Predictive Biomarkers for Immunotherapy in Gastric Cancer: Current Status and Emerging Prospects

Gastric cancer presents substantial management challenges, and the advent of immunotherapy has ignited renewed hope among patients. Nevertheless, a significant proportion of patients do not respond to immunotherapy, and adverse events associated with immunotherapy also occur on occasion, underscoring the imperative to identify suitable candidates for treatment. Several biomarkers, including programmed death ligand-1 expression, tumor mutation burden, mismatch repair status, Epstein–Barr Virus infection, circulating tumor DNA, and tumor-infiltrating lymphocytes, have demonstrated potential in predicting the effectiveness of immunotherapy in gastric cancer. However, the quest for the optimal predictive biomarker for gastric cancer immunotherapy remains challenging, as each biomarker carries its own limitations. Recently, multi-omics technologies have emerged as promising platforms for discovering novel biomarkers that may help in selecting gastric cancer patients likely to respond to immunotherapy. The identification of reliable predictive biomarkers for immunotherapy in gastric cancer holds the promise of enhancing patient selection and improving treatment outcomes. In this review, we aim to provide an overview of clinically established biomarkers of immunotherapy in gastric cancer. Additionally, we introduce newly reported biomarkers based on multi-omics studies in the context of gastric cancer immunotherapy, thereby contributing to the ongoing efforts to refine patient stratification and treatment strategies.


Introduction
Gastric cancer (GC) presents a significant global health concern, ranking fifth in incidence and fourth in mortality worldwide.In 2020 alone, there were over one million new GC cases, resulting in approximately 769,000 deaths [1].Advanced GC (AGC), also known as mid-to late-stage gastric cancer, refers to tumors that invade the muscular layer or all layers of the stomach.Previously, AGC patients had limited treatment options, primarily relying on chemotherapy with modest efficacy.In recent years, immunotherapy utilizing immune checkpoint inhibitors (ICIs) has shown therapeutic effectiveness in specific GC patients, especially those exhibiting specific traits, such as the high expression of programmed death ligand-1 (PD-L1), high microsatellite instability (MSI-H), and patients who are Epstein-Barr Virus (EBV)-positive.These patients have shown a higher response rate to immunotherapy [2].The US Food and Drug Administration (FDA) now approves pembrolizumab as a third-line treatment for recurrent or metastatic GC or gastroesophageal junction adenocarcinoma with PD-L1 expression ≥ 1%.Additionally, nivolumab in combination with chemotherapy has been approved as a first-line therapy for advanced or metastatic GC and gastroesophageal junction cancer (GEJC), irrespective of PD-L1 expression.For human epidermal growth factor receptor 2 (HER2)-positive AGC, the FDA recommends a combination of programmed death-1 (PD-1) monoclonal antibody (pembrolizumab) with HER-2 monoclonal antibody (trastuzumab) along with chemotherapy.Nevertheless, a considerable number of patients continue to demonstrate an inability to benefit from immunotherapy, and immunotherapy-related adverse reactions are observed intermittently.Further research into immune-related biomarkers for GC becomes especially crucial for the precise selection of suitable candidates for immunotherapy.
Advancements in genetic testing technology have led to the identification of specific molecular markers for GC, recognizing it as a heterogeneous disease.The Cancer Genome Atlas (TCGA) project conducted a comprehensive evaluation in 2014, classifying GC into four distinct subtypes: EBV-positive GC, MSI GC, genomically stable (GS) GC, and GC with chromosomal instability (CIN) [3].The Asian Cancer Research Group (ACRG) proposed an additional molecular typing of GC based on samples from Asian populations, further correlating it with tumor prognosis.ACRG classified GC into four subtypes: the MSI subtype, predominantly Lauren intestinal type with early clinical staging (stage I or II) and the best prognosis; MSS/EMT subtype, displaying a significantly lower age of onset than other subtypes, with the absent expression of cell adhesion gene CDH1, mostly corresponding to Lauren diffuse type, late staging (stage III or IV), and the worst prognosis among the four subtypes; MSS/TP53+ subtype, with a higher frequency of EBV infection; and the MSS/TP53-subtype, enriched with HER-2 gene amplifications [4].
While these molecular subtypes have provided some guidance in treating and predicting outcomes for GC patients, more effective predictive biomarkers, particularly for populations benefiting from immunotherapy response, are needed.Currently, HER2 and PD-L1 expression are the sole predictors guiding treatment choice in AGC patients.Targeting HER2 has shown significant improvements in overall survival for HER2-positive patients [5].However, PD-L1 expression's relationship with GC response remains controversial [6].Other potential predictive biomarkers for immunotherapy response in GC include MMR status, TMB, and EBV infection, but larger-scale studies are necessary to validate their effectiveness [7].Liquid biopsy and emerging technologies like multiplex immunohistochemistry (mIHC) and single-cell sequencing offer new tools for individualized GC research [8][9][10].Given the potential value of immunotherapy in GC treatment and the occurrence of adverse reactions associated with immunotherapy, identifying patients who may benefit from this treatment is of paramount importance.Thus, in this review we aim to summarize current clinically relevant biomarkers predicting GC immunotherapy prognosis and identifying beneficial patient populations.Additionally, we will introduce new predictive biomarkers of GC immunotherapy discovered using novel technologies.We hope this review will pave the way for new directions in GC immunotherapy biomarker research.

PD-L1 Expression as the Biomarker for GC Immunotherapy
PD-L1 is a critical ligand protein expressed on tumor cells and tumor-associated antigen-presenting cells.Its binding to programmed death-1 (PD-1) on T cells leads to T cell deactivation, enabling tumor cells to evade the host's immune response.ICIs targeting PD-L1 or PD-1 can interrupt this interaction, activating T-cell anti-tumor immunity [11].Consequently, PD-L1 expression levels serve as a crucial evaluation index for ICI response [12] (Figure 1).
Immunohistochemistry (IHC) is the primary method for detecting PD-L1 protein expression levels.To standardize PD-L1 testing, the FDA has authorized four PD-L1 diagnostic monoclonal antibodies for IHC: 22C3, 28-8, SP142, and SP263.IHC indicators such as combined positive score (CPS) and tumor proportion score (TPS) are used to evaluate PD-L1 expression.TPS represents the percentage of tumor cells stained with any intensity of PD-L1, while CPS accounts for PD-L1-positive tumor cells and tumor-associated immune cells divided by the total number of tumor cells [13].Several clinical studies have evaluated TPS or CPS as predictive biomarkers for ICI therapy in GC (    Immunohistochemistry (IHC) is the primary method for detecting PD-L1 protein expression levels.To standardize PD-L1 testing, the FDA has authorized four PD-L1 diagnostic monoclonal antibodies for IHC: 22C3, 28-8, SP142, and SP263.IHC indicators such as combined positive score (CPS) and tumor proportion score (TPS) are used to evaluate PD-L1 expression.TPS represents the percentage of tumor cells stained with any intensity of PD-L1, while CPS accounts for PD-L1-positive tumor cells and tumor-associated immune cells divided by the total number of tumor cells [13].Several clinical studies have evaluated TPS or CPS as predictive biomarkers for ICI therapy in GC (Table 1).In the third-line therapy setting, KEYNOTE-012 reported on the response of AGC patients to ICI treatment.The study evaluated the safety and efficacy of pembrolizumab in PD-L1-positive recurrent or metastatic GC or GEJC patients.The results demonstrated promising the anti-tumor activity and safety of pembrolizumab for PD-L1-positive GC/GEJC patients (TPS high) and highlighted the significance of measuring PD-L1 expression in mononuclear inflammatory cells [14].KEYNOTE-059 investigated pembrolizumab as thirdline therapy for GC or GEJC patients, showing an overall response rate (ORR) of 22.7% for patients with CPS ≥ 1, while PD-L1-negative tumors had an ORR of 8.6%.The study also explored the predictive value of the MSI status and immune-related gene expression for immunotherapy [15].Subsequently, based on the KEYNOTE-059 results, the FDA granted the accelerated approval of pembrolizumab for patients with recurrent locally advanced or metastatic GC or GEJC with CPS ≥ 1.The CheckMate032 study evaluated dual immune checkpoint inhibitors, combining nivolumab with ipilimumab, for metastatic esophagogastric cancers.The study demonstrated clinically meaningful antitumor activity for nivolumab and dual immune checkpoint inhibitors, irrespective of TPS and MSI status [16].Post hoc exploratory analyses of CheckMate032 indicated that CPS with cutoffs of ≥5 and ≥10 were more accurate predictors of immunotherapy response than TPS.Additionally, inflammatory gene signatures were associated with the response to ICI treatment [17].The ATTRACTION-2 trial showed nivolumab provided survival benefits to GC or GEJC patients, independent of TPS [18].Meanwhile, in a subset analysis of the ATTRACTION-2 study, the TPS, blood-neutrophil-to-lymphocyte ratio (NLR), and serum Na levels were proposed as predictive biomarkers for the response to nivolumab in previously treated AGC patients [19].JAVELIN Gastric 300 trial evaluated avelumab in third-line therapy for advanced GC/GEJC patients, but it did not improve overall survival (OS) or progress free survival (PFS) compared to chemotherapy.Subgroup analysis based on TPS ≥ 1 did not reveal significant differences between avelumab and chemotherapy arms in terms of OS [20].The maintenance therapy study, JAVELIN Gastric 100, involving avelumab, revealed no significant improvement in OS when compared to chemotherapy.This was observed among patients with advanced GC or GEJC, both in the overall group and specifically within the population with positive PD-L1 expression (CPS ≥ 1) [21].
The KEYNOTE-061 trial assessed anti-PD-L1 therapy as a second-line treatment for AGC.Initial results indicated that pembrolizumab did not significantly enhance OS compared to paclitaxel in patients with a CPS ≥ 1 [22].However, subsequent 2-year follow-up data unveiled encouraging trends, displaying the potential for enhanced OS with pembrolizumab when compared to paclitaxel within the CPS ≥ 1 subgroup.The 24-month OS percentages stood at 19.9% for pembrolizumab and 8.5% for paclitaxel.Notably, the efficacy of pembrolizumab in OS was more pronounced with PD-L1 enrichment: CPS ≥ 5 (24-month rate: 24.2% vs. 8.8%) and CPS ≥ 10 (24-month rate: 32.1% vs. 10.9%).However, no substantial differences in median PFS were observed across the treatment groups [47].Additionally, a recent report involving exploratory analysis within the KEYNOTE-061 trial utilized RNA sequencing data to evaluate the 18-gene T-cell-inflamed gene expression profile (TcellinfGEP) and ten non-TcellinfGEP signatures.This analysis revealed that TcellinfGEP exhibited associations with ORR and PFS specifically for pembrolizumab, without the same effects for paclitaxel.Notably, the TcellinfGEP-adjusted mMDSC signature exhibited adverse relationships with ORR, PFS, and OS for pembrolizumab, whereas such associations were absent in the case of paclitaxel [40].In a phase I/II trial (UMIN000025947) investigating second-line therapy for AGC, a combination of nivolumab, paclitaxel, and ramucirumab was studied.This trial reported a median OS of 13.8 months for CPS ≥ 1 patient and 8.0 months for CPS < 1 patient [23].Furthermore, a phase II trial conducted by the Arbeitsgemeinschaft Internistische Onkologie (AIO) evaluated the effectiveness of combining avelumab, ramucirumab, and paclitaxel as second-line therapy for GEJC patients.The trial findings indicated a median OS of 9.4 months for CPS < 5 patients compared to 14.0 months for CPS ≥ 5 patients, underscoring the prognostic value of CPS in predicting the response to immunotherapy in GC patients [24].In the MAHOGANY study, an exploration was conducted involving the amalgamation of the anti-HER2 monoclonal antibody (margetuximab) and the anti-PD-1 monoclonal antibody (pembrolizumab) in individuals afflicted by trastuzumab-resistant HER2-positive gastro-esophageal adenocarcinoma.Among patients with positive PD-L1 expression, the observed ORR was 33%; in contrast, patients with negative PD-L1 expression displayed a markedly lower ORR of 7% [25].In the CPS < 1 subgroup (n = 11), the pCR and MPR rates were 9.1% and 27.3%, respectively.In the 21 patients with CPS ≥ 1, the pCR and MPR rates were 28.6% and 57.1%, respectively.Among the 11 patients with CPS ≥ 5, the pCR and MPR rates were 27.3% and 54.5%, respectively.Among the six patients with CPS ≥ 10, the pCR and MPR rates were 33.3% and 50.0%, respectively.Two (5.6%) patients had dMMR, and one of them, with high PD-L1 expression (CPS = 68), achieved pCR, while the other patient, whose PD-L1 expression is unknown, had a non-major pathological response (TRG2).Among the two (5.6%) patients with EBV-positive status, CPS values of 10 and 0 were found, respectively, and none of them achieved a major pathological response [ The level of tumor-infiltrating lymphocytes (TILs) was measured by multiplex immunofluorescence (mIF) among these patients Intertumoral CD8+ T cells were obviously increased in CPS PD-L1 negative patients who responded to chemoimmunotherapy, compared with patients who did not respond (p = 0.011).And higher level of CD8+ TILs was demonstrated to associate with better PFS in CPS PD-L1-negative patients treated with chemoimmunotherapy (HR =23.70, 95% CI: 1.15-488.30,p = 0.04) [46] AGC: advanced gastric cancer; CPS: combined positive score; ctDNA: circulating tumor DNA; EBV: Epstein-Barr virus; EBVaGC: EBV-associated gastric cancer; EVs: Extracellular vesicles; GEJC: gastroesophageal junction cancer; HER2: human epidermal growth factor receptor 2; HR: hazard ratio; ICI: immune checkpoint inhibitor; IHC: immunohistochemistry; ISH: in situ hybridization; MPR: major pathological response; MSI: microsatellite unstable; MSI-H: high microsatellite instability; NGS: next-generation sequencing; NLR: neutrophil-to-lymphocyte ratio; ORR: overall response rate; OS: overall survival; PD-L1: programmed death ligand-1; PFS: progression-free survival; pCR: pathologic complete response; PR: partial response; SD: stable disease; TILs: tumor-infiltrating lymphocytes; TMB: tumor mutation burden; TPS: tumor proportion score; tTMB: tissue tumor mutation burden; WES: whole-exome sequencing.
In April 2021, the FDA approved nivolumab as a first-line treatment for advanced or metastatic GC patients, regardless of their PD-L1 expression.This decision was based on the findings of CheckMate 649 [26], which demonstrated significant improvements in OS and PFS for patients with a CPS of ≥1 and for all randomized patients with PD-L1 expression.However, although the combination of nivolumab and chemotherapy led to significant OS and PFS benefits for patients with a CPS of ≥5, long-term results for nivolumab plus ipilimumab versus chemotherapy did not reach statistical significance [48].In the ORIENT-16 trial, the efficacy of sintilimab (PD-1 inhibitor) combined with chemotherapy for advanced GC/GEJC patients was studied.OS benefits were consistent across CPS cutoffs (CPS ≥ 1, 5, and 10) [27].The KEYNOTE-062 study enrolled untreated, locally advanced/unresectable or metastatic G/GEJC patients with CPS ≥ 1.The result showed pembrolizumab, either alone or in combination with chemotherapy, did not show superiority in OS and PFS compared to chemotherapy alone [28].In the context of first-line therapy for patients with HER2-negative, unresectable advanced or recurrent GC or GEJC, the ATTRACTION-4 study investigated the comparative efficacy between nivolumab plus chemotherapy and placebo plus chemotherapy.The results indicated that nivolumab plus chemotherapy significantly improved PFS, but not OS, with no significant prognostic difference observed between patients with a TPS ≥ 1 and those with a TPS < 1 [29].In the KEYNOTE-659 study, researchers evaluated pembrolizumab in combination with chemotherapy in Japanese advanced GC/GEJC patients who were PD-L1-positive (CPS ≥ 1) and HER2-negative.Exploratory analysis of OS according to CPS status, in cohort 1 (pembrolizumab with S-1 + oxaliplatin), revealed that patients possessing a CPS ≥ 10 exhibited an OS of 14.9 months (95% CI, 9.5-19.1),while those with a CPS < 10 demonstrated an OS of 17.7 months (95% CI, 13.4-23.7).Within cohort 2 (pembrolizumab with S-1 + cisplatin), patients with CPS ≥ 10 experienced an OS of 15.5 months (95% CI, 11.5-22.9),whereas those with CPS < 10 showcased an OS of 21.7 months (95% CI, 8.3-NE).Similarly, the median PFS in cohort 1 was recorded at 8.1 months (95% CI, 5.5-12.6)for CPS ≥ 10 patients and 12.6 months (95% CI, 6.6-NE) for CPS < 10 patients.As for cohort 2, CPS ≥ 10 patients achieved a median PFS of 7.0 months (95% CI, 5.1-NE), and CPS < 10 patients reached a median PFS of 14.8 months (95% CI, 5.8-16.4).Overall, the combination of pembrolizumab and chemotherapy showcased a favorable combination of effectiveness and safety for patients diagnosed with PD-L1-positive, HER2-negative GC or GEJC [30].
Numerous studies have also explored immunotherapy's potential as a neoadjuvant option for locally advanced GC.In a phase II exploratory trial (NCT03878472), the effectiveness of combining an ICI (camrelizumab), antiangiogenic agent (apatinib), and chemotherapy for neoadjuvant/conversion therapy in cT4a/bN+ GC was assessed.This study employed sequential multi-omics testing encompassing whole-exome sequencing (WES), transcriptome sequencing, and T cell receptor (TCR) sequencing.This extensive analysis unveiled potential biomarkers for assessing pathological responses and dynamic alterations in tumor immune microenvironments, and T cell receptor repertoires in the context of neoadjuvant immunotherapy.Correlations were established between pathological responses and MSI status, CPS, and TMB.Notably, potential biomarkers for pathological responses were identified, including RREB1 and SSPO mutations, immune-related signatures, and a peripheral T cell expansion score [32].In the Neo-PLANET phase II trial, neoadjuvant camrelizumab was examined alongside concurrent chemoradiotherapy for locally advanced GC/GEJC.Interestingly, the study discovered that the pathological complete response (pCR) rate in PD-L1-positive tumors (defined by CPS at 1, 5, or 10 cutoffs) did not significantly differ from PD-L1-negative tumors.However, analysis of somatic mutations through the WES of treatment samples revealed a notably higher pCR rate in patients with a pretreatment TMB ≥ median level (4.04 mutations/Mb) than those with TMB < median level [33].Additionally, a phase 2 clinical trial (NCT04065282) investigated the efficacy and safety of neoadjuvant sintilimab, oxaliplatin, and capecitabine in patients with locally advanced and resectable GC or GEJC.The patients were stratified based on their CPS levels (CPS ≥ 1, CPS ≥ 5, and CPS ≥ 10).The outcomes revealed that the combination of sintilimab with oxaliplatin and capecitabine demonstrated promising effectiveness, characterized by an encouraging rate of pathologic complete response (pCR), along with a favorable safety profile within the neoadjuvant context.These findings provide support for CPS as a predictive biomarker, aiding in the identification of patients likely to benefits from neoadjuvant anti-PD-1 treatment.Within the subgroup of patients with CPS < 1, pCR and major pathological response (MPR) rates stood at 9.1% and 27.3%, respectively.In contrast, among patients with CPS ≥ 1, the corresponding rates were 28.6% for pCR and 57.1% for MPR.It is worth noting that in the CPS ≥ 5 group, pCR and MPR rates were 27.3% and 54.5%, respectively.Among patients with CPS ≥ 10, pCR and MPR rates reached 33.3% and 50.0%, respectively [34].
The role of PD-L1 expression in GC immunotherapy's significance is a topic of ongoing debate.Data from the CheckMate-649, KEYNOTE-062, and KEYNOTE-590 trials reveal that combining ICI with chemotherapy does not confer benefits for AGC patients with low PD-L1 expression (CPS 1-4 in CheckMate-649 and CPS 1-9 in KEYNOTE-062) [49].Furthermore, a systematic review noted that TPS was the primary predictor of ICI benefit in squamous cell carcinoma patients, whereas CPS emerged as the strongest predictor for AGC [50].A meta-analysis demonstrated favorable OS and PFS when ICIs were combined with first-line chemotherapy for GC/GEJC patients, irrespective of CPS status [51].Another meta-analysis proposed CPS ≥ 1 as the threshold for ICI monotherapy's survival advantage [52].A thorough analysis incorporating CPS ≥ 10 GC patients' data from KEYNOTE-059, KEYNOTE-061, and KEYNOTE-062 consistently showcased improved clinical outcomes with pembrolizumab across treatment lines [53].Similarly, a multicenter biomarker cohort study of nivolumab treatment for GC identified CPS and MSI as independent yet valuable biomarkers for nivolumab response [54].
The reasons why PD-L1 cannot effectively predict the efficacy of immunotherapy may include the following: 1.
Tumor heterogeneity: PD-L1 expression can vary within the tumor and between different tumor sites, leading to inconsistent results when assessing its predictive value.Measuring PD-L1 on circulating tumor cells (CTCs) may address heterogeneity, but its consistency with tumor cell PD-L1 remains uncertain [8].

2.
Dynamic nature of PD-L1 expression: PD-L1 expression levels can change over time, making a single assessment unreliable for predicting long-term treatment outcomes.
To enhance PD-L1's predictive value, researchers suggest using multiple biopsies for improved accuracy [55].

3.
Immune microenvironment: the presence of other immune cells and their interactions within the tumor microenvironment can influence treatment response independently of PD-L1 expression [56].4.
Immunotherapy mechanisms: immunotherapies may work through multiple mechanisms; some patients may respond even if PD-L1 expression is low or absent [57].

5.
Sample collection and testing methods: variability in sample collection and testing methods can affect the accuracy of PD-L1 assessment [58].6.
Patient-specific factors: patient-specific factors, such as their overall health, prior treatments, and individual immune profiles, can influence treatment response independently of PD-L1 status [59].
Thus, due to these complexities, PD-L1 expression alone may not be sufficient to reliably predict the effectiveness of immunotherapy, and a more comprehensive approach considering multiple factors is often necessary.

Microsatellite Instability Status and Defective Mismatch Repair as the Biomarker for GC Immunotherapy
MSI-H or defective mismatch repair (dMMR) has emerged as a significant biomarker for predicting immunotherapy response in various solid tumors.In 2017, the FDA approved pembrolizumab for treating unresectable or metastatic solid tumors with MSI-H/dMMR status [60].And in the case of MSI-H/dMMR colon cancer patients, pembrolizumab has been employed as a first-line therapy.The dysregulated expression of MMR genes can impair cellular repair function during DNA replication, leading to MSI.The accumulation of mutations in tumor cells caused by a loss of MMR gene function is thought to make them more vulnerable to recognition by immune cells, resulting in a favorable response to immunotherapy [61].
MSI-H/dMMR is a significant molecular subtype of GC [3,4].This phenotype is present in approximately 10% of GC and GEJC patients, and it is more commonly detected in early-stage GC and elderly patients aged over 85 years [62].Several clinical studies have examined the response of MSI-H/dMMR GC/GEJC patients to ICI, as presented in Table 1.In the KEYNOTE-158 trial, which enrolled patients with MSI-H solid tumors, 51 dMMR/MSI-H GC patients exhibited a sustained response to pembrolizumab, with an ORR of 31.0%and a median OS of 11.0 months [37].Similarly, in the KEYNOTE-059 trial, MSI-H/dMMR GC patients showed a higher ORR of 57.1% compared to 9% in MSS/pMMR GC patients [15].Moreover, neoadjuvant treatment utilizing nivolumab combined with ipilimumab yielded a notably high rate of pCR, reaching 59%, among patients with resectable dMMR/MSI-H GC or GEJC [36].Subsequent to a post hoc analysis of the KEYNOTE-059, KEYNOTE-061, and KEYNOTE-062 trials, it was revealed that both pembrolizumab as monotherapy and the combination of pembrolizumab with chemotherapy displayed enduring anti-tumor effects in patients with advanced dMMR/MSI-H GC [63].A comprehensive meta-analysis encompassing randomized clinical trials such as KEYNOTE-062, CheckMate-649, JAVELIN Gastric 100, and KEYNOTE-061 reported that within a cohort of 123 patients diagnosed with MSI-H gastric cancer, the hazard ratio (HR) for overall survival advantage using anti-PD-1-based treatments was 0.34 for MSI-H individuals, in comparison to 0.85 for those with MSS status [64].
Despite these findings, the current sample size is relatively small to establish MSI-H/dMMR as a direct predictor of immunotherapy effectiveness in these patients, given the relatively lower incidence of MSI-H GC [65].IHC and polymerase chain reaction (PCR) are the commonly used methods for detecting MSI status [66]; however, MSI status exhibits heterogeneity [67], which can be influenced by multiple factors such as race, sampling factors, antibodies used, and subjective judgment by pathologists [68].Moreover, MSI status as a marker for immunotherapy response is only a rough indicator, as approximately 50% of MSI-H tumor patients demonstrate inherent resistance to PD-1 inhibitors.Therefore, further exploration of key immune response indicators in MSI-H patients is necessary.For example, in this context, Kwon et al. proposed that the T-cell receptor repertoire was associated with longer progression-free survival to pembrolizumab, and an increase in PD-1 + CD8 + T cells correlated with durable clinical benefit in MSI-H GC patients [69].

Tumor Mutation Burden (TMB) as the Biomarker for GC Immunotherapy
Tumor mutation burden (TMB) is a commonly used metric to quantify the number of mutations per megabase (Mb) of genomic DNA sequencing in tumors.This measurement represents the count of nonsynonymous mutations identified in the coding regions of a tumor genome.Tumors with a high TMB (TMB-H) are believed to have a greater number of neoantigens, which can potentially be recognized and targeted by the immune system.Consequently, patients with high-TMB tumors are considered more likely to benefit from immunotherapy treatments [70].Notably, in 2020, pembrolizumab monotherapy received accelerated FDA approval for previously treated, unresectable/metastatic TMB-H solid tumors (defined as TMB ≥ 10 mutations/Mb), based on the KEYNOTE-158 trial results.However, recent studies have raised concerns about relying solely on TMB-H as the predictive biomarker for tumor immunotherapy response [71,72].Several studies have also explored the potential of high TMB as a predictive biomarker in immunotherapy for gastric cancer (GC), as presented in Table 1.
In a phase Ib/II clinical trial (NCT02915432), the safety and efficacy of toripalimab, a humanized PD-1 antibody, were evaluated in patients with AGC.The study investigated the potential of PD-L1 expression and TMB as predictive biomarkers for toripalimab response.The results showed that patients with TMB-H (defined as TMB ≥ 12 mutations/Mb) had significantly better OS compared to those with low TMB, while PD-L1 overexpression did not correlate with survival benefit [31].An exploratory analysis of the KEYNOTE-061 trial revealed a strong association between TMB and the efficacy of second-line pembrolizumab in GC/GEJC patients.This study suggested that tissue TMB (tTMB) is a significant and independent predictor of pembrolizumab response, beyond PD-L1 status.Even after excluding patients with known MSI-H tumors, tTMB still showed a significant association with the pembrolizumab response [73].In prespecified exploratory analyses of EPOC1706, which tested lenvatinib plus pembrolizumab in AGC patients either in first-line or second-line therapeutic settings, objective responses were achieved by 84% of patients with CPS ≥ 1 and 40% of patients with CPS < 1.In the subgroup of patients with CPS ≥ 10, 100% had an objective response.For patients with high TMB (TMB ≥ 10), the objective response rate was 82%, and for patients with low TMB (TMB < 10), the objective response rate was 60%.The median progression-free survival was 9.1 months in patients with CPS ≥ 1 and 5.9 months in patients with CPS < 1.The median progression-free survival was 9.8 months for patients exhibiting high TMB, while those with low TMB registered a slightly shorter median PFS of 9.5 months [38].Notably, within the exploratory analysis of the KEYNOTE-062 clinical trial, an evident association emerged between TMB levels and the clinical effectiveness of first-line pembrolizumab in combination with chemotherapy among AGC patients.However, it is pertinent to highlight that the predictive prowess of TMB underwent attenuation upon the exclusion of patients with MSI-H tumors [41].
TMB's reliability is limited by several factors.Firstly, TMB only represents the mutation burden, but not all mutations generate new antigens.Therefore, TMB can only provide indirect evidence for predicting and assessing the generation of new antigens [71].Additionally, TMB is influenced by various factors such as tumor type, genetic alterations in the tumor microenvironment, exposure to external carcinogens, and detection methods.These factors can lead to TMB heterogeneity, which affects its accuracy as a predictive biomarker.TMB detection is also impacted by sample quality, detection methods, and analytical techniques.While WES is considered the gold standard for TMB detection, its high cost, time-consuming nature, and requirement for fresh specimens limit its use.Targeted sequencing panels offer a promising alternative to WES, allowing simultaneous analysis of multiple molecular indicators in addition to TMB, but they also require high sample purity [74].Currently, researchers are exploring the combination of TMB with other molecular indicators for screening and predicting the efficacy of immunotherapy in GC.For example, the study of Chida et al. proposed low TMB as a negative predictor of PD-1 blockade responses in MSI-H/dMMR gastrointestinal tumor patients [75], and Wang et al. suggested using the combination of blood MSI and blood TMB to screen GC patients for immunotherapy benefit [76].

Epstein-Barr Virus Status as the Biomarker for GC Immunotherapy
Epstein-Barr Virus (EBV) is a common human herpesvirus that infects over 90% of the population and has been associated with various malignancies, including GC [77].EBV-associated GC (EBVaGC) is a distinct subtype, accounting for approximately 9% of all GC cases.It exhibits unique features, such as occurring in younger patients, being more prevalent in males, primarily localizing in the upper part of the stomach, showing marked immune cell infiltration, and having a favorable prognosis [3,78].EBVaGC patients also display distinct immune characteristics, including changes in immune response genes, elevated PD-L1 expression in both cancer and immune cells, increased T and NK cell infiltration, heightened expression of immune checkpoint markers, and elevated levels of some anti-tumor immunity factors [79][80][81].As a result, EBV infection has been considered a positive factor for GC immunotherapy.
Some studies have indicated a positive correlation between EBVaGC and immunotherapy response (Table 1).In a prospective phase 2 clinical trial, all EBV-positive GC patients treated with pembrolizumab achieved a partial response (PR), with a longer median duration of response of 8.5 months [39].In another prospective observational study by Xie et al., nine patients with stage-IV EBVaGC were treated with ICIs, and three patients showed PR, while five patients had stable disease, all of whom displayed positive PD-L1 expression.This study suggested that combining EBV and PD-L1 might be a more accurate biomarker combination for determining the efficacy of immunotherapy in GC [42].However, a subsequent single-arm, phase 2 prospective clinical trial enrolled six EBV-positive mGC patients treated with camrelizumab, but none of them achieved an objective response, raising doubts about EBV positivity as a reliable predictor for immunotherapy response in mGC [43].Nakayama et al. found in their study that a high EBV copy number per genome (>10 copies) correlated with PD-L1 expression in tumor cells but poor disease survival in EBVaGC [82].
While EBV-positive is an important subtype of GC, and some EBV-positive AGC patients have shown a response to ICI, the predictive value of EBV positivity for ICI response in metastatic GC patients remains uncertain [83].Large-scale studies are needed to establish the relationship between EBV infection and GC immunotherapy response.Some existing studies propose using EBV as an adjuvant predictor of immune response in GC patients [84].For example, combining PD-L1 and CD8 could identify EBVaGCs with high immunoreactivity in patients with pMMR [85].A real-world study by Yu et al. found that better ORR and PFS were observed only in EBVaGC patients when they were combined with CPS ≥ 1 [86].Further research is necessary to validate these findings and explore the potential relationship between EBV status and immune status in GC patients.
The gold standard method for detecting EBV in tissue sections is the EBV-encoded RNA (EBER) in situ hybridization (ISH).However, the EBER probe used for ISH is usually expensive and requires a certain level of sensitivity.IHC is a less costly and more convenient test for EBV infection, which detects the LMP-1 membrane protein encoded by EBV.However, IHC cannot detect the location or transcriptional quantity of the virus.Novel test methods for EBV infection have also been proposed, such as the use of NGS for detection.These new methods not only improve the accuracy of EBV detection but also provide insights into the mechanism of EBV infection and the immunological characteristics of EBVGC, which are crucial for understanding the response of EBVaGC to immunotherapy.

Liquid Biopsy-Derived Predictive Biomarkers for GC Immunotherapy
Considering the temporal and spatial heterogeneity of GC, liquid biopsy-based blood predictive biomarkers have emerged as a promising approach for predicting immunotherapy response.Liquid biopsy offers several advantages, such as being safe, convenient, and repeatable, allowing the real-time monitoring of disease progression and treatment response.It can also detect low levels of tumor cells, enhancing sensitivity and accuracy.Among the extensively investigated biomarkers in liquid biopsy for GC are circulating tumor cells (CTCs), circulating tumor DNA (ctDNA), and extracellular vesicles [8].
Various studies have emphasized the crucial role of liquid biopsy-based blood biomarkers in predicting and monitoring immunotherapy response in GC (Table 1).For instance, Ishiba et al. analyzed PD-L1 expression in circulating tumor RNA (ctRNA) from various cancer types, including GC.They found a high degree of concordance between PD-L1 protein expression in tumor tissues and PD-L1 gene expression in plasma, proposing PD-L1 expression in ctRNA as a viable assay for predicting and monitoring immunotherapy response [87].Similarly, Jin et al. utilized NGS testing to conclude that dynamic ctDNA could serve as a potential biomarker for immunotherapy response in advanced GC, with decreasing ctDNA correlating to a higher response to treatment [88].Zhang et al. revealed actionable alterations for targeted and immune therapy in Chinese AGC patients using molecular characterization with ctDNA [89].Yue et al. demonstrated that measuring the abundance of PD-L1-high CTCs at baseline and monitoring their dynamic changes could indicate early therapeutic response for solid tumor patients [90].
Extracellular vesicles (EVs) have emerged as promising plasma components for monitoring immunotherapeutic outcomes in GC.Studies have reported a correlation between extracellular vesicle PD-L1 expression and immunotherapy response in GC.For instance, a study on AGC patients treated with nivolumab found that baseline soluble PD-L1 levels had the potential to predict survival [44].Zhang et al. identified four plasma EV-derived proteins (ARG1/CD3/PD-L1/PD-L2) as an EV-score that robustly predicts immunotherapeutic outcomes at baseline and monitors disease progression with ICI treatment in GC patients [91].In a phase Ib/2 trial, elevated plasma levels of PD-L1-expressing EVs were significantly associated with higher pCR among GEJC patients with PD-L1 CPS < 10 [35].
Despite its advantages, liquid biopsy as a screening method for immunotherapy predictive biomarkers in GC also has limitations.Firstly, biomarkers in blood are often not specific to the tumor, leading to possible false-positive results.Secondly, liquid biopsy may have sensitivity limitations and may not detect low concentrations of tumor biomarkers or DNA sequences.Additionally, different techniques and methods may yield varying results, and sample quality and processing can impact the outcomes, leading to technical limitations [92].Therefore, further improvements are necessary to enhance the reliability of liquid biopsy for screening immune therapy biomarkers in GC, and more research is needed to address these issues.Furthermore, liquid biopsy can be used as a supplementary detection method in combination with other biomarkers to predict immune therapy responsiveness in GC.

Predictive Role of Gut Microbiota and Helicobacter pylori Infection in GC Immunotherapy
The gut microbiota plays a crucial role in maintaining overall health and influencing disease progression, including tumorigenesis.Recent research has highlighted its association with tumor progression and therapeutic efficacy, and it has been proposed as a potential biomarker to predict outcomes of immunotherapy for solid tumors [93,94].Despite the known roles of certain gut microbiota in the development and immune regulation of GC, there still a lack of evidence on their potential as biomarkers for GC immunotherapy efficacy.Further research is needed to explore the value of gut microbiota in the context of GC immunotherapy.
One common infecting microorganism in GC patients is Helicobacter pylori (H.pylori).Its presence is closely associated with the development of GC, and H. pylori's role in regulating the immune system has been proposed.Moreover, H. pylori can alter the immune microenvironment of GC, potentially impacting immunotherapy efficacy [95,96].Study in mouse models of colon adenocarcinoma and melanoma have demonstrated the detrimental effect of H. pylori infection on immunotherapy [97].Additionally, a retrospective analysis suggested an association between H. pylori infection and the outcome of immunotherapy for AGC patients [45].However, this study had a limited sample size, and there is a lack of prospective studies on the prognostic implications of H. pylori in GC immunotherapy.Furthermore, the impact of H. pylori infection on immunity may vary significantly among individuals due to the complex infection process.Therefore, H. pylori infection cannot be used as the sole biomarker for population screening or predicting the efficacy of GC immunotherapy beneficiaries.Further research is necessary to better understand its role in the context of immunotherapy for GC.

Predictive Biomarkers of Tumor Microenvironment in GC Immunotherapy
The effectiveness of immunotherapy for tumors is directly influenced by the TME.Recent advances in omics technologies such as genomics, transcriptomics, and metabolomics have led to the identification of multiple TME-related biomarkers for GC immunotherapy.For example, Hu et al. employed a comprehensive multi-omics approach, including transcriptomics RNA-sequencing (mRNA, LncRNA, miRNA), DNA methylation, and gene mutations, to propose two distinct molecular subtypes (CS1 and CS2) of GC.They noted that the CS2 group exhibited higher immunocyte infiltration, suggesting potential immunotherapeutic benefits within this subgroup [99].predictive capabilities for GC.They demonstrated that these molecular subtypes could forecast responses to adjuvant chemotherapy after gastrectomy and to immunotherapy in patients with metastatic or recurrent GC [107].
The efficacy of tumor immunotherapy is linked to immune cell quantity, their spatial relationships, and TME organization [108,109].Tumor-infiltrating lymphocytes (TILs), comprising T cells, B cells, and NK cells, play a pivotal role in immune responses against tumors and are associated with immunotherapy efficacy [110].The retrospective study of Tong et al. proposed intertumoral CD8+ TILs as a predictive factor for chemoimmunotherapy response in PD-L1-negative AGC [46].Distinctive TME structures, like tertiary lymphoid structures (TLS), have also been correlated with positive immunotherapy outcomes [111].TLS presence is also believed to predict responses to anti-PD1 therapy in GC [112].Novel immune checkpoints like CD96 and CD73 have also been proposed as potential GC immunotherapy biomarkers.CD96 could guide precision medicine in adjuvant chemotherapy, immunotherapy, and targeted therapies [113].High CD73 expression indicates better chemotherapeutic responsiveness but a poorer pembrolizumab response in GC [114].
mIHC and multiple immunofluorescence (mIF) are advanced techniques that significantly contribute to the understanding of TME immune status, offering insights into the immunological landscape.A meta-analysis across ten cancer types treated with ICI highlighted the superiority of mIHC/mIF over TMB, gene expression profiling, and PD-L1 expression biomarkers in predicting immunotherapy response [115].In a recent study, Chen et al. employed mIHC and multi-dimensional analyses to predict immunotherapy response in GC.Their findings suggest that the TIL signature identified via mIHC holds promise as a prognostic biomarker for anti-PD-1/PD-L1 immunotherapy response and OS prediction [116].
These findings underscore multi-omics technologies' role in understanding GC's TME characteristics, enabling sensitive tumor-biomarker discovery.Yet, limited adoption of these techniques and reliance on small cohorts or database data constrain reliability.Further indepth research is essential to uncover predictive TME biomarkers for GC immunotherapy.

Predictive Biomarkers of Immunotherapy-Related Adverse Events in GC
Tumor immunotherapy-related adverse events (irAEs) refer to adverse reactions associated with tumor immunotherapy.The mechanisms underlying the development of irAEs remain unclear at present.These adverse events are currently believed to arise from an exaggerated response of the immune system against normal tissues rather than being directly caused by the tumor itself.irAEs can affect multiple organ systems, including the skin, gastrointestinal tract, liver, and lungs, with the severity varying among individuals.Treatment typically involves the reduction in the immunotherapeutic agent's dosage or temporary suspension of treatment to alleviate these adverse reactions [117].In the context of GC immunotherapy, the incidence of irAEs is not infrequent.Moreover, severe irAEs possess the capacity to pose a threat to patients' overall survival.
Several studies have suggested a potential association between irAEs and a more favorable response to immunotherapy in patients [118,119].Additionally, research has demonstrated a correlation between the occurrence of irAEs following immunotherapy and improved survival outcomes in GC patients [120][121][122].Furthermore, certain studies have identified potential biomarkers for predicting the occurrence of irAEs [123].For example, Jiang et al. explored the utility of biomarkers for irAE prediction by focusing on EV-derived proteins.They collected dynamic plasma samples from 102 GC patients who received ICIs and screened two EV-derived proteins, namely, inducible T-cell co-stimulator (EV-ICOS) and indoleamine 2,3-dioxygenase 1 (EV-IDO1), to prognosticate the development of irAEs [124].Additionally, Jin et al. found that GC patients with genetic alterations in the CEBPA, FGFR4, MET, or KMT2B genes exhibited an increased likelihood of experiencing irAEs [88].
While research on biomarkers related to adverse events in gastric cancer immunotherapy is currently limited, investigating biomarkers for irAEs holds significant promise.This research can aid in the identification of individuals at risk for irAEs, enabling the mitigation of adverse event risks associated with immunotherapy.Additionally, it can contribute to the precise selection of the most suitable candidates for GC immunotherapy, making it one of the crucial research directions in the field of GC immunotherapy.

Conclusions
In recent years, standard treatment approaches for GC patients, including neoadjuvant therapy and advanced-stage cases, have recommended the use of single-agent and combination immunotherapy.However, clinical outcomes have demonstrated that a significant proportion of GC patients do not experience the benefits of immunotherapy, and adverse events are not uncommon.Therefore, the precise selection of GC patients who could benefit from immunotherapy is of utmost importance.In this comprehensive review, we have summarized various immunotherapy efficacy assessment markers reported in current clinical studies.These findings suggest that PD-L1 expression levels remain a widely used clinical biomarker for evaluating immunotherapy efficacy in GC.However, further attention is warranted to refine the criteria for its selection, particularly considering the variations associated with standard treatment protocols, intertumoral heterogeneity, and detection methods.While MSI status, EBV infection, and TMB levels have shown promising roles as predictive factors in certain GC immunotherapy clinical trials, large-scale clinical research is still required for further validation.The advent of novel techniques such as liquid biopsy and multi-omics studies brings renewed hope to the investigation of predictive biomarkers for GC immunotherapy efficacy.In this context, the exploration of combinations of molecular markers and personalized testing emerges as a crucial research direction in the field of GC immunotherapy.
Findings by Derks et al. demonstrated the heterogeneous nature of the tumor-immune microenvironment in GC, highlighting the pivotal role of TME in the immunotherapeutic response to GC [98].
Fu et al. compared gene expression and immune markers between low-and high-TMB groups.They developed an immune prognostic model for GC based on their findings [100].Yuan et al. integrated transcriptomics, proteomics, and metabolomics data, revealing HER2-associated metabolic heterogeneity in GC linked to responses to immunotherapy and neoadjuvant chemotherapy.Their study highlighted the quiescent and aspartate and glutamate subtypes as more likely to benefit from immunotherapy [101].Wang et al. proposed a gene-based antigen processing and presentation signature (APscore) for prognostication and prediction of response to immune checkpoint inhibitors (ICIs) in AGC patients [102].Shi et al. characterized glycometabolism and the tumor immune microenvironment to predict clinical outcomes in GC.They identified patients with a high Gluco-Immune Score as having potential for benefiting from immunotherapy [103].Zeng et al. developed an open-source TMEscore R package, validated in a prospective phase 2 clinical trial of mGC patients treated with pembrolizumab, demonstrating the TME score's superior accuracy over CPS, TMB, MSI, and EBV [104].He et al. identified molecular features correlating with tumor immunity in GC, providing potential biomarkers for stratifying GC patients who are responsive to immunotherapy [105].Chen et al. leveraged deep learning to propose immune subtypes and landscape analysis of GC based on whole-slide images, offering insights to improve immunotherapeutic strategies [106].Cheong et al. established a 32-gene signature with prognostic and

Table 1 .
Summary of biomarkers for colon cancer immunotherapy in reported clinical trials.

Table 1 .
Summary of biomarkers for colon cancer immunotherapy in reported clinical trials.