10x Genomics Spatial Transcriptomics for FFPE and Frozen Tissues: Mapping Gene Expression in Tissue Context

A tumor section under a microscope reveals cells that look identical — same morphology, same H&E stain. But their gene expression programs can be radically different, and those differences are organized in space. A T cell at the tumor-invasive margin expresses exhaustion markers that a morphologically identical T cell in the tumor core does not. A fibroblast adjacent to malignant epithelium runs a matrix-remodeling program absent from fibroblasts three hundred microns away. These spatial patterns are invisible to dissociation-based single-cell RNA sequencing, which loses all positional information the moment tissue is dissociated. Spatial transcriptomics captures both — gene expression and tissue location — and 10x Genomics Visium has become the most widely adopted platform for doing so. This guide covers the Visium and Visium HD workflows for FFPE and fresh-frozen tissues, sample preparation requirements that determine data quality, and the analysis pipelines that turn spatial count matrices into biological insight.

Figure 1: 10x Visium Platform Overview — Visium, Visium HD, CytAssist, and Tissue Compatibility Flowchart

Why Spatial Context Changes Everything

The central limitation of single-cell dissociation is not technical — it is conceptual. Knowing that a fibroblast expressed COL1A1 is informative. Knowing that the fibroblast sits within fifty microns of a tumor nest that expresses TGFB1 is transformative. Tissue architecture encodes information that no amount of single-cell sequencing depth can recover: cell-cell communication gradients, tumor-immune interface dynamics, and structured microanatomical niches that define organ function.

This is why spatial transcriptomics has grown from a niche technology into a core component of single-cell and spatial biology services. Researchers who previously sent inquiries about scRNA-seq alone are now asking about paired spatial profiling — often on the same tissue blocks. The reason is practical: a 10x spatial transcriptome sequencing experiment can be run on the same FFPE block or fresh-frozen tissue that was already collected for histology or single-cell work, bridging the gap between tissue-level organization and cell-level resolution.

The 10x Visium family addresses two fundamentally different sample realities. Fresh-frozen tissue, collected prospectively and stored at -80°C, preserves the highest RNA quality and supports probe-free poly(A)-based capture chemistry. FFPE tissue, the dominant archival format in clinical and translational research, requires probe-based chemistry due to formalin-induced RNA fragmentation — but represents the vast majority of accessible specimens in biobanks and pathology archives. The Visium platform supports both, and the 2025 release of Visium HD has pushed spatial resolution from multicellular spots to near-single-cell bins.

Figure 2: Visium vs Visium HD Resolution Comparison — 55 µm Spots vs 2 µm Bins with H&E Overlay

The Visium and Visium HD Technology Stack

The core Visium workflow consists of three stages: tissue sectioning and quality control, probe hybridization or poly(A) capture on the Visium slide, and sequencing followed by computational analysis.

Standard Visium uses slides with approximately 5,000 barcoded capture areas, each containing spatially encoded oligonucleotides arranged in 55-micrometer spots. A single spot captures transcripts from roughly 2 to 10 cells depending on tissue density and cell size. After tissue permeabilization, released mRNA binds to capture probes, and reverse transcription incorporates both a spatial barcode and a UMI. The resulting cDNA library is sequenced on an Illumina platform, and the spatial barcode maps each transcript back to its tissue coordinate.

Visium HD, launched in early 2024 and refined through Protocol 2.0 in mid-2025, replaces the 55-micrometer spot array with more than 11 million continuous 2-micrometer features across a 6.5 × 6.5 mm capture area. There are no gaps between capture regions — the entire tissue section is mapped, and data can be binned at 2-micrometer, 8-micrometer, or 16-micrometer resolution in silico. A typical 8-micrometer bin in human tissue captures single-cell or near-single-cell resolution with approximately 200 to 600 UMIs per bin — sparser per unit than standard Visium spots, but with vastly finer spatial granularity.

Protocol 2.0, released in 2025, represents a substantial improvement for FFPE samples specifically. By reworking the probe hybridization, ligation, and wash chemistry, Protocol 2.0 delivers up to a two-fold sensitivity gain for high-quality FFPE blocks — those fixed for 6 to 24 hours and stored for less than three years. The gain is not uniform: aged or low-quality FFPE specimens with low DV200 scores show minimal improvement, while fresh-frozen samples show modest but consistent gains. For researchers planning spatial studies on archival clinical cohorts, this means block selection and quality assessment are more consequential than ever — a high-quality FFPE block with Protocol 2.0 can approach the sensitivity of fresh-frozen tissue processed with earlier chemistry versions.

A critical addition in 2025 is the Visium HD 3' Assay, which uses probe-free, poly(A)-based capture on fresh-frozen tissue. Unlike the standard probe-based Visium chemistry — which targets predefined gene panels validated only for human and mouse — the 3' assay is species-independent. Any organism with a polyadenylated transcriptome is compatible. This opens Visium HD to non-model species and captures both coding and non-coding polyadenylated transcripts, including long non-coding RNAs that probe panels exclude. The trade-off is that the 3' assay is fresh-frozen only — FFPE samples lack intact poly(A) tails and require the probe-based workflow.

For researchers whose projects require Visium HD spatial transcriptomics with single-cell-scale resolution, the choice between probe-based and 3' chemistry is dictated by sample type and species: probe-based for human/mouse FFPE or fresh-frozen, 3' assay for fresh-frozen from any species.

Figure 3: Visium HD Protocol 2.0 Sensitivity Comparison — FFPE Quality Tiers and Fresh Frozen Performance

Sample Requirements That Determine Data Quality

The quality of spatial transcriptomics data is primarily determined before the tissue ever touches a Visium slide. Two parameters dominate, and they differ by sample type.

For FFPE tissue, the key metric is DV200 — the percentage of RNA fragments longer than 200 nucleotides. 10x Genomics specifies DV200 of at least 30 percent for Visium and Visium HD, but in practice, values above 50 percent produce substantially better libraries. DV200 is measured from a single 5-micrometer section using an Agilent Bioanalyzer or TapeStation. Block age matters: blocks stored for more than three years typically show declining DV200 scores due to slow, cumulative RNA degradation even under optimal storage conditions. Fixation conditions at the time of block creation — formalin concentration, fixation duration, tissue size — also affect RNA quality but are rarely documented in archival specimens.

For fresh-frozen tissue, the standard metric is RIN — RNA Integrity Number — with a recommended minimum of 7. Tissue should be embedded in OCT, snap-frozen in liquid nitrogen or isopentane-cooled on dry ice, and stored at -80°C. A common failure mode is slow freezing: when tissue freezes gradually, ice crystals shear cellular structures and degrade RNA. Isopentane chilled to -40°C to -80°C on dry ice provides faster, more uniform freezing than direct liquid nitrogen immersion, which can cause tissue cracking due to the Leidenfrost effect.

For 10x Visium FFPE or fresh-frozen spatial transcriptomics, section placement within the capture area is another critical variable. The standard Visium capture area is 6.5 × 6.5 mm; Visium HD recently added an 11 × 11 mm option (roughly three times the area) for whole-organ sections or multi-sample tissue microarrays. Sections must be placed flat and without folds or tears. Bubbles under the section create capture gaps. For FFPE, sections are typically 5 micrometers thick; for fresh-frozen, 10 micrometers — the thicker fresh-frozen section compensates for lower cellular density compared to dehydrated FFPE tissue.

The CytAssist instrument, used in all current Visium workflows, standardizes the probe hybridization and tissue registration steps. It images the tissue section, registers fiducial markers on the slide for spatial alignment, and facilitates probe transfer from the slide to the tissue — reducing the manual variability that previously plagued spatial workflows. All Visium and Visium HD protocols — FFPE probe-based, fresh-frozen probe-based, and fresh-frozen 3' assay — now pass through CytAssist. For labs without in-house CytAssist access, 10x spatial transcriptome sequencing services include CytAssist processing as part of the standard workflow.

From Raw Data to Spatial Maps: The Analysis Pipeline

Spatial transcriptomics data analysis has matured rapidly in the 2025-2026 period, with new tools and updated versions of established software addressing the unique challenges of spatially resolved count data.

Space Ranger, 10x Genomics' command-line preprocessing tool, handles alignment (STAR), spatial barcode assignment, UMI counting, and image registration. For Visium HD data, Space Ranger 4.0 and later (4.1 is the current release as of mid-2026) automatically run a cell segmentation pipeline based on a customized StarDist deep-learning model that identifies nuclei from the H&E image and expands boundaries to approximate cell margins. The output includes binned count matrices at multiple resolutions, spatial coordinates, and segmentation polygons — essentially a single-cell-resolution spatial dataset from what was originally a spot-based technology.

Downstream analysis typically begins in Seurat. The December 2025 release of Seurat v5.4 introduced native support for Visium HD segmentation data, including the ability to load polygon-level data directly via Load10X_Spatial(bin.size = "polygons"). The standard workflow — SCTransform normalization, PCA dimensionality reduction, graph-based clustering, and UMAP visualization — applies to spatial data, but the critical additional step is overlaying clusters back onto the tissue image. A cluster that looks clean on UMAP but distributes randomly across the tissue section is likely a technical artifact, not a biological population. Seurat's SpatialFeaturePlot and SpatialDimPlot render gene expression and cluster assignments directly on H&E images, and interactive Shiny-based exploration is now supported for custom spatial transcriptomics bioinformatics workflows.

For projects integrating multiple spatial technologies or modalities, Giotto Suite — published in Nature Methods in October 2025 — provides a modular, technology-agnostic R framework. Giotto handles data from Visium, Xenium, MERFISH, and Stereo-seq within a unified data structure, making it the tool of choice for multi-platform spatial studies. It also supports interactive tissue annotation, subcellular analysis, and co-registration of transcriptomics with proteomics or immunofluorescence images.

Figure 4: Spatial Transcriptomics Analysis Pipeline — From Space Ranger to Seurat/Giotto with Key QC Checkpoints

Cell-type deconvolution — inferring the cellular composition of each Visium spot or bin — is performed by integrating a matched or reference scRNA-seq dataset. cell2location, a Bayesian method implemented in Python, has become the most widely adopted deconvolution tool for spatial data in 2025-2026, offering fine-grained cell-type proportion estimates per spatial location. The method models cell-type-specific expression signatures as negative binomial distributions with spatial priors, producing per-location abundance estimates with associated uncertainty intervals — a critical advantage over deterministic methods that return point estimates without confidence bounds. RCTD (robust cell-type decomposition) and Tangram provide alternative approaches, with cell2location favored for discovery-focused projects where uncertainty quantification matters and RCTD preferred when a well-annotated, high-quality single-cell reference is available and computational speed is a priority.

Several quality checks are spatial-specific and should not be skipped. The first is visual alignment verification: do the fiducial markers on the slide image align precisely with the tissue? A rotation or offset of even 100 micrometers can misassign thousands of transcripts. The integration of high-resolution histology images with spatial molecular data is critical — methods that co-register H&E staining with transcriptomic coordinates enable precise structure-level annotation and have been shown to improve spatial assignment accuracy when fiducial-based alignment is supplemented with image-based registration (8). The second is spatial autocorrelation analysis: biologically meaningful gene expression patterns should show structured spatial variation. If no genes show significant spatial autocorrelation (e.g., by Moran's I), either the tissue is genuinely unstructured or — more likely — there is a data quality problem. The third is cluster-anatomy concordance: identified spatial clusters should correspond to recognizable anatomical structures in the H&E image. A cluster that spans necrotic tissue, adipose, and epithelium uniformly is a QC failure.

Applications: Where Spatial Transcriptomics Delivers

The strongest use case for spatial transcriptomics is the tumor microenvironment. Bulk and single-cell approaches can identify cell types present in a tumor but cannot determine whether CD8+ T cells are infiltrating malignant epithelium or restricted to the stromal periphery — a distinction with therapeutic implications. Visium-based studies have mapped immune exclusion zones, identified tertiary lymphoid structures at the tumor-invasive margin, and characterized fibroblast activation gradients radiating outward from tumor nests. A typical oncology spatial project profiles 4 to 12 tissue sections, often from paired tumor and adjacent normal tissue from the same patient, with matched scRNA-seq data used for spot deconvolution.

Figure 5: Spatial Transcriptomics Applications — Tumor Microenvironment, Neuroscience, and Developmental Biology Use Cases

Neuroscience is the second major application domain. The brain's spatial organization is intrinsically functional — layer-specific gene expression in the cortex, region-specific vulnerability in neurodegenerative disease, and spatially patterned transcriptional responses to injury. Visium HD, with its 2-micrometer bin resolution, can resolve cortical layer-specific expression patterns that standard Visium spots would average across multiple layers. A 2025 study using Visium HD on fresh-frozen mouse brain demonstrated clear separation of layer II/III, IV, V, and VI transcriptional programs within a single tissue section — resolution previously achievable only by laser-capture microdissection followed by bulk RNA-seq. In the same study, spatially restricted gene modules corresponding to the canonical hippocampal subfields CA1, CA3, and dentate gyrus were resolved from a single coronal section, each defined by fewer than 100 differentially expressed genes that would be diluted beyond detection in a bulk or standard Visium measurement. For neurodegenerative disease research, this resolution enables spatially precise mapping of amyloid-plaque-associated microglial activation signatures and tau-pathology-proximal neuronal stress programs — transcriptional gradients that radiate outward from pathological foci and fade within a few hundred microns.

Developmental biology, plant biology, and non-model organism research are emerging application areas enabled by the Visium HD 3' assay. A frog embryo at neurulation, a regenerating zebrafish fin, or a developing rice panicle can now be spatially profiled without species-specific probe design — the poly(A)-based chemistry works universally. For single-cell and spatial biology services supporting multi-species research, this universality removes what was previously a major barrier to spatial transcriptomics adoption outside of human and mouse.

Practical guidance on project scale: a typical Visium experiment profiles 4 to 8 capture areas, each corresponding to one tissue section. Sequencing depth of 25,000 to 50,000 read pairs per spot (standard Visium) or per 8-micrometer bin (Visium HD) provides sufficient coverage for gene detection and clustering. For a standard four-section Visium project, this translates to approximately 200 to 400 million read pairs total. Biological replicates — tissue sections from multiple donors or multiple regions from the same donor — are essential. A single tissue section, no matter how deeply sequenced, is an N of 1 and cannot support generalization.

FAQ

What is the difference between Visium and Visium HD?

Standard Visium uses 55-micrometer spots that capture transcripts from 2-10 cells each, with gaps between spots. Visium HD uses more than 11 million continuous 2-micrometer features with no gaps, enabling near-single-cell-resolution mapping. Visium HD also supports an 11 mm capture area and a 3' poly(A)-based assay for species-independent fresh-frozen profiling.

Can I use FFPE tissue blocks for spatial transcriptomics, or do I need fresh-frozen?

Both are compatible. FFPE tissue uses probe-based chemistry and requires DV200 of at least 30 percent, preferably above 50 percent. Fresh-frozen tissue supports both probe-based and poly(A)-based 3' chemistry and requires RIN of at least 7. Visium HD Protocol 2.0 (2025) delivers up to two-fold sensitivity improvement specifically for high-quality FFPE blocks.

What is the CytAssist instrument and do I need it?

CytAssist is the standard tissue-processing instrument for all current Visium workflows. It automates probe hybridization, tissue imaging, fiducial registration, and probe transfer — reducing the manual variability of earlier Visium protocols. All Visium and Visium HD workflows now use CytAssist.

How much does a spatial transcriptomics project cost?

Cost depends on the number of capture areas, the Visium chemistry (standard or HD), and sequencing depth. A typical four-section project ranges from roughly $5,000 to $15,000 depending on these variables. For concrete pricing based on your specific tissue type and experimental goals, contact CD Genomics directly with your project parameters.

What bioinformatics tools do I need to analyze Visium data?

The core pipeline is Space Ranger (preprocessing) followed by Seurat v5.4+ (analysis and visualization). For multi-platform or multi-modal spatial studies, Giotto Suite provides a technology-agnostic framework. Cell-type deconvolution is typically performed with cell2location or RCTD, using matched scRNA-seq data as reference.

How many tissue sections should I include in a spatial transcriptomics study?

A minimum of 3-4 sections per condition is recommended for biological replication. A single section, from a single donor, cannot support statistical generalization. For tumor studies, paired tumor and adjacent-normal sections from the same patient provide the most informative internal control, but still require replication across multiple donors.

Can spatial transcriptomics be combined with single-cell RNA-seq from the same tissue?

Yes, and this is a common experimental design. scRNA-seq from dissociated tissue provides cell-type annotations and gene signatures that are then mapped onto Visium spots or bins via deconvolution. The combination of single-cell resolution (from scRNA-seq) and spatial context (from Visium) is more informative than either approach alone.

What is the turnaround time for a spatial transcriptomics project?

A typical project of 4-8 capture areas requires 3-5 weeks from tissue receipt to sequencing completion, plus 2-4 weeks for spatial bioinformatics analysis including alignment, clustering, deconvolution, and spatial feature visualization.

What tissue types have been experimentally validated for Visium HD?

10x Genomics maintains a public database of tested FFPE and fresh-frozen tissues including human and mouse brain, kidney, lung, breast, colon, liver, and spleen. Each entry includes DV200 or RIN metrics, recommended permeabilization time, and expected gene detection benchmarks. Always consult the latest tested-tissues list for your tissue of interest before committing samples — particularly for challenging specimens such as bone, adipose, or highly pigmented tissues where protocol optimization may be required.

References:

1. Stahl PL, et al. Visualization and analysis of gene expression in tissue sections by spatial transcriptomics. Science. 2016;353(6294):78-82. https://doi.org/10.1126/science.aaf2403
2. Rademacher A, et al. Comparison of spatial transcriptomics technologies using tumor cryosections. Genome Biology. 2025;26:104. https://doi.org/10.1186/s13059-025-03624-4
3. Solano A, Yip RKH, Wang C, et al. SpatialBench: comparative cross-platform benchmarking of high-resolution spatial transcriptomics using matched mouse lymphoid tissue. bioRxiv. 2026. https://doi.org/10.64898/2026.04.29.721531
4. Ozirmak Lermi N, Molina Ayala M, Hernandez S, et al. Comparison of imaging-based single-cell resolution spatial transcriptomics profiling platforms using formalin-fixed paraffin-embedded tumor samples. Nature Communications. 2025;16:63414. https://doi.org/10.1038/s41467-025-63414-1
5. Hao Y, et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nature Biotechnology. 2024;42:293-304. https://doi.org/10.1038/s41587-023-01767-y
6. Chen JG, Chavez-Fuentes JC, O'Brien M, et al. Giotto Suite: a multiscale and technology-agnostic spatial multiomics analysis ecosystem. Nature Methods. 2025;22(10):2052-2064. https://doi.org/10.1038/s41592-025-02817-w
7. Kleshchevnikov V, et al. Cell2location maps fine-grained cell types in spatial transcriptomics. Nature Biotechnology. 2022;40:661-671. https://doi.org/10.1038/s41587-021-01139-4
8. Bergenstrahle J, Larsson L, Lundeberg J. Seamless integration of image and molecular analysis for spatial transcriptomics workflows. BMC Genomics. 2020;21:482. https://doi.org/10.1186/s12864-020-06832-3

For research purposes only, not intended for clinical diagnosis, treatment, or individual health assessments.