Peripheral nerves are increasingly recognized as active components of the tumor microenvironment (TME). Their role in non‑muscle‑invasive bladder cancer (NMIBC), and specifically in response to intravesical Bacillus Calmette‑Guérin (BCG) therapy, has been insufficiently characterized. This study quantified nerve populations in the tumor‑adjacent stroma and examined associations between stromal innervation and clinical outcomes, nerve phenotype, spatial organization, and underlying transcriptomic programs.
The authors used multispectral imaging to characterize autonomic and sensory nerve populations within the tumor‑adjacent stroma of treatment‑naive, high‑risk NMIBC. Analyses included 241 tumor regions from 141 patients. Unsupervised clustering methods were applied to resolve nerve phenotypic groups. Matched bulk transcriptomic profiling was integrated with imaging data, and findings were compared against an independent single‑cell atlas to nominate potential cellular mediators of neural–tumor interactions. The study is reported as a preprint and includes supplementary materials and code repositories referenced by the authors.
The abundance of stromal tumor‑associated nerve (STaN) varied substantially across patients. Higher global STaN abundance associated with more aggressive clinicopathological features and with clinically relevant outcomes: BCG failure and disease progression. Thus, overall innervation level in the tumor‑adjacent stroma emerged as a prognostic characteristic in this NMIBC cohort.
Unsupervised clustering of nerve features resolved distinct STaN phenotypic groups with differential clinical and spatial associations. Among these, a noradrenergic STaN population and a neurochemically heterogeneous STaN population exhibited the strongest correlations with adverse clinical outcomes. These phenotype‑level distinctions indicate that nerve subtype composition, not just quantity, is relevant to tumor behavior.
A mixed nerve population characterized by high expression of the synaptic vesicle protein synaptophysin displayed pronounced spatial remodeling. This synaptophysin‑high group showed a prominent association with BCG failure despite having limited prognostic value when assessed by abundance alone. In other words, spatial reorganization of these synaptophysin‑expressing nerves—rather than their absolute numbers—was linked to treatment outcome.
To explore potential pathways of neural–microenvironment communication, the authors integrated imaging results with an independent single‑cell atlas. This analysis nominated cancer‑associated fibroblasts (CAFs) and mast cells as candidate cellular intermediaries that could participate in or mediate interactions between nerves and other TME components. The source reports these cell types as plausible conduits of neural signaling within the bladder tumor microenvironment.
Matched bulk transcriptomic profiling revealed coordinated molecular programs that differentiated STaN‑high from STaN‑low tumors. STaN‑high tumors showed enrichment of neuronal signatures alongside extracellular matrix (ECM), lipid metabolism, and immune‑related programs. In contrast, STaN‑low tumors preferentially enriched proliferative pathways and hypoxic stress‑response programs. These coordinated transcriptional patterns suggest distinct microenvironmental states associated with innervation level and nerve phenotype.
Leveraging the transcriptional differences, the authors developed a transcriptomic classifier to distinguish STaN‑high from STaN‑low tumors. When applied to an independent test cohort, the classifier achieved an area under the receiver operating characteristic curve (AUC) of 0.87 as reported in the source. The classifier also stratified clinical outcomes in additional patient cohorts according to the authors’ analyses.
The study establishes tumor innervation as a heterogeneous and clinically relevant feature of the NMIBC microenvironment. Three related but distinct axes capture clinical behavior: innervation abundance, nerve phenotype, and spatial organization. Each axis associated with different aspects of prognosis and treatment response (notably BCG failure and progression). As the report is a preprint, findings have not been peer reviewed. Detailed methodological parameters, full statistical results, cohort demographics, and external validation cohort characteristics are reported in the source supplementary materials; if specific details are required, the source should be consulted directly.
The authors provide supplementary material and links to code repositories (GitHub) referenced in the source. Competing interests declared: one author is cofounder and Chief Science Officer of OncoRx Insights; the other authors declared no competing interests. The work was supported by several funders listed in the source. Readers should note this is a preprint and interpret results in that context.