Loss of ATOH1 in Pit Cell Drives Stemness and Progression of Gastric Adenocarcinoma by Activating AKT/mTOR Signaling through GAS1

Abstract

Gastric cancer stem cells (GCSCs) are self-renewing tumor cells that govern chemoresistance in gastric adenocarcinoma (GAC), whereas their regulatory mechanisms remain elusive. Here, the study aims to elucidate the role of ATOH1 in the maintenance of GCSCs. The preclinical model and GAC sample analysis indicate that ATOH1 deficiency is correlated with poor GAC prognosis and chemoresistance. ScRNA-seq reveals that ATOH1 is downregulated in the pit cells of GAC compared with those in paracarcinoma samples. Lineage tracing reveals that Atoh1 deletion strongly confers pit cell stemness. ATOH1 depletion significantly accelerates cancer stemness and chemoresistance in Tff1-CreERT2; Rosa26^Tdtomato and Tff1-CreERT2; Apc^fl/fl; p53^fl/fl (TcPP) mouse models and organoids. ATOH1 deficiency downregulates growth arrest-specific protein 1 (GAS1) by suppressing GAS1 promoter transcription. GAS1 forms a complex with RET, which inhibits Tyr1062 phosphorylation, and consequently activates the RET/AKT/mTOR signaling pathway by ATOH1 deficiency. Combining chemotherapy with drugs targeting AKT/mTOR signaling can overcome ATOH1 deficiency-induced chemoresistance. Moreover, it is confirmed that abnormal DNA hypermethylation induces ATOH1 deficiency. Taken together, the results demonstrate that ATOH1 loss promotes cancer stemness through the ATOH1/GAS1/RET/AKT/mTOR signaling pathway in GAC, thus providing a potential therapeutic strategy for AKT/mTOR inhibitors in GAC patients with ATOH1 deficiency.

Introduction

Gastric adenocarcinoma (GAC) is the fifth most common cancer worldwide and the third leading cause of cancer-related deaths.^[1] Chemotherapy and tumor recurrences are persistent and unresolved problems associted with GAC treatment.^[2] Gastric cancer stem cells (GCSCs) are a small population of self-renewing tumor cells isolated from GAC.^[3] As GCSCs have inherent stem cell-like properties, they play vital roles in tumor progression and therapeutic resistance. The ability of CSCs to adopt a quiescent state has emerged as an important driver of drug resistance.^[4] Unfortunately, the low efficacy of conventional 5-FU-based chemotherapy against GCSCs often leads to treatment failure.^[5] Elucidating the regulatory mechanisms of GCSCs may facilitate the development of novel targeted strategies to eliminate these cells and improve the prognosis of GAC.

Atonal basic helix-loop-helix transcription factor 1 (ATOH1) is a member of the basic helix-loop-helix (bHLH) family of transcription factors that are involved in various developmental processes.^[6] ATOH1 specifies and regulates the skin mechanosensory cells and develops the auditory system in the inner ear.^[7] To the best of our knowledge, the role of ATOH1 in gastric epithelial development has not been reported. In addition, although certain studies have demonstrated that ATOH1 participates in carcinogenesis,^[8] its specific role and mechanism in this process in GAC still need to be clarified. Therefore, this study aimed to investigate the effects of ATOH1 on the GCSC phenotype and chemotherapy resistance in GAC.

To determine the roles of ATOH1 in GAC, we established a stomach-specific Atoh1 transgenic mouse model and evaluated Atoh1 deletion as a risk factor for GAC progression. Stomach-specific Atoh1 deletion promotes stemness and chemoresistance of gastric epithelial cells. Moreover, ATOH1 downregulation results in poor GAC prognosis. ATOH1 inhibits stemness and chemoresistance in the GAC by activating growth arrest-specific protein 1 (GAS1) transcription and suppressing the RET/AKT/mTOR signaling pathway. Therefore, ATOH1 is a promising therapeutic target for the treatment of GAC.

Results ATOH1 is Downregulated in Chemoresistant GAC Tumors and GAC Pit Cells

Resistance to chemotherapy is a manifestation of GAC stemness.^[9] We, therefore, sought genes that were preferentially downregulated in GAC (vs adjacent gastric tissues) and chemoresistant GAC (vs chemosensitive GAC). We identified the expression profiles (Figure S1A, Supporting Information) of dysregulated genes in three GAC versus adjacent gastric non-tumor cohorts from Fujian Medical University Union Hospital (FJMUUH), First Affiliated Hospital of the University of Science and Technology of China (FHUSTC), and Qinghai Provincial People's Hospital (QHPH). We found 476, 727, and 319 downregulated genes with logFC < −2 and adjusted P < 0.05 in the FJMUUH, FHUSTC, and QHPH cohorts, respectively. Venn diagrams revealed 70 downregulated genes common to all three cohorts (Figure 1A). Twenty-three downregulated genes were detected in the chemoresistant and chemosensitive GAC cases from FJMUUH (Figure S1B,C, Supporting Information). These two groups overlapped only in ATOH1 (Figure 1A). ATOH1 downregulation was detected in chemoresistant cells (Figure S1D, Supporting Information). ATOH1 mRNA expression was significantly decreased in GAC versus adjacent normal tissues from the FJMUUH cohort (Figure S1E, Supporting Information). ATOH1 protein levels were significantly lower in 147 primary tumor samples than in adjacent noncancerous tissues from FJMUUH patients (Figure S1F, Supporting Information). Four out of ten wild-type C57BL6 mice developed GAC 12 months after MNU induction (Figure S1G, Supporting Information). The proportions of Atoh1⁺ cells were significantly lower in MNU-induced mouse tumors (corpus: 53.5 ± 3.6% vs 8.7 ± 2.8%, P < 0.001; antrum: 58.4 ± 4.8% vs 11.5 ± 3.1%, P < 0.001) than the normal gastric tissues (Figure S1G, Supporting Information).

We performed single-cell transcriptome sequencing (scRNA-seq) on GAC and paracarcinoma samples from the present and a previously published study (Table S4, Supporting Information) (Figure 1B; Figure S2A–C, Supporting Information).^[10] Differential gene expression analysis identified several markers associated with the cultured gastric epithelium and their expression in gastric epithelial cells are shown by t-distributed stochastic neighbor embedding (tSNE), such as PGAC, MUC5AC, and TFF1 (Figure 1B; Figure S2D, Supporting Information).^[11] Moreover, ATOH1 is barely expressed in the TFF1⁺ epithelial (pit) cells of the GAC samples, but not in the paracarcinoma samples (Figure 1C; Figure S2E,F, Supporting Information). These findings demonstrated that the number of ATOH1⁺ gastric epithelial cells decreased after oncogenic stimulation. Furthermore, the loss of Atoh1 in pit cells shaped cellular interactions and the tumor microenvironment (Figure S3A–D, Supporting Information).

ATOH1 Deletion in Mouse Stomach Pit Cells Promotes Cancer Stemness and Aggressiveness

Endogenous TFF1 was expressed in the pit regions of the gastric glands in the corpus and antrum and co-localized with Atoh1 in pit cells (Figure S4A–C, Supporting Information).^[11b,12] We generated Tff1-CreERT2; Rosa26^Tdtomato mice and confirmed that the stomachs of Tff1⁺ lineage mice contained Atoh1⁺ cells, whereas the Tff1⁺ lineage was not detected in the small intestine or colon (Figure 1D; Figure S4C,D, Supporting Information). Tamoxifen administration silenced the Atoh1 protein in Tff1⁺ cells in the gastric epithelia of Tff1-CreERT2; Atoh1^fl/fl; Rosa26^Tdtomato mice (Figure 1E; Figure S4E,F, Supporting Information). Lineage tracing showed that Tff1⁺ cells proliferated in the absence of Atoh1 (Figure 1F,G). Persistent Apc and p53 ablation led to gastric tumorigenesis 90 days after tamoxifen induction in Tff1-CreERT2; Apc^fl/fl; p53^fl/fl (TcPP) mice (Figure 1H–J). We hypothesized that Tff1-CreERT2 transgene-mediated Atoh1 ablation would enhance this effect. Hence, we administered tamoxifen to TcPP and Tff1-CreERT2; Apc^fl/fl; p53^fl/fl; Atoh1^fl/fl (TcPP; Atoh1^fl/fl) mice harboring the “floxed” ATOH1 allele (Figure S4G, Supporting Information). We observed significant increases in tumor burden and number in Atoh1^fl/fl cohort mice 90 days after tamoxifen induction (Figure 1I–K, Supporting Information).

ATOH1 Inhibits Cancer Stemness In Vivo and In Vitro

Western blot and qRT-PCR analyses revealed ATOH1 expression in various GAC cell lines. ATOH1 mRNA and protein levels (Figure 2A; Figure S5A,B, Supporting Information) were significantly reduced in the GAC cell panel compared with those in the normal gastric epithelial GES cells. We generated AGS and NCI-N87 cells overexpressing ATOH1 and used lentiviral shRNA to generate SNU-5 and Kato-III cells with endogenous ATOH1 knockdown (Figure S5C,D, Supporting Information).

Analysis of the differentiation trajectories of normal, TcPP, and TcPP; Atoh1^fl/fl cohorts revealed that the absence of ATOH1 promotes a greater proportion of cells to persist in the early stages of differentiation (Figure S6A–C, Table S5, Supporting Information). Gene scoring of the ATOH1^high and ATOH1^low cohorts using relative stem cell signatures from the GO biological process items in scRNA-seq datasets from human and mouse sample revealed a negative correlation between ATOH1 expression and pathways associated with stemness (Figure S6D,E, Supporting Information).

Gene set enrichment analysis (GSEA) of the TcPP and TcPP;Atoh1^fl/fl cohorts revealed that the absence of ATOH1 positively influenced the regulation of stem cell population maintenance (Figure S7A, Supporting Information). GSEA of the GEO, FJMUUH, TCGA, and AGS cohorts revealed enrichment of several genes and pathways regulating self-renewal and stemness in patients or GAC cells with low ATOH1 expression (Figure S7B, Supporting Information). Moreover, GSEA revealed the enrichment of signatures regulating CSCs that were also present in mouse databases (Figure S7C, Supporting Information).

We used spheroid cultures to investigate whether ATOH1 maintains GCSC properties. Preliminary experiments revealed that these culture conditions enhanced CSC-related properties,^[13] including CSC marker upregulation and enhanced tumor initiation (Figure S8A–D, Supporting Information). ATOH1 was significantly downregulated in the spheroids compared with the parental GAC cells (Figure S8D,E, Supporting Information). ATOH1 overexpression inhibited primary and secondary sphere formation (Figure 2B). Western blot and immunofluorescence showed that ATOH1 overexpression downregulated the GCSC marker CD44 and the self-renewal marker SOX2 in the spheroids (Figure 2C,D). In contrast, ATOH1 knockdown significantly increased the number and size of primary and secondary spheres (Figure 2E). Western blot and immunofluorescence confirmed that ATOH1 knockdown upregulated both CD44 and SOX2 in the spheroids (Figure S9A–C, Supporting Information). Flow cytometry analysis revealed that ATOH1 expression significantly decreased the number of CD44⁺ cells (Figure S9D, Supporting Information). However, ATOH1 knockdown exhibited the opposite effect (Figure S9E, Supporting Information). A significant inverse correlation (P < 0.001; Figure S9F, Supporting Information) between ATOH1 and CD44 expression was observed in patients according to immunohistochemical (IHC) analyses.

The limiting dilution assay reduced spheroid formation capacity in ATOH1-overexpressing cells from 1 in 1.14 to 1 in 3.11 (AGS cells, P < 0.001; Figure S9G–I, Supporting Information) and 1 in 1.23 to 1 in 3.84 (NCI-N87 cells, P < 0.001; Figure S9J–L, Supporting Information). Tumor-initiating ability is a property of CSCs.^[14] Serial tumor xenograft dilutions significantly lowered the tumor initiation capacity from 1 in 603611 AGS cells (control) to 1 in 3488397 cells (ATOH1 overexpression) (P = 0.008; Figure 2F; Figure S10A–E, Supporting Information). Furthermore, ATOH1-overexpressing NCI-N87 cells showed lower tumorigenicity and slower tumor growth than control cells (Figure S10F–J, Supporting Information). In contrast, SNU-5 cells with ATOH1 knockdown showed comparatively higher tumorigenicity and faster growth rates than control cells (Figure S10K–N, Supporting Information). Therefore, ATOH1 might regulate the GAC stemness.

Tumor-derived organoids conserve the pathophysiology of the original tumors, while maintaining cellular heterogeneity and self-renewal capacity.^[15] Organoids were established based on GAC (Figure S11A, Supporting Information) and MNU-induced mouse tumors (Figure S11B, Supporting Information). ATOH1 overexpression reduced the size and disrupted the architecture of organoids (Figure 2G). Similarly, ATOH1 overexpression significantly (P < 0.001) compromised human organoids (Figure 2H; Figure S11C, Supporting Information).

CD44 and SOX2 are considered markers of cancer stem cell-like properties.^[16,17] ATOH1 overexpression also downregulated CD44 expression in human organoids (Figure S11D, Supporting Information). We observed significant increases in the numbers of CD44⁺ and SOX2⁺ epithelial cells in tumors in the TcPP; Atoh1^fl/fl mouse cohort compared with the TcPP; Atoh1^fl/+ cohort 90 days after tamoxifen induction (CD44⁺:60.5 ± 7.0% vs 43.0 ± 5.3%, P = 0.002; SOX2⁺:82.7 ± 5.1% vs 68.6 ± 6.8%, P = 0.006) (Figure 2I).

GAS1 is a Transcriptional Target of ATOH1 and Contributes to ATOH1-Mediated GCSC Maintenance

ChIP-Seq of control and ATOH1-overexpressed AGS cells was used to identify genome-wide ATOH1-targeting sites, including 517 RefSeq genes (Table S6, Supporting Information). Through integrative analysis using RNA-Seq and ChIP-Seq data for ATOH1, we identified 25 upregulated genes, including GAS1, which bound to ATOH1 (Figure 3A). Moreover, qRT-PCR of control and ATOH1-overexpressed AGS cells showed that no compensatory molecule in addition to GAS1 was overexpressed in the growth arrest-specific protein family (Figure S12A, Supporting Information). GAS1 regulates cancer chemoresistance and tumorigenic potential.^[18] The mRNAsi-based stemness index^[19] of TCGA revealed that GAS1 expression was negatively correlated with GAC stemness (Figure S12B, Supporting Information). GAS1 weakened the spheroid-forming capacity of the GAC cells (Figure S12C, Supporting Information). GAS1 overexpression downregulated CD44 and SOX2 expression in spheroids (Figure S12D–F, Supporting Information). The data for 48 primary tumor samples indicated that GAS1 mRNA and protein expression levels were significantly reduced in tumor tissues (Figure S12G,H, Supporting Information). The results of the TCGA cohort was similar (Figure S12I, Supporting Information). In the GSE51105^[20] and GSE22377^[21] datasets, patients with GAC and high GAS1 expression showed relatively better survival (Figure S12J, Supporting Information). These results suggested that GAS1 negatively regulates GCSC.

We found that ATOH1 mRNA expression levels were positively correlated with GAS1 in 48 primary tumor samples (P < 0.001; Figure S13A, Supporting Information). Western blot and qRT-PCR revealed that in spheroids, ATOH1 overexpression upregulated GAS1, whereas ATOH1 knockdown downregulated GAS1 at both the mRNA and protein levels (Figure S13B,C, Supporting Information). Immunofluorescence staining showed that ATOH1 overexpression increased the number of GAS1⁺ cells (P < 0.001), whereas ATOH1 knockdown had the opposite effect (P < 0.001) (Figure S13D, Supporting Information).

Luciferase reporter assays revealed that ATOH1 activated the GAS1 promoter (Figure 3C). Sequence analysis revealed three putative ATOH1 binding sites in the GAS1 promoter. Sequence deletion and site-directed mutagenesis indicated that the first ATOH1 binding site is essential for ATOH1-induced GAS1 transactivation. ChIP assay confirmed the direct ATOH1 binding to the GAS1 promoters in GAC cells (Figure 3D). These results suggested that GAS1 is a direct transcriptional ATOH1 target. IHC staining of the tissue microarray (TMA) showed that ATOH1 was positively correlated with GAS1 expression in human GACs (P < 0.001; Figure 3E). ATOH1 overexpression upregulated GAS1 in human organoids (Figure 3F). Rescue experiments were performed to determine whether GAS1 contributes to ATOH1-mediated GCSC maintenance. GAS1 reduction reversed the inhibitory effect of ATOH1 overexpression on spheroid and CD44⁺ cell formation (Figures S14A,B; and S15A, Supporting Information). GAS1 upregulation significantly reduced spheroid and CD44⁺ cell formation caused by ATOH1 knockdown (Figures S14C,D; and S15B, Supporting Information). BALB/c nude mice were subcutaneously injected with AGS cells overexpressing ATOH1 alone, or both ATOH1 and shGAS1. GAS1 reduction reversed the inhibitory effects of ATOH1 overexpression on heterologous tumor growth and tumor initiation (Figure S15C–F, Supporting Information). GAS1 upregulation reduced the tumor initiation capacity of SNU-5-shATOH1 cells (Figure S15G, Supporting Information). These results suggested that ATOH1 regulates GCSCs by activating the GAS1 promoter.

ATOH1 Regulates RET/AKT/mTOR Signaling in GAC

To elucidate the downstream molecular mechanism of ATOH1 in regulating GAC, we performed GSEA on TCGA, FJMUUH, and GEO datasets. ATOH1 cohorts with high ATOH1 expression was compared with those with low ATOH1 expression, and AKT/mTOR signaling was enriched in all datasets (Figure 4A). Considering that GAS1 may be a co-receptor protein complexed with the receptor tyrosine kinase RET,^[22] we investigated whether ATOH1 affects malignancy through GAS1/RET/AKT/mTOR signaling. We performed co-immunoprecipitation (Co-IP) on AGS and NCI-N87 GAC cell lines ectopically expressing FLAG-tagged GAS1 and validated protein-protein interactions. GAS1 pull-down assay with anti-FLAG identified RET as a GAS1 binding partner (Figure 4B). Reciprocal Co-IP with anti-RET in both cell lines revealed that GAS1 was an interacting protein (Figure 4C). Thus, GAS1 may combine with RET to form a new protein complex that inhibits RET/AKT/mTOR signaling. ATOH1 overexpression significantly reduced RET, AKT, and mTOR phosphorylation levels in spheres (Figure 4D; Figure S16A, Supporting Information), and downregulated RET/AKT/mTOR phosphorylation in organoids (Figure 4E; Figure S17A–C, Supporting Information). In the xenotransplantation model, a significant inverse correlation was observed between ATOH1 expression and RET/AKT/mTOR phosphorylation (Figure S17D,E, Supporting Information). In the spheres, ATOH1 knockdown significantly augmented p-RET, p-AKT, and p-mTOR activities (Figure S16B–D, Supporting Information). GAS1 shRNA co-transfection reversed the inhibitory effect of ATOH1 overexpression on RET/AKT/mTOR phosphorylation. GAS1 overexpression reversed ATOH1 knockdown-induced upregulation of RET/AKT/mTOR phosphorylation (Figure S16E–G, Supporting Information). These results suggest that ATOH1 inhibits phosphorylation of the RET/AKT/mTOR signaling axis in a GAS1-dependent manner.

To investigate whether ATOH1 regulates GAC stemness through the AKT/mTOR signaling pathway, we added the AKT/mTOR pathway inhibitor thioridazine hydrochloride (THO) to treat developing spheroids with simultaneous ATOH1 knockdown. THO administration significantly inhibited the increase in spheroid formation caused by ATOH1 knockdown (Figure 4F). Western blot confirmed that THO downregulated p-AKT, p-mTOR, and stemness markers, which were increased by ATOH1 knockdown (Figure S18A, Supporting Information). Thus, THO offsets the growth-promoting effects of ATOH1 knockdown in heterogeneous tumors in vivo (Figure S18B,C, Supporting Information).

We observed significant increases in p-RET⁺, p-AKT⁺, and p-mTOR⁺ epithelial cells in the antra of the TcPP; Atoh1^fl/fl mice compared with those of the TcPP; Atoh1^fl/+ mice at 90 days after tamoxifen induction (p-RET⁺: 13.6 ± 3.4% vs 33.8 ± 4.0%, P < 0.001; p-AKT⁺: 70.1 ± 6.5% vs 88.7 ± 4.1%, P < 0.001; p-mTOR⁺: 76.4 ± 6.2% vs 92.3 ± 3.5%, P = 0.001) (Figure 4G–I; Figure S18D, Supporting Information). These results suggested that RET/AKT/mTOR signaling mediates ATOH1 regulation in GAC malignancy.

The ATOH1 Promoter was Hypermethylated in GAC

Using scRNA-seq datasets, we investigated DNA methylation levels in cancerous and normal epithelial tissues, revealing a remarkable elevation in DNA methylation in cancerous tissues compared with their normal counterparts. ATOH1 expression could be regulated by DNA methylation. To elucidate the mechanism by which deletion of ATOH1 regulates DNA methylation, we investigated whether ATOH1 downregulation was related to the methylation status of its promoter in the GAC. We performed bisulfite sequencing to evaluate the ATOH1 promoter methylation levels in six pairs of GAC and adjacent normal tissues. CpG islands and selected bisulfite sequencing regions of the ATOH1 promoter are shown in Figure 5A. The methylation levels of the CpG sites at −1,362 and −1,341 bp in the ATOH1 promoter were significantly higher in GAC tissues than in their adjacent noncancerous tissues (Figure 5B; Figure S19A–D, Supporting Information). ATOH1 methylation levels in GAC cell lines were significantly higher than those in normal gastric epithelial cells (Figure 5C). We treated GAC cells with the demethylation drug 5-azacytidine (5-AzaC) to determine whether ATOH1 was downregulated in response to the hypermethylation of its promoter. 5-AzaC treatment significantly increased ATOH1 mRNA and protein levels in GAC cells (Figure 5D,E). To establish the potential roles of various DNA methyltransferases (DNMTs) in mediating ATOH1 promoter methylation in GAC, we knocked down DNMT1, DNMT3A, and DNMT3B in GAC cells using specific small interfering RNAs (siRNAs) (Figure S19E, Supporting Information). Knockdown of DNMT1 but not DNMT3A or DNMT3B rescued ATOH1 expression (Figure 5F,G). DNMT1 overexpression significantly inhibited ATOH1 expression (Figure 5H,I; Figure S19F,G, Supporting Information). To determine the effects of DNA methylation on ATOH1 promoter activity and to confirm the participation of the −1,362 and −1,341 bp CpG sites in promoter regulation, we transfected wild-type ATOH1 promoter constructs or those containing site-specific CpG mutations into SNU-5 and Kato-III cells. DNMT1 overexpression significantly decreased the activity of the wild-type promoter. However, CG-to-TG mutations at the −1,362 and −1,341 bp CpG sites reversed the inhibitory effect of DNMT1 on ATOH1 promoter activity. Thus, the methylation status of the −1,362 and −1,341 bp CpG sites in the promoter region are crucial for the epigenetic regulation of ATOH1 expression (Figure 5J). These findings suggest that ATOH1 downregulation is associated with hypermethylation of its promoter in GAC.

We investigated whether pharmacological DNMT inhibition suppressed tumorigenesis by regulating ATOH1/GAS1/RET/AKT/mTOR signaling. 5-AzaC treatment inhibited AGS tumor xenograft growth (Figure S20A–C, Supporting Information). It also upregulated ATOH1 and GAS1 and significantly downregulated p-RET, p-AKT, and p-mTOR in spheroids (Figure 5K). Next, we explored whether ATOH1 upregulation inhibits 5-AzaC-mediated RET/AKT/mTOR signaling. Kato-III spheroids with ATOH1 knockdown were subjected to 5-AzaC treatment, and the effect of shATOH1 on the RET/AKT/mTOR signaling pathway was attenuated (Figure S20D, Supporting Information). 5-AzaC treatment weakened spheroid formation in SNU-5 and Kato-III cells with ATOH1 knockdown (Figure 5L). It also significantly inhibited SNU-5 tumor xenograft growth (Figure S20E,F, Supporting Information). These results indicate that the inhibition of DNMT1 activity suppresses tumor growth by regulating ATOH1/GAS1/RET/AKT/mTOR signaling in GAC.

ATOH1 Expression in Tumors is Correlated with GAC Patient Prognosis

Clinicopathological characteristics stratified by ATOH1 expression were determined using IHC of a TMA containing 379 GAC samples from FJMUUH (Table S7, Supporting Information). Low ATOH1 expression was significantly associated with advanced pT and pN stages. Similar results were obtained in the FHUSTC cohort (Table S8, Supporting Information). Kaplan–Meier survival analysis (Figure 6A,B) revealed better five-year overall survival (OS) in ATOH1^high patients than ATOH1^low patients (62.3% vs 44.3%; P < 0.001). ATOH1^high patients had a significantly higher five-year disease-free survival (DFS) than ATOH1^low patients (58.8% vs 42.4%; P = 0.002). The overall recurrence was lower in ATOH1^high patients than ATOH1^low patients (P < 0.001; Figure S21A, Supporting Information). Univariate and multivariate Cox analyses showed that ATOH1^high status was an independent protective factor against survival (Figure 6C; Figure S21B, Supporting Information). Similar results were obtained for the OS analyses of the FHUSTC cohort and GEO datasets (Figure S21C,D, Supporting Information).

We performed a Kaplan–Meier analysis to establish whether ATOH1 levels were associated with the prognosis of patients with GAC who had been administered adjuvant chemotherapy (ACT). Both OS and DFS were low in ATOH1^low patients, regardless of ACT administration (Figure 6D). ATOH1^high patients had relatively higher survival rates than ATOH1^low patients after ACT. These findings suggest that ATOH1 upregulation is associated with chemosensitivity and prognosis of patients with GAC. We evaluated the effect of combining ATOH1 expression with TNM staging on the prognostic accuracy. ATOH1 added prognostic value to clinicopathological features based on time-dependent receiver operating characteristic (ROC), C-index, and Akaike information criteria (AIC) analyses (Figure 6E).

DNMT1/ATOH1/GAS1/RET/AKT/mTOR Signaling Dysregulation Exhibits Clinical Significance

We evaluated the clinical significance of DNMT1/ATOH1/GAS1/RET/AKT/mTOR signaling in GAC. Immunohistochemistry was used to compare DNMT1, ATOH1, GAS1, p-RET, p-AKT, and p-mTOR expression in a TMA comprising 92 independent primary GAC samples and adjacent normal gastric tissue. The adjacent tissues showed upregulated DNMT1, p-RET, p-AKT, and p-mTOR and downregulated ATOH1 and GAS1 (Figure S22A,B, Supporting Information). Strong inverse correlations were observed between ATOH1 and the expression of p-RET, p-AKT, p-mTOR, and CD44 (Figure S22C, Supporting Information). This finding is consistent with our in vitro and in vivo results. The combination of DNMT1 upregulation and ATOH1 and GAS1 downregulation predicted shorter survival in GAC patients (Figure S22D,E, Supporting Information). These findings indicate that dysregulated DNMT1/ATOH1/GAS1/RET/AKT/mTOR signaling plays a critical role in disease progression and is a valuable prognostic biomarker for GAC.

THO Works Synergistically with 5-fluorouracil (5-FU) to Inhibit ATOH1-Deficient GAC Cell Growth Both In Vitro and In Vivo

Sensitivity to 5-FU differed significantly between ATOH1-low and ATOH1-high expression groups in ACRG and TCGA datasets (Figure S23A, Supporting Information). The CCK-8 assay showed that the IC₅₀ for 5-FU treatment was significantly lower in GAC cells overexpressing ATOH1 than in the vector cells. The IC₅₀ of 5-FU treatment was substantially higher in GAC cells with ATOH1 knockdown than in control cells (Figure S23B, Supporting Information). ATOH1 expression increased the 5-FU sensitivity of the xenograft tumors (Figure S23C–E, Supporting Information). Conversely, ATOH1 knockdown reduced xenograft tumor sensitivity to 5-FU (Figure S23F,G, Supporting Information).

Sixty-eight days after tamoxifen induction, TcPP; Atoh1^fl/+ and TcPP; Atoh1^fl/fl mice were treated with one 5-FU dose per week. Tissue samples from the untreated and treated mice were harvested 24 h after the final 5-FU dose (Figure 7A). In TcPP; Atoh1^fl/+ mice, the volumes of 5-FU-treated tumors were significantly lower than those of the untreated controls. However, this difference was not evident in the TcPP; Atoh1^fl/fl mice (Figure 7B,C). Sixty-eight days after tamoxifen induction, the TcPP; Atoh1^fl/fl mice were treated with one 5-FU dose per week, one THO dose twice weekly, or both 5-FU and THO for 4 weeks. The 5-FU+THO-treated tumors had the smallest volumes (Figure 7D,E). There were few proliferating cells in the mice treated with THO alone and even fewer in the 5-FU+THO-treated mice (Figure 7F–H). The divergent efficacies of various treatment regimens for ameliorating disease progression in the TcPP; Atoh1^fl/fl mouse model underscores the therapeutic value of combining AKT/mTOR inhibitors with standard chemotherapy to prevent GAC progression.

We explored the effects of altered ATOH1 expression on the chemosensitivity of human GAC organoids. ATOH1 overexpression rescued chemosensitivity in the organoids and inhibited their growth to a greater extent than treatment with 5-FU alone (Figure 7I). These results demonstrated that virus-mediated ATOH1 overexpression inhibits in vivo tumor growth and increases GAC cell sensitivity to 5-FU.

Discussion

Elucidating the molecular mechanisms underlying cancer stemness is essential for developing innovative strategies to overcome chemotherapy-resistant GAC. It is also necessary to validate these strategies using preclinical models.^[23] Lineage tracing was used to identify all progeny stemness derived from a single cell and arrange them within the lineage hierarchy.^[24] In this study, we established a stomach-specific mouse model, Tff1-CreERT2; Rosa26^Tdtomato, and empirically demonstrated by lineage tracing that Tff1 cells seldom (if ever) exhibited stemness in gastric epithelial cells. Moreover, Atoh1 deletion confers stemness to Tff1 cells in a Tff1-CreERT2; Atoh1^fl/fl; Rosa26^Tdtomato mouse model. We further elucidated the molecular mechanisms by which ATOH1 deficiency induces CSC-like properties that drive cancer progression in vitro and in vivo. ATOH1 knockout inhibits GAS1 promoter transcription and activates RET/AKT/mTOR signaling. The proposed molecular mechanism of action of ATOH1 in the GAC is shown in Figure 7J. The combination of 5-FU and the AKT/mTOR signaling inhibitor THO exhibited potential against refractory GAC, suggesting that this treatment modality might be particularly efficacious in ATOH1-deficient GAC patients.

We used stomach-specific Atoh1 mouse models to determine the functions of ATOH1 in the GAC and found that Atoh1 deficiency induced CSC-like properties and increased the tumor burden. ATOH1, a member of the bHLH transcription factor family, was initially considered an important regulator of cerebellar granule neuron precursors and cochlear hair cell development.^[6,25] Subsequent studies have indicated that ATOH1 was also implicated in cell proliferation and differentiation.^[6,8] Prior research established intestinal-specific Atoh1 mouse models and showed that ATOH1 maintained stem cell homeostasis.^[26] Although a recent study found that ATOH1 could reduce the proliferation of gastric cancer cells,^[27] the mechanism was not explored in depth. However, the conservation of ATOH1 function in different species is unknown. We further explored the mechanism by which ATOH1 regulates CSC-like properties in the GAC. In this study, we generated Tff1-CreERT2; Rosa26^Tdtomato mice in which gene modification occurred primarily in the pit cell lineages of the antral and corpus glands. However, Tff1-CreERT2 was not activated in the cecum and proximal colon. We compared Tff1-CreERT2; Rosa26^Tdtomato and Tff1-CreERT2; Atoh1^fl/fl; Rosa26^Tdtomato mice subjected to tamoxifen induction. Atoh1 knockout imparts stemness to Tff1 cells. Comparison of TcPP; Atoh1^fl/+ and TcPP; Atoh1^fl/fl mice subjected to tamoxifen induction revealed that Atoh1 knockout increased tumor burden. ATOH1 loss imparted stemness to gastric epithelial cells and contributed to GAC progression. Consistent results were observed in the established GAC cell lines. Chromatin immunoprecipitation sequencing revealed an ATOH1-binding site in the GAS1 promoter. GAS1 is a multifunctional protein that induces apoptosis and regulates cell-cycle arrest in various tissues.^[28] Here, GAS1 is identified as a novel ATOH1 target gene that regulates cancer stemness. ATOH1 may upregulate GAS1 by activating GAS1 promoter transcription. Earlier studies have reported that GAS1 expression suppresses tumor progression by inhibiting cell proliferation in GAC.^[18] Recent evidence has indicated that GAS1 regulates CSCs.^[29] GAS1 is structurally homologous with glial cell line-derived neurotrophic factor (GFRαs) receptors^[30] and complexes with RET. This complex promotes cell survival and proliferation by activating the MAPK and PI3K/AKT signaling pathways.^[31] In GAC cells, GAS1 prevents Tyr1062 phosphorylation of RET by complexation. GSEA revealed that ATOH1 might regulate GAC stemness through AKT/mTOR signaling. PI3K/AKT/mTOR signaling may be critical in various solid tumors as it regulates tumor cell growth, chemoresistance, metabolism, and CSC.^[32] This study indicates that GAS1 is vital as an ATOH1 transcription target and reduces GCSC activity and chemoresistance via the RET/AKT/mTOR signaling axis. Our preclinical model revealed that the combination of chemotherapy with drugs targeting AKT/mTOR signaling overcame ATOH1 deficiency-induced chemoresistance. The combination of 5-FU with drugs targeting CSCs may be a promising strategy to overcome chemotherapy resistance in patients with GAC. Clinically validating the safety and efficacy of molecular markers targeting ATOH1 deficiency in treating GAC and routinely utilizing this approach in routine GAC therapy are needed.

ATOH1 deficiency remains a problem in GAC progression. Nonetheless, we demonstrated that it modulates the expression of genes and pathways that regulate cellular transformation and cancer progression. Epigenetic programs regulate gene expression and CSC self-renewal and differentiation.^[33] Abnormal DNA methylation is a common epigenetic regulatory defect in various tumors.^[34] DNA hypermethylation in CpG islands may cause a loss of differentiation in state-specific gene expression and the rescue of stemness. CpG methylation is catalyzed by DNMTs including DNMT1, DNMT3a, and DNMT3b.^[35] Previous studies have shown that DNMTs maintain stem cells, progenitor cells, and CSCs.^[34a,36] However, the molecular mechanisms through which DNMTs regulate GCSCs remain unknown. Bisulfite sequencing analysis of GAC tissues revealed that DNMT1 downregulated ATOH1 and significantly increased the methylation levels of CpG sites at −1,362 and −1,341 bp in the ATOH1 promoter. DNMT1 is a methylation-maintenance enzyme that regulates the genomic integrity and transcription of certain genes and retrotransposons.^[37] We showed that DNMT1 prevented ATOH1 upregulation and suppresses the properties of CSC-like cells.

Conclusion

In summary, we determined that DNMT1-mediated hypermethylation leads to ATOH1 deficiency by blocking GAS1 promoter transcription. This, in turn, activates RET/AKT/mTOR signaling to acquire CSC-like and chemoresistant properties in GAC cells, resulting in poor GAC prognosis.

Experimental Section Animal Studies

All animal experiments were performed in accordance with the protocols approved by the Animal Experimentation Ethics Committee of Fujian Medical University (IACUC FJMU 2021-0280).

Mice: Rosa26-LSL-Tdtomato (Cat# 007914), Apc^fl/fl (Cat# 029275), and p53^fl/fl (Cat# 008462) mice were purchased from Jackson Laboratories (Bar Harbor, ME, USA). Tff1-CreERT2, Atoh1^fl/fl, and C57BL/6 wild-type mice were purchased from Cyagen Biosciences Inc. (Santa Clara, CA, USA). Mouse gene sequences are listed in Table S1 (Supporting Information). Mice were housed under specific pathogen-free conditions. Age- and sex-matched littermates ≥ 6–8 weeks old were used in the experiments. Mice were intraperitoneally injected with tamoxifen (T832955; MACKLIN, Shanghai, China) dissolved in sunflower oil at the time points indicated in the text and/or figures. Samples were analyzed at the time points indicated in the text and/or figures. Additional materials and methods are described in Supplementary Information.

N-Nitroso-N-methylurea (MNU)-Induced Mice: A mouse model of MNU-induced GAC (HY-34758; MCE, Monmouth Junction, NJ, USA) was established as previously described with slight modifications.^[38] Briefly, mice were given drinking water containing 240 ppm MNU on alternate weeks for a total of 5 weeks (total exposure of 3 weeks).

Organoid Culture

Organoid cultures of human and mouse GAC were prepared according to a previously published protocol.^[39] Briefly, tumor tissues from the stomach were washed twice with PBS containing 1 × penicillin/streptomycin (BL505A, Biosharp, Hefei, China), followed by the removal of the muscle layer and mucus using scissors, and cut into 2–3 mm pieces followed by digestion with 2.5 mg ml⁻¹ Collagenase A (Sigma Aldrich, St. Louis, MO, USA) for 30 min. Five milliliters of dissociation buffer, including d-sorbitol (Sigma Aldrich) and sucrose (Sigma Aldrich), were added to the tissue and shaken for 2 min. The final supernatant was passed through a 70 µm filter, and the crypt fraction was centrifuged at 150 g for 5 min. After washing with ice-cold PBS, the gland pellet was resuspended in Matrigel (356255, Corning, Corning, NY, USA) supplemented with standard gastric organoids [advanced DMEM/F12 (#12634010, Thermo Fisher Scientific, Waltham, MA, USA), 1× GlutaMax (#35050061, Thermo Fisher Scientific), 1× HEPES (#15630080, Thermo Fisher Scientific), 1× Penicillin/Streptomycin, 50% Wnt3a, 10% RSPO-1, 10% Noggin, 1× B27 (#17504001, Thermo Fisher Scientific), 50 ng mL⁻¹ EGF (PHG0311, Thermo Fisher Scientific), 200 ng mL⁻¹ FGF10 (#100-26, Peprotech, Rocky Hill, NJ, USA), 1 mm N-acetyl-L-cysteine (#A9165, Sigma Aldrich), 1 nm Gastrin (#G9145, Sigma Aldrich), 2 mm A83-01 (#2939/10, Tocris, Bristol, UK), 10 mm Y-27632 (#1254/10, Tocris)]. Finally, 50 µl Matrigel suspension was carefully ejected into the center of each well of a 24-well plate. Standard gastric organoid medium (1 mL) was added to each well. The organoids were cultured in a 5% CO₂ incubator at 37 °C and changed media every 2–3 days. Organoids from the second passage were infected with lentivirus with control or ATOH1 overexpression in 15 ml tubes overnight. The diameter and number of organoids in three random 100× magnification fields were measured under a light microscope 7 days after infection. For histological examination, the organoids were fixed in 4% paraformaldehyde for 1 h, embedded in 2% agarose gel, or directly fixed in Matrigel in formalin for the generation of paraffin blocks, sectioning, and staining.

Statistical Analysis

Statistical analyses were performed using SPSS software (version 22.0; IBM Corporation, Armonk, NY, USA), GraphPad Prism version 8.0 (GraphPad Software, La Jolla, CA, USA), and R software environment, version 4.2.1 (R Foundation for Statistical Computing, Vienna, Austria). Continuous variables are expressed as mean (standard deviation), and categorical variables are expressed as numbers. Differences between groups were assessed using the t-test or χ² test, as appropriate.

Overall survival (OS) was defined as the time from surgery to death from any cause. Disease-free survival (DFS) was defined as the time from surgery to recurrence or death from any cause. Survival curves were estimated using the Kaplan–Meier method, and the log-rank test was used to determine statistical significance. Prognostic factors were examined using univariate and multivariate analyses with the Cox proportional hazards model. Harrell's concordance index (C-index) was used to measure the discriminatory ability of different prognostic models.^[40] The Akaike information criterion (AIC) within the Cox regression model was used to compare the performances of different prognostic models; smaller AIC values represented a better optimistic prognostic stratification.^[41] The relative likelihood of the two models was calculated using the following formula: exp ((AIC (model A)–AIC (model B))/2). The relative likelihood represents the probability that model A minimizes information as effectively as model B and can thus be interpreted as a p-value for the comparison of both AIC values.^[42] A time-dependent receiver operating characteristic (ROC) analysis was also performed to assess the discriminatory power of the prognosis model for time-dependent disease outcomes.^[43]

Images from all representative histological experiments, western blot, and IF were obtained at least three times independently. All tests were 2-sided with a significance level of P < 0.05. *P < 0.05; **P < 0.01; ***P < 0.001, ****P < 0.0001.

Ethical Statement

Human tissue samples: All de-identified gastric adenocarcinoma (GAC) tissues were obtained from the Fujian Medical University Union Hospital (FJMUUH, Fuzhou, China), the First Affiliated Hospital of University of Science and Technology of China (FHUSTC, Hefei, China), and the Qinghai Provincial People's Hospital (QHPH, Xining, China). The institutional review committee has approved all experimental protocols using de-identified human specimens of each institution (No. 2022KY035, No. 2020-WCK-01, and No. 2020-42). Informed consent was obtained from subjects in this experiment. The study complied with the principles outlined in the Declaration of Helsinki. We constructed 3 tissue microarrays (TMA) of 379 cases of tumor tissues and 3 TMA of adjacent non-tumor gastric tissues (more than 5 cm away from tumor margin) from FJMUUH between 2010 and 2015. A total of 182 paraffin-embedded samples of GAC tissues were obtained at the FHUSTC between 2013 and 2014 and were used for validation of clinical prognostic and correlation analysis. The inclusion criteria were as follows: (a) histological identification of GAC; (b) the absence of combined malignancy and distant metastasis; (c) availability of complete follow-up data. All the cases were restaged according to the criteria described in the AJCC cancer staging manual (8th edition). We also collected GAC tumor tissues and adjacent non-tumor gastric tissues from 14 GAC patients from the FJMUUH, 8 patients from the FHUSTC, and 5 patients from the QHPH with complete clinicopathological features for transcriptomic RNA sequencing. Furthermore, fresh GAC samples were collected from 16 patients who had received chemotherapy at FJMUUH for RNA sequencing, 8 of whom were chemosensitive and 8 were chemoresistant. In this study, progressive disease or stable disease after 4 cycles of chemotherapy stipulated by the revised RECIST guideline was defined as chemoresistant; complete response or partial response after 2 cycles of chemotherapy stipulated by the revised RECIST guideline was defined as chemosensitive. According to the GAC treatment guidelines, a 5-Fluorouracil (5-FU) based chemotherapy regimen was recommended for the 16 patients.

Acknowledgements

This study was supported by the Graduate Innovation Project of the Sailing Fund of Fujian Medical University (2020QH2028), the Youth Research Planning Project of Fujian Province Health Commission Science and Technology (2023QN01010108), the National Natural Science Foundation of China (No. 81871899), and the Construction Project of Fujian Province Minimally Invasive Medical Center (No. [2021]76), and Excellent Young Scholars Cultivation Project of Fujian Medical University Union Hospital (2022XH021). The authors thank those who have devoted a lot to this study, including nurses, pathologists, further-study doctors, statisticians, reviewers, and editors, especially Li-Wei Xue. They were not financially compensated for their contributions.

Conflict of Interest

The authors declare no conflict of interest.

Author Contributions

Q.Z., H.-G.W., J.-H.Y., and R.-H.T. contributed equally to this work and should be considered co-first authors C.-M.H. had full access to all data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Q.C., Q.Z., H.W., C.H., and C.Z. contributed to concept and design. Q.Z., Q.C., H.W., J.Y., A.L., Z.L., X.H., Y.L., H.Z., G.L., Z.H., K.X., W.Q., Y.Z., M.J., Q.H., Z.S.-G., P.L., and J.X performed acquisition, analysis, and interpretation of data. Q.Z., H.W., Q.C., J.Y., R.T., and C.H. drafted the manuscript. Q.Z., H.W., and J.Y. performed statistical analysis. Q.Z., Q.C., H.W., J.Y., G.Z., Q.Z., A.L., Z.L., X.H., Y.L., G.L., Z.H., K.X., W.Q., Y.Z., J.L., R.T., Z.H., J.H., P.L., and J.X. provided administrative, technical, or material support. Q.Z., R.T., and Q.C. performed supervision.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.