QKI‐6 inhibits bladder cancer malignant behaviours through down‐regulating E2F3 and NF‐κB signalling

Abstract Quaking homolog (QKI) is a member of the RNA‐binding signal transduction and activator of proteins family. Previous studies showed that QKI possesses the tumour suppressor activity in human cancers by interacting with the 3'‐untraslated region (3'‐UTR) of various gene transcripts via the STAR domain. This study first assessed the association of QKI‐6 expression with clinicopathological and survival data from bladder cancer patients and then investigated the underlying molecular mechanisms. Bladder cancer tissues (n = 223) were subjected to immunohistochemistry, and tumour cell lines and nude mice were used for different in vitro and in vivo assays following QKI‐6 overexpression or knockdown. QKI‐6 down‐regulation was associated with advanced tumour TNM stages and poor patient overall survival. QKI‐6 overexpression inhibited bladder cancer cell growth and invasion capacity, but induced tumour cell apoptosis and cell cycle arrest. Furthermore, ectopic expression of QKI‐6 reduced tumour xenograft growth and expression of proliferation markers, Ki67 and PCNA. However, knockdown of QKI‐6 expression had opposite effects in vitro and in vivo. QKI‐6 inhibited expression of E2 transcription factor 3 (E2F3) by directly binding to the E2F3 3'‐UTR, whereas E2F3 induced QKI‐6 transcription by binding to the QKI‐6 promoter in negative feedback mechanism. QKI‐6 expression also suppressed activity and expression of nuclear factor‐κB (NF‐κB) signalling proteins in vitro, implying a novel multilevel regulatory network downstream of QKI‐6. In conclusion, QKI‐6 down‐regulation contributes to bladder cancer development and progression.


| INTRODUC TI ON
Bladder cancer remains a significant health problem in both men and women worldwide. 1 In China alone, the incidence of bladder cancer has increased in the past decade. 2 Approximately 70% of tumour lesions are non-muscle-invasive bladder cancer; however, these tumours can undergo recurrence and become invasive. 3 Thus, any tumour lesions in the urinary bladder should undergo long-term surveillance. Treatment of bladder cancer depends on the tumour stage. For example superficial tumour lesions can be surgically removed by a resectoscope 4 followed by immunotherapy with Bacillus Calmette-Guérin. 5 Bladder cancer with deep muscle invasion could be treated with cystectomy or radical cystectomy followed by radiation and chemotherapy, whereas micro-metastasized disease should be treated with neoadjuvant chemotherapy. 4 Risk factors for developing bladder cancer include tobacco smoking and occupational exposure to carcinogens (like benzidine). 6 These risk factors can alter gene expression and promote bladder carcinogenesis. 6 To date, despite improvements in surgical techniques and chemotherapeutic agents, more than 50% of bladder cancer patients still develop metastatic disease and suffer death. 7 Thus, there is an urgent need to identify biomarkers for early detection, prognosis and treatment responses and increase our understanding of the molecular mechanisms underlying bladder cancer pathogenesis.
Quaking homologue (QKI) belongs to the RNA-binding signal transduction and activator of RNA (STAR) protein family. 8 All QKI protein isoforms contain an identical GSG/STAR domain and possess identical RNA-binding specificities. 9 QKI-5 is the major nuclear isoform expressed during embryogenesis and its expression declines after birth, whereas QKI-6 and QKI-7 isoforms are expressed during late stages of embryogenesis with peak expression occurring during myelination. 10 However, these isoforms have a distinct C terminal as a result of alternative splicing of premature mRNA. 11 QKI proteins can selectively interact with a short DNA sequence in their target gene promoters, termed the QKI response element (QRE; 5'-ACUAAY…N 1-20 UAAY-3') of their targeting genes (ie QKI interactions with RNAs in the 3'-UTR via the STAR domain). However, the structure of the entire STAR domain with bound RNA remains unknown, especially in cancer cells. 12 A previous study showed that QKI directly regulates the expression of p27 and cyclinD1 to form a negative feedback loop with E2 transcription factor 1 (E2F1), indicating that QKI may be involved in cell cycle regulation and cell differentiation. 13 More recent studies demonstrated that QKI possesses tumour suppressor activity in various human cancers, including glioblastoma, and prostate, colon, lung, gastric, oral and kidney cancers. 9,11,[14][15][16][17][18] E2 transcription factor 3 (E2F3) is localized at chromosome 6p22, where gene amplification occurs in approximately 9% of bladder cancers and is associated with a higher bladder tumour stage and grade, suggesting that E2F3 is an oncogene. 19 Moreover, phosphorylation of the Rb protein permits uncomplexed E2F3 to promote cell cycle progression from the G1 to S phase in bladder cancer. 20 A previous study demonstrated that post-transcriptional regulation of E2F3 by miR-125b could suppress bladder cancer progression. 21 In addition, the NF-κB pathway involves phosphorylation, ubiquitination and degradation of IκBα and subsequent nuclear translocation of the p50 and p65 subunits of NF-κB, which facilitates target gene transcription. Thus, activation of NF-κB signalling is associated with cancer development, metastasis and chemical resistance, and abnormal constitutive activation of NF-κB signalling has been demonstrated in human cancers, including lung, prostate, breast and bladder cancers.
Mutations of various oncogenes in combination with an inflammatory microenvironment can lead to aberrant NF-κB activation. 22,23 In this study, we assayed pan-QKI expression in bladder cancer compared to normal tissue samples and determined the association of QKI-6 expression with clinicopathological data and patient survival. We then investigated the effects of QKI-6 overexpression or knockdown on bladder cancer cell malignant behaviours in vitro and in nude mouse xenografts. Our findings provide novel insight into the role of QKI-6 in bladder cancer development and progression and postulate that targeting the QKI-6-E2F3 interaction and NF-κB signalling pathway could serve as a therapeutic strategy to clinically control bladder cancer.

| Patients and tissue samples
This study was approved by the ethics committee of Shanghai General Hospital (Shanghai, China). In brief, we retrospectively collected tissue samples and the corresponding clinicopathological data (Table S1) from 223 patients with histologically diagnosed bladder cancer from our hospital (Shanghai, China). Written informed consent was obtained from all patients included in the study. These patients were diagnosed with bladder cancer according to cystoscopy criteria and staged with muscle-invasive classification. 6 Both tumour and matched adjacent normal tissues were collected and histologically confirmed by the Department of Pathology, Shanghai General Hospital. Paraffin blocks from each patient were retrieved from the Pathology Department and subjected to tissue microarray. These patients were followed for up to five years.

| Immunohistochemistry
Tissue microarray sections (4 μmol/L) were prepared for immunohistochemistry. Specifically, sections were deparaffinized in xylene and rehydrated in a series of ethanol solutions (100%-50%) and in tap water. Endogenous peroxidase activity was blocked in 0.75% H 2 O 2 in phosphate-buffered saline (PBS) for 50 minutes. The sections were then incubated in 5% bovine serum albumin (BSA) in PBS for 30 minutes at room temperature and incubated with a mouse anti-human QKI antibody (1:300; Sigma Chemicals, St. Louis, MO) or a mouse anti-human E2F3 antibody (1:500; Abcam, London, UK) at 4°C overnight. Next, immune detection followed a three-step protocol with a streptavidin-horseradish peroxidase complex and visualization by 3, 3-diaminobenzidine, according to a previous study. 14 The immunostained tissue microarray sections were reviewed and scored using a cut-off value of 30% as low versus high QKI-6 expression in these tissue specimens. Bladder cancer 5637, T24, 253J, RT4, TCCSUP and J82 cell

| Plasmid carrying QKI-6 cDNA or shRNA and lentiviral transfection
Plasmids carrying either QKI-6 shRNA or negative control shRNA were constructed and lentivirus was generated in HEK-293 cells using OBiO (Shanghai, China). The plasmid containing QKI-6 cDNA was also obtained from OBiO. To knockdown or overexpress QKI-6, bladder cancer cells were seeded at a density of 1 × 10 5 in 6-well plates and grown to approximately 75% confluency. Next, the culture medium was removed and fresh culture medium consisting of either lentiviral particles containing QKI-6 shRNA, negative control shRNA, QKI-6 cDNA or negative control was added using FuGENE 6 (Roche, Indianapolis, IN) according to the manufacturer's instructions. Cells were then further cultured in an incubator at 37°C with 5% CO 2 . After 24 hours transduction, the culture medium containing viral particles was removed and replaced with fresh medium containing an appropriate concentration of puromycin to promote growth to a sufficient cell number. The cell clones stably expressing QKI-6 shRNA and CMV-QKI-6 were selected and expanded.
Western blot and quantitative real-time polymerase chain reaction (qRT-PCR) analyses were used to evaluate infection efficiency. The positive clones were selected for in vivo experiments after 7 days.

| qRT-PCR
Total cellular RNA was isolated from cultured cells using Trizol reagent (Invitrogen) and reverse transcribed into cDNA using the M-MLV assay kit (Invitrogen) according to the manufacturer's instructions. These cDNA samples were subjected to qPCR amplification of different genes with their primer sequences (Table S1)

| Cell viability CCK-8 assay
To assess cell proliferation, we first seeded bladder cancer cells Experiments were conducted in triplicate and repeated at least three times.

| Tumour cell colony formation assay
Following QKI-6 cDNA or shRNA infection, cells were seeded in 6well plates at 500 cells/well and grown for 14 days. The cell growth medium was refreshed every three days, and at the end of the experiments cells were stained with 0.1% crystal violet for 10 minutes at room temperature. Cell colonies with more than 50 cells were counted under an inverted microscope (Leica, Wetzlar, Germany).
The experiment was conducted in triplicate and repeated at least once.

| Tumour cell transwell invasion assay
Following QKI-6 cDNA or shRNA infection, cells were seeded into the upper chambers of the Transwell (Corning, Corning, NY). The filter was precoated with Matrigel (BD Biosciences, San Jose, CA) in the growth medium without serum, whereas the bottom chamber was filled with growth medium containing 20% foetal calf serum.
The cells were grown for 24 hours. At the end of experiment, cells remaining on the top of the filter were removed using a cotton swab, and the tumour cells that invaded the low side of the filter were fixed with 10% formalin and stained with 0.1% crystal violet for 10 minutes at room temperature. The numbers of invaded cells were counted under an inverted microscope (Leica, Wetzlar, Germany).
The experiment was conducted in triplicate and repeated at least once.

| Flow cytometry cell cycle assay
To determine changes in cell cycle distribution, transfected cells were grown, washed twice with PBS, and then fixed for at least 2 hours in 300 μL of PBS and 700 μL of 100% ethanol. Next, cells were centrifuged, re-suspended in 200 μL of extraction buffer (0.1% Triton X-100, 45 mmol/L Na 2 HPO 4 , and 2.5 mmol/L sodium citrate), incubated at 37°C for 20 minutes, and then re-suspended and incubated in PBS containing 40 mg/mL propidium iodide, 0.1 mg/mL RNase (Sigma Chemicals) and 0.1% Triton X-100 at 37°C for 30 minutes in the dark.
Cell cycle distributions were analysed using the Fluorescence-activated cell sorting Accuri C6 flow cytometer (BD Biosciences, Franklin Lakes, NJ).

| Animal experiments
All animal experiments were conducted in accordance with the US Public Health Service Policy on Humane Care and Use of Laboratory Animals and were approved by the Scientific Investigation Board of Shanghai General Hospital. In brief, male nude mice (4-6 weeks old) were purchased from the Experimental Animal Center of Shanghai General Hospital. Tumour cells were grown and suspended in sterile PBS and then injected subcutaneously into each nude mouse (1 × 10 6 cells/mouse). The size and incidence of subcutaneous tumour xenografts were recorded every 7 days using the following formula: V (mm 3 ) = width 2 (mm 2 ) × length (mm)/2. After 42 days, the mice were killed by CO 2 and cervical dislocation to evaluate tumour incidence, weight and size, as well as immunostaining at the indicated time-points.

| Immunofluorescence
Bladder cancer T24 and 5637 cells were grown on coverslips overnight, washed with PBS, and then fixed in 4% formaldehyde solution for 20 minutes. For immunostaining, cells were permeabilized in 0.1% Triton X-100 for 15 minutes and then blocked in 5% normal goat serum (1:5) in PBS for 1 hour. Next, cells were incubated with a rabbit anti-QKI antibody (Sigma, Chemicals) at a dilution of 1:500 or antibodies for other regulatory proteins at room temperature for 30 minutes. Cells were washed with PBS three times and further incubated with DyLight 594 and 488-conjugated goat antirabbit or antimouse IgG (Thermo Scientific) at a dilution of 1:1000 at room temperature for 30 minutes and then counterstained with 4',6-diamidino-2-phenylindole (DAPI; Sigma Chemicals). Staining was scored under an Olympus IX71 fluorescence microscope (Olympus, Tokyo, Japan) and cell images were captured using the microscopeequipped CellSens imaging software.

| Electrophoretic mobility shift assay
Electrophoretic mobility shift assay was performed according to a previous study. 25 Briefly, oligonucleotide probes with a biotin tag at the 5'-end of the sequence (Integrated DNA Technologies) were incubated with HEK293T nuclear protein and the working reagent from the Gel shift Chemi-luminescent EMSA kit (Active Motif 37341). The wild-type E2F3 EMSA probe sequences were 5'-GGA ATA CTA ATA AGT CTT AAA AGT TC-3' and the mutant E2F3 EMSA probe sequences were 5'-GGA ATC TGC CAA GTC TGC CCA GTT C-3'. For competitor assays, an unlabelled probe was added to the reaction mixture at 100× excess. The reaction was then incubated for 30 minutes at room temperature and then loaded onto a 6% retardation gel (Thermo Fisher Scientific EC6365BOX) and run in 0.5× TBE buffer. After transfer onto a nylon membrane, the membrane was visualized with the chemiluminescent reagent as recommended.
The super shift assay was performed with 1 μg anti-QKI-6 antibody (Millipore) and incubated on ice with protein from HEK 293T for 30 minutes prior to addition of oligonucleotide probes and gel electrophoresis.

| Luciferase assay
cDNA fragments with the human QKI-6 promoter covering +1 to −2000 bp relative to the transcription initiation site were amplified using PCR primers (Table S1)

| Statistical analysis
All statistical analyses were performed with spss 17.0 software (SPSS Inc, Chicago, IL). The Pearson chi-square (χ2) test was used to correlate QKI-6 expression with clinicopathological data, whereas the Kaplan-Meier curves and Log rank test were used to analyse overall survival stratified by QKI-6 expression in bladder cancer patients. For in vitro and nude mouse results, data are expressed as mean ± SD and analysed using a one-way ANOVA. A P < 0.05 was considered statistically significant.

| QKI-6 down-regulation in bladder cancer tissues and cell lines
We first assessed QKI expression in bladder cancer tissues compared to normal tissues and bladder cancer cell lines. We found that QKI-6 expression was significantly down-regulated in bladder cancer tissues compared to the matched adjacent normal tissues in 223 patients ( Figure 1A,B). Western blot and qRT-PCR analyses showed that QKI-6 expression was also lower in bladder cancer cell lines ( Figure 1C,D).

| Association of QKI-6 expression with clinicopathological features and survival of bladder cancer patients
We next assessed whether QKI-6 expression was associated with clinicopathological features of cancer patients and found that the absence of QKI-6 expression was associated with advanced pathology stage, tumour TNM stage and depth of invasion (P < 0.0001; Table 1), but there was no association with other clinicopathological features, such as age or gender (P > 0.05; Table 1). We then correlated QKI-6 expression with overall patient survival and found that reduced QKI-6 expression was significantly associated with poor overall survival (Figure 2).

| QKI-6 inhibition of bladder cancer cell growth and invasion capacity in vitro
Our current ex vivo data demonstrated that QKI-6 expression was

| QKI-6 promotion of tumour cell apoptosis and cell cycle arrest in vitro
Next, we determined the effects of QKI-6 overexpression or knockdown on tumour cell apoptosis and cell cycle distribution. We found that the percentage of apoptotic 5637 cells was significantly lower after QKI-6 knockdown, whereas the percentage of apoptotic T24 cells was significantly higher after QKI-6 cDNA transfection compared to control cells ( Figure 4A,B). Western blot data further showed that expression of caspase-3, caspase-9 caspase-12, and PARP was up-or down-regulated in these 5637 and T24 cells respectively ( Figure 4C,D), which was also confirmed by immunofluorescence ( Figure 4E-H). Moreover, QKI-6 knockdown and up-regulation in 5637 and T24 cells induced and reduced, respectively, the proportion of cells in S phase compared to the control group ( Figure S2A,B).
Western blot data showed that cyclin D1, cyclin B, cyclin E1 and cyclin A1 expression were up-regulated, whereas levels of p27 and p21 expression were down-regulated in QKI-6-knocked down 5637 cells compared to the negative control cells. In contrast, QKI-6 overexpression in T24 cells inversely affected the tumour cells ( Figure   S2C,D). These findings suggest that QKI-6 expression promotes cell apoptosis and induces cell cycle arrest in the G2/M phase of the cell cycle.

| Anti-tumour activity of QKI-6 in vivo
To confirm the inhibitory effects of QKI in vivo, we infected bladder cancer 5637 cells with shRNA-NC or shRNA-QKI-6 lentivirus particles and T24 cells with lentivirus particles carrying vector only or QKI-6 cDNA. Transfection efficiency was verified, as shown in Figure 3A. We then subcutaneously injected these cells into nude mice and monitored tumour cell xenograft formation and growth.
We found a significant difference in tumour cell xenograft formation growth and weight ( Figure 5). Specifically, the size of tumour xenografts was larger in the shRNA-QKI-6 5637 cells compared to the negative control shRNA xenografts, whereas the tumour xenografts were consistently smaller in vector-only T24 cells compared to those of CMV-QKI-6 T24 cells ( Figure 5A,B). Immunohistochemical data showed that PCNA and Ki-67 expression were significantly up-regulated following QKI-6 knockdown in 5637 cells compared F I G U R E 1 Quaking homologue (QKI) expression is down-regulated in bladder cancer tissues and cell lines. (A), Immunohistochemistry. QKI protein expression was analysed in bladder cancer and adjacent normal tissues using immunohistochemistry. (B), Immunofluorescence. QKI protein expression was assessed in bladder cancer and adjacent normal tissues using immunofluorescence (200×). (C,D) Western blot and qRT-PCR. QKI mRNA and protein expression was analysed in different bladder cancer and normal cell lines using Western blot and qRT-PCR respectively to control cells ( Figure 5G,I). However, PCNA and Ki-67 expression were reduced in tumour cell xenografts injected with QKI-6 overexpressing T24 cells ( Figure 5H,J). Immunofluorescence staining of tissue samples showed down-regulation of Ki-67 and PCNA levels but up-regulation of PARP levels in T24 QKI-6 overexpressing tumour cell xenografts, whereas knockdown of QKI-6 expression had the opposite effects ( Figure S3A-E). These results suggest that QKI-6 suppresses tumour growth in vivo.

| QKI reduces E2F3 mRNA stability by directly binding to the E2F3 3'-untranslated region
As an RNA-binding protein, QKI post-transcriptionally regulates expression of target mRNAs. 15,18 We performed a bioinformatic analysis and found two potential QREs in the E2F3 3'-untranslated region (3'-UTR). We thereafter assessed whether QKI-6 overexpression or knockdown affects E2F3 expression using Western blot and qRT-PCR analyses (Figure 6-10A-C). Double-immunofluorescence labelling showed that QKI-6 co-localized with E2F3 and inhibited E2F3 expression and nuclear translocation ( Figure 6D). Our data showed that QKI-6 overexpression or knockdown modulated the stability of E2F3 mRNA after cells were treated with actinomycin D, an inhibitor of gene transcription ( Figure 6E,F). To further validate whether QKI-6 directly regulates E2F3, we constructed plasmids carrying the wild-type or mutant E2F3 3'-UTR ( Figure 6G) and performed a dual luciferase assay. We found that QKI-6 overexpression reduced the luciferase activity compared to the mutated controls ( Figure 6H).
Furthermore, EMSA demonstrated that QKI-6 bound to the E2F3 mRNA and generated a super-shift band ( Figure 6I). These results suggest that QKI-6 decreases stability of E2F3 mRNA by directly binding to the E2F3 3'-UTR.

| E2F3 modulation of QKI-6 transcription by binding to its promoter
As a transcription factor, E2F3 has an established role in regulating cell cycle progression. A recent study reported that amplification and overexpression of E2F3 resulted in oncogenic activity in human bladder cancer development. 26 E2F3 expression can be regulated by binding of promoter regions in target genes. 27 We therefore performed immunohistochemistry and immunofluorescence analyses and found that QKI-6 expression was inversely associated with E2F3 expression in bladder cancer tissue specimens ( Figure 7A,B), whereas E2F3 expression was inversely associated the QKI-6 mRNA and protein expression (Figure 7-10C-F).
Furthermore, we cloned the QKI-6 promoter region (−2000 to +1 bp) containing the E2F3-binding site and the mutant gene without the E2F3-binding site ( Figure 7G). We found that E2F3 enhanced QKI-6 promoter luciferase activity ( Figure 7H). The Chromatin immunoprecipitation (ChIP) assay also demonstrated that E2F3 directly binds to the QKI-6 promoter ( Figure 7I). This effect was more obvious in QKI-6 up-regulated T24 cells, but the luciferase activity was reduced in QKI-6 down-regulated 5637 cells ( Figure 7J). Our RNA pull-down assay of 5637 and T24 cell lysates with E2F3 3'-UTR showed a specific interaction with QKI-6 ( Figure 7K). These results illustrate that E2F3 directly affects transcription of QKI-6, forming a negative feedback loop.

| QKI-6 inhibits activity and expression of NF-κB signalling proteins
Transactivation of NF-κB can be initiated by a vast array of stimuli that have different biological activities, such as inflammation, immunity, differentiation, cell growth, tumorigenesis and apoptosis. [28][29][30] A previous study reported that NF-κB was frequently activated in bladder cancer. 31,32 We also have preliminary evidence demonstrating that QKI-6 inhibits bladder tumorigenesis (data not shown). In this study, we measured NF-κB expression and activity in QKI-6-knockdown or overexpressing bladder cancer cells. We found that QKI-6 knockdown in 5637 cells up-regulated the expression of phosphorylated (p)-NF-κβ, Ikk-α, Ikk-β and Ikb-α, whereas QKI-6 overexpression in T24 cells reduced expression of all of these proteins ( Figure 8A,B). Double-labelling (E-G), Immunofluorescence. Bladder cancer cell lines 5637 and T24 were infected with lentivirus carrying QKI shRNA, QKI cDNA or their negative controls, respectively, and then subjected to immunofluorescence analysis of cleaved caspase-12, caspase-9 and caspase-3 (400×).
(H), ImageJ software quantitation of E-G. Data are presented as the mean ± SD following one-way ANOVA analysis for three independent experiments. *P < 0.05 immunofluorescence staining showed that QKI-6 co-localized with p-NF-κB and inhibited p-NF-κB expression and nuclear translocation ( Figure 8C). The expression of p-NF-κB was significantly inhibited by different concentrations of PDTC, whereas QKI was not affected (Figure 8D,E). This result demonstrates that QKI is upstream of NF-κB and that QKI down-regulates NF-κB pathway activation.

| Effects of the NF-κB inhibitor PDTC on regulating tumour cell proliferation, apoptosis, cell cycle arrest and invasion
We found that 20 µmol of PDTC treatment for 6 hours significantly down-regulated p-NF-κB in both the T24 vector and QKI-6 group and in the 5637 sh-NC and sh-QKI-6 group ( Figure 9A,B). However,

| D ISCUSS I ON
Recently, altered QKI expression was implicated in various human cancers, including colorectal cancer, 33 paediatric brain tumours, 34 glioma, 35 renal clear cell carcinoma, 9 lung cancer 17 and prostate cancer. 18 In this study, we measured QKI-6 expression and determined its role in regulating bladder cancer in vitro and in vivo, and we also investigated the molecular mechanism underlying QKI-6's effect in bladder cancer cells. We found that QKI-6 expression was reduced in bladder cancer tissues and that QKI-6 down-regulation was associated with advanced bladder cancer TNM stage and poor patient overall survival. We also revealed that QKI-6 inhibits bladder  showing that QKI suppresses colon and oral tumorigenicity 11,16 and prostate cancer. 18 Although there are several QKI isoforms, 9,10 our current findings showed that QKI-6 was significantly downregulated in bladder cancer tissues and cell lines compared to other QKI isoforms.
To assess the effects of QKI-6 in bladder cancer, we knocked Cell proliferation was significantly decreased by the QKI-6 plasmid and enhanced by the sh-QKI-6 plasmid, whereas there was no significant difference after PDTC treatment. (I,J) Invasion capacity was also abolished by PDTC treatment in T24 and 5637 cells. Each assay was repeated at least three times. *P < 0.05 data in human cancer are limited, our current data further support a role for QKI-6 in different human cancers. 9 13 Other previous studies revealed that QKI could either increase or decrease the stability of target mRNAs. 13,16,38 In our current study, we treated bladder cancer cells after QKI-6 knockdown or overexpression with a gene transcription inhibitor, actinomycin D, and found that QKI-6 destabilized and reduced E2F3 mRNA and protein respectively. Furthermore, our EMSA and ChIP assays confirmed the direct binding of QKI-6 to the E2F3 mRNA 3'-UTR. Taken together, our current data revealed that QKI-6 can directly regulate E2F3, whereas QKI-6 itself was also regulated by E2F3, suggesting that QKI-6 and E2F3 form a negative feedback loop. However, the underlying mechanism of this feedback loop in bladder cancer progression requires further investigation.
A previous study reported that QKI deficiency induced inflammation in experimental endotoxemia via increasing NF-κB signalling, 39 whereas another recent study showed that NF-κB repression of QKI expression promotes cancer stem cell-like properties during the neoplastic transformation of hepatic cells in response to arsenite. 40 These two studies clearly indicate that reduced QKI-6 expression in bladder cancer tissues or cells activates the NF-κB signal pathway, which might play a key role in cancer initiation, especially in bladder cancer. 28,29 Our current data showed that up-regulated QKI-6 expression reduced the activity and expression of NF-κB signalling proteins in bladder cancer. 31,32 In conclusion, we demonstrated that QKI-6 down-regulation in bladder cancer tissues and cells promoted tumour cell malignant behaviours by down-regulating E2F3 and NF-κB signalling pathways ( Figure 9). In contrast, QKI-6 overexpression inhibited bladder cancer cell proliferation, cell cycle progression and invasion, but promoted cell apoptosis and cell cycle arrest at the G2/M phase. Furthermore, F I G U R E 1 0 Schematic illustration of the quaking homolog (QKI)-6 inhibited bladder cancer progression. Schematic representation of the function of QKI in coordinately targeting multiple cell cycle, cell apoptosis, cell differentiation and cell proliferation regulators in bladder cancer. QKI-6 targets the NF-κβ signalling pathways, which, subsequently results in cell cycle arrest, cell differentiation, cell apoptosis, decreased proliferation. Activated E2 transcription factor 3 (E2F3) transcribes QKI, which represses the activity of E2F3 reduced QKI-6 expression was associated with advanced bladder cancer TNM stage and poor survival of patients. Future studies are needed to clarify the specific mechanism by which QKI-6 regulates E2F3 and NF-κB signalling in bladder cancer development and progression. Specifically, it would be important to investigate if restoring QKI-6 expression could be a novel strategy to restrict bladder cancer.

ACK N OWLED G EM ENTS
This study was supported in part by grants from the National Natural Science Foundation of China (#81570682) and the Youth Fund Project by National Natural Science Foundation (#81602252).

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest. SJX and BMH revised the manuscript critically for intellectual content. All authors provided intellectual input and approved the final version of the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated and analysed during this study are available from the corresponding author on reasonable request.