This presentation by Dr. Fay Chen introduces groundbreaking advancements in spatial genomics, focusing on understanding cellular communities and tissue organization. The talk highlights the evolution of technologies like Slide-seq, which enables high-resolution spatial transcriptomics, and the more recent Slide-tags, a method for spatially barcoding individual cells or nuclei that is fully compatible with existing high-quality single-cell sequencing workflows (e.g., 10x Genomics) for multi-omic analysis (RNA, DNA, ATAC, TCR).
Dr. Chen demonstrates the application of these tools in cancer research, specifically using melanoma models and human metastatic lymph node biopsies. Key findings include the identification of significant transcriptional and clonal heterogeneity within tumors, the segregation of tumor subclones into distinct spatial communities with varied immune microenvironments, and the observed association of specific T-cell receptor (TCR) clonality with particular genetic tumor subclones. The presentation also details the discovery of recurrent cellular neighborhoods (niches) and signaling programs that influence gene expression within malignant cells, suggesting a complex interplay between intrinsic cellular states and extrinsic environmental factors.
Future directions emphasize the scaling of these technologies to large biological samples like whole human organs and brains, the development of spatial DNA capture for single-nucleotide variant analysis and lineage tracing, and a novel imaging-free method for reconstructing spatial information directly from sequencing data, which promises to overcome current throughput bottlenecks in array manufacturing and make spatial genomics more accessible for every single-cell experiment.
Exploring Cellular Communities in Tissues with Genomics
0:03 Introduction of Dr. Fay Chen: Anoja Pereira introduces Dr. Fay Chen, highlighting his role at the Broad Institute at MIT and Harvard, his focus on genomics and computation for understanding tissue organization and disease, and his co-invention of expansion microscopy and Slide-seq.
0:17 Family Emergency: Dr. Chen is unable to attend in person due to a family emergency and gives the presentation remotely.
0:18 Disclosures: Dr. Chen is a co-founder of Curio Biosciences, a company formed to disseminate Slide-seq and Slide-tags reagents, which has since been acquired by Takara.
0:46 Revolution in Genomics: Dr. Chen discusses the current revolution in genomics, drawing parallels to predictions made by Eric Lander 30 years ago, where now simultaneous and spatial monitoring of gene expression is possible at high resolution.
0:59 Understanding Molecular States in Tissues: The core problem addressed is understanding the molecular states of cells directly within tissues, considering cell-intrinsic programs and multicellular circuits influenced by the cell's environment.
0:06:45 Cancer as an Application: Cancer, particularly melanoma, is presented as an ideal application for spatial genomics due to its diverse cell types, states, and environmental influences.
0:08:12 Key Questions: Three key questions guide the research: defining cell neighborhoods, understanding inter-cell communication via signaling, and how the environment changes cell states.
0:09:08 Tool Development: The lab focuses on two technology types: spatial capture (barcoded arrays for DNA/mRNA) and direct in-situ sequencing.
0:10:28 Slide-seq Introduction (2019): Slide-seq is introduced as one of the first high-resolution spatial transcriptomics technologies, using 10-micron barcoded polystyrene beads to capture mRNA from tissue sections, allowing spatial mapping of RNA.
0:12:00 Mouse Hippocampus Example: An early Slide-seq experiment on mouse hippocampus demonstrates the ability to assign molecular identities to cell types and map their organization.
0:13:00 Computational Tools (SpaceXR): Tools like SpaceXR are developed to computationally define cell neighborhoods and identify spatially differential gene expression.
0:13:55 Spatial DNA Capture: Slide-seq arrays are adapted to capture DNA, using TN5 transposase to tagment DNA, allowing for spatial mapping of genomic alterations like copy number variations (CNVs) and single nucleotide variants (SNVs).
0:14:41 Metastatic Clones Example: Spatial DNA capture on mouse metastatic liver tumors reveals clonal origins and specific chromosomal amplifications (e.g., chromosome 6, 15, 19).
0:16:30 Human Melanoma Example (DNA-seq): Application to human melanoma identifies normal vs. tumor cells and distinct subclones based on CNVs (e.g., chromosome 20 gain, chromosome 8Q gains related to MYC).
0:17:15 Extrinsic Environment Impact: Slide-seq RNA data on an adjacent section is used to correlate gene expression with local tumor cell density, showing that some gene expression changes are driven by the environment (e.g., cell adhesion molecules) rather than intrinsic clonal identity.
0:21:00 Challenges with Cell Mixtures: Capture-based methods like Slide-seq suffer from cell mixture problems where individual pixels contain transcryptomes from multiple cells.
0:22:10 Slide-tags Introduction: To overcome the mixture problem and leverage single-cell data quality, Slide-tags is developed. This method spatially barcodes cells/nuclei directly in tissue, then nuclei are isolated and profiled using established single-cell platforms (e.g., 10x Genomics).
0:23:32 Slide-tags Mechanism: Photo-cleavable barcodes from the array diffuse into the tissue, associate with cell nuclei, which are then isolated and sequenced, providing single-cell RNA-seq data with spatial coordinates.
0:24:40 Slide-tags on Mouse Hippocampus: Slide-tags on mouse hippocampus shows fine-grained cell type coverage and diversity, with robust localization of cells by fitting peaks of barcode distributions.
0:26:03 Slide-tags Quality: Slide-tags does not significantly affect single-nucleus RNA-seq quality, maintaining similar cell type proportions and molecule counts. Current recovery rates are 25-30% of nuclei from tissue.
0:27:39 Slide-tags Versatility: Slide-tags is compatible with various single-cell assays beyond RNA expression, including TCR sequencing, copy number inference, and chromatin accessibility (ATAC-seq).
0:28:32 Human Lymph Node Metastasis (Slide-tags): Application to a human melanoma lymph node metastasis post-checkpoint blockade reveals distinct histological regions, diverse cell types, and spatially segregated tumor subclones with unique transcriptional states.
0:29:47 Genetic Differences and Neighborhoods: CNV analysis on single cells reveals genetic subclones. Colocalization analysis shows these subclones have different immune and stromal cell environments (e.g., CD8 T-cell enrichment/depletion).
0:31:36 TCR Clonality: TCR sequencing identifies specific, highly expanded CD8 T-cell clones enriched near one tumor subclone, demonstrating a link between T-cell clonality and specific genetic tumor subclones.
0:32:50 Melanoma Differentiation: The observed interactions might be influenced by tumor cells losing melanocytic identity and transitioning to a mesenchymal-like state, a process observable via joint ATAC and RNA sequencing.
0:36:32 Large Melanoma Cohort Study: A larger dataset (300k nuclei from 27 metastatic biopsies) is collected to identify recurrent neighborhoods and their effect on gene expression.
0:37:25 Recurrent Niches: Identified recurring cellular neighborhoods ("niches") in melanoma samples, such as immune effector hubs, myeloid/stromal niches, and tertiary lymphoid structures (TLS-like).
0:40:02 Malignant Cell Gene Programs: Recurring gene expression programs within malignant cells (proliferation, EMT, stress, apoptosis, interferon response) are identified.
0:41:00 Predicting Program Usage: Linear regression shows which stromal cell types predict malignant program usage, e.g., interferon response correlated with tumor-responsive cytotoxic T-cells and specific macrophage populations.
0:43:47 Macrophage Populations: Two distinct macrophage populations are identified: a suppressive one associated with a myloid/stromal niche and neurocrest-like signaling in malignant cells, and an effector one co-localized with interferon-stimulated cancer cells.
0:45:07 Soluble Ligands & Signaling Programs: Soluble ligands are identified as mediators of communication. A method to define recurrent extracellular signaling programs by decomposing a matrix of observed ligands is proposed.
0:46:44 Identified Signaling Programs: Three signaling programs are found: one associated with neurocrest-like tumor population, one with melanocytic state (immune), and CXL 9/10/11 defining TLS creation.
0:47:52 Rules of Tissue Organization: Understanding recurrent neighborhoods and extracellular signaling programs is crucial for learning rules of tissue organization in development, pathology, and therapy.
0:49:10 Slide-tags Summary: Single-cell data with high spatial resolution (10 microns), compatible with existing single-cell workflows, enabling diverse assays.
0:49:52 Future DNA Applications: Modifying Slide-tags for DNA capture to get 2-3% genomic coverage per cell, enabling SNV calling, mutational signatures, and local lineage information from colon samples. Also, higher coverage (80% per cell) with plate-based methods for phylogenic trees.
0:54:10 Spatial Context "For Free": Advocate for including spatial barcoding in every single-cell experiment as it adds spatial context without altering data quality, especially as sequencing costs decrease.
0:55:01 Scaling Challenges: Experimental throughput and array size are bottlenecks for large-scale studies (whole embryos, human brain, large human tissues).
0:56:52 Imaging-Free Spatial Reconstruction: A novel method inspired by Slide-tags allows reconstructing spatial information directly from sequencing data, without imaging, by measuring pairwise distances between beads. This enables arbitrarily large arrays.
1:01:17 Reconstruction Validation: The reconstructed spatial map shows high accuracy, with errors comparable to histological stretching/shrinking, making the data virtually indistinguishable from ground truth.
1:04:09 P1 Mouse Brain & Human Brain Slab: Successfully applied the reconstruction method to a P1 mouse brain (1.2 cm array) and a human brain slab (7 cm array), resolving substructured cell types at high resolution.
1:07:42 Q&A: Melanoma Cohort Data: Confirms availability of TMB and BRAF mutation status data, actively investigating how TMB and CNA influence neighborhoods and microenvironments.
1:09:55 Q&A: Circulating Biomarkers: Suggestion to link microenvironments to circulating biomarkers for diagnostic and longitudinal profiling.
1:10:23 Q&A: Gene-Specific Probes: Suggestion to use gene-specific probes for diffusion matrix to reduce complexity and target specific mutations or biomarkers.
1:11:48 Q&A: Slide-tags Resolution Limit: While 100% recovery is the theoretical limit, practical improvements are ongoing beyond the current 25-30%, but tissue-specific optimization might be needed.
1:12:47 Q&A: Tissue Transportation: Addresses challenges of transporting tissue for spatial genomics. Solutions include freezing/fixing nuclei or freezing the sectioned tissue directly on the array, with ongoing tests for stability.
1:14:46 Q&A: Multi-nucleated Cells: Acknowledges the challenge with multi-nucleated cells; they are currently treated as separate nuclei.
1:15:58 Q&A: Archival FFP Tissue: Large archival FFP tissue is a goal. DNA sequencing is closer for FFP than RNA. Flex protocol combined with large arrays could achieve good RNA results for FFP samples.