Virchows Archiv

, Volume 452, Issue 3, pp 295–304

Clinical significance of nuclear factor (NF)-κB levels in urothelial carcinoma of the urinary bladder

Authors

    • Department of PathologyNational and Kapodistrian University of Athens
  • Angelica A. Saetta
    • Department of PathologyNational and Kapodistrian University of Athens
  • Penelope Korkolopoulou
    • Department of PathologyNational and Kapodistrian University of Athens
  • Polyanthi Papanastasiou
    • Department of PathologyNational and Kapodistrian University of Athens
  • Katerina Gioti
    • Department of PathologyNational and Kapodistrian University of Athens
  • Petros Pavlopoulos
    • Department of PathologyNational and Kapodistrian University of Athens
  • Kalliopi Diamantopoulou
    • Department of PathologyNational and Kapodistrian University of Athens
  • Eupthemia Thomas-Tsagli
    • Department of PathologyNational and Kapodistrian University of Athens
  • Konstantinos Xiromeritis
    • Department of PathologyNational and Kapodistrian University of Athens
  • Efstratios Patsouris
    • Department of PathologyNational and Kapodistrian University of Athens
Original Article

DOI: 10.1007/s00428-007-0560-y

Cite this article as:
Levidou, G., Saetta, A.A., Korkolopoulou, P. et al. Virchows Arch (2008) 452: 295. doi:10.1007/s00428-007-0560-y

Abstract

Nuclear factor (NF)-κB has been reported to be constitutively activated in various human neoplasms. However, its clinical significance in bladder urothelial carcinoma (UC) remains an unresolved issue. We conducted this study trying to elucidate the role of NFκB in bladder UC and its potential prognostic significance, by quantifying immunohistochemically the levels of p65/RelA expression in paraffin-embedded tissue from 116 patients. Some of the cases had previously been stained for cellular FLICE-like inhibitory protein (c-FLIP) and bcl-2. Seventy-four cases displayed concurrent cytoplasmic and nuclear immunoreactivity, whereas 18 only nuclear immunoexpression and 21 only cytoplasmic immunoexpression, and the remaining three cases were negative for p65/RelA. Nuclear p65/RelA expression was positively associated with tumour grade and T-category (p = 0.0001 in both cases). In addition, cytoplasmic p65/RelA expression was lower in advanced T-category (p = 0.0030). Moreover, p65/RelA nuclear expression was positively correlated with c-FLIP (p = 0.0109) and bcl-2 (p = 0.0452). p65/RelA nuclear expression adversely affected survival in both univariate and multivariate analysis in superficial (Ta–T1; p = 0.0010 and p = 0.0008) as well as in muscle-invasive carcinomas (T2–T4; p = 0.0004 and p = 0.0003). Our results demonstrate that NF-κB nuclear expression is correlated with histologic grade and T category in bladder UC. Moreover, NF-κB nuclear expression emerges as an independent prognosticator of adverse significance, conveying information beyond that obtained by standard clinicopathological prognosticators.

Keywords

NFκBImmunoexpressionBladder urothelial carcinoma

Introduction

Urothelial carcinoma (UC) of the urinary bladder is considered to be the fifth most common cancer in industrialized countries [44] and the second most common cancer of the genitourinary tract [16]. Even after complete transurethral resection, 75% of patients will recur [44], a fact highlighting the necessity for the identification of molecular markers that could possibly predict disease aggressiveness and clinical course. Moreover, gaining insight into the molecular determinants of the biologic behaviour of detrusor muscle-invasive bladder cancer is imperative to the development of more rational biologically oriented treatment strategies. Current belief holds that bladder cancer development is a multi-step process, in which, apart from genetic and environmental factors several oncogenes, tumour suppressor genes, cell cycle regulators as well as perturbations in apoptotic cell death regulation are implicated [9, 44].

By virtue of its pivotal role in many biological decisions affecting cell fate, nuclear factor (NF)-κB is the point of convergence of most metabolic and oncogenic pathways [4, 5, 53]. It is a family of transcription factors, whose members share a Rel homology domain (RHD), subjecting them to a particular type of regulation centered on nuclear–cytoplasmic shuttling [13]. The RHD is an approximately 300-amino acid terminal domain, which is composed of two immunoglobulin-like repeats and serves several functions: It is the dimerization and DNA-binding domain for this class of proteins, it contains the nuclear localization signal (NLS) and most importantly, it is the binding site of NF-κB inhibitors [23, 40]. The NF-κB family consists of five RHD proteins namely, p50/NFκB1, p52/NF-κB2, c-Rel, RelB and p65/RelA [6]. Under resting conditions, ΝF-κB is localized in the cytoplasm and consists of three subunits: p50 and p65 (both DNA-binding capacity) and the inhibiting subunit IκBa, which is bound to p65. NF-κB activation may be caused by several cytokines and mitogens, as well as during oxidative stress, cell damage and viral infections [21, 38]. This activation entails the phoshorylation of IκBa by tyrosine kinases IκB kinase (IKK) [49], which in turn are phosphorylated by multiple signalling pathways, including members of the mitogen-activated protein kinase kinase family [23]. IκBa, β and ɛ contain two conserved serine residues in their N-terminal domain, the phosphorylation of which targets them to polyubiquitination and subsequent degradation by the 26S proteasome [40, 48]. This permits the liberation of the nuclear localization signal of NF-κB and consequently its translocation in the nucleus [6]. Activated NF-κB regulates the transcription of a wide variety of reporter genes, by binding to enhancers of target genes inducing their transcription [6]. The genes regulated by NF-κB include anti-apoptotic genes (such as survivin, cellular FLICE-like inhibitory protein [c-FLIP] and bcl-2), cell cycle regulators (such as cyclin D1), angiogenic factors (such as Cox-2), as well as genes encoding adhesion molecules and inflammatory cytokines (such as interleukin [IL]-8) [22, 31, 45].

Accumulating evidence highlights NF-κΒ as a major determinant of human tumourigenesis of several human neoplasms. Several investigators have reported constitutive activation of NF-κB in various tumour cells and cell lines, such as those of lymphoid origin [3, 11, 15], as well as those of ovarian [4], breast [35], pancreatic [52] and thyroid [51] origin. Moreover, its immunoexpression has been correlated with worse patients overall survival in human lung [57], prostatic [31, 45], gastric [32], esophageal [19] and tonsillar squamous carcinoma [56], as well as adenoid cystic carcinoma of salivary glands [55]. However, little information is available about the expression of NF-κB protein in urothelial bladder cancer (UC) [54], and, to the best of our knowledge, there is no published investigation addressing the potential prognostic significance of this molecule in bladder UC.

This study analyzes the immunohistochemical expression of NF-B protein in bladder UC specimens, assessing the expression of RelA/p65 subunit, in an effort to elucidate whether this molecule is correlated with tumour aggressiveness. We were also interested in the relationship of p65/RelA expression with other conventional inhibitors of apoptosis, such as the c-FLIP and bcl-2 protein. Finally, we aimed to clarify the role of NF-κB role in patients’ prognosis in the presence of standard clinicopathological prognosticators.

Materials and methods

Patients

This is a study of 116 consecutive patients with primary bladder UC diagnosed, treated and followed up at Asklepeion Voula and IKA hospitals in Athens between 1985 and 1995, for which sufficient paraffin-embedded tissue and follow-up information were available. Informed consent was obtained from all patients before enrolment in the study. The study follows the principles of the Declaration of Helsinki and was approved by the ethical committee of the University of Athens Medical School. The mean age of patients was 65.48 ± 11.90 ranging from 29 to 89 years, whereas the male-to-female ratio was 103:13. All cases were reviewed by two experienced pathologists (Levidou, Korkolopoulou) and assigned a histological grade and T-category according to the latest World Health Organization (WHO) classification [10] and the fifth edition of the TNM classification [46] taking into account the report of Mikulowski and Hellsten [34]. There were 17 tumours of low malignant potential (papillary neoplasms of low malignant potential, PNLMP), 42 low-grade carcinomas (LG) and 57 high-grade carcinomas (HG). Staging was based on combination of clinical (cystoscopy, computed tomography and ultrasound) and histological data to determine the absence (Ta, n = 17) or presence (T1, n = 34) of invasion of the lamina propria or invasion into or beyond the detrusor muscle (T2–T4, n = 65).

Follow-up information was available in all but three patients. Follow-up period for the remaining 113 patients ranged from 3 to 86.7 months (median = 36 months). During this period, 26 patients had died of their disease after a median survival of 20 months (range 7–48 months). The median follow-up for the remaining 87 patients was 41.5 months (range 3–86.7).

Immunohistochemical staining

Four-micrometre serial sections were cut from each specimen on Superfrost plus glass slides and left to dry overnight at 37°C. After deparaffinization, immunostaining was performed by using a rabbit polyclonal anti-RelA/p65 antibody (18–7308, California, USA, Zymed Laboratories) diluted 1:300. The antibody was incubated overnight at 37°C. For p65 detection, antigen retrieval with microwaving for 30 min, at 750 W, in citrate buffer, pH 6.0, was required. Immunostaining was performed using the standard ABC (Ultravision Detection system, Labvision) and visualized with 3,3′-diaminobenzidine tetrachloride solution. The slides were counterstained with hematoxylin. Known positive controls (i.e. breast carcinoma) as well as negative controls (sections in which p65/RelA antibody was substituted by non-specific rabbit serum) were also stained in each run. In addition, 51 and 49 cases were stained for c-FLIP and bcl-2 proteins, respectively, some of which were available from our previous investigations [27, 28].

Light microscopic evaluation of immunostained slides was performed independently by two pathologists without knowledge of the clinical information. If a discrepancy occurred between the two assessments, the slides were reassessed jointly without information of the previous scores. In each case, 1,000–1,500 neoplastic cells throughout the section were counted at high-power magnification. Nuclear and cytoplasmic p65/RelA expression was evaluated separately. Nuclear p65/RelA labelling index was defined as the percentage of neoplastic cells with nuclear immunoreactivity out of the total number of neoplastic cells counted. Cytoplasmic p65/RelA expression was categorized in three levels according to the percentage of neoplastic cells with cytoplasmic staining, as follows: absent/low (level 1): 0–20% of neoplastic cells with cytoplasmic staining, moderate (level 2): 21–60% of neoplastic cells with cytoplasmic staining and extensive (level 3): >60% of neoplastic cells with cytoplasmic staining.

Statistical methods

In the basic statistical analysis, p65/RelA nuclear expression was treated as a continuous variable, whereas cytoplasmic p65/RelA expression as ordinal. Correlation between nuclear and cytoplasmic p65/RelA expression was tested with Kruskal–Wallis analysis of variance (ANOVA) along with Spearman’s rank correlation coefficient. Associations of p65/RelA nuclear expression with other clinicopathological parameters, such as grade, T-category, tumour size and patients’ sex and gender, were tested with Kruskal–Wallis ANOVA, whereas the corresponding associations of cytoplasmic p65/RelA expression were tested with Chi-square or Fisher’s exact test, as appropriate. Associations of p65/RelA nuclear and cytoplasmic expression with c-FLIP and bcl-2 were calculated with Spearman’s rank correlation coefficient and Kruskal–Wallis ANOVA, respectively.

The survival analysis was performed using death of disease as the endpoint. The effect of various parameters (age, sex, tumour size, T-category, grade and p65/RelA nuclear and cytoplasmic expression) on clinical outcome was assessed by plotting survival curves according to the Kaplan–Meier method and comparing groups using the log-rank test. In univariate survival analysis, p65/RelA nuclear expression was categorized on the basis of the median value (3%). Multivariate analysis was performed using Cox’s model with stepwise forward selection to evaluate the predictive power of each power independently of others. As in basic statistical calculations, p65/RelA nuclear expression was treated as a continuous variable, whereas cytoplasmic as ordinal. Given that c-FLIP and bcl-2 were not available in many cases and their prognostic utility has been the subject of previous investigations [27, 28], these two variables were not included in survival analysis. Statistical calculations were performed using the Statistical package STATA 9.0/SE for Windows. All results with a two-sided p level less than or equal to 0.05 were considered statistically significant.

Results

Expression of p65/RelA in bladder UCs

Immunoreactivity of p65/RelA was nuclear and cytoplasmic. Seventy-four cases displayed concurrent cytoplasmic and nuclear immnoreactivity, 18 cases displayed only nuclear immunoexpression, 21 cases displayed only cytoplasmic immunoexpression and the remaining three cases were devoid of p65/RelA immunoreactivity. The overall rate of p65/RelA nuclear immunoexpression rose to 79.3% (92 of 116 cases). Positive neoplastic cells (with nuclear or cytoplasmic staining) were diffusely distributed throughout neoplastic tissue with little variation among individual cases. Normal urothelium displayed mainly a cytoplasmic p65/RelA staining with only a few scattered positive nuclei (Fig. 1a). Tumour-infiltrating lymphocytes invariably expressed p65/RelA, serving therefore as an internal positive control for each case.
https://static-content.springer.com/image/art%3A10.1007%2Fs00428-007-0560-y/MediaObjects/428_2007_560_Fig1_HTML.jpg
Fig. 1

a Normal urothelium displaying mainly cytoplasmic p65/RelA immunoreactivity. b, c, d Immunohistochemical expression of p65/RelA in a papillary neoplasm of low malignant potential, Ta (MX200, b), low grade, T1 (MX200, c), high grade, T2 (d, MX400) and high grade, T1 (MX200, e) bladder urothelial carcinomas (original magnification, ×40)

Nuclear p65/RelA immunoexpression decreased with increasing levels of cytoplasmic p65/RelA (Kruskal–Wallis ANOVA, p = 0.0049, Fig. 2), a relationship that remained significant when the analysis was restricted to muscle-invasive carcinomas (T2–T4; Kruskal–Wallis ANOVA, p = 0.0272), but failed to retain its statistical significance in superficial (Ta–T1) bladder UCs (p > 0.10). This negative correlation was confirmed with the Spearman’s rank correlation coefficient (R = −0.3012, p < 0.0001) in the entire cohort, as well as in muscle-invasive carcinomas (R = −0.3223, p = 0.0100).
https://static-content.springer.com/image/art%3A10.1007%2Fs00428-007-0560-y/MediaObjects/428_2007_560_Fig2_HTML.gif
Fig. 2

Schematic representation of the association of p65/RelA nuclear with cytoplasmic p65/RelA expression

Correlations of p65/RelA with clinicopathological parameters

Nuclear p65/RelA expression increased with increasing histological grade in the entire cohort (Kruskal–Wallis, ANOVA, p = 0.0001, Table 1, Fig. 1b,c,d, 3a), as well as in superficial (Ta–T1, Kruskal–Wallis, ANOVA, p = 0.0011) and muscle-invasive carcinomas (T2–T4 Kruskal–Wallis, ANOVA, p = 0.0001). Moreover, nuclear p65/RelA immunostaining was positively correlated with tumour T-category in the entire cohort (Kruskal–Wallis, ANOVA, p = 0.0001, Table 1, Fig. 1b,c,d, 3b) and in muscle-invasive carcinomas (Kruskal–Wallis, ANOVA, p = 0.0047), but this relationship failed to attain statistical significance when analysis was restricted to superficial (Ta–T1) bladder UCs (p > 0.10). In addition, nuclear p65/RelA was marginally correlated with tumour size (p = 0.0611, Table 1) but was unrelated to patients’ age and gender (p > 0.10).
https://static-content.springer.com/image/art%3A10.1007%2Fs00428-007-0560-y/MediaObjects/428_2007_560_Fig3_HTML.gif
Fig. 3

Schematic representation of the association of p65/RelA nuclear expression with a histologic grade and b T-category in the entire cohort. p65/RelA increases with increasing grade and T-category

Table 1

Relationship between nuclear p65/RelA expression and clinicopathological parameters in 116 patients with urothelial bladder carcinoma

 

p65/RelA nuclear expression (%)

Number

median

Range

p value

Age (years)

116

70

39–89

>0.10a

Grade

 

 

 

0.0001b

 Papillary neoplasms of low malignant potential, PNLMPs

17

1

0–5

 

 Low-grade carcinomas, LG

42

2

0–8

 

 High-grade carcinomas, HG

57

5

0–25

 

Stage

 

 

 

0.001b

 Ta

17

1

0–5

 

 T1

34

2

0–12

 

 T2

40

3

0–20

 

 T3

16

5

0–20

 

 T4

7

5

0–25

 

Tumour size (cm)

116

2

0.1–25

0.0611a

Sex

 

 

 

>0.10c

 Female

13

5

0–15

 

 Male

103

2

0–25

 

aResults of Spearman’s correlation coefficient

bResults of Kruskal–Wallis ANOVA

cResults of Mann Whitney U test

Accordingly, cytoplasmic p65/RelA reactivity was marginally correlated with histological grade in the entire cohort (p = 0.0810, Table 2). Furthermore, cytoplasmic p65/RelA immunoreactivity decreased with increasing tumour T-category in the entire cohort and in muscle-invasive carcinomas (p = 0.0030 and p = 0.0073, respectively, Table 2). The respective relationships between p65/RelA cytoplasmic expression and tumour grade or T-category in superficial bladder UCs (Ta–T1) failed to attain statistical significance (p > 0.10). However, patients’ age, sex and tumour size were unrelated to cytoplasmic p65/RelA (p > 0.10).
Table 2

Associations of cytoplasmic p65/RelA expression with clinicopathological parameters in 116 patients with urothelial bladder carcinoma (results of Fisher’s exact test)

 

Cytoplasmic p65/RelA

p value

Absent/low

Medium

Excessive

Grade

 

 

 

0.0810

 Papillary neoplasms of low malignant potential, PNLMPs

15

25

17

 

 Low-grade carcinomas, LG

5

16

21

 

 High-grade carcinomas, HG

1

6

10

 

Stage

 

 

 

0.0030

 Ta

1

6

10

 

 T1

2

19

13

 

 T2

9

12

19

 

 T3

6

5

5

 

 T4

3

3

1

 

Sex

 

 

 

>0.10

 Female

3

5

5

 

 Male

18

41

44

 

Relationship of p65/RelA with c-FLIP and bcl-2

The associations of c-FLIP and bcl2 with clinicopathological parameters in bladder UCs have been examined in detail previously [28, 27] and therefore will not be mentioned.

Nuclear p65/RelA expression was positively correlated with c-FLIP and bcl-2 expression (Spearman’s rank correlation coefficient, R = 0.3178, p = 0.0109). Moreover, nuclear expression was increased in bcl-2-positive neoplasms compared to bcl-2-negative ones (Kruskal–Wallis, ANOVA, p = 0.0452). Cytoplasmic p65/RelA expression was not associated with c-FLIP or bcl2 expression levels (Kruskal–Wallis, ANOVA, p > 0.10).

Survival analysis

Survival analysis was performed separately for Ta–T1 and T2–T4 UCs because of the substantially worse behaviour of the latter group (T4–T2 vs Ta–T1, p > 0.0001, log-rank test). In univariate analysis of Ta–T1 urothelial bladder carcinomas, higher p65/RelA nuclear expression levels (>3 vs ≤3%, Fig. 4a), higher T-category (T1 vs Ta) and histological grade (HG vs LG vs PNLMP, p < 0.0001) implied a lesser probability of survival (Table 3). The parameters affecting survival in muscle-invasive carcinomas (T2–T4) were p65/RelA overexpression (>3 vs ≤3%, Fig. 4b), low p65/RelA cytoplasmic expression (0 vs 1 vs 2,), histological grade (HG vs LG vs PNLMP, p = 0.0031) and T-category (T2 vs T3 vs T4; Table 3). The median survival time for patients whose tumours overexpressed nuclear p65/RelA was 66.6 and 30 months in superficial and muscle-invasive bladder UCs, respectively. It is to be noted that none of the patients within the low-expressor group in both Ta–T1 and T2–T4 bladder UCs had a disease-specific event, and therefore median and mean survival time for this group cannot be calculated.
https://static-content.springer.com/image/art%3A10.1007%2Fs00428-007-0560-y/MediaObjects/428_2007_560_Fig4_HTML.gif
Fig. 4

Kaplan–Meier survival curves according to p65/RelA expression in a 51 patients with superficial and b 62 patients with muscle-invasive bladder urothelial carcinoma

Table 3

Univariate analysis in Ta–T1 and in T2–T4 bladder UCs

Parameter

Ta/T1 bladder UCs, n = 51

T2–T4 bladder UCs, n = 62

Grade (PNLMP vs LG vs HG)

<0.0001

0.0031

T-category

Ta vs T1 0.0341

T2 vs T3 vs T4, p < 0.0001

Age (>35 vs ≤35)

0.1193

0.4734

Gender (1: male, 2: female)

0.7095

0.9969

Tumour size (<1.5 vs ≥1.5 cm)

0.0870

0.7681

P65/RelA nuclear LI (≤3 vs >3%)

0.0010

0.0004

P65/RelA cytoplasmic expression (0 vs 1 vs 2)

0.6546

0.0244

Analysis using the log-rank test (p values)

In multivariate analysis of superficial carcinomas (Ta/T1), p65/RelA expression was the only parameter that remained significant. High p65/RelA nuclear expression and T-category turned out to be the independent prognostic factors for muscle-invasive carcinomas. The final models are presented in Table 4.
Table 4

Cox’s proportional hazard estimation, in superficial (Ta–T1; model A) and in muscle invasive (T2–T4; model B)

Models

Parameter

Hazard ratio (HR)

p value

95% confidence interval of HR

A

p65/RelA nuclear LI

1,972

0,0080

1,194

3,257

B

p65/RelA nuclear LI

1,107

0,0003

1,034

1,184

T category T2 vs T3 vs T4

4,152

< 0,0001

2,252

7,651

Discussion

Although bladder UC is often detected at an early stage, the identification of those tumours prone to progress or recur remains a difficult and challenging issue [44]. Several molecular markers have been suggested for the stratification of patients in terms of clinical course and outcome, among which cell cycle regulators, such as p53, p27, pRb and cyclin D1 [8, 9, 26, 25, 36, 42, 44], as well as apoptotic coordinators, such as c-Flip and bcl-2 family proteins [27, 28, 44], figure prominently. Published studies, however, have yielded contradictory results, mainly because of the presence of influential methodological issues (i.e. small sample size, the ratio of superficial to muscle-invasive cases and the administration of different therapeutic protocols), which confound their outcomes and consequently hamper the identification of a universally approved prognostic marker.

Accumulating evidence suggests that NF-κB is constitutively activated and exhibits a strong transactivation potential in several human cancer tissues [4, 14, 24, 35, 41, 50, 51]. Accordingly, nuclear translocation of the p65 subunit has been reported to be markedly higher in gastric, colon and prostatic adenocarcinoma cells when compared to their normal counterparts [33, 43, 45]. However, published information regarding the role of NF-κΒ activation in tumours of the genitourinary tract is limited to a few reports describing aberrant activation in renal cell and prostatic carcinomas [31, 38, 45]. Importantly, in a recent Chinese study, NF-κB expression is higher in bladder UC cells in comparison to that of non-neoplastic mucosa [54]. In our cohort, we observed that normal urothelium displayed a mainly cytoplasmic p65/RelA immunoexpression, whereas in the majority of the studied tumours (79.3%), p65/RelA was seen in the nucleus. Cytoplasmic NF-κB immunolocalization is considered to correspond to the inactive form of NF-κB protein, whereas nuclear immunoreactivity is regarded as a surrogate marker for the activated NF-κB protein. This staining pattern is broadly in line with investigations concerning other neoplasms, such as gastric and prostatic adenocarcinomas [32, 45].

The detailed mechanisms underlying NF-κB activation in bladder UC remain unclear, and the analysis of NFκB regulating pathways and molecules for the detection of mutations that possibly promote NF-κB activation in many other human cancers is still in progress. NFκB activation has been reportedly attributed to the phosphorylation of IκB protein by IKK kinase. However, it is conceivable that NF-κB activation is a multi-step process in which several molecules interact to initiate a highly coordinated response [41]. Differential activation and expression of these molecules under different conditions could possibly account for the apparently contradictory data obtained in different tissues. For example, although NF-κB-inducing kinase plays the major role in constitutive activation of NF-κB in melanoma cells [7], in prostate cancer cells, this appears to result from the increased activity of the IKK kinase complex [12, 39, 47]. Clearly, the question of the alterations in the upstream signalling pathways that result in the constitutive activation of the NF-κB protein remains to be addressed in bladder UC cell lines.

Another interesting issue emerging from the present investigation relates to the differential nuclear expression of NF-κB during the progression of bladder UC. Thus, p65/RelA nuclear expression levels increased in higher tumour grade and advanced T-category cases, whereas cytoplasmic p65/RelA immunoreactivity decreased with increasing tumour histologic grade and invasive potential, in muscle-invasive (T2–T4) carcinomas. Taking into account that the normal urothelium displayed minimal nuclear p65/RelA immunoreactivity, our results speak in favour of the potential role of this molecule in urothelial bladder tumourigenesis as well in the acquisition of a more aggressive phenotype. It is worthy of note, however, that the significant overlap in the nuclear expression levels of p65/RelA across the various grade and stage categories hampers its use as a diagnostic tool supplementing the existing WHO criteria. Our observations are in alignment with previous studies concerning prostate and renal cell carcinomas [37, 45], in which p65/RelA nuclear immunoreactivity was related to tumour invasiveness and aggressive behaviour. Along this line, it has been postulated that NF-κB can stimulate the transcription of genes that may be essential for malignant transformation with consequent genomic instability, accelerating neoplastic progression [17]. Moreover, our results are in accordance with experimental data in a panel of human bladder UC cell lines with varying metastatic potential indicating that highly metastatic bladder UC cells constitutively expressed high levels of NF-κΒ, in comparison to that of cells with low metastatic potential [22]. In addition, the role of the NF-κΒ protein in bladder UC tumourigenesis has previously been speculated because of its relationship with interstitial cystitis [2], given that the development of bladder cancer is related to urinary tract infection [30, 52]. Thus, NF-κB is more often activated in urothelial cells from interstitial cystitis as compared to controls [2].

Perhaps, the most intriguing finding of our study was the fact that p65/RelA nuclear overexpression connotes a poor survival probability and remains a prognostic factor after multivariate analysis in both superficial and muscle-invasive UCs. It is worthy of note that this is the first attempt to link NF-κB expression with survival in bladder UC. Cytoplasmic p65/RelA expression does not seem to be informative in this regard. A plausible explanation for this finding is that the relationship between cytoplasmic p65/RelA immunoreactivity and survival that was elicited in univariate survival analysis in muscle-invasive carcinomas largely resides in its relationship with nuclear p65/RelA, the latter thus acting as a confounding factor. Our results are in line with previous studies in gastric and prostatic adenocarcinomas, in which nuclear immunolocalization of the NF-κB protein was proposed as a useful prognostic marker [31, 32]. Although the median period of surviving patients in our study was approximately 3.5 years, it should be acknowledged that the vast majority of bladder UCs progress clinically during the first 3 years after presentation [41], which allows meaningful conclusions to be drawn at the end of this period. Furthermore, our results in survival analysis recapitulate most of the parameters that have been established as determinants of clinical outcome in patients with bladder UC (i.e. histologic grade, T- category), advocating that our cohort is representative and leads to valid results. Moreover, the percentage of high-grade cases in the group of muscle-invasive carcinomas (57 in 65 cases) is broadly in line with the figure quoted in previously published studies [29]. The fact that nuclear p65/RelA expression maintained its statistical significance in multivariate survival analysis along with tumour stage in muscle-invasive carcinomas indicates that the prognostic significance of p65/RelA nuclear overexpression cannot be attributed to its relationship with this parameter and that there should be other reasons accounting for this finding, e.g. the inhibition of apoptosis that results from NF-κB activation [1, 23]. Our preliminary results of a positive relationship between p65/RelA nuclear immunoexpression with both c-FLIP and bcl-2 expression that emerged in our cohort, despite the small number of cases studied, raise the hypothesis that NF-κB may exert anti-apoptotic effects in bladder UCs. Of course, validation of these results in a larger number of cases is mandatory to verify this assumption. In this context, the NF-κB protein has been shown to induce anti-apoptotic effects in cells treated with apoptotic agents, such as tumour necrosis factor-α, ionizing radiation or chemotherapeutic agents [18, 20], therefore enhancing chemoresistance and radioresistance. Horiguchi et al. investigated the use of a dehydroxymethyl derivative of epoxuquinomycin C—a NF-κB function inhibitor in bladder cancer—and suggested that it represents a possible new treatment strategy by virtue of its role in the promotion of apoptosis against advanced bladder cancer [18]. An additional mechanism could relate to the involvement of NF-κB in the activation of a number of cell cycle regulators (i.e. cyclin D1) that have been implicated in patients’ overall survival in bladder cancer [8, 36, 42], as well as genes participating in the process of tumour metastasis (i.e. metalloproteinases 3) or angiogenesis (i.e. Cox-2 and IL-8) [21, 22].

In conclusion, our study demonstrates that p65/RelA nuclear expression and therefore NF-κB activation parallels histologic grade and T-category especially in muscle-invasive bladder UC, directly implicating NF-κΒ in the acquisition of the malignant and more aggressive phenotype in urothelial bladder carcinomas. More importantly, p65/RelA nuclear overexpression emerges as a valuable adverse prognostic marker in both superficial and muscle-invasive tumours, retaining its significance in multivariate analysis, in the presence of tumour T-category and histologic grade, thus conveying supplementing information obtained by standard clinicopathological prognosticators. These results encourage further investigations into the use of NF-κB inhibitors as an adjunct to the currently available treatment options in bladder UC [18]. Analysis of the potential role of such molecules in human urothelial bladder carcinoma cell lines is warranted to validate our observations and to provide more insight into their function in bladder UC.

Conflict of interest statement

We declare that we have no conflict of interest.

Copyright information

© Springer-Verlag 2007