Beyond Counting Immune Cells: How the Spatial Architecture of Lung Cancer May Shape Immunotherapy Response
A SiSP study using real-world clinical specimens suggests that the organization of immune cells around blood vessels may provide clues to why some lung cancers respond to immunotherapy
Immune checkpoint inhibitors have changed the treatment landscape for non-small cell lung cancer (NSCLC). For some patients, these therapies can produce long-lasting disease control. For others, however, the benefit is limited.
One of the continuing challenges in precision oncology is therefore deceptively simple:
Can we understand, before treatment, why one tumor is more likely to respond than another?
PD-L1 expression is currently one of the most widely used biomarkers to help guide immunotherapy, but it does not tell the whole story. Tumors with apparently similar biomarker profiles can behave very differently. Increasing evidence suggests that the answer may depend not only on which cells are present inside a tumor, but also on where those cells are located and how they are organized relative to one another.
In a new study from our team at the Siriraj Center of Research Excellence for Systems Pharmacology (SiCORE-PMSP/SiSP), we explored this idea using spatial biology to examine pretreatment lung cancer tissues.
Our study, “Perivascular immune organization and spatial architecture are associated with immunotherapy response in non–small cell lung cancer,” has been accepted in Cancer Immunology, Immunotherapy.
What we found points toward an intriguing idea: response to immunotherapy may be associated not simply with how many immune cells are inside a tumor, but with how those immune cells are organized—particularly around blood vessels.
Learning from real-world clinical specimens
An important aspect of this study is where the tissues came from.
These were not specimens collected specifically for a spatial biology research project under optimized experimental conditions. Instead, we studied archived pretreatment tumor specimens from patients with NSCLC who had been treated at Siriraj Hospital and subsequently received PD-1/PD-L1 blockade-based immunotherapy.
We went back to tissues obtained through clinical care and connected what we could see in those tumors with what subsequently happened to the patients during treatment. This distinction is important to us.
Real clinical specimens are not perfect research specimens. Some patients have large surgical resections; others have only small biopsies. The amount of tissue available varies. Tissue architecture may not always be sufficiently preserved for every type of spatial analysis. In our study, 10 of the 17 specimens were biopsies and seven were surgical resections.
These realities created limitations for our study, but they also reflect the environment in which precision oncology actually needs to operate.
If spatial biomarkers are eventually going to help patients, they must ultimately work with the kinds of tissues that hospitals routinely collect.
For us, this study was therefore also an opportunity to ask whether archived clinical pathology specimens could be revisited with modern spatial technologies to reveal information that was not accessible when those tissues were originally collected.
A tumor is more than a collection of cells
Traditionally, many cancer biomarkers reduce a tumor to a measurement: how strongly a protein is expressed, how many immune cells are present, or whether a particular genomic alteration can be detected.
Spatial biology asks a different set of questions.
Where are those cells?
Who are their neighbors?
Are immune cells able to reach the tumor?
And how are immune cells, tumor cells, stromal cells and blood vessels organized into a functioning tissue ecosystem?
Two tumors could contain similar numbers of cytotoxic T cells but arrange those cells very differently. One tumor might position them in areas where they can participate in an effective antitumor response, while another might organize them in a very different microenvironment.
To explore this, we used tissue-based cyclic immunofluorescence (t-CyCIF), a multiplex imaging technology that allows multiple proteins to be measured while preserving the spatial location of individual cells.
Using a 13-marker panel, we mapped tumor, immune and vascular compartments at single-cell resolution in pretreatment NSCLC tissues.
The first clue: simply counting cells was not enough
We first asked the most straightforward question:
Did tumors from patients who responded to immunotherapy simply contain more immune cells?
Surprisingly, we did not observe significant differences in the overall densities of the major lymphoid, myeloid, tumor or vascular cell populations between responders and non-responders.
In other words, counting cells alone did not clearly separate the two groups.
That observation led us to what became the central question of the study:
Perhaps it is not only how many immune cells are present—but where they are.
Looking at tumors as neighborhoods
Rather than treating every cell independently, we next examined how cells were assembled into local cellular neighborhoods.
Seven recurring neighborhood types emerged, including regions enriched for cytotoxic T cells, macrophages, tumor cells and blood vessels.
The spatial organization of these neighborhoods appeared different between the two response groups.
Non-responder tumors contained more tumor-core-associated neighborhoods. In contrast, responder tumors showed greater representation of cytotoxic T-cell-enriched neighborhoods and vasculature-associated neighborhoods. The responder tumors also showed greater interaction among immune- and vascular-associated regions.
This observation directed our attention toward something that is sometimes overlooked when thinking about cancer immunity:
the tumor vasculature.
Blood vessels may be more than plumbing
Blood vessels are often thought of primarily as the tumor's supply system, providing oxygen and nutrients.
But blood vessels are also the routes through which circulating immune cells reach tissues.
We therefore examined how immune cells were positioned relative to tumor blood vessels.
Here, an interesting pattern began to emerge.
In responder tumors, cytotoxic T cells were located closer to blood vessels. PD-1-positive cytotoxic T cells also showed greater enrichment around vessels.
In contrast, non-responder tumors showed greater enrichment of proliferating Ki67-positive tumor cells in regions surrounding vessels.
These observations raise the possibility that the space around tumor blood vessels may represent an important microenvironmental niche.
In some tumors, this perivascular space appeared more closely associated with cytotoxic immune cells.
In others, it appeared more associated with proliferating tumor cells.
We describe the former pattern as a potential perivascular immune niche.
Importantly, our data do not establish why this organization occurs—or whether it directly causes better response to immunotherapy. Blood vessels could potentially influence immune-cell recruitment, trafficking or retention, but our current study cannot determine the underlying mechanism.
For now, it is a spatial association and a biological hypothesis that deserves further investigation.
From pairs of cells to tissue architecture
Biological tissues are more complicated than interactions between two cell types.
We therefore extended the analysis to ask how blood vessels, immune cells and tumor cells were organized together.
Among the configurations examined, the relationship between blood vessels, PD-1-positive cytotoxic T cells and tumor cells showed the clearest separation between responders and non-responders.
In responder tumors, PD-1-positive cytotoxic T cells showed preferential organization within vascular-associated regions.
What interested us was that this general pattern appeared repeatedly when we looked at the tissue in different ways.
Cellular-neighborhood analysis pointed toward immune–vascular organization.
Distance analysis pointed toward immune–vascular organization.
Local proximity analysis pointed toward immune–vascular organization.
And higher-order spatial analysis again pointed toward immune–vascular organization.
Taken together, these complementary analyses suggested the presence of a coordinated perivascular immune architecture associated with immunotherapy response in this cohort.
Another sign of immune organization: tertiary lymphoid structures
We also examined tertiary lymphoid structures (TLSs)—organized aggregates of immune cells that can form within or around tumors.
TLSs are particularly interesting because they represent something beyond immune-cell abundance: they show that immune cells have assembled into an organized structure.
In our cohort, TLS density and the proportion of tumor area occupied by TLSs were higher in responders, although the significance level was still not strong, perhaps due to limited specimens (P = 0.0571).
We therefore do not interpret this as evidence that TLSs predict response in our dataset.
Instead, we see it as another observation consistent with a broader hypothesis:
perhaps responding tumors are characterized by a more organized immune ecosystem rather than simply having more immune cells.
A promising observation—but a small beginning
It is important to put these findings in perspective.
This was a small, retrospective and exploratory study.
Seventeen patients were included in the overall analysis: 10 responders and seven non-responders. Because detailed spatial analyses require sufficiently preserved tissue architecture, these analyses could only be performed using the seven surgical resection specimens—four responders and three non-responders.
We therefore do not consider the perivascular immune niche a validated biomarker of immunotherapy response, and the current results should not be used to guide treatment decisions.
Instead, the study provides something different: a hypothesis and a framework for what to investigate next.
From pathology archives to spatially informed precision medicine
One aspect of this work that particularly excites us is what it suggests about the future value of clinical pathology specimens.
Hospitals have accumulated enormous archives of tissues collected during routine patient care. Traditionally, much of the information extracted from these specimens has come from histology and a relatively small number of molecular markers.
But the tissue contains much more information.
The location of every cell is preserved.
The relationships between tumor and immune cells are preserved.
The vascular architecture is preserved.
And increasingly, spatial technologies allow us to recover and quantify that information.
When these tissues can be connected with real-world treatment histories and patient outcomes, pathology archives could become powerful resources for discovering how the organization of human tumors influences therapeutic response.
Our study is one small example of what might be possible.
A beginning, not an endpoint
Perhaps the most important lesson from this study is conceptual.
Cancer biology is often summarized by asking:
What mutations does this tumor have?
What proteins does it express?
How many immune cells are present?
Spatial biology adds another layer:
Where are those cells, who are their neighbors, and how are they organized into a tissue?
Our results suggest that this organization may contain clinically relevant information that simple measurements of cellular abundance cannot capture. Responders and non-responders in our cohort did not differ dramatically in how many major immune-cell populations they contained. What appeared different was how parts of the tumor ecosystem were arranged relative to one another.
The current dataset is modest, and these findings remain hypothesis-generating. The next steps will require larger independent cohorts, prospective studies and experiments to understand the biology underlying these spatial patterns.
But every research direction has to begin somewhere.
For us, this study represents an early step toward understanding cancer not simply as a collection of molecular alterations or individual cells, but as an organized and dynamic tissue ecosystem.
And importantly, that first step came from real patients and real clinical specimens.
There are many more clinical specimens to learn from, many more tumor ecosystems to map, and many more questions to answer.
We congratulate Dr. Romgase Sakamula and Thanaphon Likhityungyuen, our main research team, on reaching this important milestone. We thank Dr. Krittiya Korphaisarn, our clinician collaborator, for making the clinical specimens and clinical data available, and Dr. Tauangtham Anekpuritanang, our pathologist, for providing the pathology expertise that made their interpretation possible. Together with Dr. Somponnat Sampattavanich, the study's principal investigator, this team has taken a small but important first step toward understanding how the spatial organization of real human tumors may influence treatment response.
This is only a beginning. We hope it will be the first of many studies to come.
Publication: Romgase Sakamula, Thanaphon Likhityungyuen, Tauangtham Anekpuritanang, Krittiya Korphaisarn & Somponnat Sampattavanich. Perivascular immune organization and spatial architecture are associated with immunotherapy response in non–small cell lung cancer.Cancer Immunology, Immunotherapy (2026). DOI: 10.1007/s00262-026-04549-y.
