This is the first study showing that the NOX4-ROS signal contributes to the survival of human urothelial carcinoma cells via progression of the G1/S transition. This study also shows the practical use of ROS labeling in urinary cytology. Previously, we showed that a novel isoform of the DNA repair enzyme ALKBH, ALKBH-8, contributes to progression of urothelial carcinoma cells via NOX1-ROS signal-mediated resistance to apoptosis induction [14
]. Although ROS was produced in cancer cells irrespective of pathological grade or stage, immunohistochemistry showed high expression of NOX1 proteins in high-grade, superficially, and deeply invasive carcinomas (pT1 and > pT2), as well as in carcinoma in situ, but not in low-grade and non-invasive phenotypes (pTa). The present study uncovered key data for the resolution of this problem and provided a novel mechanism involved in the development of bladder cancer: NOX4 was highly expressed in near-equivalent levels in low and high grade or non-invasive and invasive urothelial carcinomas, including dysplasia, but not in normal urothelium. Moreover, NOX4 silencing reduced ROS generation and suppressed cancer cell growth via p16-dependent cell cycle arrest at the G1 phase, both in vitro and in vivo. These data indicate that NOX4-mediated ROS generation contributes to an early step in urothelial carcinogenesis and cancer cell survival. In addition to NOX4, NOX1-mediated enhancement of ROS generation might result in bladder cancer cells of a more aggressive phenotype. Dysplasia represents an early morphologic manifestation of progressive alteration between normal urothelium and carcinoma in situ, and clinicians must be attentive to the clinical development. In contrast to CIS, clinical therapy is not directly indicated for dysplasia. Therefore, pathologists should differentiate dysplasia from other flat atypical urothelial lesions. There have been a number of reports regarding molecular markers that enable accurate differential diagnosis of flat atypical urothelial lesions. CK20 and CD44 are the most useful markers for distinguishing atypia of unknown significance (AUS) from dysplasia (for example, CK20 expression is usually limited to the umbrella cells but is expressed in the deeper mucosal layer in AUS) and p53 immunostains typically highlight the dysplastic cells [19
]. However, Murata et al. [19
] demonstrated that molecular and immunohistochemical analyses, including fluorescence in situ hybridization (FISH), to detect expression of the high molecular weight cytokeratin, Ki-67 and p53 can discriminate between neoplastic and non-neoplastic lesions. However, they cannot reliably resolve the diagnostic variation of flat intraepithelial lesions. Immunohistochemical analysis of NOX4, in addition to these markers, will enable differential diagnosis of dysplasia. NOX4 knockdown was expected to have no significant effects on NOX1 expression but ROS levels were strongly decreased, and the same phenomenon was also observed in a NOX1 knockdown experiment (data not shown). Therefore, in addition to NOX1, NOX4 is required for the maintenance of intracellular ROS in urothelial carcinoma cells. NOX4 should therefore be the target molecule in bladder cancer treatment. Since the expression of NOX4 was relatively high not only in urothelial carcinoma cells but also in dysplasia as precancerous lesions, we hope to develop prophylaxis against bladder cancer occurrence and recurrence with a new strategy focusing on the NOX4-ROS signal. Various studies have been carried out on the role of NOX4 in pancreatic cancer progression. Mochizuki et al. demonstrated inhibition of NOX4 activates apoptosis via the Akt/apoptosis signal-regulating kinase 1 pathway in pancreatic cancer PANC-1 cells [20
], but apoptosis was not observed in urothelial carcinoma cells by NOX4 knockdown, unlike in NOX1 gene silencing [14
]. The biological and clinical significance of NOX may be specific to NOX isoforms, cell types, and/or a combination of NOX family members.
We demonstrated here for the first time the validity of ROS labeling for urinary cytological diagnosis. The advantages of urine cytology include high specificity; it is an established technique and minimal sample processing is required. In contrast, the high level of expertise required, significant interobserver variation, and low sensitivity (especially for low-grade phenotypes) are considered disadvantages [8
]. At present, several biomarkers are commercially available: the bladder tumor antigen (BTA) test measures urine levels of H-related protein, which is similar to H protein and is secreted in high levels by tumor cells. The BTA test outperforms cytology in sensitivity, but its specificity falls below that of cytology because of false positive results by inflammatory or infectious conditions [23
]. Nuclear matrix protein (NMP) 22, a regulator of mitosis, is known to be increased in malignant urothelium, and quantitative immunoassay of NMP22 demonstrates superior sensitivity for bladder cancer compared to urine cytology [25
]. However, the specificity is generally lower than that of urine cytology due to NMP22 dependency on the cut-off points used [26
]. Lately, FISH has attracted considerable interest for its high sensitivity (69-96%) and specificity (65-96%) [27
]. FISH identifies common alterations in the chromosomal copy number (chromosomes 3, 7, and 17) and loss of the 9p21 locus, and is largely unaffected by inflammatory conditions. Recent studies have shown that FISH could detect subclinical neoplastic changes [28
]. However, FISH analysis of urine specimens requires specialized laboratory equipment, expensive reagents, and a high level of expertise; therefore, it is not suited to high throughput testing. The International and European guideline panels no longer recommend the use of FISH tests, including UroVysion, owing to inferior specificity and a lack of reliable prospective testing [1
]. A fault common to all of these tests is that they cannot make full use of the merits of urine cytology. A new diagnostic system that counteracts the low sensitivity of cytology but retains its high specificity should be established. Our idea was to introduce labeling of ROS produced by urothelial carcinoma cells (but not by non-malignant cells) to conventional cytology: ROS-positive cells are theoretically considered to be malignant cells despite little morphological atypia. The present study clearly indicates that extraction of ROS-producing cells using the fluorescent dyes CM-DCFDA and DHE statistically improves the accuracy of urine cytology. Moreover, this system was not affected by post-surgical or therapy-related inflammation or degeneration. Sometimes, pathologists and cytologists face the problem of under- or over-diagnosis, and the number of suspicious cases that cannot be identified as malignant or benign increase due to degenerative morphological changes (for example, the size of malignant cells are miniaturized, or normal cells are distended and exhibit atypia). This problem tends to occur especially in follow-up urine samples after surgical or immune therapy. ROS-C, available in such cases, demonstrated high sensitivity and high specificity. Since the fluorescence level of ROS-reactive dyes was much higher than that of non-specific self-fluorescence, there were very few ROS-C false-positive results in the current study. Inflammatory cells such as neutrophils and macrophages were labeled by ROS dyes but were distinguished from epithelial tumor cells by cytomorphological characteristics. Inter- and intra-observer variabilities are worth considering for cytologists who lack experience in handling urine samples and/or fluorescence microscopy. We are trying to construct more objective rules to help decide whether ROS positive cells are malignant--for example, by setting a threshold of fluorescence intensity.