B-Cell Lymphoma-Extra Large Promotes Gastric Tumorigenesis via Apoptosis-Independent Wnt–β-Catenin Activation and Peroxisome Proliferator–Activated Receptor γ Suppression

Abstract

Background & Aims

The Wnt–β-catenin and Yes-associated protein pathways cooperatively drive gastrointestinal tumor progression. BCL2L1, a Yes-associated protein target gene, encodes the anti-apoptotic isoform B-cell lymphoma–extra-large, which is highly expressed in multiple cancers, but its role in gastric cancer remains unclear.

Methods

We generated Mist1-CreERT; Apc^flox/flox and bcl-xl^flox/flox mice to study gastric tumorigenesis with or without B-cell lymphoma–extra-large. Gastric organoids and xenografts were used to assess proliferation and downstream signaling, and findings were validated in human gastric cancer tissues and databases.

Results

Deletion of B-cell lymphoma–extra-large significantly reduced tumor growth and proliferation in Apc-deficient tumors, organoids, and xenografts. Nuclear β-catenin and Wnt target gene expression were attenuated, partly through derepression of Hnf4a and Pparg, indicating that B-cell lymphoma–extra-large promotes Wnt signaling. Concurrently, downregulation of Hnf4a and Pparg contributes to the inhibition of multiple metabolic pathways. In human gastric cancer, high BCL2L1 expression correlated with elevated Wnt target genes and poor prognosis, whereas PPARG expression inversely correlated with BCL2L1 and associated with favorable outcomes. Epidemiologic analysis suggested that pioglitazone, a peroxisome proliferator–activated receptor γ agonist, reduced gastric cancer incidence in diabetic patients.

Conclusions

B-cell lymphoma–extra-large promotes gastric tumorigenesis through dual mechanisms—enhancing Wnt signaling and suppressing peroxisome proliferator–activated receptor γ—and highlight B-cell lymphoma–extra-large as a potential therapeutic target while suggesting peroxisome proliferator–activated receptor γ activation may be protective.

Graphical abstract

Summary.

B-cell lymphoma–extra-large promotes gastric tumorigenesis by enhancing Wnt–β-catenin signaling and suppressing peroxisome proliferator–activated receptor γ–associated metabolic programs. Genetic or pharmacologic targeting of this pathway may inhibit gastric tumor growth, highlighting B-cell lymphoma–extra-large as a potential therapeutic target and peroxisome proliferator–activated receptor γ activation as a possible protective strategy.

What You Need to Know.

Background

B-cell lymphoma–extra-large is frequently upregulated in gastric cancer, but its functional role beyond apoptosis, particularly in regulating Wnt signaling and tumor metabolism, remains poorly understood.

Impact

This study demonstrates that B-cell lymphoma–extra-large promotes gastric tumorigenesis by enhancing Wnt–β-catenin signaling through suppression of hepatocyte nuclear factor 4 α and peroxisome proliferator–activated receptor γ, thereby sustaining proliferation and impairing metabolic differentiation.

Future Directions

Targeting B-cell lymphoma–extra-large–mediated Wnt activation or restoring peroxisome proliferator–activated receptor γ signaling, including pharmacologic peroxisome proliferator–activated receptor γ agonists, may represent novel preventive or therapeutic strategies for gastric cancer.

Gastric cancer (GC) is one of the most common cancers in the world. In recent years, the number of deaths due to GC has been declining due to a decrease in the Helicobacter pylori infection rate and improved health screening.¹ However, the 5-year survival rate for advanced GC remains low, and the 5-year survival rate for GC with distant metastasis is less than 10%.² Although several new drugs including immune checkpoint inhibitors and targeted antibodies have been applied to GC treatment and found to be effective in disease control, most advanced GCs become resistant to these drugs after several cycles and hardly show complete response. Thus, there is a demand for seeking additional therapeutic targets to control GCs that are resistant to currently available therapies.3, 4, 5, 6, 7, 8

In gastric glands, stem and progenitor cells in the isthmus continuously supply differentiated cells and maintain mucosal homeostasis. Many gastric stem cell markers have been reported, such as LGR5, CCK2R, AXIN2, FZD7, CD44, SOX2, and MIST1 for antrum and SOX9, IQGAP3, STMN1, SOX2 and MIST1 for corpus.9, 10, 11, 12, 13, 14, 15, 16, 17 Induction of genetic mutations in stem cells is sufficient to generate gastric tumors, whereas that in differentiated cells fails to develop tumors.^13,18,19 For example, deletion of Apc in the gastric stem cells results in the formation of tumors in the gastric antrum.20, 21, 22, 23 On the other hand, GC only develops in the presence of multiple mutations or a combination of mutation and additional chemical stimuli.^22,24,25 In gastric carcinogenesis, mutated stem cells are supported by the surrounding microenvironment. Many cell types are involved in such microenvironment, including inflammatory cells, fibroblasts, nerves, vascular endothelial cells, luminal bacteria, hormones, and secreted mucins, and these stromal components promote tumor growth by releasing a variety of molecules that directly activate cancer cell growth signals or restructure the surrounding area.^12,26, 27, 28, 29, 30 Among them, the Wnt–β-catenin signaling pathway has been identified to play a central role in the development and progression of GC.^21,31, 32, 33

We previously reported that the Wnt pathway and the Yes-associated protein (YAP) pathway are activated in GC cells through nerve-regulatory niche signals. YAP activation leads to upregulation of certain genes including BCL2L1 and others, but the significance and role of these gene changes remain unknown.²³ BCL2L1 is a gene encoding the Bcl2 family proteins B-cell lymphoma–extra-large (Bcl-xL) and B-cell lymphoma–extra-small (Bcl-xS), which regulate apoptosis by modulating mitochondrial membrane permeability; Bcl-xL is an inhibitor of apoptosis, whereas Bcl-xS is a promoter.34, 35, 36 Bcl-xL is reported to be involved in other biological reactions, such as autophagy,^37,38 neural network formation,³⁹ and cellular metabolism.³⁷

Bcl-xL is overexpressed in several cancers, and it has been reported that inhibition of Bcl-xL suppresses tumor growth. In pancreatic cancer, Bcl-xL was found to increase the incidence of pancreatic cancer by suppressing oncogene-induced senescence and apoptosis.⁴⁰ In colorectal and hepatocellular cancers, inhibition of Bcl-xL induced apoptosis, suggesting that combined therapies targeting Bcl-xL may increase the efficacy of chemotherapy.41, 42, 43 Overexpression of Bcl-xL and its tumor promoting role have been also reported in GCs.44, 45, 46 However, the precise mechanism of how Bcl-xL is upregulated and promotes tumor growth has not been fully elucidated. In this study, we analyzed the function of Bcl-xL in GC and explored its potential as a new therapeutic target.

Results

B-Cell Lymphoma–Extra-Large Enhances the Gastric Tumorigenesis in Part Through Activation of Wnt Pathway

To investigate the role of Bcl-xL in gastric mucosa and tumorigenesis, we analyzed tissues from Mist1-CreERT; bcl-xl^flox/flox (M-B), Mist1-CreERT; Apc^flox/flox (M-A), Mist1-CreERT; Apc^flox/flox; bcl-xl^flox/flox (M-AB), and Mist1-CreERT control (wild-type [WT]) mice. All mouse lines were generated on an LSL-TdTomato reporter background (Figure 1A and B). To characterize the expression pattern of Bcl-xL in normal and neoplastic gastric tissues, we performed fluorescent immunostaining for Bcl-xL on antral sections from each genotype. In M-B mice, Bcl-xL expression was markedly diminished in approximately 49% of glandular units relative to WT controls, indicating efficient Cre-mediated recombination in nearly one-half of the glands (Figure 1C). We next assessed the impact of Bcl-xL deficiency on tumor initiation and progression. Consistent with previous reports,¹⁰ M-A mice developed distal antral tumors within 1 month of tamoxifen (TAM) administration. At 2 months post-TAM, we compared gastric tumor sizes between M-A and M-AB mice and found that tumors in M-AB mice were significantly smaller than those in M-A mice (Figure 1D). In detail, all mice in the M-A cohort developed large tumors that covered the entire antral region and appeared as conglomerates of multiple nodular lesions. This morphology suggests that these tumors may have arisen from the fusion of multiple independent tumor foci. In contrast, mice in the M-AB cohort developed only a single, relatively small tumor per mouse. The consistent presence of a single tumor in the M-AB group, without evidence of multiple tumor formation, is likely attributable to the tumor suppressive effect of Bcl-xL deletion. Histopathologic analysis also confirmed this reduction in size, although both tumors were classified as intramucosal tumors. In contrast, M-B mice, which harbored Bcl-xL deletion alone, showed no histologic abnormalities compared with WT controls after TAM treatment (Figure 1E).

Figure 1.

The deletion of Bcl-xL reduces tumor burden in Apc-deficient tumors. (A) Gene construction schema in each mouse model (WT, M-B, M-A, and M-AB mice). (B) The protocol schema of TAM administration. (C) Bcl-xL staining in WT and M-B mice. Yellow arrows indicate the glands without Bcl-xL expression. The rates of the Bcl-xL+ cells are quantified (n = 3 mice/group). The P value was calculated using a t test. (D) Macroscopic findings of gastric tumors in M-A and M-AB mice. Tumor area is surrounded by a yellow dotted frame. The tumor diameters are quantified (n = 6 mice/group). The P value was calculated using a t test. (E) Hematoxylin and eosin (H&E) staining of WT, M-B, M-A, and M-AB mice at antrum (representative images for n = 3 mice/group). (F) Red fluorescent protein (RFP) staining in WT, M-B, M-A, and M-AB mice (representative images for n = 3 mice/group). (G) RT-PCR of Bcl2l1 (Bcl-xL) expression in WT, M-B, M-A, and M-AB mice (n = 9/group). The P value was calculated using a t test. Scale bars, 100 μm. Mean ± standard error of the mean. ∗P < .05.

Red fluorescent protein immunostaining confirmed robust TdTomato expression in both normal and tumor tissues (Figure 1F), indicating efficient Cre-mediated recombination. In M-B mice, sustained TdTomato fluorescence observed 2 months after TAM administration suggested that Bcl-xL ablation did not induce apoptosis and that Bcl-xL–deficient cells were able to survive long term. Quantitative real-time polymerase chain reaction (RT-PCR) using gastric tissue RNA further confirmed that Bcl-xL expression was decreased in M-B and M-AB mice, whereas it was elevated in M-A mice (Figure 1G).

To assess epithelial proliferative activity, we stained gastric tissues for Ki67. In both WT and M-B mice, Ki67-positive cells remained confined to the normal proliferative zones of the glands, indicating that loss of Bcl-xL alone does not disrupt baseline epithelial proliferation (Figure 2A). In contrast, M-A and M-AB mice exhibited distinct proliferation patterns in tumor regions: Ki67 expression was significantly higher in M-A tumors than in M-AB tumors (Figure 2B). These results indicate that the tumor suppressive effect of Bcl-xL loss is associated with a reduction in tumor cell proliferation. Notably, in M-A mice, immunohistochemical analyses using β-catenin and Ki67 and tdTomato and Ki67 costaining revealed an increase in Ki67⁺ cells not only among nuclear β-catenin⁺ cells and tdTomato⁺ cells but also in non–nuclear β-catenin cells and tdTomato⁻ cells. These findings suggest that hyperproliferation is not restricted to adenomatous polyposis coli (APC)-deficient cells but also extends to neighboring epithelial cells, supporting the presence of non–cell-autonomous proliferative effects induced by APC loss (Figure 2C and D).

Figure 2.

The deletion of Bcl-xL reduces tumor burden in Apc-deficient tumors. (A) Ki67 (green) staining in WT and M-B mice. The rates of the Ki67+ cells are quantified (n = 3/group). The P value was calculated using a t test. (B) Ki67 staining in M-A and M-AB mice. The rates of the Ki67+ cells are quantified (n = 3 mice/group). The P value was calculated using a t test. (C) β-catenin (green) and Ki67+ (red) staining in M-A mice. The rates of the Ki67+ cells are quantified, comparing non-nuclear and nuclear β-catenin cells (n = 3 mice/group). The P value was calculated using a t test. (D) tdTOMATO (red) and Ki67+ (green) staining in M-A mice. The rates of the Ki67+ cells are quantified, comparing tdTOMATO-positive and -negative cells (n = 3 mice/group). The P value was calculated using a t test. (E) Bcl-xL (red) and β-catenin (green) staining in WT mice (representative images for n = 3 mice/group). (F) Bcl-xL (red) and β-catenin (green) staining in M-A and M-AB mice. The rates of Bcl-xL+ cells and β-catenin–nuclear–translocated cells were quantified (n = 3 mice per group). In M-A mice, the rates of β-catenin–nuclear-translocated cells were quantified, comparing Bcl-xL-low and -high expressing cells (n = 3 mice per group). The P value was calculated using a t test. (G) β-catenin (green) staining in M-A and M-AB mice 2 weeks, 1 month, and 2 months after tamoxifen induction. The average number of β-catenin–nuclear-translocated cells per one clone were quantified (n = 10 clones per mouse, n = 3 mice per group). The P value was calculated using a t test. Scale bars, 100 μm. Mean ± standard error of the mean. ∗P < .05. DAPI, 4′,6-diamidino-2-phenylindole.

We next investigated the impact of Bcl-xL on tumorigenesis, with a particular focus on the Wnt signaling pathway, given its well-established role in GC progression.^21,31, 32, 33 In WT mice, β-catenin localization was restricted to the plasma membrane, with no clear evidence of nuclear translocation (Figure 2E). In contrast, nuclear translocation of β-catenin was observed in the tumor regions of M-A and M-AB mice, consistent with Apc deletion (Figure 2F). Moreover, M-A mice exhibited strong Bcl-xL expression throughout the tumor tissue, with particularly high levels in certain tumor cells. The Bcl-xL positivity was associated with nuclear β-catenin. In M-AB mice, Bcl-xL expression was markedly reduced in tumor regions, except for a small fraction where expression persisted, indicating efficient Bcl-xL knockout in the tumor (Figure 2F). Bcl-xL knockout also slowed down the expansion of nuclear β-catenin clones over time (Figure 2G).

Regarding SRY-box transcription factor 9 (SOX9) expression, WT and M-B mice exhibited limited expression at the basal regions of the glandular ducts. In both M-A and M-AB mice, SOX9 expression was upregulated in tumor areas, but the number of SOX9-positive cells was significantly increased in M-A mice compared with that in M-AB mice (Figure 3A and B). Most cells showing nuclear β-catenin are also positive for SOX9 (Figure 3C). M-A mice also exhibited a significantly greater number of Lgr5-positive cells, compared with M-AB mice (Figure 3D). We next generated Mist1-CreERT; Apc^flox/flox; Lgr5-DTR-green fluorescent protein (GFP) mice and examined the correlation among leucine-rich repeat-containing G protein-coupled receptor 5 (LGR5)-GFP positivity, Bcl-xL positivity, and nuclear β-catenin localization. The results showed that LGR5 expression was associated with both Bcl-xL positivity and nuclear β-catenin accumulation (Figure 3E and F). These results suggest that Bcl-xL deletion leads to a partial attenuation of Wnt pathway activation.

Figure 3.

The deletion of Bcl-xL induces the inactivation of the Wnt–β-catenin pathway. (A) SOX9 staining in WT mice (representative images for n = 3 mice/group). (B) SOX9 staining in M-A and M-AB mice. The rates of the SOX9+ cells are quantified (n = 3 mice/group). The P value was calculated using a t test. (C) SOX9 (green) and β-catenin (green) staining in M-A mice. The rates of the SOX9+ cells are quantified, comparing non-nuclear and nuclear β-catenin cells (n = 3 mice/group). The P value was calculated using a t test. (D) RNA–in situ hybridization (ISH) using Lgr5 probe in M-A and M-AB mice. The ratio of the Lgr5+ cells are quantified (n = 3 mice/group). The P value was calculated using a t test. (E) GFP (green) and Bcl-xL (red) staining in Mist1-creERT; Apc^flox/flox; Lgr5-GFP-DTR mice. The rates of GFP (=LGR5)-positive cells were quantified, comparing Bcl-xL-low and -high expressing cells (n = 3 mice per group). The P value was calculated using a t test. (F) GFP (green) and β-catenin (red) staining in Mist1-creERT; Apc^flox/flox; Lgr5-GFP-DTR mice. The rates of GFP (=LGR5)-positive cells were quantified, comparing non-nuclear and nuclear β-catenin cells (n = 3 mice/group). The P value was calculated using a t test. (G) β-catenin (green) and AQP5 (red) staining in M-A and M-AB mice 2 weeks and 1 month after TAM induction. The rates of nuclear β-catenin cells with AQP5 positivity are quantified (n = 3 mice/group). The P value was calculated using a t test. (H) β-catenin (green) and TROP2 (red) staining in M-A and M-AB mice 2 weeks and 1 month after TAM induction. The rates of nuclear β-catenin cells with TROP2 positivity are quantified (n = 3 mice/group). The P value was calculated using a t test. Scale bars, 100 μm. Mean ± standard error of the mean. ∗P < .05. DAPI, 4′,6-diamidino-2-phenylindole.

To assess the role of Bcl-xL in metaplastic and dysplastic progression, we performed immunostaining for aquaporin-5 (AQP5) and trophoblast cell surface antigen 2 (TROP2). In M-A mice, most cells with nuclear β-catenin were positive for both AQP5 and TROP2. In contrast, M-AB mice exhibited a marked reduction in AQP5- and TROP2-positive cells, accompanied by a decrease in nuclear β-catenin, suggesting a delay in metaplastic and dysplastic transformation compared with M-A mice over time. These findings indicate that Bcl-xL promotes metaplastic and dysplastic progression via activation of the WNT pathway (Figure 3G and H).

To directly assess apoptosis, we performed immunostaining for cleaved caspase-3, as Bcl-xL is known to exert anti-apoptotic effects. However, no cleaved caspase-3–positive cells were detected in the gastric tissues of WT, M-B, M-A, or M-AB mice (Figure 4A). Apoptosis was also assessed at early time points in M-B and M-AB mice following TAM administration. No clear evidence of apoptosis in epithelial cells was observed from day 7 (before tumor formation) to one month (after tumor formation) after TAM treatment (Figure 4B). Thus, although Bcl-xL is upregulated in Apc-deficient gastric tumors, its anti-apoptotic function is limited in this tumor model. Given that Bcl-xL is a transcriptional target of YAP,²³ we also analyzed YAP expression. In WT and M-B mice, YAP expression was restricted to the base of the glandular epithelium. In contrast, both M-A and M-AB mice exhibited strong YAP expression in tumor regions; however, no significant difference was observed between these groups, suggesting that Bcl-xL deletion does not influence YAP pathway activation (Figure 4C).

Figure 4.

Bcl-xL deletion suppresses Wnt signaling in APC-deficient gastric epithelium. (A) Cleaved caspase 3 staining in WT, M-B, M-A, and M-AB mice (Representative images for n = 3 mice/group). (B) Cleaved caspase 3 (red) and E-cadherin (green) staining in M-B and M-AB mice 7 days and 1 month after TAM induction (representative images for n = 3 mice/group). (C) YAP staining in WT, M-B, M-A, and M-AB mice. The rates of YAP-positive cells were quantified (n = 3/group). The P value was calculated using a t test. (D) RT-PCR of Lgr5, Axin2, Sox9, Fzd7, Ccnd1, Lgr4, and cMyc in M-A and M-AB mice (n = 9 mice/group). The P value was calculated using a t test. Scale bars, 100 μm. Mean ± standard error of the mean. ∗P < .05. DAPI, 4′,6-diamidino-2-phenylindole.

To further evaluate gene expression changes, quantitative RT-PCR was performed using complementary DNA (cDNA) synthesized from gastric tissues of M-A and M-AB mice. Wnt target genes, including Lgr5, Axin2, Sox9, Fzd7, and Ccnd1, were markedly downregulated in M-AB mice, compared with M-A mice. In contrast, Lgr4 and cMyc expression did not differ significantly between M-A and M-AB mice (Figure 4D). Collectively, these results indicate that Bcl-xL promotes gastric tumorigenesis in part through enhancing Wnt activation, thereby contributing to tumor progression, although these differences may also reflect the smaller tumor burden and lower number of APC-deficient cells in M-AB mice.

B-Cell Lymphoma–Extra-Large Activates the Wnt–β-Catenin Pathway in Tumor Organoids and Xenografts

To investigate the role of Bcl-xL in established gastric tumors in vitro, we employed a gastric organoid culture system. Organoids were derived from the normal gastric antrum of Apc^flox/flox and WT mice, infected with Cre-expressing lentivirus, and selected with puromycin to generate genetically modified organoids (Figure 5A). Initially, we aimed to generate organoids from bcl-xl^flox/flox and Apc^flox/flox; bcl-xl^flox/flox mice. However, we were unable to establish stable cultures. In contrast, Apc-deficient organoids were successfully generated. Thus, WT and Apc-deficient organoids (hereafter referred to as Apc organoids) were used for further analysis.

Figure 5.

Apc-deficient gastric organoids are sensitive to Bcl-xL inhibition. (A) The schema of the lentivirus-infected organoid development. (B–D) Treatment with Bcl-xL inhibitor (A-1155463) in WT and Apc organoids. (B) The representative images of the organoids. (C) The transitions of organoid diameters were quantified (n = 10/group). The P value was calculated using a t test. (D) Western blotting with cleaved caspase 3, caspase 3, and β actin in WT organoids with and without Bcl-xL inhibitor. Scale bars, 100 μm. Mean ± standard error of the mean. ∗P < .05.

To assess the effect of Bcl-xL on cell proliferation, we treated both WT and Apc organoids with the Bcl-xL–specific inhibitor (A-1155463) and monitored changes in organoid size. In WT organoids, Bcl-xL inhibition did not affect growth and apoptosis (Figure 5B–D and Supplementary Western blotting A–C). In contrast, Apc organoids exhibited a dose-dependent reduction in size upon Bcl-xL inhibition, indicating that Bcl-xL promotes proliferation in the Apc-deficient context (Figure 5C). There was no significant difference in organoid growth between WT and Apc-deficient organoids.

To further evaluate Bcl-xL function, we generated complex mutant organoids using LSL-Kras^G12D; LSL-Trp53^R172H; Apc^flox/flox and LSL-Kras^G12D; LSL-Trp53^R172H; Apc^flox/flox; bcl-xl^flox/flox mice infected with Cre-lentivirus and selected (hereafter referred to as KPA and KPAB organoids, respectively) (Figure 6A and B). We selected the additional Kras^G12D and Trp53^R172H mutations in the Apc-mutant GC model based on previous studies,^47,48 to enable the establishment of stable, genetically modified gastric organoids for functional analysis. Fluorescent immunostaining, Western blotting, and quantitative polymerase chain reaction (qPCR) confirmed Bcl-xL deletion in KPAB organoids (Figure 6B–E).

Figure 6.

Bcl-xL enhances the proliferation in the Wnt-dependent manner in organoid models. (A) Gene construction schema showing KPA and KPAB organoid models. (B) Hematoxylin and eosin staining in KPA and KPAB organoids. (C) RT-PCR of Bcl2l1 (Bcl-xL) in KPA and KPAB organoids (n = 4/group). The P value was calculated using a t test. (D) Western blotting of Bcl-xL and cyclin D1 in KPA and KPAB organoids are shown (n = 2/group). (E) Bcl-xL (green) andβ-catenin (red) staining in KPA and KPAB organoids. The rates of Bcl-xL+ cells and β-catenin–nuclear-translocated cells were quantified (n = 3 organoids/group). The P value was calculated using a t test. (F) RT-PCR of Axin2, Ccnd1, cMyc, and Lgr4 in KPA and KPAB organoids (n = 4/group). The P value was calculated using a t test. (G) The culture of KPA and KPAB organoids in standard medium. The ratios of organoid diameters relative to those on day 1 were quantified (n = 10/group). The P value was calculated using a t test. (H) The culture of KPA and KPAB organoids in medium lacking Wnt3a, R-spondin 3, and growth factors. The ratios of organoid diameters relative to those on day 1 were quantified (n = 10 per group). The P value was calculated using a t test. (I) Treatment with Bcl-xL inhibitor (100 nM and 1000 nM) in KPA and KPAB organoids. The ratios of organoid diameters relative to those on day 1 were quantified (n = 10 per group). The P value was calculated using a t test. (J) GFP (green) and cleaved caspase 3 (red) staining of KPA and KPAB organoids in the growth factor-free condition. The rates of cleaved caspase 3 cells were quantified (n = 10 organoids/group). The P value was calculated using a t test. Scale bars, 100 μm. Mean ± standard error of the mean. ∗P < .05.

Western blot analysis also showed reduced expression of Cyclin D1 in KPAB organoids (Figure 3D and Supplementary Western blotting D–F). β-catenin nuclear localization was markedly reduced in KPAB organoids compared with KPA organoids (Figure 6E). qPCR showed that Wnt target genes Axin2 and Ccnd1 were downregulated in KPAB organoids, while no significant changes were detected in Lgr4 or cMyc (Figure 6F).

To evaluate the impact of Bcl-xL on proliferation, we compared organoid growth between KPA and KPAB under different culture conditions. In standard medium, no significant difference in growth was observed (Figure 6G). However, under Wnt-deficient conditions (Dulbecco’s Modified Eagle Medium [DMEM]/F12 supplemented only with B27 and N2), KPAB organoids showed significantly smaller size than KPA organoids by day 5 (Figure 6H). Furthermore, treatment of KPA organoids with Bcl-xL inhibitor reduced organoid size, whereas KPAB organoids were unaffected by the inhibitor (Figure 6I). These findings indicate that Bcl-xL is dispensable for proliferation under optimal growth conditions but becomes critical for sustaining tumor growth in APC-deficient organoids under growth factor–limited conditions, which more closely reflect the in vivo tumor environment. Finally, we confirmed that there was no difference in apoptosis occurrence between KPA and KPAB organoids (Figure 6J).

Next, to assess the in vivo tumorigenic potential of Bcl-xL-deficient cells,⁴⁹ KPA and KPAB organoids were subcutaneously injected into nude mice, and tumor growth was monitored. KPAB-derived tumors were significantly smaller than KPA-derived tumors on both day 14 and day 21 (Figure 7A), indicating that loss of Bcl-xL again impairs xenografted, Apc-deleted tumor growth.

Figure 7.

Bcl-xL deletion suppresses the proliferation via inactivation of Wnt/β-catenin pathway in xenograft models. (A) Gross findings of the KPA and KPAB xenografts. The tumor sizes were quantified (n = 5/group). The P value was calculated using a t test. (B) GFP (green) and Bcl-xL (red) staining in KPA and KPAB xenografts. The rates of Bcl-xL + cells were quantified (n = 3/group). The P value was calculated using a t test. (C) GFP (green) andβ-catenin (red) staining in KPA and KPAB organoids. The rates of β-catenin–nuclear-translocated cells were quantified (n = 3/group). The P value was calculated using a t test. (D) GFP (green) and cyclin D1 (red), GFP (green) and SOX9 (red), GFP (green) and Ki67 (red), and GFP (green) and YAP (red) staining in KPA and KPAB xenograft. The rates of cyclin D1-, SOX9-, Ki67-, and YAP-positive cells were quantified (n = 3/group). The P value was calculated using a t test. (E) GFP (green) and cleaved caspase 3 (red) staining in KPA and KPAB xenografts. The rates of cleaved caspase 3–positive cells were quantified (n = 3/group). The P value was calculated using a t test. Scale bars, 100 μm. Mean ± standard error of the mean. ∗P < .05. DAPI, 4′,6-diamidino-2-phenylindole.

To further investigate molecular changes, xenograft tumors were harvested 1 month after injection and analyzed by immunohistochemical staining. GFP-positive cells, indicating organoid-derived tumor cells, were identified in both KPA and KPAB xenografts. In KPAB tumors, Bcl-xL expression was markedly reduced in GFP-positive cells, confirming successful knockout (Figure 7B). Nuclear localization of β-catenin was also decreased in KPAB tumors (Figure 7C), along with reduced expression of Wnt target genes Cyclin D1 and SOX9 and a proliferation marker Ki67, whereas no significant differences in YAP expression was observed between KPA and KPAB tumors, consistent with findings from the genetically engineered mouse and organoid models (Figure 7D). Finally, we confirmed that there was no difference in apoptosis occurrence between KPA and KPAB xenografts (Figure 7E).

These results collectively demonstrate that Bcl-xL promotes gastric tumorigenesis by enhancing Wnt pathway activity and supporting tumor cell proliferation, both in vitro and in vivo.

Negative Regulation of Hepatocyte Nuclear Factor 4 α by B-Cell Lymphoma–Extra-Large Contributes to the Activation of Wnt–β-Catenin Signaling

To comprehensively assess gene expression changes associated with Bcl-xL deficiency, RNA sequencing was performed using RNA extracted from WT, KPA, and KPAB organoids. Principal component analysis (PCA) demonstrated distinct clustering of each group, indicating unique transcriptomic profiles (Figure 8A). A focused heatmap of key gene sets (Figure 8B) revealed higher expression of Wnt-related genes in KPA organoids compared with WT and KPAB organoids. Within the Bcl2 family, Bcl2l1 was downregulated in KPAB organoids. Furthermore, the differentiation markers, including Muc6 and Tff2 for neck cell lineage and Muc5ac and Tff1 for pit cell lineage, are upregulated in KPAB organoids, compared with KPA organoids (Figure 8C). Gene Set Enrichment Analysis (GSEA) revealed downregulation of gene sets associated with the WNT pathway and epithelial cell proliferation in KPAB organoids, consistent with the results of other experiments (Figure 8D). These findings suggest that loss of Bcl-xL impairs WNT-driven epithelial proliferation, which may contribute to the reduced growth capacity observed in KPAB organoids.

Figure 8.

Comparative bulk RNA sequencing analysis of WT, KPA, and KPAB gastric organoids. (A–F) Bulk RNA sequencing comparing WT, KPA, and KPAB organoids (n = 2/group). (A) PCA plots. (B) Heatmap of gene expression. (C) Heatmap showing the Tff1, Tff2, Muc5ac, Muc6, and Mki67 expression. (D) Sigmoid curve showing the result of the GSEA with Gene Ontology Biological Process (GOBP) WNT SIGNALING PATHWAY, WP WNT SIGNALING PATHWAY. v2025, and GOBP EPITHELIAL CELL PROLIFERATIOHN gene sets. (E) Heatmaps showing the top 100 differentially expressed genes (DEGs) across the 3 groups. (F) Venn diagram of the downregulated DEGs in KPA organoids and upregulated DEGs in KPAB organoids.

Comparing the top 100 differentially expressed genes across the 3 groups, Hnf4a exhibited decreased expression in KPA organoids and increased expression in KPAB organoids (Figure 8E and F and Supplementary Table 1). Hnf4a is a transcription factor reported to regulate the Wnt signaling and cell proliferation pathways.⁵⁰ To assess its functional role, we overexpressed Hnf4a in KPA organoids using adenoviral vectors (Figure 9A). Hnf4a-overexpression (OE) suppressed cell proliferation in KPA organoids (Figure 9B). To thoroughly characterize transcriptional differences, we performed RNA sequencing on control and Hnf4a-OE KPA organoids. PCA revealed distinct clustering of the 2 groups, indicating unique transcriptomic profiles (Figure 9C). Heatmap visualization further highlighted differentially expressed gene patterns (Figure 9D). GSEA demonstrated that Hnf4a-OE significantly inhibited the activation of WNT pathway, with the downregulation of Wnt target genes (Figure 9E and F). RT-PCR analysis also revealed that Hnf4a-OE KPA organoids exhibited reduced expression of several Wnt target genes, including Sox9, Axin2, and Fzd7, compared with control KPA organoids (Figure 9G). These results suggest that Bcl-xL negatively regulates the Hnf4a expression, which activates the Wnt–β-catenin pathway and enhances the proliferation.

Figure 9.

Bcl-xL–mediated suppression of Hnf4a expression promotes activation of the Wnt–β-catenin pathway. (A) RT-PCR of Hnf4a in KPA organoids and KPA organoids overexpressing Hnf4a (n = 3/group). The P value was calculated using a t test. (B) The infection of KPA organoids with Hnf4a-overexpressing adenovirus. The ratios of organoid diameters relative to those on day 1 were quantified (n = 10 per group). The P value was calculated using a t test. (C–F) Bulk RNA sequencing comparing KPA organoids and KPA organoids overexpressing Hnf4a (n = 3/group). (C) PCA plot, (D) Heat map of top 100 differentially expressed genes (DEGs). (E) GSEA with Gene Ontology Biological Process (GOBP) WNT SIGNALING PATHWAY, WP WNT SIGNALING PATHWAY. v2025, and GOBP EPITHELIAL CELL PROLIFERATIOHN gene sets. (F) Heatmap of the WNT-target genes. (G) RT-PCR of Sox9, Axin2, and Fzd7 in KPA organoids and KPA organoids overexpressing Hnf4a (n = 3/group). The P value was calculated using a t test. Mean ± standard error of the mean. ∗P < .05.

Negative Regulation of Peroxisome Proliferator–Activated Receptor γ Expression by B-Cell Lymphoma–Extra-Large Contributes to the Tumor Growth

Focusing again on the GSEA of RNA sequencing for KPA, and KPAB organoids, the HALLMARK CHOLESTEROL HOMEOSTASIS gene set was significantly enriched in KPAB organoids, and Pparg was notably upregulated in KPAB organoids (Figure 10A and Supplementary Table 3), along with the other metabolic pathways including bile acid, fatty acid, and xenobiotic ones. Both qPCR and western blotting validated the Pparg upregulation in KPAB organoids compared with KPA organoids (Figure 10B and C and Supplementary Western blotting F–G). Additionally, immunohistochemical staining of xenograft sections revealed that GFP-positive cells from KPAB tumors strongly expressed PPARγ, whereas those from KPA tumors did not (Figure 10D). Tumor tissue in M-AB mice also showed increased expression of PPARγ compared with M-A mice. (Figure 10E).

Figure 10.

Bcl-xL suppresses PPARγ expression and multiple gene sets involved in metabolic pathways. (A) Bubble plots showing the GSEA analysis with Hallmark gene sets. Sigmoid curve showing the result with CHOLESTEROL HOMEOSTASIS gene set. (B) RT-PCR of Pparg in KPA and KPAB organoids (n = 4/group). The P value was calculated using a t test. (C) Western blotting of PPARγ in KPA and KPAB organoids (n = 2/group). (D) GFP (green) and PPARγ (red) staining in KPA and KPAB xenografts. The rates of PPARγ-positive cells were quantified (n = 3/group). The P value was calculated using a t test. (E) PPARγ (red) staining in M-A and M-AB mice. The rates of PPARγ-positive cells were quantified (n = 3/group). The P value was calculated using a t test. Scale bars, 100 μm. Mean ± standard error of the mean. ∗P < .05. DAPI, 4′,6-diamidino-2-phenylindole.

Given the observed upregulation of Pparg in Bcl-xL–deficient contexts, we next examined whether forced expression of Pparg changes the gene expressions and tumor growth. KPA organoids were infected with either Pparg-OE adenovirus or a control adenovirus. Fluorescent immunostaining and qPCR showed clear upregulation of Pparg expression in organoids infected with Pparg-OE adeno virus than control adeno virus (Figure 11A and B). Next, to comprehensively evaluate transcriptional changes, RNA sequencing was performed on KPA organoids and Pparg-OE KPA organoids. PCA revealed distinct clustering of the 2 groups, indicating unique transcriptomic profiles (Figure 11C). GSEA demonstrated that Pparg-OE significantly activated the metabolic pathways, such as fatty acid, bile acid, cholesterol, and xenobiotic ones, with the upregulation of lipid metabolism-related genes (Figure 11D and E). Measurement of organoid diameter over time revealed significant reduction of organoid growth in Pparg-OE KPA organoids compared with control KPA organoids (Figure 11F). GSEA revealed that Pparg-OE suppresses WNT pathway activation (Figure 11G), whereas Hnf4a expression is unaffected (Figure 11H), indicating that hepatocyte nuclear factor 4 α (HNF4A) does not act downstream of PPARγ and that PPARγ directly modulates WNT signaling. Moreover, Hnf4a-OE significantly impacted gene sets associated with similar metabolic pathways including fatty acid, bile acid, cholesterol, and xenobiotic ones (Figure 11I and Supplementary Table 4) and slightly upregulate the Pparg expression (Figure 11J). These data suggest that Bcl-xL may partially regulate PPARγ expression through HNF4A (Figure 11K).

Figure 11.

Bcl-xL–mediated suppression of PPARγ expression promotes tumor growth. (A–H) The infection of KPA organoids with Pparg-overexpressing adenovirus. (A) PPARγ staining in organoids infected with control adeno virus and Pparg-overexpressing adeno virus. (B) RT-PCR of Pparg expression (n = 3/group). The P value was calculated using a t test. (C–H) Bulk RNA sequencing comparing KPA organoids and KPA organoids overexpressing Pparg (n = 3/group). (C) PCA plot. (D) Sigmoid curve showing the result of the GSEA with gene sets for xenobiotic, bile acid, fatty acid, and cholesterol metabolism. (E) Heatmap of the lipid metabolism–related genes. (F) The ratios of organoid diameters relative to those on day 1 were quantified (n = 10 per group). The P value was calculated using a t test. (G) GSEA with Gene Ontology Biological Process (GOBP) WNT SIGNALING PATHWAY, WP WNT SIGNALING PATHWAY. v2025, and GOBP EPITHELIAL CELL PROLIFERATIOHN gene sets. (H) Boxplot of the Hnf4a expression based on the Bulk RNA sequencing comparing the control and Pparg-OE KPA organoids. The P value was calculated using a t test. (I) Sigmoid curve showing the result of the GSEA with gene sets for xenobiotic, bile acid, fatty acid, and cholesterol metabolism. (J) Boxplot of the Pparg expression based on the bulk RNA sequencing comparing the control and Hnf4a-OE KPA organoids. The P value was calculated using a t test. (K) Tumorigenic mechanism mediated by Bcl-xL. Scale bars, 100 μm. Mean ± standard error of the mean. ∗P < .05.

The Relationship Among B-Cell Lymphoma–Extra-Large, Peroxisome Proliferator–Activated Receptor γ, and WNT Pathway Validated by Single-Cell RNA Sequencing of Mice With Gastric Cancer

To validate the relationship between Bcl-xL, PPARγ, and the WNT pathway in gastric cancer, we performed single-cell RNA sequencing analyses using both mouse GC datasets.

For mouse GC, we utilized antral tumor data from Gp130^Y757F/Y757F mice (GSE217498).⁵¹ After manual clustering, epithelial cell clusters (clusters 2, 6, 10, and 12) positive for Epcam were isolated and further subdivided into Bcl2l1-positive and Bcl2l1-negative populations (Figure 12A–D). Bcl2l1 expression was used as a surrogate for Bcl-xL expression, as Bcl-xL is the predominant isoform encoded by Bcl2l1 in mice.⁵² Bcl2l1-positive cells exhibited reduced Pparg expression (Figure 12E), and GSEA using Hallmark gene sets revealed significant downregulation of fatty acid, bile acid, and cholesterol metabolism pathways (Figure 12F). In contrast, WNT-related gene sets were significantly enriched in the Bcl2l1-positive cell population (Figure 12G).

Figure 12.

Analysis of Bcl-xL and PPARγ expression and WNT pathway activity in single-cell RNA sequencing of mouse GC. (A–G) Single-cell RNA sequencing analysis of mouse GC (Gp130^Y757F/Y757F mice). (A) Uniform Manifold Approximation and Projection (UMAP) visualization of single-cell transcriptomic profiles. (B) Violin plot showing Epcam expression across clusters. Red arrows show the Epcam-positive clusters. (C) UMAP highlighting Bcl2l1-positive and -negative cells within epithelial cell clusters. (D) Dot plot showing Bcl2l1 expression. (E) Dot plot showing Pparg expression. (F) GSEA using Hallmark gene sets comparing Bcl2l1-positive and -negative cells. (G) Sigmoid curve showing GSEA results for the Gene Ontology Biological Process (GOBP) WNT signaling pathway.

Collectively, these single-cell transcriptomic analyses demonstrate a conserved association between Bcl2l1 expression, suppression of metabolic programs, and activation of WNT signaling in mouse GC, supporting the translational relevance of our Bcl-xL–deficient models.

B-Cell Lymphoma–Extra-Large and Peroxisome Proliferator–Activated Receptor γ Are Prognostic Factors in Human Gastric Cancer

To assess whether the findings from mouse models in vivo and in vitro are also relevant to human GC, we examined the expression of Bcl-xL (or BCL2L1) and PPARG in human GC samples.

First, immunofluorescence staining for Bcl-xL was performed using a human GC tissue array. A total of 72 cases were scored based on the intensity of Bcl-xL staining and stratified by clinical stage, showing that Bcl-xL expression increased with tumor stage, suggesting a correlation between Bcl-xL expression and cancer progression (Figure 13A).

Figure 13.

Analysis of Bcl-xL/BCL2L1 and PPARγ expression in human gastric cancer database. (A) Bcl-xL staining in human GC tissue array. The grade of the Bcl-xL positivity is quantified across the tumor stage (n = 72). (B) Box plot showing the BCL2L1 expression comparing the normal stomach (n = 34) and primary tumor tissue (n = 415) based on the TCGA database. The P value was calculated using a t test. (C) The heat map of the AXIN2, CCND1, LGR5, and SOX9 expression comparing between high- and low- BCL2L1 expressed GC based on the TCGA database (n = 225 vs 225). (D) Sigmoid curve showing the result of the GSEA with gene sets for bile acid and fatty acid metabolism, comparing between high- and low- BCL2L1 expressed GC based on the TCGA database (n = 206 vs 206). (E) Survival analysis of patients with gastric cancer (n = 875) stratified by high vs low BCL2L1 and high vs low PPARG expression. The P value was calculated using a log-rank test. (F) Analysis of Spearman’s rank coefficient between BCL2L1 and PPARG expression (n = 450). (G–J) The analysis of public single-cell RNA sequencing data for GC (GSE239676) comparing tumor epithelial samples with high BCL2L1 expression with those with low BCL2L1 expression (n = 37). (G) Uniform Manifold Approximation and Projection (UMAP) plot colored by GC patient IDs. (H) UMAP plot stratified on the basis of BCL2L1 expression. (I) PCA plot. (J) GSEA with Gene Ontology Biological Process (GOBP) WNT SIGNALING PATHWAY, WP WNT SIGNALING PATHWAY. v2025, GOBP EPITHELIAL CELL PROLIFERATIOHN, and Hallmark bile acid metabolism gene sets and downregulated differentially expressed genes (DEGs) in Hnf4a-OE KPA organoid mice. Scale bars, 100 μm. Mean ± standard error of the mean. ∗P < .05.

Using the UALCAN application and The Cancer Genome Atlas (TCGA) database,⁵³ we next evaluated BCL2L1 expression in normal gastric mucosa and GC and found that BCL2L1 expression was elevated in GC compared with normal gastric mucosa (Figure 13B). BCL2L1 expression was used as a surrogate for Bcl-xL expression, as Bcl-xL is the predominant isoform encoded by BCL2L1 in human normal tissue and cancers.⁵⁴

Next, we analyzed gene expression profiles in 450 GC samples from TCGA with the Phantasus application.⁵⁵ Cases were divided into BCL2L1-high and BCL2L1-low groups, and expression levels of Wnt target genes previously identified in mouse and organoid models (including CCND1, LGR5, AXIN2, and SOX9) were compared. This analysis revealed that CCND1, LGR5, AXIN2, and SOX9 were significantly upregulated in the BCL2L1-high group, supporting the idea that Bcl-xL enhances Wnt activation in human GC (Figure 13C).

Next, we analyzed the gene expression profiles in 412 GC samples from the PanCancer Atlas and analyzed the association between BCL2L1 and several gene sets. GSEA showed the downregulation of metabolic pathways including bile acid and fatty acid metabolism, compatible with our findings from mice experiments (Figure 13D and Supplementary Table 5).

We then explored the prognostic significance of BCL2L1 and PPARG expression using the Kaplan–Meier Plotter, a database linking gene expression to patient survival outcomes. In a cohort of 875 patients with GC, higher BCL2L1 expression was associated with a poorer prognosis, whereas higher PPARG expression correlated with better prognosis (Figure 13E). Furthermore, Spearman correlation analysis revealed a negative correlation between BCL2L1 and PPARG expression (r = −0.1591; P < .0001), suggesting a potential regulatory relationship between the 2 genes (Figure 13F).

For human GC, we utilized the Single-Cell Atlas of human GC (GSE239676).⁵⁶ GCs with high BCL2L1 expression showed significant enrichment of WNT pathway–related gene sets and of genes downregulated in Hnf4a-OE KPA organoids, whereas gene sets related to bile acid metabolism were downregulated, consistent with the findings in the mouse model.

To examine whether our findings may also apply to other gastrointestinal malignancies, we performed similar analyses using the TCGA database for esophageal cancer and colorectal cancer. In esophageal cancer, BCL2L1 expression was particularly elevated in esophageal adenocarcinoma compared with normal mucosa and squamous cell carcinoma (Figure 14A). Esophageal adenocarcinomas with high BCL2L1 expression showed activation of the WNT signaling pathway and reduced lipid metabolism compared with those with low BCL2L1 expression (Figure 14B). In colorectal cancer, BCL2L1 expression was also increased compared with normal mucosa (Figure 14C), and patients with BCL2L1-high and PPARG-low tumors had a worse prognosis than those with BCL2L1-low and PPARG-high tumors, respectively (Figure 14D). GSEA further demonstrated that BCL2L1-high colorectal cancers exhibited activation of the WNT signaling pathway and suppression of fatty acid metabolism compared with BCL2L1-low tumors (Figure 14E). Taken together, these results indicate that the association between BCL2L1 expression, WNT signaling, and metabolic pathways observed in GC may also be present in esophageal adenocarcinoma and colorectal cancer.

Figure 14.

Analysis of Bcl-xL/BCL2L1 and PPARγ expression in human esophageal and colorectal cancer database. (A) Box plot showing the BCL2L1 expression comparing the normal esophagus (n = 11), esophageal adenocarcinoma (n = 89) and esophageal squamous cell carcinoma (n = 95) based on the TCGA database. The P value was calculated using a t test. (B) Sigmoid curve showing the result of the GSEA with gene sets for WNT target genes and fatty acid metabolism, comparing between high- and low- BCL2L1 expressed esophageal adenocarcinoma based on the TCGA database (n = 44 vs 43). (C) Box plot showing the BCL2L1 expression comparing the normal stomach (n = 41) and primary colorectal cancer tissue (n = 286) based on the TCGA database. The P value was calculated using a t test. (D) Survival analysis of patients with GC (n = 1336) stratified by high vs low BCL2L1 and high vs low PPARG. The P value was calculated using a log-rank test. (E) Sigmoid curve showing the result of the GSEA with gene sets for WNT target genes and fatty acid metabolism, comparing between high- and low- BCL2L1 expressed colorectal cancer based on the TCGA database (n = 112 vs 112). Mean ± standard error of the mean. ∗P < .05.

Finally, we investigated the potential clinical relevance of PPARγ activation in GC prevention by analyzing data from 29,905 hospitalized diabetic patients in the Diagnosis Procedure Combination database. Since pioglitazone, a PPARγ agonist, is commonly used to treat diabetes, a known risk factor for GC, this analysis was restricted to diabetic patients (Table 1). Multivariate analysis adjusted for confounders showed that oral pioglitazone use was associated with a significantly reduced risk of GC (adjusted odds ratio [OR], 0.74; 95% confidence interval, 0.56–0.99; P = .0419).

Table 1.

The Association Between GC Development and Use of Pioglitazone in the Patients With Diabetes Mellitus

Factors	GC (n = 1378)	No GC (n = 28,527)	Crude OR (95% CI)	Adjusted ORa (95% CI)	P
Female	313 (2.90)	10,492 (97.10)	1	1
Male	1065 (5.58)	18,035 (94.42)	1.98 (1.74–2.25)	1.70 (1.48–1.96)	<.0001b
Age <70 y	378 (3.32)	10,996 (96.68)	1	1
Age >70 y	1000 (5.40)	17,531 (94.60)	1.66 (1.47–1.87)	1.70 (1.50–1.92)	<.0001b
Nonsmoking	635 (3.68)	16,617 (96.32)	1	1
Smoking	743 (5.87)	11,910 (94.31)	1.63 (1.47–1.82)	1.39 (1.24–1.57)	<.0001b
No H pylori eradication	1262 (4.31)	28,021 (95.69)	1	1
H pylori eradication	116 (18.05)	506 (81.57)	5.10 (4.13–6.28)	5.23 (4.23–6.48)	<.0001b
Charlson comorbidity index	3.38 ± 2.49	2.72 ± 2.13	–	–	–
No pioglitazone	1325 (4.64)	27,251 (95.36)	1	1
Pioglitazone	53 (3.99)	1276 (96.01)	0.85 (0.65–1.13)	0.74 (0.56–0.99)	.0419b
No metformin	1162 (4.69)	23,623 (95.31)	1	1
Metformin	216 (4.22)	4904 (95.78)	0.90 (0.77–1.04)	0.91 (0.79–0.86)	.267

CI, confidence interval; GC, gastric cancer; OR, odds ratio.

^aOR adjusted for age >70 years, sex, smoking, H pylori eradication, and Charlson comorbidity index.

^bP < .05.

Collectively, these findings suggest that Bcl-xL expression increases with tumor progression and promotes tumorigenesis via Wnt pathway activation in human gastric cancer, while PPARγ acts as a tumor suppressor negatively regulated by Bcl-xL. Moreover, activation of PPARγ by pioglitazone may provide a protective effect against GC development in diabetic patients.

Discussion

In this study, we demonstrate that Bcl-xL plays a previously unrecognized role in promoting gastric tumorigenesis beyond its canonical antiapoptotic function. Using genetically engineered mouse models, tumor organoids, and human clinical datasets, we show that Bcl-xL enhances GC progression primarily by augmenting Wnt–β-catenin signaling and tumor cell proliferation. Mechanistically, Bcl-xL negatively regulates the transcription factors HNF4A and PPARγ, thereby linking stem cell signaling, transcriptional control, and metabolic regulation during gastric tumor development.

Traditionally, Bcl-xL has been characterized as a key antiapoptotic member of the BCL2 family, contributing to tumor cell survival under stress conditions.⁵⁷ However, accumulating evidence suggests that BCL2 family proteins also exert noncanonical functions, including regulation of mitochondrial metabolism, transcriptional programs, and cell–cycle progression.⁵⁷ Our findings substantially expand this paradigm by identifying Bcl-xL as a positive regulator of oncogenic Wnt signaling in gastric epithelium. Notably, Bcl-xL deletion did not induce apoptosis in either normal or neoplastic gastric tissues, indicating that its tumor-promoting effects in this context are largely independent of cell death suppression. Instead, loss of Bcl-xL consistently attenuated proliferation, Wnt target gene expression, and tumor growth in vivo and in vitro, highlighting a novel and apoptosis-independent oncogenic function of Bcl-xL.

Our transcriptomic analyses revealed that Bcl-xL orchestrates broader transcriptional and metabolic programs through negative regulation of HNF4A and PPARγ. HNF4A is a key epithelial differentiation factor that has been reported to antagonize Wnt signaling and restrict proliferation in gastrointestinal tissues.⁵⁰ Restoration of HNF4A expression in tumor organoids suppressed Wnt pathway activation and inhibited growth, indicating that Bcl-xL–mediated repression of HNF4A represents a critical mechanism by which Wnt signaling is sustained in gastric tumors. In parallel, PPARγ, an essential regulator of lipid metabolism and cellular differentiation,⁵⁸ was markedly upregulated upon Bcl-xL loss. PPARγ activation might directly suppress Wnt signaling and significantly inhibited tumor growth and altered metabolic gene expression, suggesting that Bcl-xL promotes tumor progression in part by constraining differentiation-associated metabolic pathways. The partial regulation of PPARγ by HNF4A further suggests a hierarchical network linking transcriptional control and metabolic state in GC cells.

Emerging evidence indicates that stem cell signaling, Wnt–β-catenin activity, and cellular metabolism are tightly interconnected during gastrointestinal tumorigenesis.⁵⁹ Wnt signaling is not only a central driver of epithelial stemness and proliferation but also a key regulator of metabolic reprogramming, promoting anabolic pathways that support rapid cell division. In gastric and intestinal epithelia, sustained Wnt activation has been shown to enhance glycolysis and lipid biosynthesis while suppressing differentiation-associated metabolic programs.^60,61 Conversely, transcription factors such as HNF4A and PPARγ are known to promote epithelial differentiation and metabolic homeostasis and have been reported to antagonize excessive Wnt signaling, either directly or indirectly, by regulating Wnt-related gene expression and lipid metabolic pathways.^50,58 Disruption of this balance through persistent Wnt activation coupled with repression of metabolic differentiation programs favors the maintenance of a highly proliferative, stem-like tumor cell state. In this context, our findings suggest that Bcl-xL functions as a molecular integrator that reinforces Wnt-driven proliferative signaling while concomitantly suppressing HNF4A- and PPARγ-mediated metabolic programs, including fatty lipid, bile acid, xenobiotic, and cholesterol metabolism. The attenuation of these differentiation-associated metabolic pathways compromises normal epithelial differentiation, thereby creating a permissive environment for gastric tumor growth.

Several previous studies have reported the expression patterns of Bcl-xL in gastrointestinal tissues. Regarding the stomach,^38,62 Bcl-xL is broadly expressed in nearly all epithelial cells in mice based on immunohistochemistry. In humans, Bcl-xL expression was increased in gastric tumor tissues compared with adjacent normal mucosa, while some degree of Bcl-xL expression was also observed in normal tissue. In other organs, immunohistochemical analysis demonstrated diffused Bcl-xL expression in both normal and tumor tissues of the human and mice colon.^42,63 In contrast, Bcl-xL expression was relatively low in WT pancreatic tissue but progressively increased during neoplastic progression, with stronger expression in pancreatic intraepithelial neoplasia 2–3 lesions and further elevation in pancreatic ductal adenocarcinoma for both human and mice.^40,64 Taken together, these previous reports suggest that, consistent with our findings, broad Bcl-xL expression is observed in normal gastrointestinal tissues, whereas Bcl-xL expression patterns can vary depending on the tissue context and disease stage.

In addition to the tumor-intrinsic context examined in this study, the role of Bcl-xL may vary depending on the inflammatory microenvironment. Under conditions of chronic gastric inflammation, such as H pylori-associated gastritis and autoimmune gastritis,^65,66 epithelial cells are exposed to increased oxidative stress, cytokine signaling, and immune-mediated injury. In such settings, antiapoptotic factors including Bcl-xL may play a more prominent role in maintaining epithelial survival and tissue integrity. Chronic inflammatory signaling pathways, including NF-κB activation and cytokine-mediated stress responses, may increase the reliance of epithelial cells on Bcl-xL–mediated survival mechanisms. Therefore, although our current model does not demonstrate a major role for Bcl-xL in regulating apoptosis in established tumors under noninflammatory conditions, its functional contribution may be context dependent and could be enhanced during inflammation-driven gastric tumorigenesis.

Although APC deletion results in constitutive activation of Wnt–β-catenin signaling, accumulating evidence indicates that Wnt signaling output is not binary but exists along a graded spectrum.⁶⁷ Even in APC-deficient contexts, transcriptional activity of β-catenin–dependent target genes can be further modulated by additional regulatory factors. In our system, Bcl-xL enhances Wnt–β-catenin signaling; however, we did not observe impaired organoid growth or morphologic features consistent with excessive Wnt activation. This suggests that Bcl-xL-mediated augmentation likely operates within a permissive signaling range rather than exceeding the threshold that compromises proliferation. Moreover, external Wnt and R-spondin supplementation were included to maintain optimal culture conditions. APC-deficient organoids can still respond quantitatively to extracellular Wnt modulation, and signaling output is influenced by multiple layers of regulation beyond APC loss alone.

In summary, our study identifies Bcl-xL as a multifaceted regulator of gastric tumorigenesis that integrates stem cell signaling, transcriptional control, and metabolic regulation. These findings not only redefine the biological functions of Bcl-xL in GC but also highlight new therapeutic opportunities. Targeting Bcl-xL–dependent signaling networks, either directly or through modulation of downstream pathways such as Wnt or PPARγ, may represent a promising strategy for the prevention and treatment of GC.

Materials and Methods

Mice

Mist1-CreERT, Apc^flox/flox, and LoxP-stop-LoxP (LSL)-Trp53^R172H mice were provided by Columbia University, and bcl-xl^flox/flox mice were provided by Osaka University. LSL-TdTomato and LSL-Kras^G12D mice were purchased from Jackson laboratory. Lgr5-DTR-EGFP mice were obtained from Genentech (South San Francisco, CA).⁶⁸ WT and nude mice were purchased from Crea Japan. Mice other than nude mice were from the C57BL6J background. All mice were maintained in a specific pathogen-free facility at the University of Tokyo in accordance with the institutional guidelines for animal experimentation and were provided with commercially available food and water ad libitum. All animal studies and procedures were approved by the ethics committee of the University of Tokyo (ID: 24-2). Both male and female mice were used in this study. No obvious sex-based differences in phenotype were observed. For each experiment, animals were sex- and age-matched between experimental groups.

For analyses using mouse gastric tumor models, WT, Mist1-CreERT; bcl-xl^flox/flox, Mist1-CreERT; Apc^flox/flox, and Mist1-CreERT; Apc^flox/flox; bcl-xl^flox/flox mice were used. When mice were incorporated LSL-TdTomato, the genetically modified proteins were labeled with fluorescence. Lineage tracing was induced by administering TAM at 6 mg per 20 g body weight on 2 separate days, with a 1-day interval between injections to all 6-week-old mice. Thus, mice received a total of 2 TAM injections, and they were sacrificed at 2 months for analysis. For analysis using the organoid culture system, transgenic organoids were established from WT, Apc^flox/flox, LSL-Kras^G12D; LSL-Trp53^R172H; Apc^flox/flox, and LSL-Kras^G12D; LSL-Trp53^R172H; Apc^flox/flox; bcl-xl^flox/flox mice.

Organoid Culture

Organoids were generated as described previously.22, 23, 24 Mice were cervically dislocated, the gastric antrum was excised and washed with phosphate-buffered saline (PBS), then shaken in 5 mM EDTA (Invitrogen) + PBS for 45 minutes at 4 °C. Cells derived from the gastric mucosa were detached from the serosa; the required amount of cells were collected, dissolved in Matrigel (Corning), seeded in 24-well plates, and medium was added 30 minutes later. The medium was a Wnt3a and R-spondin 3 conditioned medium⁶⁹ prepared from L-WRN cells, which are modified L-Wnt3a cells derived from mouse L cells, and advanced DMEM and F12 (Invitrogen) mixed 2:1. It was then mixed with penicillin and streptomycin (Wako), Gluta MAX (Invitrogen), HEPES (Invitrogen), B27 (Invitrogen), N2 (Invitrogen), N-acetylcysteine (Wako, 1 mM), and Nicotinamide (Sigma-Aldrich, 10 mM). Epidermal growth factor (Wako, 50 ng/mL), fibroblast growth factor 10 (Peprotech, 100 ng/mL), Y-27632 (Wako, 10 μM), Gastrin (Tocris, 0.01 μM), and Primocin (Invitrogen) were used in the mixture. The medium was changed twice a week. For the quantification of organoid diameters, n = 10 refers to 10 organoids randomly selected from the same condition.

Treatment In Vitro

For the pharmacologic inhibition of Bcl-xL, A-1155463 (Selleck: S7800) was used. A-1155463 is a highly potent and selective Bcl-xL inhibitor that exhibits picomolar binding affinity for Bcl-xL, while showing more than 1000-fold weaker binding affinity for Bcl-2 and related proteins, including Bcl-w (Ki = 19 nM) and Mcl-1 (Ki > 440 nM), as described in the manufacturer’s documentation.

A-1155463 (5 mg) was dissolved in 0.7465 mL of DMSO to prepare a 10-mM stock solution, which was further diluted to generate working solutions at the indicated concentrations. During organoid passaging, the Bcl-xL inhibitor at each concentration was added to the culture medium. After 72 hours, organoids were passaged again. Starting the day after passaging, organoid images were captured, and the long-axis diameter of each organoid was measured to compare changes in organoid growth over time.

Xenograft

Organoids are composed solely of epithelial cells and are not influenced by the tumor microenvironment, such as stromal or immune components. Therefore, xenograft models are essential for evaluating the interaction between tumor cells and their microenvironment.⁴⁹ Organoid cells (1 × 10⁶ cells) were added to 100 μL of Matrigel and 100 μL of advanced DMEM and F12 and injected subcutaneously into the back of nude mice at 6 weeks after birth with a 26 G needle. The tumor diameters of the subcutaneous tumors were measured, and the volumes were calculated every week. After 1 month, the mice were sacrificed, and the subcutaneous tumors were removed for analysis.

Hematoxylin and Eosin Stain

Hematoxylin and eosin staining was performed on paraffin-embedded sections of mouse tissues and organoids fixed in 10% buffered formalin solution (pH 7.4). After deparaffinization and rinsing, nuclear staining was performed with hematoxylin. After rinsing in running water, the sections were stained with eosin, dehydrated, permeabilized, sealed with Mount Quick (DAI), and the specimens were examined.

Immunohistochemistry

Immunohistochemistry was performed on formalin-fixed and paraffin-embedded sections of mouse tissue. The sections were deparaffinized, autoclaved at 120 °C for 10 minutes for antigen activation, soaked in 3% hydrogen peroxide for 5 minutes for inactivation of endogenous peroxidase, and blocked in 10% goat serum for 30 minutes. The sample was outlined with a water-repellent pen and reacted with primary antibody diluted in PBS mixed with 2.5% Bovine Serum Albumin Cohn Fraction V (hereafter referred to as 2.5% BSA) and allowed to stand overnight at 4 °C. The next day, the antibody was washed with PBS, followed by the reaction of the secondary antibody, and allowed to stand at 4 °C for 1 hour. After washing with PBS, the antibody was incubated with avidin-biotin–labeled enzyme complex (VECTASTAIN Elite ABC Kit [Vector Laboratories]) for 20 minutes at room temperature, the avidin-biotin complex method was performed, and the reaction was carried out with enzyme substrate solution (Liquid 3,3'-diaminobenzidine [DAB] + Substrate Chromogen System, DAKO) for tens of seconds to several minutes, followed by rinsing. Nuclear staining was performed with hematoxylin, followed by rinsing, dehydration, permeabilization, sealing with MountQuick (DAI), and mirror examination. Bcl-xL (1:200, Cell Signaling Technology: 2764S), cyclinD1 (1:200, Thermo Fisher Scientific: RM-91004-SO), CD44v (1:200, BIO RAD: MCA1967), and cleaved caspase 3 (1:200, Cell Signaling Technology: 9661S) were used for immunohistochemistry.

Immunofluorescence

Fluorescent immunostaining was performed on formalin-fixed and paraffin-embedded sections of mouse tissues and organoids. The sections were deparaffinized and autoclaved at 120 °C for 10 minutes for antigen activation. The samples were surrounded with a pen for water generation, reacted with primary antibody diluted in 2.5% BSA, and left overnight at 4 °C. After washing with PBS the next day, secondary antibodies (Alexa Fluor 555, Alexa Fluor 488, Alexa Fluor 647 [Invitrogen]) diluted 1:200 in 2.5% BSA were reacted and allowed to stand at room temperature for 1 hour. Fluoro-KEEPER Antifade Reagent, Non-Hardening Type with 4′,6-diamidino-2-phenylindole (Nacalai), and then the cells were stained, sealed, and microscopically examined. Bcl-xL (1:400, Cell Signaling Technology: 2764S), β-catenin (1:400, BD Biosciences: BD61053), Ki67 (1:500, Cell Signaling Technology: 12202S), Sox9 (1;400, Merck: AB5535), YAP (total-YAP; 1:200, Cell Signaling Technology: 14074), red fluorescent protein (1:400, ROCKLAND: 600401379), GFP (1:400, Abcam: ab13970), cyclinD1 (1:400, Thermo Fisher Scientific: RM-91004-SO), cleaved caspase 3 (1:200, Cell Signaling Technology: 9661S), CD44v (1:200, Bio RAD: MCA1967), E-cadherin (1:200, BD biosciences: 610181), AQP5 (1:200, ATLAS ANTIBODY, HPA065005), TROP2 (1:200, R and D, AF650), and PPARγ (1:400, Cell Signaling Technology: 2435) were used for immunohistochemistry.

In Situ Hybridization

In situ hybridization was performed on formalin-fixed, paraffin-embedded sections of mouse tissue using the RNAscope 2.5 HD Detection Kit BROWN (ACD). The process involved 60 minutes of baking at 60 °C, 2 5-minute dips in xylene, 100% soaked in ethanol twice for 5 minutes and dried for 5 minutes; then baked at 60 °C for 30 minutes, soaked in hydrogen peroxide solution for 10 minutes, then soaked in 1× Target Retrieval Buffer (ACD) and boiled at 100 °C for 15 minutes. The cells were washed with distilled water, dehydrated with 100% ethanol, soaked in 5 to 8 drops of hydrogen peroxide (ACD), allowed to stand at room temperature for 10 minutes, soaked in Protease Plus (ACD) for 30 minutes at 40 °C, washed with distilled water, and hybridized at 40 °C for 2 hours with Lgr5 probe (ACD: 312171). The cells were washed in 1× wash buffer (ACD) for 2 minutes, immersed in 20× saline sodium citrate buffer (1000 mL distilled water with NaCl 175.3 g + sodium citrate 88.2 g adjusted to pH 7.0 and autoclaved sterile) and left overnight. After washing with 1× wash buffer for 2 minutes, AMP1 (ACD) was added dropwise to cover the area enclosed by the hydrophobic barrier pen and allowed to stand for 30 minutes at 40 °C. After washing with 1×wash buffer for 2 minutes, AMP2 (ACD) was added drop and allowed to stand for 15 minutes at 40 °C. After washing with 1× wash buffer for 2 minutes, AMP3 (ACD) was added drop and allowed to stand for 30 minutes at 40 °C. The cells were washed with 1× wash buffer for 2 minutes, AMP4 (ACD) was added drop and allowed to stand at 40 °C for 15 minutes. The cells were washed with 1× wash buffer for 2 minutes, AMP5 (ACD) was added drop and allowed to stand at room temperature for 30 minutes. The cells were washed with 1× wash buffer for 2 minutes, AMP6 (ACD) was added drop and allowed to stand at room temperature for 15 minutes. The cells were washed with 1× wash buffer for 2 minutes, then DAB-A and DAB-B (ACD) were added dropwise at 1:1 and allowed to stand for 10 minutes at room temperature. After rinsing, nuclear staining was performed with hematoxylin, soaked in 0.02% ammonia water, rinsed, soaked in 70% ethanol twice for 5 minutes, soaked in 100% ethanol twice for 5 minutes, dehydrated, permeated with xylene twice for 5 minutes, sealed with Mount Quick (DAI), and specimens were examined.

Histopathologic Analysis

All images were captured with a Leica DMi8 inverted fluorescence microscope and manually quantified after visualization using LAS X software. For each mouse, the average value was calculated from 3 randomly selected, well-oriented fields at 10× magnification. Quantification was conducted manually on fluorescence or chromogenic images by investigators blinded to the experimental conditions. Group-level values were calculated by averaging across animals and are presented as mean ± standard error of the mean. Unless otherwise indicated, analyses were performed using 3 to 5 biologically independent mice per group.

Western Blotting

Gastric tissue and organoid cells were homogenized by adding cell lysate buffer (50 mM Tris HCl, 1% Triton-X, 5 mM EDTA, 1 mM Na3VO4, 1.25 mM NaF, protease inhibitor cocktail (cOmplete Mini, Roche) and centrifuged at 12,000 g for 5 minutes to extract proteins from the supernatant. Proteins were separated by polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride membrane (Pall Corporation). The cells were blocked by shaking with 5% milk for 1 hour, washed with Tris Buffered Saline with Tween 20 (TBS-T), and reacted with the primary antibody by shaking overnight at 4 °C. The reaction was repeated at 4 °C for 1 hour. After washing with TBS-T, the antibody was reacted with horseradish peroxidase–conjugated secondary antibody (GE Healthcare) while shaking at room temperature for 1 hour. After washing with TBS-T, proteins were detected by exposure with a chemiluminescence detection system using the luminescent reagent ImmunoStar LD (Fujifilm Wako Pure Chemicals). Bcl-xL (1:1000, Cell Signaling Technology: 2764S), cyclinD1 (1:1000, Thermo Fisher Scientific: RM-91004-SO), PPARγ (1:1000, Cell Signaling Technology: 2435), total caspase-3 (1:1000, Cell Signaling Technology: 9662), cleaved caspase-3 (1:1000, Cell Signaling Technology: 9664), and β-actin (1:10,000, cell signaling technology: 5125S) antibodies were used for Western blotting.

Real-Time Polymerase Chain Reaction

RNA extraction from gastric tissue was performed using NucleoSpin RNA (Macherey-Nagel), and RNA extraction from organoids was performed using RNeasy Micro kit (Qiagen). cDNA was synthesized from RNA using ImProm-ll Reverse Transcription System (Promega). Amplification of cDNA was performed using StepOnePlus Real Time PCR System Applied Biosystems and FastStart Universal SYBR Green Master (Rox) (Roche). The results obtained were standardized and analyzed with glyceraldehyde-3-phosphate dehydrogenase. A list of primers used is given in Table 2.

Table 2.

Primer Sequences Used for Quantitative PCR

Targets	Forward or reverse	Sequences
Lgr5	Forward	5'-CCACAGCAACAACATCAGGT-3'
	Reverse	5'-AACAAATTGGATGGGGTTGT-3'
Lgr4	Forward	5'-CCTTCACCCAAGCACTGGATA-3'
	Reverse	5'-AGGTCGTTACCAGCCAGTTG-3'
Axin2	Forward	5'-TGCATCTCTCTCTGGAGCTG-3'
	Reverse	ACTGACCGACGATTCCATGT-3'
Sox9	Forward	5'-GAGGCCACGGAACAGACTCA-3'
	Reverse	5'-CAGCGCCTTGAAGATAGCATT-3'
Cd44	Forward	5'-ACAGTACCTTACCCACCATG-3'
	Reverse	5'-GGATGAATCCTCGGAATTAC-3'
Cmyc	Forward	5'-GAGCTGTTTGAAGGCTGGATTT-3'
	Reverse	5'-TCCTGTTGGTGAAGTTCACGTT-3'
Fzd7	Forward	5'-TGCAGTCATGGCGTCGCTTCTA-3'
	Reverse	5'-TTCGGCCACTGCCTTTGCCTTT-3'
Fzd5	Forward	5'-ATCCTCCGAGAGTTCTGTCCTTGG-3'
	Reverse	5'-TCGTCTCCTTCTTCCCTTTGCCT-3'
Ccnd1	Forward	5'-CGTGGCCTCTAAGATGAAGGA-3'
	Reverse	5'-TCGGGCCGGATAGAGTTGT-3'
Mist1	Forward	5'-GTTTTCCCAGGAGGTGGTG-3'
	Reverse	5'-AAGCTACGTGTCCTTGTCCC-3'
Bcl2l1	Forward	5'-GTGGAAAGCGTAGACAAGGA-3'
(Bclxl)	Reverse	5'-GTTCCCGTAGAGATCCACAA-3'
Bcl2	Forward	5'-ATGGGGTGAACTGGGGGAGGATTG-3'
	Reverse	5'-GGCCAGGCTGAGCAGGGTCTTC-3'
Bax	Forward	5'-TTGCTGATGGCAACTTCAAC-3'
	Reverse	5'-GATCAGCTCGGGCACTTTAG-3'
Bak	Forward	5'-AGCCGGGAATGCCTACGAAC-3'
	Reverse	5'-GGCCCAACAGAACCACACCA-3'
Pparg	Forward	5'-GGAGCCTAAGTTTGAGTTTGCTGTG-3'
	Reverse	5'-TGCAGCAGGTTGTCTTGGATG-3'
Gapdh	Forward	5'-AGAAACCTGCCAAGTATGATGAC-3'
	Reverse	5'-TCATTGTCATACCAGGAAATGAG-3'

Lentivirus Transduction

Organoids in culture were washed with PBS and recovered from Matrigel (Corning) basement membrane matrix using Cell Recovery Solution (Corning). We collected single-celled organoids using TrypLE Express (Thermo Fisher Scientific), added pLV-EF1-Cre-2a-vsfGFP-2a-puro lentivirus (originated in Tokyo University²⁴) at an optimized concentration for infection and the above medium (after adding a solution of 60 μL) and Trans Dux MAX (System Biosciences, 0.6 μL), centrifuged at 300 g for 30 minutes and mixed well with a pipette 3 times, dissolved in Matrigel, seeded in 24-well plates, and added medium 60 minutes later. The next day, the virus infection was confirmed by microscopic observation of GFP expression. Two to 3 days later, puromycin was added to the medium at 0.5 to 10 μg/mL to select for organoids undergoing genetic recombination. Other methods of selection included culture in medium without Wnt3A conditioned medium or R-spondin 3 conditioned medium, or by combining the method of collecting and culturing only organoids expressing GFP by genetic recombination under a microscope. We were unable to establish stable knockout organoids from either bcl-xL^flox/flox mice or Apc^flox/flox; bcl-xL^flox/flox mice following Cre-expressing lentivirus-mediated recombination, despite repeated attempts (n > 5). To analyze the cell proliferation rate of organoids, organoids of similar initial size (n = 10) were selected on day 1, and longitudinal changes in length diameter were measured using tile-scan imaging.

Adenovirus Infection

Adenoviruses were purchased in vectorbuilder (VB210912-1088ebt: pAV[Exp]-EF1A>mPparg[NM_011146.3](ns):T2A:DsRed_Express2 and VB220822-1173ydv: pAV[Exp]-CAG>mHnf4a[NM_008261.3]:IRES:mKate2) and were suspended in organoid cells, arranged for multiplicity of infection of 100, after dissolution in Matrigel. After 3 hours, organoids were seeded in 24-well plates. After 48 hours, the medium was replaced with medium without growth factors. Organoids without the selection procedure were harvested after 24 hours, and their diameters were calculated for proliferation analysis in 5 days.

The gene expression was analyzed by Western blotting, qPCR, and RNA sequencing.

RNA Sequencing

Sequences were performed on RNA extracted from organoids using the NovaSeq 6000 system from Illumina. Global normalization was performed on the obtained read information using TopHat-2.0.13 bowtie 2-2.2.3 as the alignment tool and Subio Platform as the analysis tool. Variability multipliers greater than 1.7 and P values less than .05 after correction for multiple testing using the Benjamini–Hochberg method were counted as significant variable genes. Gene Ontology analysis and Kyoto Encyclopedia of Genes and Genomes pathway analysis were performed using the Database for Annotation, Visualization, and Integrated Discovery. GSEA was also performed to analyze gene groups with significant expression variation. A false discovery rate threshold of 0.25 was used to define statistical significance. However, to illustrate similarities between several models, sigmoid curves were plotted to visualize overall trends, including gene sets that did not reach statistical significance.

Data Analysis of Human Gastric Cancer, Esophageal Cancer, and Colorectal Cancer

First, the expression status of Bcl-xL was evaluated by fluorescent immunostaining using human GC tissue microarrays. The tissue microarrays were purchased from US Biomax, and a total of 72 cases were scored according to the intensity of Bcl-xL expression and compared at each stage.

Second, using the UALCAN application and the TCGA database,⁵³ we evaluated BCL2L1 expression in normal gastric mucosa and GC, esophageal cancer, and colorectal cancer.

Third, RNA sequence data of 450 GC cases registered in TCGA were analyzed using the web tool “Phantasus” (https://artyomovlab.wustl.edu/phantasus/),⁵⁵ and the expression status of BCL2L1 was compared with WNT target genes.

Fourth, RNA sequence data of 412 GC (PanCancer Atlas), 87 esophageal adenocarcinoma (PanCancer Atlas), and 224 colorectal cancer⁷⁰ cases registered in TCGA were downloaded via Cbioportal website (https://www.cbioportal.org/) and analyzed with GSEA.

Fifth, we used the Kaplan-Meier Plotter (http://kmplot.com/analysis/index.php?p=service&cancer=gastric), a web tool for analyzing the gene expression database and the prognosis of patients with GC and colorectal cancer, to examine the contribution of each gene to survival of 875 patients with GC and 1336 patients with colorectal cancer.

Finally, we compared the occurrence of GC with gender, age, smoking, eradication of H pylori, comorbidity (Charlson comorbidity index), pioglitazone and metformin status in 29,905 hospitalized diabetic patients registered in the DPC of the University of Tokyo Hospital from April 2014 to March 2019, as described previously.71, 72, 73, 74, 75, 76 The ethics review number was 2058-[2]. Regarding the codes used to identify the disease, diabetes mellitus were E100, E101, E106, E108–E111, E116, E118–E121, E126, E128–E131, E136, E138 –E141, E146, E148, E149, E102–E105, E107, E112–E117, E122–E125, E132–E135, E137 E142–E145, and E147, and gastric cancers were C160, C161, C162, C163, C164, C165, C166, C168, and C169. Regarding the drug codes, pioglitazone was 610432040, 610432041, 621990901, 621991001, 621986301, 621986401, 622048401, 622048501, 622086001, 622086101 and metformin was 620004480, 620005570, 621676001, 621974701, 621986301, 621986401, 622242501, 622412701, 622417101, 622417201, 622421101, 622421201, 622421901, 6224 622422001, 622424401, 622424501, 622427201, 622427301, 622432601, 622432701, 622436301, 622438401, 622438501, 622448601, 622450301, 622450401, 6224 622466601, 622517101, and 622654401. Crude OR and adjusted OR (adjustment factors were age, gender, Charlson comorbidity index, smoking, and H pylori eradication) were calculated using SAS as statistical software, and P < .05 was considered significant.

The Analysis of Single-Cell RNA Sequencing

Single-cell RNA sequencing data of mice GC samples and patients with GC were obtained from GSE217498 and GSE239676, respectively. Filtered count matrices were imported to R for further analysis using the Seurat package. Dimensionality reduction was performed by PCA and Uniform Manifold Approximation and Projection embedding with 30 principal components and resolution 1.0. Epithelial cells were then subsetted based on epithelial markers. For the human dataset, BCL2L1 expression in tumor epithelial cells was obtained using the “AverageExpression” function in Seurat. Patients were then classified into BCL2L1-high or BCL2L1-low. Then we conducted differential expression and GSEA on pseudobulk workflow by using “AggregateExpression” function in the Seurat and R packages DESeq2 and fgsea.

Quantification and Statistical Analysis

Continuous variables are expressed as a means with 95% standard deviations, whereas categorical variables are expressed as numbers and frequencies (%). The differences between the means were compared using either the Student t test or the Wilcoxon test. Statistical significance was set at P < .05.