Chromosomal imbalances in human bladder urothelial carcinoma: similarities and differences between biopsy samples and cancer stem-like cells

Bladder cancer is the fourth most common cancer in men and the eighth in women in both incidences and mortality and over 90% of bladder tumors are transitional cell carcinomas (TCCs). The existence of two distinct groups of tumors with different clinical features is a remarkable feature of TCC. More than 70% are in fact low-grade (LG) non-infiltrating (NI) cancers at diagnosis, that can be treated endoscopically by transurethral resection (TUR) alone, recur at least once and 10-20% progress in stage and grade. Thus repetitive and costly follow-up based on urine cytology, cystoscopy and imaging studies of the urinary tract is required, even if the prognosis is usually good. On the other hand, about 20% of tumors show muscle invasion (IN) at diagnosis and have a poor prognosis with <50% survival after 5 years [1]. A model for at least two major pathways has emerged based on the existence of these two distinct groups of lesions [2, 3]. LG NI tumors are generally characterized by constitutive activation of the receptor tyrosine kinase–Ras pathway, and they have activating mutations in the proto-oncogene FGFR3[4, 5]; in contrast, inactivating mutations of the tumor suppressor pathways of TP53, RB1 or PTEN were found in muscle-invasive lesions [5, 6]. Array comparative genomic hybridization (aCGH) studies have been instrumental in delineating genomic regions that are targeted by copy number changes, called Copy Number Alterations (CNAs). Several aCGH studies of bladder cancer have been published to date and provide the identification of a number of genomic regions of DNA amplification that contain known or candidate oncogenes including cyclin D1 (CCND1) on 11q13 [7, 8], ERBB2 on 17q21 [9], MDM2 on 12q14–q15 [10], and E2F3 on 6p22 [11]. Similarly, deletions of genomic regions containing tumor suppressor genes, such as CDKN2A, DBC1 and TSC1 (at 9p21, 9q33 and 9q34, respectively), PTEN on 10q23, RB1 on 13q14, and TP53 on 17p13 [12]. Some of these aberrations have been associated with the pathological stage and/or outcome of bladder cancer. Several studies evidenced exclusive genomic alterations in LG and HG tumors [13], with a significant increase in CNAs and genomic instability with increasing stage/grade and with outcome [14, 15]. In addition, the over-representation of focal amplifications, such as at chromosome 6p22, was significantly associated with HG IN tumors [16] and recurrent cases [17].

Since it is currently believed that bladder cancer is derived from a common cancer stem cell (CSC) likely derived by transformation of urothelial cells of the basal layer, bladder CSCs have been isolated based on basal cell markers such as CD44. Nevertheless, in this way, CSCs have only been identified in HG IN tumors [18‐21] suggesting that a distinct progenitor cell type exists for LG NI. A recent study of Dancik et al. provides evidence of the existence of distinct progenitor cells in NI and IN tumors, supporting new conceptual framework for investigating and understanding bladder cancer [22]. CSCs are responsible for treatment failure and cancer recurrence since they exhibit specific stem cells features, such as growth as nonadherent spheres in a stem cell medium, unlimited self-renewal, multipotency and lineage-specific differentiation [23]. Understanding the origins and supporting mechanisms of these cells and their relation to the bulk population has a great relevance for improving the knowledge of cancer pathogenesis and therapeutics [23].

In the first step of this study we performed conventional chromosome analysis on TCC biopsies with different histotypes (LG vs HG) in order to detect rough differences between them. We subsequently performed aCGH analysis on another set of 20 biopsies to look for chromosomal imbalances and smaller differences. In a second step, we compared the global pattern of CNAs in 16 of these biopsies with the corresponding isolated CSCs in order to reveal specific genomic aberrations that would provide them with growth advantages and a more aggressive phenotype.

This study focuses on the complexity of the bladder cancer genome, and for the first time to our knowledge, on the possibility to compare genomic alterations of in vitro selected cancer stem-like cells and their original biopsy in order to identify different genomic signature already present in the early stages of tumorigenesis of LG and HG tumors. Despite the low number of cases enrolled and therefore, at the moment, the interpretation may be only hypothetical, the findings of this study may assume a very important significance to those with closely related research interest. Firstly, we performed a conventional karyotype on 20 biopsies (11 LG and 9 HG) in order to delineate the status of ploidy in bulk tumors. Despite this technique is far from being innovative, we observed that LG tumors generally have near diploid metaphases, while HG tumors have a tendency to triploidy with a greater dispersion in the number of chromosomes per cell. Furthermore, we revealed that only two samples out of nine, both non-invasive LG, show a strong tendency to Y chromosome loss. Conversely, the others maintain or, at most, show an opposite tendency to acquire multiple copies of this chromosome. Although the number of cases is very low, this observation is a little in contrast with a recent study showing that Y chromosome losses are equally frequent in urothelial bladder cancer of all grades and stages [29]. Our data seem to agree with another study that showed a significant association of Y polysomies with HG invasive tumors [30]. The overall observations obtained by conventional chromosomal analysis and FISH have confirmed a greater aggressiveness of HG tumors than LG ones; in addition, although a larger number of cases must be studied, we believe that the involvement of the Y chromosome has yet to be clarified and that it is no coincidence that men are more affected by this type of cancer [31].

In this work we present a comprehensive catalog of CNAs across 20 tumor fresh biopsies providing an overview of their common alterations. The overall data evidenced a general chromosomal instability, especially in HG tumors, with a general CNAs ratio of 4:1 respect to LG tumors, and a ratio of 10:1 considering only gains. A previous study by HR-CGH analysis had reached the same conclusion [13]. In addition, in the present work we unveiled an opposite situation analyzing the genomic profiles of CSCs, as in HG tumors they were less affected than in LG tumors, with a general ratio of 1:3 (1:6 if we consider only the losses). To understand this anomalous behavior it would be useful to compare the genomic alterations of the original biopsy with their isolated cells because this approach may focus on the alterations most involved in tumorigenesis of TCCs. The overview of CNAs per chromosome (Figure 3) evidences a better conservation between cells and biopsies of HG tumors than LG tumors, even if isolated cells of HG group are less altered than their original biopsies while the situation is reversed in LG group. In two HG samples the complex pattern of CNAs involving chromosome 6 is consistent with “a chromothripsis like event”. Morrison et al. have recently reported the chromothripsis phenomenon in muscle-invasive TCCs [32]. These authors postulated that chromothripsis is related to a defective replication-licensing complex and that it could lead to intratumoral mutational heterogeneity. However, a recent work suggests that more stringent criteria must be used to identify chromothripsis and that it cannot be distinguished from other complex genomic rearrangements [33]. According to these authors, great caution should be exercised when labeling complex rearrangements as chromothripsis from genome hybridization and sequencing experiments. In our study we identified several shared gain/amplifications in chromosome 6 between the biopsy and the isolated cells of the same tumor, providing evidence in favor of a non-progressive mechanism.

The differences between HG and LG tumors also emerged by GO analysis, especially for two functional GO classes: transcription and apoptosis. Additional file 10: Table S7 shows a preponderance of GO terms for transcription class derived from gained genes of biopsies and isolated cells of HG tumors. Although some LG tumors exhibited a similar behavior (i.e. 34 sample), statistical significance for these tumors is determined by lost genes, especially in isolated cells where it has been shown many lost regions. Similarly, GO terms for apoptosis class were derived from lost regions of isolated cells of LG tumors, while the same GO terms were linked to gained regions of isolated cells of HG tumors. Thanks to the comparative analysis between biopsies and CSCs isolated from them, we can speculate that the driving forces of tumorigenesis are quite different in HG and LG tumors, even showing a complementary behavior. Specifically, we confirmed a good correlation between the total number of CNAs and genomic instability with increasing stage and grade of the biopsy [15]. Moreover, we found that CSCs isolated from LG biopsies accumulate a disproportionate number of genomic losses, so the isolated cells would be much altered respect to their original biopsy. This phenomenon was not observed in HG tumors, then it would not seem the result of purely random alterations due to culture conditions, but to essential characteristics which diversify the two types of tumor. Furthermore, GO stat analysis and aCGH data evidenced a subgroup of LG tumors where this paradox seems to be more evident. It would be interesting to verify if this subgroup of LG tumors could have a more aggressive potential and a greater propensity to progress and invade.

In this study we provide not only an overview of changes in the CNAs of HG vs LG tumors, but for the first time to our knowledge we also make a comparison of the overall CNAs in biopsies and CSCs isolated from them. Our approach indicate that LG differ from HG regarding their respective genomic profile also in the early stage of tumorigenesis; moreover it has been identified a subgroup of LG samples in which the tendency to loss of genomic regions is significantly higher. These findings provide valuable information to deeper study TCC carcinogenesis and may be applicable in the clinic for the identification of patients who will benefit from a more aggressive therapy.