Bladder Cancer Atlas Reveals Drug Targets Switch On and Off as Tumors Mature

One of the largest single-cell atlases of muscle-invasive bladder cancer ever assembled is rewriting how scientists think about a promising class of drugs. By profiling 157,449 individual cells from 21 patients across three independent cohorts, researchers have shown that the molecular targets of antibody-drug conjugates—among the most celebrated weapons in modern oncology—are not fixed labels on tumor cells but dynamic traits that rise and fall as cancer cells travel along a differentiation trajectory. The finding, published in the Journal of Translational Medicine, carries immediate implications for how patients are selected for these expensive and sometimes toxic therapies.

Antibody-drug conjugates, or ADCs, are often described as guided missiles: a monoclonal antibody engineered to recognize a specific surface protein is chemically linked to a potent cytotoxic payload, allowing the drug to deliver its lethal cargo preferentially to cells bearing the target. Three such targets have already demonstrated clinical efficacy in urothelial carcinoma—NECTIN4, TACSTD2 (better known as TROP-2), and ERBB2 (HER2). Yet the authors of the new study noted a conspicuous gap: few investigations had systematically mapped how these targets behave at single-cell resolution in relation to tumor differentiation states, intrinsic biological programs, and the surrounding immune microenvironment. That gap matters, because a target that glows brightly in one region of a tumor may be nearly invisible in another, even within the same patient.

To build their atlas, the team integrated single-cell RNA sequencing data from three public cohorts, applying Harmony batch correction to merge datasets generated in different laboratories without erasing genuine biological differences. Quality-control metrics, benchmarking against an alternative integration method, and copy-number-variation-based validation of malignancy all supported the robustness of the merged dataset. Within the epithelial compartment, the researchers assigned consensus molecular subtype classifications at single-cell resolution, distinguishing basal, luminal, and neuroendocrine-like tumor populations. They then applied pseudotime trajectory inference using the Slingshot algorithm to order tumor cells along a basal-to-luminal differentiation axis, effectively reconstructing a maturation timeline from the static snapshot of sequencing data.

What emerged from that trajectory analysis was a striking divergence in how the three ADC targets are distributed along the differentiation axis. TROP-2, encoded by TACSTD2, was expressed broadly across the entire trajectory, making it the most uniform of the targets—a property reflected in its low inter-patient heterogeneity, with a Gini index of just 0.27 compared with 0.67 for NECTIN4, the most variable target. ERBB2/HER2, by contrast, was sharply localized to a narrow window along the trajectory, and that window coincided with elevated activity of ABC transporter genes—a family of drug efflux pumps with obvious relevance to chemotherapy resistance. NECTIN4 told a third story: it was virtually absent in early basal cells and rose progressively through the partial-epithelial-to-mesenchymal-transition zone, the transitional state that many tumor biologists associate with plasticity and metastatic potential.

These expression patterns were not merely descriptive. Patient-level analyses linked TROP-2 expression to elevated epithelial-mesenchymal transition program activity and HER2 expression to elevated ABC transporter activity, with modest associations to interferon-gamma signaling. The team validated their findings across species, showing that broad TROP-2 expression enriched in more differentiated cells is conserved in normal human and mouse urothelium, suggesting it reflects a fundamental physiological property of urothelial differentiation rather than a tumor-specific quirk. Bootstrap analyses resampling the 21 patients 200 times confirmed that the breadth-versus-focality differences between targets were statistically supported, though the peak positions of NECTIN4 and HER2 along pseudotime could not be cleanly separated from one another.

Perhaps the most provocative result came from the CellChat analysis of tumor-immune communication. The researchers, including a custom-curated ligand-receptor pair, found that NECTIN4-TIGIT ranked as the top predicted tumor-to-immune inhibitory checkpoint interaction, with tumor epithelial cells serving as the exclusive predicted source of NECTIN4-mediated TIGIT engagement. TIGIT is an inhibitory receptor on T cells and natural killer cells, and the finding implies that the very molecule drug developers exploit to shuttle toxins into tumors may simultaneously be helping the cancer suppress the immune cells that ADCs rely on for antibody-dependent cellular cytotoxicity. If confirmed experimentally, this dual role would provide a mechanistic rationale for combining NECTIN4-directed ADCs with TIGIT blockade.

The communication analysis also surfaced an unexpected player: the APP-CD74 axis, in which amyloid precursor protein—a molecule famous for its role in Alzheimer’s disease—emerged as a tumor-derived signal engaging CD74 on immune cells. The researchers detected 26 significant intercellular interactions along this axis, and validation in the TCGA-BLCA bulk transcriptomic cohort of 403 to 412 patients showed that tumors with high APP and low CD74 expression had the worst overall survival, with a hazard ratio of 2.00 (95 percent confidence interval 1.31 to 3.04, p equals 0.001). The authors were careful in their framing: the data suggest that disruption of this axis is associated with worse survival, not that it is causally responsible, but the association was strong enough to flag APP-CD74 as a novel prognostic communication axis worthy of functional follow-up.

The differentiation theme extended to patient outcomes in the TCGA-BLCA validation cohort. When the researchers distilled their pseudotime analysis into a differentiation signature and applied it to bulk tumor data, they found that early-differentiation tumors—those dominated by basal-like cells—carried significantly worse overall survival, with a hazard ratio of 1.58 (95 percent confidence interval 1.12 to 2.24, p equals 0.009), a result that held in multivariate models adjusting for molecular subtype, pathologic stage, and age. These early-differentiation tumors were enriched for epithelial-mesenchymal transition activity (Spearman rho of negative 0.398 with the differentiation index), interferon-gamma response (rho of negative 0.430), and proliferation signatures (rho of negative 0.502). Intriguingly, they also harbored elevated tumor mutational burden, hinting that despite their aggressive biology, these tumors might still respond to immune checkpoint inhibitors—a nuance that could matter for sequencing therapy in the clinic.

For the ADC field, the takeaway is that target expression breadth should inform biomarker-guided stratification. A target like TROP-2, expressed broadly across the differentiation spectrum and consistently across patients, offers a wide therapeutic window in principle; a target like HER2, confined to a narrow differentiation window and coupled to drug-efflux machinery, may require more careful patient selection and combination strategies to overcome resistance. NECTIN4’s dependence on differentiation state helps explain why it is the most heterogeneous target between patients, and why a biopsy from one region of a tumor might substantially overestimate or underestimate the true burden of target-positive cells. The pseudotime-derived differentiation signature, the authors propose, could serve as a biomarker to guide ADC therapy selection in urothelial carcinoma, complementing the molecular subtyping frameworks already in clinical discussion.

The study is not without caveats, several of which the authors acknowledge in their supplementary analyses. The atlas, while among the largest for this disease, derives from 21 patients, and pseudotime inference reconstructs a trajectory that cells may not literally traverse in vivo. The NECTIN4-TIGIT and APP-CD74 findings are computational predictions that will need wet-lab confirmation, and the published version remains subject to final editorial revision. Still, the work exemplifies a broader shift in oncology: from viewing tumors as uniform masses bearing static molecular flags, to reading them as ecosystems in which drug targets, immune evasion programs, and differentiation states are woven together—and in which the most effective therapies will be those designed with that dynamism in mind.

Subject of Research: Single-cell transcriptomic mapping of antibody-drug conjugate target dynamics and tumor-immune checkpoint interactions in muscle-invasive bladder cancer

Article Title: Single-cell transcriptomic profiling of muscle-invasive bladder cancer reveals differentiation-dependent ADC target dynamics and tumor-immune checkpoint interactions

Article References: Chang, P. M.-H., Yen, C.-C., Wu, W.-C., & Lai, J.-I. (2026). Single-cell transcriptomic profiling of muscle-invasive bladder cancer reveals differentiation-dependent ADC target dynamics and tumor-immune checkpoint interactions. Journal of Translational Medicine. https://doi.org/10.1186/s12967-026-08916-2

Image Credits: AI Generated

DOI: 10.1186/s12967-026-08916-2

Keywords: bladder cancer, single-cell RNA sequencing, antibody-drug conjugates, NECTIN4, TROP-2, HER2, TIGIT, APP-CD74 axis, pseudotime trajectory, tumor microenvironment, biomarkers, urothelial carcinoma