Abstract P2-11-05: Accurate and cost-effective reconstruction of TNBC heterogeneity through 3D digitization of core needle biopsies

Abstract Objective: Increasing evidence suggests that the tumor microenvironment (TME) of patients with triple negative breast cancer (TNBC) is hallmarked by extensive inter and intra-tumor heterogeneity. Deeper insights into the biological basis of this heterogeneity are needed to elucidate the aggressive nature of TNBC. Early attempts to stratify TNBC patients using bulk or single-cell transcriptomics have shown promise to understand differences in clinical outcomes and provide a personalized medicine approach. However, most examples limit themselves to a single 2D tissue slice per patient. Unfortunately, such an approach represents only a fraction of a 3D core needle biopsy, which is the only available tissue source to examine the treatment naïve TNBC microenvironment. Recently, spatial omics technologies demonstrated superior results for determining heterogeneity as these tools simultaneously profile the intact TME architecture and spatially resolved transcriptomes. Constructing 3D datasets by sectioning tissue and profiling each slice (n > 40) is feasible, but cost-prohibitive and impractical. An innovative digitization workflow that infers transcript localization across 3D datasets without probing each tissue slice would reduce costs and maximize tissue usage, whilst subsequently enhancing our understanding of tissue composition in the treatment naïve TNBC biopsy. Our goal is to facilitate building a comprehensive and large-scale map depicting TNBC heterogeneity from difficult-to-access tissue sources that could shed the much-needed light on TNBC. Method: We sectioned and imaged a treatment-naive core needle biopsy of TNBC into 48 serial slices, with 8 for spatial transcriptomics (Visium, 10X Genomics) equally spaced and 40 for H&E staining. Image coloration was normalized in a semi-automatic manner, and sections were registered on the associated H&E images using VALIS to create a 3D representation of the tissue. Transcripts from spatial transcriptomic sections were enhanced at super-resolution and mapped across serial sections using iSTAR. 3D spatial transcriptomic data analysis and quantitative comparisons with in-silico 2D slices was performed using the Giotto Suite ecosystem. Results: We generated a fully digital and super-enhanced (20 µm pixels) 3D spatial transcriptomics dataset from a single core needle TNBC biopsy by integrating image registration with spatial transcriptomics. Transcripts were projected across tissue sections using spatial interpolation and deep learning methods that learn the spatial and histology relationships for genes, thereby providing a comprehensive and super-resolved 3D representation. We observed that specific 3D structures, such as stromal plasma infiltrates and tertiary lymphoid structures, can be more accurately identified from the full 3D digital biopsy. We also illustrate how spatial gene expression patterns, key signaling pathways, and cellular architecture are organized in three dimensions, which cannot be easily inferred from 2D datasets alone. Importantly, systematic analysis comparing 3D versus 2D in-silico generated slices provides a quantitative framework that can be used to determine the limitations and power of spatial transcriptomic data analysis in 2D. Conclusion: This study demonstrates the feasibility of incorporating spatial transcriptomics and 3D registration methods into standard spatial analyses, enabling a detailed understanding of the spatial organization within a core needle TNBC biopsy and leading to informed insights on tissue architecture and function. Our approach facilitates a data-driven approach to better understand the strengths and limitations of profiling tumor heterogeneity in core needle biopsies of breast cancer. Finally, our approach enables researchers to make more informed research designs for large-scale projects, maximizing the usage of precious and often difficult-to-obtain patient samples. Citation Format: Jeffrey Sheridan, Jaiji George Chen, Junxiang Xu, Christina Ennis, Emma Kelley, Neal Kewalramani, Wonyl Choi, Jason Weiss, Gerald Denis, Naomi Ko, Ruben Dries. Accurate and cost-effective reconstruction of TNBC heterogeneity through 3D digitization of core needle biopsies [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P2-11-05.