High-throughput single-cell genomics allows for the clustering of cells into distinct types based on gene expression and protein data, enabling ambitious projects like profiling every cell type in the human body. Advances in spatial transcriptomics provide unbiased gene expression analysis with spatial context, combining genomics, imaging, and tissue pathology. These technologies are complementary; single-cell methods lack spatial context, while spatial methods may need single-cell data integration for detailed cell type information. While progress has been made in integrating these data types, high-resolution analysis of cell-cell interactions and definitive transcript assignment to cells with spatial context at high gene plexy is lacking. An ideal solution would provide high-plex, high-throughput, multimodal readouts with spatial context and subcellular resolution, compatible with both fresh frozen (FF) and formalin-fixed paraffin-embedded (FFPE) tissues. Several high-plex in situ technologies have emerged, but integrating these with whole transcriptome single-cell or spatial data remains a challenge. This study aims to address this by integrating single-cell, spatial, and in situ technologies on serial sections of FFPE-preserved breast cancer to explore tumor heterogeneity.

The existing literature highlights the power of single-cell and spatial transcriptomics in understanding cellular heterogeneity and spatial organization within tissues. Studies using single-cell RNA sequencing (scRNA-seq) have generated comprehensive cell atlases for various tissues and organs, including the human breast, revealing complex cell type compositions and their dynamic interactions. Spatial transcriptomics technologies, such as Visium, have further advanced this understanding by providing spatial context to the gene expression data. However, existing methods for integrating these technologies, particularly in the context of FFPE tissues, have limitations in resolution and the ability to identify rare cell populations. This study builds upon these previous efforts by leveraging the capabilities of multiple high-resolution technologies.

The study uses three complementary technologies on serial sections of FFPE breast cancer tissue blocks: 1) Chromium Single Cell Gene Expression Flex (scFFPE-seq) for whole transcriptome single-cell analysis; 2) Visium CytAssist for whole transcriptome spatial analysis; and 3) Xenium In Situ for high-resolution, targeted in situ gene expression analysis. For scFFPE-seq, FFPE tissue curls were dissociated using the Miltenyi FFPE Tissue Dissociation Kit, and single-cell libraries were prepared using the Chromium Single Cell Gene Expression Flex workflow. For Visium CytAssist, adjacent 5 µm sections were H&E stained, then processed to transfer analytes to Visium slides. The Xenium In Situ technology used a targeted panel of 313 genes, selected based on single-cell atlas data. The Xenium workflow involved probe hybridization, ligation, rolling circle amplification, and multiple cycles of fluorescent probe hybridization and imaging. Post-Xenium H&E and immunofluorescence (IF) staining were performed. Data integration was performed using computational methods such as t-SNE dimensionality reduction, differential gene expression analysis, and cell type annotation. Specific techniques like spot interpolation were employed to integrate Visium and Xenium data.

The integrated analysis revealed significant heterogeneity within the breast cancer samples. scFFPE-seq and Visium identified 17 clusters, allowing for cell type annotation. Visium pinpointed the spatial location of three tumor domains: two distinct DCIS subtypes (DCIS #1 and DCIS #2) and invasive tumor. Xenium provided high-resolution data for a targeted gene panel, identifying 167,885 cells and 36,944,521 transcripts. Integration of scFFPE-seq and Xenium data accurately identified cell types in the Xenium data (86% unambiguous identification). Comparison of DCIS #1 and DCIS #2 ROIs revealed differences in cell type composition, particularly in myoepithelial cell populations and the presence of invasive cells within DCIS #2. A small triple-positive (ERBB2+/ESR1+/PGR+) region was identified using Xenium, and integration with Visium data provided whole transcriptome information on this region, revealing 48 differentially expressed genes compared to PGR-negative DCIS. In a second sample, Xenium identified a small population of "boundary cells" expressing both tumor and myoepithelial markers, which were subsequently identified in scFFPE-seq data. Differential gene expression analysis revealed genes such as CX3CL1, CCL28, PROMI, and KLKS highly expressed in these boundary cells.

This study demonstrates the power of integrating single-cell, spatial, and in situ technologies to comprehensively characterize the tumor microenvironment at high resolution. The integration strategy successfully revealed subtle differences in cell type composition and molecular signatures between distinct tumor regions and identified rare cell populations that might play crucial roles in cancer progression. The findings highlight the limitations of relying on single-technology analysis and the importance of combining complementary approaches to gain a holistic understanding of cancer heterogeneity. The identification of boundary cells with a mixed tumor/myoepithelial profile suggests a potential transitional state during the progression from DCIS to invasive carcinoma. The ability to identify such rare cell types and their molecular profiles opens new avenues for research on cancer progression and the development of targeted therapies.

This study demonstrates that integrating single-cell, spatial, and in situ transcriptomic technologies provides a powerful approach for high-resolution mapping of the tumor microenvironment. The combination of these techniques revealed significant tumor heterogeneity not apparent from single modality analyses, including the identification of rare cell populations involved in tumor progression. These findings highlight the importance of multimodal approaches for a more comprehensive understanding of cancer biology and for the development of improved diagnostic and therapeutic strategies. Future research could focus on expanding the use of this integrated approach to a larger cohort of samples and investigating the functional roles of the identified cell populations and molecular signatures.

The study is limited by the use of only two FFPE tissue blocks. While the results are highly suggestive, validation on a larger, more diverse cohort is needed. Additionally, the Xenium technology used in this study was a prototype instrument, and further refinement of the technology may improve resolution and throughput. The interpretation of some findings, such as the functional role of boundary cells, requires further investigation.