Single-Cell Map Tracks Rogue T Cell Clones Through Space and Time in Graft-Versus-Host Disease

Allogeneic hematopoietic cell transplantation remains one of the most powerful curative options for patients with leukemia and other hematologic malignancies, yet its promise is shadowed by a devastating complication: acute graft-versus-host disease, in which donor immune cells turn against the recipient’s own tissues. A new study published in Nature Immunology has now delivered the most detailed picture to date of how this immunological betrayal unfolds in human patients, following individual donor T cell clones across time, tissues and disease states with unprecedented resolution.

The research, led by Lingting Shi, Ajna Uzuni and Ximi K. Wang under the joint supervision of Elham Azizi and Ran Reshef at Columbia University, followed 31 transplant recipients using an integrated spatiotemporal framework. The team combined longitudinal T cell antigen receptor profiling with single-cell RNA sequencing paired with TCR sequencing and spatial transcriptomics, allowing them to track the fate of specific alloreactive T cell clones from the bloodstream into the gut epithelium, where the most lethal manifestations of graft-versus-host disease occur.

A central methodological innovation was DecompTCR, a computational tool developed by the team to resolve the temporal dynamics of T cell clonal expansion from bulk TCR repertoire data. The researchers also adapted existing computational approaches to map clone phenotypes and their tissue niches, including StarfyshHD for spatial deconvolution of high-resolution spatial transcriptomic data and DecipherTCR for joint representation of clonal states. Together, these tools allowed the investigators to ask not merely which T cell clones were present, but when they expanded, what programs they expressed and where they accumulated within damaged tissue.

One of the study’s most clinically consequential findings concerns post-transplantation cyclophosphamide, a widely used prophylactic strategy in which high-dose chemotherapy is administered shortly after infusion of the donor graft. The time-resolved modeling revealed that cyclophosphamide does indeed selectively deplete alloreactive T cell clones, validating the biological rationale behind the intervention. However, the data also exposed a critical vulnerability: when early expansion of alloreactive clones is insufficient, the depletion is incomplete, and residual surviving clones can seed severe disease. This finding reframes the drug’s success or failure as a matter of clonal population dynamics rather than simple toxicity.

Patients who went on to develop severe graft-versus-host disease were distinguished by persistent expansion of alloreactive clones in the blood, accompanied by increased clonal diversity among the expanding populations. The team’s analyses showed that severe disease was also marked by a rewiring of homeostatic cell types, suggesting that the alloimmune attack does not simply destroy tissue but actively remodels the cellular ecosystem of the affected organ. These early-repertoire dynamics, the authors propose, could serve as biomarkers to identify high-risk patients before clinical symptoms escalate.

Perhaps the most striking discovery emerged from tracking donor-derived CD8-positive clonotypes as they migrated into epithelial tissue. During this migration, the clones underwent a remarkable phenotypic transformation, diversifying and acquiring Hobit-positive tissue-resident memory T cell programs, marked by expression of the transcription factor ZNF683. This plasticity means that the same alloreactive clone can exist in fundamentally different functional states depending on its location, shifting from a circulating effector into a tissue-embedded resident cell that is far harder to eliminate with systemic therapies.

Spatial transcriptomic analysis then revealed where these transformed cells wreak their damage. Using spatial deconvolution, the researchers identified hubs of CD8-positive effector and Hobit-positive tissue-resident memory T cells clustered near intestinal stem-cell-rich crypt bases and in regions of crypt loss. This anatomical positioning is devastating in its implications: intestinal stem cells are essential for regenerating the epithelial barrier, and their destruction by neighboring cytotoxic T cells undermines the gut’s capacity for repair. The study thereby links tissue-instructed tissue-resident memory remodeling directly to localized epithelial injury, providing a mechanistic bridge between clonal immune dynamics and the clinical pathology of gastrointestinal graft-versus-host disease.

The framework also documented collaborative immune responses within crypt-loss regions, where multiple cell types appeared to act in concert to drive tissue destruction. By resolving these interactions at clonotype level, the study moves the field beyond the traditional view of graft-versus-host disease as a diffuse inflammatory process and toward a model in which discrete, spatially organized immune hubs determine where and how severely tissue damage occurs. The burden of these spatial hubs, the authors suggest, may itself serve as a measurable biomarker of disease severity.

The translational implications are substantial. If early clonal expansion dynamics in the blood can predict which patients will progress to severe disease, clinicians could intensify prophylaxis or intervene earlier in those at highest risk, while sparing others unnecessary immunosuppression. The identification of Hobit-positive tissue-resident memory programs as drivers of epithelial injury also nominates new therapeutic targets, since strategies that prevent T cells from adopting residency programs or that dislodge established resident populations could complement existing approaches such as costimulation blockade with abatacept and other prophylactic regimens under clinical evaluation.

All data generated in the study have been deposited in the Gene Expression Omnibus under accession number GSE307215, and the analysis code is publicly available, enabling the broader research community to build on this clonotype-resolved framework. As single-cell and spatial technologies continue to mature, this work offers a template for dissecting other immune-mediated conditions, from inflammatory bowel disease to solid organ transplant rejection, where the same principles of clonal tracking, phenotypic plasticity and spatial mapping are likely to illuminate how destructive immune responses take root in human tissue.

Subject of Research: Spatiotemporal single-cell profiling of alloreactive T cell clonal dynamics and phenotypic plasticity in human graft-versus-host disease

Article Title: Spatiotemporal single-cell profiling reveals T cell clonal dynamics and phenotypic plasticity in human graft-versus-host disease

Article References: Shi, L., Uzuni, A., Wang, X. K., Pressler, M., Harle, D. W., Chakrabarti, S., Macedo, R., Belay, K., Gordillo, C. A., McMahon-Skates, T., Raps, E., Zhang, J. Y. A., Nazaret, A., Fan, J. L., Jin, Y., Shen, X., Fuller, J. S., Azad, T., Huang, J., … Reshef, R. (2026). Spatiotemporal single-cell profiling reveals T cell clonal dynamics and phenotypic plasticity in human graft-versus-host disease. Nature Immunology. https://doi.org/10.1038/s41590-026-02631-2

Image Credits: AI Generated

DOI: 10.1038/s41590-026-02631-2

Keywords: graft-versus-host disease, hematopoietic cell transplantation, single-cell RNA sequencing, T cell receptor profiling, tissue-resident memory T cells, spatial transcriptomics, alloreactive T cells, cyclophosphamide, intestinal stem cells, clonal dynamics, transplant immunology, Hobit ZNF683