Errata: Erratum Volume 117, Issue 21, 5021, Article first published online: 11 April 2011
Bladder urothelial carcinoma has high rates of mortality and morbidity. Identifying novel molecular prognostic factors and targets of therapy is crucial. Mammalian target of rapamycin (mTOR) pathway plays a pivotal role in establishing cell shape, migration, and proliferation.
Tissue microarrays were constructed from 132 cystectomies (1994-2002). Immunohistochemistry was performed for Pten, c-myc, p27, phosphorylated (phos)Akt, phosS6, and 4E-BP1. Markers were evaluated for pattern, percentage, and intensity of staining.
Mean length of follow-up was 62.6 months (range, 1-182 months). Disease progression, overall survival (OS), and disease-specific survival (DSS) rates were 42%, 60%, and 68%, respectively. Pten showed loss of expression in 35% of bladder urothelial carcinoma. All markers showed lower expression in invasive bladder urothelial carcinoma compared with benign urothelium with the exception of 4E-BP1. Pten, p27, phosAkt, phosS6, and 4E-BP1 expression correlated with pathologic stage (pathological stage; P<.03). Pten, 4E-BP1, and phosAkt expression correlated with divergent aggressive histology and invasion. phosS6 expression inversely predicted OS (P = .01), DSS (P = .001), and progression (P = .05). c-myc expression inversely predicted progression (P = .01). In a multivariate analysis model that included TNM stage grouping, divergent aggressive histology, concomitant carcinoma in situ, phosS6, and c-myc expression, phosS6 was an independent predictor of DSS (P = .03; hazard ratio [HR], −0.19), whereas c-myc was an independent predictor of progression (P = .02; HR, −0.38). In a second model substituting organ-confined disease and lymph node status for TNM stage grouping, phosS6 and c-myc remained independent predictors of DSS (P = .03; HR, −0.21) and progression (P = .03; HR, −0.34), respectively.
Despite recent multidisciplinary advances in its treatment, urothelial carcinoma of the bladder continues to carry unacceptably high rates of mortality and morbidity, with a 10-year survival rate of 40% to 50%.1 Variable biologic behavior in bladder urothelial carcinoma appears to be related to differences in oncogenic pathways alterations. Whereas superficial papillary neoplasms are driven by gain-of-function mutations in oncogenes such as H-RAS and FGFR3, flat carcinoma in situ and muscle-invasive tumors usually carry loss-of-function mutations, affecting tumor suppressor genes such as p53, phosphatase, and tensin homologue (PTEN).2 Improvements in understanding bladder urothelial carcinoma oncogenesis have fueled a current search for biological markers that can be targeted for therapy or bare prognostic significance.
Inactivation of PTEN tumor suppressor gene triggers the phosphatidylinositol 3′-kinase (PI3K)-protein kinase B (AKT) pathway that leads to Akt phosphorylation and activation (phosAkt). AKT plays a central role in orchestrating interaction between different growth regulating pathways and represents a major feedback control point.3, 4 PhosAkt promotes cell cycle progression through p27kip1 (p27) depletion,5 cell proliferation through c-myc up-regulation,6 and protein translation through mammalian target of rapamycin (mTOR) activation via its main downstream effectors, phosphorylated S6 protein (phosS6) and eukaryotic translation initiation factor 4E-binding protein-1 (4E-BP1).7
The current study evaluates the expression status and reciprocal interplay of 6 of the above biomarkers (Pten, phosAkt, phosS6, 4E-BP1, p27, and c-myc), aiming to be the first to evaluate mTOR pathway status as it relates to outcome in a well-characterized uniform cohort of bladder urothelial carcinoma treated by cystectomy.
MATERIALS AND METHODS
Patient Cohort and Tissue Microarray Construction
We retrieved 144 cystectomy specimens performed at the Johns Hopkins Hospital between 1994 and 2002. All sections were reviewed for confirmation of original diagnoses and staged, according to the 2002 American Joint Committee on Cancer-TNM classification,8 by 2 urologic pathologists (R.A. and G.J.N.). Paraffin blocks were available in 132 cases, including 120 pure bladder urothelial carcinoma cases and 12 bladder urothelial carcinoma cases with divergent aggressive histology. The latter included 2 tumors with extensive squamous differentiation, 5 with sarcomatoid features (carcinosarcoma), 1 micropapillary urothelial carcinoma, and 1 urothelial carcinoma with plasmacytoid features and 3 undifferentiated carcinomas. Tissue microarrays (TMAs) were constructed using a Beecher (Silver Spring, Md) instrument as previously described by Fedor and De Marzo.9 Triplicate tumor samples and paired benign urothelium were spotted from each specimen. Each TMA spot was further categorized as invasive bladder urothelial carcinoma, high-grade noninvasive papillary bladder urothelial carcinoma, or carcinoma in situ (CIS).
All pertinent clinicopathological data were retrieved from electronic medical records. These included patient demographics and preoperative information such as diagnostic procedure, precystectomy treatment, and clinical stage. Follow-up data on disease progression, postoperative chemotherapy, and/or radiotherapy, disease-specific survival (DSS), and overall survival (OS) were also obtained (Table 1). Because all patients underwent cystectomy with curative intent, pelvic recurrence as well as recurrent metastatic disease were considered progression events. Three cases that recurred in urinary tract locations other than bladder (eg, renal pelvis) were excluded from the progression and survival analyses. Both tumor pathological stage and TNM stage grouping were analyzed during statistical analysis.
Table 1. Demographic and Clinicopathological Characteristics of 132 Cystectomy Patients
TMA indicates tissue microarray.
Age at cystectomy, y
Pathologic stage at cystectomy
TMA spot sampling
Noninvasive papillary/carcinoma in situ component
Radiotherapy or systemic chemotherapy
Standard immunohistochemistry (IHC) analysis was performed for mTOR pathway members Pten, phosAkt, phosS6, and 4E-BP1. IHC analysis was also performed for AKT regulated markers c-myc and p27.
Immunostaining was performed on formalin-fixed paraffin embedded tissue sections using a Bond Max-Leica autostainer (Leica Microsystems, Bannockburn, Ill). Sections were deparaffinized, rehydrated, and subjected to heat-induced antigen retrieval with a buffer solution using a steamer. Sections were then incubated with appropriate primary antibody. After the application of a secondary polyclonal rabbit antibody (except for c-myc, for which we used a CSA kit), slides were developed using 3-3′-diaminobenzidine chromogen and counterstained with hematoxylin. Table 2 lists all pertinent markers information, including vendor, clone, dilution, pretreatment and incubation conditions, and detection kit. TMA spots with artifactual folds or lacking tissue target representation were omitted from further analysis. The latter accounted for any variability in number of total evaluable spots/cases among markers. Tumor and benign TMA spots stained with each marker were evaluated for pattern of staining (nuclear vs cytoplasmic), extent (percent of positive cells), and intensity (0 to 3+ score). A final H-score was generated for each marker as the sum of the products of each intensity category × extent of immunoexpression. H-scores were used during statistical analyses for all markers. In addition, Pten analysis was performed using extent scores. During multivariate analysis, a cutoff was used for phosS6 expression based on mean tumor H-score (H-score ≥27) and for c-myc expression based on the 90th percentile tumor H-score (H-score ≥15).
Table 2. Summary of Antibodies Used in the Immunohistochemical Analysis of mTOR Pathway
Findings were analyzed using the Stata 9.2 (StataCorp, College Station, Tex) software package. Equality of population medians among groups was tested using Kruskal-Wallis test for nonparametric 1-way analysis of variance (ANOVA) by ranks. ANOVA analysis of variance was used when comparing categorical variables. Pairwise correlation coefficients were calculated to test relations among variables. A Cox regression model was used during multivariate analysis.
Mean patient age at cystectomy was 68 years; the male to female ratio was 4:1, and mean length of follow-up was 62.6 months (range, 1-182 months). In our cohort, disease progression rate was 42% (25% local recurrence and 17% distant metastases). Overall survival (OS) and disease specific survival (DSS) rates were 60% and 68%, respectively.
Fifty (38%) patients received preoperative intravesical therapy. Two patients received mitomycin, and the remaining 48 were treated with Bacillus Calmette-Guérin. Six (4%) patients received neoadjuvant systemic chemotherapy and/or neoadjuvant radiotherapy.
The TMA sampling was limited to noninvasive bladder urothelial carcinoma in 27 cases. These included 12 pTa/pTis cystectomies and 15 (11%) higher stage lesions where the invasive component was no longer present in the TMA spot.
Biomarkers Expression Status in Relation to Clinicopathological Parameters
Expression levels of all 6 mTOR pathway-related biomarkers are summarized in Table 3 and depicted in Figure 1.
Table 3. Summary of All 6 mTOR-Related Biomarkers' Expression (Mean H-Score) in the Categorized Pathological Parameters
Nuclear and or cytoplasmic Pten expression was evaluated in 119 tumors. Pten expression levels were significantly lower in bladder urothelial carcinoma compared with benign urothelium (H-score, P = .0001; extent, P = .0000). Complete lack of Pten expression was seen in 50 (42%) of 119 bladder urothelial carcinomas. Additional 9 (7.5%) tumors showed low Pten expression levels (extent, <20%). All corresponding benign urothelium showed high Pten expression (extent, >50%).
On univariate analysis, significantly lower Pten expression extent was seen in tumor TMA spots with invasion compared with noninvasive spots (P = .0004) and in tumors of divergent aggressive histology compared with pure bladder urothelial carcinoma (P = .009). Pten expression showed an inverse correlation with pathological stage and TNM stage grouping (P = .0001 and P = .0004, respectively).
Nuclear phosAkt was evaluated in 113 tumors. PhosAkt expression was significantly lower in bladder urothelial carcinoma compared with benign urothelium (P = .0000). Fifty (44%) tumors were negative for phosAkt, and an additional 20 (18%) bladder urothelial carcinomas showed low phosAkt expression (H-score ≤5).
On univariate analysis, lower levels of phosAkt expression significantly correlated with presence of invasion at TMA spot (P = .03) and divergent aggressive histology (P = .04). PhosAkt expression inversely correlated with pathological stage (P = .0316). Noninvasive bladder urothelial carcinoma (pTa/pTis) had significantly higher levels of phosAkt compared with invasive (pathological stage 1+) bladder urothelial carcinoma (P = .004).
Cytoplasmic phosS6 was evaluated in 114 bladder urothelial carcinomas. Forty-two (37%) tumors were negative for phosS6, and an additional 10 (9%) tumors showed low phosS6 expression (H-score ≤5). phosS6 expression was significantly lower in bladder urothelial carcinoma compared with benign urothelium (P = .0000).
On univariate analysis, phosS6 expression did not correlate with presence of invasion at the TMA spot or with divergent aggressive histology (P = not significant [NS]). However, phosS6 expression inversely correlated with pathological stage (P = .02). Non–muscle-invasive bladder urothelial carcinoma (pathological stage <2) had higher phosS6 levels compared with pathological stage ≥2 (P = .0030).
Cytoplasmic 4E-BP1 was evaluated in 114 bladder urothelial carcinomas. Thirty-seven (32%) tumors were negative for 4E-BP1. We found significantly higher levels of 4E-BP1 expression in bladder urothelial carcinoma compared with benign urothelium (P = .0000).
On univariate analysis, significantly lower 4E-BP1 expression levels were seen in tumor TMA spots with invasion compared with noninvasive spots (P = .005) and in tumors of divergent aggressive histology compared with pure bladder urothelial carcinoma (P = .004). 4E-BP1 expression did not correlate with overall pathological stage or TNM stage grouping (P = NS). A trend toward lower expression levels in noninvasive bladder urothelial carcinoma (pTa/pTis) was found compared with pathological stage ≥1 (P = .07).
Nuclear p27 was evaluated in 115 tumors. One hundred three (90%) of 115 tumors were positive for the marker. p27 expression levels were significantly lower in bladder urothelial carcinoma compared with benign urothelium (P = .002).
On univariate analysis, tumors with divergent aggressive histology had significantly lower levels of p27 (P = .04). p27 showed an inverse correlation with pathological stage and TNM stage grouping (P = .04 and P = .003, respectively).
Nuclear c-myc was evaluated in 114 tumors. Seventy-seven (67.5%) of 114 bladder urothelial carcinomas were negative for c-myc, and an additional 16 (14%) tumors had low levels (H-score ≤5) of c-myc expression. c-myc expression was significantly lower in bladder urothelial carcinoma compared with benign urothelium (P = .0000).
On univariate analysis, lower c-myc expression correlated with presence of invasion at TMA spot (P = .002).
Correlation among tested biomarkers
A weak but statistically significant positive correlation was present between phosAkt and Pten bladder urothelial carcinoma expression levels (coefficient of correlation [cc], 0.19; P = 0.03) as well as between phosAkt and p27 (cc, 0.25; P = .008) bladder urothelial carcinoma expression levels. Pten and p27 bladder urothelial carcinoma expression also showed positive correlation (cc, 0.29; P = .001). Bladder urothelial carcinoma phosS6 expression showed strong positive correlation with bladder urothelial carcinoma c-myc expression (cc, 0.42; P = .000) and only weak positive correlation with p27 (cc, 0.20; P = .03) bladder urothelial carcinoma expression levels.
Univariate and Multivariate Outcome Analyses
On univariate analysis, TNM stage grouping and presence of divergent aggressive histology significantly predicted DSS, OS, and disease progression. Concomitant CIS predicted OS (P = .008). Pathological stage alone was of borderline significance in predicting outcome (Table 4). When pathological stage was categorized as organ confined (pathological stage <3) versus non-organ confined (pathological stage ≥3), organ-confined status significantly predicted all 3 outcome parameters (DSS, OS, and disease progression). Presence of lymphovascular invasion as well as nodal status were also predictive of outcome (see Table 4). The ratio of number of positive nodes to total number of examined nodes (positive lymph node density) was only of borderline significance in predicting disease progression (P = .07) and DSS (P = .06). Preoperative therapy, including neoadjuvant chemotherapy or radiation, did not correlate with any outcome parameter (P = NS).
Table 4. Results of Univariate Analysis of Clinicopathological Parameters and the Expression Levels of 6 mTOR-Related Biomarkers in Relation to DSS, OS, and Disease Progression
To evaluate the prognostic role of the above clinicopathologic parameters before factoring in the input of mTOR pathway markers, 2 multivariate models were adopted (Table 5). The first included TNM stage grouping, presence of concomitant CIS, and aggressive histology; TNM stage grouping remained an independent predictor of DSS, OS, and disease progression (P = .001, P = .001, and P = .007, respectively), and concomitant CIS remained a predictor of OS (P = .04) in this first model. In a second clinicopathologic parameter model that substituted organ-confined status and lymph node status for TNM stage grouping, only nodal status remained an independent predictor of disease progression (P = .03).
Table 5. Two Multivariate Cox Regression Analysis Models Correlating Clinicopathological Parameters and Expression Levels of mTOR Related Biomarkers With DSS, OS, and Disease Progression
Model 1 includes TNM stage grouping, whereas Model 2 includes organ-confined disease status and lymph node status as a substitute. Both models include aggressive histology concomitant CIS, phosS6, and c-myc.
P=.03; HR, −0.19; 95% CI, −0.37 to −0.01
P=.02; HR, −0.38; 95% CI, −0.70 to −0.06
TNM stage grouping
P=.004; HR, 0.11; 95% CI, 0.03 to 0.18
P=.01; HR, 0.10; 95% CI, 0.02 to 0.18
Divergent aggressive histology
P=0.03; HR, −0.21; 95% CI, −0.40 to −0.02
P=.03; HR, −0.34; 95% CI, −0.67 to −0.02
Organ-confined disease status
P=.03; HR, 0.22; 95% CI, 0.02 to 0.42
Lymph node status
Divergent aggressive histology
Clinicopathological parameters and mTOR-related biomarkers expression
Among all 6 tested biomarkers, phosS6 was a significant predictor of DSS (P = .001), OS (P = .01), and disease progression (P = .05), whereas c-myc expression was a significant predictor of disease progression (P = .01) but not of DSS or OS (P = NS) on univariate analysis (see Table 4).
In the first of 2 multivariate analysis models, which included TNM stage grouping, presence of divergent aggressive histology, concomitant CIS, phosS6, and c-myc expression, phosS6 was an independent predictor of DSS (P = .03; hazard ratio [HR], −0.19), whereas c-myc was an independent predictor of progression (P = 0.02; HR, −0.38). In the second multivariate analysis model, we substituted organ-confined disease status and lymph node status for TNM stage grouping. phosS6 remained an independent predictor of DSS (P = .03; HR, −0.21), and c-myc remained an independent predictor of progression (P = .03; HR, −0.34).
mTOR pathway is a key regulator of protein translation and cell proliferation that has been shown to be up-regulated in several solid malignancies, including thyroid carcinoma,10 small cell carcinoma of lung,11 gastrointestinal tumors,12, 13 and clear cell renal cell carcinoma.14 The main downstream effectors of mTOR pathway (phosS6 and 4E-BP1) have been shown to be independent predictors of prognosis in several types of solid tumors, including renal cell,14 ovarian,15 liver,16-19 and mammary carcinomas.20
Although the search for prognostic biomarkers in bladder urothelial carcinoma initially focused on cell cycle regulators, p53, and retinoblastoma alterations,21, 22 more recent efforts have also included tyrosine kinase receptors, mTOR, and microenvironment interaction pathways23 as potential prognosticators. Several studies have addressed the prognostic potential of individual mTOR pathway members in bladder cancer.24-26 Comprehensive simultaneous assessment of key members of PI3K-mTOR pathway and its downstream affected biomarkers, in relation to clinicopathologic parameters and outcome, in a well-characterized cystectomy cohort is needed.
PTEN loss leading to activation of AKT/mTOR axis has been reported in other genitourinary tract tumors such as prostate adenocarcinoma,27, 28 renal cell carcinoma,29 and upper urinary tract bladder urothelial carcinoma.30 These findings offered a rationale for the limited but promising response to mTOR inhibitors therapy in such settings. However, Yoo et al,31 using a mouse model that conditionally deletes PTEN in urogenital epithelium, found AKT/mTOR pathway highly activated in prostate tumors, but not in bladder epithelium. Moreover, PTEN knockout mice were shown to develop urothelial carcinomas in the upper urinary tract, but not in the bladder.30 In contrast, synchronous deletions of both PTEN and p53 tumor suppressor genes have been shown to lead to the development of invasive bladder cancer in mouse models.32
Our findings of frequent loss of Pten (42%) and phosAkt (44%) expression in bladder urothelial carcinoma are in line with prior studies. Pten and phosAkt loss of expression has been previously reported in up to 50% and 70% of bladder urothelial carcinoma, respectively, and lower Pten expression has been linked to invasive behavior.24-26 In our cohort, both Pten and phosAkt expression levels inversely correlated with pathological stage and aggressive divergent histology, and their loss was more likely to be seen in invasive compared with noninvasive tumor component within a given TMA spot. Our findings of lack of correlation of either Pten or phosAkt tumor expression with disease outcome have also been previously illustrated by Harris et al and others.26, 33 The generally lower levels of mTOR pathway members in bladder urothelial carcinoma with divergent histology suggests down-regulation of mTOR pathway in less differentiated urothelial tumors.
The positive but weak correlation between Pten and phosAkt expression is intriguing and may suggest a nonconditional association between the loss of inhibitory effect of PTEN tumor suppressor gene and phosAkt expression (Akt activation) in bladder urothelial carcinoma. Bose et al34 found no correlation between Pten and phosAkt expression in mammary carcinoma, despite finding both markers to be altered, suggesting that Akt activation is not always directly linked to PTEN loss.
AKT regulates its downstream target mTOR, which in turns operates through 2 distinct complexes: mTORC1 and mTORC2. mTORC2 directly activates AKT in a feedback fashion, whereras mTORC1 pathway regulates cell growth through its main downstream effectors: ribosomal S6 kinase and 4E-BP1. Activation of mTORC1 is thought to stimulate translation through phosphorylation of S6 and inhibition of 4E-BP1.35 Our findings of lower phosS6 and higher 4E-BP1 expression levels in bladder urothelial carcinoma compared with benign urothelium are consistent with the lower activation of AKT found in our bladder urothelial carcinoma cohort, resulting in an overall down-regulation of mTORC1 downstream events. The inverse correlation between phosS6 expression and tumor pathological stage in our cohort is in keeping with our finding of a favorable prognostic effect of higher phosS6 levels in bladder urothelial carcinoma, but is in contrast with prior findings of unfavorable prognostic effect of phosS6 expression in other solid tumors.19
Our finding of down-regulation of phosS6 in more aggressive bladder urothelial carcinoma tumors could theoretically be related to their hypoxic tolerant phenotype,36 given prior evidence for down-regulation of mTOR pathway through hypoxia inducible factor 1 (HIF1)α activation in hypoxic states,37-39 and given the association of HIF1α overexpression with poor outcome in bladder urothelial carcinoma.40-44 We are in the process of evaluating HIF1α expression in the current cohort in relation to mTOR pathway expression, hoping to test this hypothesis.
In addition to promoting translation, phosS6 is recognized to repress PIK3-AKT pathway through the inhibition of insulin receptor substrates 1 and 2.7 Accordingly, we found phosS6 expression, but not 4E-BP1 to correlate with other AKT-regulated members p27 and c-myc. The strongest correlation among pathway members was between c-myc and phosS6. Furthermore, high c-myc expression was an independent predictor of disease progression in our cohort. These results are also in keeping with a recently suggested c-myc dependent mechanism for phosS6 translation.45
Lower p27 expression in our bladder urothelial carcinoma tumors and its inverse correlation with pathological stage and aggressive divergent histology are in agreement with prior reports.46-48 The significantly positive correlation between p27 and Pten expression levels in our bladder urothelial carcinoma cohort could be interpreted as additional evidence of a potential oncogenic role for down-regulation of p27 in PTEN deficient bladders, as has recently been pointed by Yoo et al.31 Unlike some of the prior studies on p27 in bladder urothelial carcinoma,46, 48 we did not find loss of p27 expression to be an unfavorable predictor of disease progression or DSS. One possible reason for the difference could be that these studies had a larger proportion of superficial tumors. Interestingly, recent studies have pointed to a favorable effect on outcome for loss of p27 in bladder urothelial carcinoma, especially in combination with other cell cycle markers.21, 49 Additional large cohort studies are needed to better resolve the prognostic role if any of p27 in bladder urothelial carcinoma patients.
In summary, our study represents the first attempt at simultaneous assessment of key members of PI3K-mTOR pathway and its downstream affected biomarkers, in relation to clinicopathologic parameters and outcome. We found phosS6 expression to be an independent predictor of DSS (P = .03) and c-myc expression to be an independent predictor of disease progression (P = .02) in addition to TNM pathologic stage grouping in our cohort. Our intriguing novel findings of a statistically significant prognostic role for phosS6 and c-myc in a multivariate model that included established clinicopathologic prognostic parameters are promising and certainly warrant further confirmation in an independent cystectomy cohort, preferably in a prospective setting and in combination with cell cycle markers evaluation. Additional studies to address the potential value of mTOR pathway markers expression in predicting therapeutic response to mTOR inhibitors are also warranted.
We thank Helen Fedor and Marcella Southerland, Johns Hopkins Tissue Microarray Core Facility.
CONFLICT OF INTEREST DISCLOSURES
Supported in part by National Cancer Institute/National Institutes of Health grant PO1# CA077664.