Predictive value of progression-related gene classifier in primary non-muscle invasive bladder cancer

Bladder cancer is a genetic disorder driven by the progressive accumulation of multiple genetic and epigenetic changes. At the molecular level, these genetic changes result in uncontrolled cell proliferation, decreased cell death, invasion, and metastasis. The specific alterations in gene expression that occur as a result of cross-talk between various cellular pathways determine the biologic behavior of the tumor, including growth, recurrence, progression and metastasis, and may influence patient's survival. While several molecular markers for the development, recurrence and progression of bladder cancer, such as p53 and Rb, have been studied [1‐3], the limited value of these established prognostic markers created the need for new molecular indicators of bladder cancer outcomes.

New high-throughput microarray technology makes it possible to gain comprehensive insight into the molecular basis of human diseases [4, 5]. With this technology, the RNA expression levels of hundreds or even thousands of genes in a tumor can be surveyed simultaneously. The use of high throughput technologies to assess gene expression patterns in tissues, exfoliated cells in urine, or molecules in serum and in circulating cells for many malignancies, including bladder cancer, has been reported [6, 7]. These studies open a door to the possibility of rapidly assessing gene expression patterns in individual tumors to determine tumor classification [8], or to predict clinical outcomes [9, 10] and response to chemotherapy [11, 12]. In fact, gene expression profiling is currently being tested in clinical trials to define populations of patients with breast cancer who should receive chemotherapy [10, 12]. Such trials were launched in Dutch academic centers and in the United States [13].

Many different genetic or epigenetic changes that lead to aberrant gene expression have been identified in bladder cancer [6, 7]. Thus, gene expression profiling in bladder cancer represents a potentially useful way to discriminate between good and poor prognosis. Microarray gene expression analysis could be used to facilitate the identification of molecular prognostic markers that correlate with bladder cancer outcomes. In the current study, we identified genetic signatures that are associated with disease progression in patients with non-muscle invasive bladder cancer (NMIBC).

Although there have been reports of the value of gene expression profiles for cancer prognosis [8‐12], only limited data are available on the prognostic value of expression profiles in human bladder cancer in relatively large-scale study populations with long-term follow-up [25‐27]. With regard to the microarray studies that have been performed in the bladder cancer field to date, Sanchez-Carbayo et al. [25] used cDNA microarrays to facilitate the classification scheme of diagnostic and prognostic utility for stratifying advanced bladder cancer. This technology allowed them to identify poor outcome profile could assist in selecting patients who may benefit from more aggressive therapeutic intervention. Additionally, Dyrskjot et al. [26, 27] also reported the clinical usefulness of molecular markers for the prediction the clinical course of patients with NMIBC. These studies enhance the importance of the genome-wide studies in bladder cancer fields. In concordance with earlier reports, our study should support the potential usefulness of microarray study in these fields [25‐27]. However, both the validity and the reproducibility of microarray-based clinical research have been challenged [28]. To develop and validate a method of classification, it is necessary to start with a sufficiently large set of samples to analyze an independent test set and a validation set [5]. Furthermore, the results of microarray analysis in general should be validated using other techniques for quantifying RNA expression and by an independent cohort to reduce the false discovery rate [29, 30]. In the present study, we used microarray data to identify several types of prognosis-related gene classifiers. We were able to validate progression-related gene classifier in primary NMIBC by RT-PCR in a relatively large-scale long-term follow-up independent study population, but not for the other prognosis-related gene classifiers. These findings emphasize the importance of validation to support the reliability of microarray-based results.

The prediction of disease progression for patients with bladder cancer is a major clinical challenge. A number of molecular markers of cancer that have been identified to date have been explored as predictors of disease progression. Some of these, such as p53, have been suggested to be independent markers, while others do not appear to have a role as prognostic indicators [1‐3]. Because multiple genetic alterations are required for the transformation of a normal cell into a cell with a malignant and ultimately metastatic phenotype, assessment of multiple markers as a whole might better describe the biological phenotype of a particular cancer.

We identified the following eight genes as predictors of progression in patients with NMIBC: S100 calcium binding protein A8 (S100A8), 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 4 (PFKFB4), heme oxygenase 1 (HMOX1), methylthioadenosine phosphorylase (MTAP), kinesin family member 1A (KIF1A), coagulation factor C homolog (COCH), EGF LAG seven-pass G-type receptor 3 (CELSR3) and chromosome 16 open reading frame 74 (MGC17624). Overexpression of S100A8 is associated with stage progression, invasion, metastasis and poor survival in human bladder cancer [31]. Thus, the expression of S100A8 in NMIBC could be an early indicator for a sub-group of tumors with a propensity for muscle invasion. PFKFB4 is strongly induced by hypoxia through an hypoxia inducible factor 1 alpha (HIF1A) subunit dependent mechanism, and might contribute significantly to the Warburg effect observed in malignant gastric, pancreatic, breast and colon tumors [32, 33]. HMOX1 is an essential enzyme in heme catabolism. HMOX1 participates in the processes of angiogenesis and vasculogenesis [34]. There are significant differences in the distribution of the HMOX1 genotype in patients with different-stage urothelial carcinoma [35]. MTAP encodes an enzyme that plays a major role in polyamine metabolism. Many cancers, including bladder cancer, are deficient of the MTAP enzyme because the gene is co-deleted along with the tumor suppressor p16 [36]. KIF1A, a member of the kinesin family, appears to play a critical role in the development of axonal neuropathies resulting from impaired axonal transport [37]. However, the function of KIF1A related to cancer is unknown. COCH is a cell adhesion molecule [38]. The specific functions of CELSR3 and MGC17624 have not been determined. Although the identification of genes linked to disease progression suggests potential therapeutic interventions based on their mechanism of action, the lack of a biological context for these genes does not diminish their potential as clinical biomarkers. Many biomarkers, such as prostate-specific antigen and carcinoembryonic antigen, have unknown functions but are nonetheless useful as diagnostic or prognostic markers for disease [5].

Because some genetic events occur early in the disease process or before progression, molecular diagnosis may enable the prediction of disease progression before the onset of symptoms or overt radiographic evidence. Thus, from a clinical point of view, the most promising application for gene classifiers is the early prediction of tumor progression. In addition, new relevant biomarkers must provide cost-effective, non-invasive monitoring of low-risk patients and identify high-risk refractory tumors before they progress. The main goal of all of these efforts is to develop more accurate prognostic tools that can be used to direct treatment [5]. In the current study, we identified a specific subset of genes that predicted progression in NMIBC. NMIBC is a heterogeneous group including patients with the enormous array of clinical and pathological risk factors involved, such as number of tumors, tumor size, prior recurrence rate, T-category, carcinoma in situ, grade, intravesical therapy, and other factors [2, 3, 39]. Thus, multivariate analysis including these factors needs to control confounding effects. However, in the present study, multivariate analysis of disease progression could not be carried out due to the fact that any patients with NMIBC in the good-prognosis signature group did not experience cancer progression. Nonetheless, this gene set of progression-related markers has favorable potential to contribute to the understanding of bladder cancer behavior and the future treatment of patients with urothelial cancer.

The progression-related gene classifier identified in the current study represents a potentially valuable tool for the stratification of heterogeneous bladder cancer patient populations into risk groups that can be used to guide clinical decision-making, including observation versus adjuvant therapy. We derived clear clinical results using gene classifier for progression in NMIBC. The results of the current study suggest that the gene expression profiling to determine progression can be used to provide customized clinical care.