1. Introduction
Urothelial carcinoma accounts for 90% of all bladder cancer cases [
1]. Approximately 75% of cases are primarily identified as NMIBC [
2], whereas 50 to 70% of NMIBC patients experience tumor recurrence, and up to 30% of them progress to muscle-invasive disease (MIBC) [
3]. Due to its high recurrence rate and intensive follow-up, bladder cancer is one of the most expensive cancer types to treat [
4]. In NMIBC, the commonly used scoring system is the “European Association of Urology (EAU) Risk Stratification for NMIBC”. This system categorizes patients into low-, intermediate-, high-, and very-high-risk groups for treatment, considering various histopathological parameters, such as TNM stage, WHO04/16 grade, and the presence of carcinoma in situ (CIS) [
5]. Bacillus Calmette–Guérin (BCG) immunotherapy is the standard treatment for intermediate- and high-risk patients with NMIBC. It prevents tumor recurrence and stage progression by activating anti-cancer immunity. However, up to 40% of patients fail to respond to BCG treatment, experiencing tumor recurrence and stage progression, while only suffering from severe side effects [
3,
6]. Therefore, it is important to distinguish between BCG responders and BCG non-responders. However, the current TNM and WHO04/16 classifications for bladder cancer have limitations when it comes to predicting disease course and treatment response [
5]. To overcome this challenge, there is a need for an improved classification approach. Combining genetic alterations and histopathological features could improve the risk stratification of patients with NMIBC.
Urothelial carcinomas are highly heterogeneous tumors with a high mutation rate [
7,
8]. Previous studies in bladder cancer have found significant associations between genetic alterations, histopathological parameters, and clinical outcomes [
9]. The papillary and non-papillary molecular classification system, which distinguishes tumors arising from hyperplasia and carcinoma in situ (CIS), respectively, is widely recognized for its distinct genetic alterations and clinical outcomes [
10]. In general, NMIBC has a more favorable prognosis compared to MIBC, although it has a high recurrence rate [
11]. It is well known that receptor tyrosine kinases (RTKs), particularly FGFRs, play a crucial role in the development of bladder cancer. Previously, it has been demonstrated that Ta LG tumors were often enriched in FGFR3 mutations and were associated with good prognosis [
11,
12]. However, FGFR3 mutations have also been linked to an increased risk of recurrence in NMIBC [
13]. Another member of the FGFR family, FGFR1, has been less extensively studied compared to FGFR3. FGFR1 alterations have been associated with epithelial–mesenchymal transition (EMT) and a higher stage and grade in NMIBC [
11,
14,
15]. Recently, other RTKs, such as ERBB2 and ERBB3, have gained attention in NMIBC as well. Although ERBB2 is best known for its prognostic and predictive roles in breast cancer, studies have also emphasized its prevalence in bladder cancer. In NMIBC, ERBB2 alterations were associated with a higher tumor stage and grade and linked to worse PFS [
16,
17,
18]. Furthermore, others have identified two luminal subgroups in NMIBC: FGFR3-enriched tumors with a favorable prognosis and ERBB2/ERBB3-enriched tumors with poorer outcomes [
19,
20]. Moreover, cell cycle dysregulation is one of the main drivers in tumorigenesis and metastasis. The MYC family of oncogenes, including MYC and MYCN, promotes tumor growth and metastasis by upregulating key cell cycle genes, such as CCND1 [
21]. Alterations in MYC and CCND1 have been frequently observed in bladder cancer, correlating with aggressive phenotypes and worse prognosis [
22,
23], while MYCN alterations were most common in neuroblastomas [
24]. Additionally, TP53 mutations have been detected in both the luminal subtype of NMIBC and the basal subtype of MIBC, associated with adverse outcomes [
25]. Previous research has shown that TP53 mutations were mutually exclusive with FGFR3 mutations in bladder cancer [
26], whereas their co-occurrence with ERBB2 alterations was associated with a worse prognosis in breast cancer [
27]. Genetic alterations have also impacted the phosphatidylinositol 3-kinase (PI3K) signaling pathways in NMIBC [
28]. It has been reported that PIK3CA was frequently mutated in NMIBC, especially in LG tumors, and was often associated with FGFR3 mutations and better patient outcomes [
29,
30,
31]. Our previous research has shown that high levels of CD25+ T regulatory cells (Tregs) were associated with shorter PFS in NMIBC [
32]. Furthermore, studies have reported that FGFR3 and MYC overexpression led to an increase in Tregs, inducing an immunosuppressive tumor microenvironment (TME) [
33], while ERBB2 overexpression was linked to a decrease in Tregs [
34,
35,
36]. Cytokeratins are intermediate filaments that indicate the differentiation status of tumor cells. Specifically, KRT20 alterations have been associated with luminal subtypes in both NMIBC and MIBC, indicating a worse prognosis in NMIBC but a better prognosis in MIBC [
10]. In our study, we used p53 and CK20 as surrogate markers for TP53 and KRT20, respectively [
37].
The primary objective of this study is to investigate the association between molecular alterations, such as single-nucleotide variants (SNVs), short insertions and deletions (INDELs), and copy number variants (CNVs), and both stage progression and tumor recurrence in a retrospective, single-institution, population-based study of NMIBC patients. Additionally, we aim to enhance the risk stratification of NMIBC patients by integrating genetic alterations with histopathological, IHC, and proliferation features.
4. Discussion
We aimed to investigate the association between genetic alterations and histopathological and IHC features to improve the risk stratification of patients with NMIBC.
In our study, distinct genetic profiles within Ta, T1, LG, and HG tumors were identified. We found that FGFR3 alterations were significantly associated with Ta LG tumors. This aligns with prior studies, showing that FGFR3 alterations were more prevalent in less aggressive, early-stage bladder cancers [
11,
47]. Additionally, our results showed a significant association between ERBB2 alterations and T1 HG tumors. Clinical enrichment analysis further revealed that ERBB2 alterations were associated with high expression of CK20, p53, and a high proliferation rate. Hedegaard et al. similarly reported that Ta LG tumors were enriched with FGFR3 alterations, while T1 HG tumors were enriched with KRT20, TP53, and ERBB2 alterations [
19]. Although PIK3CA alterations were more common in Ta tumors in our cohort, as confirmed by others [
28], we found no significant association with the WHO04/16 grade. Similarly, Duenas et al. also did not find any association between PIK3CA alterations and tumor grade. However, they found a higher frequency of PIK3CA alterations in T1 tumors, with a Ta:T1 ratio of 1:1, suggesting an important role in tumor aggressiveness [
29]. Discrepancies between our findings and theirs may be due to the differences in the ratio of Ta:T1 stages, which was 4:1 in our study. Moreover, FGFR1 and CCND1 alterations were found in 3% and 8% of our cohort, respectively. FGFR1 alterations promote tumor invasion through EMT mechanisms [
48], while CCND1 regulates cell cycle progression through cyclin pathways [
49]. In our cohort, HG tumors had a significantly higher frequency of FGFR1 and CCND1 alterations compared to LG tumors. Previous studies have also confirmed that FGFR1 and CCND1 alterations were linked to HG tumors [
15,
50]. In somatic interaction analysis, FGFR1 and CCND1 were found to be significantly mutually exclusive with FGFR3 alterations. These findings may indicate that alterations in FGFR1 and CCND1 represent early events in tumorigenesis and act as drivers for tumor progression in NMIBC.
In the assessment of stage progression, FGFR3 alterations were associated with improved PFS, while alterations in ERBB2, MYC, and MYCN were linked to shorter PFS. Others have also reported that FGFR3 expression is associated with better PFS in NMIBC, while frequent amplifications of ERBB2 and MYC were linked to poor prognosis [
51,
52]. In line with other publications, our results revealed mutual exclusivity between ERBB2 and FGFR3 [
17]. To the best of our knowledge, limited literature has investigated the association between MYCN and patient outcomes in NMIBC. MYCN, a transcription factor in the MYC family, is known to be frequently amplified in neuroblastoma, causing an aggressive disease course [
53]. In NMIBC, Hedegaard et al. published that MYCN alterations were associated with cell cycle process and a more aggressive luminal subtype [
19]. Consistent with other findings, we could not find a significant association between PIK3CA alterations alone and PFS [
30]. However, PIK3CA in combination with ERBB2 or MYC was inversely associated with PFS. Although PIK3CA alterations are common in bladder cancer, their prognostic significance is unclear and may depend on interactions with other genetic alterations [
54]. These findings suggest that identifying gene sets may more accurately capture the biological complexity of patient outcomes, providing greater predictive power than single-gene analyses. In the assessment of tumor recurrence, FGFR3 alterations were significantly associated with shorter RFS. Some studies have also confirmed that FGFR3 alterations were linked to worse RFS, but better PFS [
52,
55]. Additionally, RTKs within the RTK-RAS pathway, including FGFRs, ERBBs, PDGFRA, RET, ALK, and MET, were also linked to shorter RFS. This finding may be due to a potential overrepresentation of FGFRs among RTKs. In summary, FGFR3 alterations play a crucial role in tumor initiation but may not be the primary drivers of stage progression. Instead, there seems to be a potential shift toward ERBB2 dominance in higher-stage tumors.
Recent efforts have focused on the molecular classification of NMIBC. Hurst et al. emphasized the need to distinguish between Ta (non-invasive) and T1 (submucosal invasive) tumors due to significant molecular heterogeneity within these stages [
9]. Our analysis identified two subgroups in Ta tumors: one with FGFR3 alterations linked to LG tumors and favorable outcomes, and another with alterations in ERBB2, PIK3CA, and ERBB3 mutations, and FGFR1, CCND1, and MYC amplifications, associated mostly with HG tumors and poorer prognosis. In T1 tumors, FGFR3 alterations were linked to worse RFS, and MYCN amplifications were associated with worse PFS. Our findings are similar to the molecular subtypes identified by the Aarhus group classification and the Lund taxonomy applied to NMIBC: Class I/UroA is enriched with FGFR3, Class 2a and 2b with ERBB2, ERBB3, and FGFR1, and UroC/Genomically Unstable (GU) with MYCN and PIK3CA alterations [
56].
The EAU guidelines recommend intravesical BCG as the primary treatment for intermediate- and high-risk NMIBC [
5]. However, the biological mechanisms driving tumor recurrences are still not well understood. In our BCG-treated patient group, FGFR3 and PIK3CA alterations were significantly associated with shorter RFS and PFS, respectively. Previous publications have demonstrated that downregulation of FGFR3 was associated with BCG responders [
57], while FGFR3 overexpression was linked to BCG non-responders [
58]. In accordance with prior studies, our results suggest a significant association between FGFR3 and PIK3CA alterations, and poor response to BCG treatment [
59,
60]. Interestingly, our study found that genetic alterations in AR were associated with shorter RFS, a surprising result given that AR amplifications are typically associated with castration-resistant prostate cancer [
61]. However, research has demonstrated that AR and the androgen signaling pathway play a role in the etiology and progression of bladder cancer [
62], and combining antiandrogens with BCG immunotherapy could improve its efficacy [
63,
64,
65]. Furthermore, the current BCG shortage and issues of BCG unresponsiveness emphasize the urgent need for better treatment options. Frequent alterations in RTKs make them promising targets for more personalized treatments [
66]. Upon analyzing amino acid variants of FGFR3, the most common variants were Ser249Cys and Arg248Cys in the extracellular domain, and Tyr373Cys in the transmembrane domain, causing constant downstream signaling without the presence of ligands. Recently FDA-approved, erdafitinib, a pan-FGFR (FGFR1-4) tyrosine kinase inhibitor (TKI), has shown efficacy in treating advanced bladder cancer and intermediate- and high-risk NMIBC by blocking tyrosine kinase signaling [
67,
68]. In our cohort, ERBB2 amplification was more prevalent compared to missense alterations, consistent with observations in other cancer types [
69,
70]. Despite success in breast and gastric cancers, clinical trials with anti-ERBB2 treatments have shown limited efficacy in bladder cancer. Furthermore, it has been previously reported that inhibition of FGFR3 can trigger compensatory ERBB2 signaling, emphasizing the importance of targeting both pathways simultaneously [
71]. In our study, the most frequent variants of PIK3CA were Glu545Lys, Glu542Lys, and His1047Arg. These alterations are associated with the catalytic subunit encoded by the PIK3CA gene and can lead to constant activation of the PI3K pathway [
72]. In general, PI3K inhibitors, such as alpelisib, are effective treatment options in bladder cancer but can cause very severe side effects [
73].
In the cancer genome, SBS signatures reveal distinct patterns of alterations associated with specific mutagenic processes. In our analysis of mutational signatures, we observed that C > T (35%) was the most frequent base substitution linked to the APOBEC mutational signature. It has been similarly documented that in NMIBC, the APOBEC mutational signature contributes up to 30% of all alterations [
74,
75]. APOBEC enzymes induce hypermutation by deaminating cytosine bases, resulting in C > T and C > G substitutions [
76]. Moreover, we found that the APOBEC mutational signature was significantly associated with PIK3CA alterations. In bladder cancer, previous reports have also observed that C > T base substitutions were most common in FGFR3 and PIK3CA [
77]. Additionally, we did not find a significant association between the APOBEC mutational signature and either RFS or PFS. However, others have demonstrated that enrichment of APOBEC mutational signatures was linked to high-risk NMIBC and poor prognosis [
19,
20]. Furthermore, we also identified mutational signatures, such as SBS15 and SBS7, which were associated with defective DNA mismatch repair and smoking mutational processes, respectively. However, our mutational signature results must be interpreted carefully, as SBS signatures are based on whole-genome sequencing (WGS) data, whereas our study used a targeted NGS panel, resulting in fewer alterations.
To the best of our knowledge, our study is among the largest single-institution, population-based studies focused exclusively on patients with NMIBC and is the largest to separately analyze Ta tumors (n = 227). Although we examined a limited number of genes, we validated previous findings and identified key tumor-driver genes in NMIBC. A significant strength of our research is the utilization of a well-established NGS panel that is routinely used in diagnostic settings. This enhances the clinical relevance of our findings and underscores their potential for integration into standard clinical practice.
In NMIBC, differentiating between LG and HG tumors is challenging because there are no standardized and reproducible methods available. In our research, we could differentiate between Ta and T1, as well as LG and HG tumors, based on their distinct molecular profiles. Previously, both our study and others have shown a significant association between IHC markers and proliferation features, and TNM stage, WHO04/16 grade, and patient outcomes [
32,
39,
78]. We further established that IHC markers and MAI were significantly linked to distinct genetic alterations. Confirming our earlier findings, MAI was the best predictor for stage progression and was associated with a more aggressive molecular phenotype. Moreover, we also found that distinct molecular profiles were associated with RFS and PFS. In summary, our results confirm that integrating histopathological, IHC, and proliferation features with genetic alterations enhances the objective and reproducible classification of patients into appropriate risk groups. To enhance the robustness of our findings, validation with larger gene panels, including genes involved in cell cycle regulation (CDKN2A and RB1), chromatin remodeling (KDM6A, ARID1A, and STAG2), DNA repair (ERCC2), tumor suppression (MTAP), and metabolic pathways (PPARG), as well as expanded cohorts, is warranted.