Spatial transcriptomics maps compartmental remodeling in Hashimoto thyroiditis
Spatial transcriptomics has opened a new way to study Hashimoto thyroiditis without tearing the tissue apart first. In this pilot analysis, thyroid sections from two patients with histologically confirmed Hashimoto thyroiditis and two normal controls were profiled with 10x Genomics Visium FFPE technology, allowing epithelial, stromal, and immune compartments to be examined in place. The study was intentionally descriptive and hypothesis-generating, and its own technical framing matters: the control sections and the Hashimoto sections were processed on separate Visium slides, so any direct comparison across disease groups remains confounded by slide-level batch effects.
Even with that constraint, the atlas is useful because it shows how much spatial information is lost in bulk profiling. Follicular epithelium, stromal support structures, and immune aggregates were not treated as abstract cell lists; they were observed as organized tissue domains with distinct transcriptional programs. In autoimmune thyroid disease, that spatial context is the biology. Thyroid follicles are not only targets of immune injury; they also sit inside a remodelled microenvironment where fibroblast-like stromal cells, antigen-presenting cells, T cells, and B cells can occupy different neighborhoods and behave differently depending on where they sit in the gland.
Follicular epithelial domains showed altered marker distribution
Unsupervised clustering produced 16 transcriptionally distinct domains, later annotated as follicular epithelial-associated, stromal-associated, and immune-enriched regions. Within the epithelial compartment, the Hashimoto sections showed altered distribution patterns with reduced PAX8 and GPX3 signals and increased representation of Cluster 5, a domain defined by thyroid hormone synthesis pathways. In practical terms, that pattern is compatible with a gland in which the architecture of hormone-producing follicles is being reorganized rather than simply erased.
The epithelial finding is biologically plausible. PAX8 is a key thyroid lineage marker, and GPX3 is tied to oxidative handling in the gland. A fall in those signals suggests a change in differentiated follicular identity, while the redistribution of a hormone-synthesis-associated cluster points to uneven preservation of thyroid functional programs across the tissue. In autoimmune thyroiditis, that kind of patchy remodeling fits the well-known transition from intact follicles to disrupted glands infiltrated by lymphocytes and gradually losing functional reserve.
The data do not establish causality. They do, however, provide a spatial readout of what chronic autoimmune injury looks like at the tissue level: not a uniform collapse, but a mosaic of altered epithelial states.
Stromal programs shifted toward matrix and signaling pathways
The stromal-associated regions carried a different signature. SMOC2, CRABP1, and MYL9 were among the markers used to annotate those domains, and pathway enrichment pointed toward ECM–receptor interaction and PI3K–Akt signaling. This is the part of the atlas that speaks to tissue remodeling rather than immune attack alone.
Fibrotic and matrix-associated programs often accompany chronic inflammation because stromal cells respond to repeated injury by changing adhesion, contractility, and extracellular matrix behavior. In the thyroid, such remodeling can alter follicular spacing, vascular relations, and local cell communication. The PI3K–Akt axis is especially relevant because it links survival, growth, and inflammatory signaling; once activated in a remodeling niche, it can help sustain a tissue state that resists return to normal architecture.
One sentence captures the practical meaning of these findings: the stromal compartment was not passive background tissue. It carried its own remodeling program.
Immune-enriched niches carried distinct T-cell, B-cell, and myeloid signatures
The immune-enriched domains were divided into myeloid-, T-cell-, and B-cell-enriched clusters, each with its own pathway signature. Myeloid-enriched regions showed TNF signaling. T-cell-enriched domains showed TCR, Th17, PD-1/PD-L1, and related immune checkpoint pathways. B-cell-enriched regions were linked to cell cycle and DNA replication programs.
This compartmentalization matters because autoimmune thyroiditis is not driven by one immune mechanism. Myeloid cells can shape local cytokine tone and antigen presentation. T cells can sustain chronic tissue injury through persistent activation and helper polarization. B cells, especially when they are proliferating, suggest an organized local humoral response rather than a bystander presence. The presence of checkpoint-related pathways alongside T-cell signals also suggests that the immune compartment is not simply “on” or “off”; it is likely being regulated and counterregulated inside the gland itself.
Again, these are spatial observations rather than proven mechanisms. The study did not perform orthogonal validation, and it does not support causal claims about which cell type initiates the disease. Even so, the atlas makes a strong case that immune activity in Hashimoto thyroiditis is topographically organized, with different immune programs occupying different tissue niches.
A pilot atlas with a hard boundary on inference
The cohort was deliberately small: two Hashimoto thyroiditis cases and two controls, all female, all profiled from one section per specimen. The Hashimoto patients had elevated TSH values of 12.3 and 8.9 mIU/L and very high TPO antibody levels above 1300 IU/mL in both cases, with TgAb also elevated. The controls had adenomatous hyperplasia and no histologic thyroiditis in the profiled sections. None of the participants had received glucocorticoids or immunosuppressants before surgery.
The technical platform was a strength and a limitation at once. Visium FFPE preserves architecture and is well suited to map diseased tissue, but this was a single-platform, four-section pilot with no orthogonal confirmation. The slide arrangement is especially important: both normal controls were placed on one slide and both Hashimoto samples on another, making NC–HT differences vulnerable to batch effects. The investigators stated this directly, and they were right to do so. A spatial atlas that cannot separate biology from processing differences must remain descriptive.
What the study contributes is a tissue map, not a therapeutic conclusion. It shows that Hashimoto thyroiditis can be read as coordinated remodeling across epithelial, stromal, and immune compartments, with each compartment expressing its own molecular logic. The next step would require independent, batch-controlled cohorts and orthogonal validation before any claim about mechanism, progression, or translational use can be defended.