Nkx2.8 Inhibits Epithelial–Mesenchymal Transition in Bladder Urothelial Carcinoma via Transcriptional Repression of Twist1 Free

The prognosis of metastatic bladder urothelial carcinoma is extremely poor, with a median survival time of less than 15 months, even with systematic therapy (1). However, the mechanism underlying urothelial carcinoma metastasis is not clear. Epithelial-to-mesenchymal transition (EMT), a process of cell phenotypic change, is characterized by decreased E-cadherin expression and weakened adhesive cell–cell or cell–stroma attraction, and subsequently facilitates cell migration (2–4). EMT has been implicated in the conversion of early-stage tumors into invasive malignancies (2–4), which suggests that a better understanding of the mechanism underlying EMT may be important with respect to the clinical management of urothelial carcinoma.

Twist, a highly conserved transcriptional factor, has been well known as a key EMT inducer (5). Twist downregulates E-cadherin expression, upregulates fibronectin and vimentin expression, and subsequently facilitates EMT (5–7). Yang and colleagues (6) have found that forced expression of Twist results in decrease of E-cadherin–mediated adhesion and induction of cell motility, which suggests that Twist1 plays a promoting role in EMT. It is also well known that Twist1 participates in regulating the expression of Bmi-1, miR-200, and miR-205 (7–8), which are all involved in EMT. In urothelial carcinoma, Twist has been found to be associated with tumor grade and progression, and is inversely correlated with E-cadherin expression (9–11). Moreover, in bladder cancer tissues, Twist is always been found to be negatively associated with the expression of E-cadherin (10–11). In addition, Twist is reported to be involved in urothelial carcinoma invasiveness (9–10, 12). However, how Twist expression levels are regulated in urothelial carcinoma remains a mystery.

Human Nk2 homeobox 8 (Nkx2.8), which acts as a transcription factor, is a member of the NK-2 gene family (13). Nkx2.8 usually binds to DNA sequences containing 5′-(C/T)AAG-3′ motifs (14–16). The findings of several studies have suggested that Nkx2.8 acts as a tumor suppressor in human carcinogenesis (17–19). In lung cancer, Harris and colleagues (17) found most tumors had low expression of Nkx2.8 and enforced expression of Nkx2.8 can inhibit proliferation of lung cancer cells. Lin and colleagues (18) reported that downregulation of Nkx2.8 can activate NF-κB and promote angiogenesis in esophageal cancer cells. Qu and colleagues (19) also found a tumor-suppressive role of Nkx2.8 in human liver cancer. Our previous study established that Nkx2.8 expression was markedly reduced in urothelial carcinoma tissues and that Nkx2.8 negativity was associated with lymph node metastasis and prognosis in urothelial carcinoma patients (20). However, the role of Nkx2.8 in metastasis and the regulatory mechanisms underlying this phenomenon are largely unknown. In the current study, we investigated the role of Nkx2.8 in EMT, the relationship between Nkx2.8 and Twist1 in urothelial carcinoma, and the mechanisms underlying the effects of Nkx2.8 in urothelial carcinoma.

The bladder cancer cell lines T24, 5637, and J82 were obtained from the ATCC in 2013. The BIU87 and EJ were obtained from the Institute of Urology at the First Affiliated Hospital of Peking University as a gift in 2012. All cell lines were maintained in RPMI 1640 (Invitrogen) supplemented with 10% FBS (HyClone), penicillin (100 units/mL), and streptomycin (100 units/mL) and tested to ensure mycoplasma free. All cell lines used in this study were authenticated 3 months before the beginning of the study (2013) based on viability, recovery, growth, morphology, and isoenzymology by the supplier, and all the cell lines have not been in culture for more than 2 months. The pBabe-Nkx2.8 and pSuper-retro-Nkx2.8 RNAi(s) were generated as described previously (20).

The wild-type human TWIST1 promoter and the TWIST1 promoter with a deletion or mutation of the Nkx2.8 binding sites were individually cloned into the pGL3 luciferase reporter plasmid (Promega). Urothelial carcinoma cells with endogenous silencing of Nkx2.8 and cells with the forced expression of exogenous Nkx2.8 were generated as previously described. The T24 cells exhibited no expression of Nkx2.8 and were infected with retroviruses carrying pBabe-Nkx2.8. The BIU87 cells showed high expression levels of Nkx2.8 and were infected with retroviruses carrying pSuper-retro-Nkx2.8-shRNAs. The 5637 cells showed moderate expression levels of Nkx2.8 and thus were infected with retroviruses carrying either pBabe-Nkx2.8 or pSuper-retro-Nkx2.8-shRNAs. Stable cell lines were selected with 0.5 μg/mL puromycin for 10 days after transfection. Cell lysates, which were prepared from pooled populations of cells in sample buffer, were fractionated by SDS–PAGE to confirm Nkx2.8 protein levels.

Total RNA samples from cultured cells were extracted using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. The sequences of the primers are listed in Supplementary Table S1. Quantitative real-time PCR (qRT-PCR) was performed using an ABI PRISM 7500 Sequence Detection System (Applied Biosystems). The housekeeping gene glyceraldehyde-3 phosphate dehydrogenase (GAPDH) was used as an internal quantitative control.

Western blot and immunofluorescence analyses were performed according to the standard methods as described previously (20) using anti-Nkx2.8 (Santa Cruz Boitechnology, Inc.), anti-Twist1 (Abcam), anti–E-cadherin, anti–α-catenin, anti-fibronectin, and anti-vimentin (BD Transduction Laboratories) antibodies. For the Western blot assays, anti–α-tubulin antibody (Sigma) was used as a loading control. For immunofluorescence analysis, the coverslips were counterstained with 4', 6-diamidino-2-phenylindole and imaged with a confocal laser-scanning microscope (Olympus FV1000). The quantification of immunofluorescence staining was measured using Olympus FV10-ASW software.

The study has been approved by Institutional Review Board of Sun Yat-sen University Cancer Center, and the study was performed in accordance with Declaration of Helsinki. Written informed consent was obtained from the patients before the study began. Twist1 expression in the paraffin-embedded bladder cancer tissues of the previously studied 161 cases was detected by immunohistochemistry (IHC) using an anti-Twist1 antibody. Patient information and the method for scoring Nkx2.8 have been previously reported (20). For the Twist1 labeling index, we used the following scoring system. In brief, the proportion of positive cells in the stained sections was evaluated at ×200 magnification, and the mean value of 10 representative fields analyzed from each section was recorded. The proportion of positive cells was scored as follows: <25%, 1; 25% to 50%, 2; 50% to 75%, 3; and >75%, 4. We judged the intensity of Twist1 staining according to 4 categories: negative, 0; weak, 1; moderate, 2; and strong, 3. We used the product of the staining intensity score and the proportion of positive tumor cells as the staining index. Then, the scores were divided into 2 groups (0–4, low expression; 5–12, high expression).

Cells infected with vector, Nkx2.8, or Nkx2.8 RNAi were seeded in 24-well dishes coated with Growth Factor–Reduced Matrigel (BD Biosciences) at a density of 1 × 10⁴ cells/well and were covered with growth medium supplemented with 2% Matrigel. The medium was exchanged with 2% Matrigel every 3 to 4 days. Images were captured via microscopy at 2-day intervals for 2 to 3 weeks. The quantification of 3D morphogenesis Matrigel culture was represented by mean spheroid area measured using Olympus cellSens Standard 1.9.

Cells infected with vector, Nkx2,.8 or Nkx2.8 RNAi were seeded in 6-well plates and grown under permissive conditions until reaching 90% confluence. The cells were then serum starved for 24 hours, and a linear wound was created in the confluent monolayer using a pipette tip. Wounds were photographed immediately (time 0 hour) and thereafter at 10 and 20 hours. Wound size was measured randomly at 10 sites perpendicular to the wound. Each experiment was repeated at least 3 times.

Cells infected with vector, Nkx2.8, or Nkx2.8 RNAi were serum starved for 24 hours. Then, 2 × 10⁴ cells were plated into the upper chamber of a polycarbonate transwell filter chamber coated with Matrigel (BD Biosciences) and incubated for 20 hours in serum-free medium. The lower chamber contained the usual serum-containing medium as chemoattractant. At the end of the 20-hour incubation, cells inside the chamber were removed with cotton swabs. The invaded cells that remained on the lower surface of the filter were fixed in 1% paraformaldehyde, stained with hematoxylin, and counted (10 random 100× fields per well). Cell counts were expressed as the mean number of cells per field of view. Three independent experiments were performed, and the data were presented as the average ± SEM.

The chromatin immunoprecipitation (ChIP) assay was performed according to the protocol described previously. In brief, cells were plated at a concentration of 2 × 10⁶ cells per 100-mm diameter dish and cross-linked with 1% formaldehyde for 10 minutes. The cells were trypsinized and resuspended in lysis buffer. The nuclei were then isolated and sonicated to shear the DNA into 500-bp to 1-kb fragments, which was verified by agarose gel electrophoresis.

Equal aliquots of chromatin supernatants were incubated with 1 μg of anti-Nkx2.8 antibody (Santa Cruz Biotechnology, Inc.) or an anti-IgG antibody (Millipore) overnight at 4°C with rotation. DNA was extracted, and the TWIST1 promoter was amplified with a qPCR assay. All the ChIP assays were performed 3 times, and representative results were presented. All the sequences of the PCR primers are shown in Supplementary Table S2.

Twenty thousand cells were seeded in triplicate in 48-well plates and allowed to settle for 24 hours. Using the Lipofectamine 2000 reagent according to the manufacturer's recommendations, 100 ng of luciferase reporter plasmids containing different fragments of the TWIST1 promoter, or the control luciferase plasmid, plus 1 ng of pRL-TK Renilla plasmid (Promega) was transfected into the cells. The luciferase and Renilla signals were measured 24 hours after transfection using the Dual Luciferase Reporter Assay Kit (Promega) according to a protocol provided by the manufacturer.

BALB/c-nu mice (4–5 weeks of age, female, 18–20 g) were purchased from Charles River Laboratories. All the experimental procedures were approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University. For the tail vein injection model, the BALB/c nude mice were randomly divided into 5 groups (n = 6/group). Each group of mice was intravenously injected in the tail vein with 5637/vector cells, 5637/Nkx2.8 cells, 5637/scrambled cells, or 5637/Nkx2.8 RNA interference cells (5 × 10⁶) per mouse. After 8 weeks, the animals were euthanized, and tumors were excised, weighed, and embedded in paraffin. For the orthotopic xenograft bladder cancer model (21), the bladder was washed with PBS and then scratched with the catheter tip before instilling 100 μL of 2 × 10⁶ cells through a small catheter. The urethra was temporarily closed with a single, sterile suture at the distal part of the urethra, thus retaining the cells in the bladder for 2 hours. The BALB/c nude mice were randomly divided into 5 groups (n = 6/group). Each group of mice was instilled with T24/vector cells, T24/Nkx2.8 cells, 5637/scrambled cells, 5637/Nkx2.8 RNAi cells, or 5637/Nkx2.8 RNAi/Twist1 RNAi cells transfected with luciferase. Xenograft implantation was confirmed by the presence of bioluminescence activity 1 week after cell implantation. For bioluminescent evaluation, each mouse received 150 mg luciferin/kg body weight through an intraperitoneal injection. Imaging of the mice was then conducted in anesthetized conditions with the IVIS Lumins III. After 8 weeks, the bioluminescence activity was detected again, the animals were sacrificed, and bladder were excised, weighed, and embedded in paraffin. Serial 6.0-mm sections were cut and subjected to hematoxylin and eosin (H&E) staining with Mayer's hematoxylin solution. Images were captured using the AxioVision Rel.4.6 Computerized Image Analysis System (Nikon Eclipse 80i).

Statistical analyses were conducted using the SPSS 11.0 statistical software package. Statistical tests for the data analysis included the Fisher exact test, the log-rank test, χ2, and Student two-tailed t test. The data were presented as the mean ± SD. A P value of ≤0.05 was considered statistically significant.

To determine the role of Nkx2.8 in urothelial carcinoma metastasis, we evaluated whether Nkx2.8 can influence the EMT phenotype in urothelial carcinoma cells. We established forced exogenous Nkx2.8-expression cells and Nkx2.8-silenced cells and examined the expression levels of several EMT-related proteins. Our Western blot results showed that the expression levels of E-cadherin and α-catenin, which mediate cell–cell and cell–stroma adhesion, were much higher in Nkx2.8-overexpression cells than in vector cells. Accordingly, the expression levels of the mesenchymal markers fibronectin and vimentin were markedly reduced in Nkx2.8-overexpression cells compared with vector cells (Fig. 1A). Immunofluorescence analysis in 2D and 3D cultured cells confirmed the above findings (Fig. 1B). Consistent with the above results, Nkx2.8-silenced cells displayed decreased E-cadherin and α-catenin expression and enhanced fibronectin and vimentin expression (Fig. 1A and Supplementary Fig. S1). Our data showed that Nkx2.8 inhibits the EMT phenotype in urothelial carcinoma cells.

EMT can augment tumor cell motility and lead to tumor cell invasion into the basement membrane, which leads to advanced metastasis. Thus, we investigated the effect of Nkx2.8 on urothelial carcinoma cell invasion and metastatic potential. Matrigel-coated Boyden chamber invasion assay, whose results were represented as the number of migrated cells, showed that the invasiveness of Nkx2.8-overexpression cells was much weaker than that of vector cells (Fig. 2A). 3D morphogenesis cultures revealed that Nkx2.8 overexpression reduced the numbers of irregular branched structures that characterize the invasive phonotype. Mean spheroid area was significantly smaller in Nkx2.8-overexpression cells compared with vector cells (Fig. 2B). Furthermore, wound-healing assay showed that cell motility was dramatically hampered by Nkx2.8 overexpression (Fig. 2C). These findings suggested that Nkx2.8 overexpression inhibited urothelial carcinoma cell motility and invasion.

We next evaluated the in vivo effects of Nkx2.8 on invasion and metastasis using an experimental metastasis assay, in which we injected cells transfected with Nkx2.8 or vector into the lateral tail veins of nude mice and evaluated cell growth in the lungs after 8 weeks. Fewer metastatic nodes were found on the lung surfaces of the Nkx2.8-overexpression group than on the lung surfaces of the vector group. In addition, H&E staining showed that there were smaller and fewer microscopic metastatic nodules in the Nkx2.8-overexpression group than in the vector group (Fig. 2D). Furthermore, orthotopic xenograft bladder cancer model also provided data about Nkx2.8's role on bladder cancer cells invasiveness. As shown in Supplementary Fig S2A–S2C, mice bladder implanted with T24/Nkx2.8 cells showed submucosa infiltration lesion, whereas those with T24/vector cells showed muscle-invasive disease. These data indicate that Nkx2.8 acts as a negative regulator of the aggressive metastasis of urothelial carcinoma cells.

To investigate the impact of Nkx2.8 on EMT of urothelial carcinoma cells further, we assessed the invasiveness and metastatic potential of Nkx2.8-silenced cells. Both Matrigel-coated Boyden chamber invasion assay and 3D morphogenesis cultures indicated that ablation of endogenous Nkx2.8 induced urothelial carcinoma cell invasiveness (Fig. 3A and B). Wound-healing assay also revealed that urothelial carcinoma cells with Nkx2.8 ablation exhibited significantly enhanced mobility compared with control cells (Supplementary Fig S3A). Taken together, these results suggested that silencing Nkx2.8 dramatically promoted urothelial carcinoma cell motility and invasiveness.

To determine the role of Nkx2.8 in urothelial carcinoma cell metastasis in vivo, we constructed a lung metastasis model by injecting cells with silenced endogenous Nkx2.8 or control cells into the tail veins of 6-week-old nude mice. After 8 weeks, many more metastatic nodes were observed on the lung surfaces of the Nkx2.8-silenced group than on the lung surfaces of the control group. H&E staining confirmed that both the numbers and the volumes of microscopic metastatic lesions were markedly increased in the lungs of mice injected with Nkx2.8-silenced cells compared with the lungs of control mice (Fig. 3C). Moreover, as shown in Supplementary Fig S3B–S3D, orthotopic xenograft bladder cancer model, mice bladder implanted using 5637/Nkx2.8 shRNA#2 cells showed cancer cells infiltrated into submucosa, whereas those with 5637/scrambled cells showed dysplastic cells with nuclear hyperchromatism within the mucosa. These results confirmed that Nkx2.8 exerts negative regulatory effects on urothelial carcinoma cell metastasis.

To explore the underlying mechanism of Nkx2.8-inhibiting metastasis, we detected the relationship between Nkx2.8 and EMT-inducing transcription factors, including TWIST1, TWIST2, SNAIL, SLUG, ZEB1, and ZEB2, using qPCR (22). We discovered that Nkx2.8 downregulated TWIST1 mRNA expression but had no significant effect on other above-mentioned gene mRNA expression. Twist1 is a key regulator of EMT and plays an important role in urothelial carcinoma metastasis (5–11). Further study revealed an inverse relationship between Nkx2.8 and Twist1 expression in cultured urothelial carcinoma cell lines (Fig. 4A). Western blotting analysis also revealed that Twist1 expression was significantly decreased in cells forced to express exogenous Nkx2.8 compared with vector control cells; however, Twist1 expression was markedly upregulated in Nkx2.8-silenced cells compared with vector control cells (Fig. 4B and C). Twist1, which also acted as transcriptional factor, was reported to be a direct regulator of BMI1 and E-cadherin, two important elements in EMT occurrence (7). Therefore, we detected the effect of Nkx2.8 on BMI1 and E-cadherin. As shown in Fig. 4D, BMI1 was downregulated, whereas E-cadherin was upregulated in Nkx2.8-overexpression cells. Consistent with these findings, BMI1 was upregulated, and E-cadherin was downregulated in Nkx2.8-silenced cells. These results are in accordance with the results pertaining to the effects of Twist1 on Bmi1 and E-cadherin. Our findings indicated that Nkx2.8 inhibited Twist1 expression.

As a transcriptional factor, Nkx2.8 binds to special DNA sequences in promoters. Interestingly, we identified two potential binding sites for Nkx2.8 in the TWIST1 promoter, each of which included three adjacent core sequences (Fig. 5A). Therefore, we speculated that Nkx2.8 binds to the promoter locus and regulates TWIST1 transcription. We conducted ChIP assay to verify this speculation. As shown in Fig. 5B, we detected 13 TWIST1 promoter loci. As expected, Nkx2.8 bound to the 2nd and 10th loci of the TWIST1 promoter (Fig. 5B). These loci extend from –1510 bp to –1472 bp and from +774 bp to + 801 bp, respectively (Fig. 5A). This binding was restrained when Nkx2.8 was silenced (Fig. 5C), which confirmed that Nkx2.8 directly targets the TWIST1 promoter.

Furthermore, we found that overexpressing Nkx2.8 decreased the luciferase activity driven by the wild-type TWIST1 promoter, whereas silencing Nkx2.8 increased the luciferase activity driven by the TWIST1 promoter. However, neither overexpression nor knockdown of Nkx2.8 had any effect on the luciferase activity levels of TWIST1 promoters containing a deleted or mutated 2nd or 10th locus (Fig. 5D), indicating that both the 2nd and the 10th loci are needed for Nkx2.8 to target the TWIST1 promoter.

To further analyze the functional correlation between Twist1 and Nkx2.8, we tested whether the ablation of Twist1 expression in Nkx2.8-silenced urothelial carcinoma cells could restore the expression of factors downstream of Twist1 and the invasive phenotype of these cells. As expected, silencing Twist1 expression with TWIST1 shRNA in Nkx2.8-silenced cells decreased Bmi1 expression and increased E-cadherin expression (Fig. 6A). Moreover, the Boyden chamber invasion assay results suggested that Twist1 knockdown markedly restrained the restored invasiveness of the Nkx2.8-silenced cells (Fig. 6B). The results of 3D Matrigel cultures and a wound-healing assay confirmed the restoration of the motility of Nkx2.8-silenced cells via the expression of TWIST1 shRNA (Supplementary Figs. S4 and S5).

To decipher the functional correlation between Nkx2.8 and Twist1 in vivo, we performed xenograft tumor experiments. As shown in Fig. 6C, Nkx2.8-silenced cells caused mice to exhibit many more metastatic nodes on their lung surfaces, whereas Twist1 knockdown in Nkx2.8-silenced cells caused mice to exhibit greatly reduced numbers of metastatic nodes on their lung surfaces. H&E staining confirmed that the tissues comprising Nkx2.8-silenced cells in which Twist1 was knocked down displayed significantly less and smaller microscopic metastatic lesions than the tissues comprising Nkx2.8-silenced cells expressing Twist1. In addition, in orthotopic xenograft bladder cancer model, mice bladder implanted using 5637/Nkx2.8 shRNA#2/Twist1 shRNA#2 cells illustrated atypical hyperplasia cells within the mucosa, whereas those with 5637/Nkx2.8 shRNA#2 cells showed cancer cells infiltrated into submucosa (Supplementary Fig S6A–S6C). These results suggest that Twist1 ablation could restore the inhibitory effect of Nkx2.8 on EMT and the metastatic potential of Nkx2.8-silenced urothelial carcinoma cells.

To confirm the findings derived from the above in vitro and animal experiments, we analyzed the levels of Nkx2.8 and Twist1 in 15 freshly collected clinical urothelial carcinoma samples. Real-time RT-PCR analyses revealed that NKX2.8 mRNA levels were inversely correlated with TWIST1 expression levels (r² = 0.589, P = 0.0008; Fig. 7A). IHC analysis of 161 tissue specimens also showed that Nkx2.8 expression was inversely correlated with Twist1 expression (P < 0.0001; Fig. 7B and C). The data pertaining to these cases have been published previously. Patients with high Twist1 expression levels have much worse survival than patients with low Twist1 expression (Supplementary Fig S7). Further analysis showed that patients with both Nkx2.8 positivity and low Twist1 expression had the best survival rate (Fig. 7D and E). In summary, our results reveal that an inverse relationship exists between Nkx2.8 and Twist1 expression and that Nkx2.8 positivity and low Twist1 expression lead to better outcomes in urothelial carcinoma patients than Nkx2.8 negativity or high Twist1 expression.

EMT is an initial event in cancer cell invasion and metastasis characterized by repressed E-cadherin expression and subsequently facilitates cell migration and invasion (2–4). Non–muscle-invasive urothelial carcinoma carries a high risk of progressing to muscle-invasive urothelial carcinoma. This process is accompanied by decreased E-cadherin expression, suggesting that EMT is involved in the process (23). Muscle-invasive urothelial carcinoma has a high incidence of lymph node or distal metastasis, again suggesting that EMT is involved in its development (24). Nkx2.8 has been reported to be involved in the carcinogenesis and progression of several human cancers (17–19). Our previous study showed that Nkx2.8 acted as a tumor suppressor in urothelial carcinoma by inhibiting cancer cell proliferation through the MEK/ERK pathway. Moreover, we found that negative Nkx2.8 expression was associated with lymphatic metastasis (20). In this study, we explored the correlation between Nkx2.8 expression and EMT in urothelial carcinoma. We found that Nkx2.8 overexpression inhibited urothelial carcinoma cell EMT, whereas Nkx2.8 silencing promoted urothelial carcinoma cell EMT in vitro and urothelial carcinoma cell metastatic potential in vivo. Thus, our present study has demonstrated that Nkx2.8 can function as a novel urothelial carcinoma EMT inhibitor.

Twist1, which suppresses E-cadherin expression, has been considered an important promoter in EMT (5–8). Numerous studies have shown that Twist1 plays an important role in urothelial carcinoma (9–11). However, the upstream regulatory mechanism underlying the effects of Twist1 in EMT is not well illuminated. Twist1 has been found to be a downstream target of NF-κB–like transcription factor in both Drosophila and vertebrates (25–27). Tan and colleagues (28) reported that Twist1 expression is induced by the embryonic factor high mobility group A2 (HMGA2), which causes mesenchymal transition, in mouse mammary epithelial cells. It has also been reported that Twist1 stability can be regulated by the ubiquitin-proteasome system in embryos. (29). However, little is known regarding the mechanism underlying Twist1 regulation in EMT in bladder cancer. Shiota and colleagues (12) found that Foxo3a can negatively regulate Twist1 and positively regulate E-cadherin in bladder cancer cells. Our study has provided evidence that Nkx2.8 can transcriptionally repress Twist1 expression by directly binding to the Twist1 promoter and has thus also provided strong evidence that this process represents the mechanism by which Nkx2.8 regulates Twist1 in EMT. Furthermore, our study has demonstrated the existence of a novel pathway (Nkx2.8-Twist1-E-cadherin) that participates in urothelial carcinoma EMT.

Nk2 family proteins are characterized by three highly conserved domains, one of which is homeodomain, a region for specific DNA binding that usually binds to DNA sequences containing 5′CTTG3′ or 5′CAAG3′ (14, 16). Kajiyama and colleagues found that Nkx2.8 bound to the active AFP promoter and that antisense inhibition of Nkx2.8 mRNA translation selectively reduced endogenous human AFP gene expression (30). Nkx2.8 can also repress AKIP1 expression by directly targeting the AKIP1 promoter and then inhibiting NF-κB activation in esophageal cancer (18). Interestingly, we found that the promoter region of TWIST1 contains two potential binding sites for Nkx2.8. This finding indicates that Nkx2.8 binds directly to the TWIST1 promoter. Our ChIP analysis and luciferase assay showed that Nkx2.8 binds to the two suspected regions mentioned above and represses activation of the TWIST1 promoter. Interestingly, Nkx2.8 must bind both sites to repress TWIST1 promoter activation, as a TWIST1 promoter featuring deletions or mutations of the 2nd or 10th locus is not affected by Nkx2.8. These two loci, which each contain three adjacent core sequences, are located on either side of the transcription start site. Thus, the TWIST1 promoter differs from the other known Nkx2.8 targets, which contain only one site for Nkx2.8 (18, 30). The above result is illustrative of a new mechanism by which Nkx2.8 exerts its regulatory effects and has thus provided us with important information that can be used to identify new targets of Nkx2.8. However, the precise mechanism by which Nkx2.8 regulates TWIST1 transcription warrants further investigation.

Previous studies have demonstrated the prognostic significance of Nkx2.8 and Twist1 in urothelial carcinoma (9–11, 20), but no study has explored the relationship between Nkx2.8 and Twist1. Here, we showed that an inverse relationship exists between Nkx2.8 and Twist1 expression in urothelial carcinoma patients, a result consistent with those of previous studies showing that Nkx2.8 represses Twist1 expression in vitro. This inverse relationship between Nkx2.8 and Twist1 expression has an impact on the prognoses of urothelial carcinoma patients, as patients displaying Nkx2.8 positivity and low Twist1 expression have a better prognosis than patients displaying Nkx2.8 negativity or high Twist1 expression. In our study, Nkx2.8 failed to repress Twist1 expression in approximately 45% of cases with positive Nkx2.8 expression, a finding that may be attributable to the fact that Nkx2.8 is universally expressed at low levels in urothelial carcinoma. The prognosis of these patients was not significantly different from that of patients with negative Nkx2.8 expression, which indicates that the Twist1 repression is the main mechanism by which Nkx2.8 exerts its effects. Fifteen percent of cases with negative Nkx2.8 expression showed low Twist1 expression, implying that Twist1 inhibition occurs independently of Nkx2.8 expression. Thus, additional studies are needed to identify other upstream regulators of Twist1.

In conclusion, the findings of our study serve as an extensive explanation of the mechanism underlying the effects of Nkx2.8 on urothelial carcinoma EMT, as we elucidated the crucial role of the Nkx2.8-Twist1 pathway in urothelial carcinoma EMT and may have thus identified novel therapeutic targets for the treatment of urothelial carcinoma.

No potential conflicts of interest were disclosed.

Conception and design: C. Yu, Z. Zhang, F. Zhou

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Yu, Z. Liu, Q. Chen, Y. Li, L. Jiang, F. Zhou

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Yu, Q. Chen, F. Zhou

Writing, review, and/or revision of the manuscript: L. Jiang, Z. Zhang, F. Zhou

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Yu

Study supervision: Z. Liu, Z. Zhang, F. Zhou

This study was supported by grants from the National Natural Science Foundation of China (nos. 81272810, 81402114, 81300597, 81672530, 81472385, 81772726, and 81772716) and Natural Science Foundation of Guangdong Province (no. 2016A030 310213).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.