Stage-dependent prognostic value of programmed death ligand-1 expression in gastric cancer: Implications for personalized immunotherapy

Abstract

The integration of immune checkpoint inhibitors has fundamentally altered treatment strategies for advanced gastric and gastroesophageal junction cancer. However, the prognostic and predictive utility of programmed death ligand-1 (PD-L1) expression remains a subject of intense debate due to its inherent biological complexity. While large-scale meta-analyses generally suggest that high PD-L1 expression correlates with poorer overall survival, clinical outcomes in trials of PD-1 blockade show significant heterogeneity. This review reappraises PD-L1 not as a static marker, but as a dynamic reflection of the tumor microenvironment. We discuss the biological framework of the three major immune phenotypes - immune-inflamed, immune-excluded, and immune-desert - as the basis for understanding clinical response. Furthermore, we examine the influence of molecular subtypes, such as Epstein-Barr virus and microsatellite instability, and evaluate evidence from pivotal phase III trials in both metastatic and perioperative settings. By moving beyond simplified scoring systems and acknowledging the stage-dependent and subtype-specific biological realities, we propose a more personalized approach to immunotherapy in gastric cancer.

Key Words: Gastric cancer; Programmed death ligand-1; Immune checkpoint inhibitors; Tumor microenvironment; Biomarkers; Individualized medicine

Core Tip: This review reappraises programmed death ligand-1 expression in gastric cancer and gastroesophageal junction cancer as a dynamic reflection of the tumor microenvironment. We examine the stage-dependent prognostic shift of programmed death ligand-1 and immune phenotypes (inflamed, excluded, and desert). By integrating phase III trial data with molecular subtypes like Epstein-Barr virus and microsatellite instability, we highlight limitations of simplified scoring. We propose personalized immunotherapy, driven by spatial analysis and dynamic monitoring, to optimize outcomes across the complex biological landscape of gastric cancer.

INTRODUCTION

Gastric cancer (GC) and gastroesophageal junction cancer (GEJC) represent a formidable global health burden, consistently ranking as a leading cause of cancer-related mortality worldwide[1]. Despite the implementation of multimodal therapeutic strategies combining surgery with perioperative or adjuvant regimens, the five-year survival rates for patients with locally advanced disease remain stubbornly suboptimal[2]. Historically, the management of these malignancies was largely dictated by pathological staging and anatomical location. However, the paradigm has shifted toward a more sophisticated approach rooted in rigorous molecular characterization, including the assessment of human epidermal growth factor receptor 2 (HER2) and mismatch repair status[3,4]. Even with these advancements, systemic recurrence remains a frequent and devastating outcome, underscoring the urgent need for more effective systemic interventions[5,6].

The clinical application of targeted therapies has provided some progress in precision oncology, yet the intrinsic biological complexity and clonal evolution of gastric tumors often limit the long-term efficacy of such strategies[7,8]. In the setting of refractory disease, it has become increasingly evident that the tumor microenvironment (TME) exerts a profound influence on therapeutic response and overall prognosis[9]. We discuss the significant findings by Zhang et al[10] regarding the stage-dependent prognostic impact of programmed death ligand-1 (PD-L1) expression in gastric or gastroesophageal junction cancer. The emergence of immunotherapy, particularly agents targeting the PD-1 and PD-L1 axis, has fundamentally revolutionized the treatment landscape. While initial successes in the metastatic setting were robust, the transition of immune checkpoint inhibitors (ICIs) into earlier stages of the disease has proven to be biologically and clinically complex.

Recent evidence has highlighted a provocative stage-dependent shift in the prognostic value of PD-L1 expression. Specifically, high PD-L1 expression appears to function as a negative prognostic factor in stage III disease, whereas this association is notably absent or even reversed in stage II disease[10]. This phenomenon is not unique to the gastrointestinal tract; similar observations in non-small cell lung cancer suggest that PD-L1 serves as a dynamic marker of the host-tumor interface[11,12]. In the early stages of malignancy, PD-L1 expression often reflects a state of “adaptive immune resistance”, where the marker is upregulated in response to active infiltration by tumor-infiltrating lymphocytes[13,14]. Conversely, in advanced or metastatic stages, persistent PD-L1 expression typically signals established immune evasion and terminal T-cell exhaustion[15].

BIOLOGICAL AND CLINICAL HETEROGENEITY IN THE TME

The biological significance of PD-L1 expression in GC/GEJC is characterized by significant heterogeneity, which complicates its utility as a standalone biomarker. Meta-analyses involving thousands of patients have generally concluded that high PD-L1 expression correlates with poorer overall survival, yet the high degree of inter-study variability suggests that clinical stage and molecular subtype are critical confounding variables[16]. The prognostic significance of PD-L1 is intimately tied to the anatomical extent of the disease and the specific compartment - tumor cells vs immune cells - where the expression occurs[17,18]. Furthermore, the temporal stability of PD-L1 is questionable, as post-operative changes in PD-L1 status have been observed, potentially offering additional prognostic insights during the surveillance period[19]. In the early-stage setting, some studies suggest that high PD-L1 expression identifies a subgroup of patients with a more favorable clinicopathological profile, characterized by higher lymphocytic density and a more robust anti-tumor response[20].

To conceptualize this complexity, it is essential to categorize the TME based on its immune architecture. The TME can be categorized into three distinct immune phenotypes - immune-inflamed, immune-excluded, and immune-desert - based on the spatial distribution of CD8+ T cells and the mechanisms of immune evasion[21,22]. These biological distinctions are visualized in Figure 1. The “immune-inflamed” phenotype is characterized by a high density of CD8+ T cells and reactive PD-L1 expression, typically predicting a favorable response to ICIs. In contrast, the “immune-excluded” phenotype features immune cells that are relegated to the stroma, unable to penetrate the tumor nests due to physical or chemical barriers, while the “immune-desert” phenotype represents a complete failure of immune recognition[21,22].

Figure 1 Conceptual schematic of the three immune phenotypes in the gastric tumor microenvironment. A: Immune-inflamed: Characterized by high CD8+ T-cell infiltration and reactive programmed death-ligand 1 expression driven by interferon-gamma. This phenotype typically predicts favorable immune checkpoint inhibitor benefit; B: Immune-excluded: T cells are present in the stroma but are prevented from infiltrating tumor nests, often by a dense extracellular matrix; C: Immune-desert: Characterized by a lack of significant T-cell infiltration and low programmed death-ligand 1 expression, signaling primary resistance to immune checkpoint inhibitor monotherapy. TME: Tumor microenvironment; PD-L1: Programmed death-ligand 1; IFN-γ: Interferon-gamma; ICI: Immune checkpoint inhibitor; CD8+: Cluster of differentiation 8-positive.

Molecular subtypes further refine our understanding of PD-L1. Patients with Epstein-Barr virus-positive tumors or microsatellite instability-high status frequently exhibit constitutive PD-L1 expression and improved survival outcomes compared to those with chromosomal instability or genomically stable subtypes[23,24]. These biomarkers are now considered essential for accurate risk stratification in both Western and Asian populations[25-27]. From a mechanistic perspective, PD-L1 expression is not a static trait but is primarily regulated by the interferon-gamma pathway via the Janus kinases-signal transducers and activators of transcription signaling cascade[28-30]. Understanding these fundamental principles of immune checkpoint blockade is critical for the development of anti-PD-1 therapies[31-33]. The failure of immune surveillance is often the result of specific intrinsic signaling pathways, such as beta-catenin activation, which can actively exclude T cells from the TME[34]. In GC, the clinical response to PD-1 blockade is often predicated on the existence of a pre-existing interferon-gamma-related mRNA profile, which serves as a surrogate for an active immune TME[35,36]. Advanced spatial analysis through multiplex immunohistochemistry and immunofluorescence has confirmed that the physical proximity between PD-L1+ cells and tumor-infiltrating lymphocytes is more predictive of response than simple expression density[37,38]. However, persistent exposure to chronic inflammation eventually leads to terminal exhaustion, a state driven by alternative checkpoints such as Galectin-9 and Tim-3[39].

CLINICAL TRIAL EVIDENCE AND THE PERIOPERATIVE CHALLENGE

The clinical landscape of GC/GEJC has been rapidly reshaped by a series of pivotal phase III trials (Table 1). In the metastatic setting, the ATTRACTION-2 trial established nivolumab as a viable option for late-line patients[40]. Similarly, the KEYNOTE-059 trial demonstrated the efficacy of pembrolizumab monotherapy in PD-L1-positive disease[41]. As these agents moved into earlier lines, the results became more nuanced. The KEYNOTE-061 trial showed that while pembrolizumab did not improve survival over paclitaxel in the overall population, a clear benefit emerged in patients with high PD-L1 Combined Positive Score (CPS)[42]. This was followed by KEYNOTE-062, which suggested that pembrolizumab monotherapy was non-inferior to chemotherapy in the CPS ≥ 1 population, although the combination failed to meet superiority endpoints for overall survival[43].

Table 1 Representative phase III trials of immune checkpoint inhibitors in perioperative and advanced gastric cancer.

Trial name	Phase	Patient population	Treatment arms (intervention vs control)	PD-L1 cut-off (CPS)	Primary endpoint/result	Ref.
KEYNOTE-585	III	Perioperative (LAGC)	Pembro + Chemo vs Placebo + Chemo	Analysis ≥ 1, ≥ 10	pCR: Significantly improved; EFS/OS: Not met in the overall population	Shitara et al[52], 2024
MATTERHORN	III	Perioperative (LAGC)	Durva + FLOT vs Placebo + FLOT	All comers	pCR: Significantly improved (19% vs 7%); EFS: Significantly improved (HR 0.71); OS immature	Janjigian et al[53], 2025
CheckMate 649	III	Advanced (1st line)	Nivo + Chemo vs Chemotherapy alone	≥ 5, ≥ 1	OS and PFS: Significantly improved in CPS ≥ 5 (standard of care)	Janjigian et al[44], 2021
KEYNOTE-062	III	Advanced (1st line)	Pembro ± Chemo vs Chemotherapy alone	≥ 1, ≥ 10	OS: Pembro monotherapy non-inferior to chemo (CPS ≥ 1); combination failed superiority	Shitara et al[43], 2020
ATTRACTION-4	III	Advanced (1st line)	Nivo + Chemo vs Placebo + Chemo	All comers	PFS: Significantly improved; OS: Not met	Kang et al[46], 2022

CPS: Combined Positive Score; EFS: Event-free survival; LAGC: Locally advanced gastric cancer; OS: Overall survival; pCR: Pathological complete response; PFS: Progression-free survival; FLOT: Fluorouracil + leucovorin + oxaliplatin + docetaxel; Pembro: Pembrolizumab; Chemo: Chemotherapy; Durva: Durvalumab; Nivo: Nivolumab.

The gold standard for first-line treatment of advanced HER2-negative GC/GEJC was established by the CheckMate 649 trial. This study demonstrated that the addition of nivolumab to chemotherapy significantly improved both progression-free survival and overall survival in patients with a CPS ≥ 5[44]. Comprehensive meta-analyses have since validated the safety and efficacy of this combination[45]. Interestingly, the ATTRACTION-4 trial showed a significant improvement in progression-free survival regardless of PD-L1 status, but failed to show an overall survival benefit, likely due to subsequent lines of therapy[46,47]. Translating these successes to the perioperative setting has proven more difficult[48,49]. Biomarkers like CPS and tumor mutational burden are central to these efforts, yet require further refinement[50,51].

A notable setback occurred with the KEYNOTE-585 trial, which failed to show a significant improvement in event-free survival when pembrolizumab was added to perioperative chemotherapy in the overall population[52]. However, the MATTERHORN trial provided a more optimistic outlook, reporting a significant increase in pathological complete response with the addition of durvalumab to the FLOT regimen (fluorouracil, leucovorin, oxaliplatin, docetaxel)[53]. Despite these pathological complete response gains, the translation into a meaningful overall survival benefit remains to be seen[54]. In the HER2-positive subgroup, dual PD-1 and HER2 blockade has shown promising efficacy in early-phase trials[55,56]. The biological variability observed in these trials necessitates a critical re-evaluation. In stage II patients, intensifying neoadjuvant chemotherapy could be counterproductive by inducing systemic immunosuppression before the ICI can exert its effect[57,58]. Upfront surgery followed by adjuvant immunotherapy might leverage existing immune priming more effectively[59,60].

FUTURE PERSPECTIVES: INTEGRATED PRECISION MEDICINE

The significance of PD-L1 can no longer be evaluated in isolation. The future of individualized medicine in GC/GEJC will depend on moving beyond static scoring toward a holistic appraisal of the tumor’s biological state. This transition will be driven by the integration of spatial analysis, composite biomarkers, and dynamic monitoring into a single clinical framework[50,51]. Advanced techniques such as multiplex immunohistochemistry allow clinicians to evaluate the precise orientation of PD-L1-expressing cells relative to effector T cells[37,38]. Furthermore, sophisticated models combining PD-L1 with microsatellite instability-high status, Epstein-Barr virus positivity, and tumor mutational burden are essential[23,24]. The clinical paradigm must shift toward dynamic monitoring through liquid biopsies and ctDNA kinetics, allowing for real-time observation of therapeutic resistance and detection of minimal residual disease[59,60].

CONCLUSION

The role of PD-L1 in GC/GEJC is shifting from a simple predictive marker to a complex indicator of TME dynamics. Future research must prioritize stage-specific models that account for biological differences between early and advanced disease. By moving beyond uniform trial designs and embracing the stage-dependent biological reality, we can finally transition to a truly individualized immunotherapeutic strategy.