How is spatial transcriptomics used in cancer?

Spatial transcriptomics maps gene expression across a tumor while preserving each measurement’s tissue location, used to characterize baseline tumor biology, confirm a drug candidate’s mechanism of action once treatment begins, and investigate why resistance or relapse occurred afterward. A phase 1 clinical trial in hepatocellular carcinoma has already used it to identify a transcriptional program distinguishing treatment responders from non-responders.

Can spatial transcriptomics reveal drug resistance?

Yes. It can show where within a tumor resistance-associated gene expression is concentrated, sometimes even within a single patient who otherwise responded well to treatment. One study identified spatially heterogeneous expression of BCL2L1, encoding BCL-XL, during the emergence of drug-tolerant persister cells in EGFR-targeted lung cancer therapy, and found that inhibiting that gene mitigated resistance.

What does spatial transcriptomics show in tumors?

It shows which genes are active in specific tumor regions and cell populations, revealing gene expression programs, such as a tertiary lymphoid structure signature linked to immunotherapy response, that would be invisible in bulk tissue measurements. It can also reveal that a tumor classified as a single treatment outcome actually contains spatially distinct regions with different underlying biology.

Spatial transcriptomics in cancer drug development

Highlights in this article

Spatial transcriptomics in the oncology pipeline
Mapping tumor gene programs
Mechanism-of-action studies
Resistance and relapse biology
Integrating with other data

Applying spatial transcriptomics to an oncology drug development question means asking not just which genes a tumor expresses, but where within the tumor those genes are active, and how that spatial pattern shifts once treatment begins. That shift, a gene expression program appearing, disappearing, or relocating within a tumor over the course of therapy, is frequently the most direct evidence a drug development team can get for why a treatment worked in one patient and not another.


Key takeaways A 2023 study applied spatial transcriptomics directly within an active phase 1 clinical trial, identifying a specific transcriptional program that distinguished responding from non-responding hepatocellular carcinoma tumors. Within that same trial, one patient classified as a treatment success by standard criteria still contained a spatially localized pocket of resistance biology invisible to non-spatial assessment. A separate study identified a specific, druggable gene, BCL2L1, whose spatially heterogeneous expression tracked the emergence of drug-tolerant persister cells during EGFR-targeted therapy. A 30-gene tertiary lymphoid structure signature, identified spatially in triple-negative breast cancer, predicts immunotherapy response and stratifies patients into distinct therapeutic vulnerability groups. Spatial transcriptomics is increasingly integrated with bulk sequencing and protein-level data rather than used as a standalone readout.

Key takeaways

A 2023 study applied spatial transcriptomics directly within an active phase 1 clinical trial, identifying a specific transcriptional program that distinguished responding from non-responding hepatocellular carcinoma tumors.
Within that same trial, one patient classified as a treatment success by standard criteria still contained a spatially localized pocket of resistance biology invisible to non-spatial assessment.
A separate study identified a specific, druggable gene, BCL2L1, whose spatially heterogeneous expression tracked the emergence of drug-tolerant persister cells during EGFR-targeted therapy.
A 30-gene tertiary lymphoid structure signature, identified spatially in triple-negative breast cancer, predicts immunotherapy response and stratifies patients into distinct therapeutic vulnerability groups.
Spatial transcriptomics is increasingly integrated with bulk sequencing and protein-level data rather than used as a standalone readout.

Spatial transcriptomics in the oncology pipeline

Spatial transcriptomics entered oncology research faster than most other spatial methods for a specific reason: gene expression, unlike protein abundance, can be measured genome-wide from a single tissue section without requiring a preselected antibody panel, which makes it a natural first step for a discovery-stage question before committing to a narrower, targeted protein assay.

Our colleagues at Technology Networks set out the broader methods landscape this pipeline draws on in Spatial Biology: A Complete Guide to Methods, Technologies, and Applications, which covers how spatial transcriptomics platforms have evolved from fluorescence in situ hybridization and immunohistochemistry, methods limited to one or two targets at a time, into genome-wide, tissue-preserving assays. Within a drug development pipeline specifically, that capability tends to enter at three distinct points: characterizing baseline tumor biology before treatment, confirming a candidate’s mechanism of action once treatment has begun, and investigating why resistance or relapse occurred after treatment appeared to work.

Mapping tumor gene programs

Before a drug enters the picture at all, spatial transcriptomics is used to establish which gene expression programs exist within a tumor and where they are active, work that increasingly identifies programs with direct therapeutic relevance rather than purely descriptive biology.

A study applying spatial transcriptomics to 92 patients with triple-negative breast cancer identified a 30-gene tertiary lymphoid structure signature, enriched in B cell markers and lymphoid priming genes, that predicts improved survival and response to immunotherapy in triple-negative breast cancer and other cancer types. Spatially, these tertiary lymphoid structures were predominantly intratumoral or stromal, and by integrating spatial transcriptomics with bulk RNA sequencing, the study identified nine distinct spatial archetypes, including one enriched for tertiary lymphoid structures and immune activation, and a separate immunosuppressive archetype. These archetypes stratify patients by therapeutic vulnerability, which turns a descriptive gene expression map directly into a patient-selection tool.

That kind of signature, a defined gene set with a demonstrated link to treatment response, is what separates mapping tumor gene programs as an academic exercise from mapping them as a drug development asset. A named signature can be checked in a new patient sample and used to make an actual treatment decision; a general description of tumor heterogeneity cannot.

Mechanism-of-action studies

The clearest current demonstration of spatial transcriptomics used directly within an active drug development program, rather than retrospectively on banked samples, comes from a phase 1 clinical trial in hepatocellular carcinoma.


A named transcriptional program distinguishing response from non-response A 2023 study, registered as ClinicalTrials.gov NCT03299946, applied spatial transcriptomics to resection specimens from 15 patients treated with neoadjuvant cabozantinib, a multi-tyrosine kinase inhibitor primarily blocking VEGF signaling, combined with nivolumab, a PD-1 inhibitor. Five of the 15 patients achieved a pathologic response. Spatial transcriptomics showed that responding tumors were enriched for immune cells and cancer-associated fibroblasts exhibiting pro-inflammatory signaling relative to non-responders. More specifically, the cancer-immune interactions enriched in responding tumors were characterized by activation of the PAX5 module, a known regulator of B cell maturation, and this activation colocalized spatially with regions of increased B cell marker expression, indicating genuine functional B cell activity rather than incidental gene expression. Separately, cancer-associated fibroblast interactions in responders were associated with extracellular matrix remodeling, reflected in elevated FOS and JUN activation in fibroblasts positioned adjacent to the tumor. This is mechanism-of-action evidence in the fullest sense: a specific transcriptional program, named and spatially located, that plausibly explains why some patients responded to this combination and others did not.

A named transcriptional program distinguishing response from non-response

A 2023 study, registered as ClinicalTrials.gov NCT03299946, applied spatial transcriptomics to resection specimens from 15 patients treated with neoadjuvant cabozantinib, a multi-tyrosine kinase inhibitor primarily blocking VEGF signaling, combined with nivolumab, a PD-1 inhibitor. Five of the 15 patients achieved a pathologic response. Spatial transcriptomics showed that responding tumors were enriched for immune cells and cancer-associated fibroblasts exhibiting pro-inflammatory signaling relative to non-responders.

More specifically, the cancer-immune interactions enriched in responding tumors were characterized by activation of the PAX5 module, a known regulator of B cell maturation, and this activation colocalized spatially with regions of increased B cell marker expression, indicating genuine functional B cell activity rather than incidental gene expression. Separately, cancer-associated fibroblast interactions in responders were associated with extracellular matrix remodeling, reflected in elevated FOS and JUN activation in fibroblasts positioned adjacent to the tumor. This is mechanism-of-action evidence in the fullest sense: a specific transcriptional program, named and spatially located, that plausibly explains why some patients responded to this combination and others did not.

Resistance and relapse biology

The same hepatocellular carcinoma trial supplies a second, distinct finding directly relevant to relapse. Among the patients with a major pathologic response, one patient experienced early recurrence, and spatial transcriptomics of that specific tumor revealed a distinctive immune-poor region resembling the non-responding tumor microenvironment seen across the broader patient cohort, characterized by cancer cell-fibroblast interactions and expression of cancer stem cell markers. That is a plausible spatial mechanism for early immune escape specific to a single patient, and it demonstrates something a non-spatial assessment of that same patient’s overall pathologic response would have missed entirely: a treatment classified as successful by standard criteria can still contain a spatially confined pocket of resistance biology.

A separate study of lung adenocarcinoma resistance to EGFR-targeted therapy illustrates how spatial transcriptomics can point directly at a druggable countermeasure rather than only describing resistance descriptively. Comparing untreated tumor tissue, tissue reflecting a short-term drug-tolerant persister state under EGFR-TKI treatment, and tissue from eventual recurrence, the study identified spatial heterogeneity in expression of BCL2L1, the gene encoding BCL-XL, within tumor cells specifically. Both genetic ablation and pharmacological inhibition of BCL2L1/BCL-XL mitigated the onset of therapy resistance in this system, which converts a spatial observation about where a resistance-associated gene is expressed into a specific, testable combination therapy hypothesis: pairing EGFR inhibition with a BCL-XL inhibitor to prevent the persister population from ever establishing itself.

Integrating with other data

None of the three findings above rested on spatial transcriptomics alone. The hepatocellular carcinoma trial combined spatial transcriptomics with clinical pathologic response data; the tertiary lymphoid structure signature study integrated spatial transcriptomics with bulk RNA sequencing to derive its nine spatial archetypes; and the EGFR resistance study connected a spatial expression pattern to a functional genetic and pharmacological validation step.

That pattern, spatial transcriptomics supplying the where and a second data type supplying confirmation or scale, recurs across essentially every drug-development-relevant application of the technology, for three practical reasons.

Spatial platforms often sample less deeply per position than bulk sequencing. Integrating with bulk RNA sequencing recovers sequencing depth the spatial assay alone may not reach at every tissue position.
A spatial expression pattern is a correlation until it is functionally tested. The BCL2L1 example above shows the natural next step: genetic or pharmacological perturbation of the gene identified spatially, to confirm it is not merely associated with resistance but causally involved.
Clinical relevance requires clinical data. A gene expression program is a research finding until it is checked against an actual treatment outcome, which is precisely what elevated the hepatocellular carcinoma finding from a descriptive result to genuine mechanism-of-action evidence.

How tumor heterogeneity itself, beyond the single-patient example described above, complicates a spatial transcriptomics program built around a single expected gene expression pattern is developed in full in Tumor Heterogeneity and Spatial Clonal Architecture.

For the broader tumor microenvironment argument this spoke builds on, see Spatial Biology in Oncology: Decoding the Tumor Microenvironment, and for where spatial transcriptomics 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.