Spatial Transcriptomics: Mapping the Cellular Atlas

Introduction

Spatial transcriptomics represents a paradigm shift in molecular pathology and systems biology. Unlike traditional single-cell RNA sequencing (scRNA-seq) which dissociates tissues and loses all positional information, spatial transcriptomics preserves the anatomical context of gene expression. This technology enables researchers to construct cellular atlases that map the transcriptional landscape of tissues at single-cell resolution. For veterinary medicine, the ability to localize gene expression within intact tissue sections provides unprecedented insights into host-pathogen interactions, developmental biology, and disease pathogenesis across diverse animal species.

The fundamental biophysical challenge of spatial transcriptomics is the capture and measurement of RNA molecules from defined spatial coordinates within a tissue section. Two principal methodological categories exist: sequencing-based approaches that capture transcripts from spatially barcoded arrays, and imaging-based approaches that directly visualize RNA molecules in situ via fluorescence microscopy. Each approach carries distinct trade-offs in resolution, throughput, and sensitivity.

Biophysical and Chemical Principles

Sequencing-Based Spatial Transcriptomics

Sequencing-based methods rely on a spatially barcoded capture surface. A tissue section is placed onto a microarray slide functionalized with oligonucleotide probes. Each probe contains a poly-T sequence to capture polyadenylated mRNA, a unique spatial barcode, a unique molecular identifier (UMI), and a sequencing primer binding site. The physical position of each barcode on the array is known. After tissue permeabilization, mRNA molecules diffuse downward and hybridize to the capture probes. Reverse transcription generates cDNA that incorporates the spatial barcode, linking each transcript to its original location. The cDNA is then pooled, amplified, and sequenced. Computational reconstruction maps sequencing reads back to their spatial coordinates using the barcode information.

The resolution of sequencing-based methods is determined by the center-to-center distance between capture spots. Modern arrays achieve spot diameters of approximately 10 micrometers with inter-spot distances of 10 to 100 micrometers. Each spot captures mRNA from multiple cells, resulting in a multcellular resolution. Computational deconvolution algorithms can infer cell-type compositions at each spot by comparing transcriptomic profiles to scRNA-seq reference datasets.

Imaging-Based Spatial Transcriptomics

Imaging-based methods achieve true single-molecule resolution. Two major subcategories exist: single-molecule fluorescence in situ hybridization (smFISH) and in situ sequencing. In smFISH, multiple oligonucleotide probes targeting a single mRNA species are hybridized to the tissue section. Each probe carries a fluorophore. The summed fluorescence from all bound probes produces a diffraction-limited spot that can be detected by microscopy. Multiplexing is achieved through sequential rounds of hybridization, imaging, and probe stripping. Error-robust barcoding schemes assign each mRNA species a unique sequence of fluorophore signals across rounds.

In situ sequencing approaches use padlock probes or sequencing-by-ligation chemistry to read the RNA sequence directly within the tissue. Padlock probes circularize upon hybridization to a target RNA. The circularized probe is then amplified by rolling circle amplification to generate a nanoball of DNA that is localized to the original RNA molecule. Sequencing-by-ligation cycles read the barcode on the amplified product, identifying the target gene. This approach permits high multiplexing and subcellular localization of transcripts.

Computational Pipelines for Data Analysis

Spatial transcriptomics generates high-dimensional data requiring specialized computational pipelines. The analysis workflow includes preprocessing, alignment, normalization, dimensionality reduction, clustering, and downstream interpretation.

Preprocessing and Alignment

Raw sequencing data from sequencing-based methods must be demultiplexed based on the spatial barcode and UMI sequences. Reads are aligned to the reference transcriptome using splice-aware aligners. The output is a count matrix where rows represent spatial coordinates and columns represent genes. For imaging-based data, computational image processing identifies punctate fluorescence signals, segments cells based on nuclear staining or membrane markers, and assigns transcripts to individual cells.

Normalization and Integration

Spatial transcriptomics data exhibits technical variation due to differences in capture efficiency, permeabilization efficiency, and sequencing depth across spots. Normalization methods include SCTransform, scran, or spatial-specific approaches that account for spatial autocorrelation. Integration of spatial transcriptomics with scRNA-seq data from the same tissue type enables cell-type annotation of spatial spots using label transfer methods or reference-based deconvolution.

Spatial Domain Identification

Identifying regions of coherent gene expression, known as spatial domains, is a core analytical task. Methods such as BayesSpace, stLearn, and SpaGCN incorporate spatial coordinates into clustering algorithms. These approaches use Markov random fields or graph neural networks to model spatial dependencies. The resulting spatial domains often correspond to histological structures such as lymphoid follicles, tumor margins, or granuloma cores.

Mermaid diagram: Spatial transcriptomics pipeline

graph TD
    A[Fresh Frozen or FFPE Tissue Section], > B[Sequencing Capture Array or In Situ Hybridization]
    B, > C{Method Type}
    C, >|Sequencing Based| D[barcoded cDNA Library Preparation]
    C, >|Imaging Based| E[Multiplexed Fluorescence Microscopy]
    D, > F[High Throughput Sequencing]
    F, > G[Demultiplexing and Read Alignment]
    G, > H[Spatial Count Matrix]
    H, > I[Normalization and Integration]
    I, > J[Cell Type Deconvolution]
    I, > K[Spatial Domain Detection]
    J, > L[Cellular Atlas Construction]
    K, > L
    E, > M[Image Processing and Spot Detection]
    M, > N[Single Moleule Localization]
    N, > O[Sequential Barcode Decoding]
    O, > P[Single Cell Expression Matrix]
    P, > Q[Cellular Atlas with Subcellular Resolution]
    L, > R[Biological Interpretation]
    Q, > R
    R, > S[Differential Expression, Ligand Receptor Analysis]
    S, > T[Downstream Validation and Hypothesis Generation]

Applications in Veterinary Medicine

Spatial transcriptomics has been applied across multiple veterinary species and disease contexts. The following sections detail representative studies and their implications for clinical veterinary medicine.

Ocular Atlas and Retinal Artery Occlusion in Mice

Du and colleagues [1] generated a comprehensive single-cell and spatial transcriptomic atlas of the mouse eye. This study integrated data from multiple time points and tissue processing conditions to map the spatial organization of retinal cell types. The authors applied this atlas to a model of retinal artery occlusion (RAO), a condition analogous to ocular vascular occlusive disease in companion animals. Spatial transcriptomics revealed region-specific transcriptional responses to ischemia, including upregulation of inflammatory pathways in the inner retina and metabolic stress markers in photoreceptors. These findings have direct relevance to veterinary ophthalmology, where conditions such as sudden acquired retinal degeneration syndrome in dogs and ischemic optic neuropathy in horses remain poorly understood at the molecular level. The spatial atlas provides a reference framework for interpreting transcriptional changes in these naturally occurring diseases.

Host Response to Oropharyngeal Candidiasis

Nabeela and colleagues [2] used spatial transcriptomics to map the host transcriptional response to oropharyngeal candidiasis in a murine model. This study demonstrated the power of spatial transcriptomics to resolve the architecture of host-pathogen interactions at the mucosal surface. The authors identified distinct zones of transcriptional activity extending from the epithelial surface into the underlying lamina propria. Epithelial cells directly adjacent to fungal hyphae showed upregulation of antimicrobial peptides and alarmins, while deeper stromal cells exhibited type 17 immune signatures. For veterinary clinicians, these insights are directly applicable to oral candidiasis in poultry, particularly in the context of antibiotic-associated dysbiosis and immunosuppressive viral infections. The spatial mapping approach could be extended to other mucosal pathogens affecting veterinary species, including Candida albicans infections in psittacine birds and Gallibacterium anatis salpingitis in laying hens.

Tuberculous Pericarditis

Xiong and colleagues [3] constructed a single-cell and spatial transcriptomic atlas of human tuberculous constrictive pericarditis. This study identified spatially organized immune aggregates, including granuloma cores surrounded by fibrotic capsules. The transcriptional profiles of macrophages within granuloma cores showed a distinct activation state intermediate between M1 and M2 polarization. For veterinary medicine, these findings have direct comparative value for bovine tuberculosis caused by Mycobacterium bovis. The spatial organization of granulomas in cattle lymph nodes and lung tissue is a hallmark of infection surveillance in eradication programs. Spatial transcriptomics could enable detailed characterization of bovine tuberculous lesions, identifying biomarkers of active infection versus latent carriage. This would improve diagnostic accuracy in herds where tuberculin skin testing yields ambiguous results.

Technical Considerations for Veterinary Tissues

The application of spatial transcriptomics to veterinary tissues requires optimization of tissue handling protocols. Fresh frozen tissue remains the gold standard for preserving RNA integrity. However, many veterinary diagnostic samples are formalin fixed paraffin embedded (FFPE). Recent advances in spatial transcriptomics for FFPE tissues use probe-based capture targeting specific gene panels rather than full transcriptome profiling. This approach is compatible with degraded RNA and expands the utility of archived pathology specimens.

Tissue quality metrics differ between species. High melanin content in equine and bovine skin can interfere with fluorescence based imaging methods. Fatty tissues, such as those from adipose depots in obese companion animals, require modified permeabilization protocols to ensure efficient RNA capture. The presence of endogenous biotin in some mammalian tissues may produce background signals in streptavidin based detection systems.

Future Directions

The integration of spatial transcriptomics with proteomic and metabolomic measurements represents the next frontier. Multi-omic spatial platforms that simultaneously capture RNA, protein, and metabolite distributions in the same tissue section will provide a more complete picture of cellular function. For veterinary infectious disease research, spatial transcriptomics combined with pathogen detection probes could map the direct interface between host cells and invading microorganisms.

Computational challenges remain in the development of robust deconvolution methods for samples with mixed cell populations and in the handling of technical artifacts such as tissue folding, tears, and uneven permeabilization. The generation of reference atlases for common veterinary species, including dogs, cats, cattle, and poultry, will accelerate translational research and improve diagnostic precision.

Conclusion

Spatial transcriptomics enables the construction of cellular atlases that preserve the anatomical organization of gene expression. This technology provides a bridge between traditional histopathology and molecular biology, offering new insights into disease mechanisms at the tissue level. For veterinary medicine, spatial transcriptomics has the potential to transform our understanding of host-pathogen interactions, developmental abnormalities, and neoplastic progression across a wide range of animal species. Continued methodological refinement and the establishment of species-specific reference atlases will be essential for translating these discoveries into clinical diagnostic applications.

References

[1] Du C, Li Y, Li Z, et al. Single-Cell Annotation and Localization via Integrating Spatial Transcriptomics Maps the Mouse Ocular Atlas and RAO Dynamics. Adv Sci (Weinh). URL: https://pubmed.ncbi.nlm.nih.gov/42220076/

[2] Nabeela S, McSwiggin H, Magalhaes RDM, et al. A spatial transcriptomic atlas of the host response to oropharyngeal candidiasis. mBio. URL: https://pubmed.ncbi.nlm.nih.gov/40586608/

[3] Xiong F, Qi Y, Wang S, et al. A single-cell and spatial transcriptomic atlas of human tuberculous constrictive pericardium. EBioMedicine. URL: https://pubmed.ncbi.nlm.nih.gov/41875497/