KRT14 marks a subpopulation of bladder basal cells with pivotal role in regeneration and tumorigenesis

The urothelium is a specialized epithelium that lines the urinary tract. It consists of three different cell types, namely, basal, intermediate and superficial cells arranged in relatively distinct cell layers. Normally, quiescent, it regenerates fast upon injury, but the regeneration process is not fully understood. Although several reports have indicated the existence of progenitors, their identity and exact topology, as well as their role in key processes such as tissue regeneration and carcinogenesis have not been clarified. Here we show that a minor subpopulation of basal cells, characterized by the expression of keratin 14, possesses self-renewal capacity and also gives rise to all cell types of the urothelium during natural and injury-induced regeneration. Moreover, these cells represent cells of origin of urothelial cancer. Our findings support the hypothesis of basally located progenitors with profound roles in urothelial homoeostasis.

T he urothelium is a slowly cycling tissue consisting of basal, intermediate and superficial or umbrella cells that form the urine-blood barrier 1 . Tissue regeneration following microbial or chemical injury relies upon proliferation of progenitor cells 2,3 . Whether the repair process is mediated by a single basal progenitor co-expressing sonic hedgehog (SHH) and keratin 5 (KRT5) 4 , or by distinct basal and intermediate progenitors that regenerate the basal and umbrella layers, respectively 5,6 , without lineage crossing, has become a controversial issue.
In humans, cells expressing KRT14 (keratin 14; KRT14 pos ) are considered the most primitive population in bladder cancer 7,8 , and are enriched upon consecutive rounds of chemotherapy 9 . In a mouse model of invasive bladder cancer, KRT14 pos cells are preferentially amplified upon STAT3 overexpression 10 . Nevertheless, KRT14 pos cells are not yet described in normal human urothelium, while definitive proof that KRT14 pos cells correspond to urothelial progenitors in mice remains elusive. Moreover, potential roles of these cells in tissue homoeostasis and regeneration are yet to be investigated.
Here we provide unequivocal evidence that a small subset of basal cells of embryonic origin characterized by KRT14 expression are the stem cells of the bladder. Using in vivo lineagetracing experiments in mice, and in vitro clonogenic and explant cultures, we show that KRT14 pos cells participate both in natural and injury-induced bladder regeneration by giving rise to all layers. Finally, upon neoplastic transformation, KRT14 pos cells give rise to a spectrum of tumours, implicating them as the cells of origin of bladder cancer. These findings will inspire future studies regarding their role in normal bladder homoeostasis and disease, and their use in regenerative medicine applications.
Within 6 h of chemical injury with cyclophosphamide (CPP) 2 , damage and exfoliation of KRT20 pos cells occurs ( Supplementary  Fig. 1b), to be followed by a marked increase of KRT14 pos cell numbers, peaking at 48 h post CPP injection to 22.3 ± 2.2% and declining soon after tissue repair (Table 1 Table 1). Interestingly, between 18 and 24 h, when the umbrella cell layer is largely absent, the mitotic index of KRT14 pos cells is approximately threefold higher than that of KRT14 neg cells. As proliferation seems to be spreading to non-basal cells by 48 h, this difference drops to a still statistically significant 1.4-fold ( Fig. 1f; Supplementary Fig. 1c; Supplementary Table 2).
Genetic labelling and lineage tracing of KRT14 pos cells. To perform lineage-tracing experiments in vivo, we generated a knock-in CreERT2 recombinase line into the Krt14 locus (Fig. 2a). CreERT2 insertion disrupts the open reading frame of the locus leading to a null allele. Tamoxifen administration in Krt14 CreERT2/ þ ;R26 tdTomato/ þ bitransgenic mice identifies a subset of basal cells that co-express KRT14 and Tomato indicating faithful CreERT2 expression (Fig. 2b). As indicated by the existence of KRT14 pos Tomato neg cells, the R26 tdTomato allele is not recombined in all KRT14 pos cells. The most obvious explanation for this discrepancy is that either tamoxifen local concentration or Krt14 expression levels fail to reach an effective threshold. Vehicle-treated control mice fail to produce Tomatopositive cells ( Supplementary Fig. 2), indicating a tightly regulated Cre recombinase driver.
KRT14 pos cells give rise to all urothelial lineages. Tamoxifen administration followed by a single CPP injection and recovery of Krt14 CreERT2/ þ ;R26 tdTomato/ þ bitransgenic mice shows a significant increase of Tomato pos cells in the basal layer, and for the first time in the umbrella layer (Fig. 2c,d; Table 2). Of note, after a single injection with CPP, the immediate descendants of KRT14 pos cells initially remain basal/intermediate, as indicated by the sharp increase in Tomato pos /KRT5 pos frequency (from 3.89 ± 1.25% to 17.33 ± 3.07%) and their relative absence from the umbrella layer (Fig. 2c,d; Table 2). Given the fact that upon CPP treatment, all umbrella cells need to be replenished, the scarcity of Tomato pos /KRT20 pos cells ( Fig. 2d; Table 2) implies the existence of a non KRT14 cell population that initially mediates umbrella layer regeneration. Upon repeated cycles (5 Â ) of CPP injection and recovery; however, Tomato pos cells become quite abundant in all cell layers (Fig. 2c,d; Table 2), indicating that KRT14 pos cells are primitive cells that outlast and can give rise to all other cell types.
In utero labelling followed by chase through adulthood reveals that postnatal (P5) KRT14 pos cells are derived directly from their embryonic counterparts (Fig. 2e). Moreover, Tomato pos descendants of embryonically labelled KRT14 pos cells repopulate CPP-injured bladders and give rise to all cell types (Fig. 2f). Eight-month long chase experiments in Krt14 CreERT2/ þ ; R26 tdTomato/ þ bitransgenic mice injected with tamoxifen at the age of 8 weeks indicate that KRT14 pos cells participate in the natural regeneration of all urothelial layers (Fig. 2g). Altogether, our data indicate that Krt14 expression marks an embryonic subpopulation of cells that persists through adulthood and participates both in natural cycling, and repair upon injury. The KRT14 pos subpopulation gives rise to all cell types in the mouse urinary bladder.
KRT5 pos basal cells regenerate the umbrella layer. Our observation that KRT14 pos basal cells participate in umbrella layer repair is in agreement with findings, showing that SHH pos cells in the basal layer give rise to umbrella cells upon chemical and uropathogenic bacteria-induced damage 4 . It contradicts, however, a previous report showing that KRT5 pos cells, which encompass the KRT14 pos subpopulation, do not contribute to the umbrella layer regeneration 5 . To clarify this issue, we performed lineage-tracing experiments using a Krt5 CreERT2 transgenic mouse line and found that KRT5 pos cells contribute to umbrella layer regeneration following a single challenge with CPP ( Fig. 3a,b; Supplementary Table 3). Given the fact that both studies have used the same Krt5 CreERT2 driver 13 , we hypothesize that the R26 tdTomato reporter used in this study is more prone to recombination than the R26 Tomato/gfp used in the Gandhi et al study. In support of this hypothesis, variable labelling efficiency has been reported with the Krt5 CreERT2 driver (60 and 39% in two different studies 5,14 with Krt5 CreERT2 ;R26 Tomato/gfp mice, and an even lower 29% with Krt5 CreERT2 ;R26 LacZ mice 14 ), while we observe a 63.5% labelling ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11914 efficiency (all percentages are calculated as the fraction of reporter-positive cells that immunostain with anti-KRT5 antibodies). The existence of a small subpopulation of KRT5 pos cells, which, however, express lower levels of Krt5 (KRT5 low ) 15 and therefore label more poorly, is a plausible explanation for this difference. In support of this hypothesis, immunofluorescence experiments identify a small population of KRT5 low/neg / KRT20 neg Tomato-labelled intermediate cells that could correspond to the 'I' cells that were previously described as umbrella layer progenitors 5 (Fig. 2c).
KRT14 pos cells support ex vivo growth of bladder explants. To assess the proliferative potential of KRT14 pos cells and their contribution to tissue growth, we employed an ex vivo assay using bladder tissue explants 16 . When seeded onto polyester filters, these explants produce outgrowths spreading and covering the filter surface within days. We dissected and grew explants from Krt14 CreERT2/ þ ;R26 tdTomato/ þ mice injected with tamoxifen before sacrifice. We observed a massive expansion of the Tomato-labelled population, both on the explanted tissue and the newly formed outgrowth (Fig. 4a). Ki67 and KRT14 immunostaining showed that outgrowth cells are actively proliferating and are nearly all KRT14 pos , as well as KRT5 pos ( Fig. 4b; Supplementary Fig. 3a). Extensive Tomato labelling of KRT14 pos outgrowths (Fig. 4a,b) indicates that this population represents the lineage of in vivo-labelled KRT14 pos basal cells, rather than a population with newly acquired Krt14 expression. This is supported by the fact that bladder explants from Krt14 CreERT2/ þ ;R26 tdTomato/ þ mice, which were treated with vehicle, present extremely rare Tomato-positive cells corresponding to 'leaky' recombination, while explants from the same bladder treated with 4-hydroxytamoxifen (4OHT) in vitro produce explants with extensive Tomato fluorescence (Supplementary Fig. 3b).
Conditional ablation of KRT14 pos cells in tissues explanted from Krt14 CreERT2/ þ ;R26 DTR/ þ mice 17 Table 1 and in Supplementary Tables 1 and  2, respectively. For b, d and e, multiple comparison using Kruskal-Wallis test was also performed and P values were 0.0007, o0.0001 and 0.0001, respectively. Dash lines represent the basement membrane. Scale bars, 50 mm. N/T, not treated. ARTICLE wild-type counterparts (Krt14 CreERT2/ þ or R26 DTR/ þ ) in the absence of DT (Fig. 4c).
KRT14 pos cells originate from themselves during injury. To trace the origin of KRT14 pos cells in vivo during regeneration, Krt14 CreERT2/ þ ;R26 DTR/ þ mice were challenged with CPP upon DT-mediated ablation of KRT14 pos cells. Despite the obvious tissue damage, no proliferation was observed, while the KRT14 population was practically extinct (Fig. 4d). This implies that the KRT14 cell pool is regenerated exclusively from KRT14 pos cells. Unfortunately, premature mouse dying due to fatal complications in other tissues expressing KRT14 prevented us from monitoring mice through a complete round of injury and recovery, and thus to assess the effect of KRT14 pos cell absence in bladder regeneration.
Clonogenic and differentiation capacity of KRT14 pos cells. To measure the clonogenic capacity of KRT14 pos cells in vitro, we  generated bladder single cell suspensions from Krt14 CreERT2/ þ ;R26 tdTomato/ þ mice that were injected with tamoxifen before tissue digestion. Fluorescent-activated cell sorting (FACS) indicated that Tomato pos cells represent 1.2% of the total population (Fig. 5a). When seeded on Matrigel in clonogenic densities, these cells produce perfect spheres within 2 weeks (Fig. 5a,b). Fluorescence microscopy revealed that the sphereforming capacity of KRT14 pos (Tom pos ) cells is significantly higher (Po0.0001) than KRT14 neg (Tom neg ) cells (9.21±0.61% versus 0.56 ± 0.09%; Fig. 5c). After 4 weeks in culture, expression of KRT14 is restricted in the outer layer of spheres, while KRT5 and Tomato are expressed throughout (Fig. 5d). This illustrates that in vitro, KRT5 pos KRT14 neg cells differentiate from KRT14 pos cells, and this stratification is reminiscent of what is observed in vivo. Passaging of both populations and reculturing on Matrigel showed that KRT14 pos cells retain their in vitro proliferative capacity (Fig. 5e). These data indicate that the clonogenic capacity of urothelial cells reside by large within the KRT14 compartment. Combined, our data indicate that KRT14 pos cells give rise to themselves and other cell types both in vivo and in vitro.
Wnt/b-catenin signals support KRT14 pos cell proliferation. Previous reports have implicated the Wnt/b-catenin signalling pathway in regulating basal cell proliferation during repair 4 . In vivo administration of the nonsteroidal anti-inflammatory drug indomethacin, which inhibits the Wnt/b-catenin pathway, before CPP-induced injury led to significant decrease in KRT14 pos cell proliferation and, consequently, numbers (Fig. 6a). Moreover, in vitro assays showed significantly reduced clonogenic capacity (Fig. 6b), while Wnt/b-catenin inhibitors prevented the KRT14 pos cell proliferation and explant tissue growth in a dose-dependent manner (Fig. 6c,d). While indomethacin is not a specific Wnt/b-catenin inhibitor, knockdown of b-catenin with small hairpin RNAs (shRNAs) confirmed these observations (Fig. 6e,f).
KRT14 pos cells are cells of origin of bladder cancer. Previous reports have implicated KRT5 pos and SHH pos cells as cells of origin in bladder cancer 14,15,18,19 . To investigate the role of KRT14 pos cells in tumour initiation, we employed the wellestablished model of chemical carcinogenesis with N-butyl-N-(4-hydroxybutyl)nitrosamine (BBN). A cohort (n ¼ 11) of male and female littermates (Krt14 CreERT2/ þ , Krt14 CreERT2/ þ ; R26 tdTomato/ þ and R26 tdTomato/ þ ) were exposed to BBN for a maximum of 6 months. In comparison to age-matched control mice, a marked increase of KRT14 pos cell numbers was observed in mice exposed to the carcinogen for 4 months (Fig. 7a,b). After 6 months on BBN, animals developed invasive tumours that almost exclusively expressed Krt5 as previously described 15 . Krt5 expression absolutely coincided with Krt14 expression ( Fig. 7c; Supplementary Fig. 4). We consider Krt14 positivity an indication of the cell of origin rather than an acquired property. It is important in this aspect to emphasize that occasional squamous metaplasia marked by Krt10 expression was also observed ( Supplementary Fig. 4). In support of this hypothesis, lineagetracing experiments in Krt14 CreERT2/ þ ;R26 tdTomato/ þ mice (n ¼ 7) that were injected with tamoxifen before BBN exposure showed that initial expansion of Tomato pos cells (Fig. 7d) was followed by the development of neoplasms half of which (12/24) showed Tomato fluorescence (Fig. 7e-i). We observed no significant difference in regard to contribution of KRT14 pos cells to different tumour subtypes (Supplementary Table 4). Because cohort mice were injected with tamoxifen before BBN exposure, Tomato positivity is enough proof that Tomato pos tumours originated from KRT14 pos cells. However, clonality is difficult to establish in this experimental set-up, and therefore, additional experiments will be required to assess the contribution of individual bladder populations in chemical-induced tumorigenesis.

Discussion
It has been postulated for years that basal cells are responsible for the regeneration of all urinary bladder layers, including the umbrella layer 4,20,21   Given the fact that not all KRT14 pos cells are genetically labelled with Tomato, these data rather underestimate the contribution of KRT14 pos cells in bladder regeneration. The apparent contradiction between our findings and published research is not straightforward to explain. A possible explanation could be that different genetic tools for conditional cell labelling show variable degree of recombination and consequently labelling efficiency. Because our results do not eliminate the possibility that stem cells dedicated to umbrella layer regeneration actually exist, we hypothesize that these cells, if existent, can only be short-term urothelial stem cells (USCs) sufficient to regenerate mildly injured bladders, or even undisturbed aging bladders. KRT14 pos cells, on the other hand, represent USCs with long-term repopulating capacity (urothelial stem cells) with clear roles in tissue repair, as well as tumorigenesis. This stem cell pool is mobilized under conditions of repeated and/or chronic injury and fully regenerates the bladder urothelium. While in vivo clonogenic assays through orthotopic transplantation would be required to prove that a single cell can generate a fully functional bladder, all our findings indicate that KRT14 pos cells likely have that capacity.
In agreement with the hypothesis that stem cells could represent cells of origin of cancer, this subpopulation expands in size in response to chemical carcinogens and undergoes neoplastic transformation that leads to the development of invasive cancer. In this respect, we believe that future studies should focus on the validation of KRT14 pos cells as tools in regenerative medicine and targets in cancer intervention.

Methods
Mice. Newly developed Krt14 tm(CreERT2) (Krt14 CreERT2 ) heterozygotes (see below) were crossed to Gt(ROSA)26Sor tm9(CAG-tdTomato)Hze (R26 tdTomato ) and to Gt(ROSA)26Sor tm1(HBRGF)Awai (R26 DTR ) to produce doubly heterozygous Krt14 CreERT2/ þ ;R26 tdTomato/ þ or Krt14 CreERT2/ þ ;R26 DTR/ þ , respectively. Krt14 CreERT2/ þ littermates were used as controls for in vitro cell ablation experiments. Heterozygous Tg(Krt5-Cre/ERT2)2lpc (Krt5 CreERT2 ) mice where crossed to R26 tdTomato/ þ to produce doubly heterozygous Krt5 CreERT2/ þ ;R26 tdTomato/ þ . Male mice between 6 and 9 weeks of age were used for all experiments, except in BBN carcinogenesis experiments where a mixed cohort was used. Wild-type mice were in all cases of C57Bl/6 background. Animals were housed in individually ventilated cages under specific pathogen-free conditions in full compliance with FELASA (Federation of Laboratory Animal Science Associations) recommendations in the Animal House Facility of the Biomedical Research Foundation of the Academy of Athens (BRFAA, Greece). All procedures for the care and treatment of the animals were approved by the Institutional Committee on Ethics of Animal Experiments and the Greek Ministry of Agriculture.
Generation of Krt14-CreERT2 mice. Krt14 homologous arms were PCR-amplified from mouse 129/Sv genomic DNA as template. The 5 0 arm (3.8 kb) including the KRT14 ATG, which was precisely fused into the CreERT2 ATG, and the 3 0 arm (4.4 kb) flanked a CreERT2-loxP-Neo R -loxP cassette in a pBluescriptSK þ backbone. The targeting construct was linearized and electroporated into 129/Sv W4 embryonic stem (ES) cells that were selected with G418 (150 mg ml À 1 ) for 8 days. A total of 288 clones were picked and analysed by Southern blot for homologous recombination. Positive clones were injected into C57BL/6 blastocysts. Male chimeras were crossed to C57BL/6 females and offspring was genotyped to assess germline transmission.
Urothelial injury. Chemical injury of the urothelium was induced by intraperitoneal injection of a CPP (Sigma) solution in phosphate-buffered saline (PBS; 250 mg kg À 1 ). Bladders were collected at the indicated time points after administrating CPP. In the case of multiple rounds of injury, mice were left to recover for 14 days before CPP was re-administered.
Lineage-tracing studies. Eight-week-old Krt14 CreERT2/ þ ;R26 tdTomato/ þ or Krt5 CreERT2/ þ ;R26 tdTomato/ þ were injected intraperitoneally with 3 mg tamoxifen (Sigma) daily, for 5 consecutive days. Labelling of KRT4 pos or KRT5 pos cells without injury was assessed 72 h after the last tamoxifen injection. For lineage tracing of labelled cells post injury, the injurious chemical was administered at least 72 h after the last tamoxifen injection. For embryonic labelling of KRT4 pos cells, 1 mg of tamoxifen was injected once intraperitoneally to pregnant mothers at gestation day 16.5.
Urothelial tissue explant culture. The procedure has been previously described 22 . In brief, bladders were collected, rinsed in PBS and cut sagittally. The two halves were further cut into B3 mm 2 pieces and the urothelium separated from the muscle layer carefully using forceps. Tissue fragments were spread onto 12 mm diameter, 0. Dulbecco's modified essential medium (Sigma), supplemented with 0.1 mM ethanolamine (Sigma), 0.1 mM phosphoethanolamine (Sigma), 0.5 mg ml À 1 hydrocortisone (Sigma), 5 mg ml À 1 insulin (Sigma), 15 mg ml À 1 adenine (Sigma), 100 U ml À 1 penicillin and 100 mg ml À 1 streptomycin, was added to the lower compartment so that the medium was just in contact with the porous membrane, and explants were grown on the air-liquid interface. Media were changed every other day. For ex vivo CreERT2 activation, 4OHT (Sigma) was supplemented to the medium at 0.5 mM and medium was exchanged with 4OHT-free medium after 12 h. For KRT14 pos cell ablation experiments, DT (Sigma) was supplemented to the medium at 50 ng ml À 1 and fresh DT-supplemented medium was changed daily. suspended in 1 mM ethylenediaminetetraacetic acid (EDTA) in PBS at a density of 1,000 cells ml À 1 , appropriate volume of which was mixed with 40-ml ice-cold Matrigel (Corning) and plated onto glass coverslips in 24-well tissue culture plates. After allowing Matrigel to solidify for 20 min at 37°C, a 1:1 mixture of MDCB153/ advanced Dulbecco's modified essential medium (described above) and V79 lung fibroblast conditioned medium 4 was added. Medium was changed every 2 days. Cells were isolated from three biological replicates consisting of two bladders each and plated at least four wells from each replicate. For the Wnt/b-catenin signalling inhibition, indomethacin was supplemented to the medium at 100 mM and medium was changed every other day. Data presented are mean values ± s.e.m.
Quantitative PCR with reverse transcription. Tissues were frozen in liquid nitrogen, pulverized with a mortar and pestle and total RNA was isolated using Nucleospin RNA (Macherey-Nagel). In the case of Matrigel-grown spheres, Matrigel was first digested using dispase II (5 mg ml À 1 ) and spheres collected by centrifugation. Complementary DNA samples were prepared using Superscript II (Invitrogen), and quantitative PCR reactions were performed using KAPA SYBR Pharmacological treatment. Indomethacin (Santa Cruz) was administered by intraperitoneal injection at 2.5 mg kg À 1 every 12 h. Mice were given four doses of indomethacin before CPP administration. Dosing scheme described above was continued after CPP administration, until mice were sacrificed at indicated time points. DT (0.04 mg kg À 1 ) was injected intraperitoneally for two consecutive days and mice killed 24 h later. CPP administration to DT-treated mice was at 24 h after initial DT treatment.
shRNA knockdown of b-catenin. shRNA sequence (CTAACCTCACTTGCA ATAATccatggATTATTGCAAGTGAGGTTAG) was cloned into pLKO.1/IRE-Segfp. Lentiviral supernatants were generated in HEK293T cells using standard procedures and concentrated using Amicon Ultra-15 centrifugal filters (Millipore). Using serially diluted lentiviral samples and primary mouse bladder cells cultured in Matrigel, we titrated our viral preparations based on abundance of green fluorescent protein expression. Equivalent titres of shRNA and scrambled sequence-expressing particles were then used to transduce primary mouse bladder cells. In short, cells were incubated with the viral particles at 37°C for 1 h, then mixed with Matrigel and plated. Lentiviral particles were maintained in the medium for 24 h before being removed and medium replaced. Seventy-two hours after initial transduction, all cells expressed green fluorescent protein. To eliminate any untransduced background, cells were selected using 1.5 mg ml À 1 puromycin for 5 days.
BBN carcinogenesis. BBN was diluted to 0.05% and administered to mice through drinking water. Mice were put on BBN water at least 72 h after the last tamoxifen injection. As separate tumours, we considered lesions within the same bladder if they were separated by normal-looking urothelium, or were uniformly positive or negative for Tomato (where relevant), or were of distinctive histology and/or marker expression pattern. All tumours with at least 50% Tomato pos cells were counted as Tomato pos .
Imaging. All images were captured on Leica TCS SP5 inverted confocal, Leica HC, Leica DM IRE2, Leica DM LS2 or Nikon SM2800. Image processing and cell counts were performed using ImageJ and Adobe Photoshop CS3.
Statistical analysis. Animals were randomly assigned into different groups. Group allocation and outcome assessment was not blinded. All quantitation was performed on at least three independent biological samples, using the ImageJ software. For quantitation of urospheres in Matrigel, a lower cutoff of 0.002 mm 2 of sphere surface was used. Data presented are mean values ± s.e.m. Statistical analysis was performed using the GraphPad Prism software v6. In two group comparisons, statistical significance was determined using a two-tailed Student's t-test, considering a value of Po0.05 as significant. Multiple comparisons were performed using the Kruskal-Wallis statistical test. All sample sizes met the minimum requirements of the respective statistical test used.
Data availability. Data supporting the findings of this study are available within the article and its supplementary information files and from the corresponding author upon reasonable request.