High expression of ERBB2 is an independent risk factor for reduced recurrence-free survival in patients with stage T1 non-muscle-invasive bladder cancer

DeepKEGG: a multi-omics data integration framework with biological insights for cancer recurrence prediction and biomarker discovery

Deep learning-based multi-omics data integration methods have the capability to reveal the mechanisms of cancer development, discover cancer biomarkers and identify pathogenic targets. However, current methods ignore the potential correlations between samples in integrating multi-omics data. In addition, providing accurate biological explanations still poses significant challenges due to the complexity of deep learning models. Therefore, there is an urgent need for a deep learning-based multi-omics integration method to explore the potential correlations between samples and provide model interpretability. Herein, we propose a novel interpretable multi-omics data integration method (DeepKEGG) for cancer recurrence prediction and biomarker discovery. In DeepKEGG, a biological hierarchical module is designed for local connections of neuron nodes and model interpretability based on the biological relationship between genes/miRNAs and pathways. In addition, a pathway self-attention module is constructed to explore the correlation between different samples and generate the potential pathway feature representation for enhancing the prediction performance of the model. Lastly, an attribution-based feature importance calculation method is utilized to discover biomarkers related to cancer recurrence and provide a biological interpretation of the model. Experimental results demonstrate that DeepKEGG outperforms other state-of-the-art methods in 5-fold cross validation. Furthermore, case studies also indicate that DeepKEGG serves as an effective tool for biomarker discovery. The code is available at https://github.com/lanbiolab/DeepKEGG.

Homologous recombination mRNAs (RAD21, RAD50 and BARD1) have a potentially poor prognostic role in ERBB2-low bladder cancer patients

Human epidermal growth factor receptor 2 (HER2/ERBB2) factor is known to be implicated in many malignancies and the potential of it as a prognostic biomarker was reported years ago. Molecular subtypes of HER2/ERBB2 negative and positive with distinct clinical outcomes have been identified in recent years; however, it is still under investigation for bladder cancer. This study evaluates the biological and prognostic significance of RAD21, RAD50 and BARD1 (homologous recombination biomarkers) mRNA levels with ERBB2 low and high expression to explore their impact on bladder cancer patient survival and cancer aggressiveness. The expression of ERBB2, RAD21, RAD50 and BARD1 mRNA levels was assessed in The Cancer Genome Atlas (TCGA) bladder cancer dataset along with four validation cohorts. Outcome analysis was evaluated using disease-free survival (DFS) and overall survival (OS). Univariate and multivariate analysis were used to evaluate the relationship between RAD21, RAD50, BARD1 and ERBB2 expression and clinicopathological variables. A significant increase in mRNA expression levels of RAD21, RAD50 and BARD1 was noticed in ERBB2-low patients compared to ERBB2-high patients. This overexpression of the homologous recombination repair transcripts was associated with poor outcome in ERBB2-low tumors, not in ERBB2-high tumors. Furthermore, the combined expression of high RAD21/RAD50, high RAD21/BARD1 or high RAD50/BARD1 were significantly associated with worse DFS and a better outcome for those with low co-expression in the ERBB2-low cohort. High expression of either RAD21/RAD50 or RAD21/BARD1 in ERBB2-low cohort associated with higher chance of metastasis. In addition, gene expression of BARD1 alone or in combination with RAD50 acted as an independent prognostic factor for worst survival. The data presented in this study reveal a connection between RAD21, RAD50, BARD1 and ERBB2 and patient survival. Importantly, it provided novel findings and potential prognostic markers, particularly in ERBB2-low bladder cancer.

HER2 Expression in Bladder Cancer: A Focused View on Its Diagnostic, Prognostic, and Predictive Role

Bladder cancer (BC) is a heterogeneous disease from a molecular, morphological, and clinical standpoint. HER2 is a known oncogene involved in bladder carcinogenesis. Assessing HER2 overexpression as a result of its molecular changes in a routine pathology practice using immunohistochemistry might be a useful adjunct in several scenarios, namely (1) to correctly identify flat urothelial lesions and inverted urothelial lesions in the diagnostic setting; (2) to provide prognostic hints in both non-muscle invasive (NMI) and muscle invasive (MI) tumors, thus supplementing risk stratification tools, especially when evaluating higher-risk tumors such as those with variant morphology; (3) to improve antibody panels as a surrogate marker of BC molecular subtyping. Furthermore, the potential of HER2 as a therapeutic target has been only partly explored so far, in light of the ongoing development of novel target therapies.

Targeted Next-Generation Sequencing of Flat Urothelial Lesions Reveals Putative Pathobiological Pathways, Potential Biomarkers, and Rational Therapeutic Targets

Flat urothelial lesions are controversial diagnostic and prognostic urologic entities whose importance relies mainly on their ability to progress to muscle-invasive tumors via urothelial carcinoma in situ (CIS). However, the carcinogenetic progression of preneoplastic flat urothelial lesions is not well established. Moreover, predictive biomarkers and therapeutic targets of the highly recurrent and aggressive urothelial CIS lesion are lacking. Using a targeted next-generation sequencing (NGS) panel of 17 genes directly involved in bladder cancer pathogenesis, we investigated alterations of genes and pathways with clinical and carcinogenic implications on 119 samples of flat urothelium, including normal urothelium (n = 7), reactive atypia (n = 10), atypia of unknown significance ( n = 34), dysplasia ( n = 23), and CIS (n = 45). The majority of the flat lesions were tumor-associated but grossly/microscopically or temporally separated from the main tumor. Mutations were compared across flat lesions and concerning the concomitant urothelial tumor. Associations between genomic mutations and recurrence after intravesical bacillus Calmette-Guerin treatment were estimated with Cox regression analysis. TERT promoter mutations were highly prevalent in intraurothelial lesions but not in the normal or reactive urothelium, suggesting that it is a critical driver mutation in urothelial tumorigenesis. We found that synchronous atypia of unknown significance-dysplasia-CIS lesions without concomitant papillary urothelial carcinomas had a similar genomic profile that differed from atypia of unknown significance-dysplasia lesions associated with papillary urothelial carcinomas, which harbored significantly more FGFR3, ARID1A, and PIK3CA mutations. KRAS G12C and ERBB2 S310F/Y mutations were exclusively detected in CIS and were associated with recurrence after bacillus Calmette-Guerin treatment (P = .0006 and P = .01, respectively). This targeted NGS study revealed critical mutations involved in the carcinogenetic progression of flat lesions with putative pathobiological pathways. Importantly, KRAS G12C and ERBB2 S310F/Y mutations were identified as potential prognostic and therapeutic biomarkers for urothelial carcinoma.

The Prognostic Value of the Circulating Tumor Cell-Based Four mRNA Scoring System: A New Non-Invasive Setting for the Management of Bladder Cancer

Bladder cancer (BC) is one of the most expensive lifetime cancers to treat because of the high recurrence rate, repeated surgeries, and long-term cystoscopy monitoring and treatment. The lack of an accurate classification system predicting the risk of recurrence or progression leads to the search for new biomarkers and strategies. Our pilot study aimed to identify a prognostic gene signature in circulating tumor cells (CTCs) isolated by ScreenCell devices from muscle invasive and non-muscle invasive BC patients. Through the PubMed database and Cancer Genome Atlas dataset, a panel of 15 genes modulated in BC with respect to normal tissues was selected. Their expression was evaluated in CTCs and thanks to the univariate and multivariate Cox regression analysis, EGFR, TRPM4, TWIST1, and ZEB1 were recognized as prognostic biomarkers. Thereafter, by using the risk score model, we demonstrated that this 4-gene signature significantly grouped patients into high- and low-risk in terms of recurrence free survival (HR = 2.704, 95% CI = 1.010−7.313, Log-rank p < 0.050). Overall, we identified a new prognostic signature that directly impacted the prediction of recurrence, improving the choice of the best treatment for BC patients.

Combined exome and transcriptome sequencing of non-muscle-invasive bladder cancer: associations between genomic changes, expression subtypes, and clinical outcomes

BACKGROUND

Three-quarters of bladder cancer patients present with early-stage disease (non-muscle-invasive bladder cancer, NMIBC, UICC TNM stages Ta, T1 and Tis); however, most next-generation sequencing studies to date have concentrated on later-stage disease (muscle-invasive BC, stages T2+). We used exome and transcriptome sequencing to comprehensively characterise NMIBCs of all grades and stages to identify prognostic genes and pathways that could facilitate treatment decisions. Tumour grading is based upon microscopy and cellular appearances (grade 1 BCs are less aggressive, and grade 3 BCs are most aggressive), and we chose to also focus on the most clinically complex NMIBC subgroup, those patients with grade 3 pathological stage T1 (G3 pT1) disease.

METHODS

Whole-exome and RNA sequencing were performed in total on 96 primary NMIBCs including 22 G1 pTa, 14 G3 pTa and 53 G3 pT1s, with both exome and RNA sequencing data generated from 75 of these individual samples. Associations between genomic alterations, expression profiles and progression-free survival (PFS) were investigated.

RESULTS

NMIBCs clustered into 3 expression subtypes with different somatic alteration characteristics. Amplifications of ARNT and ERBB2 were significant indicators of worse PFS across all NMIBCs. High APOBEC mutagenesis and high tumour mutation burden were both potential indicators of better PFS in G3pT1 NMIBCs. The expression of individual genes was not prognostic in BCG-treated G3pT1 NMIBCs; however, downregulated interferon-alpha and gamma response pathways were significantly associated with worse PFS (adjusted p-value < 0.005).

CONCLUSIONS

Multi-omic data may facilitate better prognostication and selection of therapeutic interventions in patients with G3pT1 NMIBC. These findings demonstrate the potential for improving the management of high-risk NMIBC patients and warrant further prospective validation.