As we know, except for tumor cells, PD-L1 is highly expressed on the surface of tumor infiltrating immune cell, such as neutrophils, macrophages, and mast cells. Since GC is a solid tumor with relatively weak immunogenicity, we found that there are many studies related to immunotherapy of GC target the immune cells infiltrating around the tumor. Figure 2 briefly describes the expression of PD-L1 on immune cells in the microenvironment of GC.

In GC, the expression of PD-L1 of neutrophils can be induced by tumor derived granulocyte-macrophage colony-stimulating factor (GM-CSF), which can be activated through Janus kinase (JAK) - signal transducer and activator of transcription 3 (STAT3) pathway, to a large extent, inhibiting tumor immunity and promote progression of GC (35). Similarly, while chemokine colony stimulating factor 2 (CSF-2) secreted by GC cells, macrophages are stimulated to secrete lots of chemokine CXCL8. In this autocrine form, macrophages continue to receive positive feedback, resulting PD-L1 highly expresses on the cell surface, to play the above tumor immunosuppression (36). In the tumor microenvironment, in some degree cytokines and chemokines are responsible for recruiting immune cells. Among them, GC cells can recruit mast cells through CXCR4-CXCL12 receptor-ligand binding. Then, under the condition that GC cells secrete a large amount of tumor necrosis factor a (TNF-a), while mast cells express TNF-a II ligand to receive such signals, and highly express PD-L1 through NF-kB pathway, to further promote the malignant progression of GC by cascade (37).

Induction and activation of JAK2/STAT3 signal by interferon γ (IFN- γ) can promote the upregulation of PD-L1 expression in GC and various kinds of solid tumors (38–40). In GC, in addition to the EGFR mentioned above, there are c-Met, mTOR, PI3K-Akt, JAK2/STAT3 and NF KB signaling pathways or key molecules (41, 42) that can make a difference to PD-L1. When it comes to STAT signal related genes, some researchers found that some DNA double strand breaks can activate STAT signal pathway and enhance the expression of PD-L1 (43). Besides, cytokine TNF- α and IL-6 also activate NF-KB and STAT3 signaling pathways in regulating the expression of PD-L1 (44). Other inflammatory cytokines, such as toll like receptor 3(TLR-3), transforming growth factor β (TGF-β), IFN- α/β and IL-4/6/17/27 have been shown to enhance the expression of PD-L1 mRNA on tumor cells or tumor associated stromal cells. In tissue samples from patients with GC, the expression of PD-L1 on tumor cell membrane was positively correlated with the presence of CD8 positive T cells and IFN- γ in matrix (45, 46). Overexpressing PD-L1 GC cell could recruit and enrich myeloid-Derived Suppressor Cells (MDSCs) by releasing lots of cytokines and promote the occurrence of gastric tumors by inhibiting tumor infiltrating CD8 + T cells (47). There are reports in the document that IFN- γ could restrain autophagy in the mechanism of promoting PD-L1 gene transcription (38, 40). Lately, Wang et al. found that autophagy, one specific mode of cell death, can directly down regulate PD-L1 through p62/SQSTM1-NF-κB pathway in GC (48).

The increased expression of PD-L1 in GC is related to the epithelial mesenchymal transformation phenotype (EMT), which can further increase the potential of tumor metastasis. The causal role of PD-L1 in promoting EMT has been confirmed in gastric carcinoma that IFN- γ upregulates the expression of PD-L1 in solid tumor cells through JAK signal transducer and transcriptional pathway activator, and weakens the cytotoxicity of tumor antigen specific CTL to tumor cells. After pretreatment with IFN- γ, cells were treated with anti-PD-L1 mAb, the decreased anti-tumor CTL activity by IFN- γ reached a level higher than that of non-therapeutic control target (45). Mir-940/Cbl-b/STAT5A axis up regulates the expression of PD-L1 and promotes the proliferation and EMT of GC cells (63). In another study, it is also confirmed that EMT can enhance the migration and invasion of GC cells through nuclear factor κB (NF-κB) pathway upregulates PD-L1 (64). Another study showed that the induction of EMT and PD-L1 expression was bidirectional. The induction of EMT upregulates the expression of PD-L1 mainly by activating PI3K-Akt pathway. These results demonstrate that the high expression level of PD-L1 is directly related to the phenotype of epithelial mesenchymal transformation.

The high expression of PD-L1 in GC can increase the potential of gastric CSCs and further promote the proliferation and progression of GC. Gastric CSCs also express PD-L1 (65). It is reported that GC cells were heterogeneous, and that PD-L1 in GC cells had different reactivity to GCMSCs and PD-L1 associated with CTCF to contribute to the stemness and self-renewal of GC cells, and GC mesenchymal cells (GCMSCS) enhance the CSC like characteristics of GC cells through PD-L1, resulting in the resistance of GC cells to chemotherapy (62). In vivo, PD-L1 positive GC cells have greater stem cell potential and tumorigenicity than PD-L1 negative GC cells (62). We find that the specific mechanism between PD-L1 and CSC in tumor cells in the latest study this year, that, the interaction between PD-L1 and frizled 6 receptor up-regulates β-Catenin targeted gene expression and promotes cancer progression through the maintenance and expansion of CSC (66).

The most famous biological function of PD-1/PD-L1 axis is to promote apoptosis of effector T cells, the basic principle behind which, is tumor immune escape. In 1992, Ishida et al. found PD-1 expressed transiently during inducing T cell death apoptosis in mice (67). In 2002, Minato et al. found that PD-L1 and PD-1 signaling pathways are involved in the regulation of tumor immunity for the first time (68). It’s reported that endogenous PD-L1 expression in gastric adenocarcinoma epithelial cells can promote T cell apoptosis (69). In another study, exogenous PD-L1 from GC can also induce Jurkat T cell apoptosis and inhibit T cell activation (50). PD-L1 can activate epidermal growth factor receptor (EGFR) through mir-429 to resist apoptosis induced by tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) in GC (20). The most classic mechanism of PD-L1 promoting apoptosis and inhibiting T cell activation can be explained as follows: after PD-L1 binding to PD-1, tyrosine residues located in ITSM of PD-1 are phosphorylated and protein tyrosine phosphatases (PTPs), such as SHP2, are recruited. These PTPs can dephosphorylate various key signal kinases and resist the positive signal events caused by the co-stimulation of TCR-CD28 receptor during T cell activation. To some extent, they preferentially inhibit the pathways driven by TCR rather than the CD28 mediated pathway. We also find an important role of PD-L1 in another cell death, autophagy. In general, autophagy is negatively controlled by the PI3K/Akt/mTOR pathway, and PD-L1 induced activation of this pathway seems to lead to the down regulation of autophagy. It is recently reported that autophagy can inhibit the expression of PD-L1 in GC cells reversely (48). Another study has explored the mechanism that autophagy downregulates PD-L1 by reducing the expression of histone deacetylase (HDAC) (70).

A recurring observation among phase III gastric and esophageal trials combining chemotherapy and immunotherapy has been differential activity based on PD-L1 expression (71). Some studies have reported that PD-L1 can reduce the cytotoxicity sensitivity to cytokine induced killer cell (CIK) therapy in GC (72). The expression of PD-L1 can also enhance cisplatin resistance in GC (73). Low dose Diosbulbin-B (DB) (12.5 μ M) activates tumor-intrinsic PD-L1/NLRP3 signaling pathway to restore the sensitivity of the cisplatin resistant GC (CR-GC) cells to cisplatin, the level of PD-L1 downregulates simultaneously in GC cells (74). These results further confirm that PD-L1 can reduce the sensitivity of GC cells to cisplatin. However, one research reports that non-responders to XELOX chemotherapy, all of whom were PD-L1 negative, failed to upregulate PD-L1 expression on treatment or favorably remodel their TME and exhibited minimal change in T-cell infiltration but rather an increase in LAG3 expression and B-cell infiltration (75). The relationship between PD-L1 expression and cytotoxic chemotherapy response has not been examined in detail, but response is numerically lower (41% vs. 46%) in PD-L1–negative patients in the large (n = 1,581) phase III CheckMate-649 trial (76).

In different clinical trials of PD-1/PD-L1 inhibitors, most clinical trials have benefited from OS and PFS to varying degrees. ATTRACTION-4 and CHECKMATE-649 trials suggest that the combination of nivolumab with chemotherapy might be more effective than chemotherapy alone (76, 81). KEYNOTE-062, a phase 3 randomized clinical trial demonstrated that pembrolizumab was noninferior to chemotherapy for OS in patients with CPS of 1 or greater, but pembrolizumab monotherapy was not superior to chemotherapy in patients with CPS of 1 or greater (82). However, there have also been certain cases of failure. For example, the use of avelumab alone in the third-line treatment of GC patients can not significantly improve the overall survival (OS) of patients compared with the chemotherapy drugs selected by doctors. However, the results of Javelin gastric 300 (clinical trial NCT02625610) provide an important reference for us to continue to explore the role of avelumab in the treatment of GC. According to different drug regimens of PD-L1/PD-1 inhibitors, we divided them into first-line, second-line, third-line, and late treatment according to the combination of Europe, the USA, and Asia guideline for gastric cancer, as shown in Table 1 .

Chromosomal events in patients with CIN GC include HER2 + or ERBB2 amplification, FGFR2b expression, FGFR amplification, EGFR amplification and MET amplification. Trastuzumab plus chemotherapy is the first-line drug for GC patients with HER-2 (+) and ERBB2 amplification in the current medical guidelines ( Figure 3 ). Trastuzumab Deruxtecan DS-8201 (China) has been reported as a third-line drug for HER2 + GC (83). ADC targeting HER2 and immune checkpoint inhibitor are important potential biomarkers for clinical treatment, as well. Pembrolizumab currently be used in the second-line treatment of MSI GC. Tumor tissues were collected from patients participating in ITACA-S, and randomized adjuvant chemotherapy trials were carried out. According to MSI status, inflammatory response, and the expression of PD-L1, the prognosis and regressive analysis of MSI GC were analyzed. The results show that MSI status may be a meaningful prognostic biomarker in patients with radical resection stage II-III GC. The first-line regimens of PD-1/PD-L1 inhibitors currently used in GC patients expressing PD-L1 are: (1) nivolumab plus chemotherapy (CPS ≥ 5%); (2) Sintilimab plus chemotherapy (China). Third-line therapeutic schedule is pembrolizumab (CPS ≥ 1%). PD-1 monoclonal antibody may also be used as a potential first-line regimen for the treatment of MSS/TMB-high GC. At present, there is no suitable PD-1/PD-L1 agents for EBV positive GC, but PD-1 monoclonal antibody may be a potential first-line treatment for EBV cytotoxic T lymphocytes. Nivolumab (Asia) and Apatinib (China) can be used as a third-line regimen for some GC patients without clear molecular classification. At the same time, PD-1 monoclonal antibody combined with multiple kinase TKI (Stivarga/Lenvatinib) and PD-1 monoclonal antibody combined with PARP inhibitor can also be used as candidate immunotherapy schemes for this kind of groups.

As many of the patients with GC are PD-L1 negative, combination therapy is attractive. Furthermore, even some patients who are PD-L1 positive have quite limited benefit from anti-PD-1 therapy alone and relapse/recurrence frequently occurs due to various resistance mechanisms (84). The limitation of anti PD-L1/PD-1 monotherapy is due to the complex tumor microenvironment (TME) and multiple immunosuppressive factors in each step of the cancer immune cycle (85). In order to achieve effective tumor control and eradication, three levels of immunotherapy capabilities are required. First, CD4+T, CD8+T and B lymphocytes are activated to become antigen specific effector cells. Second, the inhibitory factors in the turbine microenvironment, represented by PD-L1 and others, are neutralized. Third, tumor cells that express related targets are recognized and killed (86). Therefore, increasing the benefit of checkpoint-immune blockade through combinations to achieve long-term efficacy may be helpful (87, 88). At present, PD-1/PD-L1 inhibitors combined with various chemotherapeutic drugs or targeted regimens have been reported in clinical trials.

In terms of mentioned above dual ICIs treatment scheme, the application of immune checkpoint inhibitors can amplify the immune response, T lymphocytes were over-activated, leading to the immune-related adverse events (irAEs). Gastrointestinal tract, lung, joint and other organs could be injured, such as inflammatory arthritis, synovitis, synovitis and Sjogren’s syndrome, antinuclear antibody positive and other immune diseases would appear. Patients treated with anti-PD-1/PD-L1 were at greater risk of hypothyroidism, hepatotoxicity, and pneumonia.

The current application of PD-L1/PD-1 inhibitors in patients with advanced GC has greatly improved the prognosis and quality of life of patients (105). Moreover, we found that about 12.5% of the patients who received immunotherapy could benefit which based on the data of patients who met the qualification criteria of ICI treatment in 2018, The proportion of patients who benefit from this is less in GC. Therefore, the drug regimen of PD-L1/PD-1 inhibitor is recommended only for some patients with advanced GC with high expression of PD-L1. In these patients, acquired drug resistance also occurs in the later stage of treatment due to different reasons such as tumor heterogeneity, epitope change, PD-L1 blocking and so on (106).

In addition, the use of ICIs and the benefits of patients may also be related to the intestinal microbial environment. Recently, it has been reported that intestinal flora may affect the therapeutic effect of anti PD-1 inhibitors on patients through the level of tumor microenvironment (107), which has been further verified in patients with melanoma (108). Intestinal microbiome analysis showed a significant increase in bifidobacterial species in mice with slow tumor growth and the consequent increased benefit of PD-1 therapy (107, 109). The intestinal diversity of patients who responded to PD-1 treatment significantly increased their microbiota, some of which were relatively rich, such as Clostridium and Fecal bacilli (108). At present, it is not clear which specific composition of intestinal microbiota is most conducive to promoting anti-tumor immune response or promoting immunotherapy where the research on intestinal microbial ecology and tumor microenvironment is in full swing.

Immunotherapy such as anti-PD-1/PD-L1 further leads to the rapid progression of some tumors, which is called hyper progressive disease (HPD). The incidence rate of HPD in AGC is not clear, but the risk is certain. It was previously reported that when nivolumab was administered to AGC patients, liver metastasis and size of tumor at baseline were significantly associated with HPD. Some researchers explained that it may be necessary to further replace the concept of HPD with super HPD (naturally rapid growth of tumor after treatment) (110). At the same time, a separate case reported the occurrence of HPD in patients with advanced GC treated with nivolumab (111). However, it has been reported that there is no statistical correlation between PD-L1 expression in GC and hyper-progressive diseases (112).

In addition to the corresponding side effects, it has also been reported that new nanoparticles composed of platinum loaded complexes can produce positive synergistic effects in a large dose range, thereby achieving higher immunotherapeutic effects and lower side effects (113). MSI-H/dMMR phenotype, high tumor mutational burden and EBV etiopathogenesis appear to be more reliable indicators of tumor immunogenicity and sensitivity to PD-1L/PD-1 inhibitors. Indeed, PD-L1/PD-1 inhibition can substantially improve the outcome of these kinds of GC patients, the problem to be solved is to better identify the patients and the corresponding inclusion criteria for PD-L1/PD-1 therapy, while further clinical trials are needed to explore such specific biomarkers (114).