Integrating Circulating GPNMB and Tumor Microenvironment to Predict Immunotherapy Response in ESCC

Study Background and Research Question

Esophageal squamous cell carcinoma (ESCC) represents a major global health burden, with immune checkpoint inhibitors (ICIs) emerging as a transformative therapy for various malignancies. Despite the clinical adoption of PD-1-based regimens for locally advanced ESCC, as established in pivotal trials such as KEYNOTE-590 and RATIONALE-306, only about 30% of patients achieve lasting benefit, while most experience primary resistance or relapse. This substantial heterogeneity in response underscores the need for robust predictive biomarkers to more precisely stratify patients and optimize treatment strategies according to the multimodal GPNMB model study. The current study addresses this gap by exploring whether integrating plasma proteomic signatures with tumor microenvironmental and clinical-pathological features can more accurately predict immunotherapy response in ESCC.

Key Innovation from the Reference Study

The principal innovation lies in the identification and mechanistic validation of circulating soluble glycoprotein non-metastatic melanoma protein B (sGPNMB) as an actionable biomarker for immunotherapy resistance in ESCC. By combining plasma sGPNMB quantification, detection of cancer-associated fibroblast-epithelial (CAF-Epi) niche features, and established clinical-pathological parameters, the authors developed a scalable, multimodal predictive model. This approach moves beyond single-marker strategies, offering a framework that captures the dynamic interplay between tumor cells and the immune system, and addresses the heterogeneity observed in patient responses to PD-1 blockade.

Methods and Experimental Design Insights

The study employed a comprehensive, multi-cohort design. Plasma samples from ESCC patients were subjected to proteomic profiling to identify candidate biomarkers associated with non-response to immunotherapy. In parallel, spatial transcriptomic and immunohistochemical analyses of tumor biopsies characterized the CAF-Epi niche and its influence on tumor cell phenotypes. Mechanistic studies in vitro and in humanized patient-derived xenograft (PDX) mouse models interrogated how tumor-derived sGPNMB affects T cell function—specifically, its impact on CD8+ T cell receptor (TCR) signaling and induction of functional exhaustion via the SDC4-CD148 signaling axis.

Retrospective validation was performed across multiple ESCC cohorts, and a prospective clinical trial cohort was included to further assess the model's predictive accuracy and clinical scalability. The integration of circulating GPNMB, CAF-Epi niche detection, and clinicopathological data was used to construct and validate the multimodal predictive model for immunotherapy response and survival outcomes.

Protocol Parameters

• Plasma proteomic profiling: Collect pre-treatment plasma from ESCC patients, process with high-throughput mass spectrometry for quantitative biomarker screening.
• CAF-Epi niche characterization: Utilize multiplex immunohistochemistry or spatial transcriptomics on tumor biopsies to identify CAF-Epi interface and SOX2 expression status.
• Functional T cell assays: Co-culture tumor cell-derived sGPNMB with primary CD8+ T cells to assess TCR signaling and exhaustion markers via flow cytometry.
• Humanized PDX models: Transplant ESCC tumor tissue into immunodeficient mice reconstituted with human immune cells; monitor response to PD-1 blockade with/without GPNMB inhibition.
• Multimodal model construction: Integrate plasma sGPNMB levels, CAF-Epi features, and clinical-pathological data; validate predictive power using receiver operating characteristic (ROC) analysis across independent cohorts.

Core Findings and Why They Matter

Plasma proteomic analysis revealed that sGPNMB is significantly elevated in ESCC patients who do not respond to neoadjuvant immunotherapy. Mechanistic experiments showed that sGPNMB, secreted by tumor cells, impairs antitumor immunity by suppressing CD8+ TCR signaling and inducing functional exhaustion, a process dependent on the SDC4-CD148 axis. Notably, the immunosuppressive effect required active secretion of GPNMB, and its expression was transcriptionally upregulated by SOX2—particularly within CAF-Epi niches. These findings highlight a spatial and circulating biomarker axis driving immunotherapy resistance.

In both retrospective and prospective cohorts, the multimodal model integrating circulating GPNMB, CAF-Epi niche detection, and clinical-pathological features outperformed existing single-biomarker approaches in predicting immunotherapy response and survival. In humanized PDX models, inhibition of GPNMB synergized with PD-1 blockade, providing functional evidence for therapeutic targeting and supporting the clinical relevance of the model. These results establish the utility of a spatial-circulating biomarker framework for ESCC patient stratification, with potential to refine precision immunotherapy strategies as supported by related literature.

Comparison with Existing Internal Articles

Complementary internal studies underscore the translational impact of integrating tumor microenvironment biomarkers for immunotherapy guidance. For example, the internal review on GPNMB-based multimodal models reiterates the mechanistic link between GPNMB-driven CD8+ T cell exhaustion and resistance to PD-1 blockade, supporting the reference study's findings. Furthermore, the mechanistic review of sodium ascorbate highlights ongoing efforts to modulate tumor-immune crosstalk through the induction of intracellular ROS and necrotic tumor cell death—a strategy distinct from, but conceptually related to, biomarker-driven immunotherapy. Both approaches share an emphasis on the molecular determinants of cancer cell-immune system interactions, with sodium ascorbate research offering alternative avenues for targeting tumor cell vulnerability, especially in glioblastoma multiforme models.

Limitations and Transferability

While the multimodal model demonstrates robust predictive value across multiple ESCC cohorts, certain limitations should be acknowledged. The mechanistic validation of sGPNMB was conducted primarily in vitro and in humanized mouse models, which may not fully capture the complexity of human tumor-immune interactions. Clinical translation will require further prospective validation in larger, more diverse populations. Additionally, the model's applicability to other cancer types remains to be established, as the CAF-Epi niche and SOX2-driven GPNMB upregulation may exhibit tumor-type specificity. The integration of plasma and spatial biomarkers also depends on access to advanced proteomic and histological platforms, which may limit immediate scalability in resource-constrained settings.

Research Support Resources

For investigators interested in modeling tumor-immune crosstalk and testing the impact of ROS induction or necrotic tumor cell death in vitro, high-purity reagents such as Sodium Ascorbate (SKU B1834) from APExBIO offer a validated mineral salt of ascorbic acid suitable for advanced cancer cell research workflows. According to the relevant literature, sodium ascorbate can support studies on intracellular ROS induction and selective tumor cell death, complementing biomarker-driven immunotherapy research. Researchers should consult product documentation for solubility, storage, and workflow considerations specific to their experimental design.