Systematic single-cell dissecting reveals heterogeneous oncofetal reprogramming in the tumor microenvironment of gastric cancer

Oncofetal reprogramming of the tumor microenvironment is clinically relevant. This study used the non-negative matrix factorial (NMF) algorithm for single-cell RNA sequencing data of gastric cancer (GC) based on embryonic stem genes. Pseudotime analysis, cell–cell interaction analysis, and SCENIC analysis revealed that cancer-associated fibroblasts (CAFs), tumor-associated endothelial cells (TECs), and tumor-associated macrophages (TAMs) have different oncofetal reprogramming that affects cell function, enhances intercellular communication, and activates multiple transcription factors in these cells. Furthermore, based on the signatures of the newly defined oncofetal cell subtypes and expression profiles of large cohorts in GC patients, we determined that GJA1 + TEC-C2, IFITM1 + CAF-C3, PODXL + TEC-C1, SFRP2 + CAF-C2, and SRSF7 + CAF-C1 are crucial prognostic factors for GC patients and predictors of immune checkpoint blockade in GC. Cell subtypes were validated by immunohistochemical methods. Our novel, profound, and systematic analysis of the oncofetal reprogramming of GC may facilitate the development of improved drugs for treating GC.


Introduction
Gastric cancer (GC) is a major cause of cancer-related morbidity and mortality worldwide [1]. Despite several studies on the prevention and treatment of gastric cancer, it remains a significant health burden with poor prognosis, especially in East Asia, East Europe, and parts of South America [2]. Therefore, there is an urgent need for a deeper understanding of GC molecular biology. Fortunately, various genomic approaches may help unravel the complexity and heterogeneity of this disease and lead to new individualized strategies for GC treatment.
Oncofetal reprogramming of the tumor microenvironment is attracting increasing attention from researchers. Oncogenesis and embryogenesis share many similarities, including rapid cell division, adaptive plasticity, and a highly vascularized environment [3]. Oncofetal reprogramming is characterized by the re-expression of fetal genes and proteins by malignant cells [4]. Oncofetal reprogramming also occurs in the tumor microenvironment (TME), which is composed of T cells, B cells, tumor-associated macrophages (TAMs), cancer-associated fibroblasts (CAF) and endothelial cells [5,6]. The TME is characterized by a Shaocong Mo, Xin Shen and Yulin Wang have contributed equally to the work.
tumor-promoting and immunosuppressive phenotype [7]. Further investigation of oncofetal reprogramming could contribute to understanding cancer development and evolution, stratifying patient prognosis, and guiding therapies targeting specific components in the TME. Fortunately, advances in single-cell transcriptomics facilitate the confirmation of embryonic development and tumor formation, as well as the discovery of complex intercellular communication among the diverse oncofetal reprogrammingassociated subtypes of TME cells and tumor cells, which remains a poorly researched area.
In this study, we applied non-negative matrix factorization (NMF) based on embryonic stem genes to reveal oncofetal reprogramming in the GC tumor microenvironment using GC single-cell RNA sequencing data. Cell-cell interaction and SCENIC transcription factor analyses revealed the particularity of key oncofetal cell subtypes, such as SRSF7 + CAF-C1 and GJA1 + TEC-C2, which was verified through IHC. Moreover, combined with the bulk expression profiles of large GC cohorts, we verified that multiple oncofetal cell subtypes impacted cancer prognosis and the outcome of immune checkpoint blockade therapy in GC patients.

Data acquisition and processing
Single-cell RNA sequencing data (GSE167297) were downloaded from the Gene Expression Omnibus (GEO) database (www. ncbi. nlm. nih. gov/ geo). The Seurat pipeline was applied to GSE167292. The Seurat objects underwent normalization, scaling, and dimensional reduction. We referred to common tumor microenvironment cell markers for cell-type annotation. Epithelial cells, T cells, B cells, myeloid cells, fibroblasts, and endothelial cells were identified. The expression profile of The Cancer Genome Atlas (TCGA) was downloaded from UCSC Xena in the format of fragments per kilobase of exon model per million mapped fragments (FPKM) along with the survival information. The data were further transformed into transcripts per kilobase of exon model per million mapped reads (TPM) and logged. The expression profiles and survival data of the Asian Cancer Research Group (ACRG) gastric cancer cohort were downloaded from the GEO database. The embryonic stem gene set was obtained from Msigdb (https:// www. gsea-msigdb. org/ gsea/ msigdb/ human/ genes et/ BHATT ACHAR YA_ EMBRY ONIC_ STEM_ CELL). Immunohistochemical (IHC) staining images were obtained from The Human Protein Atlas (HPA) database (https:// www. prote inatl as. org) [8].

Single-cell non-negative matrix factorization (NMF)
For single-cell NMF, normalized data were extracted from the Seurat object [9]. Embryonic stem genes were used for the NMF [10]. Genes that were not expressed in any cell type were deleted. In addition, cells without embryonic stem gene expression were removed. A maximum of ten clusters was tolerable with the NMF method set to snmf/r. The NMF result was further added to the Seurat object for dimensional reduction. Feature genes of each NMF cluster were screened using FindAllMarker. Only clusters with embryonic stem genes with log2(fold change) greater than 1 were named "Gene + Cell type". Clusters with no characteristic embryonic stem genes were named "Non-Embryonic-Cell type". Clusters with characteristic embryonic stem genes with log2(fold change) less than 1 were named "Unclear-Cell type".

Pseudotime analysis, cell-cell interaction analysis, and transcription factor analysis
Pseudotime analysis was conducted with the Monocle2 package, as in previous studies [11]. Briefly, size factors and dispersions were estimated, and highly variable features were identified in the Monocle object. Subsequently, the dimensions were reduced, and the cells were ordered to perform pseudotime visualization. Cell-cell interactions were analyzed using Cellchat [12,13]. Secreted signaling in humans was included in the cell-cell interaction analysis [14]. The SCENIC package [15] was used for transcription factor analysis. CisTarget database of [16] hg19-tss-centered-10 kb-7species.
Further, mc9nr.feather was used as a reference file. For each transcription factor, genes with the top 50 weights were retained to construct the co-expression network. The area under the curve (AUC) of each regulon was further calculated and averaged for comparison between oncofetal subtypes.

Gene set scoring
The GSVA package was applied for gene set scoring in bulk RNA sequencing or microarray. Single-sample gene set enrichment analysis (ssGSEA) was performed using the GSVA algorithm [14,17,18]. For single-cell RNA sequencing data, GSVA was used to evaluate the known CAF subtype from a previous study [19]. The AddModuleScore was used to assess the M1-like/M2-like polarization markers obtained from previous studies [20]. The scMetabolism package was used to evaluate the metabolic scores of the cells. Additionally, the Kyoto Encyclopedia of Genes and Genomes (KEGG) was conducted on the feature genes using the clusterProfiler package.

Statistical analysis
The Wilcoxon test was used to compare the continuous variables. Cox regression analysis was conducted using the survival package, and the glm function was applied for logistic regression. The log-rank P test was used for the Kaplan-Meier analysis. Statistical significance was set at P < 0.05. All analyses were performed using R (4.1.1).

Embryonic stem genes shared high heterogeneity among cells
The overall design of the study is shown in the workflow chart (Fig. 1A). First, using TCGA tumor samples and matched normal samples, we discovered that the embryonic stem cell score was significantly altered in tumor samples with ssGSEA, indicating the importance of oncofetal reprogramming in gastric cancer carcinogenesis (Fig. 1B). Subsequently, we applied GSE167297, which contained scRNA-seq of matched normal and tumor tissues of GC, to explore oncofetal genes at the single-cell level. A heatmap revealed that many embryonic stem genes demonstrated high heterogeneity among cells. For instance, SERPINH1 shared higher expression levels in endothelial cells and fibroblasts, but showed lower expression levels in other cells, indicating the necessity for further investigation of subpopulations of cell types (Fig. 1C). Six genes were differentially expressed between the normal and tumor groups, which indicated the need for further subtyping and exploration of these oncofetal genes by focusing on a single cell type (Fig. 1D). We also showed that FABP5, another embryonic stem gene, was highly expressed in myeloid and endothelial cells of tumor samples, which could also serve as a potential subtyping feature (Fig. 1E).

Cancer-associated fibroblasts (CAFs) showed different directions of oncofetal reprogramming
CAF has recently been recognized as one of the core factors in the TME [24]. Thus, we focused on the oncofetal reprogramming of CAFs. Interestingly, pseudotime analysis revealed that embryonic stem genes were expressed at different stages of development ( Fig. 2A). For example. FABP5 is a feature of early development, whereas SRSF7 is a feature of late development, indicating the crucial role and complexity of the oncofetal reprogramming of CAFs. Then, using NMF clustering, we identified SRSF7 + CAF-C1, SFRP2 + CAF-C2, IFITM1 + CAF-C3, and non-embryonic-CAF-C4 subtypes in the CAFs (Fig. 2B, Table S1). Using the Cellchat package, we determined that SRSF7 + CAF-C1, SFRP2 + CAF-C2, and IFITM1 + CAF-C3 had more powerful interactions with other cell types than C4 (Fig. 2C,  Fig. S1A). Thus, we hypothesized that SRSF7 + CAF-C1 and SFRP2 + CAF-C2 IFITM1 + CAF-C3 might have some specific activation of transcription factors (TFs) to promote cell talk. We then discovered that specific TFs such as ATF4, NFIB, and STAT1 were activated in SRSF7 + CAF-C1, SFRP2 + CAF-C2, and IFITM1 + CAF-C3, respectively, whereas non-embryonic-CAF-C4 lacked the function of TFs (Fig. 2D, Fig. S1B). To further investigate the functions of the newly identified oncofetal CAF subtypes, we calculated the average GSVA scores of the known CAF subtypes. SFRP2 + CAF-C2 had the highest scores for pan-iCAF, pan-dCAF, and pan-pCAF [25,26] (Fig. 2E). Finally, we compared key CAF phenotype markers, such as pro-inflammatory genes, neo-angiogenic genes, and MMPs. We found that pro-inflammatory genes were predominantly expressed in the SRSF7 + CAF-C1 and SFRP2 + CAF-C2 subtypes (Fig. 2F). Together, we demonstrated the different directions of oncofetal reprogramming in CAFs.

Tumor-associated macrophages (TAMs) exhibit differences in metabolism and polarization during oncofetal reprogramming
Next, we investigated whether oncofetal reprogramming influences the functions and phenotypes of TAMs. First, we successfully identified TAMs from myeloid cells that had high expression levels of TAM marker genes (CD68, C1QA, APOC1), but low expression levels of monocyte Fig. 1 Embryonic stem gene expression was highly heterogeneous among cells. A Workflow of the study. B GSVA scoring of embryonic stem gene set between normal stomach tissue and GC tissue. C Expression of embryonic stem genes in all types of cells in GSE167297. D Differentially expressed embryonic stem genes between cells from normal tissues and those from tumors in GSE167297 (P value < 0.05). E Expression of an example embryonic stem gene (FABP5) in different types of cells markers (FCN1, S100A9, S100A12) (Fig. 4A, Fig. S3A, S3B). Pseudotime analysis indicated that embryonic stem genes were expressed at different stages of macrophage development (Fig. S3C). NMF with embryonic stem genes separated TAMs into SMS + Mac-C1, FABP5 + Mac-C2, TUBB + Mac-C3, non-embryonic-Mac-C4, and Unclear-Mac-C5 (Fig. 4B, Table S3). Further, SMS + Mac-C1, FABP5 + Mac-C2, and TUBB + Mac-C3 had relatively more robust interactions with other cell types (Fig. 4C,   Fig. S3D). TFs analysis revealed that compared with nonembryonic-Mac-C4, oncofetal TAMs C1-C3 all exhibited high TF activity (Fig. 4D). As TAMs undergo metabolic reprogramming in the tumor microenvironment, and protumor macrophages tend to have an overactivated metabolism [28], we applied scMetabolism to evaluate the metabolic heterogeneity in TAM subtypes. Interestingly, we determined that SMS + Mac-C1 and TUBB + Mac-C3 had the highest metabolic activity compared to the other groups, while Embryonic + Mac-C4 and Unclear-Mac-C5 presented a metabolic downturn (Fig. 4E). Furthermore, we calculated the M1-like/M2-like polarization scores of different oncofetal TAM subtypes. Notably, SMS + Mac-C1 had the highest M2 polarization score, while Unclear-Mac-C5 tended to be M1-like, indicating that TAMs could be induced to the M2 phenotype during oncofetal reprogramming (Fig. 4F).

Multiple oncofetal cell subtypes influenced the prognosis of GC patients
Furthermore, we aimed to determine whether the aforementioned newly defined oncofetal subtypes indeed influenced the survival of patients with GC. First, we showed the cell-cell communication between all oncofetal subtypes, which described all the potential relationships  (Fig. 5A). In the TCGA dataset, we first used ssGSEA to calculate the enrichment score of each oncofetal subtype. We discovered that GJA1 + TEC-C2, IFITM1 + CAF-C3, PODXL + TEC-C1, SFRP2 + CAF-C2, and SRSF7 + CAF-C1 could distinguish patient survival (Fig. 5B, Fig. S4A). Notably, in the ACRG cohort, all cell subtypes could distinguish survival; exemplar genes are shown in Fig. 5B (Fig. 5B, Fig. S4B). Using univariate Cox regression, we obtained the hazard ratio of each oncofetal subtype in TCGA and ACRG cohorts, which showed that GJA1 + TEC-C2, IFITM1 + CAF-C3, PODXL + TEC-C1, SFRP2 + CAF-C2, and SRSF7 + CAF-C1 were poor prognostic factors (Fig. 5C, Table S4, Table S5). Subsequently, we used multivariate COX regression to combine the five subtypes into a prognostic signature trained in the TCGA cohort (Table S6). It was reassuring to observe that the signature also showed prognostic value in the validation ACRG cohort (Fig. 5D).

Oncofetal cell subtypes predicted drug sensitivities and outcomes of immune checkpoint blockade (ICB) therapy
As oncofetal reprogramming influences the tumor environment, we next tried to determine whether oncofetal subtypes might predict the outcomes of chemotherapeutic agents and immune checkpoint blockade (ICB) therapy. First, we used the oncoPredict method to assess the impact of oncofetal reprogramming cell subpopulations on the sensitivity of more than 200 small molecule chemotherapeutic agents and targeted therapeutics. We discovered that the cell subpopulations GJA1 + TEC-C2, IFITM1 + CAF-C3, PODXL + TEC-C1, SFRP2 + CAF-C2, and SRSF7 + CAF-C1 were associated with increased sensitivity to the chemotherapeutic agent PI3β inhibitor AZD6482 and decreased sensitivity to the EGFR inhibitor Afatinib (Fig. S4C).Then, we applied the TIDE algorithm to assess the ICB treatment response of each patient in TCGA and ACRG. Then, we noticed that in the TCGA cohort, SRSF7 + CAF-C1, SFRP2 + CAF-C2, IFITM1 + CAF-C3, PODXL + TEC-C1, GJA1 + TEC-C2, FABP5 + TEC-C3, SRSF7 + TEC-C4 were all downregulated in responders, indicating that these cells might be associated with ICB resistance, which was verified in the ACRG cohort (Fig. 6A,B). Using logistic regression, the odds ratio value to predict responses was obtained, revealing the adverse role of these cells (Fig. 6C, Table S7, Table S8). To evaluate the functions of oncofetal subtypes, we also used the IMvigor210 bladder cancer cohort to compare cell abundance between SD/PD and CR/PR patients. We also found that SFRP2 + CAF-C2 and PODXL + TEC-C1 were downregulated in CR/PR patients (Fig. 6D). Collectively, some oncofetal cells had stable predictive values for the outcomes of ICB treatment. To verify the presence of the above cell types, we used IHC to observe cellular  A An abundance of oncofetal cell subtypes, such as SRSF7 + CAF-C1 and PODXL + TEC-C1, differed between responders and non-responders of ICB in TCGA, as predicted by TIDE. B As predicted by TIDE, an abundance of oncofetal cell subtypes differed between responders and non-responders of ICB in ACRG. C Odd ratio produced by logistic regression of different subtypes in TCGA and ACRG cohorts. D Multiple oncofetal cell subtypes showed different infiltration scoring between SD/PD and CR/PR patients in the IMvigor210 cohort. E IHC staining images of four crucial embryonic stem cell genes. Panel1: GJA1 + endothe-lial cells were marked with a black arrow. Protein was mainly expressed on cell membrane (source HPA ID: GCAB010753. Patient ID: 2066, male, age 76). Panel2: IFITM1 + fibroblasts were marked with a black arrow. Protein was mainly expressed on cell membrane (source HPA ID: HPA004810. Patient ID: 2105, male, age 62 years). Panel3: PODXL + endothelial cells were marked with a black arrow. Protein was mainly expressed on cell membrane (source HPA ID: HPA002110. Patient ID: 2626, female, age 79 years). Panel4: SRSF7 + fibroblasts were marked with black arrows. Protein was mainly expressed in the nucleus (source HPA ID: HPA056926. Patient ID: 650, male, age 68 years) localization of GJA1, IFITM1, PODXL and SRSF7. We discovered that some of the vascular endothelial cells in stromal did express GJA1. Interestingly, compared to other vascular endothelium, GJA1 + endothelial cells were rich in immune cells inside, which was consistent with the aforementioned results of cell-cell communication (Fig. 6E, first  panel). Similarly, PODXL + TEC was observed, especially in some micro vessels, showing significantly strong positivity, from which immune cells were seen to exude (Fig. 6E,  third panel). IFITM1 was expressed on the cell membrane of some CAFs, with vascular endothelium nearby (Fig. 6E,  second panel). As a splicing factor, SRSF7 was expressed in the nucleus of some CAFs. As a potential iCAF, we also observed a high abundance of immune cells around SRSF + CAFs in pathological sections (Fig. 6E, fourth panel).

Discussion
Oncofetal reprogramming of the TME is of growing concern because of its contribution to cancer promotion, immune evasion, and cancer metastasis [4]. However, few studies have investigated the potential cancer-promoting role and cell type-specific patterns of embryonic stem genes using single-cell genomics. In this study, we comprehensively explored oncofetal reprogramming in the main components of TME. We revealed the intricate intercellular interaction between oncofetal reprogramming-associated TME cell subtypes with single-cell sequencing as validate the cell types through IHC. Our study provides a novel perspective for understanding how the cell-specific expression pattern of embryonic stem genes shapes the TME, thus affecting the prognosis and outcomes of immune checkpoint blockade therapy in individual GC patients.
CAFs, as highly versatile and plastic cell components in the TME, are critical in cancer progression, possibly through complicated communications with other components in the TME [29]. Based on their specific molecular characteristics, CAFs are classified into pan-myCAFs, pan-dCAFs, pan-iCAFs, pan-nCAFs, and pan-pCAFs [19]. To date, no study has reported the expression patterns of embryonic stem genes in CAFs. Our study identified three oncofetal reprogramming-related CAF subtypes: SPSF7-CAF-C1, SFRP2 + CAF-C2, IFITM1 + CAF-C3, and non-embryonic-CAF-C4, as well as the interaction between these subtypes and other components in the TME. We observed that oncofetal reprogramming-related fibroblasts manifested more extensive communication with other components than non-oncofetal reprogramming-related fibroblasts. We highlighted SFRP2 + CAF-C2 because of its high correlation with pan-dCAF, pan-iCAF, and pan-pCAF, along with elevated expression levels of IL-7, IL33, and CXCL12, which are well-recognized pro-inflammatory genes [30][31][32]. In addition, we emphasized three oncofetal reprogrammingrelated CAF subtypes for their outstanding performance in distinguishing the survival of patients with GC. Therefore, we hypothesized that oncofetal reprogramming might affect the function and phenotype of CAFs, which subsequently facilitates the immunosuppressive TME to accelerate the malignant progression and metastasis of GC.
TECs play critical roles in cancer progression and therapy efficacy [33]. Embryonic stem genes participate in various stages of TEC development. The metabolic regulation of TECs is emerging as a research hotspot because of its critical role in nourishing cancer cells and facilitating tumor metastasis [34]. Notably, the LGALS9 pathway was specifically activated in oncofetal TEC subgroups, which is consistent with previous reports associating the LGALS9 pathway with metastasis and immunosuppression [35]. In addition, the MDK pathway, a well-characterized pathway for enhancing the proliferation of endothelial cells and inducing cancer angiogenesis [36], was also elevated in the oncofetal TEC subgroups.
Furthermore, oncofetal TEC subtypes exhibit a proinflammatory phenotype. In previous studies, proinflammatory TECs have been shown to promote cancer cell intravasation, lung colonization, and postsurgical metastasis [37]. Taken together, oncofetal reprogramming may reshape TECs into a more oncogenic phenotype that promotes malignant transformation, tumor invasion, and metastasis.
Recent research is increasingly concerned with oncofetal reprogramming in immune cell components within the TME, especially tumor-associated macrophages (TAMs) [38]. High infiltration of TAMs usually indicates poor prognosis and therapy resistance [39]. Using NMF clustering, we classified TAMs into five subclusters, and all five subclusters showed broad communication with other TME constituents. The metabolic characteristics and functions of TAMs are critical for tumor progression. We demonstrated the divergent metabolic statuses of the five oncofetal TAM subtypes.
Interestingly, SMS + Mac-C1 and TUBB + Mac-C3, members of the oncofetal reprogramming-related TAM subtypes, showed significant activation of various metabolic pathways, including amino sugar and nucleotide sugar metabolism, butanoate metabolism, pentose, and glucuronate interconversion, and steroid hormone biosynthesis. M1-like/ M2-like phenotype scores were calculated, indicating diverse pro-oncogenic or tumor-suppressive roles among the different TAM subtypes. In summary, we demonstrated that embryonic stem genes modulate the functional and metabolic characteristics of TAMs, indicating a mechanism of immune evasion mediated by TAMs in GC.
We analyzed TFs at the single-cell level to investigate cell-specific gene regulatory networks. In general, each subtype of CAFs, TAMs, and TECs exhibited distinct TFs characteristics. For CAFs, TFs such as ATF4, NFIB, and STAT1 are specifically activated in different oncofetal reprogramming-related CAF subclusters. Among them, NFIB has been identified as a putative target of oncofetal miRNAs and is associated with tumor aggressiveness in lung adenocarcinoma [40], indicating that the embryonic stem gene signature in CAFs, particularly IFITM1 + CAF-C3, plays a role in tumor progression. Furthermore, for TECs, high activities of ATF3, IRF1, FOSB, EGFR1, JUNB, JUN, ATF4, JUND, and ELF1 were observed in oncofetal reprogramming-related TEC subgroups. Notably, JUNB, JUN, and JUND have been reported to regulate IGF2BP1, an oncofetal protein [41], thus facilitating epithelial-mesenchymal transition in various cancer types. Moreover, TAMs, including NFE2L2, JUND, REL, NFKB1, CEBPB, and BHLHE40, were upregulated in oncofetal TAM subgroups. Another notable one was NFE2L2, which was reported to be related to higher expression levels of the oncofetal proteins AFP and GPC3, along with a poorer prognosis in hepatoblastoma [42]. In summary, oncofetal cell subtypes may orchestrate complex TF regulatory networks to mold an immunosuppressive microenvironment, thus promoting GC development. Notably, oncofetal cell subtypes, including CAFs, TECs, and TAMs, demonstrated more and more robust network connections than their non-embryonic or unclear cell counterparts, suggesting that oncofetal reprogramming occurring in various cell types might potentially contribute to the formation of an immunosuppressive TME.
Finally, we comprehensively evaluated the predictive value of oncofetal TME subtypes for GC prognosis and outcomes of ICB therapy using public bulk RNAseq cohorts. Oncofetal TME subtypes, especially GJA1 + TEC-C2, IFITM1 + CAF-C3, PODXL + TEC-C1, SFRP2 + CAF-C2, and SRSF7 + CAF-C1, showed favorable performance in differentiating the survival and ICB response of GC patients as poor prognostic and ICIresistant factors. Therefore, further investigation into the oncofetal reprogramming of the TME has practical application value for patients with GC.

Conclusions
In this study, we performed a comprehensive analysis of the oncofetal reprogramming of GC at the single-cell level. Novel oncofetal cell subtypes were screened based on the NMF algorithm, providing crucial markers for the prognosis and ICB therapy in patients with GC. This study greatly deepens our understanding of oncofetal reprogramming in the tumor microenvironment of GC.

Shortcomings and future directions
Our findings were mainly restricted by the low sequencing depth of scRNA-seq and suboptimal sample size. Thus, our findings need to be validated in larger cohorts of patients with GC. Compared with bulk RNA-seq data, our data may have clustering bias because of the considerable zero observations of some embryonic stem genes. Nonetheless, our scRNA-seq analysis still provides a novel perspective to reveal the characteristics of oncofetal reprogramming in various TME cell types that promote immune suppression and malignant progression. These distinctive cell features in TME may serve as predictive markers for survival and ICI efficacy in GC patients and may be clinically targeted in the future.