Blog
Apr 1, 2025 / Oncology

Harnessing spatial biology to better understand drug response and optimize clinical trial design

Josh Azevedo

Evidence-based medicine. There’s never been a time in our history where three little words have held such a big promise for cancer treatment. As more and more promising biomarkers are being discovered, identifying and selecting the right ones for drug development, patient stratification, and monitoring treatment responses are a critical next step forward for optimizing patient outcomes. However, in spite of this potential, these biomarkers are often not tested, or applied, in newer clinical trials to examine their real-world utility in stratifying patients and guiding therapeutic approaches.

A general workflow highlighting how patients with different molecular and/or genetic phenotypes can be stratified into different treatment groups based on biomarkers.
A general workflow highlighting how patients with different molecular and/or genetic phenotypes can be stratified into different treatment groups based on biomarkers.

Particularly, biomarkers identified from retrospective studies are rarely applied prospectively in clinical trials to see whether they can enrich treatment groups for responders. One such biomarker—a tumor inflammation signature (TIS) derived from 18 cancer- and inflammation-associated genes—has shown particular promise as a prognostic marker across multiple tumor types, but had not been used in a prospective study.

Grande et al.—the researchers highlighted in this blog—sought to test the utility of TIS scoring to guide treatment in muscle-invasive bladder cancer (MIBC; 1) in the DUTRENEO trial (EudraCT: 2017-002246-6). While the prospective profiling of patients with TIS was unsuccessful, they were able to perform a retrospective analysis with 10x Genomics’ sequencing- and imaging-based spatial transcriptomics to identify novel spatial biomarkers. These promising spatial biomarkers were better associated with patient responses to neoadjuvant immune checkpoint and chemotherapy-based treatments, as well as treatment non-responders.

Testing the utility of TIS scoring for patient stratification and predicting treatment outcomes

The researchers noted they wanted to specifically address the prospective value of TIS scoring in MIBC patients to guide treatment strategies for the clinical trial. Their initial approach was to first divide patients into groups based on whether they had a high or low TIS score (i.e., whether they had “hot” or “cold” tumors, respectively). 

All 21 patients assessed with cold tumors were treated with neoadjuvant chemotherapy (NAC). Of the 46 patients assessed with hot tumors included in the study, 22 were treated with NAC and 24 were treated with immune checkpoint inhibitors (ICI). 

In the cold tumor group, 55% of patients had a complete response (i.e., all signs of cancer disappeared). In the hot tumor group, 39% and 30% of patients had a complete response to ICI and NAC, respectively.

Notably, there were no significant differences in response rate or overall survival between treatment groups, and the authors noted these response rates were similar to other studies. Combined, this suggested TIS scoring alone didn’t tilt the balance towards pairing MIBC patients with better treatments—so how could the researchers improve on this method?

Identifying response-associated markers with sequencing-based spatial transcriptomics

The team first turned to bulk RNA-seq to characterize transcriptional signatures associated with responders in both ICI- and NAC-treated patients. Next, using sequencing-based Visium Spatial Gene Expression, they identified the spatial localization of response-associated signatures and compared their average distance from signatures associated with relevant (e.g., stromal and/or immune) cell types.

They found that positive responders to ICI treatment had a statistically significantly shorter distance between these stromal and response signature areas compared to non-responders. Notably, there was no significant change between immune and response signature areas, nor was any change in distance observed in NAC-treated patients regardless of response.

The researchers again used the Visium platform to analyze intratumoral heterogeneity (ITH), a phenomenon associated with therapeutic resistance. Focusing first on genetic ITH, they saw that NAC responders in both hot and cold tumor groups exhibited greater copy number variation heterogeneity than non-responders.

The group next focused on the variability of intratumoral cancer cell phenotypes (phenotypic ITH). Notably, phenotypic ITH alone was not associated with treatment responses in any group. However, the researchers found that high phenotypic ITH paired with low genetic ITH was associated with (and seen exclusively in) NAC-treated non-responders. Combined, these results suggest a spatial biomarker (distance between stromal area and response signatures) is associated with positive responses to NAC chemotherapy, while high-genetic and low-phenotypic ITH is more likely to exhibit chemotherapeutic resistance.

Adding “where” to the biomarker toolkit with Xenium single cell spatial imaging

Seeking more information for factors that were associated with ICI response, the researchers again turned to their bulk RNA-seq data. The authors found that ICI responders showed significantly increased expression of both the PDL1 and CTLA4 genes versus non-responders, suggesting that bulk RNA signatures could distinguish response—but they wanted to do better.

Using imaging-based spatial transcriptomics with the Xenium platform, the team examined what cell types were closest to epithelial cells in tumors. Consistent with their earlier experiments using the Visium platform, patients who responded to ICI demonstrated a shorter distance between immune cells and epithelial cells than non-responders.

Interestingly, overall cell composition was largely unchanged between the groups, consistent with the notion that immune cell spatial localization—not immune cell abundance—is associated with ICI success.

The group also took advantage of the Xenium platform to identify cell types that expressed PDL1 and CTLA4, two of their binding partners (PD1 and CD86), and two immune checkpoint genes (LAG3 and TIM3). They next determined the relative contribution of each cell type to their overall gene expression and found that patients who responded to ICI had higher expression of these genes in relevant cell types. Importantly, the team noted that, while bulk RNA-seq of this six-gene signature was associated with ICI response, the expression of these genes within specific, relevant cell types provided better response predictions.

Finally, given cellular and tumor microenvironments can influence cell interactions, the researchers examined whether the expression of pairs of immune checkpoint genes and their binding partners varies depending on the local microenvironment. In ICI-treated patients, the authors noted differential cell type contributions to overall gene expression. Responders tended to have more “focused” expression in specific communities while non-responders tended to have broader expression of these genes, underscoring the spatial context of immune activity that may underlie varying treatment responses.

Looking forward: Turning retrospectives into prospectives

While TIS scoring alone was not an effective predictor of NAC or ICI treatment response in MIBC patients, this study demonstrates how critical iterative analysis of biomarkers can be. The authors’ retrospective analyses revealed multiple avenues to further refine biomarkers of NAC and ICI response or resistance.

Finally, this work exemplifies the increasing recognition that biomarkers are not just a “what” but also a “where,” and understanding where the cells are positioned in the tumor—not just what they’re expressing—can play a large role in biomarker discovery and refinement for future testing. It also serves as evidence that novel spatial transcriptomics approaches can unlock a new dimension in biomarker analysis and development, and that these tools will be increasingly important for researchers to improve their clinical trial design, execution, and biomarker discovery.

Looking for more ways spatial transcriptomics can influence drug development for clinical trials and beyond? Check out our new resource hub curated specifically to help pharma researchers discover the impact these technologies can have across the entire drug development pipeline. You can also view the Visium and Xenium platform pages for a deeper dive into these groundbreaking technologies.

References:

  1. Grande E, et al. Spatial biomarkers of response to neoadjuvant therapy in muscle-invasive bladder cancer: the DUTRENEO trial. medRxiv (2025). doi: 10.1101/2025.02.07.25321742

About the author:

Josh earned a PhD in Neuroscience from the University of Michigan. His thesis work focused on two separate projects: one on disruptions in the microRNA regulatory network in human mood and anxiety disorders, and the second on the cellular basis of transcriptional dysregulation in an animal model of depression. After a postdoc centered on microglial alterations in Alzheimer’s disease, he began his career as a technical writer in the biotech industry. A longtime advocate of better scientific communication, he now uses his skills to make single cell and spatial tools engaging and intuitive, with a particular emphasis on technological comparisons.
Josh earned a PhD in Neuroscience from the University of Michigan. His thesis work focused on two separate projects: one on disruptions in the microRNA regulatory network in human mood and anxiety disorders, and the second on the cellular basis of transcriptional dysregulation in an animal model of depression. After a postdoc centered on microglial alterations in Alzheimer’s disease, he began his career as a technical writer in the biotech industry. A longtime advocate of better scientific communication, he now uses his skills to make single cell and spatial tools engaging and intuitive, with a particular emphasis on technological comparisons.