All models were grown with HBLAK human bladder progenitor cells (CELLnTEC, Switzerland) according to the online CELLnTEC protocol (https://cellntec.com/products/resources/protocols/culture/). HBLAK are spontaneously immortalised, non-transformed cells, originally derived from a male patient having benign prostatic hyperplasia surgery; if used within the correct passage window (see below), the cells behave like progenitor cells and can differentiate into a stratified urothelium. Briefly, the frozen commercial vial, which is cryo- preserved at approximately passage 25, was thawed and transferred to a cell culture vessel containing pre-equilibrated CnT-Prime medium (CnT-PR, CELLnTEC), incubated at 37°C and 5% CO₂. HBLAK cells were passaged when they reached 70 to 90% confluency, detached using only Accutase and seeded at the recommended density (4000 cells cm²). During the course of cell culture, no serum, antibiotics and/or antifungals were used, and care was taken that cells never exceeded 90% confluency, as this can impair subsequent differentiation. Cells were propagated to passage 4 (otherwise known as passage 29, if including passages prior to receipt from the company), then frozen vials were prepared following the CELLnTEC freezing protocol.

Human 3D bladder cultures were prepared as described previously (Hubbard et al., 2019), with some modifications. Briefly, HBLAK cells were used only within passage 8-12 after thawing the above-mentioned passage 4 stock and subsequent propagation (equivalent to passages 33-37 if including passages prior to vial receipt). Cells were seeded onto polycarbonate Transwell inserts (in a 12-well plate) with a pore size of 0.4 µm (VWR, United Kingdom) at the recommended seeding density (2 x 10⁵ cells per insert) (CELLnTEC, 3D Culture Protocol). 1.5 ml of CnT-PR medium was added to the basolateral chamber of the inserts and 0.5 ml to the apical chamber so that the medium levels were equal. Because the membranes are translucent and confluency cannot be confirmed by microscopy, we checked confluency by sacrificing a parallel well and staining with 4′,6-diamidino-2-phenylindole (DAPI) or equivalent to allow visualisation of cell confluence. Rare cultures that did not reach 100% confluence after 48 h were discarded, as our experience is that these will not terminally differentiate. Once cells were confluent, CnT-PR was replaced with 3D Barrier Medium (CnT-PR-3D, CELLnTEC) and incubated for 15-16 h so cells could form intercellular adhesion structures. 3D cultures were initiated by completely replacing apical CnT-PR-3D with filter-sterilized human urine (pooled gender) (BioIVT, UK) and fresh CnT-PR-3D medium in the basolateral chamber (tapering in the urine is not required), with the same volumes as before. Urine and CnT-PR-3D medium were changed at regular intervals (three times per week, or twice if ThinCert cell culture plates were used) and 3D cultures were fully stratified and differentiated on days 18-20. The cultures remained urine-tolerant throughout this period as assessed by their low shedding, healthy morphology and ultrastructure in subsequent experiments (see below results).

To profile 3D-UHU cluster of differentiation (CD) cell surface markers, terminally differentiated cultures were dissociated using Accutase solution (Merck, UK). Cells were washed with PBS and filtered (100 µm) to prepare single cell suspensions. BD Horizon Fixable Viability Stain 780 solution (1:1000) was added to the cell suspension and incubated at room temperature (RT) for 15 min. Cells were washed and blocked with blocking buffer plus Fc receptor blocking antibody (Clone Fc1.3216, BD Biosciences, UK). Cells were incubated at RT for 10 min then centrifuged (800 g, 3 min). BD Horizon Brilliant Stain Buffer was added to each sample followed by the following antibodies at the optimised volume per test ( Supplementary Table 1 ): CD9 (BV421, Clone M-L13), CD44 (Alexa Fluor 700, Clone G44-26), CD47 (BV786, Clone B6H12), CD49b (Alexa Fluor 647, Clone AK-7), CD59 (BUV 395, Clone p282), CD63 (PE-Cy7, Clone H5C6), CD95 (BUV737, Clone DX2), CD104 (BV480, Clone 439-9B), CD227 (BV650, Clone HMPV), CD271 (BV711, Clone L128). Cells were incubated at 4°C in the dark for 30 min then washed and resuspended in wash buffer with 1% formaldehyde solution. Data acquisition was performed using Cytek Aurora equipped with 5 lasers. The flow cytometry results were analysed using FlowJo™ v10.8 Software (BD Life Sciences).

Prior to immunofluorescence staining, 3D-UHU cultures were fixed with 4% methanol-free formaldehyde (Thermo Fisher Scientific, UK) overnight at 4°C. Membranes were excised with a scalpel and placed in a 24-well plate and permeabilised in 0.2% Triton-X100 (Sigma-Aldrich, UK) in PBS for 20 min at RT. Membranes were washed with PBS then blocked with 5% normal goat serum (Thermo Fisher Scientific) at RT for 1 h. 3D cultures were incubated at 4°C overnight with primary antibodies in 1% bovine serum albumin (BSA)/PBS as follows: rabbit anti-uroplakin-1A (UPK1A) polyclonal antibody (1:100 dilution, PA5-49668, Thermo Fisher Scientific), rabbit anti-uroplakin-II (UPKII) polyclonal antibody (1:100 dilution, NBP2-38904, Novus biologicals); mouse anti-uroplakin-III (UPKIII) monoclonal antibody (1:50 dilution, sc-166808, Santa Cruz); mouse anti-cytokeratin 8 (CK8) monoclonal antibody (1:50 dilution, MA1-06317, Thermo Fisher Scientific); rabbit anti-cytokeratin 13 (CK13) polyclonal antibody (1 µg/ml, PA5-83165, Thermo Fisher Scientific); rabbit anti-cytokeratin 20 (CK20) polyclonal antibody (1:100 dilution, PA5-22125, Thermo Fisher Scientific); rabbit anti-E-cadherin polyclonal antibody (1:200 dilution, PA5-32178, Thermo Fisher Scientific); rabbit anti- zonula occludens-1 (ZO-1) polyclonal antibody (2.5 µg/ml, 40-2200, Thermo Fisher Scientific); mouse anti-claudin 1 monoclonal antibody (2 µg/ml, 37-4900, Thermo Fisher Scientific); rabbit anti-claudin 3 polyclonal antibody (1:100 dilution, PA5-16867, Thermo Fisher Scientific); mouse anti-chondroitin sulfate antibody (1:100 dilution, ab11570, Abcam); rabbit anti-TLR-2 polyclonal antibody (10 µg/ml, PA5-20020, Thermo Fisher Scientific), rabbit anti-TLR-4 monoclonal antibody (1:500 dilution, 76B357.1, Thermo Fisher Scientific), and rabbit anti-TLR-5 polyclonal antibody (10 µg/ml, 36-3900, Thermo Fisher Scientific).

Post incubation, membranes were washed with 1% BSA/PBS and incubated with 1:500 dilution of secondary antibodies (goat anti-mouse or goat anti-rabbit) conjugated with Alexa Fluor-488 (Thermo Fisher Scientific) at RT for 1.5 h. Membranes were washed then stained with Alexa Fluor-633 phalloidin (1:500 dilution, Thermo Fisher Scientific) at RT for 30 min to label filamentous actin followed by DAPI nucleic acid stain (DAPI dihydrochloride, 300 nM in PBS, Invitrogen, UK) at RT for 5 min. Filter membranes were washed then mounted with ProLong Glass antifade mountant (Invitrogen) and imaged on a Leica SP8 microscope. Images were acquired using 20x, 40x or 63x magnification with 0.3 µm Z step size giving 50% image overlap required for 3D image reconstruction. Images were processed using LAS X software version 3.5.7.

To analyse urothelial cell shedding, apical supernatants from a 12-well plate was collected and centrifuged at 300 g for 5 min. Pellets were re-suspended in 100 μl of PBS and cytocentrifuged onto a glass slide using a Shandon Cytospin 2 at 800 rpm for 5 min. Cells were fixed with 4% methanol-free formaldehyde for 15 min at RT then washed 3x in Hanks’ Balanced Salt solution (HBSS) (Thermo Fisher Scientific). Cells were stained with 5 µg/ml of Wheat Germ Agglutinin (WGA), Alexa Fluor-488 conjugate (Thermo Fisher Scientific) for 1 h at RT. Cells were washed with HBSS and stained with 2 µg/ml Hoechst (33342, Thermo Fisher Scientific) for 15 min at RT, and mounted as described above.

To aid in closer inspection of any aspect of the confocal reconstructions, all raw tiff data have been deposited into the open access University College London Data Repository (doi: 10.5522/04/22193692).

The TEER of the 3D-UHU model was measured using the EVOM3 with a STX2-Plus electrode (World Precision Instruments, UK) according to the manufacturer’s instructions. Briefly, a 1000 Ω test resistor was used to set up the equipment then blank handling mode was selected to allow subtraction of the blank control from the current resistance measurement of 3D cultures. Each measurement was recorded and stored on the device for further analysis.

Barrier integrity of 3D cultures was assessed using Fluorescein Isothiocyanate (FITC)-Dextran (MW 4,000, Sigma-Aldrich, UK). FITC-dextran solution (1 mg/ml) was prepared in urine and added to the apical chamber. Aliquots of medium from the basolateral side were collected in 0, 5, and 24 h post-infection (p.i.) (multiplicity of infection; MOI 10) and fluorescence was measured by a Tecan Spark microplate reader (Tecan, Switzerland) at an excitation of 485 nm and emission of 538 nm. The FITC-conjugated dextran concentration was presented as relative fluorescence unit (RFU).

To examine the barrier disruption and innate immune response to bacterial co-culture, several uropathogenic strains was employed. Uropathogenic Escherichia coli (UPEC) strains UTI89 (Mulvey et al., 2001), CFT073 (Mobley et al., 1990) (ATCC 700928), and E. coli 83972 (bei Resources) associated with asymptomatic bacteriuria (ABU) were grown in Luria-Bertani (LB, Sigma Aldrich) broth statically at 37°C for 48 h to induce expression of type 1 pili. Previously published clinical isolates were also tested: Enterococcus faecalis from a patient with UTI (E. faecalis 36 [EF36] (Lau et al., 2020); E. faecalis from an asymptomatic healthy male E. faecalis 77 [EF77] (Sathiananthamoorthy et al., 2019) and Streptococcus agalactiae from an asymptomatic healthy female (Sathiananthamoorthy et al., 2019) (Group B Streptococcus), were grown overnight in Tryptone Soya Broth (TSB, Thermo Fisher Scientific) at 37°C.

3D-UHU cultures were co-cultured with bacterial isolates at MOI 10. Initial seeding density was normally used to calculate MOI to aid in reproducibility; it should be noted, however, that the cells proliferate after this time, but equally the apical interface contains fewer cells (approximately 25,000-30,000 cells, as roughly calculated by surface area and cell size observed in microscopy images). To measure the expression of pro-inflammatory cytokines/chemokines, the apical supernatants were collected at 24 h post infection. Human Interleukin-8 (IL-8), IL-6, IL-1β, and TNF-α was measured by enzyme-linked immunosorbent assay (ELISA) per manufacture’s instruction (Biolegend, ELISA MAX Deluxe Set) and CXCL-1/GRO-α was measured using a R&D systems DueSet ELISA kit.

Data were analysed using GraphPad Prism 9. At least three independent biological replicates in technical replicates of two were performed for statistical analysis. Differences in TEER, FITC-dextran measurements and cytokine/chemokine expression between experimental samples and controls were analysed using 2way-ANOVA. Tukey’s multiple comparisons test was used to compare 3D-UHU TEER measurements in different passage numbers. Dunnett’s multiple comparisons test was used to compare each variable with the untreated control.

To assess 3D-UHU morphologically, we grew the 3D cultures for 18-20 days and then fixed and stained them for confocal microscopy. As shown in Figure 1A , we obtained stratified human bladder 3D models achieving up 7 cell layers thickness ( Figure 1A ). The single optical slices showed densely packed, very small basal cells in the basal layer ( Figure 1B ) followed by several layers (4-6) of slightly larger intermediate cells ( Figure 1C ), and large hexagonal umbrella-like cells forming the top layer ( Figure 1D ) ( Supplementary Movies 1, 2 ). The movies revealed that there appeared to be a second layer of large cells directly below the apical layer that might also be umbrella-like cells, or intermediate cells on a continuum with umbrella cells, although uroplakin staining suggested only the most apical cells were terminally differentiated (see below). Supplementary Figure 1 shows an example of how layer numbers were estimated. We imaged 3-6 randomly selected low-power (x20) microscopy fields derived from models created from four independent experiments, fixed and stained over more than a year’s worth of experiments, to assess whether the differentiated umbrella-like cell layer was, as our observations suggested, routinely homogeneous ( Supplementary Figure 2 ). This assessment suggested that the umbrella-like cell layer was indeed homogeneous, with nearly 100% differentiation without the extensive zones of hypertrophy and undifferentiated monolayers seen in our previous version of this model (Horsley et al., 2018).

We examined epithelial cell shedding by collecting spent apical supernatants at regular intervals which were centrifuged onto glass slides prior to fixation and staining. A very low number of shed cells (~ 25 cells/well based on nuclei count, rest presumed cell debris) were detected on day 18 onwards ( Figure 1E ), suggesting that the model retains most of its apical cell layer (with shed cells representing approximately 0.1% of the predicted apical cell number).

In summary, the 3D-UHU model showed three distinct cell layers with discernible cell size differences and appropriate thickness, reminiscent of a human urothelium and remained largely intact throughout the culture process.

We identified stratified and terminally differentiated 3D-UHU cell types based on their marker expression using flow cytometric analysis. Dissociated single cells were first gated by size and granularity to exclude cellular debris. From the “gated cells”, singlets were sub-gated by their FSC-A/FSC-H properties. Single cells were plotted with the viability dye and cells considered “live” were selected ( Figures 2A–C ). Our choice of markers was informed by a paper from Liu and colleagues (Liu et al., 2012), who used a combination of transcriptomics and immunohistochemistry on human bladder tissue to comprehensively catalogue the cell surface markers expressed by different cell types in the human urothelium. From the “live” population, cells that were positive for CD9 and CD59 markers were selected ( Figure 2D ). CD9 and CD59 have been reported to be expressed by all three cell types in humans: basal, intermediate, and umbrella cells (Liu et al., 2012) ( Supplementary Table 2 ). CD9, CD59 gated cells were plotted with CD44, CD104, and CD271 ( Figures 2E, F ). CD44 and CD104 have been described to be expressed strongly by basal cells in humans although intermediate cells similarly express these markers in low levels, whereas CD271 is expressed exclusively by basal cells (Liu et al., 2012) ( Supplementary Table 2 ). The basal cells in 3D-UHU model were identified by expression of CD271 ( Figure 2F ). Intermediate cells have been reported to express a number of CD molecules in humans (e.g., CD9, CD46, CD49b, CD55, CD59, CD95, CD116, CD147) (Liu et al., 2012); however, these markers are also expressed by basal and umbrella cells to varying degrees of intensity ( Supplementary Tables 2A, B ). CD49b and CD95 markers (Liu et al., 2012) were used to label the intermediate cells ( Figure 2G ). To detect the umbrella-like cells, CD9/CD59 gated cells were plotted with CD47, CD63, and CD227 ( Figures 2H, I ). CD47 and CD63 markers have been shown to be expressed weakly by intermediate and basal cells in humans ( Supplementary Table 2B ). We identified the umbrella-like cells by CD227, the marker that is reported to be expressed exclusively by these cells in humans (Liu et al., 2012) ( Figure 2I ).

We used unstained, fluorescence minus one (FMO), and CD57 marker expressed by urothelial nerve sheath (data not shown) controls to set gate limits and rule out non-specific binding of the panel antibodies to the cellular surface. The number of each cell type was calculated from acquired flow cytometry data anlaysed by FlowJo ( Figure 2J ). Approximately 42,000 3D-UHU cells were acquired in total of which ~16,000 were CD271⁺ basal cells. Intermediate cells (CD95⁺/CD49b⁺) comprised the largest proportion of cell numbers (~20,000), which is consistent with there being multiple layers of this cell type, while a smaller number of cells (~5,000) comprised the CD227⁺ umbrella-like population, which is consistent with them forming one layer of very large cells. The median fluorescence intensity (MFI) of CD markers expressed by human bladder 3D model showed a high CD104 and a relatively low CD227 expression ( Supplementary Figure 3 ).

We characterised the model further by examining the expression of urothelial-specific markers with confocal microscopy. The 3D-UHU model showed the expression of all three uroplakin proteins integral to formation of the AUM; uroplakin-1A (UPK1A) ( Figure 3A ), uroplakin-II (UPKII) ( Figure 3B ), and uroplakin-III (UPKIII) ( Figure 3C ) were expressed in a subset of umbrella-like cells with UPKIII exhibiting the highest surface expression followed by UPK-1A. Although studies have indicated that cultured urothelial cells do not express UPII (Hu et al., 2005), our model showed a modest expression of this marker. Next, we examined the GAG layer expression. The GAG layer consists of glycoproteins and glycolipids and is mainly composed of chondroitin/dermatan sulfate, heparan sulfate, keratan sulfate, and hyaluronic acid (Klingler, 2016). Although not all of these components were tested due to a lack of commercially available reagents, a chondroitin-rich GAG layer was detected at the umbrella-like cell layer ( Figure 3D ).

The 3D-UHU model also exhibited the correct spatial expression of cytokeratin (CK)8, 13, and 20 ( Figure 4 ). CK8 was expressed throughout the basal, intermediate, and umbrella-like cell layers, though detected most strongly at the apical surface ( Figure 4A ) ( Supplementary Figure 4 ). CK13, a differentiation marker indicating a switch from basal cells to intermediate cells which also has been reported to indicate cytodifferentiation of late intermediate cell layer (Varley et al., 2004; Varley and Southgate, 2008; Fishwick et al., 2017), appeared to be detected on the late differentiated intermediate cell layers ( Figure 4B ) ( Supplementary Figure 5 ), not on the surface, while CK20, a terminal differentiation marker, was detected only at the apical surface layer, presented by umbrella-like cells ( Figure 4C ).

The expression of adherens junction (AJ) E-cadherin and tight junction (TJ) proteins – ZO-1 and claudin 1, 3 – were examined by confocal microscopy ( Figures 5A–D ). In the literature, the umbrella cell layer is the only urothelial layer reported to form detectable tight and adherens junctions (Dalghi et al., 2020). The 3D-UHU model showed both membranous and cytoplasmic E-cadherin expression in the umbrella-like cells ( Figure 5A ). A strong ZO-1 expression was detected in cytoplasm of the umbrella-like cell layer with a faint staining at cell borders ( Figure 5B ), while claudin 1 and claudin 3 exhibited a discontinuous membrane and cytoplasmic staining and the expression was distributed diffusely throughout the strata ( Figures 5C, D ).

We determined the barrier integrity by measuring the transepithelial electrical resistance (TEER) in a time-dependent manner ( Figure 5E ). TEER was assessed on day 2 (48h post-seeding), day 3 (CnT-PR-3D addition to initiate cell differentiation), and days 4-18 (cell exposure to urine). Although the TEER time-course showed some passage differences around day 2 and 7 (p < 0.05), a stable increase in TEER was detected from day 12 to 18 reaching to ~1000 Ω.cm² showing no significant inter-passage differences by the end.

To examine the expression of TLRs, 3D-UHU models were stained for TLR-2, -4, and -5 ( Figure 6 ). A moderate to strong expression of TLR-2 and TLR-4 was detected in 3D-UHU models ( Figures 6A, B ), while TLR-5 exhibited a low expression ( Figure 6C ).

Next, we wanted to evaluate the effect of bacterial infection on barrier integrity by assessing the paracellular permeability ( Figure 7 ). A FITC-dextran (4 kDa) marker was used to measure tight junction permeability in response to uropathogenic E. coli CFT073 and UTI89, E. faecalis 36 and E. faecalis 77 clinical UTI isolates, E. coli 83972 (asymptomatic bacteriuria [ABU] isolate), and S. agalactiae isolated from urine of a healthy female (multiplicity of infection; MOI 10). At 5 h post-infection (p.i.), a similar but slight disruption in barrier integrity was detected with all strains compared with uninfected control cells. However, at 24 h p.i., the highly pathogenic CFT073 strain showed a significant (p <0.0001) disruption compared with the uninfected control, as did E. faecalis 36 (p < 0.01), and UTI89 (p < 0.05). E. coli 83972, E. faecalis 77, and S. agalactiae showed a barrier disruption that was not statistically significant compared with uninfected control.

To determine the 3D-UHU innate immune response to uropathogenic E. coli UTI89 and E. faecalis 36 (MOI 10), we measured proinflammatory IL-8, IL-6, CXCL-1, TNF-α, and IL-1β cytokine/chemokine levels in collected supernatants ( Figures 8A–E ). UTI89 caused a significant IL-8 (p < 0.05) ( Figure 8A ), IL-6 (p < 0.001) ( Figure 8B ), CXCL1 (p < 0.001) ( Figure 8C ), TNF-α (p < 0.05) ( Figure 8D ), and IL-1β (p < 0.001) ( Figure 8E ) response. Although E. faecalis 36 showed a significant IL-6 (p < 0.01) ( Figure 8B ) and TNF-α (p < 0.05) ( Figure 8D ) release, it caused an IL-8, CXCL-1, and IL-1β response ( Figures 8A, C, E ; respectively) that was not statistically significant compared with uninfected control.