Muscle-invasive bladder cancer remains an aggressive disease where better patient selection for perioperative treatment is still needed. Cisplatin-based neoadjuvant chemotherapy has long been used for eligible patients, and immune checkpoint inhibitor-based strategies are increasingly being integrated and evaluated in perioperative treatment approaches.
One of the main challenges is identifying which patients are most likely to benefit from immunotherapy. Bulk gene-expression signatures have been proposed to define inflamed tumors, but their prospective clinical utility remains uncertain.
The article is titled “Spatial architecture contributes to failure of bulk biomarker-guided neoadjuvant immunotherapy selection in bladder cancer: The DUTRENEO study” and was published online in Cell Reports Medicine on June 23, 2026.
Authors: Enrique Grande, Mustafa Sibai, Daniela Grases, Elena Andrada, Oscar Reig, Marc Escobosa, Ainara Azueta, Daniel Castellano, Javier Puente, Jaime Martínez de Villarreal, Albert Font, Teresa Alonso-Gordoa, Raquel Benítez, Ane Moreno-Oya, Mario Álvarez-Maestro, Javier Burgos, M. Angel Climent, Mario Domínguez, Patricia Galván, Isabel Galante, Juan F. García, Elena Perez, Xavier García del Muro, Félix Guerrero-Ramos, Miriam Marqués, Pablo Maroto, Jesús M. Paramio, Alvaro Pinto, Aleix Prat, Núria Malats, Ignacio Durán, Eduard Porta-Pardo, and Francisco X. Real.
You can read more about Bladder Cancer: Symptoms ,Causes, Stages, Diagnosis and Treatment on OncoDaily.
Why DUTRENEO Was Conducted
The DUTRENEO study tested whether a retrospectively validated 18-gene tumor inflammation signature could improve treatment selection in muscle-invasive bladder cancer.
The signature was designed to classify tumors as TIS-high, or “hot,” and TIS-low, or “cold,” based on interferon-gamma-related immune activity. The clinical hypothesis was that TIS-high tumors could identify patients more likely to benefit from neoadjuvant immune checkpoint inhibition. At the same time, patients with TIS-low tumors were assigned to standard cisplatin-based neoadjuvant chemotherapy.
Study Design
DUTRENEO was a multicenter, open-label phase 2 study conducted across 11 Spanish tertiary hospitals. Eligible patients had muscle-invasive bladder cancer after transurethral resection of bladder tumor, were fit for cisplatin-based treatment, had ECOG performance status 0–1, had planned cystectomy, and had tumor tissue available for baseline TIS testing.
A total of 101 patients were screened. Among them, 73 were eligible and received treatment.
After TIS testing, patients with TIS-low, or cold, tumors received standard cisplatin-based neoadjuvant chemotherapy. Among patients with TIS-high, or hot, tumors, the first 6 received durvalumab plus tremelimumab in a safety run-in, and the remaining 46 were randomized to receive either durvalumab plus tremelimumab or standard cisplatin-based neoadjuvant chemotherapy.
Standard chemotherapy consisted of gemcitabine plus cisplatin or dose-dense MVAC, according to investigator choice. The immunotherapy arm received durvalumab 1500 mg plus tremelimumab 75 mg every 4 weeks for three cycles before surgery.
The primary endpoint was pathological complete response after cystectomy, defined as ypT0N0.
Primary Endpoint Was Not Met
The study did not meet its primary endpoint. Among patients with hot tumors, pathological complete response was observed in 10 of 26 patients treated with durvalumab plus tremelimumab, corresponding to 39%. In comparison, pathological complete response occurred in 6 of 20 patients with hot tumors treated with neoadjuvant chemotherapy, corresponding to 30%.
This difference was not statistically significant.
In the cold tumor group treated with neoadjuvant chemotherapy, 11 of 20 patients achieved pathological complete response, corresponding to 55%.
Downstaging was observed in 62% of patients treated with immunotherapy, 50% of patients with hot tumors treated with chemotherapy, and 60% of patients with cold tumors treated with chemotherapy. There were no significant differences in progression-free survival or overall survival across treatment groups.
These findings showed that the 18-gene TIS signature did not sufficiently enrich for patients who responded to neoadjuvant immune checkpoint inhibition.
Read more about Success Rate of Immunotherapy for Bladder Cancer on OncoDaily.
TIS Captured Inflammation, Not Response
The failure of TIS was not because the biomarker failed to detect inflammation. The analysis showed that TIS-high tumors had higher immune-cell fractions, including B cells, CD4 T cells, Tregs, and CD8 T cells. Hot tumors also had shorter distances between epithelial cells and nearby immune cells.
This means that the TIS signature did capture an inflamed tumor phenotype.
However, inflammation alone was not enough to predict clinical benefit. The study showed that the key difference between responders and non-responders was not simply how much immune activity was present, but where immune cells were located inside the tumor.
Spatial Transcriptomics and Biomarker Failure
To understand why bulk TIS testing failed, the study included multi-omics analysis of pretreatment tumor tissue. Tumor samples were analyzed using whole-exome sequencing, bulk RNA sequencing, and high-resolution spatial transcriptomics.
The spatial transcriptomics dataset included 377 genes across approximately 5.4 million single cells from 20 patients. This analysis showed that response to immunotherapy was linked to tumor spatial architecture. Responders had adaptive and innate immune cells located closer to epithelial cancer cells. In contrast, non-responders showed features of immune exclusion, including higher fibroblast abundance and fibroblast-rich immune-excluded communities.
In responders, CD8+ T cells, Tregs, dendritic cells, M1 macrophages, and NK cells were enriched in epithelial-rich communities. In non-responders, fibroblasts were more prominent within several immune-rich neighborhoods, supporting the role of stromal exclusion.
Spatial Checkpoint Signaling in Responders
The study also examined immune checkpoint expression and its spatial context. PD-L1 protein expression by immunohistochemistry was not associated with response. However, bulk RNA sequencing showed higher CD274 and CTLA4 transcript levels in responders.
Spatial analysis added another layer. In responders, checkpoint signaling was more localized to specific immune-tumor neighborhoods. The co-expression of CD86-CTLA4 around CD8+ T cells was higher in responders, and PDL1-PD1 co-expression around non-regulatory CD4+ T cells was also higher.
This suggested that response was associated with local checkpoint engagement within epithelial-rich communities, rather than diffuse immune activation across the tumor.
In non-responders, checkpoint expression appeared more broadly distributed across heterogeneous niches and was often located away from epithelial tumor communities. This pattern suggested immune activation without effective tumor contact.
CD8+ T Cell Proximity as a Key Feature
Among the spatial features evaluated, the median shortest distance between CD8+ T cells and cancer cells was the strongest correlate of response.
This CD8-to-cancer cell distance achieved an area under the curve of 0.90 for distinguishing responders from non-responders. Leave-one-out LASSO analysis selected CD8+ T cell proximity as the most stable feature, with a selection frequency of 90.9%.
The study also tested how spatial biomarkers could be translated into more scalable clinical assays.
The analysis suggested that a panel of approximately 77 genes and tissue regions of at least 3 mm in diameter could preserve most of the predictive spatial signal. In contrast, standard 1-mm tissue microarray cores may be too limited for validating spatial neighborhood biomarkers.
Read more about What Changed Immunotherapy in Bladder Cancer on OncoDaily.
Limitations
The spatial analyses were exploratory and associative. They were performed on pretreatment TURBT specimens and included a smaller spatial transcriptomics subset of 20 patients.
The study does not establish an immediately ready clinical biomarker. Instead, it provides a framework for understanding why bulk inflammation-based biomarkers may fail and how spatial biology could improve future biomarker development.
Takeaway
The DUTRENEO study shows that an 18-gene bulk tumor inflammation signature was not sufficient to guide neoadjuvant immunotherapy selection in muscle-invasive bladder cancer.
Although TIS identified inflamed tumors, it did not reliably identify patients who would respond to durvalumab plus tremelimumab. The spatial analysis showed why: response depended less on the amount of inflammation and more on the organization of the tumor microenvironment. CD8+ T cell proximity to cancer cells, localized checkpoint signaling in epithelial-rich neighborhoods, and fibroblast-poor niches were key spatial features associated with response.
DUTRENEO supports the development of spatially informed biomarkers that can capture where immune activity occurs inside the tumor, not only how much inflammation is present.