The study, led by Marxa L. Figueiredo of Purdue University, integrated publicly available single-cell RNA sequencing datasets spanning multiple preclinical mouse models of inflammatory arthritis, together with 12 healthy controls and 15 rheumatoid arthritis samples. The models were chosen deliberately to capture complementary facets of the disease: collagen-induced arthritis replicates the chronic, autoimmune, T-cell and B-cell-dependent phases of systemic joint destruction; serum transfer-induced arthritis bypasses adaptive priming to mimic the acute, antibody- and innate immune-mediated effector phase; antigen-induced arthritis models localized, antigen-specific, T-cell-driven joint inflammation; and the glucose-6-phosphate isomerase model represents metabolic-immune crosstalk with rapid systemic pathology. By converging on the programs shared across these diverse etiologies, the team sought a conserved landscape of synovial remodeling rather than a model-specific snapshot.

The bioinformatic pipeline was substantial. Count matrices were processed through Trailmaker and ICARUS v3, with background noise removed using transcripts-per-cell thresholds, dead cells filtered by mitochondrial content, doublets excluded with scDblFinder, and batches corrected using Harmony. Unsupervised Leiden clustering with UMAP visualization resolved the integrated mouse synovium into 25 distinct clusters, which the researchers organized into functional niches: a structural and fibroproliferative compartment, a vascular and endothelial compartment, a myeloid immune compartment, a lymphoid and mast compartment, and even a peripheral glial niche containing Schwann cells. Within the structural niche alone, profiling distinguished universal Pi16-positive sublining progenitors, specialized Calb1-positive stromal cells, mechanosensitive Dkk3-positive fibroblasts, fibrogenic Postn-positive fibroblasts, activated fibroblast-like synoviocytes, myofibroblasts, inflammatory progenitors, and joint-destructive Prg4/Il6-positive lining fibroblasts, alongside osteoblast and chondrocyte lineages.

Differential expression analysis comparing inflamed and healthy joints revealed a pronounced transcriptional imbalance. Upregulated genes in arthritic synovium were strongly enriched for innate immune cascades, including neutrophil degranulation and platelet activation and aggregation, as well as extracellular matrix organization, supramolecular fiber organization, focal adhesion, regulation of proteolysis, IL-17A signaling, and TNF production. Protein-protein interaction network analysis identified interconnected modules centered on collagen formation, ECM organization, and actin filament regulation. Conversely, downregulated genes represented a loss of homeostatic programs: tissue development, blood vessel development, glial cell fate specification, biomineral tissue development, and fatty acid metabolism. The picture that emerges is a joint shifting away from protective, regenerative maintenance toward aggressive immune activation and pathological matrix remodeling, a dual failure that helps explain why suppressing inflammation alone often cannot stop structural damage.

The study’s most striking mechanistic finding came from cell-cell communication modeling with CellChat, which inferred ligand-receptor interactions across the diseased synovium. Vascular mural cells, sublining progenitors, fibrogenic sublining fibroblasts, and lining fibroblasts emerged as strongly upregulated signaling hubs. Within the Notch family, the highest-strength disease-state links were autocrine Jag1-Notch3 signaling within vascular mural cells and endothelial-to-mural Dll4-Notch3 communication from activated endothelial cells. Beyond this structural feedback loop, mural Jag1-Notch2 and endothelial Dll4-Notch2 signaling maintained strong communication with infiltrating inflammatory monocytes and sublining macrophages, with an additional Dll4-Notch1 interaction targeting monocytes. This positions the vascular-mural unit as an actionable upstream trigger that simultaneously drives structural remodeling and macrophage activation in the inflamed joint.

Downstream of this perivascular trigger, regulatory network inference with SCENIC revealed a synchronized, stress-induced transcriptional program. Vascular mural cells, chondrocytes, inflammatory progenitors, specialized sublining subsets, and macrophage populations all displayed coordinated activity of the Atf3 and Fos regulons, transcription factors of the AP-1 family that govern inflammatory stress responses, clustering with the immediate-early gene Egr1 and Krüppel-like factors Klf4 and Klf10. The researchers suggest this shared module may function as a unified regulatory hub coordinating myofibroblast differentiation and the aggressive fibrotic response characteristic of chronic joint disease. Replicate-aware pseudobulk DESeq2 modeling confirmed that these cluster-specific patterns mirror the global trends, supporting the findings under a statistically rigorous framework with a Benjamini-Hochberg adjusted threshold.

Crucially, the mouse findings translated to human disease. The team analyzed a high-resolution human rheumatoid arthritis synovial dataset generated with 10x Genomics Visium HD spatial transcriptomics, processed through the VST-DAVis pipeline. A single sample yielded 474,081 tissue-associated 8-micrometer bins with a mean of 197.2 UMIs per bin, allowing unsupervised clustering of spatially resolved tissue domains. Human vascular mural lineages separated into NOTCH3, ACTA2, and TAGLN-enriched mural cells and MYH11-expressing perivascular fibroblasts, mirroring the mouse compartments. Weighted gene co-expression network analysis returned a single dominant module, SM01, a stress and matrix-remodeling hub containing VIM, FN1, MMP2, CTSB, PLA2G2A, CD74, COL6A2, SELENOP, and COL3A1. This module was selectively and highly enriched in human activated FLS and lining FLS, the functional homologs of the mouse sublining fibroblast and mural hubs, suggesting the Atf3/Fos stress-response pathway and its downstream matrix-remodeling program are conserved drivers of the inflammatory stromal phenotype in human RA.

The spatial data then did what conventional single-cell sequencing cannot: it placed these circuits in physical context. Activated FLS emerged as a central spatial coordinator, acting as the primary source of outgoing Collagen, FN1 (fibronectin), Tenascin, and THBS (thrombospondin) ligands while remaining highly receptive to incoming NOTCH, THBS, and FN1 signaling. Spatial chord diagrams confirmed perivascular fibroblasts presenting JAG1 to NOTCH3 on vascular mural cells, alongside a multicellular structural network mediated by THBS2, THBS4, and COMP linking lining and activated FLS to SDC1 and CD47 receptors on plasma cells and activated FLS. Most tellingly, spatial overlay maps showed that FN1 and THBS signaling networks co-localize at high density within a distinct inflammatory niche restricted to the synovial tissue’s edge, heavily populated by activated FLS, lining FLS, and perivascular mural cells. The disease, in other words, has an address.

The authors are careful to note the limitations. The spatial transcriptomics dataset, while high-resolution and derived from untreated RA tissue, comes from a single human donor and serves as a proof-of-concept bridge that will benefit from broader validation. CellChat infers potential microenvironmental signaling rather than confirming direct physical cell-cell contact in vivo, so functional assays will be needed to validate the predicted pathways. Even so, the model that emerges is coherent and therapeutically provocative: vascular-derived Jag1 and Dll4 activate Notch3 and Notch2 on adjacent stromal and immune cells, initiating an Atf3/Fos transcriptional cascade that drives both myeloid activation and a matrix-remodeling program within a localized, self-amplifying spatial niche at the tissue’s edge.

That model points toward a stroma- and vascular-directed therapeutic paradigm that differs fundamentally from conventional immunomodulatory biologics. Rather than broadly suppressing systemic immunity, selective blockade of Notch3/2 activation or endothelial ligand presentation could specifically uncouple vascular-to-stromal and vascular-to-immune crosstalk, preventing both pathogenic fibroblast activation and immune infiltrate recruitment. Disrupting this structural trigger could also destabilize the self-amplifying niche at the tissue’s edge. The authors propose that combining perivascular Notch pathway inhibition with downstream targeted therapies against matrix-remodeling networks such as THBS or FN1, or against AP-1 transcription factors like Atf3 and Fos, may represent a synergistic strategy to simultaneously halt chronic inflammation and structural joint destruction. For the many patients whose disease resists today’s immune-targeted drugs, that multi-tier approach, aimed at the joint’s own cellular architecture rather than only its inflammatory guests, offers a genuinely new map of where to intervene.

Subject of Research: Conserved stromal remodeling and Notch/AP-1 signaling pathways in inflammatory arthritis identified through cross-species single-cell and spatial transcriptomics

Article Title: Cross-species single-cell and spatial transcriptomics map the conserved stromal remodeling landscape of inflammatory arthritis

Article References: Figueiredo, M. L. (2026). Cross-species single-cell and spatial transcriptomics map the conserved stromal remodeling landscape of inflammatory arthritis. Frontiers in Molecular Medicine, 6, 1951540. https://doi.org/10.3389/fmmed.2026.1951540

Image Credits: AI Generated

DOI: 10.3389/fmmed.2026.1951540

Keywords: rheumatoid arthritis, single-cell RNA sequencing, spatial transcriptomics, synovial fibroblasts, Notch signaling, AP-1 transcription factors, vascular mural cells, extracellular matrix remodeling, fibroblast-like synoviocytes, CellChat, Visium HD, mouse models

The study, led by Vincent Hammer and Stephan Ossowski, tackles a problem that has grown more urgent as spatial transcriptomics has matured. Modern platforms such as Stereo-seq and Visium HD can, in principle, sequence the entire transcriptome at nearly subcellular resolution, revealing which genes are active at which positions within a tissue slice. But most existing analysis tools focus on identifying tissue domains or layers, treating the boundary itself as an afterthought. REBOSEP, by contrast, is boundary-centric: it takes any existing tissue annotation, measures the physical distance of every data point from an annotated border, and represents each gene’s expression as a continuous signal across that border.

The technical challenge is considerable. Although Stereo-seq spots are minuscule, the number of RNA molecules captured in any single spot is too low for meaningful analysis, so spots are merged into bins, typically of 30 by 30 spots, roughly the size of one to four cells. These bins are formed without regard to where cells actually sit, and bins straddling a tissue edge inevitably contain transcripts from both sides. The result is noisy, sparse data in which clustering algorithms assign random labels to isolated bins and even to empty cavities where RNA has diffused during sample preparation. REBOSEP confronts this by letting users filter out small spurious regions, manually correct annotations, and interactively select the exact stretch of boundary they want to study, then reconstructing long continuous borders even across short gaps.

Once a boundary is defined, the workflow computes the average expression of every gene at each distance interval from the border, producing a table that reads like a cross-section of transcriptional activity. Genes whose profiles are too sparse or inconsistent across replicate samples are filtered out using dynamic time warping, a technique borrowed from time-series analysis that aligns curves by locally stretching or compressing them. A restrictive warping window allows the method to tolerate minor positional shifts of about one bin, roughly 30 micrometers, without conflating genuinely different patterns. Hierarchical clustering then groups genes with similar cross-boundary profiles, and gene ontology enrichment analysis reveals the biological processes each group represents.

To demonstrate the approach, the researchers applied REBOSEP to publicly available Stereo-seq data from E16.5 mouse embryos, analyzing five brain boundaries: dorsal pallium-meninges, midbrain-meninges, lower and upper hindbrain-meninges, and subpallium-meninges. Boundary lengths ranged from about 1,740 to 3,667 micrometers, and expression was traced up to 300 micrometers into the brain and 150 micrometers into the surrounding meninges. The resulting atlas captures the expression pattern of every gene the platform detected, including long non-coding RNAs and poorly characterized Riken genes, many of which now carry a boundary-specific expression signature that offers the first clue to their function.

The deepest analysis focused on the dorsal pallium, the embryonic precursor of the neocortex, and its long border with the meninges, the protective membranes that envelop the brain and play active roles in its development. Clustering revealed striking functional homogeneity. Genes expressed constitutively across all pallium layers were overwhelmingly tied to nervous system and neuron development. Genes active throughout the meninges were enriched for extracellular matrix organization, while genes peaking in the meningeal layers closest to the border showed strong enrichment for angiogenesis, the formation of blood vessels, a process closely linked to the developing blood-brain barrier. On the brain side, a distinct cluster of genes sharply upregulated only very near the border was enriched for synaptic signaling and transmission, and another, with a more gradual falloff into deeper pallium layers, was associated with neuron projection and neuron development.

The researchers then examined how specific signaling pathways are spatially regulated at this border. Wnt signaling, a canonical developmental pathway, showed clear boundary-dependent modulation: the receptors Fzd1 and Fzd2 were inhibited at the border, the Wnt-promoting factor Rspo3 peaked around 75 micrometers, and the cell-cycle regulator Ccnd2, a downstream target, was confined to a narrow band within about 20 micrometers of the interface. Notch signaling, essential for cell fate decisions and neurogenesis, was broadly suppressed across the outer brain layers, with Notch receptors highly expressed in the meninges but reduced toward the border. Cell adhesion molecules mediating neuron-neuron and growth cone-axon interactions were largely confined to a thin band of roughly 50 to 150 micrometers within the brain, with genes such as Cntnap2, L1cam, Ncam1, Nrxn1, Nrxn2, and Ptprd gradually decreasing toward deeper layers.

Robustness checks supported the reliability of these patterns. Repeating the clustering a hundred times with random seeds produced vastly higher similarity within clusters than between them, even when artificial noise was added. Comparing Leiden and Louvain clustering algorithms yielded a Spearman correlation of 0.939 between the resulting pattern distances, and varying the bin size had little effect except at the coarsest resolution of 100 bins, well beyond what is typically used. Crucially, the team validated their findings in an independent Visium HD dataset from E15 to E15.5 embryos, a different technology, different specimens, and a slightly earlier developmental stage. The Visium HD profiles closely reproduced the Stereo-seq patterns, preserving the positions of expression peaks, their relative amplitudes, and their rising or decaying trends.

Perhaps the most provocative finding concerns disease. Many of the genes with strong boundary-specific expression, including Apc, Ccnd2, Tle4, Ctnnd2, L1cam, Nrcam, Cadm1, Cntn1, Nlgn1, and Nrxn2, are known players in cancer biology, autism spectrum disorder, or broader neurodevelopmental syndromes. Tumor malignancy is, in essence, the breakdown of a tissue boundary, and boundary roughness is a recognized marker of invasiveness and poor prognosis. The fact that genes governing adhesion and Wnt signaling are so precisely localized at healthy embryonic borders suggests that tumors may co-opt the same molecular machinery to remodel and invade across tissue frontiers, making these boundary profiles a rich source of candidate mechanisms and therapeutic targets.

The authors are candid about limitations. Gene-level interpretations remain exploratory and hypothesis-generating, the quality of boundary definitions depends on upstream tissue segmentation, current capture rates prevent true single-cell resolution, and all analyzed datasets came from female mice, so sex-specific effects could not be assessed. Still, the resource and the accompanying code, both publicly available on GitHub, lower the barrier for any laboratory to interrogate boundaries in their own spatial transcriptomics data. As spatial omics continues to spread through developmental biology, oncology, and neuroscience, tools like REBOSEP signal a shift in perspective: the most interesting biology may lie not within tissues, but precisely at the lines where they meet.

Subject of Research: Spatial transcriptomics analysis of gene expression patterns across tissue boundaries in the developing mouse brain

Article Title: Spatial transcriptomics profiling across tissue boundaries in the developing mouse brain

Article References: Hammer, V., & Ossowski, S. (2026). Spatial transcriptomics profiling across tissue boundaries in the developing mouse brain. iScience, 29(11), Article 117765. https://doi.org/10.1016/j.isci.2026.117765

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117765

Keywords: spatial transcriptomics, tissue boundaries, REBOSEP, developing mouse brain, Stereo-seq, Visium HD, meninges, dorsal pallium, Wnt signaling, Notch signaling, cell adhesion, neurodevelopment

The problem CellART attacks is fundamental to how modern spatial platforms work. Instruments such as 10x Genomics’ Xenium and Visium HD, Vizgen’s MERFISH and BGI’s Stereo-seq can now resolve gene expression at subcellular scale, but they pay a price for that resolution. Some capture only sparse transcript counts per spot, while others measure a limited panel of genes. In imaging-based assays, transcripts are detected as individual dots scattered across a tissue image, and deciding which dots belong to which cell is far from trivial. On array-based platforms like Visium HD, the data arrive in tiny bins that frequently straddle cell boundaries, so a single bin may contain RNA from two or three neighboring cells. Without a reliable way to draw cell boundaries and assign transcripts, the single-cell picture that biologists crave remains frustratingly blurred.

CellART, described in a Brief Communication published in Nature Computational Science on 5 October 2026, tackles both challenges at once. The framework performs cell segmentation and cell-type annotation simultaneously, fusing three complementary data streams: the staining images that accompany spatial assays, the spatial transcriptomics data themselves, and single-cell RNA sequencing references that catalog the expression signatures of known cell types. The key insight is that these modalities carry mutually reinforcing information. Histology images reveal where nuclei and membranes lie; transcript distributions show where gene products accumulate; and reference atlases constrain what combinations of cell types are biologically plausible in a given tissue. By integrating deep learning with probabilistic modeling, CellART lets each data source correct the weaknesses of the others.

Technically, the segmentation module builds on ideas from computer vision, drawing on architectures such as feature pyramid networks and residual networks that have proven their worth in object detection and image recognition. These networks learn to delineate cell boundaries from the staining images and the spatial patterns of transcript density. On top of that segmentation layer sits a probabilistic model in the spirit of deep generative approaches to single-cell transcriptomics, which assigns each segmented cell a cell-type label by comparing its expression profile with the scRNA-seq reference. This joint design means that segmentation errors and annotation errors can be caught and corrected in a single pass rather than compounding across separate pipeline stages, a common failure mode when researchers chain together independent tools.

The breadth of validation is one of the study’s most striking features. The team applied CellART to datasets spanning at least four distinct spatial platforms: Xenium, Visium HD, MERFISH and Stereo-seq, across tissues including human lung cancer, human breast cancer, human colorectal cancer, human skin, and mouse brain. On the Xenium 2.0 human lung dataset, where 10x Genomics provides a ground-truth segmentation derived from mixed-fluorescence imaging, CellART’s boundaries were quantified against established methods including Baysor, Cellpose and ProSeg using F1-scores and Jaccard indices. On Visium HD mouse brain data, where the recommended 16-micrometer bins do not align with cells in the histology image, CellART and Bin2Cell were compared by transcript coverage rate and by the correlation between cytoplasmic and nuclear expression profiles of individual cells, covering nearly sixty thousand cells in one benchmark alone.

Accuracy in cell typing proved equally important. In a human lung section profiled with both Xenium and post-Xenium Visium HD, CellART recovered consistent cell-type labels across the two platforms, suggesting that its annotations reflect genuine biology rather than platform-specific artifacts. On mouse brain data from all four platforms, the framework reconstructed the layered architecture of the cortex and reproduced the hippocampal organization documented in the Allen Mouse Brain Atlas, with the extracted cells correlating strongly with the single-cell reference. The team also measured computational efficiency by subsampling tissues of increasing size, and CellART remained tractable where some competing approaches struggled; one alternative method, TopACT, did not complete at full tissue size, a point the authors extrapolated rather than measured.

The biological payoff is vividly illustrated in cancer tissue. In the Xenium breast cancer dataset, CellART resolved ten cell types with high consistency between replicates, and its sharper boundaries prevented a classic pitfall: transcripts from neighboring cells bleeding into incorrectly drawn boundaries. Where a standard pipeline combining 10x segmentation with scANVI misassigned cells as tumor because ERBB2 transcripts from adjacent malignant cells drifted inside their borders, CellART produced crisper ERBB2 spatial patterns with less transcript diffusion. In breast cancer, where ERBB2 defines the HER2-positive subtype and guides therapy, that distinction between a tumor cell and its neighbor is not a cosmetic detail; it can change how a tumor’s molecular profile is read.

Perhaps the most compelling demonstration came from Visium HD colorectal cancer data, where CellART’s single-cell resolution outperformed spot-level deconvolution by RCTD, which classified the majority of more than 429,000 spots as doublets containing mixed cell populations. Working at true single-cell scale, CellART separated macrophage subpopulations that the coarser analysis blurred together. It distinguished SELENOP-positive macrophages scattered among tumor cells from SPP1-positive macrophages clustered at specific regions, one of which was recovered only by CellART. Defining tumor-associated macrophages as macrophages lying within fifty micrometers of a tumor cell, the team found these cells concentrating along tumor boundaries and forming a transcriptionally distinct cluster between macrophages and tumor cells, with upregulated immune-related and tumor-promoting genes, particularly MMP12.

That last finding hints at the kind of discovery CellART makes routine. Ligand-receptor analysis revealed strong interactions between tumor-associated macrophages and tumor cells, including the MMP12-PLAUR, SPP1-ITGAV and ITGB5, and APOE-SORL1 pairs, with their spatial expression co-localized at tumor margins. The urokinase plasminogen activator system, to which PLAUR belongs, has long been implicated in cancer metastasis and prognosis, and integrins are established drivers of invasion. Seeing these interaction axes light up precisely where macrophages and malignant cells touch provides a spatially grounded view of the tumor microenvironment that neither dissociated single-cell sequencing nor coarse spatial deconvolution can deliver. For immunotherapy research, where the dialogue between immune cells and tumor cells is everything, tools of this kind could reshape how tumor sections are interpreted.

Accessibility may determine whether CellART achieves the adoption its designers hope for. The software is written in Python, released under the permissive MIT licence on GitHub, and archived on Zenodo, with simulation and evaluation scripts, documentation and worked examples included. Its outputs are compatible with widely used community tools, including the SpatialData framework for spatial omics, easing integration into existing analysis workflows. All the spatial and single-cell reference datasets analyzed in the study are publicly available from 10x Genomics, Vizgen, BGI, the Allen Institute, the CZI CELLxGENE portal and the Allen Brain Atlas, and the processed inputs and simulated benchmark data are deposited for reuse. In a field where proprietary pipelines and incompatible file formats often fragment progress, a unified, open, platform-agnostic tool that turns raw high-resolution spatial data into trustworthy single-cell atlases could become as indispensable as the sequencing machines that generate the data in the first place.

Subject of Research: A unified computational framework for single-cell segmentation and annotation in high-resolution spatial transcriptomics

Article Title: CellART: a unified framework for extracting single-cell information from high-resolution spatial transcriptomics

Article References: Chen, Y., Liu, Y., Wang, Z., Zeng, Y., Chao, Z., Jiang, P., Chen, H., Wang, J., Xiao, J., & Yang, C. (2026). CellART: a unified framework for extracting single-cell information from high-resolution spatial transcriptomics. Nature Computational Science. https://doi.org/10.1038/s43588-026-01054-1

Image Credits: AI Generated

DOI: 10.1038/s43588-026-01054-1

Keywords: spatial transcriptomics, CellART, single-cell analysis, cell segmentation, deep learning, probabilistic modeling, tumor microenvironment, macrophages, Xenium, Visium HD, MERFISH, computational biology

Alveolar echinococcosis is rare but devastating. The parasite’s larvae proliferate in the liver in a tumor-like fashion, provoking granulomatous inflammation and driving progressive scarring of the surrounding tissue. Clinicians have long observed that patients with advanced disease show suppressed immune activity and extensive fibrosis, but the cellular choreography that carries the liver from acute inflammation to chronic remodelling has remained poorly defined. The central problem, the authors note, is spatial: immune cells do not act in isolation but within structured niches, and conventional bulk methods dissolve that architecture into averages. The new study set out to preserve the geography.

The team used an experimental model in which mice received a secondary hepatic infection with E. multilocularis protoscoleces, the larval forms that initiate disease. Liver tissue was collected from uninfected animals and from mice at two timepoints: two weeks after infection, representing the early inflammatory phase, and three months after infection, representing the chronic phase. Four complementary techniques were layered on top of one another: flow cytometry to quantify immune populations, immunofluorescence microscopy to place specific markers in tissue, single-cell RNA sequencing to profile individual cells, and Visium HD spatial transcriptomics to map gene expression across intact liver sections at high resolution.

The flow cytometry results traced the arc of the immune response with striking clarity. Two weeks after infection, eosinophils, macrophages, and CD4-positive T cells surged in the infected livers, a signature of vigorous anti-parasitic mobilization. By three months, these populations had declined, yet eosinophils and macrophages remained elevated above levels seen in uninfected control animals, hinting that the immune system never fully stands down even as the disease settles into its chronic state. This partial, persistent activation is a key piece of the puzzle of how alveolar echinococcosis smolders for years, sometimes decades, in human patients.

The single-cell and spatial data then revealed what those flowing populations were actually doing. At the two-week mark, distinct stromal cell subsets had already infiltrated the granulomatous areas that form around parasite tissue. Among them were three transcriptionally distinct populations of endothelial cells, labeled ECs1, ECs2, and ECs3, alongside a marked accumulation of myofibroblasts, the contractile, matrix-producing cells best known for driving wound healing and fibrosis. By the chronic three-month phase, those myofibroblasts had significantly ramped up expression of Timp1 and Spp1, two well-established pro-fibrotic markers, indicating that the parasite-associated lesions had become active engines of scar formation.

The myeloid compartment proved equally heterogeneous. Monocyte-derived macrophages diversified into two recognizable subtypes, one characterized by Arg1 and Spp1 expression and another marked by Ccr7 and Itgax, profiles associated respectively with tissue remodeling and inflammatory activation. Meanwhile, Kupffer cells, the liver’s resident macrophages, showed a different trajectory: during chronic infection, their expression of inflammatory cytokines fell significantly. The juxtaposition of pro-inflammatory monocyte-derived cells with quieted resident macrophages suggests that the parasite-associated environment reshapes the myeloid landscape in ways that may blunt effective anti-parasitic immunity while promoting tissue damage and repair gone awry.

Perhaps the most intriguing finding concerned eosinophils, the granulocytes typically associated with anti-parasitic defense and allergic disease. In this model, eosinophils were specifically enriched in the cores of hepatic granulomas two weeks after infection. But their gene expression profile was not purely inflammatory. Instead, they displayed an immunoregulatory transcriptional program, and the spatial maps placed them in close proximity to cells expressing IL-33, an alarmin cytokine with well-documented roles in type 2 immunity and tissue repair. The colocalization raises the possibility that granuloma-core eosinophils are not simply attacking the parasite but actively sculpting a tolerance-promoting microenvironment, a hypothesis the authors suggest merits direct testing in future studies.

T cells told a complementary story about geographic segregation. CD8-positive T cells split into two functionally distinct subsets: an Ifng-expressing effector population, capable of the antiviral and antiparasitic cytotoxicity associated with interferon-gamma, and a Gzmk-positive subset bearing migratory and homeostatic characteristics. Notably, both populations were found predominantly outside the granulomatous areas, in the surrounding non-granulomatous liver. The cytotoxic arm of adaptive immunity, in other words, appears to be kept at the perimeter of the parasite-associated lesions, potentially unable to reach the parasite itself, which sits protected at the granuloma’s center.

By combining spatial deconvolution, a computational technique that estimates which cell types occupy each spot on a tissue map, with cell-cell communication analysis, the team identified the collagen-rich granuloma as a critical signaling niche. Their predictions pointed to myofibroblast-derived collagen signals, exemplified by the COL1A1–CD44 ligand-receptor pair, interfacing with lymphocytes at the granuloma boundary. This offers a concrete molecular hypothesis for how fibrotic stromal cells communicate with immune cells and potentially enforce the immunosuppressive character of chronic alveolar echinococcosis. Collagen-rich matrices are known to physically impede T cell migration in cancer, and the parallel here is difficult to ignore.

The authors frame the work as a spatially resolved cellular framework for understanding the immune-stromal regulatory network during E. multilocularis infection, and as a foundation for future studies of disease pathogenesis. The implications reach beyond parasitology. If myofibroblasts are confirmed as central architects of both fibrosis and immune suppression in alveolar echinococcosis, they become candidate targets for adjunctive therapies aimed at slowing disease progression, complementing the surgery and benzimidazole chemotherapy that form today’s standard of care. With the collaboration between advanced single-cell technology and high-resolution spatial mapping now demonstrated in this infection model, the field has, for the first time, a genuine atlas of the terrain on which this devastating parasite reshapes its host.

Subject of Research: Immune–stromal cell interactions in Echinococcus multilocularis metacestode infection mapped by single-cell and spatial transcriptomics

Article Title: Integrating single-cell RNA sequencing with Visium HD spatial atlas reveals immune–stromal cell interactions in Echinococcus multilocularis metacestode infection

Article References: Yan, M., Qi, W., Zhang, G., Wang, X., Wu, C., Tian, M., Geng, A., Wang, H., Zhang, C., Li, J., & Zhang, W. (2026). Integrating single-cell RNA sequencing with Visium HD spatial atlas reveals immune–stromal cell interactions in Echinococcus multilocularis metacestode infection. Parasites & Vectors. https://doi.org/10.1186/s13071-026-07591-y

Image Credits: AI Generated

DOI: 10.1186/s13071-026-07591-y

Keywords: alveolar echinococcosis, Echinococcus multilocularis, single-cell RNA sequencing, spatial transcriptomics, Visium HD, myofibroblast, granuloma, liver fibrosis, eosinophils, macrophages, immune evasion, parasitic liver disease

The framework, developed by Seowon Chang, Alexander Fleischmann and Ying Ma, addresses a problem that has grown acute as spatial assays have diversified. Transcriptome-plus-protein platforms such as SPOTS and spatial CITE-seq, epigenome-transcriptome methods that jointly capture chromatin accessibility or histone marks alongside RNA, metabolomic imaging and multiplexed approaches such as spatial-Mux-seq each carry modality-specific signals that generic integration tools tend to wash out. Earlier software could typically merge only two modalities, struggled with the massive spot counts of Visium HD, and returned a single consensus answer with no indication of where the model was guessing. SCIGMA’s designers set out to build a system that was scalable, generalizable across platforms and honest about its own uncertainty.

Architecturally, SCIGMA combines two ideas that have proven powerful in machine learning but had not been married in quite this way for spatial biology. The first is a multiview graph neural network built on graph attention layers, which represents each spatial location as a node connected to its tissue neighbors, letting the model learn from both molecular measurements and the physical arrangement of the tissue. Each modality is encoded through its own branch, so transcriptomic, proteomic, epigenomic and metabolic information is transformed into a shared latent space without being forced into a single flattened feature matrix. The second idea is an uncertainty-aware contrastive learning objective: by treating temperature as a learnable, uncertainty-linked quantity, the model learns to pull together representations of the same location seen through different molecular lenses while pushing apart mismatched views, and it modulates how strongly it aligns views depending on how reliable the data at that location appear to be.

This uncertainty machinery is more than a statistical nicety. After training, SCIGMA produces spatially resolved uncertainty maps that highlight regions of biological or technical heterogeneity — tumor margins, regions with mixed cell populations, or spots where assay quality degrades. The authors show that uncertainty estimates flag locations where feature reconstruction error is highest, meaning researchers can see precisely where the integrated representation is least trustworthy rather than accepting a smoothed-over consensus. Interpretability was a design goal throughout: the framework preserves modality-specific signals within its joint embedding, so users can trace which molecular layer drives a given spatial domain and recover regulatory programs that are visible only in, say, the chromatin channel and not the transcriptome.

The evaluation was unusually broad. The team benchmarked SCIGMA across 19 datasets spanning eight modalities, ten tissue types and nine technological platforms, ranging from spatial epigenome-transcriptome profiling of the postnatal mouse brain to protein-plus-RNA measurements of the mouse spleen, single-cell-resolution Xenium Prime datasets from human ovarian and cervical cancers, and enormous Visium HD sections of mouse intestine and human colorectal cancer with more than one million spots each. Where ground truth allowed comparison, SCIGMA outperformed existing methods on spatial domain detection, preservation of modality-specific information, feature reconstruction and reproducibility across repeated runs — a metric that matters given growing concern about the fragility of machine learning models in biomedical data science.

The biological case studies illustrate what multimodal integration buys that single-modality analysis cannot. In the mouse brain, combining chromatin accessibility or histone modification data with transcriptomics let SCIGMA refine cortical layers and white matter boundaries beyond what RNA alone revealed, and to link transcriptional identities with the regulatory elements that govern them. In the spleen, joint transcriptome-proteome analysis sharpened the demarcation of follicles, marginal zones and T cell zones, tying protein-level markers such as CD19 and CD3 components to their local transcriptional contexts. Analyses of Xenium Prime tumor sections demonstrated that the framework resolves intratumoral heterogeneity, distinguishing tumor epithelial niches, stromal compartments and immune infiltrates while its uncertainty layer highlights the unstable boundary regions where tumor biology is most contested.

Scalability was addressed head-on. Naive graph neural networks choke when every spot in a Visium HD section becomes a node in a graph with millions of vertices. SCIGMA incorporates efficient sampling strategies and nearest-neighbor retrieval, borrowing ideas from the Faiss library, to keep training tractable, and the authors demonstrate full-length analysis of Visium HD datasets exceeding one million spatial locations. Equally important is extensibility: the framework is modular, so when a future technology profiles six, seven or more molecular layers simultaneously, additional modality encoders can be attached without redesigning the core model. The team demonstrated this flexibility on spatial-Mux-seq mouse brain data, simultaneously integrating five modalities into a coherent joint representation.

Software accessibility rounded out the release. The SCIGMA package, along with scripts reproducing every published analysis and a suite of tutorials, is openly available through GitHub and archived on Zenodo, supporting the reproducibility standards the paper explicitly engages with. Funding came from the National Science Foundation and the National Institutes of Health, and the work was carried out at Brown’s Center for Computational Molecular Biology with computational support from the university’s Center for Computation and Visualization. The authors report no competing interests, and the article underwent peer review by researchers including Mengjie Chen, Zhaoheng Li and Wei Sun.

For the field, SCIGMA arrives at a moment when the volume and variety of spatial data are outpacing analysis tools. High-resolution platforms such as Visium HD and Xenium Prime are making million-spot, subcellular-resolution datasets routine, and multiplexed assays are stacking molecular layers that earlier software simply could not combine. A framework that handles five modalities, scales to the largest current datasets, runs across platforms without platform-specific tuning, and quantifies its own confidence offers a kind of common analytical ground that spatial biology has lacked. If it holds up in broad community use, the practical effect will be that researchers spend less time wrestling incompatible data formats into fragile pipelines and more time asking biological questions — about how tumors orchestrate their microenvironments, how cortical architecture is patterned, and how genomic regulation plays out across the geography of living tissue.

Subject of Research: Development of SCIGMA, a scalable uncertainty-aware deep learning framework for integrating spatial multiomics data across modalities and platforms

Article Title: Scalable, generalizable and uncertainty-aware integration of spatial multiomics across diverse modalities and platforms with SCIGMA

Article References: Chang, S., Fleischmann, A., & Ma, Y. (2026). Scalable, generalizable and uncertainty-aware integration of spatial multiomics across diverse modalities and platforms with SCIGMA. Nature Genetics, 58(9), 2284-2302. https://doi.org/10.1038/s41588-026-02706-8

Image Credits: AI Generated

DOI: 10.1038/s41588-026-02706-8

Keywords: spatial omics, multiomics integration, deep learning, graph neural networks, contrastive learning, uncertainty estimation, spatial transcriptomics, tumor heterogeneity, spatial domain detection, Visium HD, Xenium Prime, computational biology