Prognostic Significance of CREB-Binding Protein and CD81 Expression in Primary High Grade Non-Muscle Invasive Bladder Cancer: Identification of Novel Biomarkers for Bladder Cancer Using Antibody Microarray

High-grade (HG) bladder cancers (BCs) are genetically unstable and have an unpredictable course. The identification of prognostic factors in HG non-muscle invasive BC (NMIBC) is crucial for improving patients’ quality of life and preventing BC-specific mortality. Here, we used an antibody microarray (AbM) to identify novel candidate biomarkers in primary HG NMIBC and validated the prognostic significance of the candidate biomarkers. Three pairs of tissue samples from primary HG NMIBC and normal urothelium were analyzed using an AbM kit containing 656 antibodies, and differentially expressed proteins were identified. Among the 42 upregulated and 14 downregulated proteins with statistical significance in BC tissues, CREB-binding protein and CD81 were selected as representative upregulated and downregulated candidate biomarkers, respectively. We then validated the expression of these candidate biomarkers in primary human urothelial cells and BC cell lines by western blotting and immunofluorescence assays, and the results were consistent with the AbM expression profiles. Additionally, Kaplan-Meier survival using immunohistochemical data from an independent primary HG NMIBC cohort comprising 113 patients showed that expression of the 2 biomarkers was significantly associated with recurrence-free and progression-free survival. In multivariate analysis, the 2 biomarkers remained significant predictors for recurrence-free survival. Taken together, our findings suggest that expression of CREB-binding protein and CD81 in BC tissue specimens may have prognostic value in patients with primary HG NMIBC.

Introduction [16]. All TURBT surgeries were performed with a curative intent using a standard technique as previously described [17].

AbM profiling of human bladder tissues
For the AbM analysis, paired tissue samples of BC and normal urothelium were acquired from 3 patients who were diagnosed with primary HG NMIBC following complete TURBT. Exclusion criteria were LG BC, recurrent HG BC, an absence of proper muscle in the TURBT specimen, and the presence of other malignancies. Given the possible field effect [18] of normalappearing bladder mucosa surrounding the BC, patients with concomitant carcinoma in situ were also excluded. After the tissue samples were obtained intraoperatively, they were immediately frozen in liquid nitrogen and stored at -70°C until protein extraction.
The AbM was prepared and analyzed using an AbM assay kit with 656 antibodies (Fullmoon Biosystems, Sunnyvale, CA) according to the manufacturer's guidelines, as previously described [19]. Briefly, protein was extracted using lysis beads in a protein extraction buffer (Fullmoon Biosystems) containing 1% protease inhibitor cocktail (Sigma, St. Louis, MO) and 1% phosphatase inhibitor cocktail. After extraction, the protein solution was purified using the gel matrix column included in the AbM kit. The concentration of purified sample was then measured using the BCA Protein Assay Kit (Pierce, Rockford, IL) and a NanoPhotometer (Implen, UK). The protein sample was then labeled using a biotin/dimethylformamide solution. The AbM slide was treated with 30 mL of blocking solution in a petri dish and incubated on a shaker for 1 hour at room temperature. After blocking, the slide was rinsed with Milli-Q grade water. The labeled samples were then mixed with 6 mL of coupling solution and incubated with the blocked microarray slide in a coupling dish on a shaker for 2 hours at room temperature. After coupling, the slide was washed 6 times with 30 mL of washing solution. Then, 30 μL of 0.5 mg/mL Cy3-streptavidin (GE Healthcare, Piscataway, NJ) was mixed with 30 mL of detection buffer. The slides were scanned at 10-μm resolution with a GenePix 4000B scanner (Molecular Devices LLC, Sunnyvale, CA), and the scanned images were gridded and quantified using GenePix software. The numeric data were analyzed using Genowiz 4.0 (Ocimum Biosolutions, India). Global normalization was used for data analysis. A 2-class t-test or paired t-test was performed to compare each experiment-reference group combination. The averages of normalized ratios were calculated by dividing the average normalized BC sample intensity by the average normalized normal urothelium sample intensity. The benchmarks for upregulation and downregulation were 1.3-fold and 0.77-fold, respectively. After analysis, the data were annotated using protein information in the UniProt database. A hierarchical clustering algorithm was used to evaluate the association of protein profiles and the study individuals, with Pearson correlation as the distance metric, where distance = (1-Pearson correlation value)/2, and the average linkage method, as described previously [12]. The array data were deposited in the NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE59440 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE59440).

Cell culture
Primary human urothelial cells were purchased from ScienCell Research Laboratories (Carlsbad, CA) and cultured in Urothelial Cell Medium (ScienCell Research Laboratories) according to the manufacturer's recommendations. BC cell lines (RT4, T24, and TCC-SUP) were obtained from the Korean Cell Line Bank (Seoul, South Korea) or American Type Culture Collection (ATCC, Manassas, VA) and were cultured at 37°C in Dulbecco's Modified Eagle's Medium (Welgene, South Korea) or Eagle's minimal essential medium supplemented with 10% fetal bovine serum (Gibco BRL, Grand Island, NY).

Western blotting and immunofluorescence assay (IFA)
Western blotting and IFA were performed as described previously [20]. Polyclonal rabbit anti-CREBBP (Bioss Inc., Woburn, MA), anti-CD81 (Thermo Scientific, Rockford, IL) and anti-βtubulin (Sigma) antibodies were used as primary antibodies. For IFA, samples were examined using a Nikon Eclipse E400 microscope (Nikon Instruments Inc., Melville, NY). Images were captured using a Nikon Digital Sight DS-U2 camera, and analyzed using NIS element F.

Pathology evaluation and immunohistochemistry
To validate the prognostic values of the selected candidate proteins, immunohistochemical studies were performed on tissue specimens from 113 patients who had been diagnosed with primary HG NMIBC following complete TURBT between 2000 and 2011 at Eulji University Hospital. The patients in the validation cohort underwent surveillance according to our followup protocol [9,21]. Patients with short-term follow-up periods (less than 6 months) were excluded. All pathology slides prepared from each TURBT specimen were thoroughly reevaluated by a single uropathologist (JHK). Lymphovascular invasion (LVI) was defined as the observation of tumor cells in the luminal space lined by endothelial cells on hematoxylin and eosin staining [17].

Statistical analysis
Outcomes were recurrence-free survival (RFS) and progression-free survival (PFS). Tumor recurrence was defined as the presence of pathological evidence of similar-or lower-stage disease by bladder biopsy or TURBT, and progression was defined as a pathological shift to more advanced-stage disease. Fisher's exact test was used to evaluate the association between categorical variables. The Kaplan-Meier method was used to calculate RFS and PFS rates, and differences were evaluated with the log-rank test. For the analysis, biomarker expression based on immunohistochemical score was appropriately dichotomized because such groupings showed the most significant survival difference in the Kaplan-Meier analysis. The prognostic significance of candidate biomarkers and clinical parameters was assessed by univariate and multivariate Cox regression analysis. All tests were 2-sided, and p values less than 0.05 were considered significant. All statistical analyses were performed using Stata/SE software, version 12.1 (Stata Corporation, College Station, TX).

Differential protein expression identified by AbM profiling
Differentially expressed proteins between BC and normal urothelium were selected based on the fold change criteria (>1.33-fold or <0.77-fold). We found 42 upregulated proteins and 14 downregulated proteins with statistical significance in the BC tissues (Fig 1). While some proteins (shown in red in Fig 1) such as CREBBP, progesterone receptor, CD56, caspase 7, caspase 1, and PDGFR-β were more highly expressed in BC tissue than in normal mucosa, other proteins (shown in green in Fig 1), including caldesmon, desmin, myeloperoxidase, and CD81, had lower expression in BC tissue. Some proteins that appeared to not be associated with cancer development, such as IgA and light chain, were excluded as potential candidate biomarkers.
After an extensive, critical literature review focused on the association of the aforementioned differentially expressed proteins with cancer development and prognostic value in solid cancers, including BC [23][24][25][26][27][28][29][30][31], we identified CREBBP and CD81, which have been investigated in a limited number of BC studies (less than 3), as representative upregulated and downregulated candidate biomarkers, respectively.

Evaluation of candidate protein expression in human cell lines
To validate the expression of the candidate biomarkers in human cells, we performed western blotting and IFA (Fig 2A and 2B). The levels of CREBBP expression in 3 BC cell lines (RT4, T24, and TCC-SUP) were higher than those in primary human urothelial cells. The expression of CD81 in all tested BC cell lines, except the LG cell line RT4 [32,33], were significantly lower than that in primary urothelial cells. Similar expression patterns were observed by IFA ( Fig  2B). CREBBP expression was mainly detected in the nucleus, and its expression was higher in T24 cells than in primary human urothelial cells. CD81 expression was prominent at the cell surface of primary human urothelial cells, whereas it was only detected in a few T24 cells.
Immunohistochemical staining for candidate biomarkers CREBBP and CD81 expression was also examined in human tissues by immunohistochemistry (Fig 3). CREBBP was expressed primarily in the nuclei of BC cells, whereas its expression was weak or absent in normal urothelium. CD81 was highly expressed on the cytoplasmic membrane of normal urothelium, whereas its expression was much lower in BC cells, although there was some variation in BC tissues. The immunohistochemical scores for the 2 biomarkers, based on staining area and intensity, were as follows: CREBBP expression was negative in 73 patients (64.6%), mild in 31 patients (27.4%), moderate in 5 patients (4.4%), and strong in 4 patients (3.5%); CD81 expression was negative in 8 patients (7.1%), mild in 21 patients (18.6%), moderate in 22 patients (19.5%), and strong in 62 patients (54.9%). Based on these data, CREBBP expression was dichotomized as negative or mild (designated as "positive"), and CD81 expression was dichotomized as moderate (designated as "low") or strong (designated as "high").

Relationship between immunohistochemical biomarker expression and clinicopathological characteristics
To evaluate the relevance of CREBBP and CD81 as clinical biomarkers in BC, we analyzed expression in an independent HG NMIBC cohort by immunohistochemistry. Table 1 summarizes the baseline characteristics of the validation cohort. The median age of the patients was 68 years (range, 28-85 years). Positive CREBBP expression was significantly associated with adverse pathological characteristics, including multifocality, LVI, and higher T stage, whereas high CD81 expression was significantly associated with favorable pathological characteristics, including less carcinoma in situ, less LVI, papillary morphology, and lower T stage ( Table 2). A significant inverse correlation was noted between CREBBP and CD81 expression (Kendall's τ = -0.258, p = 0.006).

Prognostic value of the candidate biomarkers
The median follow-up was 48.0 months (mean, 49.7 months; range, 6-138.6 months). During surveillance, recurrence and progression were observed in 45 (39.8%) and 20 (17.7%) patients, respectively at median times of 14.0 and 32.9 months, respectively. The overall 5-year RFS and PFS rates were 65.6% and 82.1%, respectively. In the Kaplan-Meier survival analysis, positive CREBBP expression was significantly associated with lower RFS and PFS rates than negative expression (Fig 4A and 4B), whereas high CD81 expression was significantly associated with higher RFS and PFS rates than low expression (Fig 4C and 4D). Similarly, univariate Cox regression analyses showed that CREBBP and The expression of CREBBP and CD81 were examined by immunofluorescence assay. CREBBP expression was mainly detected in the nucleus, and its expression was higher in T24 cells than in primary human urothelial cells. CD81 expression was prominent at the cell surface of primary human urothelial cells, but was only detected in a few T24 cells. Representative images from 3 independent experiments are shown. Nuclei and target proteins were detected using immunofluorescence staining and observed with a Nikon Eclipse E400 microscope at 400× magnification. UC, primary human urothelial cells.  CD81 expression were significant predictors for tumor recurrence and progression (Table 3). Compared to negative expression, positive CREBBP expression was associated with a 2.417-fold increased risk for tumor recurrence (p = 0.003) and a 3.213-fold increased risk for progression (p = 0.011). Conversely, compared to low CD81 expression, high CD81 expression was associated with a 0.650-fold decreased risk for tumor recurrence (p < 0.001) and a 0.630-fold decreased risk for progression (p = 0.005).
In a multivariate analysis including both biomarkers, the 2 biomarkers remained significant predictors for RFS. CREBBP and CD81 alterations were associated with a 2.029-fold (p = 0.048) and 0.727-fold (p = 0.018) increased risk for tumor recurrence, respectively (Table 3). In multivariate analysis with each biomarker alone, similar findings were obtained (S1 Table). In multivariate analysis for PFS, intravesical therapy only was significant for PFS (HR = 0.275, p = 0.029), and the 2 biomarkers were not significant (Table 3). Survival analysis according to the 2 prognostic biomarkers indicated that RFS significantly decreased with an increased number of marker alterations (Fig 5).

Discussion
This study showed, for the first time, the use of an AbM to identify novel candidate biomarkers that can differentiate between BC and normal urothelium. The 2 candidate proteins chosen from the AbM analysis, CREBBP and CD81, were validated in BC cell lines and human tissue samples. In addition, we confirmed the prognostic significance of the candidate proteins in primary HG NMIBC. These findings indicate potential applicability of AbM profiling for the identification of novel biomarkers in primary HG NMIBC.
AbM profiling has the advantage that it directly measures protein expression. Although cDNA microarrays are commonly used to evaluate mRNA expression, gene expression profiling may not represent the cellular proteome because posttranslational modifications cannot be detected. We found that the protein expression from the AbM analysis was relatively reliable correlated with the results of western blotting/IFA of cell lines and immunohistochemistry on BC tissue samples. However, the expression of the candidate proteins identified by the AbM profiling still needed to be confirmed in additional experiments. Because AbMs are based on protein extracts, the microarrays do not provide information about cell type (e.g., epithelial or stroma cells) or the location (e.g., cytosol or nucleus) of the protein. For example, PDGFR-β expression, which was recently suggested as a prognostic biomarker in NMIBC [31], was not detected in BC tissues or in normal urothelial tissues and cells even though this protein was upregulated in our AbM profiling (Fig 1). Interestingly, immunohistochemical staining showed that PDGFR-β was only located in BC stromal tissues (S1 Fig). CREBBP is highly homologous and functionally related to p300, which was originally identified as a protein that interacts with the adenoviral E1A oncoprotein [34,35]. CREBBP is involved in multiple signal transduction pathways where is functions as a transcription cofactor for genes that encode many different proteins, such as oncoproteins, transforming viral proteins, and tumor-suppressor proteins. CREBBP has been implicated in the carcinogenesis of several cancers [23][24][25]. For example, genomic alteration of CREBBP is associated with esophageal cancer [24] and small cell lung cancer [25]. CREBBP expression has also been shown to be associated with prognosis of small cell lung cancer [25]. Positive CREBBP expression was associated with a significantly lower survival than negative CREBBP expression. To date, only one study has examined the expression of CREBBP in human BC cell lines and radical cystectomy tissue specimens [26]. The study [26] included both NMIBC (40% of the patient population) and muscle-invasive BC (stage T2 or higher; 60% of the patient population), positive CREBBP expression was noted in 86% of patients (47/55), which was significantly higher than that in ours (35.4%). Although differences in patient populations preclude direct comparison between the two studies [26], these findings suggest that MIBC has a higher proportion of CREBBP positivity than NMIBC. No previous study has examined the prognostic value of CREBBP expression in BC patients. Our results suggest that positive CREBBP expression is significantly associated with adverse pathological characteristics and poor prognosis in primary HG NMIBC. Since CREBBP is a transcription cofactor for many different proteins, further studies are needed to elucidate the underlying mechanism in BC.
CD81, a member of the tetraspanin family, is a cell-surface protein that is involved in a broad range of cellular functions, including development, activation, growth, and motility. With regard to cancer pathogenesis, in several tumors, CD81 has been shown to play a role in cancer cell migration and progression; in particular, decreased expression of CD81 is associated with metastasis in hepatocellular carcinoma and with the growth and survival of gastric cancer  [27,28]. In a recent study, a direct interaction between the C-terminal cytoplasmic domain of CD81 and Rac has been suggested as a possible mechanism for regulating cell migration [29]. Studies on the role of CD81 in BC carcinogenesis and its prognostic value are sparse; however, one study reported heterogeneous CD81 and CD82 expression in human BC cell lines [30], but its clinical significance was not investigated. To our knowledge, this study is the first to demonstrate an inverse correlation between CD81 expression and adverse pathological characteristics and its prognostic value in NMIBC. Further studies are required to reveal the exact mechanisms underlying the role of CD81 in BC carcinogenesis and prognosis. Our validation cohort results clearly show that CREBBP and CD81 have prognostic value for predicting the clinical outcomes of 113 patients with primary HG NMIBC. Significantly, our RFS analysis with both prognostic biomarkers (Fig 5) suggested that the 2 biomarkers may have a cumulative effect on patient prognosis, even though the markers are involved in different cellular processes. The lack of statistical significance in the PFS analysis was likely due, at least in part, to the relatively small number of progression events in our validation cohort.
In summary, this study demonstrated the potential applicability of AbM profiling for identifying novel biomarkers in NMIBC. Alterations in the expression of 2 biomarkers, CREBBP and CD81, identified in our study were associated with biologic aggressiveness of tumors and poor prognosis in primary HG NMIBC. Although further external validations are required in larger cohorts, we suggest that evaluation of the expression of CREBBP and CD81 in TURBT specimens has useful prognostic value in patients with primary HG NMIBC and that patients with positive CREBBP and low CD81 expression may require closer surveillance and more aggressive treatment.