Immunotherapy for diffuse gastric cancer: challenges and new avenues.

Introduction
In 2022, the number of newly diagnosed gastric cancer (GC) cases worldwide was estimated to be 968350, and 659853 patients died of this malignancy, making GC the fifth most common cancer1. In recent decades, the overall incidence of GC has continued to decline owing to economic development, which has improved living conditions and reduced the prevalence of Helicobacter pylori (H. pylori) infection2. However, the temporal trends of incidence rates for GC vary between histological subtypes. Nearly sixty years ago, according to its histological and morphological characteristics, GC was divided into intestinal, diffuse and mixed types3,4. The success of prevention and screening strategies has led to a decrease in the incidence rate of intestinal GC (IGC), whereas the burden of diffuse GC (DGC) has remained relatively constant or has been increasing5–8. Therefore, along with its aggressiveness and poor prognosis, DGC may become a major challenge in the field of human health, necessitating more attention and research.
Since the end of the last century, the standardization of surgical procedures and the development of perioperative chemotherapy have dramatically improved the survival of patients with resectable localized GC. However, the outcome of patients with advanced GC remains dismal9,10. In recent years, immunotherapy, which functions to reactivate the immune system to kill tumor cells, has revolutionized the oncology landscape and changed the management of many cancer types11,12. In patients with GC, several clinical studies have tested the survival benefits of immunotherapy under various conditions, ranging from neoadjuvant therapy to the use of more than three lines. However, only a minority of these studies achieved positive results and supported the approval of several immune checkpoint inhibitors (ICIs) for GC treatment. Even in clinical trials meeting their primary endpoints, only a minority of the included patients responded to immunotherapy, not to mention the proportions of patients with a durable response. One of the major reasons is the heterogeneous histology of GCs, such as DGC, which is less responsive to many therapies for GC than IGC is. However, the responsiveness of DGC to immunotherapy has not been seriously considered in either clinical trials or routine practice.
Therefore, in this review, we aimed to discuss the current status of immunotherapy in DGC, focusing on ICIs. First, we briefly review the mechanisms of carcinogenesis and clinicopathological features of DGC. The possible mechanisms underlying the poor responsiveness of DGC to immunotherapy will be discussed. Next, we comprehensively summarize the available evidence on the efficacy of ICIs in DGC, mainly from subgroup analyses of phase III clinical trials. Finally, we propose how this knowledge could refine our future approaches to improve the efficacy of immunotherapy in DGC.

Overview of the carcinogenesis and clinicopathological features of DGC
Unlike IGC, which is intimately associated with H. pylori infection and follows a stepwise cascade of events from metaplasia to dysplasia before becoming malignant, the mechanisms driving the carcinogenesis of DGC are not very clear, and most understanding of DGC is derived from studies on patients or mouse models that carry germline mutations in the CDH1 gene, which encodes the adhesion junction protein E-cadherin13. E-cadherin dysfunction and other alterations that may involve the organization of the actomyosin cytoskeleton or the activation of signaling pathways associated with adhesion junctions trigger the delamination of cells into the mucosa13. Additional genetic alterations, including mutations in RHOA and CTNNB1, and changes in the tumor microenvironment (TME), such as increased stromal fibrosis and reduced immune cell infiltration, promote these intramucosal lesions into undifferentiated, invasive, and migratory stages13. Compared to IGC, the TME in DGC is characterized by a higher abundance of cancer-associated fibroblasts (CAFs) and a lower presence of cytotoxic T cells, contributing to its aggressive phenotype13. Although sporadic DGC tumors are indistinguishable from hereditary DGC (HDGC) histologically, HDGC is primarily driven by germline CDH1 mutations, whereas sporadic DGC may arise from somatic alterations in CDH1 or other genes (e.g., RHOA)14. Whether these stepwise events also occur in sporadic DGC or whether secondary acquisition of DGC-associated genetic alterations transforms IGC into DGC remains to be determined.
Pathologically, DGC manifests as diffusely infiltrating growth, characterized by poorly cohesive cancer cells scattered in stroma filled with rich fibrous components, leading to more aggressive tumor biology15,16. With respect to molecular characteristics, DGC often shows frequent mutations in CDH1, RHOA, and CTNNB1, which encode proteins involved in cell‒cell adhesion and the actomyosin network17. Clinically, DGC is enriched in younger populations, characterized by larger cancer masses, poorly differentiated and advanced stages, and tends to progress through direct invasion and peritoneal dissemination4,18–20. These clinicopathological characteristics result in significantly less favorable survival in DGC patients than in their IGC counterparts21. In addition to these negative clinicopathological characteristics, a low response to various therapies also contributes to an unfavorable prognosis. Generally, DGC is an independent prognostic factor in patients following radical gastrectomy and responds poorly to chemotherapy, chemoradiation and targeted therapy, which are routinely used for treatment in GC patients regardless of their histological subtype22.

Mechanisms underlying the poor responsiveness of DGC to immunotherapy
Although subgroup analyses in a few clinical trials revealed that ICIs may work in patients with DGC23–25, as no studies have focused on patients with DGC or stratified patients by the Lauren classification when randomization is performed, confounding factors, such as histological heterogeneity, cannot be accounted for, especially when the Lauren classification is determined by biopsy specimens. Accurate Lauren classification depends on comprehensive examination of all cancer tissue sections, which can be achieved only through radically resected samples. Currently, only the ATTRACTION-5 study included patients who all have undergone radical gastrectomy and reported no survival benefits of ICIs in patients with DGC26. Therefore, although conclusions cannot be drawn from these findings, patients with DGC may not be ideal candidates for ICIs, as the cold nature of the tumor immune microenvironment (TIME), shaped by the histological and molecular features of DGC, determines their poor responsiveness to immunotherapy (Fig. 1).

The cold nature of the DGC TIME
Carcinogenesis is not only determined by intrinsic genetic or epigenetic alterations but also sculpted by the immune system, which is referred to as immunoediting and proceeds through three phases: elimination, equilibrium, and escape27. Clinically diagnosed cancers have always developed immune escape mechanisms to form visible lesions. Therefore, to effectively treat cancer with immunotherapy, the immunoediting process must reoccur at the elimination stage, which relies on immune effector cells, especially cytotoxic T cells (CTLs), that preexist in the TME or are replenished from systemic immunity28. Therefore, immune components in the TME, referred to as the TIME, have been extensively studied in various dimensions of GC, with the potential to predict the efficacy of immunotherapy29. A recent study classified the TIME into four subtypes via RNA-seq data associated with immune and stromal cells. Among these subtypes, the “immune-enriched, nonfibrotic” subtype presented the highest response rate, whereas the “fibrotic” and “desert” TIMEs presented the lowest response rate to ICIs30. When GC patients from The Cancer Genome Atlas (TCGA) were defined by this classification system, the TIMEs of DGC were more likely to be the “fibrotic” and “desert” subtypes31. Using proteomic data, Shi et al. compared the relative abundance of immune cells in DGC and IGC patients with the endothelial subtype and reported that DGC patients presented greater numbers of CD4+ memory T cells, CD8+ T cells, and Th1 cells and fewer common lymphoid progenitor, natural killer (NK), and Th2 cells than IGC patients did. Interestingly, DGC and IGC patients with this immune subtype have opposite prognostic trends, with DGC patients showing poorer survival32. Compared with IGC, advanced DGC is more likely to spread through peritoneal metastasis (GCPM). Using single-cell sequencing, a recent study comprehensively assessed the TIME of GCPM and reported that the GCPM ecosystem features a distinct stroma–myeloid niche composed of tumor-associated macrophages (TAMs), endothelial cells and matrix cancer-associated fibroblasts (mCAFs), which underlies the resistance of GCPM to immunotherapy33. Additionally, via single-cell sequencing, Zou et al. reported that the immune context associated with DGC ascites was either immune ‘desert’ (immune cells barely present) or exhausted T cells with increased expression of immune checkpoint-related genes34. Finally, compared with IGC, peritoneal metastasis from DGC has a distinct TIME with less infiltration by CTLs, monocytes, NK cells, and dendritic cells (DCs)20.
Despite these findings obtained by advanced techniques, some studies have also investigated the TIME of DGC using clinically available measures. In one of these studies, immune cells, including tumor-infiltrating lymphocytes (TILs) and TAMs, were assessed via immunohistochemistry (IHC), and the results revealed that DGC had the lowest number of infiltrating immune cells, corresponding to cold tumors35. Peripheral blood mononuclear cells (PBMCs) and TILs were quantified by flow cytometry and IHC in a cohort of patients with untreated GC. Patients with diffuse and mixed GC have significantly lower rates of NK cells and regulatory T cells in circulation and less CD8+ TIL infiltration within cancer tissues than patients with IGC. The authors concluded that this cold TIME may be associated with poor survival36. In contrast, another study revealed more abundant intratumoral CD8+ TILs in cancer tissues from patients with DGC than in those from patients with IGC. However, these CD8+ TILs in DGC patients were less functional and may be different from their counterparts in IGC patients, as CD8+ TILs only predicted better survival in IGC patients but not in DGC patients37.
Overall, DGC has a suppressive or desert TIME with less anticancer effector cell infiltration or is shaped by distinct mechanisms that cannot be overcome by currently approved immunotherapy methods. An in-depth understanding of the histological and molecular features that lead to immunotherapy resistance can help in the design of rational combinations or new strategies to improve the efficacy of immunotherapy in DGC.

Histological and molecular barriers to effective immunotherapy in DGC
To achieve effective immunotherapy, the activation and maintenance of an anticancer immune response in the TME is essential, which involves complex processes and is determined by various factors, including the mutational landscape, antigen processing and presentation machinery (APM), immune-evasive oncogenic signaling pathways and trafficking of immune effector cells. Disruptions in any of these factors can lead to the failure of immunotherapy. Unfortunately, DGC exhibits many histological and molecular features that can disrupt these factors, which are essential for immunotherapy efficacy.
Numerous studies have established the predictive role of the tumor mutational burden (TMB) for immunotherapy, as neoantigens generated from these mutations have the potential to serve as effective targets for anticancer immunity38,39. We did not find any studies that reported the mutational landscape of DGC; however, according to the TCGA classification, DGC is enriched mainly in the genomically stable (GS) subtype, which has a lower TMB40. In a separate study, GC patients were classified according to ferroptosis patterns, and DGC was enriched in subtypes with higher ferroptosis subtype scores, which are associated with lower TMB and are less likely to respond to immunotherapy41. In addition, cancers with deficient mismatch repair (dMMR) or high microsatellite instability (MSI-H) exhibit high TMB and respond well to immunotherapy. dMMR/MSI-H cancers account for 4–10% of IGC cases but only 0–3% of DGC cases, indirectly reflecting the low TMB of DGC42–44. To expose neoantigens to the immune system, the APM plays critical roles, and cancer cells can evade attack from immune cells by downregulating the expression of the APM, especially components of major histocompatibility complex (MHC) class I. Although no studies have explored the expression of MHC class I molecules in DGC, lower expression of the HLA-DR antigen, an MHC class II molecule, has been observed in DGC45. MHC class II expression has also been shown to be a potential biomarker of the response to immunotherapy in other cancers, and disruption of its function may lead to immunotherapy resistance46,47 (Fig. 1A).
Interferon-γ (IFN-γ) signaling plays an important role in mediating the effective anticancer immune response, and disruption of the IFN-γ signaling pathway, including Jak1, Stat1, Ifngr1, Ifngr2, and Jak2, is mostly relevant to immunotherapy resistance48,49. There are no studies reporting alterations in the IFN-γ signaling pathway in DGC specifically; however, in the TME, IFN-γ activates Janus kinase (JAK)-signal transducer and activator of transcription (STAT) signaling, which increases PD-L1 levels (Fig. 1B). Therefore, PD-L1 status may be an indirect surrogate of the activation of the IFN-γ signaling pathway and productive anticancer immune response, although various other mechanisms also participate in the regulation of PD-L1 expression50,51. However, the PD-L1 levels in DGC were relatively lower than those in IGC. In a meta-analysis, the authors pooled data from 4 studies including 1101 patients and reported that PD-L1 was overexpressed in 18.79% of DGC patients and in 28.14% of IGC patients52. In large phase III clinical trials, the proportion of patients with high PD-L1 expression was also found to be lower in DGC. For example, in the CheckMate 649 study, patients with a combined positive score (CPS) ≥ 5 accounted for 52.75% of DGC patients and 64.38% of IGC patients23,53. Similarly, in the KEYNOTE-859 study, the proportions of patients with CPSs ≥1 and ≥10 were both lower in DGC25.
Oncogenic signaling pathways is not only the foundation of cancer initiation and progression but also relevant to tumor immunity, leading to an immunosuppressive TME. In the TCGA study, DGC was mostly classified as a GS tumor, which presented high levels of CDH1 mutations (37%), RHOA mutations (15%), and CLDN18-ARHGAP fusions (15%)40. CDH1 encodes the tumor suppressive protein E-cadherin, which functions as the main component of adhesion junctions and sequesters β-catenin at the cell membrane. In DGC, the loss of E-cadherin expression leads not only to disruption of adhesion junctions but also to activation of the WNT/β-catenin pathway54. For example, a recent study reported that WNT gene set scores were associated with decreased expression of CDH134. The WNT/β-catenin signaling pathway has recently been demonstrated to be a mechanism responsible for immune evasion and immunotherapy resistance in various cancers55. In a transplanted tumor model, tumors with CDH1 loss presented an immunosuppressive phenotype characterized by high infiltration of exhausted T cells and myeloid-derived suppressor cells (MDSCs)34. In addition, RHOA mutations and CLDN18-ARHGAP fusions activate the Yes-associated protein 1 (YAP1), phosphoinositide 3-kinase (PI3K)-AKT, and β-catenin signaling pathways, which have been found to have negative effects on anticancer immunity and the efficacy of immunotherapy56 (Fig. 1C).
Compared with IGC, histological examination of DGC frequently revealed a rich fibrous stroma, which was infiltrated by poorly cohesive cancer cells16. This ‘supermodule’ of stroma-related genes is associated with transforming growth factor-β (TGF-β) signaling, which has potent immunosuppressive effects, conferring resistance to immunotherapy16,57. TGF-β ligands induce the phenotypic transition of normal fibroblasts to CAFs, which impedes anticancer immunity through the secretion of cytokines, chemokines and growth factors and the synthesis and remodeling of the tumor stroma58. In the stroma of GCPM, accumulated CAFs recruit tissue-resident macrophages and induce their tumor-associated macrophage (TAM) transformation, which is associated with immunotherapy resistance33. Furthermore, the rich and compressive stroma shaped by the TGF-β signaling pathway and CAFs creates a physical barrier that limits immune effector cell infiltration and anticancer drug delivery to cancer cells59 (Fig. 1D).
Generally, the abovementioned signaling pathways, such as the WNT/β-catenin, TGF-β and YAP1 pathways, have the potential to induce epithelial‒mesenchymal transition (EMT) and cancer stem cell (CSC) phenotypes60,61 (Fig. 1C). Therefore, DGC is often enriched in molecular subtypes with EMT and CSC features40,62,63. For example, clustering analysis of the GC cohort from the TCGC identified clusters with EMT activation, which consisted of a high proportion of patients with DGC (59%)64. Cancer cells with the EMT or CSC phenotype are resistant to traditional cytotoxic therapies. In recent years, evidence has also accumulated to support the roles of EMT and stemness in immunotherapy resistance56.

Evidence of the efficacy of ICIs in DGC
Although several clinical trials have tested the efficacy of ICIs in patients with GC, none have focused on DGC or stratified patients by the Lauren classification when randomization is performed. However, in some large phase III clinical trials, subgroup analyses according to the Lauren classification have been reported (Table 1). Although these findings are unlikely to provide a definitive conclusion, indirect evidence on the efficacy of ICIs in patients with DGC can be obtained from these landmark trials. Consistent with the traditional pattern of anticancer drug development, immunotherapy for GC also began in later lines and then advanced to first-line and adjuvant therapy, which determines the sequence of the following evidence summaries.

Second- or more-line therapy
Initially, the ATTRACTION-2 trial investigated the efficacy of nivolumab, an inhibitor of programmed death-1 (PD-1), in Asian patients with advanced gastric or gastroesophageal junction cancer (GC/GEJC) who had experienced more than one previous chemotherapy regimen65. When included, patients were randomly treated with nivolumab monotherapy or placebo. At the data cutoff, nivolumab monotherapy was found to prolong the overall prognosis of the included patients. Although survival benefits of nivolumab were also observed in patients with DGC, the results did not reach statistical significance, with a hazard ratio (HR) and 95% confidence interval (CI) of 0.82 (0.57–1.17)65. In 2019, the authors updated the results of this trial, which confirmed the primary conclusions after a 2-year follow-up66. In the second-line setting, the KEYNOTE-061 study compared the efficacy of pembrolizumab monotherapy with that of paclitaxel in GC/GEJC patients who progressed on first-line chemotherapy. The results of this study revealed that pembrolizumab monotherapy is not superior to paclitaxel alone in improving overall survival (OS) as a second-line therapy for patients with advanced GC/GEJC, regardless of the Lauren classification (IGC or DGC), PD-1 ligand 1 (PD-L1) CPS value (overall or ≥1%) and time point (the primary report or the 2-year update)67,68.

First-line therapy
Because of the intensive immunosuppressive state in GC patients who have experienced previous intensive treatment, which may impede the efficacy of immunotherapy, investigators have conducted several large clinical trials to test ICIs in patients with newly diagnosed advanced GC/GEJC. The KEYNOTE-062 study is one of the earliest, which included only patients with a PD-L1 CPS ≥ 1. Patients were randomized to pembrolizumab, pembrolizumab in combination with chemotherapy, or chemotherapy plus placebo. The results indicate that although pembrolizumab was noninferior, it was also not superior to chemotherapy in terms of OS and progression-free survival (PFS). An intriguing finding of subgroup analyses is that in patients with DGC but not IGC, the margins of HR were smaller than 1, especially in patients with a PD-L1 CPS ≥ 1 and who received immunotherapy combined with chemotherapy24. Following KEYNOTE-062, KEYNOTE-859 further compared the efficacy of pembrolizumab in combination with chemotherapy and that of chemotherapy alone. In this study, pembrolizumab in combination with first-line chemotherapy significantly prolonged the OS and PFS of patients with advanced GC/GEJC. Subgroup analyses revealed that the difference in the objective response rate (ORR) was 6.5 (95% CI, −1.2–14.2) in patients with DGC, and the HR for PFS was 0.85 (95% CI, 0.71–1.02); however, the HR for OS was 0.76 (95% CI, 0.64–0.90), and the magnitude of benefit for OS was greater in patients with higher PD-L1 expression25.
Another commonly used ICI, nivolumab, has also been extensively studied as a first-line therapy in patients with advanced GC/GEJC. ATTRACTION-4 was carried out in Asian patients to test the efficacy of the addition of nivolumab to standard first-line chemotherapy. This study revealed that nivolumab plus oxaliplatin-based chemotherapy significantly prolonged PFS but not OS. However, improvements in both PFS and OS were not achieved in patients with DGC69,70. The CheckMate 649 study applied this therapeutic regimen globally, which significantly prolonged PFS and OS in patients with newly diagnosed advanced GC/GEJC. In patients with DGC, nivolumab in combination with chemotherapy did not result in a greater OS benefit than chemotherapy alone. However, when DGC patients were enriched with a PD-L1 CPS ≥ 5, an improvement in OS was observed23,53.
All the abovementioned studies were conducted in patients with human epidermal growth factor receptor 2 (HER2)-negative GC/GEJC. In HER2-positive patients, the standard first-line therapy is the combination of trastuzumab with cytotoxic chemotherapy. KEYNOTE-811 was designed to investigate whether pembrolizumab can further improve the survival of HER2-positive patients receiving standard first-line therapy. From the first to third interim analyses, pembrolizumab was found to significantly improve the ORR and PFS when combined with first-line trastuzumab and chemotherapy. However, in DGC patients, the difference in ORR was 18.1 (95% CI −8.5–42.5) in the first interim analysis, and the HRs for PFS were 0.72 (95% CI, 0.48–1.07) and 0.71 (95% CI 0.47–1.07) in the second and third interim analyses, respectively71,72.

(Neo)adjuvant therapy
In a report that retrospectively collected available data from an ongoing prospective clinical trial that aimed to test the efficacy of neoadjuvant tislelizumab plus chemotherapy, only patients with IGC achieved better pathological regression but not patients with DGC, and resistance occurred in patients with intestinal-to-diffuse transition following treatment31. In recent years, the results of two large clinical trials exploring immunotherapy as (neo)adjuvant therapy have also been published. ATTRACTION-5 was a randomized, multicenter and placebo-controlled phase III trial conducted in East Asia. The included patients were pathologically confirmed to have stage III GC/GEJC after radical resection. The findings of this trial did not support the combination of nivolumab with chemotherapy as adjuvant therapy in patients following radical gastrectomy. In the subgroup analysis for DGC, the HRs for relapse-free survival (RFS) and OS were 0.79 (95% CI 0.57–1.08) and 0.75 (95% CI 0.53–1.05), respectively26. The KEYNOTE-585 study investigated the anticancer activity of pembrolizumab in combination with chemotherapy as perioperative therapy in patients with locally advanced resectable GC/GEJC. Although the pathological complete response (pCR) was significantly improved by perioperative immunochemotherapy, it did not translate to significant event-free survival (EFS) benefits. Consistent results were also reported in patients with DGC73.

Perspectives on therapeutic implications
On the basis of these findings, immunotherapy, especially with ICIs alone, may not be a superior strategy in patients with DGC. Although not comprehensive, the understanding of the possible mechanisms underlying poor responsiveness suggests that maximizing the efficacy of immunotherapy in DGC patients needs rational combination or new strategies. In the following context, we introduce several such strategies, the majority of which are extrapolated from other disease settings, and whether they can work in DGC needs further validation (Fig. 2).

Promotion of progression of the cancer-immunity cycle
An active anticancer immune response and dynamic adaptation to cancer evolution require the iteration of a series of events, including antigen release and presentation, T-cell priming and trafficking, and recognition and killing of cancer cells74. As a cancer type that mostly presents with a cold TIME and is genomically stable, the induction of immunogenic cancer cell death (ICD) in DGC is the most critical step in initiating the cancer-immunity cycle. The ability of chemotherapy and radiotherapy to induce ICD has been demonstrated; however, both therapies have shown poor efficacy in patients with DGC22. The mechanisms of resistance to these traditional cytotoxic therapies are not yet understood, the elucidation of which may help improve the synergistic effects between them and immunotherapy. Most advanced DGC patients present with peritoneal metastasis and are candidates for peritoneal-directed regional therapy, such as hyperthermic intraperitoneal chemotherapy (HIPEC)75. HIPEC-associated mutational signatures were identified, and combination with immunotherapy improved the survival of patients with peritoneal metastasis from other cancer types76,77. Oncolytic virotherapy was also demonstrated to induce ICD and systemic and potentially durable anticancer immunity78. When oncolytic virus was injected into peritoneal metastases from GCs, the TIME improved79. These preliminary findings suggest that peritoneal-directed regional therapy may promote the cancer-immunity cycle and, in combination with the currently approved immunotherapy, may expand the arsenal of treatments for DGC (Fig. 2A).
Throughout the cancer-immunity cycle, cell trafficking plays a decisive role but is limited by the histological and molecular barriers in DGC. Therefore, to attract more immune effector cells to the TME of DGC, improving the accessibility of the cancer stroma and creating an inflamed TME are two promising approaches. Several strategies have been explored to improve the accessibility of the tumor stroma to increase immune cell infiltration and anticancer drug delivery. These strategies include CAF depletion, the use of enzymatic agents targeting matrix molecules, TGF-β blockade, and vasculature normalization, some of which have shown promising results in preclinical studies80. As a type of cancer rich in fibrous stroma, DGC may be an ideal candidate for testing the efficacy of these stroma-targeting therapies. In addition to physical barriers, the lack of immune cell infiltration is also attributed to other factors, such as the absence of chemokines, activation of oncogenic pathways, and high expression of immunosuppressive molecules. In combination with currently available drugs, drugs targeting these factors may improve the efficacy of immunotherapy, which has been demonstrated in other cancers81. For example, epigenetic therapy was shown to improve the immunogenicity of cancer cells, and innate immune response modulators can induce an inflamed TIME82. Both therapies may hold promise in DGC, as low immunogenicity and a cold TIME are two of the greatest obstacles impeding the success of immunotherapy in this cancer type (Fig. 2B).

Biomarker-directed patient stratification
Although patients with DGC included in clinical trials were less likely to benefit from immunotherapy, some of them achieved good responsiveness, indicating that DGC is not a homogeneous entity. For example, at the single-cell transcriptomic level, DGC can be divided into two major subtypes with different immune landscapes, which may respond differently to immunotherapy34. In addition, although rare, MSI-H tumors can still be detected in patients with DGC42–44. However, whether DGC patients with MSI-H tumors benefit equally to those with other MSI-H tumors is unclear. PD-L1 status is a key biomarker for immunotherapy in GC. While not routinely stratified by Lauren classification, subgroup analyses suggest that high PD-L1 expression may overcome the typical resistance of DGC, even with some trials showing survival benefits in DGC but not in IGC23,25. This highlights the importance of biomarker-driven patient selection, particularly in DGC’s heterogeneous population. Although clinically accessible, these IHC-based methods have been limited by their low dimensionality. Therefore, high-throughput, high-dimensional, single-cell approaches have been extensively studied to develop multidimensional biomarker signatures and significantly outperform traditional biomarkers in predicting the efficacy of immunotherapy83. Nevertheless, the exploration of these advanced methods in DGC is limited, which warrants further studies to better select DGC patients with the greatest likelihood of benefitting from immunotherapy (Fig. 2B).

Alternative immunotherapy strategies and targets
The majority of clinical trials and approved drugs for the treatment of GC have focused on targeting PD-1/PD-L1; however, DGC may not rely heavily on this immune checkpoint to escape anticancer immunity. Therefore, in addition to combination with other strategies to improve the efficacy of ICIs, the exploration of other immunotherapy strategies and targets may be an alternative direction.
In addition to the PD-1/PD-L1 axis, cancer cells use various mechanisms to escape immune surveillance, such as other T cell inhibitory receptors, immunosuppressive cells and metabolic alterations, representing a broad spectrum of potential targets to further modulate anticancer immunity. To date, exciting results have been reported in trials targeting these mechanisms84. However, the extent to which these alternative immune evasion mechanisms are exploited by DGC has not been investigated, which warrants further studies.
To overcome the inherent limitations of immunotherapy, which unleash endogenous T-cell immune activity, adaptive cell therapy and immune cell engagers have been developed, the success of which requires the expression of tumor-specific targets85,86. Compared with IGC, DGC was found to express some cell surface molecules at higher levels, such as claudin 18.2, fibroblast growth factor receptor-2 isoform IIIb (FGFR2-IIIb) and MET22. Claudin 18.2, a splice variant of claudin 18, is found mainly in the normal gastric epithelium. In DGC, the surface expression of claudin 18.2 is relatively high, indicating that claudin 18.2 is an appealing target for immunotherapy87. For example, T-cell engagers designed to target both claudin 18.2 and CD3 have been tested in clinical trials and have shown promising results87. In a phase I trial, patients with claudin 18.2-positive gastrointestinal malignancies were treated with claudin 18.2-specific chimeric antigen receptor (CAR) T-cell therapy. The authors reported an ORR of 48.6% and a disease control rate of 73.0%, with 44.8% of responses lasting ≥6 months88. The reasons for this uncommon success in solid tumors are unknown, and elucidation of these factors may greatly improve the management of patients with DGC, as the proportion of patients with high claudin 18.2 expression is high, but other effective treatments are limited (Fig. 2C).

Conclusions
With the decreasing incidence of IGC associated with H. pylori eradication, DGC has become a great healthcare challenge. Despite the promising advancements in immunotherapy for GC in recent years, compared with IGC, DGC responds poorly to this widely approved therapy. The mechanisms underlying this lower responsiveness are complex. Together, low TMB, disruption of antigen presentation, activation of specific oncogenic pathways and rich fibrous stroma create a cold TIME in DGC, leading to low responsiveness to immunotherapy. Therefore, strategies targeting these factors are promising for synergizing with current immunotherapies. However, none of them have been investigated in DGC, although they are worthy of further consideration. Alternatively, exploring other immunotherapy strategies and targets beyond the PD-1/PD-L1 axis, such as CAR-T cells targeting claudin 18.2, may expand the arsenal of treatments for DGC.
The most prominent feature of modern medical oncology is personalization and precision. However, such an approach has not been established for immunotherapy in patients with GC. In addition to considering MSI and PD-L1 expression in some settings, currently available ICIs are approved for all GC patients, reflecting the limited number of strategies, with only drugs targeting PD-1/PD-L1 being approved. In addition, even in DGC, heterogeneity is inevitable. Therefore, a detailed and high-dimensional understanding of the dynamic histological, molecular and immune landscape in patients with DGC will be the decisive step forward in improving the efficacy of currently available immunotherapies and developing new approaches to overcome the immense health burdens caused by this GC subtype.