What is the tumor immune microenvironment?

The population of immune cells present within and around a tumor, along with their type, density and spatial location relative to tumor cells and to each other. Research since 2006 has shown that this spatial arrangement, not simply the total number of immune cells present, predicts clinical outcome and response to immunotherapy.

How does spatial biology inform immunotherapy?

By revealing whether immune cells are absent, physically blocked from reaching the tumor core, or actively present throughout the tumor, three distinct phenotypes with different therapeutic implications. Spatial methods can also identify the specific molecular or structural mechanisms responsible for exclusion, such as a collagen receptor that physically walls off tumors from T cell infiltration, pointing directly at new therapeutic targets.

What is immune exclusion?

A tumor phenotype in which immune cells, particularly T cells, are present at the tumor periphery or in surrounding stroma but are physically prevented from penetrating the tumor core. It is distinct from an immune-desert tumor, which lacks lymphocytes entirely, and the distinction matters because exclusion points toward a physical or structural barrier that may be directly targetable.

Mapping the tumor immune microenvironment

Highlights in this article

What the immune microenvironment determines
Immune contexture and exclusion
Spatial signatures of response
Applications in IO discovery
Emerging spatial IO targets

Reading the tumor immune microenvironment spatially rests on an idea that is nearly two decades old: it is not simply how many immune cells infiltrate a tumor that predicts outcome, but their type, their density, and specifically where within the tumor they are located. That idea predates the spatial biology platforms now used to measure it directly, and understanding its history is what makes the current wave of spatial immuno-oncology findings legible rather than a string of disconnected discoveries.


Key takeaways The type, density and location of immune cells within a tumor, the immune contexture, has been shown to predict clinical outcome since 2006, well before modern spatial platforms existed. Tumors are classified into three spatial immune phenotypes: immune desert, immune-excluded, and inflamed, a framework established in 2017 that still organizes most spatial immuno-oncology work. In one 280-patient head and neck cancer study, median survival differed sharply by phenotype: 37 months for immune desert, 61 for excluded, and 85 for inflamed. A 2021 study identified a specific collagen receptor mechanism physically excluding T cells from tumors, and a humanized antibody targeting it reached a Phase 1 clinical trial. Spatial signatures of response increasingly combine multiple layers of evidence, cell density, direct interaction, and molecular mechanism, rather than resting on any single measurement.

Key takeaways

The type, density and location of immune cells within a tumor, the immune contexture, has been shown to predict clinical outcome since 2006, well before modern spatial platforms existed.
Tumors are classified into three spatial immune phenotypes: immune desert, immune-excluded, and inflamed, a framework established in 2017 that still organizes most spatial immuno-oncology work.
In one 280-patient head and neck cancer study, median survival differed sharply by phenotype: 37 months for immune desert, 61 for excluded, and 85 for inflamed.
A 2021 study identified a specific collagen receptor mechanism physically excluding T cells from tumors, and a humanized antibody targeting it reached a Phase 1 clinical trial.
Spatial signatures of response increasingly combine multiple layers of evidence, cell density, direct interaction, and molecular mechanism, rather than resting on any single measurement.

What the immune microenvironment determines

The claim that immune cell position within a tumor matters, not just immune cell abundance, is old enough to predate the spatial biology field entirely. A landmark 2006 study in Science examined the type, density and location of immune cells within human colorectal tumors and found that this immune contexture predicted clinical outcome, in some analyses more reliably than the tumor staging systems pathologists had relied on for decades. That finding, established using conventional immunohistochemistry on tissue sections rather than any high-throughput spatial platform, is the conceptual root of essentially everything spatial immuno-oncology has built since.

What the immune microenvironment determines, in practical terms for a drug discovery program, is whether an immune-based therapy has any physical opportunity to work at all. A checkpoint inhibitor that successfully unleashes an exhausted T cell accomplishes nothing if that T cell was never able to reach the tumor in the first place, and that distinction, between an immune response that is suppressed and one that is physically excluded, is precisely what position-aware measurement reveals and abundance-only measurement cannot. Our colleagues at Technology Networks develop this further in Spatial Biology in Cancer and the Tumor Microenvironment, covering how spatially defined gene expression programs and structures such as tertiary lymphoid aggregates correlate with immunotherapy response.

Immune contexture and exclusion

The framework that organizes most current spatial immuno-oncology work classifies tumors into three spatial immune phenotypes, established in a widely cited 2017 analysis and now close to a standard vocabulary in the field.

Phenotype	Spatial pattern	Clinical implication
Immune desert	Few or no lymphocytes present in either the tumor parenchyma or its periphery	Immunotherapy has little existing immune activity to unleash; often the poorest prognosis of the three
Immune-excluded	Lymphocytes present but confined to the tumor periphery or stroma, unable to penetrate the tumor core	The immune response exists but is physically blocked, a distinct and separately targetable problem from immune desert
Inflamed	Lymphocytes present throughout the tumor parenchyma, including the core	Generally the most favorable phenotype and the population most likely to respond to checkpoint blockade

Table 1. The three-phenotype framework for spatial immune classification. Distinguishing excluded from desert tumors is only possible with position-aware measurement, since both can show low overall lymphocyte counts if location is not assessed.

That distinction between desert and excluded tumors is the framework’s single most practically important contribution, because the two phenotypes can look similar on a simple lymphocyte count and require entirely different therapeutic strategies. A study of 280 patients with locally advanced head and neck squamous cell carcinoma demonstrates how concretely this classification predicts outcome. Using a simple immunohistochemical algorithm evaluating CD8-positive cytotoxic T cell density in the intraepithelial and stromal tumor compartments, tumors were classified as immune desert, excluded, or inflamed, and this classification produced median overall survival of 37, 61, and 85 months respectively, a substantial and clinically meaningful separation driven entirely by where the T cells were rather than simply how many were present.

Spatial signatures of response

Predicting immunotherapy response from spatial data is increasingly a matter of layering different kinds of positional evidence rather than relying on any single measurement, and the layers available have expanded considerably beyond simple lymphocyte density and location.

Three distinct types of spatial evidence now feed into response prediction.

Cell density and location. The foundational layer described above: how many of which immune cell types are present, and specifically where within the tumor architecture.
Direct molecular interaction. Techniques that confirm two proteins are physically interacting, such as checkpoint receptor-ligand pairs, rather than simply co-located in the same tissue region, a distinction covered in depth in Spatial Biology in Oncology: Decoding the Tumor Microenvironment.
Structural and mechanistic barriers. Physical features of the tissue itself, such as extracellular matrix architecture, that can explain why immune cells are excluded rather than simply documenting that they are, which is the subject of the following section.

That third layer, structural mechanism, is where some of the most actionable recent findings have emerged, because a mechanism, unlike a descriptive spatial pattern, points directly at a therapeutic target.

Applications in IO discovery

The clearest demonstration of moving from a descriptive spatial pattern to a mechanistic, targetable one comes from work identifying exactly how some tumors physically block T cell entry, rather than simply documenting that exclusion occurs.


From a spatial observation to a Phase 1 drug candidate A 2021 study in Nature identified the extracellular domain of discoidin domain receptor 1, a collagen receptor, as a specific mechanism by which tumors align collagen fibers into a physical barrier that excludes T cells. Ablation of the receptor in mouse models of triple-negative breast cancer restored intratumoral T cell penetration and eliminated tumor growth, and in human triple-negative breast cancer tissue, expression of the receptor correlated negatively with T cell abundance within the tumor, exactly as the mechanism would predict. The extracellular domain, rather than the receptor’s intracellular signaling function, was shown to be responsible for the effect, which made the mechanism a tractable target for an antibody rather than a small molecule. A subsequently developed humanized antibody targeting this domain disrupted collagen fiber alignment and increased T cell infiltration in tumor models, and per its 2023 publication had entered a Phase 1 clinical trial. That is a complete pipeline: careful spatial assessment identified a structural exclusion mechanism, the mechanism was validated functionally in animal models and correlated in human tissue, and a therapeutic candidate targeting it reached clinical testing.

From a spatial observation to a Phase 1 drug candidate

A 2021 study in Nature identified the extracellular domain of discoidin domain receptor 1, a collagen receptor, as a specific mechanism by which tumors align collagen fibers into a physical barrier that excludes T cells. Ablation of the receptor in mouse models of triple-negative breast cancer restored intratumoral T cell penetration and eliminated tumor growth, and in human triple-negative breast cancer tissue, expression of the receptor correlated negatively with T cell abundance within the tumor, exactly as the mechanism would predict.

The extracellular domain, rather than the receptor’s intracellular signaling function, was shown to be responsible for the effect, which made the mechanism a tractable target for an antibody rather than a small molecule. A subsequently developed humanized antibody targeting this domain disrupted collagen fiber alignment and increased T cell infiltration in tumor models, and per its 2023 publication had entered a Phase 1 clinical trial. That is a complete pipeline: careful spatial assessment identified a structural exclusion mechanism, the mechanism was validated functionally in animal models and correlated in human tissue, and a therapeutic candidate targeting it reached clinical testing.

Emerging spatial IO targets

The DDR1 example above illustrates a category of target that spatial biology is particularly well suited to surface: structural and stromal mechanisms of immune exclusion that would be invisible to any assay lacking positional information, since the relevant biology is defined entirely by where cells and matrix components sit relative to each other rather than by abundance alone.

Three features distinguish this emerging target category from more conventional immuno-oncology targets.

The target is often structural rather than purely a signaling molecule. Collagen fiber architecture, rather than a receptor-ligand pair alone, was the operative barrier in the DDR1 example, which is a different kind of drug target than most checkpoint biology addresses.
The therapeutic hypothesis is combination-oriented by nature. Removing a physical barrier to infiltration does not itself kill tumor cells; it is intended to let an existing or separately administered immune response reach cells it previously could not, which argues for pairing exclusion-targeting therapies with checkpoint or cellular therapies rather than deploying them alone.
Validation requires spatial evidence at every stage. Unlike a target whose relevance can be established from expression data alone, a structural exclusion mechanism can only be confirmed, from initial observation through to clinical biomarker development, using methods that preserve tissue architecture.

The broader question of which spatial targets and signatures are being pursued across immuno-oncology, and how they are being combined into therapeutic strategies, is developed in full in Spatial Biology in Immuno-Oncology: Targets, Signatures, and Combinations. The specific question of predicting immunotherapy response using spatial data, including how the evidence layers described above are combined into a usable predictive signature, is treated in Predicting Response to Immunotherapy With Spatial Signatures.

For the broader tumor microenvironment argument this spoke builds on, see Spatial Biology in Oncology: Decoding the Tumor Microenvironment, and for where immuno-oncology sits within the full spatial biology pipeline, Spatial biology in drug discovery: From target discovery to translational medicine.

This article was produced under Drug Discovery News’s editorial policies.