e12582 Background: The goal of our study is to improve the diagnosis and treatment of DCIS through the identification of the factors driving its invasion and metastasis. Methods: We utilized the Mouse INtraDuctal (MIND) model to investigate the mechanism(s) by which some DCIS become invasive while others remain indolent. This model closely replicates the human condition by injecting non-invasive epithelial cells from patients into mouse mammary ducts to form in situ lesions. In some cases, these cells will display invasive behavior by overcoming the natural barriers of myoepithelium. To compare these two processes, we performed single-cell (sc) RNA-sequencing analysis of 60,000 cells from the pre-transplant time-point representing 17 DCIS (11 became invasive, 6 remained non-invasive). We also performed scATAC/RNA sequencing of a DCIS with associated IDC. Results: Our analysis revealed that the proportion of endothelial cells was higher in cases that became invasive, while the proportion of plasma and T cells was higher in those that remained non-invasive (t-test; P<0.05). This finding suggested that invasive cells reprogrammed the immune microenvironment to exclude immune cells as a means of evading immune surveillance. Additionally, we performed cell-cell interaction (CCI) analysis by CellPhoneDB. This analysis revealed highly ranked CCIs in invasive cases, including luminal cell interactions with pericytes via Wnt5A-SFRP1. On the other hand, highly ranked CCIs in non-invasive cases included HR-positive luminal cell interactions with T cells via CD226-Nectin-2 and CD96-Nectin1. Notably, CD226 and CD96 are T cell receptors that enhance CD8+ T cell activity. These results support the potential role of T-cell interactions with DCIS luminal cells in preventing their invasive transition. We conducted additional analysis of sc-RNA-sequencing data using a stem cell-based gene signature algorithm called CytoTRACE. Our analysis revealed that invasive cells were enriched in cell clusters with significantly higher stem cell scores than non-invasive cells. In another analysis of a patient DCIS with associated IDC using sc-ATAC and RNA sequencing, we discovered that the open chromatin regions in the stem cell cluster were enriched for FOXA1 binding motifs. This cluster also had the highest expression of two pioneer factors (FOXA1 and GRHL2), NEAT1 (paraspeckle protein), ERBB2, 3,4, and ANKRD30A (cell surface protein). Our findings suggest that the pioneer factors FOXA1 and GRHL2 play a crucial role as transcriptional drivers of stemness. Additionally, the ANKRD30A could be used to isolate stem cell clusters to validate their self-renewal potential. Conclusions: DCIS progression is facilitated by interactions between stromal, immune, and epithelial cells, which lead to epigenetic and transcriptional changes. This results in the formation of stem-like cells that are characterized by their plasticity and invasive capacity.

Combating chemotherapy resistance in TNBC is critical to improving quality of care and reducing fatality. To understand chemoresistance, it has been proposed that the spatial interactions between cell types in the tumor microenvironment (TME) could offer insights into the differential response to therapy and metastatic potential. Thus, this study utilizes spatially resolved transcriptomics (SRT) to profile the spatial interactions in TNBC TME. With the goal of understanding chemoresistance and providing a proof-of-concept for new technology, we launched a longitudinal SRT study of a TNBC PDX of residual disease before, during, and after adriamycin and cyclophosphamide (AC) treatment (Tx). SRT by 10X Visium was performed on Veh, AC-residual tumor (21 d post-Tx), and AC-regrown tumor (50 d post-Tx when tumors regrew). This study is divided into two parts. 1. Due to the lack of specialized tools for processing xenograft reads from SRT, we developed the Xenomake pipeline, which combines xenograft sorting and spatial barcode demultiplexing to assign reads into the host and graft organisms per spatial spot. It permits clustering the spots into stroma-rich (mouse), and epithelial-rich (human) regions. We show that it can find differential cytokines in the stroma (S) and epithelium (E), and finding SS, EE, and SE interactions. 2. Although using PDX necessitates immune-compromised mice, several types of stromal populations are present and analyzable. Thus, using Xenomake, we inferred the localization of stromal cell types in Visium spots including cancer associated fibroblasts (CAF), macrophages (MP), endothelial (ENDO), monocytes, and perivascular cells. We compared the spatial localization profiles (SLP) of stromal cell types across samples. With human reads, we computed the spatial pathway activity map (SPAM) for HALLMARK pathways. We correlated SPAM with stroma cell type SLP to survey stroma-epithelial interactions. Stroma cells are distributed in the tumor periphery in vehicle and AC50, while in AC21 there is a notable increase in stroma abundance and stroma infiltration within tumor mass. SLP of MP (Cd68+ and Csf1r+) is correlated with ENDO (Pecam1+), and CAF (Acta2+ and Pdgfrb+). Additionally, all 3 cell types are correlated with Vim and Cd44 expression. Although all samples show enrichment of MP, ENDO, and CAF, they interact differently with pathway activities of adjacent epithelial cells. In vehicle and AC50, the 3 cell types are colocalized with OXPHOS, MYC targets, E2F pathway, while in AC21, a switch to colocalization with EMT is observed. Further, hypoxia response and glycolysis display anti-correlation with stroma cell types, meaning that these pathway activities are further away from stroma and occupy distinct territories. The results suggest stroma-tumor metabolic crosstalk and ways of targeting residual disease. Citation Format: Benjamin Strope, Katherine Pendleton, William Bowie, Gloria Echeverria, Qian Zhu. A spatial transcriptomic study of a triple-negative breast cancer (TNBC) patient-derived xenograft (PDX) model of residual disease refractory to conventional chemotherapy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3876.

Abstract TNBC is one of the most aggressive subtypes of breast cancer. Combating chemotherapy resistance is critical to improving quality of care and reducing fatality among TNBC patients. In order to understand the mechanisms of chemoresistance in TNBC tumors, it has been proposed that the spatial interactions between cell types in the tumor microenvironment (TME) could offer insights into the differential response to therapy among tumors, and metastatic potential of certain tumors over others. Thus, this study utilizes spatially resolved transcriptomic (SRT) technology to profile the spatial interactions in the TNBC TME. With the goal of better understanding chemoresistance in TNBC and providing a proof-of-concept for new technology, we have launched a longitudinal SRT study of a TNBC PDX of residual disease before, during, and after adriamycin and cyclophosphamide (AC) treatment (Tx). SRT by 10X Genomics Visium was performed on vehicle, AC-treated residual tumor (21 days post-Tx), and AC-treated regrown tumor (50 days post-Tx when tumors regrew to starting tumor volume). This study is divided into two parts: 1) development of a computational pipeline; 2) spatial colocalization analysis of residual and regrown tumors to understand chemoresistance. PART 1: Due to the lack of specialized tools for processing and sorting xenograft reads from SRT data, we developed the Xenomake pipeline, which combines a xenograft sorting algorithm (Xengsort) and spatial barcode demultiplexing pipeline to assign reads into the host and graft organisms for each spatial spot. RESULTS: Xenomake permits clustering the spatial spots into stroma-rich (enriched for mouse mRNAs), and epithelial-rich (enriched for human mRNAs) regions. We show that Xenomake can find differential cytokine production in the stroma and epithelium. Since PDX data separate the tumor into stroma and epithelium by organism, the pipeline enables fine-tuned downstream analysis such as stromal-stromal and stromal-epithelial interactions. Xenomake is thus generally applicable for SRT involving PDX samples. PART 2: Previously we detailed patterns of tumor regression into a residual tumor state, followed by uncontrolled regrowth in the absence of treatment in multiple PDX models of TNBC. Although using PDX models necessitates immune-compromised mice, several types of stromal populations are present and analyzable in these models. Thus, using organism-assigned reads processed by Xenomake, we computationally inferred the localization pattern of stromal cell types in Visium spots including cancer associated fibroblasts (CAF), macrophages (MP), endothelial cells (ENDO), monocytes, and perivascular-like cells. We compared the spatial localization profiles (SLP) of stromal cell types across samples. With human reads, we computed the spatial pathway activity map (SPAM) for the ~30 HALLMARK pathways across samples. We correlated the patterns of SPAM with stroma cell-type SLP to survey the stroma-epithelial interactions across samples. RESULTS: Stroma cells are generally distributed in the tumor periphery in vehicle and AC50, while in AC21 there is a notable increase in stroma abundance and stroma infiltration within tumor mass. SLP of MP (Cd68+ and Csf1r+) is correlated with ENDO (Pecam1+), and with CAF (Acta2+ and Pdgfrb+). Additionally, all 3 cell types are correlated with Vim and Cd44 expression. Although all samples show enrichment of MP, ENDO, and CAF, they interact differently with pathway activities of adjacent epithelial cells. In vehicle and AC50, the 3 cell types are colocalized with OXPHOS, MYC targets, E2F pathway, while in AC21, a switch to colocalization with EMT is observed. Furthermore, hypoxia response and glycolysis display anti-correlation with stroma cell types, meaning that these pathway activities are further away from stroma and occupy distinct territories. The results suggest stroma-tumor metabolic crosstalk and ways of targeting residual disease. Citation Format: Benjamin Strope, Katherine Pendleton, William Bowie, Gloria Echeverria, Qian Zhu. A spatial transcriptomic study of a triple-negative breast cancer (TNBC) patient-derived xenograft (PDX) model of residual disease refractory to conventional chemotherapy [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO2-28-08.

Abstract An increasing number of cancer research studies employ spatially resolved transcriptomics (SRT) to investigate the composition of tumor microenvironment in a cancer type of interest. These studies have defined tumor microenvironment (TME) states and spatial domains based on clustering spatial gene expression patterns in SRT in an unbiased manner, yet a more thorough delineation of TME states requires the incorporation of the tumor’s histology image. Here, we develop MultiNMF, a multiview factorization approach that is suitable for cancer research studies where joint profiles of spatial multi-omics and tumor histology images are available. We apply MultiNMF to analyze a set of TNBC SRT primary tumor samples and reveal TME states in the stromal, epithelial, and immune enriched compartments, defined by distinct histomorphological features. We further illustrate the ability of the approach to extend to paired spatial ATAC-seq and histology dataset that is recently published on HER2 breast cancer. MultiNMF thus permits an automated and data-driven decomposition of SRT and spatial ATAC data supported by histomorphological evidence. Context In SRT by 10X Visium, the hematoxylin eosin (H&E) staining image is automatically aligned to the slide where the tissue section is mounted. Image patches can be easily extracted from areas under the SRT barcoded spots. For these images, I will apply a pre-trained convolutional neural network (CNN) model to extract high dimensional features from spot images. The image features can be added as a second view after gene expression view for joint analysis under MultiNMF. Results MultiNMF can factorize the SRT data into components well-supported by histological evidence. It has identified T-cell infiltrating regions, and EMT-enriched regions with distinct immune and stroma morphological characteristics. The T-cell infiltration neighbors a region that has an appearance of necrosis according to the pathologist evaluation. These domains further demonstrate enrichments of modules with motif enrichment (known as regulons) according to SCENIC analysis. Analysis of spatial ATAC illustrates not only HER2 but also its amplicon amplification. From MultiNMF components of spatial ATAC, we derive regulatory regions for T-cell/B-cell marker genes. Overall, MultiNMF is a novel model that has not been applied to SRT field. It provides an multiomic extension to the traditional NMF. Citation Format: William Bowie, Stacy Wang, Benjamin Strope, Qian Zhu. MultiNMF: multiview factorization for joint modeling of spatial multi-omics and histology images [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2335.