Nuclear Factor-κB-Mediated Transforming Growth Factor-β-Induced Expression of Vimentin Is an Independent Predictor of Biochemical Recurrence after Radical Prostatectomy

Prostate cancer is the most common malignancy in American males and is second only to lung cancer as the leading cause of cancer-specific mortality (1). Metastases are responsible for mortality in prostate cancer patients. Post-treatment serum prostate-specific antigen (PSA) values have been used to identify patients at risk for metastases. In fact, depending on the definition used, the 5-year PSA recurrence rates following either radical prostatectomy or radiation therapy were reported to be up to 31%, and the 10-year clinical recurrence rates in these patients were reported to be ∼75% (2–4). A short time interval to biochemical recurrence, rapid PSA doubling times, and high Gleason scores are all considered high-risk factors for prostate cancer-specific mortality (5, 6). However, these clinical characteristics have not been proven to be useful predictors of clinical outcome in patients with low-grade disease (Gleason score ≤6; refs. 7–9). Therefore, additional biomarkers are needed to predict recurrence after radical prostatectomy.

Biomarkers expressed in each tumor reflect the clinical signature of the disease and may determine the clinical course. Although other methods of metastases have recently been described, the epithelial-to-mesenchymal transition (EMT) has been considered one of the major mechanisms mediating invasion and metastasis of cancer cells (10). The role of transforming growth factor-β (TGF-β) signaling in EMT has been established for many normal and transformed cell lines (11, 12). Recently, the expression of vimentin has been linked to the EMT in prostate cancer and metastasis in vitro (13). Similarly, increased vimentin expression levels have been associated with poorly differentiated prostate cancer (14). Although TGF-β is implicated in invasion in prostate cancer cell invasion by up-regulating vimentin expression (15), the precise mechanism and downstream mediators remain incompletely defined (16). From an intracellular signaling perspective, TGF-β may induce the EMT and invasion via a Smad-dependent pathway by activating Jag1 and Hey1 (17). There is evidence to indicate that other intracellular signaling pathways also play a role in EMT (16). For example, nuclear factor-κB (NF-κB) may be activated by TGF-β signaling through activation of TAK1 (11). Also, Ras may act with TGF-β to activate NF-κB (18). However, although many of these TGF-β-interacting proteins have been associated with the EMT and invasion, they have not specifically been implicated in prostate cancer metastasis. Therefore, the purpose of the present study is 2-fold: (a) to elucidate the mechanisms underlying TGF-β-induced EMT and (b) to explore factors involved in this pathway as potential markers for evaluating biochemical recurrence in prostate cancer patients following radical prostatectomy.

Prostate cancer cell lines. Four variants of human prostate cancer PC-3 cell lines were kindly provided by Drs. Fidler and Pettaway of the M. D. Anderson Cancer Center (19). These variants included PC-3 (wild type), PC-3M-Pr04 (selected for cancer cells confined to the prostate), PC-3M (selected for metastasis), and PC-3M-LN4 (selected for metastasis to lymph nodes). These variants are ideal for studying the role of TGF-β and NF-κB in tumor migration and invasion. PC-3 and PC-3M-Pr04 are less aggressive than PC-3M and PC-3M-LN4. As a negative control, cells were rendered insensitive to TGF-β by introducing a dominant-negative TGF-β type II receptor vector (TβRIIDN; ref. 20; Fig. 1A, top right). The infection efficiency was >95.9% for the TβRIIDN vector. Cells were cultured in RPMI 1640 (Life Technologies) supplemented with 10% heat-inactivated fetal bovine serum (Life Technologies). Cells were treated with or without TGF-β (10 ng/mL, 24 h). A NF-κB inhibitor, BAY11-7085 (E)-3-[(4-t-butylphenylsulfony1)]-2-propenenitrile (BIOMOL Research Laboratories), was used at 20 μmol/L, 24 h.

TGF-β1 ELISA. All variants (PC-3, PC-3M-Pr04, PC-3M-LN4, and PC-3M) and the corresponding TβRIIDN-transfected cell lines were cultured in fresh serum-free medium for 24 h (1.0 × 10⁷ per T75 flask). The pooled conditioned medium was collected and concentrated by using YM-3 Centriprep Centrifugal Filter Devices (Millipore). TGF-β1 ELISA was carried out using the Quantikine Human TGF-β1 Immunoassay Kit from R&D Systems. The total number of cells in each flask was counted using a Coulter counter and the levels of TGF-β1 were reported as pg/10⁵ cells/48 h.

[³H]thymidine incorporation assay. All cells were grown in culture for 48 h. Cells were then exposed to a medium containing [³H]thymidine (0.5 μCi/mL; Amersham Biosciences) for an additional 5 h. The experiment was terminated by washing with warm serum-free medium. NaOH (0.1 mol/L) was added to all cell culture wells (1 mL). An aliquot of 100 μL was removed for measurement of the protein content and the remainder was used for the determination of the radioactivity as in counts/min. Thymidine incorporation was expressed as the fraction of counts found in cells of untreated controls.

Western blot analysis. Cell lysates were prepared by adding lysis buffer (50 mmol/L Tris-HCl, 1% NP-40, 0.25% sodium deoxycholate, 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L Na₃VO₄, and 1 mmol/L NaF) to cell pellets. An aliquot of ∼30 μg total protein was loaded onto 10% acrylamide gel in Tris-HCl (Bio-Rad). Electrophoresis was carried out in Tris-glycine-SDS running buffer and transferred to a polyvinylidene difluoride membrane. Blots were probed for the total and phosphorylated Smad2 (Santa Cruz Biotechnology) at 1:500, vimentin at 1:500 (DAKO), and cytokeratin 18 (CK18; DAKO) at 1:500 and then stripped and reprobed for glyceraldehyde-3-phosphate dehydrogenase (Advanced) with a monoclonal antibody at 1:300. Proteins of interest were detected with the enhanced chemiluminescence detection kit (Amersham Biosciences) followed by exposure to Kodak X-OMAT AR film.

Immunofluorescence and microscopy. Immunofluorescent studies were done on all PC-3 variants as described previously (21). For colocalization of vimentin and NF-κB, cells were analyzed by using nucleus-vimentin-NF-κB triple staining. For colocalization of TβRIIDN (with GFP) and vimentin or NF-κB, cells were analyzed by nucleus-vimentin or nucleus-NF-κB double staining. All slides were deparaffinized and blocked by normal serum. Stained slides were viewed with Nikon TE2000-U fluorescent microscopy (Nikon). Images were digitized by Photoshop 7.0 with a PC computer.

Cell invasion assay. Cell invasion assay (Matrigel invasion assay) was done in a 24-well Transwell chamber (8 μm pore size; CytoSelect; Cell Biolabs). Cells were plated at a density of 0.5 × 10⁶ to 1.0 × 10⁶/mL in serum-free medium. TGF-β1 and/or BAY11-7085 were added directly to the cell suspension, and 24 h later, the suspension was aspirated and the invaded cells were counted with a light microscope under high magnification objective (×100; Olympus) and measured at A_{560 nm} in a plate reader after treatment with the extraction solution.

Construction of tissue microarrays and clinical outcome assessment. A series of prostate tissue microarray was constructed with formalin-fixed, paraffin-embedded prostatectomy specimens as described previously (22). At least three cores were taken from each donor tissue block. A total of 287 radical prostatectomy specimens were obtained from the Prostate SPORE Tissue Bank at Northwestern University. All patients (n = 287) for whom there were tissue microarray samples were included in the study. Power calculation using the pwr.r.test function of PWR package in R showed that a sample size of 287 patients provides 99% power to detect a correlation with correlation coefficient of 0.25 (or -0.25) at the significance level of 0.05.

All enrolled subjects provided written informed consent and the study was approved by the institutional review board of Northwestern University. These specimens included 165 cases with low Gleason score (≤6), 55 cases with intermediate Gleason score (7), and 67 cases with high Gleason score (≥8; Table 1). A retrospective analysis of outcome assessment was done based on clinical information linked to the tissue microarray specimens. None of the patients included in this study underwent additional adjuvant therapies (e.g., hormone and radiation). Biochemical recurrence was defined as a postoperative increase in serum PSA >0.2 ng/mL. All patients had at least 1 year follow-up.

Immunohistochemistry. All antibodies were first tested and optimized on whole-tissue sections and test arrays. Once an appropriate dilution was determined, a set of three slides containing all patient samples were stained for each antibody using standard two-step indirect immunohistochemistry. After deparaffinization in xylenes, the tissue array sections were rehydrated in graded alcohols. Endogenous peroxidase was quenched with 3% hydrogen peroxide in methanol at room temperature (25°C). The sections were placed in a 95°C solution of 0.01 mol/L sodium citrate (pH 6.0) for antigen retrieval. Normal goat serum (5%) was applied for 30 min to block nonspecific protein-binding sites. Primary mouse monoclonal antibodies were applied for 30 min at room temperature at the following dilutions: TGF-β1 at 1:200 (DAKO), TβRI (TGF-β type I receptor; Santa Cruz Biotechnology) at 1:100, TβRII (TGF-β type II receptor; Upstate) at 1:200, p-Smad2 (Santa Cruz Biotechnology) at 1:400, vimentin (DAKO) at 1:500, CK18 (DAKO) at 1:500, and NF-κB at 1:200. Detection was accomplished with the DAKO Envision System followed by chromogen detection with diaminobenzidine. Sections were counterstained with Harris's hematoxylin followed by dehydration and mounting. Negative controls were done using identical array sections stained in the absence of the primary antibody. Semiquantitative assessment of all tissue microarrays was done independently by two pathologists (M.P. and X.J.Y.) who were blinded to all clinical information. The frequency of nuclear-positive cells (range, 0-100%) in prostatic glandular epithelium was scored for each tissue microarray core. Results of immunohistochemical staining were expressed as the percentage of cancer cells that stained positive for a given antibody and were scored on a scale of 0 to 3: 0 (<20% cell staining), 1 (20-49% cell staining), 2 (50-74% cell staining), and 3 (75-100% cell staining; refs. 23, 24).

Statistical analysis. Numerical data were expressed as mean ± SD. ANOVA and multiple range test were done to determine differences of means among treatment groups. P < 0.05 was considered as statistically significant. The SPSS 10.0.7 software package (SPSS) was used for the analysis. Kaplan-Meier survival curves were analyzed by the log-rank test using R, a statistical computing language and environment. To investigate an association with biochemical recurrence, the log-rank test was first used to determine whether various markers showed any significant effect on biochemical recurrence, which was defined as PSA > 0.2 ng/mL (25). We tested variables, including all tissue microarray markers, and TGF-β1, clinical stage, clinical (pathologic) Gleason score grouped as 4-6, 7, and 8-10, surgical margin status, and patient age. Clinical stages were divided into patients with T₁ disease or patients with ≥T₂ disease. Age was stratified as follows: <60, 60 to 69, and ≥70. tissue microarray Gleason score was averaged across all the tissue microarrays and divided into three: groups 4-6, 7, and 8-10. All markers were scored with a scale of 0, 1, 2, or 3 before performing recurrence analysis (23, 24).

NF-κB is intermediate mediator for TGF-β-induced EMT, which confers invasive potential in human prostate cancer cells. Following treatment with TGF-β1, the overall morphology of all PC-3 variants changed. We found that cells treated with TGF-β1 show elongated cell processes and have reduced numbers of cell-cell contact when compared with the untreated cells. Both changes are characteristic of EMT (26). After rendering cells insensitive to TGF-β by infection with a dominant-negative construct, TβRIIDN, or with a NF-κB inhibitor, BAY11-7085, they regained cell-cell contact and decreased the number of elongated fibroblastic processes (Fig. 1A, left). Western blot showed that, after infection of TβRIIDN, the activity of p-Smad2 of all PC-3, PC-3M-Pr04, PC-3M-LN4, and PC-3M was down-regulated significantly (Fig. 1B). Compared with PC-3 and PC-3M-Pr04, PC-3M and PC-3M-LN4 secreted higher baseline levels of TGF-β1 and showed increased invasive potential (Supplementary Fig. S1A). Although there was no significant difference in the original proliferation rate among PC-3 variants, PC-3 and PC-3M-Pr04 cells were less aggressive than PC-3M and PC-3M-LN4 and more sensitive to growth inhibition (>30%) by TGF-β1 by [³H]thymidine incorporation assay. In contrast, PC-3M-LN4 and PC-3M were only inhibited by TGF-β1 by <7.5%. The growth-inhibitory effect of TGF-β1 disappeared when these cells were rendered insensitive to TGF-β by infection with TβRIIDN (Supplementary Fig. S1B).

There was a dose-dependent relationship between TGF-β1 exposure and the expression of vimentin (positive correlation) and CK18 (negative correlation; Fig. 1C). Untreated PC-3M and PC-3M-LN4 cells constitutively express high levels of vimentin but down-regulated or devoid of CK18 expression in comparison with PC-3. In contrast, the untreated PC-3 and PC-3M-Pr04 expressed relatively low levels of vimentin and relatively high levels of CK18 (Fig. 1D; Supplementary Fig. S2). In addition, the level of NF-κB expression in these cells was significantly increased following treatment with TGF-β1 (Fig. 2). However, following transfection with TβRIIDN, the level of vimentin expression (Fig. 1D) and NF-κB (Fig. 2) was significantly reduced, whereas CK18 increased simultaneously. This was observed in all cell lines, except for PC-3M-LN4, the most aggressive among the PC-3 variants. Interestingly, on treatment with NF-κB inhibitor, the effect of TGF-β1 on EMT was abrogated (Fig. 1D). The role of NF-κB in TGF-β-induced EMT was further supported by coexpression of vimentin and NF-κB in all PC-3 cell lines. Only those cells expressing NF-κB showed expression of vimentin. As expected, when cells were transfected by TβRIIDN, the level of expression of both vimentin and NF-κB was reduced significantly (Fig. 3A).

Using a Transwell chamber assay, before TGF-β1 treatment, PC-3M-LN4 and PC-3M had a higher invasive potential than that for PC-3 and PC-3M-Pr04 (Fig. 3B-D), which may due to the higher baseline level of TGF-β1 secreted by PC-3M-LN4 and PC-3M (Fig. 1B, top). With the treatment of TGF-β1, there was a further increase in invasion in each group. Invasion of all PC-3 variants was inhibited by transfection with TβRIIDN or by NF-κB inhibitor. Importantly, the inhibition of invasion by the NF-κB inhibitor could not be reversed by TGF-β1 treatment (Fig. 3B-D). This observation suggested that NF-κB was downstream of TGF-β, and NF-κB and TGF-β could synergistically mediate EMT and invasion in PC-3 cells. Taken together, these results are consistent with the notion that NF-κB is a mediator for the TGF-β-induced EMT, leading to an invasive phenotype in PC-3 variants.

TGF-β-induced EMT correlates with clinical characteristics. To evaluate the association of TGF-β with the induction of the EMT in prostate cancer specimens, we performed immunohistochemistry on tissue microarrays. Expression levels of TGF-β signaling components including TGF-β1, TβRI, TβRII, p-Smad2, and EMT-associated factors, such as vimentin, CK18, and NF-κB, were determined using specific antibodies. Using tissue microarray specimens, we found that a high level of expression of TGF-β1 and EMT-related factors coupled with a low level of expression of TβRI, TβRII, and p-Smad2 was associated with adverse pathologic features, such as increased Gleason grade (Fig. 4A). For instance, high levels of TGF-β1 expression (75-100% positively stained cancer cells) were identified in 36.7%, 6.7%, and 6.0% for tumors with high (≥8), intermediate (7), and low (≤6) Gleason scores, respectively. High levels of NF-κB expression were found in 60.6%, 10%, and 5.2%, vimentin expression in 23.9%, 3.6%, and 1.8%, and CK18 expression in 47.1%, 67.2%, and 68% of high, intermediate, and low Gleason score cancers, respectively (Fig. 4B). TGF-β signaling components were directly correlated with increasing Gleason grade. The regression coefficient (R) and the significance level (P) for these parameters were as follows: TGF-β1 (R = 0.423, P = 1.152411e-13), NF-κB (R = 0.512, P = 2.2e-16), and vimentin (R = 0.375, P = 1.45e-10). TβRI (R = -0.285, P = 1.14), TβRII (R = -0.14, P = 0.019), p-Smad2 (R = -0.181, P = 0.034), and CK18 (R = -0.221, P = 0.015) were negatively correlated with increasing Gleason score (Supplementary Table S1). Correspondingly, we found decreased expression of TβRI and TβRII in PC-3M and PC-3M-LN4, which are more aggressive phenotypes, compared with PC-3 and PC-3M-Pr04, which are less aggressive phenotypes (Supplementary Fig. S3). Because of a relatively low degree of correlation between the level of TGF-β1 expression and that of p-Smad2 activity but a relatively high degree of correlation between TGF-β1 and expression of EMT factors, we postulated that a Smad-independent pathway was probably responsible for the TGF-β-induced EMT.

Using tissue microarray specimens, we also found a relationship between levels of expression of TGF-β1 and EMT-related factors. In all tissue microarrays analyzed, there was a significant correlation between the expression of TGF-β1 and vimentin (R = 0.284, P = 2.38e-06), TGF-β1 and NF-κB (R = 0.313, P = 4.27e-07), vimentin and NF-κB (R = 0.429, P = 3.04e-12), and vimentin and CK18 (R = -0.164, P = 0.010; Supplementary Table S1). Taken together, these results suggest that the level of TGF-β expression is significantly associated with the expression levels of NF-κB and many other EMT-associated proteins.

We found that there was a correlation between age and expression of TGF-β1 (R = 0.204, P = 0.002206915) and CK18 (R = -0.305, P = 9.981004e-06), respectively. Unexpectedly, we found a positive correlation between patient age and expression of TGF-β1 (R = 0.204, P = 0.002) but a negative correlation between patient age and expression of CK18 (R = -0.305, P = 9.98e-06). Prostate cancer specimens of older patients (ages ≥70 years) showed a higher level of expression of TGF-β1 but a lower level of CK18, a condition conducive for EMT induction. Another finding of interest was that levels of expression of the above tumor markers correlated significantly with the percentage of cancer in the prostate, which was calculated using the total amount of tumor in the frozen tissue as the percentage of the entire prostate. The expression level of TGF-β1 (R = 0.295, P = 0.007), NF-κB (R = 0.393, P = 0.001), vimentin (R = 0.445, P = 3.19e-05), and CK18 (R = -0.362, P = 0.003) in the tumor all correlated significantly with the percentage of cancer (Supplementary Table S1). These data support the concept the induction of EMT is TGF-β-dependent but Smad-independent in prostate cancer.

NF-κB-mediated TGF-β-induced EMT predicts biochemical recurrence in prostate cancer patients. We next compared the clinical outcomes to the above-mentioned markers. Expression levels of TGF-β1 (P = 0.0158), vimentin (P = 0.004), CK18 (P = 0.006), NF-κB (P = 0.030), patient age (P = 1.99e-05), pathologic stage (P = 6.5e-07), and clinical Gleason score (P = 0.005) all correlated significantly with biochemical recurrence, defined as a PSA >0.2 ng/mL after radical prostatectomy (Fig. 5A; Supplementary Table S2). However, clinical stage, surgical margin status, and expression of TβRI, TβRII, and p-Smad2 did not correlate with biochemical recurrence (P > 0.05). A Kaplan-Meier curve was generated for each of the above significant variables. To determine the best model for predicting biochemical recurrence, Cox proportional hazards model was fit to include all the significant variables (Fig. 5A) and backward selection method was used to eliminate nonsignificant variables. The final selected model included vimentin, grouped as scale <3 or 3 (log-rank P = 0.049; hazard ratio, 2.03; 95% confidence interval, 1-4.12), and pathologic stage grouped as <T_3a and ≥T_3a (log-rank P = 0.031; hazard ratio, 10.1; 95% confidence interval, 1.23-82.8; Fig. 5B and C; Supplementary Table S3). Tumors with a vimentin expression score of 3 had a 2.03 higher risk of recurrence than patients with lower scores of vimentin expression in the tumor. There was a high risk of recurrence if high levels of expression of TGF-β1, vimentin, and NF-κB and low level of CK18 were noted in the tumor. This was particularly true for vimentin, which is independent of patients' Gleason score. These findings suggest that proteins associated with the EMT predict biochemical recurrence in prostate cancer patients. Clinical Gleason score did not add any predictive value to the proportional hazards model despite that clinical Gleason score sum alone had a significant effect on biochemical recurrence. Also, we did not find any correlation between PSA doubling time with the level of expression of EMT markers, such as TGF-β1, vimentin, and NF-κB. Taken together, markers of the EMT were stronger predictors of PSA recurrence than variables such as clinical Gleason score and PSA doubling time.

Results of the present study suggest that TGF-β can mediate the EMT in prostate cancer cells, which results in an aggressive phenotype. This is possibly mediated by NF-κB, which is considered to be an essential factor for EMT induction (16, 18, 27–30). Ao et al. have proposed that TGF-β promotes invasion in tumorigenic, but not nontumorigenic, prostatic epithelial cells and that TGF-β regulation of vimentin expression is dependent on Akt (15). In the present study, we have confirmed that NF-κB is an integral downstream factor of TGF-β-induced EMT. Because Akt is known to promote NF-κB activity via IKKs (11), in combination with the Ao et al. study, we connected this downstream effector between TGF-β-induced EMT in prostate cancer. More importantly, we found that the detection of EMT in tumor specimens obtained at the time of radical prostatectomy may serve as a predictor of biochemical recurrence in prostate cancer patients following radical prostatectomy.

Transition of cells from an epithelial to a mesenchymal phenotype confers a loss of sensitivity to growth inhibition by TGF-β and manifests an invasive and metastatic phenotype (31). Our results show that TGF-β inhibits prostate cancer cells with a less aggressive phenotype, such as PC-3 and PC-3M-Pr04, but not with the highly aggressive cancer cells, such as PC-3M-LN4 and PC-3M. These latter two cell lines have escaped the growth inhibition by TGF-β, which is linked to a strong induction of EMT and a reduced expression of TGF-β receptors. For example, we observed similar trends in specimens obtained. High levels of expression of TGF-β, vimentin, and NF-κB were identified in 36.7%, 66.6%, and 23.9% of tumors with high, intermediate, and low Gleason grades, respectively. Furthermore, in tumors with intermediate or low Gleason grade, the percentage of positively stained cancer cells for these markers was relatively reduced. There is a significant positive correlation between expression of TGF-β1 and expression of EMT-related factors such as vimentin, NF-κB, and CK18. Consistent with observations obtained from cell lines, TGF-β is a mediator for EMT in prostate cancer.

In the present study, low levels of p-Smad2 were found in tumors with high Gleason score. This finding suggests that these tumors have escaped the growth inhibition by TGF-β. The signaling pathways in TGF-β-induced EMT can be divided into Smad-independent (32–36) and Smad-dependent (37) pathways. Recent reports indicated that phosphorylation of Smad3 at the linker region may be involved in the EMT, which is TβRI and p-Smad2 independent (31, 37). Results of the present study showed that, in prostate cancer, TGF-β, rather than Smads, mediated the activation of NF-κB. Furthermore, we have shown NF-κB to be the major downstream mediator of TGF-β-induced EMT in prostate cancer cells. Our results indicate that blockade of either TGF-β or NF-κB reverses the EMT process. Overexpression of TGF-β1 by tumor cells will result in the induction of NF-κB and vimentin expression, which confers an aggressive phenotype. Although other factors such as Ras, Jag1, and Hey1 may be involved in the process (17, 18), NF-κB is an obligatory switch for TGF-β-induced EMT in prostate cancer.

The present results indicate that TGF-β-induced EMT correlates with tumor invasiveness and biochemical recurrence in prostate cancer patients. NF-κB is the essential intermediate mediator for TGF-β-induced EMT in prostate cancer. Within 7 years following radical prostatectomy, variables including TGF-β1, NF-κB, vimentin, CK18, and patient age correlated significantly with biochemical recurrence. We confirmed that high stage is associated with higher recurrence rate than those undergone surgery at a lower pathologic stage (38). We show that patients with a higher tissue level of vimentin have higher biochemical recurrence rate than those with a lower tissue level of vimentin. Furthermore, the relationship between vimentin expression and biochemical recurrence appears to be independent of pathologic stage or PSA doubling time. This indicates that a high level of TGF-β-induced expression of vimentin (scale 3: 75-100% positively stained cancer cells) in prostatectomy specimens may be used to predict biochemical recurrence in prostate cancer patients. Our result indicates that vimentin in combination with pathologic stage are good predictors for disease outcome. The exact mechanism of this observation remains unclear, but tumors from older patients have higher expression levels of TGF-β1 and lower levels of CK18 expression than those of their younger counterparts, which indicated that EMTs are more likely to occur in former group. Based on our results, during progression of prostate cancer, an attenuation of expression of TGF-β receptors facilitates tumor cells escaping from the growth inhibition by TGF-β, which is Smad dependent. Meanwhile, the Smad-independent pathway, such as NF-κB signaling, mediates the EMT process, which could be reversed by the blockade of TGF-β signaling by infection with TβRIIDN or by the use of the NF-κB inhibitor. Variables such as tumor expression of EMT markers and TGF-β1, or age of patients, may be useful in predicting clinical outcome following radical prostatectomy. High levels of vimentin (75-100% positively stained cancer cells) and pathologic stage (≥T_3a) proved to be risk factors for biochemical recurrence in prostate cancer patients regardless of clinical Gleason score.

In summary, our findings have indicated that NF-κB-mediated TGF-β-induced EMT may be used to predict patients at high risk for biochemical recurrence following radical prostatectomy. This event is independent of Gleason score or PSA doubling time. Transition of cancer cells from an epithelial to a mesenchymal phenotype induces loss of sensitivity to growth inhibition by TGF-β and confers an invasive phenotype (31). It is conceivable that the EMT-related factors may be used as prognostic markers for other types of cancer.

No potential conflicts of interest were disclosed.