Integrative Genomic Profiling of Tumor Heterogeneity and Microenvironments in Breast Cancer and Their Clinical Associations

Abstract

Breast cancer (BC) represents a major clinical challenge due to its extensive molecular and clinical heterogeneity, underscoring the necessity of comprehensive characterization to improve patient outcomes. This dissertation addresses this complexity through an integrative multi-omics approach that combines whole-genome and whole-exome sequencing, transcriptomic profiling, and long-read genomic analyses. The primary objective is to elucidate the molecular mechanisms underlying tumor heterogeneity, therapeutic resistance, and divergent clinical trajectories. By integrating data across multiple platforms and clinical subtypes, this research establishes an integrative framework for understanding BC biology and contributes to the development of precision therapeutic strategies. Chapter 1 investigates the genomic and transcriptomic features of triple-negative apocrine carcinoma (TNAC) in comparison with triple-negative breast cancer (TNBC). Although TNAC has traditionally been regarded as a rare histological subtype within TNBC, this analysis revealed that TNAC exhibits distinct molecular features and significantly more favorable outcomes than TNBC with low Ki-67 (LK-TNBC). Genomic profiling identified TP53 as the most frequently altered driver gene in TNAC, followed by PIK3CA, ZNF717, and PIK3R1. Mutational signature profiling showed that TNAC was enriched for DNA mismatch repair defect-related signatures (SBS6 and SBS21) as well as SBS5, whereas LK-TNBC exhibited a stronger APOBEC-associated signature, driven primarily by SBS13. Intrinsic subtyping revealed that most TNAC samples aligned with luminal A, followed by luminal B and HER2-enriched, while LK-TNBC was predominantly basal-like. TNAC had significantly better survival than LK-TNBC, and within TNAC, luminal A and normal-like subtypes were associated with the most favorable outcomes. These findings establish TNAC as a biologically and clinically distinct subtype of TNBC, characterized by unique molecular profiles and potential therapeutic implications, including possible responsiveness to immunotherapy and a reduced need for cytotoxic chemotherapy. Chapter 2 presents an integrative multi-omics analysis of residual BC following neoadjuvant chemotherapy (NAC), aiming to uncover molecular and immunologic determinants of divergent post-treatment outcomes. A total of 91 residual BCs were analyzed using whole genome/exome sequencing, bulk RNA sequencing, and spatial transcriptomics. In HR⁺HER2⁻ tumors, the basal-like intrinsic subtype emerged as the principal determinant of poor prognosis, characterized by TP53 mutations, elevated homologous recombination deficiency (HRD) scores, and strong SBS3 mutational signatures. HER2⁺ residual tumors with poor prognosis exhibited ERBB2 point mutations, structural amplifications, and extrachromosomal DNA–associated oncogene amplification, suggesting potential mechanisms underlying resistance to anti-HER2 therapy. Spatial and deconvolution analyses of non-TNBCs revealed the coexistence of basal-like cancer cells and SPP1⁺ macrophages within immune-excluded microenvironments associated with poor prognosis, in contrast to immune-permissive contexts in favorable cases. In TNBC, genomic features were largely comparable between prognostic groups; however, immune landscapes differed considerably. CXCL9⁺ macrophages, positively correlated with CD8⁺ T-cell infiltration, were associated with improved survival and served as independent prognostic factors alongside residual cancer burden. Together, these findings highlight subtype-specific prognostic determinants in residual BC following NAC. Chapter 3 broadens the scope of this dissertation by exploring the technical frontier of cancer genomics, applying paired long-read (PacBio HiFi) and short-read (Illumina) whole-genome sequencing of tumor–normal pairs to achieve more accurate and haplotype-resolved insights into the BC genome. Through multi-tool structural variant (SV) calling and integrative analyses, long-read sequencing was found to detect a greater number of complex and insertion-type SVs, while several short-read-specific events were identified as false positives, reflecting the fragmentation of complex rearrangements due to limited read length. Long-read data uniquely detected clinically relevant variants, including a 319 bp deletion in RAD51B and a 57 bp insertion in ESR1, both of which were missed by short-read data. Copy-number profiling further showed markedly reduced noise and improved concordance with SV breakpoints in long-read data, especially within repetitive and low-mappability regions. Haplotype-resolved analysis additionally uncovered allele-specific chromothripsis events on chromosomes 5 and 8 that were undetectable by short-read sequencing, underscoring the ability of long-read sequencing to reconstruct complex rearrangements across haplotypes. By refining the structural and allelic interpretation of the cancer genome, long-read sequencing serves not only as a technical enhancement but also as a powerful analytical approach that complements biological findings from earlier chapters, ultimately enabling a more comprehensive understanding of BC genomics. Together, these chapters provide a multidimensional understanding of BC, encompassing molecular subtype characterization, treatment response biology, and high-resolution genomic architecture. Through this integrative approach, the dissertation bridges biological insight and technological advancement, establishing a comprehensive framework for precision oncology in BC.