Bioinformatic Approach to Identify Positive Prognostic TGFB2-Dependent and Negative Prognostic TGFB2-Independent Biomarkers for Breast Cancers

Breast cancer is highly heterogeneous, with multiple subtypes that differ in molecular and clinical characteristics. It remains the most common cancer among women worldwide. We conducted a hypothesis-generating study using a bioinformatics approach in order to identify potential prognostic biomarkers for breast cancer patients across multiple molecular subtypes. Given the influential role of the transforming growth factor beta (TGFB) pathway in shaping the immune microenvironment, we focused on the isoform, transforming growth factor beta 2 (TGFB2), which is upregulated in tumors, to identify TGFB2-dependent and -independent biomarkers for breast cancer patients’ overall survival (OS) responses. We evaluated the impact of TGFB2 mRNA expression, in conjunction with other potential prognostic markers, on overall survival (OS) in breast cancer patients using The Cancer Genome Atlas (TCGA) and KMplotter databases. We employed a multivariate Cox proportional hazards model to compute hazard ratios (HRs) for TGFB2 mRNA expression, integrating an interaction term that accounts for the multiplicative relationship between TGFB2 and marker gene expressions while controlling age at diagnosis and cancer subtype and differentiating between patients receiving chemotherapy alone and those undergoing alternative therapeutic interventions. We used the KMplotter database to confirm TGFB2-independent prognostic markers from TCGA data. In cases dependent on TGFB2, increased mRNA expression of TGFB2 alongside higher levels of GDAP1, TBL1XR1, RNFT1, HACL1, SLC27A2, NLE1, or TXNDC16 was correlated with improved OS among breast cancer patients, of which four genes were upregulated in tumor tissues (SLC27A2, TXNDC16, TBL1XR1, GDAP1). Future studies will be required to confirm breast cancer patients could improve OS outcomes for patients expressing high levels of TGFB2 and the marker genes in prospective clinical trials. Additionally, multivariate analysis revealed that the elevated expression of six genes (ENO1, GLRX2, PLOD1, PRDX4, TAGLN2, TMED9) were correlated with increases in HR, independent of TGFB2 mRNA expression; all except GLRX2 were identified as druggable targets. Future investigations assessing protein expression in breast cancer tumors to confirm the results of our retrospective analysis of mRNA levels will determine whether the protein products of these genes represent viable therapeutic targets. Protein–protein interaction (STRING) analysis indicated that TGFB2 is associated with EGFR and MYC from the PAM50 breast cancer gene signature. These findings suggest that correlation of TGFB2-related markers could potentially complement the PAM50 signature in the assessment of OS prognosis in breast cancer patients, but further validation of the TGFB2/EGFR/MYC proteins in tumors is warranted.