23.1 Single-cell analysis and spatial transcriptomics

Single-Cell Analysis and Spatial Transcriptomics

Single-cell analysis lets you measure gene expression in individual cells rather than averaging across thousands of them. This matters because tissues that look uniform under a microscope often contain distinct subpopulations with very different molecular profiles. Spatial transcriptomics builds on this by preserving where each cell sits within a tissue, so you can connect gene expression patterns to tissue architecture and cell-cell communication.

Together, these two approaches have reshaped how we study development, disease, and tissue organization.

Single-Cell Analysis

Principles of scRNA-seq

Single-cell RNA sequencing (scRNA-seq) captures a snapshot of which genes are active in each individual cell. Here's how the workflow breaks down:

1. Cell isolation — Individual cells are separated from a heterogeneous tissue using microfluidics or droplet-based methods. Platforms like 10X Genomics encapsulate single cells into nanoliter-sized droplets, each containing a barcoded bead that tags all mRNA from that cell with a unique identifier.
2. Reverse transcription — mRNA within each droplet is converted to complementary DNA (cDNA). This step captures the cell's transcriptional state at the moment of isolation.
3. Amplification — Because a single cell contains only picograms of RNA, the cDNA must be amplified (via PCR or in vitro transcription) to generate enough material for sequencing.
4. Sequencing — Amplified cDNA libraries are sequenced on high-throughput platforms (typically Illumina), generating millions of reads that can be traced back to individual cells using the barcodes from step 1.
5. Bioinformatic analysis — Reads are aligned to a reference genome, and gene expression is quantified per cell. Cells are then normalized, dimensionally reduced (using methods like PCA and UMAP), and clustered based on similar expression profiles to reveal distinct cell types and states.

Advantages vs. Limitations of Single-Cell Techniques

Advantages of single-cell analysis:

• Resolves cellular heterogeneity — You can detect rare subpopulations (e.g., cancer stem cells making up <1% of a tumor) that bulk sequencing completely misses by averaging their signal with thousands of other cells.
• Reveals stochastic and transient states — Gene expression varies from cell to cell even within the same type. scRNA-seq captures this variation, which is especially useful for studying processes like differentiation where cells pass through brief intermediate states.
• Enables trajectory reconstruction — By ordering cells along a pseudotime axis based on gradual shifts in their expression profiles, you can reconstruct developmental lineages without needing to sample the same cell at multiple time points.

Limitations of single-cell analysis:

• Isolation artifacts — Dissociating tissue into single cells can stress or damage them, altering gene expression. Cell doublets (two cells captured in one droplet) also introduce noise.
• Dropout events and amplification bias — Low-abundance transcripts frequently fail to be captured or amplified, creating false zeros in the data. This is a major source of technical noise in scRNA-seq.
• Loss of spatial context — Once you dissociate a tissue, you lose all information about where each cell was located and which cells were its neighbors.
• Cost and computational demands — Profiling tens of thousands of cells generates high-dimensional datasets that require significant computational resources and specialized bioinformatic pipelines.

Advantages of spatial transcriptomics:

• Preserves tissue architecture — Gene expression is measured in situ, so you can map which cell types sit next to each other and study how position influences function.
• Reveals microenvironments — You can identify niches (e.g., stem cell niches, immune-excluded tumor regions) where local cell-cell interactions regulate behavior.
• Links expression to morphology — Overlaying molecular data onto histological images lets you correlate gene expression patterns with visible tissue structures.

Limitations of spatial transcriptomics:

• Lower cellular resolution — Many spatial methods (e.g., Visium from 10X Genomics) capture expression from spots containing multiple cells, so the data represents a local mixture rather than a true single-cell readout. Newer methods like MERFISH and Slide-seq are pushing toward single-cell or subcellular resolution, but trade-offs remain.
• Gene panel constraints — Some approaches (particularly imaging-based ones like MERFISH or seqFISH) require you to preselect a panel of target genes, so you won't capture the full transcriptome.
• Tissue quality sensitivity — RNA degrades quickly, and sample preparation steps like fixation and permeabilization can compromise transcript detection.

Spatial Transcriptomics

Concept of Spatial Transcriptomics

Spatial transcriptomics refers to a family of techniques that measure gene expression while retaining information about where transcripts are located within a tissue section. The core idea is straightforward: instead of dissociating cells and losing their positions, you profile them in place.

There are two broad strategies. Sequencing-based methods (like Visium) use arrays of barcoded capture spots on a glass slide; a tissue section is placed on top, mRNA diffuses to the spots, and each spot's barcode encodes its spatial coordinates. Imaging-based methods (like MERFISH or seqFISH) use repeated rounds of fluorescent probe hybridization to directly visualize individual RNA molecules within intact tissue.

Both approaches enable several key analyses:

1. Mapping cell type distributions — Using marker genes and clustering algorithms, you can assign cell identities to spatial locations and visualize how different types are organized across a tissue.
2. Characterizing cell-cell interactions — Knowing which cells are physically adjacent lets you infer signaling relationships, such as tumor-stroma interactions at the invasive margin of a carcinoma.
3. Studying spatial heterogeneity — Gene expression often varies with position, whether that's anterior-posterior patterning during embryonic development or gradients of hypoxia within a solid tumor.
4. Integrating with histology — Spatial transcriptomic data can be overlaid onto H&E-stained images, directly linking molecular profiles to tissue morphology visible under the microscope.

Applications in Complex Biological Systems

Tumor heterogeneity

Tumors are not uniform masses. Spatial transcriptomics has revealed that distinct subpopulations of cancer cells occupy specific regions within a tumor, and their spatial arrangement matters for clinical outcomes.

• Mapping immune cell infiltration patterns shows that some tumors have "immune-hot" regions rich in cytotoxic T cells while other zones are immune-excluded, which helps explain variable responses to immunotherapy.
• Drug-resistant persister cells often cluster in particular microenvironments (e.g., near hypoxic cores), suggesting that spatial context contributes to resistance mechanisms.
• Characterizing tumor-stroma boundaries at the molecular level reveals signaling crosstalk (e.g., between cancer-associated fibroblasts and tumor cells) that bulk approaches cannot resolve.

Developmental processes

During embryonic development, cells differentiate and organize into tissues with precise spatial patterns. Combining scRNA-seq with spatial transcriptomics provides a powerful toolkit for studying this:

1. Trajectory reconstruction — scRNA-seq pseudotime analysis orders cells along differentiation paths, revealing the sequence of transcriptional changes during lineage commitment.
2. Discovery of novel cell states — Transient progenitor populations that exist only briefly during organogenesis can be identified by their unique expression signatures, even when they're morphologically indistinguishable from neighboring cells.
3. Spatial patterning — Spatial transcriptomics maps how signaling gradients (e.g., Wnt, BMP, Notch) establish anterior-posterior or dorsal-ventral axes during organ formation.
4. Cell-cell signaling in context — By knowing which cells are in physical contact, you can predict ligand-receptor interactions (e.g., Delta-Notch signaling between adjacent cells) with much greater confidence than from dissociated data alone.