How is spatial biology used in immuno-oncology?

Spatial biology measures checkpoint proteins, exhaustion markers, and immune cell populations while preserving their tissue location, revealing whether the physical interactions a checkpoint drug depends on are actually occurring. This has produced multi-marker signatures, such as combined PD-1, LAG-3, TIGIT, and TIM-3 co-expression, that predict treatment resistance independent of any single marker measured alone.

What are spatial IO biomarkers?

Predictive signatures built from spatially resolved measurements of multiple immune markers, cell types, or molecular layers rather than a single protein readout such as PD-L1 alone. A 2024 study combining four exhaustion markers produced a validated 25-gene signature predicting checkpoint inhibitor response with high accuracy across several clinical trial datasets.

Can spatial data guide combination immunotherapy?

Yes. Identifying which additional checkpoints or resistance markers are co-expressed within a patient’s tumor can directly justify which second agent to add to a checkpoint inhibitor backbone. This logic underlies the FDA-approved combination of relatlimab, targeting LAG-3, with nivolumab for metastatic melanoma, the first approved combination built around a checkpoint beyond PD-1/PD-L1 and CTLA-4.

Spatial biology in immuno-oncology

Highlights in this article

Why IO is a spatial problem
Spatial targets beyond checkpoints
Response and resistance signatures
Rationalizing combinations
The biomarker opportunity

Building immuno-oncology biomarkers that actually predict which patients respond has proven far harder than the initial success of checkpoint blockade suggested it would be. More than a decade after the first approvals, the field has grown into a substantial and still-expanding drug development effort, and spatial biology is increasingly where its next generation of targets and patient-selection tools is coming from, because it can finally show where the interactions these drugs act on are actually happening.


Key takeaways A single biomarker, such as PD-L1 expression alone, has proven structurally insufficient for reliably predicting checkpoint inhibitor response. A 2024 study found that co-expression of four exhaustion markers, PD-1, LAG-3, TIGIT, and TIM-3, predicted checkpoint inhibitor resistance independent of PD-L1 status, and yielded a validated 25-gene predictive signature. Relatlimab, targeting LAG-3, combined with nivolumab is already FDA-approved for melanoma, showing that biology of this kind can translate directly into an approved combination regimen. Rationalizing a combination increasingly means showing, spatially, which specific exhaustion or resistance markers coexist in the same patient’s tumor before committing to which second agent to add. The biomarker opportunity in spatial immuno-oncology lies less in discovering new individual targets and more in combining multiple spatially resolved signals into a single predictive readout.

Key takeaways

A single biomarker, such as PD-L1 expression alone, has proven structurally insufficient for reliably predicting checkpoint inhibitor response.
A 2024 study found that co-expression of four exhaustion markers, PD-1, LAG-3, TIGIT, and TIM-3, predicted checkpoint inhibitor resistance independent of PD-L1 status, and yielded a validated 25-gene predictive signature.
Relatlimab, targeting LAG-3, combined with nivolumab is already FDA-approved for melanoma, showing that biology of this kind can translate directly into an approved combination regimen.
Rationalizing a combination increasingly means showing, spatially, which specific exhaustion or resistance markers coexist in the same patient’s tumor before committing to which second agent to add.
The biomarker opportunity in spatial immuno-oncology lies less in discovering new individual targets and more in combining multiple spatially resolved signals into a single predictive readout.

Why IO is a spatial problem

Immuno-oncology has grown into one of drug development’s largest and most active areas, and our earlier coverage in Immuno-oncology drug development growth traces the scale of that expansion across the pipeline. Scale alone has not solved the field’s central practical problem, though: predicting which patients will respond to a given checkpoint inhibitor, or a given combination, remains difficult, and a large share of that difficulty traces back to biomarkers built on a single measurement taken without regard to where in the tumor it was made.

A checkpoint interaction, almost by definition, is a spatial event: it requires a receptor on one cell and its ligand on a neighboring cell, physically close enough to engage. Measuring either protein’s abundance in a homogenized tissue sample says nothing about whether that physical engagement is actually occurring, or where, or how often, which is precisely the information a spatial platform is built to supply.

Spatial targets beyond checkpoints

The most direct route to a better biomarker, current evidence suggests, is not necessarily a new individual target but a more complete spatial picture of several targets measured together in the same tissue.


Four markers, measured together, tell a different story than any one alone A 2024 study analyzed tumor samples from 166 patients with lung adenocarcinoma using multiplex immunofluorescence, quantifying tumor-infiltrating lymphocytes expressing four separate exhaustion markers simultaneously: PD-1, LAG-3, TIGIT, and TIM-3. Co-expression of these markers was associated with resistance to immune checkpoint inhibition, and critically, this association held irrespective of PD-L1 status, the single biomarker most current checkpoint therapies are selected on. From this multi-marker data, the study derived a 25-gene signature indicative of CD8-positive T cell exhaustion, validated using several datasets from various clinical trials, with high predictive accuracy for checkpoint inhibitor response. The finding directly challenges the traditional view that a single exhaustion marker, measured in isolation, adequately captures which patients are likely to resist treatment. It does not; the combination measured together, in place, carries information no individual marker supplies.

Four markers, measured together, tell a different story than any one alone

A 2024 study analyzed tumor samples from 166 patients with lung adenocarcinoma using multiplex immunofluorescence, quantifying tumor-infiltrating lymphocytes expressing four separate exhaustion markers simultaneously: PD-1, LAG-3, TIGIT, and TIM-3. Co-expression of these markers was associated with resistance to immune checkpoint inhibition, and critically, this association held irrespective of PD-L1 status, the single biomarker most current checkpoint therapies are selected on.

From this multi-marker data, the study derived a 25-gene signature indicative of CD8-positive T cell exhaustion, validated using several datasets from various clinical trials, with high predictive accuracy for checkpoint inhibitor response. The finding directly challenges the traditional view that a single exhaustion marker, measured in isolation, adequately captures which patients are likely to resist treatment. It does not; the combination measured together, in place, carries information no individual marker supplies.

That finding reframes what counts as a spatial IO target. It is not only a new protein or structural mechanism, of the kind covered in Mapping the Tumor Immune Microenvironment for Immuno-Oncology's DDR1 example. A defined combination of markers, measured in the same location on the same cells, can itself function as the target-adjacent biomarker a drug development program needs.

Response and resistance signatures

The 25-gene signature above is one specific instance of a broader pattern: the strongest current spatial immuno-oncology signatures combine several distinct measurements rather than resting on any single one, whether that combination spans multiple exhaustion markers, multiple cell types, or multiple molecular layers.

Signature type	What it measures	What it adds over a single marker
Multi-marker exhaustion co-expression	Simultaneous expression of several exhaustion receptors, such as PD-1, LAG-3, TIGIT, and TIM-3, on the same cells	Predicts resistance independent of any single marker’s status, including PD-L1
Cell-type composition signatures	The relative abundance and spatial arrangement of multiple immune and stromal cell types within the tumor	Captures microenvironment context a single protein readout cannot represent
Cross-layer molecular signatures	Combined transcriptomic and proteomic evidence from the same tissue	Confirms that a transcriptional signal corresponds to an actual functional protein state, not only gene expression

Table 1. Three types of spatial signature currently used in immuno-oncology, and what each adds beyond a single-marker readout. The pattern across all three is combination rather than substitution.

That table’s pattern, combination rather than substitution, is worth stating plainly as the field’s current direction. No single new marker discovered to date has replaced PD-L1 as a biomarker; instead, the field has moved toward measuring several markers together and accepting that the combination, not any individual component, is what actually predicts outcome.

Rationalizing combinations

Multi-marker spatial evidence has a direct, practical use beyond biomarker development: it can justify which second agent to add to a checkpoint inhibitor backbone, and this is not a hypothetical application.

Relatlimab, an antibody targeting LAG-3, in combination with nivolumab, an anti-PD-1 antibody, was validated by the RELATIVITY-047 trial and subsequently approved by the FDA for first-line treatment of metastatic melanoma, the first clinical validation of a checkpoint target beyond PD-1/PD-L1 and CTLA-4. That approval demonstrates concretely that biology of the kind the exhaustion co-expression study above describes, evidence that a second checkpoint is meaningfully active alongside PD-1, can translate all the way into an approved, prescribed combination regimen rather than remaining a research finding.

The practical logic connecting spatial evidence to combination selection follows a specific sequence.

Identify which additional checkpoints or resistance markers are co-expressed in a given patient population. The 2024 exhaustion study demonstrates the method: measure several candidates together, in place, rather than one at a time.
Confirm the co-expression is spatially and functionally meaningful, not incidental. Markers present in the same tissue region on the same cells, ideally with evidence they are functionally engaged, carry more weight than markers simply detected somewhere in the same sample.
Select a second agent targeting the co-expressed marker with the strongest resistance association. This is precisely the logic that supported LAG-3 as the second target in relatlimab plus nivolumab, ahead of other candidate checkpoints.

That sequence is a considerably more evidence-based route to combination selection than testing candidate second agents empirically without first establishing which resistance mechanism is actually present and active in the relevant patient population.

The biomarker opportunity

The clearest opportunity spatial biology currently offers immuno-oncology is not a single undiscovered target waiting to be found, but a more complete way of combining evidence that already exists into a signature that predicts response more reliably than any one component does alone.

Three practical implications follow for a biomarker development program built on this pattern.

Single-marker biomarkers should be treated as a floor, not a ceiling. PD-L1 testing remains clinically standard, but the exhaustion co-expression evidence above shows it captures only part of what determines resistance, so a program aiming for a more predictive biomarker should plan to measure additional markers in the same tissue from the outset.
Validation across independent clinical trial datasets is what separates a real signature from a promising correlation. The 25-gene signature’s validation across several trial datasets is precisely the kind of reproducibility check that should be expected of any spatial signature before it informs a real development decision.
A biomarker strategy and a combination therapy strategy are, in practice, the same exercise. Identifying which resistance mechanism is spatially present in a patient’s tumor is simultaneously a patient-selection tool and a rationale for which second agent that patient is most likely to benefit from.

The broader question of how spatial biomarkers move toward regulatory and clinical use, including companion diagnostic development specifically, is developed in full in Spatial Biomarkers and Companion Diagnostics: The Next Frontier].

For the broader tumor immune microenvironment argument this spoke builds on, see Mapping the Tumor Immune Microenvironment for Immuno-Oncology and 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 AI editorial policies.