The research team, led by Romgase Sakamula and corresponding author Somponnat Sampattavanich, set out to address a well-recognized limitation in clinical oncology. PD-L1 expression, the protein biomarker currently used to guide immunotherapy decisions in lung cancer, has limited predictive performance on its own. Many patients with high PD-L1 scores fail to respond, while some with low scores experience remarkable tumor regression. Emerging evidence has pointed toward a more nuanced explanation: therapeutic response depends not merely on the abundance of immune cells within a tumor but on their spatial organization, the intricate architecture of cellular neighborhoods that determines whether killer T cells can actually reach and engage their malignant targets.

To capture this architecture, the investigators employed tissue-based cyclic immunofluorescence, abbreviated t-CyCIF, a technique that repeatedly stains the same tissue section with fluorescently labeled antibodies, imaging one round of markers after another and stripping the signal between cycles. By iterating this process, the team built a 13-marker panel capable of single-cell phenotyping across preserved archival tissue. This approach allowed them to identify individual cell types, including cytotoxic T cells, regulatory immune populations, and tumor cells, while retaining the exact positional relationships among them, something that dissociative methods such as single-cell sequencing of suspended cells cannot preserve.

The cohort consisted of seventeen pre-treatment specimens from patients with non-small cell lung cancer who subsequently received immune checkpoint inhibitors in a real-world clinical setting. Ten of these patients responded to therapy and seven did not. When the researchers first compared global measures, the density of immune cells and tumor cells across whole samples, they found no significant differences between the two groups. This null result is itself instructive, reinforcing the growing consensus that bulk cellular counts obscure the biologically meaningful features of the tumor microenvironment. The decisive information, the study suggests, resides in spatial relationships rather than simple abundance.

Because spatial analysis requires intact tissue architecture, the team focused their most detailed geometric investigations on a subset of resection specimens with preserved structure, comprising four responders and three non-responders. Within these samples they examined several layers of organization: lymphoid structures, cellular neighborhoods, pairwise distances between cell populations, and higher-order tissue architecture. Metrics related to tertiary lymphoid structures, organized aggregates of immune cells that form within tumors and are thought to serve as sites of local immune activation, showed a trend toward higher values in responders, hinting at greater immune organization within their tumors, although the small sample size means this observation requires validation in larger cohorts.

The cellular neighborhood analysis produced some of the study’s most striking findings. In responder tumors, regions enriched for cytotoxic T cells and for vasculature were significantly expanded, whereas non-responder tumors displayed a predominantly tumor-centered spatial architecture, with malignant cells dominating the landscape and immune populations relegated to the periphery. This distinction implies that the gross topological layout of a tumor, whether immune activity is woven into its core or pushed to its margins, may fundamentally shape how the tumor responds to checkpoint blockade, which works by releasing the brakes on T cells that must be in physical proximity to cancer cells to kill them.

Perhaps the most compelling discovery centered on the perivascular niche, the zone immediately surrounding tumor blood vessels. Responders exhibited shorter distances between vessels and cytotoxic T cells, along with greater enrichment of cytotoxic and PD-1-positive cytotoxic T cells in these perivascular regions. Higher-order spatial analysis further highlighted the preferential organization of PD-1-positive cytotoxic T cells within vascular-associated niches in responder tumors. This arrangement carries mechanistic significance: PD-1 is the receptor targeted by checkpoint inhibitors such as pembrolizumab and nivolumab, and PD-1-positive T cells positioned near vessels are plausibly the cells most accessible to circulating antibody drugs and best positioned to traffic between the bloodstream and the tumor parenchyma. A tumor that stations its exhausted-but-revivable killer cells along its vascular highways may be primed for checkpoint release in a way that a tumor with sequestered, disorganized immune infiltrates is not.

These findings carry substantial implications for biomarker development. If validated, perivascular immune organization could serve as a complementary predictive framework, layered alongside PD-L1 testing to refine patient selection for immunotherapy. The technical requirements are notable but feasible: the analysis relies on formalin-fixed paraffin-embedded archival tissue, the standard specimen type held by pathology departments worldwide, and cyclic immunofluorescence can in principle be implemented on existing clinical material. Spatial biology platforms of this kind are advancing rapidly across oncology, and this study demonstrates their application to a clinically urgent question using a modest number of samples and a real-world patient population rather than a curated trial cohort.

Important caveats temper the enthusiasm. The study analyzed seventeen patients overall, with spatial analyses restricted to seven resection specimens, numbers far too small to support clinical decision-making. The reported trends, including the tertiary lymphoid structure metrics, will need confirmation in larger, independent, and ideally prospective cohorts before spatial biomarkers can enter routine practice. The retrospective design, approved by the Siriraj Institutional Review Board with the informed consent requirement waived, also means that treatment regimens and follow-up were not standardized as they would be in a controlled trial. Nevertheless, the consistency of the spatial signal, emerging across multiple independent analytical approaches, lends credibility to the central conclusion.

The study, funded by the National Research Council of Thailand, the Siriraj Foundation, and the Siriraj Research Development Fund, adds to a rapidly accumulating body of evidence that cancer immunology must be read in three dimensions. The tumor microenvironment is not a well-mixed soup of cells but a structured tissue, with vascular corridors, lymphoid aggregates, and territorial neighborhoods that govern immune surveillance. As the authors demonstrate, the difference between a tumor that succumbs to immunotherapy and one that resists it may be written into that geography before the first dose of treatment is ever given. For patients with non-small cell lung cancer, the road to personalized immunotherapy may soon run through the perivascular spaces of their own tumors.

Subject of Research: Spatial organization of perivascular immune cells as a biomarker for immunotherapy response in non-small cell lung cancer

Article Title: Perivascular immune organization and spatial architecture are associated with immunotherapy response in non-small cell lung cancer

Article References: Sakamula, R., Likhityungyuen, T., Anekpuritanang, T., Korphaisarn, K., & Sampattavanich, S. (2026). Perivascular immune organization and spatial architecture are associated with immunotherapy response in non-small cell lung cancer. Cancer Immunology, Immunotherapy. https://doi.org/10.1007/s00262-026-04549-y

Image Credits: AI Generated

DOI: 10.1007/s00262-026-04549-y

Keywords: non-small cell lung cancer, immunotherapy, immune checkpoint inhibitors, tumor microenvironment, spatial biology, perivascular immune niche, cyclic immunofluorescence, PD-L1, cytotoxic T cells, tertiary lymphoid structures, biomarker, tumor vasculature