Aldo-keto reductase 1C1 induced by interleukin-1β mediates the invasive potential and drug resistance of metastatic bladder cancer cells

In treating bladder cancer, determining the molecular mechanisms of tumor invasion, metastasis, and drug resistance are urgent to improving long-term patient survival. One of the metabolic enzymes, aldo-keto reductase 1C1 (AKR1C1), plays an essential role in cancer invasion/metastasis and chemoresistance. In orthotopic xenograft models of a human bladder cancer cell line, UM-UC-3, metastatic sublines were established from tumors in the liver, lung, and bone. These cells possessed elevated levels of EMT-associated markers, such as Snail, Slug, or CD44, and exhibited enhanced invasion. By microarray analysis, AKR1C1 was found to be up-regulated in metastatic lesions, which was verified in metastatic human bladder cancer specimens. Decreased invasion caused by AKR1C1 knockdown suggests a novel role of AKR1C1 in cancer invasion, which is probably due to the regulation of Rac1, Src, or Akt. An inflammatory cytokine, interleukin-1β, was found to increase AKR1C1 in bladder cancer cell lines. One particular non-steroidal anti-inflammatory drug, flufenamic acid, antagonized AKR1C1 and decreased the cisplatin-resistance and invasion potential of metastatic sublines. These data uncover the crucial role of AKR1C1 in regulating both metastasis and drug resistance; as a result, AKR1C1 should be a potent molecular target in invasive bladder cancer treatment.


Establishment of primary and metastatic cancer cells by orthotopic human bladder cancer xenograft.
To investigate the molecular mechanisms of bladder cancer metastasis, we established an orthotopic xenograft model using human bladder cancer UM-UC-3 cells that were stably expressing firefly luciferase 2 and tdTomato red fluorescent protein (designated as UM-UC-3-tdTomato-luc2). The cells were inoculated into the bladders of two nude mice through a urethral catheter, and the labeled cancer cells were successfully detected by bioluminescent imaging (BLI) (Fig. 1a). The primary bladder xenograft gradually grew and after 45 days, metastatic tumors could be detected in the lungs, liver, bone, and lymph nodes (Fig. 1a,b), which were confirmed histologically (Fig. 1c). A significant increase in the invasiveness of the UM-UC-3-bladder, UM-UC-3-lung, and UM-UC-3-liver cells was observed compared with the parental cells (control) (Fig. 1d), which is similar to the increased motility according to the chemotaxis assay ( Supplementary Fig. 1a). The growth rates of these sublines were approximately equal, except for the UM-UC-3-bone cell line, which displayed slower growth ( Supplementary Fig. 1b). Kaplan-Meier analysis of the mice with orthotopically implanted UM-UC-3-lung (n = 5) and UM-UC-3-liver (n = 4) cells demonstrated significantly worse survival than those with the control cells (n = 8, P < 0.01; Fig. 1e, Supplementary Fig. 1c), suggesting the metastatic tumors have aggressive features. It should be noted that metastatic bladder cancer models could not be established using the other bladder cancer cell lines J82 and TCC-SUP ( Supplementary Fig. 1d). AKR1C1 and EMT-and stemness-related genes were increased in metastatic cancer cells. As invasive potential and aggressiveness were increased in the metastatic cancer cells, we next investigated the role of EMT and the stemness profile in metastatic cell lines. Immunoblot analysis revealed that the UM-UC-3-bladder cells had higher expression levels of Snail and fibronectin, EMT markers, compared with the control cells (Fig. 2a). An increase in the fibronectin expression level was also observed in all metastatic sublines, whereas the UM-UC-3-bone cells had additional enhancement of Slug and N-cadherin (Fig. 2a). To verify EMT, the vimentin expression levels were increased in lung and liver micrometastatic tumor cells compared to primary bladder cancer according to immunohistochemistry (Fig. 2b). The expression levels of MMP-2 and stemness markers, such as Nanog and CD44, were also increased in the three metastatic cell lines (Fig. 2c,d). To assess their stemness features, decreased mRNA levels of cytokeratin CK18 and increased mRNA levels of p63 were observed by microarray analysis (data not shown).
To identify the molecules responsible for bladder cancer metastasis, gene expression profiles were investigated by microarray analysis using the control UM-UC-3 line, the primary UM-UC-3-bladder line, and three metastatic UM-UC-3 sublines. Increased expression levels of EMT-related molecules, such as Slug, Zeb2, MMP-3, MMP-1, TGFBR1, BMP2, and NRP1; metastasis-related cytokines, such as IL-6, IL-8, Cox2, IL-1β , and CXCR7; and other metastasis-related genes, such as VEGFA, EGF, CD74, and CD24, were observed in the primary and metastatic tumors (Fig. 2e). We focused on the genes that are commonly up-regulated more than two-fold in all three metastatic tumor cells compared with the primary bladder tumor-derived cells (Fig. 3a), and eight genes were identified (Fig. 3b). Among them, increased mRNA expression of aldo-keto reductase family 1, member C1 (AKR1C1) was confirmed in metastatic sublines by qRT-PCR (Fig. 3c), whereas the levels of other genes were not significantly elevated ( Supplementary Fig. 2a,b). The expression levels of the AKR1C1 protein were also elevated in the metastatic tumors, particularly in UM-UC-3-liver and UM-UC-3-bone cells (Fig. 3d). The AKR1C1 expression level in primary bladder tumors was elevated 2.4-fold compared to that in control cells (Fig. 3e).

IL-1β induces AKR1C1 expression and elimination of AKR1C1 decreases UM-UC-3 cell invasion.
Next, the molecular mechanisms underlying the increased AKR1C1 expression levels were analyzed. We focused on IL-1β , IL-6, and TGF-β 1 because these are well-known cancer progression-related factors for many malignancy types, including bladder cancer 14 . The qRT-PCR analyses confirmed that the IL-1 and IL-6 expression levels and those of their corresponding receptors were increased in metastatic tumors (Fig. 4a,b). To assess a potential cells labeled with tdTomato-luc2 were orthotopically implanted into two athymic mice. Tumor growth was monitored using a bioluminescence imaging system on the indicated days. (b) After 45 days, the mice were sacrificed with bioluminescent imaging photons that showed the primary bladder tumor and lymph nodes, lung, liver, and bone metastases. Fluorescent microscopy demonstrated tdTomato-expressing tumor cells isolated in the culture dish. (c) Excised tumors were fixed and stained with H&E. Primary bladder tumors with a magnified insert (top) as well as lung (middle) and liver (bottom) metastatic tumors. N: normal tissue, T: tumor. (d) Matrigel invasion assay. In UM-UC-3, UM-UC-3-tdTomato-luc2 (control), primary bladder tumor cells, and lung, liver, and bone metastatic cells were seeded onto Matrigel-coated transwell chambers. After 24 h of incubation, the invading cells under the filter were counted and depicted as the means ± SD. Representative photographs are shown. *P < 0.05 versus WT. (e) Kaplan-Meier curves were constructed to compare survival between the mice implanted with control cells (control: n = 8) versus metastatic cells (meta; lung: n = 5, liver: n = 4). The overall survival of mice implanted with the metastatic cells was significantly worse (P < 0.05).
role for these factors in AKR1C1 expression, control UM-UC-3 cells were treated with IL-6 (10 ng/mL), IL-1β (20 ng/mL), or TGF-β 1 (5 ng/mL), and qRT-PCR was used to confirm the up-regulation of AKR1C1 mRNA in response to IL-6 and IL-1β stimulation (Fig. 4c). For verification, the cytokine effects on other bladder cancer cell lines, such as TCC-SUP and 5637, were examined, and IL-1β increased the AKR1C1 levels in both cell lines ( Supplementary Fig. 3).
To examine the role for AKR1C1 in bladder cancer metastasis, we eliminated AKR1C1 in the UM-UC-3 cells using a siRNA technique. The knockdown efficiency of AKR1C1 was validated by both the mRNA and protein levels (Fig. 4d). UM-UC-3-liver cells had a spindle shape, suggesting mesenchymal features, whereas the AKR1C1 knockdown cells had a flat, cobblestone-like morphology (Fig. 4e). A Matrigel invasion assay demonstrated that the numbers of invading cells were significantly reduced by AKR1C1 knockdown in the control UM-UC-3, UM-UC-3-bladder, and UM-UC-3-liver cells (Fig. 4f). As the enzymatic activity of AKR1C3 has been shown to be suppressed by non-steroidal anti-inflammatory drugs (NSAIDs) 31 , we tested the effect of one NSAID, flufenamic acid (FFA), and found that the invasion potential was significantly suppressed in the control and metastatic UM-UC-3 cells (Fig. 4f). Similar results were obtained using another siRNA for AKR1C1 (si-AKR1C1 #2, Supplementary Fig. 4a-c). Notably, siRNA-based AKR1C1 depletion remarkably diminished the expression of Rac1 and phosphorylation levels of Akt (S473), Src (Y416), and FAK (Y397) in the metastatic UM-UC-3-liver cells (Fig. 4g) in addition to numbers of focal adhesion like paxillin (Fig. 4h, Supplementary Fig. 4d), which might
interest to evaluate the role of AKR1C isoforms in the chemotherapeutic response of cisplatin in bladder cancer cells. In the UM-UC-3, 5637, and J82 cell lines, long-term treatment with cisplatin significantly increased the expression levels of all AKR1C isoforms ( Fig. 5a and Supplementary Fig. 5a,b), although the basal AKR1C1 expression levels were different among these cells ( Supplementary Fig. 5c). For the orthotopic xenograft model, the metastatic tumor cells, particularly the liver and bone cells, possessed significantly higher expression levels of AKR1C1, AKR1C2 and AKR1C3 (Figs 3c and 5b), corresponding with increased cisplatin resistance (Fig. 5c,d). Furthermore, the pharmacological AKR1C inhibitor, FFA, significantly restored the sensitivity of the metastatic tumor cells (liver and bone) to cisplatin (Fig. 5e), suggesting a requirement of AKR1C enzymatic activity for cisplatin resistance. These results illustrate that AKR1C1 confers chemoresistance to cisplatin, and FFA can suppress AKR1C1-dependent resistance in metastatic cells, resulting in increased sensitivity to chemotherapy.

Enhancement of AKR1C1 expression in metastatic human bladder cancer lesions.
To ensure clinical significance, we examined the AKR1C1 levels using surgically resected specimens derived from bladder cancer patients, including 33 primary and 5 metastatic lesions involving the lymph nodes (n = 3), lungs (n = 1), and liver (n = 1). RT-PCR analysis demonstrated that the metastatic tumors exhibited higher levels of AKR1C1 mRNA compared with the primary sites (P = 0.008; Supplementary Fig. 6). For verification, immunohistochemical analysis was performed using 25 matched-paired samples of primary and metastatic human bladder cancer lesions (Supplementary Table 2). The AKR1C1 expression level was evaluated as the sum of the intensity score (0-3) and the proportion score (0-3) for AKR1C1 immunostaining (Supplementary Fig. 7a,b). In 17 out of 25 cases, higher AKR1C1 expression was observed in the metastatic lesions affecting the lymph nodes, lung, and liver compared with the paired primary tumors (Fig. 6a-c and Supplementary Fig. 7c).

Discussion
To advance human bladder cancer treatment, it is necessary to understand the molecular mechanisms underlying tumor invasion, metastasis, and drug resistance. In this study, we evaluated orthotopic mouse xenograft models in which cancer cells were inoculated through the urethra. The cells spontaneously attached to the bladder mucosa surface and formed solid tumors, resulting in natural tumor growth and metastasis to other organs. We utilized the bladder cancer cell line, UM-UC-3, possessing a typical MIBC gene profile with LOH of chromosome 9, p53 mutation, and ERBB2 overexpression without FGFR3 mutation 32,33 . In our model, circulating cancer cells were observed in the bloodstream of the mice 34 , and we could establish three metastatic sublines from the nodules found in the lungs, liver, and bone.
In these three cell lines, EMT markers, including Snail, Slug, ZEB2, N-cadherin, and vimentin, were elevated and had increased in vitro invasion potential (Figs 1d and 2a,b,e). Injected bladder cancer cells grown in the mucosa also exhibited enhanced invasiveness with their EMT profiles compared to wild type UC-UM-3 cells. Cancer stem cells also play a major role in tumor recurrence and metastasis in various cancers, including bladder cancer 35,36 . In our xenograft model, CD44 and Nanog, as stemness markers, were significantly up-regulated in the lung metastatic subline compared with the control cells (Fig. 2d). As ERBB2 is frequently overexpressed in metastatic lesions 37 , mild enhancement of ERBB2 was observed in all three metastatic cell lines by microarray analysis (data not shown).
The UM-UC-3-lung and UM-UC-3-liver metastatic sublines exhibited aggressive invasion features (Fig. 1d) and in vivo tumor formation (data not shown). Kaplan-Meier analysis of the mice implanted with metastatic sublines (lung: n = 5, liver: n = 4) demonstrated significantly worse survival than those implanted with the control cells (n = 8, P < 0.01; Fig. 1e). Additionally, the UM-UC-3-bone metastatic cells demonstrated less invasiveness in the Matrigel invasion assay (Fig. 1d) and lower growth potential in vitro ( Supplementary Fig. 1b) and in vivo (data not shown), although the cells exhibited higher expression levels of EMT markers and metastasis-associated genes (Fig. 2e). This discrepancy between the expression profile and biological phenotype of the UM-UC-3-bone cells can be from the microenvironment-dependent growth of the cells. UM-UC-3-bone cells can grow inside the surrounding physically hard materials, including calcium phosphate in bone; as a result, they may not properly move in conventional in vitro assays. Further investigation is required to delineate the precise characteristics of metastatic bone cells.
In our metastatic model system, AKR1C1 was identified as one of the commonly up-regulated genes in all three metastatic cell lines. In this study, immunohistochemistry demonstrated that AKR1C1 was highly expressed in metastatic lesions of human bladder cancer patients (Fig. 6a-c and Supplementary Fig. 7c). Although a higher AKR1C3 level was reported in the metastatic bone tumors of the prostate cancer 38 , the molecular mechanism of AKR1C1-dependent bladder cancer metastasis had not been previously described.
We investigated the mechanism for the increased AKR1C1 levels in the bladder cancer cell lines and examined whether pro-inflammatory mediators and growth factors, which are related to tumor progression, regulate AKR1C1 expression in bladder cancer cells 26,39 . Our microarray and qRT-PCR data suggested that metastatic tumor cells possess higher expression levels of endogenous IL-6 and IL-1β and their receptors (Figs 2e and 4a,b). IL-1β enhanced the expression of AKR1C1 in the three bladder cancer cell lines, UM-UC-3, TCC-SUP, and 5637 cells ( Fig. 4c and Supplementary Fig. 3). The significance of this IL-1β dependent increase in AKR1C1 was confirmed in clinical specimens of human bladder cancer tissues (Supplementary Fig. 8).
The molecular machinery for the AKR1C1-dependent up-regulation of EMT and invasiveness was examined with AKR1C1 knockdown bladder cancer cells that exhibited less in vitro invasiveness (Fig. 4f). Steroid hormones may affect cancer cell migration; for example, 17β -estradiol suppresses IL-6-dependent tyrosine phosphorylation of Src, p130 Cas , and paxillin in gastric cancer cells 40 . As a result, inhibition of 17β -estradiol by AKR1C1 may recover cell motility in cancer cells. In fact, for the AKR1C1 knockdown cells, Src and FAK phosphorylation was decreased (Fig. 4g). In addition, AKR1C family proteins are implicated in the metabolism of small GTPases, Scientific RepoRts | 6:34625 | DOI: 10.1038/srep34625 such as Ras, Rho, and Rac, through regulating prenylation 41 . AKR1C may produce geranylgeranyl pyrophosphate as the substrate of geranyl-geranyl transferase, providing higher levels of mature small GTPases, such as Rac 41 . The decreased concentration of Rac1 in AKR1C1 knockdown cells (Fig. 4g) may reflect the involvement of this pathway.
Recently, AKR1C2 was reported to regulate cisplatin resistance in bladder cancer 42 . The established liver and bone metastatic cells in this study exhibited higher AKR1C1, AKR1C2 and AKR1C3 expression levels (Figs 3c,d  and 5b), and this overexpression appeared to correlate with cisplatin resistance (Fig. 5c,d). It is speculated that in cisplatin-resistant cells, AKR1C may reduce the levels of cytotoxic lipid peroxidative products from an aldehyde, such as 4-hydroxy-trans-2-nonenal (HNE) and 4-oxo-2-nonenal (ONE) 24 . To reverse this resistance, we examined the effect of a chemical inhibitor of the AKR1C family, flufenamic acid (FFA), and found that FFA could reverse the cisplatin resistance of metastatic liver and bone UM-UC-3 cells (Fig. 5e). Furthermore, AKR1C1 expression was associated with cancer stem cells in lung and thyroid cancer 43,44 . Therefore, AKR1C1 contributes to the development of chemoresistance and induction of stemness-like features in metastatic bladder cancer.
In primary localized bladder cancer, particularly after cisplatin exposure, the surrounding tumor microenvironment may promote IL-1β production by inflammatory lymphocytes. These inflammatory microenvironments with IL-1β production may up-regulate the AKR1C1 levels in a specific cell clone, resulting in cisplatin-resistance and increased invasiveness, which allows the cells to penetrate into blood vessels and circulate throughout the entire body. Finally, these cells can attach to organ tissues, such as the lungs, liver, and bone, to form metastatic nodules. A limitation of this study is the use of a single cell line, UM-UC-3, due to lack of tumor forming potentials of other cells, including T24, J82, and TCC-SUP. As FFA suppressed cisplatin resistance and cell invasion in vitro, FFA may have clinical significance in reversing drug resistance and suppressing metastasis, improving the prognosis of bladder cancer patients (Fig. 6d).

Establishment of UM-UC-3 cells stably expressing tdTomato-Luc2. UM-UC-3 cells were sta-
bly transfected with pCSII-CMV-tdTomato-Luc2 (kindly provided by Dr. Kyoko Hida, Hokkaido University, Sapporo), which was followed by selection using 0.5 mg/mL bleomycin. The resulting live cells were sorted by fluorescence-activated cell sorting (FACS; Aria II, BD, Japan), and the cells with excessive tdTomato expression were isolated. Luciferase activity was examined with a luciferase assay (Promega, Madison, WI).
Orthotopic xenograft model. We have previously reported on the methods for generating this murine model 28 . Six-to eight-week-old female nude mice, Balb/cA Jcl nu/nu (Clea Japan, Inc., Tokyo, Japan), were used for an orthotopic xenograft model. After the mice were anesthetized with either 1.75% isoflurane or intraperitoneal sodium pentobarbital (60 mg/kg), the bladder was pretreated with 100 μ L of 0.1 mg/mL poly-L-lysine (Sigma) for 20 min through the urethra with vascular catheters (24G; Terumo, Tokyo, Japan). After that, the bladder was flushed with 100 μ L of phosphate-buffered saline (PBS), and 5 × 10 6 UM-UC-3-tdTomato-luc2 cells in 50 μ L of PBS were inoculated into the bladder. For the 3 h before catheter removal, mice were placed under anesthesia. Bioluminescent imaging (BLI) was performed with the IVIS Spectrum imaging system (Caliper Life-Sciences, Hopkinton, MA) using a post-intraperitoneal injection of VivoGlo Luciferin In Vivo Grade (Promega). Xenograft-derived cells were collected from the primary site, and three metastatic lesions from the lungs, liver, and bone were used to establish the following sublines: UM-UC-3-bladder-1 and UM-UC-3-lung from mouse 1 and UM-UC-3-bladder-2, UM-UC-3-liver, and UM-UC-3-bone from mouse 2. All animal experiments were conducted in accordance with the guidelines of Hokkaido University Manual for Implementing Animal Experimentation and approved by Institutional Animal Care and Use Committee at Hokkaido University Graduate School of Medicine (Number 12-0092). All researchers who performed procedures using live animal were pre-approved by the Animal Welfare Committee of Hokkaido University based on their completion of required animal use and care training and acceptable previous experience with animal experiments. All ethical issues related to animal experiments and human pathological specimens were discussed and approved by the Ethics Committee of Hokkaido University Graduate School of Medicine.

Establishment of primary and metastatic tumor cells in an orthotopic xenograft model. At
45 days post-inoculation with UM-UC-3 cells, the primary and metastatic tumors were identified with the IVIS imaging system. The mice were euthanized, and tumor tissues were excised and incubated in medium with collagenase for 30 min at 37 °C. After filtration, the cells were cultured with bleomycin to select tumor cells for two weeks, and the primary and metastatic tumor cells were established (Mouse #1; bladder-1 and lung metastatic tumors; Mouse #2; bladder-2, liver-and bone metastatic tumors). The established tumor cells were verified using a fluorescence microscope to detect tdTomato 45 . RNA extraction and quantitative real-time RT-PCR. Total RNA isolation and first-strand cDNA synthesis were performed as previously described 46 . The primer sequences are listed in Supplementary Table S1. Quantitative real-time polymerase chain reaction (qRT-PCR) was performed using the StepOne real-time PCR Scientific RepoRts | 6:34625 | DOI: 10.1038/srep34625 system (Applied Biosystems, Foster City, CA). Data were normalized to the glyceraldehyde 3-phosphate dehydrogenase (GAPDH) expression level and are expressed as the fold change relative to control. Microarray analysis. Total RNA was extracted from parental UM-UC-3 and its sublines derived from orthotopic xenografts (bladder-1, bladder-2, lung, liver, and bone), and microarray analysis was performed using SurePrint G3 Human Gene Expression 8 × 60K v2 (Agilent Array ® ). Gene expression was analyzed in bladder-1 versus lung and bladder-2 versus liver and bone cell lines.
Cell viability assay. Cells (1 × 10 5 ) suspended in complete DMEM were plated in 60-mm dishes. After 24 h, cisplatin and FFA were added to the culture medium. Three days after treatment, the cells were trypsinized and counted.
Invasion assay. The invasion assay was performed as previously described 47 . Briefly, 1 × 10 5 cells suspended in serum-free medium were seeded onto the upper chamber. Medium containing 10% FBS as a chemoattractant was added to the lower chamber. After 22 h, invaded cells were counted at 200x magnification in at least 3 randomly selected fields.
Immunoblotting. Cells were washed twice with cold PBS and lysed in a buffer containing 0.5% NP-40, 10 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA, 50 mM NaF, 1 mM PMSF, and 1 mM Na 3 VO 4 , and the lysate was clarified by centrifugation at 15,000 rpm for 10 min. The supernatants were then subjected to 8-12% SDS-PAGE. Separated proteins were transferred to polyvinylidene difluoride membranes and blocked using Tris-buffered saline containing Tween-20 (TBS-T) with 5% skim milk at room temperature for 1 h. Primary antibodies were incubated at 4 °C overnight and then washed with TBS-T, which was followed by incubation with secondary antibodies. The membranes were washed 3 times in TBS-T, and the signal was developed using ECL (GE Healthcare, Little Chalfont, UK), which was followed by detection using the LAS4000 (GE Health Care).
Immunofluorescence of paxillin. Immunofluorescence of paxillin was performed according to the previous report 47 . Briefly, cells were fixed with 3% paraformaldehyde for 15 min at room temperature, permeabilised with 0.1% Triton X-100 in PBS for 4 min at room temperature, and then incubated with 1% bovine serum albumin to block nonspecific binding of antibodies. The cells were further incubated with anti-paxillin antibody (1:1000 dilution) overnight at 4 °C, after which immune complexes were detected by incubation for 1 h at room temperature in the dark with AlexaFluor488-conjugated secondary antibodies (1:250 dilution). Images were acquired with an FV-10i confocal microscope (Olympus, Tokyo, Japan).

Immunohistochemical analysis.
Formalin-fixed paraffin-embedded tissues, including human bladder cancer specimens (surgically resected in Department of Urology in Hokkaido University Hospital, Sapporo, Japan) that were obtained with informed consent, were sectioned and stained with hematoxylin and eosin (H&E). Immunohistochemistry (IHC) was performed using an antibody to AKR1C1 (Lifespan Biosciences). All ethical issues related to the animal experiments and human pathological specimens were discussed and approved by the Ethics Committee of Hokkaido University Graduate School of Medicine and Hokkaido University Hospital, respectively. Experiments were performed in accordance with the relevant Ethical Guidelines.
Statistical analysis. Data are presented as the mean and standard deviations (S.D.) for at least 3 independent experiments. Student's t-test, Mann-Whitney U-test, or Spearman's test were used to analyze significant differences. JMP ® version 10 (SAS Institute, Inc., Cary, NC) was used for all calculations. P < 0.05 was considered statistically significant.