Summary
A bioRxiv preprint uses spatial multi-omics and 3D imaging to map melanoma tumours treated with neoadjuvant immunotherapy. The study links organised immune structures, including tertiary lymphoid structures and germinal centres, with treatment response and disease-free survival.
A bioRxiv preprint reports a spatial and three-dimensional analysis of how melanoma tissue responds to neoadjuvant immune checkpoint blockade (ICB), a form of immunotherapy given before surgery. The researchers profiled more than 112 million cells across three melanoma ICB regimens, combining spatial transcriptomics, multiplexed imaging and single-cell RNA sequencing.
The work identifies organised immune and stromal structures associated with response, including mature tertiary lymphoid structures (TLS), germinal centres and TCF7-positive stem-like T cells. It also introduces computational tools intended to quantify these features in tissue sections and routine pathology images. The report was posted on September 18, 2026, as a bioRxiv preprint and has not yet undergone peer review.
Mapping the melanoma immune ecosystem
The study examined the tumour microenvironment at single-cell resolution while retaining information about where cells were located. MERFISH spatial transcriptomics measured gene-expression patterns in tissue, multiplexed immunofluorescence mapped multiple cellular markers, and single-cell RNA sequencing provided additional molecular profiles.
The authors describe the resulting dataset as the largest neoadjuvant ICB spatial multi-omics cohort to date. Their analysis found a complex immune architecture rather than a simple accumulation of immune cells. Mature TLS, which are organised immune-cell assemblies that can include germinal centres, appeared alongside TCF7-positive T cells and distinct neighbourhoods of myeloid cells. Germinal centres are specialised areas where B-cell responses mature.
The analysis also identified fibroblasts expressing the signalling molecules CCL19 and CCL21 as a stromal scaffold for these immune hubs. Stromal cells are structural and support cells within tissue; in this setting, the findings suggest that some fibroblasts help organise the local environment in which immune responses develop.
To quantify the architecture, the researchers developed three tools. SCIRA measures receptor–ligand interactions between cells across whole slides. GCSCAN uses molecular information to delineate TLS and germinal-centre structures. PathNet-TLS is designed to detect TLS automatically in haematoxylin-and-eosin (H&E) images, the routine stained sections commonly used in pathology.
Immune structures associated with response
Applying these tools across the cohort, the researchers found differences between responders and non-responders in immune composition, stromal composition and the organisation of cellular neighbourhoods. The analysis also showed regimen-specific architectures. In particular, the authors reported a substantially stronger mature TLS and germinal-centre response with IPI-NIVO than with NIVO-RELA.
TLS and germinal-centre density measured by GCSCAN each stratified disease-free survival in the cohort. Among responders, those without germinal centres had a higher risk of relapse. This makes the density and maturity of these structures candidate biomarkers for further study in treatment-response assessment, rather than established clinical tests.
The researchers then extended the analysis from two dimensions to three. Open-top light-sheet imaging and CODA-based 3D reconstruction revealed interconnected tunnels linking germinal centres and TLS. These connections were not visible in standard two-dimensional histopathology, which examines thin tissue sections and can therefore provide only a limited view of how structures connect through a tumour.
Together, the findings connect molecular measurements at single-cell resolution with tissue-level patterns that may be detectable using computational pathology. The proposed workflow could support future biomarker studies across cancer types, although the current report is an early preprint. The supplied abstract does not provide the cohort size, the detailed composition of all three regimens or numerical effect estimates for the survival analyses, and the clinical value of the tools will depend on validation in additional cohorts.