer in HBLAK cells. Cells cultured for 14 days on filter inserts with high calcium medium in the basal and apical compartments were comprised of only one layer, showing no stratified organoid formation, little heparan sulphate, and no detectable expression of chondroitin sulphate (Fig. 4d). Moreover, these cells appeared to exhibit a degree of anaplasia, with a more 'spindle-like' multipotent progenitor cell-like morphology (data not shown). In contrast, HBLAK cells cultured in 50% or 100% human urine at the apical surface produced zones of organoid formation between 3 and 6 cell layers thick (Fig. 4e,f) as previously seen in the experiments presented in Fig. 1, and both stimulated the expression of a heparan and chondroitin sulphate-rich GAG layer at the urine-umbrella cell interface (Fig. 4e,f). Taken together with the data in Fig. 3, these experiments suggest that urine is an indispensable effector of urothelial differentiation in this model. Nevertheless, despite a lack of GAG layer and of the structures that appear to be asymmetric unit membrane plaques, undifferentiated cells remain viable for long periods in urine, which suggests that these "barriers" are not required for urine tolerance. The HBLAK organoid is a potential model for studying Enterococcus infection. Our laboratory studies the uropathogen Enterococcus faecalis. This bacterium is commonly implicated in chronic UTI in the elderly and is frequently associated with multi-drug resistance, hospital-acquired infection, and catheter-associated biofilm formation; we previously demonstrated that it exhibits intracellular invasion in patient cells [54][55][56][57][58][59][60][61][62] . To determine whether the organoid is a good model for studying host-pathogen interactions with this bacteria, we infected the HBLAK organoid with patient-isolated E. faecalis 55 . When we inspected the cultures two hours after infection, we found that E. faecalis exhibited robust tropism in this model, with the resulting adherent colonies relatively loosely packed (Fig. 5a). As with human patients suffering from acute and chronic UTI, the apical cell layer was shed in response to bacterial insult 55 , leaving an uneven surface of basal and intermediate cells (Fig. 5a,d; also compare the uneven surface of orthogonal views in Fig. 5d to the smooth surface evident in orthogonal views of the intact, uninfected HBLAK organoid in Figs 1d and 2d). Inspection of supernatants post-infection revealed extensive shed umbrella-like cells (Fig. 5b) reminiscent of those seen in the urine of infected patients (Fig. 5c and 55 ). Significantly, in the tissue that remained, we saw frequent examples of large intracellular bacterial colonies (Fig. 5d, arrows) within the superficial layer of cells, showing a similar loosely-packed morphology to those we previously observed in shed urinary umbrella cells from patients 55 . Image analysis showed that 7.3% of cells harboured intracellular colonies with a mean of 37.8 (+/−SD of 11.5) bacteria per cell. A gentamicin protection assay performed on organoids infected with E. faecalis for two hours (Fig. 5e)   of 4.8 × 10 3 cfu/ml in the lysate (approx. 55% bacterial recovery). A non-parametric Friedman test of differences among repeated measures was conducted on the median cfu/ml data from the untreated (control, N = 3) and gentamicin treated (N = 3) organoids. A statistically significant difference was found among the gentamicin treated organoids (χ 2 = 6, p = 0.04) whereas no difference was found among the untreated control data (χ 2 = 3.8, p = 0.15). Taken together, these findings suggest that the model recapitulates some key aspects the host/pathogen interaction of E. faecalis with its human host.
In summary, this model, to our knowledge, represents the first long-term urine-tolerant human bladder organoid produced in vitro. This tissue is reminiscent of normal human bladder urothelium and expresses a number of key markers in the correct spatial compartments in response to exposure to urine. As with patients, this organoid rapidly sheds the apical cell layer in response to bacterial insult. Moreover, E. faecalis displayed invasive phenotypes, supporting our previous findings in shed patient cells 55 .

Discussion
The development and use of in vitro human tissue mimetics is thought to be accelerating the drug discovery process 63 and improving our understanding of human tissue morphogenesis. Due to improved physiological relevance, such models could even reduce the use of animal models in the coming years 64 . Here, we present a urine-tolerant, three-dimensional urothelial organoid derived from human progenitors that is easy to grow from commercially-sourced, quality-controlled materials, that displays key hallmarks of the human urothelium, and which may serve as an alternative to the murine model. If desired, it should also be possible to create similar urine-tolerant organoids from fresh human biopsies using the protocols and defined medium we describe. An even more physiological response could well be achieved by embedding the organoid into a perfusion-based bioreactor allowing the flow of apical urine, which would introduce relevant shear forces and metabolite flux 45 . Further improvements could be achieved by seeding the organoid onto different scaffolds; by co-culturing cells that could secrete an appropriate extracellular matrix; and/or by introducing a stretch parameter 65 .
As this organoid model was generated in a similar manner to others that have been reported 21 , it is not clear why it has superior urine tolerance. Additional work in our laboratory has not revealed any differences in outcome depending on the source of the human urine. The primary cells, and their HBLAK derivatives, were obtained from standard healthy adult human biopsy of the trigome region, and we do not anticipate this being materially different from other biopsies; indeed, we achieved good results with similar cells sourced from a different company. Therefore, the most likely factor is the specialized medium, which is free of serum and other growth factor supplements that could impair differentiation. We are currently using an empirical approach to determine which aspects of the growth media and urine are important for differentiation.
Morphologically, the HBEP and HBLAK cells produced tissue with encouraging similarities to human urothelium 17,18,21,25 . HBEP cells produced tissue ~3 cell layers thick whereas HBEP cell-derived tissue elaborated ~5-7 layers. In contrast to mice, higher mammals such as humans have multiple intermediate cell layers, a feature which may favor the use of HBLAK cells 20 . However, rate of cell division could have played a role in this outcome, as could cell senescence in the HBEP population 66 . Both culture types developed an apical layer of enlarged, flattened cells that appeared umbrella-cell-like 20 in morphology. Based on size and visible distension of cells in the tissue, the differentiated layer seemed more akin to a stretched (urine-filled) state rather than a relaxed (empty bladder) state. Multiple nuclei were not very common in these cells, as would be expected from the mouse urothelium. We could find no consensus in the literature about the expected percentage of multinucleation in the human urothelium, with studies being scant, and it is also not clear whether the multinucleated state plays any functional role. Further work is needed to understand how tissue thickness and multinucleation is regulated in both cell types.
Phenotypic analysis of both the HBEP and HBLAK urothelial organoid tissue elucidated the presence of some important urothelial markers. Uroplakins are a group of highly conserved glycoproteins which are unique to mammalian urothelium 48 . Uroplakin-III (UP3) was expressed on the apical surface throughout the differentiated tissue and its expression was morphologically similar to that of patient-shed urothelial cells -with the caveat that these patient cells were infected, and expression might alter after exfoliation. Laguna et al. (2006) 17 found that the normal human urothelium expresses a range of cytokeratins in relation to the level of cytodifferentiation. Our 3D culture mimicked these findings with cytokeratin-8 (CK8) found throughout the tissue and cytokeratin-20 (CK20) being a preferential phenotype of well-developed umbrella cells 17 . To our knowledge, this is the first human bladder organoid demonstrating the correct spatial expression of CK20 34 .
Patients with overactive bladder symptoms are frequently treated with antimuscarinics which target muscarinic receptors in the detrusor and urothelium 46 . Crucially, rodent bladder urothelial cells do not express muscarinic receptors 47 . In the case of the HBEP and HBLAK tissue grown in this study, muscarinic receptors M2 and M3 were detected throughout the urothelial cell layers, further supporting its physiological relevance and potential use in the development of novel therapeutic agents for the bladder symptoms of MS and other neurogenic disorders.
More detailed analysis revealed several interesting aspects to our HBLAK organoid model. First, EM showed that the tissue was not homogenously differentiated; instead, the surface consisted of three discrete zones: 'valleys' of undifferentiated monolayer; 'plateaus' of fully differentiated 3D tissue; and 'mountains' of hyperplasia. Second, inspection of GAG layer markers revealed that the presence of urine correlates with its elaboration, in parallel with differentiation and organoid formation. Third, focusing on the differentiated plateaus, which comprised about two-thirds of the total area, our EM imaging shows expected key features, including morphologically distinct layers, and the enlarged flat nature of the distended umbrella-like cells. We also noted structures consistent in shape, size and spacing with the characteristic hinges and plaques associated with the AUM, and white lacunae that might possibly represent the fusiform vesicles responsible for trafficking during bladder filling and emptying phases. These structures were entirely absent from the zone of undifferentiated cells, showing a possible correlation with differentiation status. However, higher resolution EM imaging and immunostaining is needed to confirm their identity.
Intriguingly, when we tested the barrier function of the organoid by several methods, including transepithelial resistance and fluorescent dextran permeability (data not shown), we found a lack of what is traditionally thought of as urothelial "barrier function". We presume that this result was caused by the sporadic presence of undifferentiated cells across the tissue, which in essence short-circuit the apicobasal electrical potential difference. This result, taken together with the fact that the non-differentiated zone devoid of AUM-like structures and GAG can grow for several weeks in the presence of urine, strongly suggests that AUM and GAG are not required for urine tolerance, and that something intrinsic in the cells themselves confer resistance to its toxic effects. It also suggests that apical urine is necessary for full differentiation. Whilst it may be the case that urine is not sufficient for differentiation, given the presence of the undifferentiated 'valleys' , it may also be the case that a subset of HBLAK cells are subtly different. Further exploration into the nature of urine tolerance and the effect of urine and other media components on differentiation and hyperplasia are warranted. Genomic analysis of cells from various regions, compared with starting cells, would also be interesting. In the meantime, researchers performing image-based studies should be able to focus on areas of the organoid corresponding to their desired level of differentiation.
Our model shows great promise for studying the host/pathogen interactions of UTI in a human-cell system. The majority of host/pathogen interaction studies in UTI focus on the most common uropathogen E. coli, but very little is known about other uropathogens such as Enterococcus faecalis, which is more common in certain cohorts, such as the elderly, the hospitalized and those using urinary catheters. We previously reported the discovery of intracellular colonies of E. faecalis harbored within the urothelial cells of chronic UTI patients 55 . This previous report also slowed sloughing of umbrella cells from the epithelial lining into the urine, which is known in both mice and humans to be a common response to infection [67][68][69][70][71] . These findings were echoed in our urothelial organoid, where E. faecalis formed significant intracellular colonies within the intermediate and basal cells of the urothelial mimetic after its umbrella-cell layer had been jettisoned. These results further support the notion that E. faecalis exhibits an intracellular phenotype.
In conclusion, current advances in 3D tissue culture enabled us to grow physiologically relevant, organotypic human models of the bladder. Human bladder biomimetics could be used as a reproducible test bed for chronic infective disease formation, treatment, and resolution in humans.

Materials and Methods
Human Primary progenitor cell expansion and handling in 2D. Commercially available human bladder epithelial progenitor cells (HBEP, Cell N Tec) 72 and their spontaneously immortalised, non-transformed counterparts (HBLAK, Cell N Tec) were supplied in frozen aliquots containing ~5 × 10 5 cells at passage 2 and ~0.5 × 10 5 at passage 25 respectively. Cells were isolated from bladder trigone biopsies from male patients undergoing surgery for benign prostatic hyperplasia. HBEP cells are guaranteed to grow for a further 15 population doublings before senescing whereas HBLAK cells, although spontaneously immortalised, should not be differentiated into 3D cultures once they have exceeded a passage number of 40-50. Both cell types were cultured identically, with the exception of slight differences in incubation time between passages, due to the slightly increased rate of cell division exhibited by HBLAK cells.
Thawed cells were seeded (~300 cell clumps/cm 2 ) into pre-warmed and equilibrated low-calcium, high-bovine pituitary extract, primary epithelial medium (CnT-Prime, Cell N Tec) in 9 cm polystyrene dishes and incubated at 37 °C in a humidified incubator under 5% CO 2 . Culture medium was replaced after overnight incubation to remove residual dimethyl sulfoxide (DMSO). Antibiotics were not added to culture medium at any point due to adverse effects on cytodifferentiation, metabolism and morphology 73 . Furthermore, trypsin is known to damage primary cells, so Accutase solution (Innovative Cell Technologies) was used to detach cells at all stages of experimentation 74 . Cells were allowed to expand to ~70% confluency before freezing batches of cells at a density of ~1 × 10 6 cells/ml in defined freezing medium (CnT-CRYO-50, Cell N Tec) in preparation for later experiments. Cells were not allowed to become fully confluent during cell expansion in an effort to maintain a proliferative phenotype.

Differentiation of 3D human urothelium in vitro.
In preparation for organotypic culture, previously frozen progenitor cells were thawed and expanded on 9 cm culture dishes as above. Once 70-80% confluent, the cells were washed briefly with calcium-and magnesium-free phosphate buffered saline (PBS, Sigma-Aldrich) and incubated at 37 °C in ~3 ml of pre-warmed Accutase solution for 2-5 min. The dishes were lightly tapped and detached cells re-suspended in 7 ml of warm CnT-Prime. After centrifugation at 200 × g for 5 min, the supernatant was removed and the pellet re-suspended in fresh CnT-Prime. This cell suspension was counted whilst allowing the cells to equilibrate for 3 min at room temperature. 2 × 10 5 cells in 400 µl of CnT-Prime (internal medium) were added to 6 12 mm 0.4 µm pore polycarbonate filter (PCF) inserts (Millipore) standing in 6 cm culture dishes containing ~3 ml of fresh pre-warmed CnT-Prime medium (external medium, level with insert filters). A further 8 ml of CnT-Prime medium was added to the 6 cm dish (external to the filter inserts) until internal and external fluid levels were the same. The 3D culture inserts were incubated for 3-5 days until 100% confluent. Confluency was determined through the fluorescent staining of 1 insert and visualisation under epi-fluorescence microscopy (see section below). Once deemed confluent, internal and external medium was removed and replaced with low-BPE, calcium-rich (1.2 mM) differentiation barrier medium (CnT-Prime-3D, Cell N Tec) to promote differentiation. Subsequent to overnight incubation, the internal medium (apical surface of cell culture) was removed and replaced with filter-sterilised human urine pooled from healthy volunteers of both genders to aid terminal differentiation into umbrella cells. The external CnT-Prime-3D medium and the internal human urine were replaced every 3 days and the culture incubated for 14-24 days at 37 °C in 5% CO 2 . To explore the effect of urine on differentiation in 2D, the HBLAK cells were seeded on 8-well permanox Lab-Tek slides (Sigma-Aldrich) and grown to confluency. The cells were then exposed to CnT-Prime-3D medium alone, 25%, or 50% sterile human urine diluted in CnT-Prime-3D for 72 hours at 37 °C in 5% CO 2 . To analyse the effect of urine on HBLAK organoid formation, cells were grown to confluency on filter inserts as above. The basal compartment was treated with CnT-Prime-3D throughout, however, the apical compartment was filled with either CnT-Prime-3D alone, 50% sterile human urine diluted in CnT-Prime-3D or 100% urine. The specified medium or urine was changed every 3 days and the culture incubated for 14 days at 37 °C in 5% CO 2 .
Characterisation of the 3D urothelium. Prior to fluorescent staining and immunofluorescence (IF), filter inserts were carefully transferred to 8-well plates (Nunc) and submerged in 4% methanol-free formaldehyde (Thermo Scientific, Fisher Scientific) in PBS overnight at 4 °C. After fixation, the filter inserts were kept at 4 °C in 1% formaldehyde in sealed containers in preparation for processing.
In preparation for imaging, filters were carefully removed from inserts using a scalpel, mounted with FluorSave reagent (Calbiochem), and a coverslip fixed in place with clear nail varnish. Lab-Tek slide wells and gaskets were carefully removed prior to the addition of FluorSave and a coverslip as above.

Electron microscopy. Electron microscopy was conducted by the Division of Medicine, University College
London electron microscopy unit at the Royal Free Campus, Hampstead, London.
For transmission electron microscopy (TEM), samples were fixed in Karnovsky's fixative (2.5% glutaraldehyde/2% paraformaldehyde) and then washed in 1 M PBS 3 × 10 min, followed by a post-fixation in 1% Osmium tetroxide for 1 hour at room temperature. Tissue was rinsed with distilled water 3 × 10 min. Samples were dehydrated in an ethanol series (30,50,70,90 and 100%) then treated with resin-ethanol (1:1) overnight. Subsequently, samples were embedded in 100% LEMIX resin and incubated at 65 °C for 24 hours. Ultra-thin sections of the resin block were cut and post stained with 2% Uranyl acetate and Lead citrate.
For scanning electron microscopy (SEM), the samples were fixed and dehydrated as above. Samples were then incubated in Tetramethylsilane for 10 min and air dried before mounting on stubs and sputter-coated with gold.
Experimental infection of the human urothelial organoid and the gentamicin protection assay. A single strain of Enterococcus faecalis (E. faecalis) originally derived from a patient with chronic UTI 55 was grown aerobically in a shaking incubator at 37 °C for 24 hours. Once a batch of 6 HBLAK 3D urothelial cultures had reached 14 days of growth, 1.6 × 10 7 colony-forming units of each bacteria (MOI of 25) were added to the filter-sterile human urine at the apical liquid-liquid interface of each culture. The experimentally infected cultures were incubated for 2 hours at 37 °C under 5% CO 2 . The 3D culture filter inserts were washed with PBS before fixation and staining as above. For the gentamicin protection assay, after the infection was completed, the cultures were washed 3 times in PBS before the addition of 3D barrier medium containing 2000 µg/ml of gentamicin (Sigma-Aldrich) to the apical and basal compartments of the culture filter. The organoids with antibiotics were incubated for a further 2 hours at 37 °C under 5% CO 2 to kill any extracellular bacteria. Post incubation the supernatant was serially diluted in PBS (undiluted, 1:100, 1:1000, 1:10000) and 25 µl of each dilution spread on a quartile of a Columbia blood agar plate (CBA, Oxoid) before being incubated for 24 hours aerobically at 37 °C to enumerate live bacteria. The organoids were washed a further 3 times in PBS before being lysed with 1% Triton-X100 in PBS for 10 minutes at RT. The lysate was added to CBA plates and incubated as above to detect intracellular bacteria. All experiments were completed in triplicate. Analysis of shed epithelial cells. Patient samples were collected with informed consent and analyzed in accordance with a protocol approved by the East Central London Regional Ethics Committee (REC1) (Ref: 11/ H0721/7).
Experimentally infected organoid supernatants, or patient urine specimens, were collected, cytocentrifuged on to glass slides and fixed and stained prior to imaging 55 . Briefly, 80 µl of supernatant or patient urine was cytocentrifuged using a Shandon Cytospin 2 cytocentrifuge at 800 rpm (≈75 g rcf) for 5 minutes. The cellular deposit was circumscribed with a ImmEdge pen (Vector Laboratories) before fixing, staining and mounting as above.
Imaging and Analysis. We performed epi-fluorescence microscopy on an Olympus CX-41 upright microscope, and confocal laser scanning microscopy on Leica SP5 and SP2 microscopes. Images were processed and analysed using Infinity Capture and Analyze V6.2.0, ImageJ 1.50 h 75 and the Leica Application Suite, Advanced Fluorescence 3.1.0 build 8587 Software.
TEM was conducted using a Jeol 1200-Ex digital image capture system with a side mount 2 Kv AMT camera. SEM was performed using a Jeol JSM-5300 fitted with a Semafore digital image capture system.
The proportion of the HBLAK organoids exhibiting evidence of differentiation was calculated by analyzing low-power SEM micrographs (N = 3) using ImageJ automatic thresholding and measure tools 75 . The number of bacteria per cell was calculated using nearest neighbour 3D connectivity analysis with the ImageJ Object coun-ter3D plugin 75,76 . The DAPI channel of 3D laser scanning confocal constructs were analysed using Differential Voxel Filters allowing the enumeration of mammalian nuclei and bacteria.
Statistical analysis. Data were analysed using IBM SPSS Statistics version 24. Non-parametric Friedman tests of differences among repeated measures were performed due to non-normal distributions. 3 experimental replicates were performed for statistical testing.
Data availability. All data generated or analysed during this study are included in this published article.