Oncology · Medical Literature
Spatial architecture contributes to failure of bulk biomarker-guided neoadjuvant immunotherapy selection in bladder cancer: The DUTRENEO study.
Abstract
Predictive biomarkers for immune checkpoint inhibitors (ICIs) are largely identified retrospectively, but their prospective clinical utility remains unproven. Here, we report DUTRENEO, a prospective randomized phase 2 trial testing whether a retrospectively validated 18-gene bulk tumor inflammation signature (TIS) can guide neoadjuvant ICI therapy in muscle-invasive bladder cancer. The trial does not meet its primary endpoint, and this demonstrates that bulk gene-expression stratification does not sufficiently enrich for responders. To define the biology underlying this failure, we generate single-cell spatial transcriptomic profiles of 377 genes in ∼5.4 million cells across large tissue areas. Response is governed by spatial architectures invisible to bulk assays, including CD8+ T cell proximity to cancer cells, localized checkpoint co-expression within epithelial cancer-rich neighborhoods, and fibroblast-rich immune-excluded communities in non-responders. We provide a quantitative framework showing that ≥77 genes and ≥3-mm tissue diameter regions preserve predictive spatial signal at scalable throughput. The registration details of the trial are EudraCT 2017-002246-68.
Clinical relevance
Can a prospectively applied, retrospectively validated 18-gene bulk Tumor Inflammation Signature (TIS) reliably select patients with muscle-invasive bladder cancer (MIBC) who will benefit from neoadjuvant dual immune checkpoint blockade (durvalumab + tremelimumab) over cisplatin-based neoadjuvant chemotherapy? __SNAPSHOT__ {"studyDesign":"Prospective, biomarker-driven randomized phase 2 trial with embedded single-cell spatial transcriptomic correlative analysis","setting":"Multicenter; Spain (EudraCT 2017-002246-6; NCT03472274); institutional academic hospitals","sampleSize":"n=73 patients treated in total. TIS-high ('hot') tumors randomized 1:1 to ICI (durvalumab + tremelimumab) or neoadjuvant chemotherapy (NAC); TIS-low ('cold') tumors assigned to NAC. Spatial transcriptomics performed on ~5.4 million cells profiling 377 genes across large tissue areas.","population":"Adults with localized muscle-invasive urothelial bladder cancer (MIBC), cisplatin-eligible, planned for radical cystectomy, stratified by TIS score into 'hot' (top two-thirds of score distribution) or 'cold' (lower third) prior to randomization","intervention":"TIS-high arm: neoadjuvant durvalumab (anti-PD-L1) + tremelimumab (anti-CTLA-4); TIS-low arm: neoadjuvant cisplatin-based chemotherapy (NAC)","comparator":"TIS-high arm comparator: cisplatin-based NAC (MVAC, gemcitabine/cisplatin, or gemcitabine/cisplatin/paclitaxel). TIS-low arm: NAC (non-randomized, assigned by biomarker score)","primaryOutcome":"Pathologic complete response (pCR) rate at radical cystectomy","followUp":"Not reported (survival data reported but duration of follow-up not specified in the abstract)","fundingSource":"Mixed: unrestricted educational grant from AstraZeneca (industry); NanoString provided reagents for TIS scoring; additional support from Fundación Científica AECC, la Caixa Foundation, and Spanish Ministry of Science (government/non-profit)"}
Key findings
The trial did not meet its primary endpoint. Pathologic complete response (pCR) rates were 38.5% in TIS-high patients treated with ICI (durvalumab + tremelimumab), 30.0% in TIS-high patients treated with NAC, and 55.0% in TIS-low ('cold') patients treated with NAC (p = 0.349), indicating no statistically significant difference across groups and no enrichment of ICI responders by TIS stratification. No statistically significant differences were observed in progression-free survival across treatment groups (p = 0.7; hot NAC reference: hot ICI HR = 1.50, 95% CI 0.56–4.04; cold NAC HR = 1.43, 95% CI 0.48–4.26) or in overall survival (p > 0.6; hot ICI HR = 2.16, 95% CI 0.44–10.76; cold NAC HR = 1.45, 95% CI 0.20–10.59). TIS score did not significantly differentiate responders from non-responders in any treatment arm (p = 0.097 for a trend toward higher TIS in ICI responders, not statistically significant). Post-hoc analysis showed that higher TIS thresholds would have increased the proportion of ICI responders but at the cost of excluding the majority of patients. NNT is not calculable given the non-significant primary endpoint and the directionally unfavorable pCR rates. Regarding safety, grade 3–4 TRAEs (neutropenia, thrombocytopenia, asthenia, mucosal inflammation) were noted at a higher incidence in the NAC arm versus ICI; surgical delays attributable to durvalumab + tremelimumab were reported at approximately 13% (cited from external systematic review). Single-cell spatial transcriptomic analysis (377 genes, ~5.4 million cells) revealed that ICI response was governed by spatial architecture features invisible to bulk assays: proximity of CD8+ T cells to cancer cells, localized checkpoint co-expression within epithelial cancer-rich neighborhoods, and fibroblast-rich immune-excluded tumor microenvironments in non-responders. A quantitative framework derived from this dataset suggests that a minimum of ≥77 genes and ≥3 mm tissue diameter regions are required to preserve predictive spatial signal at scalable throughput.
Practice implications
DUTRENEO establishes — prospectively and for the first time in a randomized design — that the 18-gene bulk TIS, as applied in this trial, does not adequately enrich MIBC patients for response to neoadjuvant dual ICI over standard cisplatin-based chemotherapy. This is a critical negative finding because TIS had been retrospectively validated across multiple tumor types. The pCR rate in TIS-selected ICI recipients (38.5%) was not superior to unselected historical ICI cohorts, confirming that bulk stratification failed to meaningfully concentrate responders. Notably, TIS-low ('cold') patients treated with chemotherapy achieved the numerically highest pCR rate (55%), underscoring that TIS-based re-direction away from NAC may have disadvantaged some patients. The mechanistic correlative analyses provide a compelling biological explanation: response to ICI is determined by spatially organized features of the tumor immune microenvironment — including CD8+ T cell-to-cancer cell proximity and fibroblast-mediated immune exclusion — that are fundamentally undetectable by bulk gene-expression assays. This study does not establish which patients should receive neoadjuvant ICI in MIBC; it establishes what does NOT work as a selection tool. Findings apply specifically to cisplatin-eligible MIBC patients and may not extend to cisplatin-ineligible populations or other ICI regimens. The spatial framework (≥77 genes, ≥3 mm tissue sampling) is hypothesis-generating and requires prospective validation before clinical application. IMPORTANT: Only the abstract was available for detailed statistical review; additional outcomes, subgroup analyses, and full safety data may be present in the complete publication that could affect interpretation. __LIMITATIONS__ Small sample size (n=73 total treated) substantially limits statistical power for PFS and OS endpoints and increases the risk of type II error for the primary pCR endpoint.,Open-label design without blinding of treatment assignment introduces potential performance and assessment bias.,Approximately 17% of patients were excluded from the primary analysis (noted in systematic review), representing a serious risk of bias per ROB-2 assessment.,The TIS score threshold (top two-thirds vs. lower third) was selected from retrospective data; the post-hoc analysis showing higher thresholds improve prediction raises concerns about the a priori threshold calibration.,The trial enrolled only cisplatin-eligible patients with pure or predominantly urothelial histology, limiting generalizability to cisplatin-ineligible patients, variant histologies, or non-urothelial MIBC.,All patients in the ICI arm had pure urothelial tumors, while mixed histology patients (who had numerically higher pCR rates) were concentrated in the chemotherapy arm, introducing a potential confound.,PFS and OS data are immature given the wide and crossing confidence intervals; the trial was not powered for survival endpoints.,The spatial transcriptomic findings (CD8+ proximity, fibroblast exclusion communities) are correlative and hypothesis-generating; no prospective validation of spatial biomarkers was performed.,Partial industry funding (AstraZeneca unrestricted grant; NanoString reagents) introduces potential sponsorship bias, though the trial tested a negative outcome for the sponsored modality.,Durvalumab + tremelimumab is not an approved standard-of-care neoadjuvant regimen in MIBC, potentially limiting the translational relevance of the specific agent pair to clinical practice.,Summary is based on the published abstract and supplementary data available; full manuscript details (complete adverse event tables, subgroup analyses, translational methodology) were not fully accessible for review. __BOTTOM_LINE_CATEGORY__ potentially_practice_informing __BOTTOM_LINE__ DUTRENEO is a pivotal negative prospective randomized trial demonstrating that bulk 18-gene TIS-based selection fails to enrich MIBC patients for neoadjuvant ICI benefit over cisplatin chemotherapy, with no significant difference in pCR (38.5% vs. 30.0% vs. 55.0%; p = 0.349) or survival across groups. The accompanying spatial transcriptomic analysis provides strong mechanistic evidence that ICI response is determined by spatially organized immune architecture — particularly CD8+ T cell proximity to tumor cells and fibroblast-mediated immune exclusion — features undetectable by bulk transcriptomics. Clinicians should not use bulk TIS or equivalent bulk inflammation signatures alone to select MIBC patients for neoadjuvant immunotherapy; spatially resolved biomarker strategies require prospective validation before guiding therapeutic decisions.