Stereo-seq in Action: Five Recent Discoveries Across Biomedicine

Inquiry

Figure 1. Five recent Stereo-seq-powered studies spanning developmental biology, oncology, cardiovascular research, evolutionary neuroscience, and immunology.

Summary

Spatial transcriptomics is moving fast. In the past eighteen months, Stereo-seq — a sequencing-based platform that combines 500 nm subcellular resolution with centimeter-scale capture areas — has powered discoveries published in Nature Cell Biology, Nature Methods, Science, Advanced Science, and Nature Communications. These are not pilot studies or proof-of-concept demonstrations. They are full-scale research programs that used Stereo-seq to answer questions that bulk sequencing, single-cell RNA-seq, and targeted spatial methods could not resolve alone. This article rounds up five of the most instructive recent examples, covering early human development, pan-cancer lncRNA biology, aortic dissection, vertebrate brain evolution, and cutaneous lupus. Each section explains what the study found and why Stereo-seq was the enabling technology.

Key Takeaways

• Stereo-seq now supports discovery across a wide range of biomedical fields. The five studies covered here span developmental biology, cancer genomics, cardiovascular disease, evolutionary neuroscience, and autoimmune dermatology.
• Each study used a different Stereo-seq capability. These include 3D reconstruction from serial sections, pan-cancer multi-cohort integration, cross-species multi-omics comparison, FFPE-compatible total RNA profiling, and large-field-of-view whole-organ mapping.
• These are not small-scale pilots. The largest study integrated 28 spatial transcriptomic datasets and 24 scRNA-seq datasets across 13 cancer types. Another processed 110 aortic tissue samples from 80 individuals. The embryo study reconstructed 82 serial cryosections into a 3D atlas.
• The common thread is discovery of spatially organized biology that single-cell methods alone cannot resolve. In every case, spatial context — where cells sit relative to one another in tissue — turned a descriptive catalog into a mechanistic finding.

Why Stereo-seq Is Gaining Research Traction

The platform dynamics that make a spatial transcriptomics technology "ready for discovery" have changed. Three years ago, most spatial studies were method benchmarking exercises — does this technology detect enough genes? Can it resolve cell types? Is the signal reproducible? Those questions are now largely settled for sequencing-based spatial methods like Stereo-seq. The technology has entered its application phase, where the same platform features that were once novel — subcellular resolution, whole-transcriptome capture, large-area imaging — are being deployed as standard tools in ambitious biological projects.

Several platform features recur across the studies below. The combination of 500 nm spot spacing and centimeter-scale chip formats (standard 1 × 1 cm, expandable to 13.2 × 13.2 cm) means that researchers can capture tissue architecture at near-single-cell resolution without sacrificing the field of view needed for whole-organ mapping. The non-targeted, poly(A)-based capture chemistry means that Stereo-seq detects coding and non-coding RNA alike — critical for the lncRNA study — and works across any species with a polyadenylated transcriptome, from human to lamprey. The availability of FFPE-compatible protocols (Stereo-seq V2/OMNI) extends the platform to archived clinical specimens, as demonstrated in the lupus and aortic dissection studies. For researchers evaluating whether Stereo-seq fits their project, the When to Choose Stereo-seq decision guide covers platform selection criteria in detail.

A Human Embryo Blueprint at Gastrulation

Cui L, Lin S, Yang X, et al. Nature Cell Biology. 2025;27:360–369.

Human gastrulation — the process that establishes the three germ layers and the body axis — occurs during the third week of development, a stage so early and so difficult to access that it has been called a developmental black box. The Carnegie stage 7 (CS7) embryo, approximately 15–17 days post-fertilization, captures the moment when ectoderm, mesoderm, and endoderm are first specified and spatially organized. Until now, no three-dimensional molecular map of this stage existed.

The study team obtained a single intact CS7 human embryo and cut 82 serial cryosections, processing every section with Stereo-seq. They reconstructed a 3D spatial transcriptomic atlas defining 11 major tissue clusters, including epiblast, amniotic ectoderm, mesoderm subtypes, endoderm, and yolk sac. A second embryo was used for immunofluorescence validation.

Three findings stand out. First, the study distinguished paraxial, lateral plate, and extraembryonic mesoderm subtypes — lineages that are notoriously difficult to separate by transcriptome alone but occupy distinct spatial domains, making Stereo-seq's spatial resolution essential. Second, the anterior visceral endoderm (AVE), a signaling center critical for anterior-posterior axis establishment, was localized to the anterior midline — the first direct spatial confirmation of AVE positioning in a human embryo. Third, primordial germ cells (PGCs) were mapped to the connecting stalk rather than the posterior embryo, a finding that revises the accepted model of early human germ cell localization. The study also captured HSC-independent hematopoiesis in the yolk sac, pushing the timeline of human blood formation earlier than previously documented.

This study is a clear example of what 3D Stereo-seq reconstruction can deliver that single-cell RNA-seq cannot: the spatial coordinates that transform a list of cell types into an anatomical map.

Figure 2. 3D reconstruction of a Carnegie stage 7 human embryo from 82 serial cryosections processed with Stereo-seq (Cui et al. 2025).

Pan-Cancer lncRNA Diversity Mapped Spatially

Prakrithi P, Vo T, Xiong Z, et al. Nature Methods. 2026;23:1236–1249.

Long non-coding RNAs are abundant in the human genome, but most remain unannotated and functionally uncharacterized. The problem is compounded by the fact that lncRNAs are typically expressed at lower levels than mRNAs, show stronger tissue and cell-type specificity, and are frequently missed by standard scRNA-seq pipelines optimized for protein-coding genes. Adding spatial context — knowing where a lncRNA is expressed within a tumor — had not been attempted at pan-cancer scale.

This study addressed that gap directly. The team integrated 28 spatial transcriptomic datasets (Stereo-seq, Xenium, and Takara Bio Seeker) with 24 scRNA-seq datasets, covering 13 cancer types. They identified 219,442 cancer-associated transcribed regions (cuTARs), of which 94,795 were previously unannotated — roughly doubling the catalog of tumor-associated lncRNA candidates.

The core finding is that these cuTARs are not randomly distributed. They show strong cancer-type specificity, cell-type specificity, and spatial heterogeneity within tumors. Several cuTARs co-vary spatially with coding genes involved in cell proliferation, immune response, angiogenesis, and metastasis, suggesting they are not transcriptional noise but functional participants in tumor biology. Some cuTARs outperformed traditional coding-gene markers in patient prognosis models. The team also built SPanC-Lnc, an interactive database that makes the full pan-cancer atlas accessible to the research community.

This study illustrates Stereo-seq's value for discovery of non-coding biology. Because the platform captures total polyadenylated RNA — not a preselected gene panel — it detects lncRNAs without requiring prior annotation. For research teams working on non-coding RNA in tissue contexts, this is a differentiating capability that targeted spatial methods cannot match.

Metabolic-Immune Axis Driving Aortic Dissection

Tao J, Yang H, Yong J, et al. Advanced Science. 2026;e75509.

Aortic dissection is a catastrophic cardiovascular event with mortality rates exceeding 50% without emergency intervention. The disease involves breakdown of the aortic wall, but the cellular sequence — what happens first, and what drives progression — has been difficult to establish because the process involves multiple cell types (smooth muscle cells, fibroblasts, immune cells) interacting across tissue layers that cannot be resolved by bulk or dissociated-cell methods.

This study built one of the largest single-cell and spatial atlases of thoracic aortic disease to date: 110 tissue samples from 80 individuals, covering normal aorta, aortic aneurysm, and aortic dissection, yielding 767,018 high-quality cells analyzed by both scRNA-seq and Stereo-seq.

The key discovery is a metabolic-immune axis that connects vascular smooth muscle cell (vSMC) dysfunction to immune-mediated tissue destruction. Under hypoxic conditions within the diseased aortic wall, vSMCs undergo ENO1-driven glycolytic reprogramming. This metabolic shift causes them to lose their contractile phenotype and begin secreting macrophage migration inhibitory factor (MIF). MIF, in turn, recruits and polarizes macrophages toward a pro-inflammatory state. These activated macrophages upregulate proteolytic and fibrinolytic pathways that degrade extracellular matrix, accelerating wall breakdown in a positive-feedback loop. The study further showed that ENO1 knockdown in vivo reduces vSMC phenotypic switching, dampens macrophage inflammation, and slows disease progression.

Stereo-seq was indispensable here because it placed each cell state in its correct anatomical layer — intima, media, or adventitia — revealing that vSMC metabolic reprogramming initiates in the media before immune amplification occurs in the adventitia. That spatial sequence, from media to adventitia, is the mechanistic spine of the story and could not have been inferred from scRNA-seq alone.

Figure 3. Spatial organization of the ENO1-MIF metabolic-immune axis in aortic dissection — from vSMC reprogramming in the media to macrophage-driven ECM degradation in the adventitia (Tao et al. 2026).

An Ancestral Blueprint of Vertebrate Brains

Wu H, Chen D, Li J, et al. Science. 2026;392(6804):eaaea2535.

How did the vertebrate brain evolve from a simple ancestral structure into the complex organ found in modern mammals, birds, and fish? The lamprey — a jawless vertebrate that diverged from the jawed-vertebrate lineage over 500 million years ago — offers a unique window into the ancestral brain organization that predates the emergence of jaws, myelin, and the elaborate cortical structures of mammals.

The research team constructed a 3D whole-brain molecular atlas of the adult lamprey (Lethenteron reissneri) by combining Stereo-seq spatial transcriptomics with single-nucleus RNA-seq. They identified 14 major brain regions and 209 cell clusters, then performed cross-species comparisons against eight species' single-cell data and five species' spatial transcriptomic data.

Two major insights emerged. The first is conservation: brain regions including the olfactory bulb, thalamus, and hindbrain retain spatial organization features that are recognizably shared between lamprey and mouse, indicating that a surprisingly mature brain blueprint was already present in the common vertebrate ancestor. The second is specialization: despite this deep conservation, lineage-specific innovations are evident — particularly in forebrain laminar organization and midbrain cell populations — suggesting that vertebrate brain complexity arose not by inventing new brain regions wholesale but by elaborating and refining an ancient structural scaffold. The study also identified cell populations in the lamprey hindbrain with molecular features resembling cerebellar neurons, implying that a cerebellar ground plan predates the anatomical cerebellum of jawed vertebrates.

This study, published as a Science cover article, demonstrates Stereo-seq's species-agnostic capability: lamprey brain tissue is as accessible to Stereo-seq as mouse or human tissue, because the capture chemistry requires only polyadenylated RNA — no species-specific probe design.

Pathogenic Niches Shaping Lupus Skin Lesions

Zhou W, Huang Y, Lei Y, et al. Nature Communications. 2026.

Cutaneous lupus erythematosus manifests as chronic, disfiguring skin lesions that resist conventional treatment. Histologically, the lesions show immune infiltrates, but the spatial logic — which cells drive inflammation, from where, and through which signaling pathways — has remained unclear. This gap has limited both the biological understanding of lesion pathogenesis and the identification of spatially defined therapeutic targets.

The study constructed the first single-cell-resolution spatial transcriptomic atlas of lesional skin from patients with discoid lupus erythematosus (DLE) and systemic lupus erythematosus (SLE), using Stereo-seq on FFPE tissue sections. A key technical achievement was the successful application of Stereo-seq to archived FFPE dermatology specimens, demonstrating the platform's compatibility with clinically derived samples.

The defining finding is the identification of a pathogenic keratinocyte-immune interface at the epidermal-dermal junction. A population of stress keratinocytes positioned precisely at this boundary upregulates CXCL9, CXCL10, and CXCL11 — chemokines that signal through CXCR3 on T cells and plasma cells — via the IFN-γ–JAK–STAT axis. This chemokine gradient actively recruits inflammatory cells into the skin, and the effect is significantly stronger in active SLE lesions than in chronic DLE, explaining the more aggressive clinical course of SLE skin involvement.

A second major finding was the presence of organized tertiary lymphoid structures (TLS) within chronic lesions. These TLS-like niches contain age-associated B cells, pre-germinal center B cells, Tregs, T follicular helper/peripheral helper cells, memory T cells, and CCL14+ vascular endothelial cells — a cellular composition that closely resembles lymphoid follicle organization. These structures likely sustain chronic inflammation and contribute to lesion persistence and recurrence. This discovery reclassifies lupus skin lesions from "diffuse immune infiltrates" to "spatially organized pathogenic ecosystems," with immediate implications for therapeutic targeting strategies.

Figure 4. Pathogenic niches in cutaneous lupus — stress keratinocytes at the epidermal-dermal junction recruit immune cells via CXCL9-11/CXCR3, while organized tertiary lymphoid structures sustain chronic inflammation (Zhou et al. 2026).

What These Five Studies Reveal About the Platform

Looking across these five studies, several patterns emerge that matter for researchers planning spatial transcriptomics projects.

Serial section 3D reconstruction is becoming routine. The embryo study used 82 sections; the brain study used full-organ serial sections. The ability to reconstruct three-dimensional molecular atlases from consecutive tissue slices — aligning each section's spatial coordinates into a unified volume — is moving from technical feat to standard workflow. For researchers studying developing organs, tumor margins, or any biology where the third dimension matters, organ-scale Stereo-seq study design has become a practical option.

FFPE compatibility unlocks retrospective cohorts. The aortic dissection and lupus studies both used clinical FFPE specimens. This means that archived pathology blocks — often accompanied by years of outcome data — are now accessible for spatial transcriptomic reanalysis. The ability to go back to existing cohorts rather than waiting for prospective collections substantially shortens the timeline from question to answer. Note that specific project timelines vary depending on sample availability, tissue quality, and experimental design — contact CD Genomics for a feasibility assessment and timeline estimate for your spatial transcriptomics project.

Multi-omics integration is the norm, not the exception. Every study paired Stereo-seq with at least one complementary modality: scRNA-seq, snRNA-seq, immunofluorescence, or long-read sequencing. Spatial data is most powerful when it contextualizes cell states that have been defined at single-cell resolution. The studies here treat this integration as a standard part of the experimental design, not an afterthought.

Non-coding and non-model-organism applications are expanding. The lncRNA study showed that Stereo-seq captures non-coding transcripts that targeted panels miss. The lamprey study showed that the platform works across species without probe redesign. Together, these expand the addressable research space beyond well-annotated coding genes in standard model organisms.

The common thread is that in each case, the biological conclusion depended on knowing where cells and transcripts are located in tissue — not just which cells and transcripts are present. That is what Stereo-seq is uniquely positioned to deliver.

From Published Discovery to Your Research Project

The studies reviewed here demonstrate what Stereo-seq can achieve when applied to well-designed biological questions with appropriate sample preparation, sequencing depth, and bioinformatics support. Translating these published successes into a practical research project requires careful attention to several variables: sample type (fresh-frozen vs. FFPE), tissue quality (RIN ≥ 7 for frozen, DV200 ≥ 30% for FFPE), chip format selection (from 0.5 × 0.5 cm to 13.2 × 13.2 cm), sequencing depth, and the choice of complementary modalities — scRNA-seq, snRNA-seq, immunofluorescence, or long-read sequencing — that provide the single-cell reference maps needed for spatial data interpretation.

CD Genomics provides end-to-end Stereo-seq services that cover this entire workflow: sample feasibility assessment, tissue sectioning, library preparation, sequencing on MGI platforms, SAW pipeline data processing, and customized bioinformatics for cell type annotation, spatial domain identification, and multi-omics integration. The service supports fresh-frozen and FFPE specimens across species, with study design consultation informed by the same experimental logic that produced the studies featured in this article. For researchers ready to explore whether Stereo-seq fits their tissue and research question, CD Genomics Stereo-seq Spatial Transcriptomics service provides a starting point for feasibility discussion and project planning.

Figure 5. CD Genomics spatial transcriptomics service overview — from sample feasibility review and Stereo-seq library preparation through sequencing, SAW data processing, and customized bioinformatics deliverables.

FAQ

Q: What distinguishes Stereo-seq from other spatial transcriptomics platforms for the types of studies covered here?
Stereo-seq combines subcellular resolution (500 nm spot spacing) with centimeter-scale capture areas and non-targeted poly(A) transcriptome capture. This means it can map whole organs at near-single-cell resolution, detect coding and non-coding RNA, and work across any species without probe redesign. These features were essential for the five studies discussed: 3D embryo reconstruction, pan-cancer lncRNA discovery, cross-species brain comparison, and FFPE clinical specimen analysis each required a combination that targeted or smaller-field-of-view platforms cannot provide simultaneously. For a systematic comparison, see the Stereo-seq platform selection guide.

Q: Are these studies reproducible, or do they represent exceptional cases from highly optimized labs?
The five studies come from independent research groups across China, Australia, and Europe, covering multiple tissue types, species, and disease contexts. The common element is not a single optimized protocol but the platform's inherent capability to capture spatial transcriptomic data at scale. The publication venues — Science, Nature Cell Biology, Nature Methods, Nature Communications, Advanced Science — represent standard peer review at the highest level. These are not isolated success stories; they reflect growing adoption across the spatial biology community.

Q: Can Stereo-seq be used with FFPE samples?
Yes. The lupus and aortic dissection studies both used FFPE tissue sections with Stereo-seq V2/OMNI chemistry, which supports total RNA profiling from formalin-fixed paraffin-embedded samples. FFPE compatibility is significant because it opens access to archived clinical specimen collections with associated longitudinal outcome data. Tissue quality requirements (DV200 ≥ 30%) still apply, and not every FFPE block will yield high-quality data, so feasibility review with a service provider is recommended before committing archived specimens.

Q: How much tissue and how many sections do typical Stereo-seq studies use?
It varies by study design. The embryo study used 82 serial sections for 3D reconstruction of a single specimen. The aortic dissection study processed 110 samples from 80 individuals across disease states. The lamprey brain study used full-organ sections from adult animals. The lupus study used standard FFPE dermatology sections. Stereo-seq chip formats range from 0.5 × 0.5 cm to 13.2 × 13.2 cm, so the platform accommodates both small precious specimens and large organ sections. For study-specific guidance, see the Stereo-seq study design resource.

Q: What bioinformatics infrastructure is needed to analyze Stereo-seq data?
Stereo-seq data is processed through the SAW (Stereo-seq Analysis Workflow) pipeline, which handles mapping, spatial barcode decoding, expression matrix generation, image registration, clustering, and reporting. Downstream analysis — differential expression, cell type annotation, spatial domain identification, cell-cell communication, and multi-omics integration — follows standard spatial transcriptomics workflows but benefits from the high spatial resolution and large field of view of Stereo-seq data. For a detailed walkthrough, see the Stereo-seq data analysis guide.

References

1. Cui L, Lin S, Yang X, et al. Spatial transcriptomic characterization of a Carnegie stage 7 human embryo. Nature Cell Biology. 2025;27:360–369.
2. Prakrithi P, Vo T, Xiong Z, et al. Unraveling lncRNA diversity at a single cell resolution and in a spatial context across different cancer types. Nature Methods. 2026;23:1236–1249.
3. Tao J, Yang H, Yong J, et al. Integrated single-cell and spatial analysis reveals a metabolic-immune axis driving aortic dissection. Advanced Science. 2026;e75509.
4. Wu H, Chen D, Li J, et al. Lamprey 3D single-cell transcriptomics reveals ancestral and specialized features of the vertebrate brain. Science. 2026;392(6804):eaaea2535.
5. Zhou W, Huang Y, Lei Y, et al. Spatial characterization of skin lesions in discoid and systemic lupus erythematosus. Nature Communications. 2026.

Research Use Only Statement

The information provided in this article is for research use only and is not intended for use in diagnostic or therapeutic procedures. CD Genomics provides sequencing and bioinformatics services for research purposes. Researchers should consult the appropriate regulatory guidelines for their specific applications.