Calcium-modulated GJB6-GRHL3 positive feedback loop attenuates ESCC progression through AKT signaling pathway inhibition Authors: Jing Ren, Zhinan Zhou, Yanyan Dong, Yingzhen Gao, Xin Yang, Yingying Zhang, Mengyuan Yang, Xiaolong Cheng, Yongping Cui, Pengzhou Kong Published: 2026 ABSTRACT Esophageal squamous cell carcinoma (ESCC) is an aggressive malignancy characterized by dismal outcomes, but its molecular pathogenesis remains incompletely understood. To uncover critical driver genes in ESCC, we conducted a functional screen using a CRISPR/Cas9 sgRNA library targeting 406 copy-number alteration (CNA)-related genes. Among the candidates, gap junction protein beta 6 (GJB6), a major connexin family member, emerged as a tumor suppressor that inhibits malignant phenotypes. Clinically, reduced GJB6 expression significantly correlates with ESCC development and predicts unfavorable outcomes. Mechanistically, we delineated a reciprocal regulation between GJB6 and its downstream gene GRHL3: GJB6 enhances GRHL3 expression via Ca²⁺-dependent signaling, while GRHL3 transcriptionally upregulates GJB6, thereby forming a positive feedback circuit to exert their tumor-suppressive functions. Disruption of the GJB6/GRHL3 reciprocal regulatory axis activates AKT signaling, thereby driving ESCC pathogenesis. Notably, the oncogenic effects of GJB6 ablation could be pharmacologically reversed by the AKT inhibitor capivasertib, suggesting a potential therapeutic strategy for GJB6-deficient ESCC patients. Collectively, our findings establish GJB6 as both a critical suppressor and a clinically actionable prognostic biomarker, highlighting the potential of drug repurposing approaches for ESCC treatment. Introduction Esophageal cancer is a highly aggressive gastrointestinal malignancy, ranking seventh in morbidity and sixth in mortality rates with dismal clinical outcomes. The majority of esophageal cancer cases in China are histologically classified as esophageal squamous cell carcinoma (ESCC). According to previous reports, approximately 50% of global ESCC cases occur in China1,2. Despite rapid advances in diagnosis and treatment achieved over recent years, ESCC still has an unfavorable prognosis and lacks therapeutic targets as well as their relevant predictor biomarkers. Hence, it is imperative to conduct integrative studies focusing on critical genes and signaling pathways during ESCC tumorigenesis. We previously depicted genome-wide profiles of ESCC and identified numerous focal copy-number alterations (CNAs) segments3,4. Here, we used a CRISPR/Cas9 sgRNA library to identify critical CNA-related genes. Among the candidate genes, gap junction protein beta 6 (GJB6), a major connexin protein, was selected for further investigation due to its largely unknown biological role and clinical significance in ESCC. Intercellular communication is essential for maintaining tissue homeostasis through the exchange of metabolites and signaling molecules. Connexins are the main components of gap junctional intercellular communication (GJIC), which mediates vital physiological functions encompassing cell proliferation, migration, apoptosis, and survival5–7. Gap junctions, formed by connexin hexamers (connexons), facilitate intercellular communication, but their dysfunction and aberrant connexin expression are linked to cancer8,9. Located at chromosome 13q12.11, GJB6 encodes one of the connexin proteins. GJB6 abnormalities can lead to hearing loss, dermatological and nervous system disorders in cochlea and brain10–12. In addition, some researchers have explored the potential underlying mechanisms in malignant tumors. GJB6 serves as a biomarker for prognostic evaluation for breast cancer stratification and gastric cancer intestinal phenotype13,14. Functioning as a suppressor, GJB6 profoundly inhibits proliferation and migration by modulating IGF-1R or downstream kinases pAkt/pErk in glioma, suggesting it may be a novel therapeutic target15. What’s more, GJB6 could impede self-renewal through the repression of glioma stem cell markers, such as CD133 and c-Myc16. To date, studies on GJB6 in ESCC progression and prognosis have not been reported. In this study, we employed a CRISPR/Cas9 sgRNA library to uncover functionally pivotal CNA-associated genes and found that GJB6 acts as a tumor suppressor in ESCC. Integrated analysis of our data with public datasets positioned GRHL3, a grainyhead-like transcription factor, as a downstream target of GJB6. Furthermore, we revealed a positive feedback loop of GJB6/GRHL3, wherein GJB6 enhances GRHL3 expression by augmenting intracellular Ca2+ flux, and conversely, GRHL3 transcriptionally upregulates GJB6. Inhibition of the tumor suppressors GJB6 and GRHL3 activates the AKT signaling pathway, thereby promoting ESCC tumorigenesis. Clinically, our findings demonstrated that the Food and Drug Administration (FDA)-approved AKT inhibitor capivasertib (AZD5363, CAPI) might serve as a promising drug repurposing strategy with therapeutic potential for ESCC patients harboring GJB6 loss. Taken together, our work elucidated the prognostic value of GJB6, dissected its molecular mechanisms, and proposed targeted therapeutic strategies. This highlights the potential of AKT inhibition as a therapeutic intervention for ESCC patients with low GJB6 expression. Results Low GJB6 expression correlates with poor prognosis of ESCC To identify novel ESCC-associated drivers, we designed a CRISPR/Cas9 library targeting 406 genes with copy-number variation, including 140 amplified and 266 deleted genes, all of which were high-frequency CNA genes identified through preliminary large-scale sequencing in our own cohorts3,4. ESCC cell lines KYSE150 and KYSE450 were infected with lentiviruses harboring CRISPR-associated nuclease 9 (Cas9) to establish stable cell lines, and then reverse-transfected with the single guide RNA (sgRNA) library. After 72 hours of incubation, the cell growth was monitored by CCK-8 assay (Fig. 1A). Genes were considered significant if they showed a ≥ 1.5-fold increase or ≤ 0.5-fold decrease in proliferation relative to the control in the sgRNA library (Fig. 1B). In particular, GJB6 was noted for further research due to its unclarified function and mechanism, and its obvious effect in suppressing tumor cell growth in KYSE150-Cas9 and KYSE450-Cas9. To reconfirm the relationship between copy-number variation and gene expression for GJB6, which is located within a region of frequent deletions (Figure S1A), we conducted the linear regression analysis. As expected, there was a strong correlation between GJB6 copy number and its mRNA expression in 155 clinical samples (GSA accession number HRA003107)17 (Fig. 1C). Moreover, GJB6 deletion indicated poor outcomes of ESCC patients based on our in-house sequencing data from 490 ESCC patients (GSA accession number HRA000021)3 (Fig. 1D). Then a pan-cancer analysis using GEPIA 2.0 demonstrated widespread GJB6 expression divergence, particularly in esophageal cancer (ESCA) (P < 0.05, Figure S1B). Immunohistochemistry (IHC) studies were also performed to reveal higher expression of GJB6 in normal esophageal tissues compared with that in matched tumors (Fig. 1E). In addition, we analyzed GJB6 mRNA levels in the transcriptomic data from two cohorts: 155 in-house ESCC samples and 179 ESCC samples from the GSE53625 dataset. Both paired and unpaired t-tests demonstrated that GJB6 mRNA levels were markedly reduced in ESCC tissues compared to adjacent normal tissues (Fig. 1F, G, and Figure S1C, D). With the progression of clinical stage, the expression level of GJB6 was also reduced (Figure S1E). For the intrinsically complex microenvironment of ESCC, single-cell analysis from GSE188900 was further adapted to explore GJB6 expression in various cell clusters. It was indicated that GJB6 was mainly distributed in epidermal cells, which suggests that it may be a valuable marker in ESCC (Figure S1F-G).Fig. 1The downregulation of GJB6 correlates with its copy-number deletion in ESCC.A High-throughput screen of sgRNA library revealing genes critical for ESCC proliferation. B Heatmap demonstrating the significant genes in KYSE150-Cas9 and KYSE450-Cas9. C Correlation analysis of GJB6 copy number and mRNA expression (r = 0.2413, P = 0.0025). D Kaplan-Meier analysis of overall survival for 490 ESCC patients (GSA accession number HRA000021)3 stratified by GJB6 copy-number level. E Representative images of GJB6 expression in primary ESCC and matched tumor-adjacent tissues (n = 5). Decreased GJB6 mRNA levels in ESCC tissues compared with adjacent normal specimens for 155 cases using an unpaired (F) and paired t-test (G). H Kaplan-Meier analysis grouped by high and low GJB6 expression in 155 ESCC patients. I Multivariate Cox regression analysis for overall survival in 155 ESCC patients. Error bars, SDs; *P < 0.05; **P < 0.01; ****P < 0.0001. The downregulation of GJB6 correlates with its copy-number deletion in ESCC. A High-throughput screen of sgRNA library revealing genes critical for ESCC proliferation. B Heatmap demonstrating the significant genes in KYSE150-Cas9 and KYSE450-Cas9. C Correlation analysis of GJB6 copy number and mRNA expression (r = 0.2413, P = 0.0025). D Kaplan-Meier analysis of overall survival for 490 ESCC patients (GSA accession number HRA000021)3 stratified by GJB6 copy-number level. E Representative images of GJB6 expression in primary ESCC and matched tumor-adjacent tissues (n = 5). Decreased GJB6 mRNA levels in ESCC tissues compared with adjacent normal specimens for 155 cases using an unpaired (F) and paired t-test (G). H Kaplan-Meier analysis grouped by high and low GJB6 expression in 155 ESCC patients. I Multivariate Cox regression analysis for overall survival in 155 ESCC patients. Error bars, SDs; *P < 0.05; **P < 0.01; ****P < 0.0001. Next, we examined the association between GJB6 expression levels and clinical parameters in 155 ESCC patients (Table 1). The chi-square test revealed that GJB6 expression was significantly associated with ESCC grade (P < 0.001), TNM stage (P = 0.002), LN classification (P = 0.001), and overall survival (OS) event (P = 0.008). Then we evaluated the impact of GJB6 expression on survival time in the same cohort. The Kaplan-Meier curve presented that a high level of GJB6 was strongly linked to better overall survival in ESCC (Fig. 1H). Similarly, we investigated the predictive value of GJB6 by stratified Kaplan-Meier survival analysis. Low GJB6 expression was relevant to shorter overall survival with males, age ≤ 60, no smoking history, TNM stage Ⅲ-Ⅳ, and patients with lymph node metastasis (Figure S2A, B, E, H, J, P < 0.05). It was notable that patients with lower tumor location, no family history, and T3 could survive significantly longer when accompanied by high GJB6 expression (Figure S2C, D, I, P < 0.01). These data confirmed that ESCC patients with increased expression of GJB6 have a better prognosis. Finally, univariate and multivariate Cox regression analyses of GJB6 expression and clinical features were presented to show that high GJB6 expression was an independent protective factor and indicated a better prognosis in 155 patients with ESCC (Fig. 1I and Figure S1H). These findings above suggest that the reduced expression of GJB6 in ESCC may be partly due to copy-number loss and low expression of GJB6 may indicate unsatisfied outcome of ESCC.Table 1Correlation between GJB6 expression and clinical features in ESCCClinical featuresGJB6lowGJB6highP valueAge≥6040 (40.4%)59 (59.6%)0.114<6030 (53.6%)26 (46.4%)GenderMale43 (41.7%)60 (58.3%)0.229Female27 (51.9%)25 (48.1%)LocationLower18 (46.2%)21 (53.8%)0.902Middle49 (45.4%)59 (54.6%)Upper3 (37.5%)5 (62.5%)Smoking statusnever40 (46%)47 (54%)0.817yes30 (44.1%)38 (55.9%)Drinking statusnever46 (43.4%)60 (56.6%)0.516yes24 (49%)25 (51%)Tumor grade12 (9.1%)20 (90.9%)<0.001235 (41.7%)49 (58.3%)333 (67.3%)16 (32.7%)TNM stageⅠ1 (16.7%)5 (83.3%)0.002Ⅱ31 (40.3%)46 (59.7%)Ⅲ28 (45.2%)34 (54.8)Ⅳ10 (100%)0 (0%)T classificationT1-T221 (51.2%)20 (48.8%)0.363T3-T449 (43%)65 (57%)Lymph node metastasisNo32 (39%)50 (61%)0.104Yes38 (52%)35 (48%)LN classificationN032 (39%)50 (61%)0.001N115 (36.6%)26 (63.4%)N213 (59.1%)9 (40.9%)N310 (100%)0 (0%)VitalstatusDecreased47 (54.7%)39 (45.3%)0.008Living23 (33.3%)46 (66.7%)P-values of these bold values were < 0.05, indicating statistical significance. Table 1Correlation between GJB6 expression and clinical features in ESCCClinical featuresGJB6lowGJB6highP valueAge≥6040 (40.4%)59 (59.6%)0.114<6030 (53.6%)26 (46.4%)GenderMale43 (41.7%)60 (58.3%)0.229Female27 (51.9%)25 (48.1%)LocationLower18 (46.2%)21 (53.8%)0.902Middle49 (45.4%)59 (54.6%)Upper3 (37.5%)5 (62.5%)Smoking statusnever40 (46%)47 (54%)0.817yes30 (44.1%)38 (55.9%)Drinking statusnever46 (43.4%)60 (56.6%)0.516yes24 (49%)25 (51%)Tumor grade12 (9.1%)20 (90.9%)<0.001235 (41.7%)49 (58.3%)333 (67.3%)16 (32.7%)TNM stageⅠ1 (16.7%)5 (83.3%)0.002Ⅱ31 (40.3%)46 (59.7%)Ⅲ28 (45.2%)34 (54.8)Ⅳ10 (100%)0 (0%)T classificationT1-T221 (51.2%)20 (48.8%)0.363T3-T449 (43%)65 (57%)Lymph node metastasisNo32 (39%)50 (61%)0.104Yes38 (52%)35 (48%)LN classificationN032 (39%)50 (61%)0.001N115 (36.6%)26 (63.4%)N213 (59.1%)9 (40.9%)N310 (100%)0 (0%)VitalstatusDecreased47 (54.7%)39 (45.3%)0.008Living23 (33.3%)46 (66.7%)P-values of these bold values were < 0.05, indicating statistical significance. Correlation between GJB6 expression and clinical features in ESCC P-values of these bold values were < 0.05, indicating statistical significance. GJB6 represses malignant phenotypes of ESCC To explore whether GJB6 might exert roles in ESCC, we carried out experiments to assess the key cellular functions, including cell proliferation, migration, and invasion. Based on GJB6 expression patterns in ESCC cells (Figure S3A), we proceeded to modulate its expression. KYSE30 and TE-1 cells transfected with siRNA showed a significant reduction in GJB6 expression (Figure S3B–D), while KYSE150 cells ectopically transfected with overexpression plasmids successfully elevated expression of GJB6 at both transcriptional and translational levels (Figure S3E, F). CCK-8 and clonogenic assays could serve to quantify cell proliferation. CCK-8 assays showed that cell proliferative activity was markedly different between treatment and control groups. GJB6 knockdown by siRNA dramatically accelerated cell growth rate compared with the control group of KYSE30 and TE-1 cells respectively (Fig. 2A, B), while GJB6 over-expression remarkably suppressed the proliferation of KYSE150 cells (Fig. 2C), which was confirmed by colony-forming assays (Fig. 2D–F). Transwell and wound healing assays were employed to probe the metastatic potential of tumor cells. As presented in Fig. 2G–I, GJB6 knockdown using siRNA noticeably increased the numbers of migrating and invading cells of KYSE30 and TE-1, while GJB6 upregulation led to an apparent decline of the numbers in KYSE150. Similar results were observed in the wound healing assays (Fig. 2J–L). Employing pre-validated siRNA target sequences, lentiviral shRNA constructs were generated and proved potent functional efficacy in cells (Figure S3G–J). To verify the anti-oncogenic role of GJB6 in vivo, we carried out nude mouse xenograft experiments with subcutaneous tumor implantation. The results from the lentiviral knockdown model showed that GJB6 knockdown considerably promoted tumor growth, as indicated by tumor morphologies, weights and volumes (Fig. 2M–O). Immunohistochemical staining confirmed that suppressing GJB6 promoted tumor cell proliferation, as indicated by elevated Ki-67 expression (Fig. 2P). These data demonstrate that GJB6 serves as a tumor-suppressor gene in ESCC.Fig. 2GJB6 decreases the capability of proliferation, migration and invasion in ESCC cells.OD450nm values of ESCC cells untreated, transfected with siRNA, or expressing GJB6-encoding constructs at the specified time points in KYSE30 (A), TE-1 (B), and KYSE150 (C). Colony-forming assay measuring GJB6-silenced and GJB6-upregulated ESCC cells in KYSE30 (D), TE-1 (E), and KYSE150 (F). G–I Control, GJB6 siRNA-knockdown, and GJB6 vector-transfected cells were subjected to Boyden chamber assays. Representative images for KYSE30, TE-1, and KYSE150 cells and quantification from three random microscopic fields are shown in (G, H, and I), respectively. Representative photographs and statistical graphs of the wound healing scratch assay from GJB6-silenced or overexpressing KYSE30 (J), TE-1 (K) and KYSE150 cells (L). M–P Subcutaneous xenografted nude mouse model. Macroscopic appearance (M), tumor weight (N), volume (O), and representative H&E and immunohistochemistry staining (P) of xenografts from GJB6-knockdown KYSE30 cells (n = 6 per group). Scale bar, 100 μm. Error bars, SDs; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. GJB6 decreases the capability of proliferation, migration and invasion in ESCC cells. OD450nm values of ESCC cells untreated, transfected with siRNA, or expressing GJB6-encoding constructs at the specified time points in KYSE30 (A), TE-1 (B), and KYSE150 (C). Colony-forming assay measuring GJB6-silenced and GJB6-upregulated ESCC cells in KYSE30 (D), TE-1 (E), and KYSE150 (F). G–I Control, GJB6 siRNA-knockdown, and GJB6 vector-transfected cells were subjected to Boyden chamber assays. Representative images for KYSE30, TE-1, and KYSE150 cells and quantification from three random microscopic fields are shown in (G, H, and I), respectively. Representative photographs and statistical graphs of the wound healing scratch assay from GJB6-silenced or overexpressing KYSE30 (J), TE-1 (K) and KYSE150 cells (L). M–P Subcutaneous xenografted nude mouse model. Macroscopic appearance (M), tumor weight (N), volume (O), and representative H&E and immunohistochemistry staining (P) of xenografts from GJB6-knockdown KYSE30 cells (n = 6 per group). Scale bar, 100 μm. Error bars, SDs; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. GJB6 is a member of the gap junction protein family, also known as the connexin family. Both GJIC and cell surface hemichannel functions were characterized as canonical roles for connexins historically. Besides, a diverse range of novel biological functions has emerged and been documented over time18. Accordingly, we utilized the scrape-loading dye transfer (SLDT) technique to detect the canonical GJIC function of GJB6. Compared with control cells, transfection of siGJB6 into KYSE30 cells increased the number of GJ-coupled dye-spreading cells, as indicated by green fluorescence (Figure S4A). On the contrary, the communication distance was shorter in the GJB6 overexpression group compared with the control group in KYSE150 cells (Figure S4B). The data indicated that GJB6 participates in gap junction communication and functions as a repressor of intercellular signal transmission. Nevertheless, malignant tumor phenotypes are inseparable from signal transduction and substance exchange, which requires a close relationship with gap junction. Homeostatic balance is regulated through essential cell-cell communication pathways. To further verify whether GJB6 inhibits the malignant phenotypes of ESCC via a GJIC-dependent mechanism, we used the gap junction inhibitor 18α-glycyrrhetinic acid (18α-GA) and the activator all-trans-retinoic acid (ATRA). The growth curve assay indicated that cell proliferation was partly restored by 18α-GA (10 μM, 1 h) in the GJB6 knockdown group of KYSE30 cells (Figure S4C), and similar effects were observed on cell migration and invasion (Figure S4D, E). Intriguingly, we also found that ATRA (1 μM, 24 h), the gap junction enhancer, increased the proliferation and metastasis abilities of KYSE150 cells in the GJB6 overexpression group compared with the DMSO control group (Figure S4F–H). Collectively, these findings reveal that GJB6 exerts tumor-suppressive effects through GJIC-dependent cell coupling. AKT inhibitors reverse the effects of GJB6 ablation To figure out the mechanisms underlying GJB6-mediated ESCC pathogenesis, we performed RNA-seq to delineate the downstream regulatory networks by comparing the gene expression profiles between GJB6-knockdown and control cells. KEGG pathway enrichment analysis (http://www.sangerbox.com/) revealed several significantly enriched pathways related to AKT signaling, including FOXO translocation and PI3K-AKT signaling pathways (Fig. 3A). Gene Set Enrichment Analysis (GSEA) (https://www.xiantaozi.com/) based on TCGA data further corroborated this finding (Figure S5A). Subsequently, we performed Western blot analysis to experimentally validate these bioinformatics predictions. Downregulation of GJB6 by shRNA promoted AKT pathway activation, as indicated by diminished PTEN expression and enhanced p-AKT levels. Conversely, overexpression of GJB6 inhibited AKT pathway activity (Fig. 3B).Fig. 3Capivasertib-mediated AKT inhibition reverses GJB6 knockdown-induced effects in vitro.A KEGG pathway enrichment analysis indicating the activation of AKT-related signaling pathways in GJB6-knockdown cells. B Western blot analysis results showing the expression levels of marker proteins in the PI3K/AKT pathway in ESCC cells. C The chemical structure of CAPI. Western blot measuring the inhibitory effect of CAPI on GJB6-knockdown KYSE30 (D) and TE-1 (E) cells. Half-maximal inhibitory concentration (IC50) curves of CAPI in GJB6-knockdown and control KYSE30 (F) and TE-1 (G) cells. IC50 values were calculated by nonlinear regression. CCK-8 proliferation analysis using the AKT inhibitor CAPI (3 μM, 24 h) in GJB6-knockdown KYSE30 (H) and TE-1 (I) cells. Boyden chamber assay conducted to evaluate the impact of CAPI (3 μM, 24 h) on the migratory and invasive capacity of GJB6-knockdown KYSE30 (J) and TE-1 (K) cells. Error bars, SDs; ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Capivasertib-mediated AKT inhibition reverses GJB6 knockdown-induced effects in vitro. A KEGG pathway enrichment analysis indicating the activation of AKT-related signaling pathways in GJB6-knockdown cells. B Western blot analysis results showing the expression levels of marker proteins in the PI3K/AKT pathway in ESCC cells. C The chemical structure of CAPI. Western blot measuring the inhibitory effect of CAPI on GJB6-knockdown KYSE30 (D) and TE-1 (E) cells. Half-maximal inhibitory concentration (IC50) curves of CAPI in GJB6-knockdown and control KYSE30 (F) and TE-1 (G) cells. IC50 values were calculated by nonlinear regression. CCK-8 proliferation analysis using the AKT inhibitor CAPI (3 μM, 24 h) in GJB6-knockdown KYSE30 (H) and TE-1 (I) cells. Boyden chamber assay conducted to evaluate the impact of CAPI (3 μM, 24 h) on the migratory and invasive capacity of GJB6-knockdown KYSE30 (J) and TE-1 (K) cells. Error bars, SDs; ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Akt signaling is often aberrantly activated in many cancers, thus rendering Akt a promising target for cancer treatment19. As an FDA-approved first-in-class AKT inhibitor, capivasertib (AZD5363, CAPI) (Fig. 3C) has been clinically implemented for cancer treatment. Western blot analysis showed that CAPI specifically downregulated p-AKT expression in KYSE30 and TE-1 shGJB6 cells, with no effect on GJB6 or total AKT protein levels (Fig. 3D,E). We then examined drug sensitivity and found that shRNA-mediated knockdown of GJB6 sensitized both KYSE30 and TE-1 cells to CAPI relative to controls (Fig. 3F, G). The results showed administration of CAPI (3 μM, 24 h) to GJB6-silenced cells resulted in marked suppression of proliferative, migratory, and invasive capacities in KYSE30 and TE-1 cells (Fig. 3H–K). Furthermore, another AKT inhibitor, MK-2206 (1 μM, 24 h), yielded similar outcomes (Figure S6A–H). Statistical analysis revealed that GJB6-depleted cells showed significantly greater sensitivity to CAPI or MK-2206 treatment than the control cells (Figure S6I–K). Next, we evaluated the efficacy of AKT inhibitors in vivo. Following subcutaneous inoculation with shGJB6 or control KYSE30 cells, nude mice were randomized to receive intraperitoneal treatment with vehicle, MK-2206 (50 mg/kg), or CAPI (50 mg/kg). As expected, treatment with either MK-2206 or CAPI demonstrated more potent anticancer activity compared to vehicle treatment (Fig. 4A). Accordingly, mice treated with MK-2206 or CAPI had reduced tumor weights relative to the vehicle-treated controls (Fig. 4B, C). Moreover, we monitored the mouse body weights throughout the treatment period and found no significant differences among the various groups (Fig. 4D). AKT inhibitor administration did not induce any treatment-associated alterations in gross anatomical features or histological architecture in pulmonary, splenic and hepatic tissues (Figure S7A, B). Furthermore, histological analysis through hematoxylin-eosin (H&E) staining and IHC assessment demonstrated a marked reduction in Ki-67 and p-AKT expression following MK-2206 or CAPI administration (Fig. 4E). To determine the impact on metastasis, we employed a tail vein metastasis model to investigate GJB6 knockdown and AKT inhibitor treatment (Fig. 4F). GJB6 knockdown elevated metastatic rate of ESCC cells, which were attenuated by MK-2206 or CAPI (Fig. 4G). Body weight was comparable among all groups throughout the treatment course (Fig. 4H). H&E staining of lung tissues confirmed that shGJB6 cells led to metastatic lesions, whereas this phenotype was reversed by AKT inhibitors (Fig. 4I). Furthermore, no treatment-related changes were observed in spleen or liver morphology or histology (Figure S7C, D). Together, our results suggest that the AKT-inhibiting compounds safely and effectively rescue the effects caused by reduced GJB6 expression within a defined dose range.Fig. 4Targeting AKT halts ESCC progression driven by GJB6 knockdown in vivo.A Timeline of the experiment and macroscopic appearance of xenografts of GJB6-knockdown KYSE30 cells treated with MK-2206 (50 mg/kg) or CAPI (50 mg/kg) (n = 5 per group). B Tumor volumes measured at indicated time points in xenografts derived from GJB6-knockdown KYSE30 cells following treatment with MK-2206 or CAPI. C Tumor weights of GJB6-knockdown KYSE30 xenografts treated with MK-2206 or CAPI. D Body weights of the mice measured every three days. E Histopathological analysis of the tumors and representative images of IHC staining of Ki-67, GJB6, and p-AKT. Scale bar, 100 μm. F Schematic of the experimental design for the ESCC metastasis mouse model (n = 5 per group). G Metastasis rates in mice receiving tail vein tumor cell injection and AKT inhibitor treatment (MK-2206/CAPI, 50 mg/kg). H Effect of AKT inhibitors on body weight in KYSE30 cell-based lung metastasis model mice. I Representative H&E staining of lung tissue following AKT inhibitor administration. Scale bars: 200 µm (upper panel) and 50 µm (lower panel). Targeting AKT halts ESCC progression driven by GJB6 knockdown in vivo. A Timeline of the experiment and macroscopic appearance of xenografts of GJB6-knockdown KYSE30 cells treated with MK-2206 (50 mg/kg) or CAPI (50 mg/kg) (n = 5 per group). B Tumor volumes measured at indicated time points in xenografts derived from GJB6-knockdown KYSE30 cells following treatment with MK-2206 or CAPI. C Tumor weights of GJB6-knockdown KYSE30 xenografts treated with MK-2206 or CAPI. D Body weights of the mice measured every three days. E Histopathological analysis of the tumors and representative images of IHC staining of Ki-67, GJB6, and p-AKT. Scale bar, 100 μm. F Schematic of the experimental design for the ESCC metastasis mouse model (n = 5 per group). G Metastasis rates in mice receiving tail vein tumor cell injection and AKT inhibitor treatment (MK-2206/CAPI, 50 mg/kg). H Effect of AKT inhibitors on body weight in KYSE30 cell-based lung metastasis model mice. I Representative H&E staining of lung tissue following AKT inhibitor administration. Scale bars: 200 µm (upper panel) and 50 µm (lower panel). GRHL3 is one of the downstream targets of GJB6 Subsequently, we sought to investigate intrinsic mechanisms underlying the role of GJB6 in AKT signaling pathway. We re-analyzed RNA-seq from GJB6 knockdown ESCC cells to identify the downstream target genes (Fig. 5A). In addition, public database search based on TCGA-ESCA was performed to discover genes that have significant relevance to GJB6. The results from the LinkedOmics database indicated that 20,104 genes were correlated with GJB6 expression in 96 ESCC samples (Spearman correlation analysis, P < 0.05). Gene heatmaps were generated, displaying the top 50 positively and negatively correlated genes (Figure S5B). In order to narrow down the correlated gene lists, we utilized another database: cBioPortal. We screened out 126 genes under the criteria of P < 0.05 and the absolute value of the Spearman correlation coefficient > 0.75 simultaneously. We then selected 32 GJB6 co-expressed genes overlapping among the two databases mentioned above and constructed interaction networks by STRING database and Cytoscape software (Fig. 5B, C). Intriguingly, Grainyhead-like 3 (GRHL3) was a prominent overlapping gene, as it was both downregulated upon GJB6 knockdown in RNA-seq and positively correlated with GJB6 expression in the TCGA-ESCA cohort. Meanwhile, GRHL3 is an upstream regulatory factor of the AKT pathway20. Given the above, GRHL3 was selected as a candidate target gene of GJB6 in the follow-up investigation. Next, we analyzed transcriptomic data from 155 ESCC samples and the TCGA-ESCC cohort, as well as proteomic data from 124 ESCC samples (IPX0002501000). In these cohorts, we detected a statistically significant positive correlation between GJB6 and GRHL3 at both the mRNA and protein levels (P < 0.0001, Fig. 5D–F). Besides, GJB6 was also correlated with GRHL3 in LUSC, HNSC and CESC by mining UCSC Xena website (P < 0.0001, Figure S8A–C). The results suggest that the relevance between GRHL3 and GJB6 is universal in SCC, supporting the hypothesis that GRHL3 is positively regulated by GJB6. Next, we proceeded to the experimental verification of expression. Consistent with the bioinformatics prediction, GRHL3 mRNA and protein levels were decreased upon GJB6 knockdown via shRNA in KYSE30 and TE-1 cells, while they were increased in GJB6-overexpressing KYSE150 cells (P < 0.0001, Fig. 5G–J). Downregulation of GJB6 activated the PI3K/AKT pathway in KYSE30 and TE-1 cells, as evidenced by decreased PTEN and increased p-AKT, whereas overexpression of GJB6 inhibited the pathway in KYSE150 cells (Fig. 5J). Similar results were observed in the IHC staining of xenotransplanted tumors (Fig. 5K). Additionally, Western blot analysis revealed that GRHL3 overexpression counteracted GJB6 knockdown-induced PI3K/AKT activation, whereas GRHL3 suppression reactivated this pathway in GJB6-overexpressing cells (Fig. 5L). In summary, these data imply that GRHL3 might be a core hub on the GJB6/PI3K/AKT signaling axis.Fig. 5GRHL3 functions in the GJB6/PI3K/AKT signaling axis.A Volcano plot displaying differentially expressed genes after GJB6 knockdown in KYSE30 cells. B Venn diagram of genes overlapping in the cBioPortal database and the LinkedOmics database. C Protein-protein interaction (PPI) network for GJB6 visualized by Cytoscape software. Correlation analyses of GJB6 and GRHL3 expression in the transcriptome (D), TCGA-ESCC (E), and IPX0002501000 (F) cohorts. The mRNA expression levels of GJB6 and GRHL3 examined in GJB6-downregulated KYSE30 (G) and TE-1 (H) cells by qRT-PCR. I qRT-PCR results showing GJB6 and GRHL3 expression in GJB6-overexpressing KYSE150 cells. J Western blot analysis showing expression of the PI3K/AKT pathway hallmark target genes and the candidate downstream gene GRHL3 in ESCC cells. K Representative H&E and IHC staining of Ki-67, GJB6, GRHL3, and p-AKT in KYSE30 xenografts. Scale bar, 100 μm. L Western blot analysis of AKT signaling induced by GJB6 modulation following GRHL3 restoration. Error bars, SDs; ***P < 0.001; ****P < 0.0001. GRHL3 functions in the GJB6/PI3K/AKT signaling axis. A Volcano plot displaying differentially expressed genes after GJB6 knockdown in KYSE30 cells. B Venn diagram of genes overlapping in the cBioPortal database and the LinkedOmics database. C Protein-protein interaction (PPI) network for GJB6 visualized by Cytoscape software. Correlation analyses of GJB6 and GRHL3 expression in the transcriptome (D), TCGA-ESCC (E), and IPX0002501000 (F) cohorts. The mRNA expression levels of GJB6 and GRHL3 examined in GJB6-downregulated KYSE30 (G) and TE-1 (H) cells by qRT-PCR. I qRT-PCR results showing GJB6 and GRHL3 expression in GJB6-overexpressing KYSE150 cells. J Western blot analysis showing expression of the PI3K/AKT pathway hallmark target genes and the candidate downstream gene GRHL3 in ESCC cells. K Representative H&E and IHC staining of Ki-67, GJB6, GRHL3, and p-AKT in KYSE30 xenografts. Scale bar, 100 μm. L Western blot analysis of AKT signaling induced by GJB6 modulation following GRHL3 restoration. Error bars, SDs; ***P < 0.001; ****P < 0.0001. GJB6 deficiency promotes tumor progression by regulating GRHL3 As a transcription factor, GRHL3 has been documented as essential for epidermal morphogenesis and skin barrier integrity, as well as in various cancers, including lung squamous cell cancer (LUSC) and cutaneous squamous cell carcinoma (SCC)21–24. However, the detailed mechanisms remain obscure. Thus far, research of GRHL3 focused on ESCC remains limited. To further elucidate the alterations of malignant progression caused by GRHL3 in ESCC, gain-of-function and loss-of-function experiments were performed. First, cells were transiently transfected with siGRHL3 and pENTER-GRHL3 plasmids to alter the expression of GRHL3 (Figure S9A-D). The results of functional experiments suggested that GRHL3 knockdown via siRNA promoted aberrant cell proliferation, migration, and invasion in KYSE30 and TE-1 cells compared to control cells (Fig. 6A, B, D, E). Contrarily, GRHL3 overexpression suppressed the malignant properties of KYSE150 cells (Fig. 6C, F). Furthermore, survival analysis suggested that patients with higher tumor GRHL3 expression survived significantly longer than patients with lower expression utilizing the GSE53625 dataset (P = 0.0187, Fig. 6G and Table 2). Survival analysis of ESCC patients stratified by T stage confirmed that GRHL3high patients had better prognoses (Fig. 6H). To identify independent prognostic factors, all relevant variables were subjected to both univariate and multivariate Cox regression analyses. It was demonstrated that TNM stage and GRHL3 expression were independent predictors of overall survival (Figure S10A, B and Table 2). The above results demonstrated GRHL3’s tumor-suppressor function in ESCC.Fig. 6GRHL3 is a downstream target of GJB6 and has biological and clinical significance in ESCC.Growth curves of KYSE30 (A) and TE-1 (B) cells after reduction of GRHL3. C Growth curve of KYSE150 cells after GRHL3 upregulation. Boyden chamber assay showing the effect of GRHL3 knockdown on the migration and invasion of KYSE30 (D) and TE-1 (E) cells. F Boyden chamber assay showing the effect of GRHL3 overexpression on the migration and invasion of KYSE150 cells. G Kaplan-Meier survival curves for 179 patients with ESCC stratified by GRHL3 expression. H Stratified survival analysis according to T stage. Western blot assays showing GJB6 and GRHL3 expression following GJB6 inhibition with or without GRHL3 overexpression (I) or following GJB6 upregulation with or without GRHL3 inhibition (J). Growth curves measured by CCK-8 assay of ESCC cells following GJB6 inhibition with or without GRHL3 overexpression (K) or following GJB6 upregulation with or without GRHL3 inhibition (L). Boyden chamber assays of ESCC cells following GJB6 inhibition with or without GRHL3 overexpression (M) or following GJB6 upregulation with or without GRHL3 inhibition (N). Error bars, SDs; ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.Table 2Correlation between GRHL3 expression and clinical features in ESCCClinical featuresGRHL3lowGRHL3highP valueAge≥6047 (53.4%)41 (46.6%)0.424<6054 (59.3%)37 (40.7%)GenderMale83 (56.8%)63 (43.2%)0.81Female18 (54.5%)15 (45.5%)LocationLower37 (59.7%)25 (40.3%)0.121Middle57 (58.8%)40 (41.2%)Upper7 (35%)13 (65%)Smoking statusnever36 (55.4%)29 (44.6%)0.832yes65 (57%)49 (43%)Drinking statusnever40 (54.8%)33 (45.2%)0.715yes61 (57.5%)45 (42.5%)Tumor grade1 + 284 (57.1%)63 (42.9%)0.6783 + 417 (53.1%)15 (46.9%)T classificationT1-T221 (53.8%)18 (46.2%)0.713T3-T480 (57.1%)60 (42.9%)Lymph node metastasisNo40 (48.2%)43 (51.8%)0.039Yes61 (63.5%)35 (36.5%)Vital statusDecreased66 (62.3%)40 (37.7%)0.058Living35 (47.9%)38 (52.1%)P-values of these bold values were < 0.05, indicating statistical significance. GRHL3 is a downstream target of GJB6 and has biological and clinical significance in ESCC. Growth curves of KYSE30 (A) and TE-1 (B) cells after reduction of GRHL3. C Growth curve of KYSE150 cells after GRHL3 upregulation. Boyden chamber assay showing the effect of GRHL3 knockdown on the migration and invasion of KYSE30 (D) and TE-1 (E) cells. F Boyden chamber assay showing the effect of GRHL3 overexpression on the migration and invasion of KYSE150 cells. G Kaplan-Meier survival curves for 179 patients with ESCC stratified by GRHL3 expression. H Stratified survival analysis according to T stage. Western blot assays showing GJB6 and GRHL3 expression following GJB6 inhibition with or without GRHL3 overexpression (I) or following GJB6 upregulation with or without GRHL3 inhibition (J). Growth curves measured by CCK-8 assay of ESCC cells following GJB6 inhibition with or without GRHL3 overexpression (K) or following GJB6 upregulation with or without GRHL3 inhibition (L). Boyden chamber assays of ESCC cells following GJB6 inhibition with or without GRHL3 overexpression (M) or following GJB6 upregulation with or without GRHL3 inhibition (N). Error bars, SDs; ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Table 2Correlation between GRHL3 expression and clinical features in ESCCClinical featuresGRHL3lowGRHL3highP valueAge≥6047 (53.4%)41 (46.6%)0.424<6054 (59.3%)37 (40.7%)GenderMale83 (56.8%)63 (43.2%)0.81Female18 (54.5%)15 (45.5%)LocationLower37 (59.7%)25 (40.3%)0.121Middle57 (58.8%)40 (41.2%)Upper7 (35%)13 (65%)Smoking statusnever36 (55.4%)29 (44.6%)0.832yes65 (57%)49 (43%)Drinking statusnever40 (54.8%)33 (45.2%)0.715yes61 (57.5%)45 (42.5%)Tumor grade1 + 284 (57.1%)63 (42.9%)0.6783 + 417 (53.1%)15 (46.9%)T classificationT1-T221 (53.8%)18 (46.2%)0.713T3-T480 (57.1%)60 (42.9%)Lymph node metastasisNo40 (48.2%)43 (51.8%)0.039Yes61 (63.5%)35 (36.5%)Vital statusDecreased66 (62.3%)40 (37.7%)0.058Living35 (47.9%)38 (52.1%)P-values of these bold values were < 0.05, indicating statistical significance. Correlation between GRHL3 expression and clinical features in ESCC P-values of these bold values were < 0.05, indicating statistical significance. We next examined whether GRHL3 is essential for the biological function of GJB6 in ESCC. To this end, GRHL3 was further overexpressed in GJB6 knockdown cells or interfered with in GJB6-upregulated cells (Fig. 6I, J). Further functional experiments illustrated that GRHL3 overexpression successfully attenuated cellular growth and dissemination of GJB6-depleted KYSE30 cells, whereas GRHL3 inhibition abrogated the effects of GJB6 overexpression on these cellular functions (Fig. 6K–N). Collectively, these findings highlight the tumor-suppressive role of GRHL3 and its potential as a prognostic marker in ESCC. Furthermore, we demonstrate that GRHL3 acts as a key downstream mediator of GJB6’s tumor-suppressor function. GJB6 increases intracellular Ca2+ flux to enhance GRHL3 expression We next sought to explore the factors responsible for GJB6-mediated regulation of GRHL3. Considering that the connexins constitute specialized intercellular channels to permit rapid intercellular transfer of small molecules, such as second messengers IP3 and Ca2+ (the maximum molecular weight is 1 kDa)25, we speculated that signaling molecules may trigger the expression of genes downstream. GO and KEGG analyses of the differentially expressed genes were performed via TCGA-ESCC cohort, which implied that calcium signaling may be involved in molecular basis of GJB6 regulation (Figure S5C). Since normal connexin operation requires Ca²⁺-dependent fine-tuning26, a Fura-2-AM probe was adopted to monitor calcium mobilization. The results suggested that GJB6 knockdown using shRNA decreased the oscillation of intracellular calcium ions in KYSE30 and TE-1 cells (Fig. 7A, B). To elucidate the mechanism of action of GJB6, we cloned the GRHL3 promoter region into a luciferase reporter, spanning from -2000 to +100 bp relative to the transcriptional start site. As presented in Fig. 7C, D, GJB6 knockdown with shRNA markedly suppressed the luciferase signal of GRHL3 promoter reporter in KYSE30 and TE-1 cells. Interestingly, after the addition of CaCl2 (1.8 mM, 48 h) to augment calcium signaling, transcriptional activity of GRHL3 was significantly elevated in the GJB6 knockdown groups, but not in the control group. Similar experiments were conducted in both KYSE150 and 293 T cells in the presence and absence of BAPTA-AM, a cytosolic Ca2+ chelator. Dual-luciferase assays revealed that GJB6 overexpression enhanced transcriptional activity of GRHL3 promoter. Conversely, this enhanced signal was attenuated by BAPTA-AM (10 μM, 24 h) (Fig. 7E). Furthermore, signal recovery was comprehensively assessed not only at the transcriptional level but also through mRNA and protein levels. It was indicated that intensified calcium signaling did not alter the expression of GJB6, whereas it markedly induced GRHL3 expression at both mRNA and protein levels (Fig. 7F–K). These findings collectively highlight that GJB6 regulates GRHL3 expression through a calcium-dependent mechanism.Fig. 7GJB6 modulates the expression of GRHL3 in a Ca2+-dependent manner.Measurement of Ca2+ levels in KYSE30 (A) and TE-1 (B) cells with GJB6 expression manipulation. Relative luciferase activities driven by the GRHL3 promoter in GJB6-depleted KYSE30 (C) and TE-1 (D) cells with or without CaCl2 treatment (1.8 mM, 48 h). E Relative luciferase activities driven by the GRHL3 promoter in GJB6-overexpressing KYSE150 cells in the presence or absence of BAPTA-AM (10 μM, 24 h). qRT-PCR analysis showing GJB6 and GRHL3 mRNA levels in KYSE30 (F) and TE-1 (G) cells with or without CaCl2 induction (1.8 mM, 48 h). H qRT-PCR analysis showing GJB6 and GRHL3 mRNA levels in KYSE150 cells with or without the addition of BAPTA-AM (10 μM, 24 h). Western blot analysis showing GJB6 and GRHL3 protein levels in KYSE30 (I) and TE-1 (J) cells with or without CaCl2 supplementation (1.8 mM, 48 h). K Western blot analysis showing GJB6 and GRHL3 protein levels in KYSE150 cells with or without the addition of BAPTA-AM (10 μM, 24 h). Error bars, SDs; ns not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. GJB6 modulates the expression of GRHL3 in a Ca2+-dependent manner. Measurement of Ca2+ levels in KYSE30 (A) and TE-1 (B) cells with GJB6 expression manipulation. Relative luciferase activities driven by the GRHL3 promoter in GJB6-depleted KYSE30 (C) and TE-1 (D) cells with or without CaCl2 treatment (1.8 mM, 48 h). E Relative luciferase activities driven by the GRHL3 promoter in GJB6-overexpressing KYSE150 cells in the presence or absence of BAPTA-AM (10 μM, 24 h). qRT-PCR analysis showing GJB6 and GRHL3 mRNA levels in KYSE30 (F) and TE-1 (G) cells with or without CaCl2 induction (1.8 mM, 48 h). H qRT-PCR analysis showing GJB6 and GRHL3 mRNA levels in KYSE150 cells with or without the addition of BAPTA-AM (10 μM, 24 h). Western blot analysis showing GJB6 and GRHL3 protein levels in KYSE30 (I) and TE-1 (J) cells with or without CaCl2 supplementation (1.8 mM, 48 h). K Western blot analysis showing GJB6 and GRHL3 protein levels in KYSE150 cells with or without the addition of BAPTA-AM (10 μM, 24 h). Error bars, SDs; ns not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. GRHL3 transcriptionally upregulates GJB6 expression to form a positive feedback loop Prior experimental findings revealed a significant correlation between GJB6 and GRHL3. It’s noteworthy that GJB6 was identified as one of the overlapping genes associated with GRHL3 based on the integration of microarray data from siGRHL3 (GSE37049) and RNA-seq analysis of GRHL3 (GSE122175)27,28 (Fig. 8A). Based on the aforementioned evidence, we reasoned that GJB6 might be a downstream target of GRHL3 in turn. To validate this hypothesis, we assessed the impact of GRHL3 expression on GJB6. GRHL3 siRNA decreased GJB6 expression at both the mRNA and protein levels, whereas GRHL3 overexpression upregulated its expression (Fig. 8B-E). GRHL3 belongs to the GRHL family, which is conserved and consists of three genes: GRHL1, GRHL2, and GRHL329. By examining the motifs obtained from the JASPAR database (https://jaspar.elixir.no/) and previous studies30, we found that the binding motif sequences of these three molecules exhibited a high degree of similarity (Fig. 8F). We further analyzed the DNA-binding domains of GRHL1/2/3 in human and found that they were completely identical (Fig. 8G). As presented in Fig. 8H, the core GRHL3 binding DNA sequence on the promoter of GJB6 exhibited remarkable evolutionary conservation across diverse vertebrates. To decipher the molecular basis of GRHL3-dependent transcriptional regulation, we identified and subcloned three functionally distinct promoter regions, designated as Segment A (-1108 bp ~ +76 bp), B (-944 bp ~ +76 bp), and C (-590 bp ~ +76 bp) for further characterization. In addition, three putative binding sites were predicted through bioinformatic analysis within these genomic regions (Fig. 8I). Dual-luciferase reporter assays showed that GRHL3 overexpression effectively activated the luciferase signal driven by both GJB6 Segment A and Segment B but not by Segment C in KYSE150 and HEK293T cells, whereas GRHL3 knockdown by siRNA significantly suppressed the dual-luciferase reporter activity (Fig. 8J–L). Therefore, the genomic segment spanning from -1108 to -590 bp might be important for GRHL3-mediated GJB6 activation. To detect whether GRHL3 targets GJB6 directly, we carried out chromatin immunoprecipitation (ChIP) assay in KYSE30 cells. ChIP DNA electrophoresis and qPCR quantification demonstrated that GRHL3 bound to the GJB6 promoter at sites 1–3, identifying GJB6 as a direct target of GRHL3 (Fig. 8M, N). Taken together, these data demonstrate that GRHL3 serves as a transcriptional activator of GJB6 expression, thereby forming GJB6-GRHL3 positive feedback loop.Fig. 8GJB6 is a direct target of GRHL3.A Venn diagram displaying potential downstream molecules of GRHL3 based on GSE37049 and GSE122175. qRT-PCR analysis conducted to evaluate the mRNA expression levels of GRHL3 and GJB6 in GRHL3-downregulated KYSE30 (B) and TE-1 (C) cells. D The expression profiles of GRHL3 and GJB6 in GRHL3-overexpressing KYSE150 cells demonstrated by qRT-PCR. E Western blot analysis showing the protein expression levels of GRHL3 and GJB6 after altering GRHL3 expression. F Motifs enriched in members of GRHL family. G DNA-binding domains of GRHL family members. H GRHL3 binding motifs exist in the GJB6 promoter across species. I Schematic diagram of the putative binding sites in the promoter region of GJB6. Relative luciferase activities driven by the indicated segments of the GJB6 promoter in KYSE30 (J), KYSE150 (K), and HEK293T (L) cells. ChIP-PCR (M) and ChIP-qPCR (N) of sites in the GJB6 promoter with either GRHL3 antibody or IgG. Error bars, SDs; ns, not significant; ***P < 0.001; ****P < 0.0001. GJB6 is a direct target of GRHL3. A Venn diagram displaying potential downstream molecules of GRHL3 based on GSE37049 and GSE122175. qRT-PCR analysis conducted to evaluate the mRNA expression levels of GRHL3 and GJB6 in GRHL3-downregulated KYSE30 (B) and TE-1 (C) cells. D The expression profiles of GRHL3 and GJB6 in GRHL3-overexpressing KYSE150 cells demonstrated by qRT-PCR. E Western blot analysis showing the protein expression levels of GRHL3 and GJB6 after altering GRHL3 expression. F Motifs enriched in members of GRHL family. G DNA-binding domains of GRHL family members. H GRHL3 binding motifs exist in the GJB6 promoter across species. I Schematic diagram of the putative binding sites in the promoter region of GJB6. Relative luciferase activities driven by the indicated segments of the GJB6 promoter in KYSE30 (J), KYSE150 (K), and HEK293T (L) cells. ChIP-PCR (M) and ChIP-qPCR (N) of sites in the GJB6 promoter with either GRHL3 antibody or IgG. Error bars, SDs; ns, not significant; ***P < 0.001; ****P < 0.0001. Discussion CNAs play essential roles in various cancer subtypes. While some reports have uncovered the genetic basis underlying ESCC carcinogenesis3,4,31, a systematic exploration of CNA targets remains largely lacking. To address these issues, we implemented a genome-wide CRISPR/Cas9 screening and identified GJB6 as one of the critical regulators in ESCC tumorigenesis. Decreased expression of GJB6 is associated with poor prognosis and boosts the proliferation and motility of ESCC cells. More importantly, GJB6 and GRHL3 form a positive feedback loop through calcium modulation, thereby attenuating the activity of the downstream AKT signaling pathway. Therapeutically, the AKT inhibitor capivasertib may alleviate ESCC progression in patients characterized by low GJB6 expression or GJB6 deletion. Accumulating studies have deepened our insight into the broad impacts of connexins and GJIC upon a range of cellular and molecular activities, which have potential clinical applications in cancer treatment32,33. Although connexins 26, 32, and 43 have been extensively studied due to their widespread expression, GJB6 (connexin 30) has received limited attention owing to its restricted tissue distribution. A previous report has shown that GJB6 expression was low in multiple tissue types, except in brain and skin34, where its physiological functions remain to be fully elucidated. Our study revealed GJB6 as a tumor-suppressor gene with copy-number alterations and uncovered that ESCC patients with diminished GJB6 expression exhibited more aggressive clinicopathological features and poorer survival. Analysis of our previous large-scale sequencing data indicated that the GJB6 locus at 13q12.11, a region characterized by frequent deletions, was significantly associated with adverse outcomes, providing substantial support for our current results3. The clinical significance of GJB6 extends beyond ESCC, with studies reporting its potential as a prognostic marker in other tumors. For instance, Artesi et al. observed that the copy number loss frequency of GJB6 was 25.8% among the 751 glioma patients35. Additionally, GJB6 functioned as a potential marker for outcome and stage or grade in breast cancer and gastric cancer13,14. These collective findings revealed GJB6 as a novel driver gene and potential prognostic indicator in ESCC and other malignancies, offering innovative perspectives on molecular stratification for ESCC prognosis prediction. Dysregulated gap junction expression plays multifaceted roles in both tumor initiation and progression18,36. Thus, we performed gain- and loss-of-function experiments and found that GJB6 reduced the proliferation and motility of ESCC cells. The finding is partially consistent with the ability of GJB6 to suppress astrocytic migration37. Subsequently, we established a xenograft tumor model using stable GJB6-knockdown ESCC cell lines. While GJIC represents the most canonical function of connexins, emerging evidence has revealed a diverse array of novel functions. These cellular functions can be categorized into GJIC-dependent and GJIC-independent mechanisms38,39. Our results implied that GJB6 exerted its tumor-suppressive function via GJIC-mediated cell coupling in ESCC. Despite evidence supporting GJB6 as a tumor-suppressive factor, the molecular mechanisms underlying its role in ESCC pathogenesis remain unclear. Here, we revealed that transcription factor GRHL3 was a key downstream target for GJB6 in inhibiting proliferation and metastasis in ESCC. According to previous literature reports, the role of GRHL3 was controversial, with cancer-suppressive or cancer-promoting functions depending on cancer types40–42. In squamous cell cancers such as head and neck squamous cell carcinoma (HNSCC), GRHL3 functions as a tumor suppressor. It has been demonstrated that GRHL3 directly transcriptionally targets FLG and GSK3B43,44. Nevertheless, the upstream regulatory mechanisms of GRHL3 remain poorly understood, and its functional role and prognostic value in ESCC have yet to be fully elucidated. We found that elevated GJB6 increased the expression of GRHL3 at both the mRNA and protein levels. Consistent with its role in other squamous cell carcinomas, GRHL3 acted as a tumor suppressor in ESCC and predicted poor prognosis. Mechanistically, it was clarified that GJB6 regulated GRHL3 expression in a calcium-dependent mechanism. Connexin channels enable the intercellular diffusion of calcium ions and other second messenger molecules. Disrupted Ca2+ homeostasis has emerged as a newly recognized characteristic of cancer, contributing significantly to tumorigenesis and disease progression45. Hopkin et al. have proven that calcium-mediated signaling promotes keratinocyte differentiation in human epidermal cells27, resulting in elevated GRHL3 expression levels. Our study provides additional insights into the multi-level regulatory mechanisms of calcium ions on GRHL3 expression. Intriguingly, we have uncovered a reciprocal positive feedback loop between GJB6 and the tumor-suppressor GRHL3. In this circuit, GRHL3 directly transcriptionally upregulates GJB6, which in turn promotes GRHL3 expression via Ca²⁺-dependent signaling. This mutually reinforcing interaction amplifies their tumor-suppressive effects. Our findings identify GJB6 as a novel upstream regulator of GRHL3, revealing a previously unrecognized reciprocal regulatory axis. Transcriptome sequencing analysis and experimental data demonstrated that the GJB6/GRHL3 feedback loop exerted its tumor-suppressive effects by modulating the AKT signaling pathway. Previous studies have established the PI3K-AKT pathway as a downstream signaling cascade of GRHL3 in ESCC and hepatocellular carcinoma20,46,47, and our current findings strongly corroborate these reports. Notably, Darido et al. demonstrated that GRHL3 directly activates PTEN transcription20. Given PTEN’s role as a key negative regulator of the PI3K/AKT pathway, this suggests that the GJB6/GRHL3 positive feedback loop modulates AKT activity through PTEN. The AKT signaling pathway represents an evolutionarily conserved central regulatory network, and orchestrates diverse cellular processes such as survival, proliferation, migration and invasion48. Dysregulation or genetic alterations of this pathway critically contribute to human tumorigenesis49. Accordingly, considerable scientific interest has focused on the design and assessment of small-molecule inhibitors against core kinases of the AKT pathway. Remarkably, capivasertib, a selective and potent inhibitor, has been FDA-approved and has emerged as the most promising therapeutic agent targeting all AKT isoforms in advanced solid tumors, particularly in breast cancer50–52. In the CAPItello-291 trial, capivasertib plus fulvestrant doubled progression-free survival (PFS) versus fulvestrant alone in HR + , HER2- breast cancer patients with resistance to aromatase and CDK4/CDK6 inhibitors53. Here, we revealed that capivasertib treatment significantly suppressed cellular proliferation, migration and invasion induced by GJB6 deficiency, with effects comparable to those of another AKT inhibitor, MK-2206. Our findings underscore the potential of AKT inhibitors as a therapeutic option for ESCC patients. Whether AKT inhibitors could be applied as adjuvant therapy for ESCC patients, particularly those with low GJB6 expression, warrants further investigation. However, our study has several limitations. Firstly, although we found that GJB6 modulates the expression of GRHL3 via a calcium-dependent mechanism, increased Ca²⁺ influx could activate numerous signaling cascades. Hence, pinpointing the specific transcription factor that binds to the GRHL3 promoter necessitates further in-depth investigation. Secondly, the potential of AKT inhibitors for ESCC patients with low GJB6 expression or GJB6 deletion requires clinical validation. Moving forward, clinical trials are needed to establish the translational relevance and precision therapeutic value of our findings. Therefore, these aspects merit further exploration. In summary, we presented robust in vitro and in vivo evidence indicating that GJB6 suppresses carcinogenesis in ESCC and identified capivasertib, an FDA-approved AKT inhibitor, as a therapeutic agent for ESCC treatment. Mechanistically, GJB6 and GRHL3 form a mutually reinforcing positive feedback loop, thereby inhibiting the downstream AKT signaling pathway (Fig. 9). Taken together, these findings imply that GJB6 could serve as a potential therapeutic target as well as a novel prognostic marker for ESCC, which offers valuable insights for advancing treatment approaches against this lethal disease.Fig. 9Schematic diagram summarizing the contribution of GJB6 and GRHL3 to ESCC carcinogenesis via the AKT signaling pathway.GJB6 deficiency accelerates the proliferation and motility of ESCC cells. Mechanistic studies showed that GJB6 represses ESCC progression via the AKT signaling pathway in a manner of reciprocal positive regulation between GJB6 and GRHL3. GJB6 increases GRHL3 expression by enhancing intracellular Ca2+ flux, while GRHL3 transcriptionally activates GJB6 expression in turn. The AKT inhibitor CAPI restores the phenotypes induced by GJB6 suppression. Schematic diagram summarizing the contribution of GJB6 and GRHL3 to ESCC carcinogenesis via the AKT signaling pathway. GJB6 deficiency accelerates the proliferation and motility of ESCC cells. Mechanistic studies showed that GJB6 represses ESCC progression via the AKT signaling pathway in a manner of reciprocal positive regulation between GJB6 and GRHL3. GJB6 increases GRHL3 expression by enhancing intracellular Ca2+ flux, while GRHL3 transcriptionally activates GJB6 expression in turn. The AKT inhibitor CAPI restores the phenotypes induced by GJB6 suppression. Methods Samples and clinical data This study incorporated two earlier ESCC cohorts from our laboratory: A. 155 ESCC tumor tissues and adjacent normal pairs (approved by the ethical committee of Shanxi Medical University [No. 2019LL245]; GSA accession number HRA003107)17; B. 508 ESCC tumor tissues and adjacent non-tumor specimens (approved by the ethical committee of Shanxi Medical University [No. 2016LL106]; GSA accession number HRA000021)3. A total of 490 cases with complete GJB6 data were included in the survival analysis. The individuals received no prior anticancer therapy and provided written informed consent. All patient cases were classified under WHO criteria and diagnosed independently by three pathologists using hematoxylin-eosin (H&E) staining. Clinical staging of ESCC patients was performed based on the American Joint Committee on Cancer (AJCC, 8th ed.) criteria. The collection and use of specimens complied with the Declaration of Helsinki and its later amendments. Cell culture and drugs All cell lines employed in this study (KYSE30, KYSE150, KYSE450, TE-1, TE-5, TE-9 ESCC lines and HEK-293T) were commercially acquired from the Chinese Academy of Sciences’ Cell Bank of Type Culture Collection (CBTCC, Shanghai). ESCC cell lines were maintained in RPMI-1640 medium (VivaCell, Shanghai, China) while HEK-293T cells were cultured in high-glucose DMEM medium (VivaCell, Shanghai, China). All cell media were supplemented with 10% fetal bovine serum (FBS; VivaCell, Shanghai, China). Cells were cultured at 37 °C in a 5% CO₂ atmosphere and passaged based on cell states. The CCK-8 assay kit was purchased from Boster Biotech (Wuhan, China); 18α-glycyrrhetinic acid (18α-GA) was acquired from Sigma-Aldrich (St. Louis, MO, USA) and MedChemExpress (MCE; Monmouth Junction, NJ, USA); lucifer yellow CH dipotassium salt, all-trans-retinoic acid (ATRA), BATPA-AM, MK-2206 dihydrochloride and capivasertib (AZD5363) were all obtained from MCE. CRISPR Cas9 screening in vitro A crRNA library targeting 406 genes was applied to screen for CNA driver genes related to ESCC (Dharmacon, Boston, USA). Stable Cas9-expressing cell lines were generated by transducing KYSE150 and KYSE450 cells with LentiCas9-puro virus (Genechem Co., Ltd, Shanghai, China). Following 14-day selection by puromycin, the cells were continuously cultured in puromycin-supplemented media for several passages before library infection. The reverse transfection of arrayed crRNA was conducted according to the instruction manual. After the cultivation of 72 h, the reverse-transfected cells were subjected to CCK-8 assay. Genes with a relative proliferation value of ≥ 1.5-fold (upregulation) or ≤ 0.5-fold (downregulation) compared to the control were considered significant. Plasmids, siRNA and lentiviral production To achieve targeted gene overexpression, the full-length coding sequence (CDS) of GJB6 was cloned into pcDNA3.1 vector and the CDS of GRHL3 was cloned into pENTER vector. The constructs were validated by sequencing, qRT-PCR, and Western blot. Gene-specific small interfering RNAs (siRNAs) for GJB6 and GRHL3 knockdown were commercially obtained from RiboBio (Guangzhou, China), with all siRNA sequences provided in Supplementary Table S2. Transfections of plasmids and siRNA were performed via LipofectamineTM 2000 (Invitrogen, California, USA) following the manufacturer’s recommendations. The knockdown efficiency was verified by quantitative real-time PCR (qRT-PCR) and Western blot. Lentiviruses expressing shRNAs targeting GJB6 were prepared based on siGJB6 sequences that efficiently knocked down GJB6 expression. The shRNA sequences were cloned into the vector U6-MCS-Ubiquitin-Cherry-IRES-Puro by Genechem Biosciences (Genechem Co., Ltd, Shanghai, China). The cells were cultured under puromycin selection for stable expression and subsequent experiments. Cell viability assays To evaluate cell viability, Cell Counting Kit-8 (CCK-8) assay was conducted on the specified esophageal cancer cells in a 96-well plate. Cells were treated with 20 μL CCK-8 reagent (Boster, Wuhan, China) per 200 μL medium and incubated at 37 °C for 1 h in the dark. Then OD450 values were determined spectrophotometrically. At least three replicate wells were set up for each experimental group. For growth curve measurement, 3 × 103 cells were seeded into each well, and cell viability was assessed at 24-hour intervals for five days. For dose-response curve assessment, 5 × 103 cells per well were seeded, treated with serial concentrations of AKT inhibitors for 48 h, and then subjected to the CCK-8 assay. The half-maximal inhibitory concentration (IC50) was determined from dose-response curves using GraphPad Prism 8.0. Colony formation assay The colony formation assay was used to evaluate cell proliferation potential. Cells were seeded at 8×102 cells/well in 6-well plates and maintained for 14 days. For visualization, 4% PFA-fixed colonies were stained with 0.25% crystal violet dye. Those containing > 50 cells were enumerated under light microscopy. Transwell assay Cell migration and invasion capacities were assessed by Boyden chamber transwell assays in 24-well plates with or without Matrigel matrix (Corning, New York, USA). 8 × 10⁴ cells suspended in serum-free medium were seeded into the upper chamber, and 600 μL of medium supplemented with 10% FBS was added to the lower chamber. After 24-hour incubation, migrated cells on the lower membrane were fixed using 4% paraformaldehyde and subsequently stained with 0.25% crystal violet. Excess stain was removed by gentle rinsing with distilled water. Five random microscopic fields were imaged to quantify migrated cells. Wound healing assay The wound healing assay was used to compare collective migration rates between experimental groups. Cells were plated into the 6-well plates to reach 100% confluency overnight. Vertical scratch wounds were created with a 20 μL Eppendorf pipette tip. Wound closure was imaged immediately and 24 hours after the initial scratch to quantify the relative migration area. Scrape-loading dye transfer (SLDT) assay The SLDT assay was performed to detect GJIC activity according to established protocols54,55. Briefly, cells were plated at an optimized density to achieve 90-95% confluency after 24 hours of culture. The cells were gently rinsed twice with sterile phosphate-buffered saline (PBS) and stained with 0.5 mg/mL Lucifer Yellow. Scratches were generated in confluent monolayers using a sterile surgical blade. Following a 10-min incubation at room temperature, the dye solution was aspirated and washed away by PBS to remove the residual fluorescent dye. Finally, cells were fixed in 4% paraformaldehyde for 5-10 minutes and then washed again. Fluorescence microscopy was conducted using an inverted system microscope (Olympus, Tokyo, Japan) with appropriate filter sets. RNA sequencing RNA sequencing was conducted using KYSE30 cells transfected with siNC and siGJB6#1, respectively. Total RNA was isolated from cultured cells using RNAiso Plus reagent (Takara, Dalian, China) and subsequently processed by BGI Genomics (Shenzhen, China) for quality assessment, cDNA library construction, and sequencing. Based on the reference genome GRCh38.hg38, transcript abundance was presented by fragments per kilobase of transcript per million mapped reads (FPKM). After significance and FDR analysis, differentially expressed genes (DEGs) were filtered by the DESeq2 package based on the following criteria: 1) fold change > 1.5; 2) FDR < 0.05. After P values correction (Benjamini-Hochberg method), KEGG pathway enrichment was analyzed using the ‘clusterProfiler’ package in R. RNA isolation and quantitative real-time PCR Analysis (qRT-PCR) Total RNA was extracted by commercial RNAiso Plus reagent (Takara, Dalian, China) in accordance with the standardized protocol. cDNA synthesis was conducted with PrimeScript RT Master Mix (Takara, Dalian, China), and qRT-PCR analysis was subsequently performed using SYBR Green Premix Ex Taq II (Takara, Dalian, China) on an ABI StepOne Plus instrument (Foster City, CA, USA). Target gene expression levels were normalized to β-actin and calculated according to the formula R = 2 -ΔΔCt. Western blot Cells were lysed in ice-cold RIPA buffer (50 mM Tris-HCl [pH 7.6], 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS) for 30 min. Notably, protease inhibitor and phosphatase inhibitor cocktails (Roche, Basel, Switzerland) were added in accordance with manufacturer’s instructions. Cell lysates were centrifuged for 30 min at 12,000 rpm at 4 °C. Supernatants were harvested and subjected to protein quantification using the BCA assay. 50 μg of protein was separated on SDS-PAGE gels, transferred to a PVDF membrane (Millipore, Billerica, MA, USA), and subsequently blocked in 5% skim milk for 2 h at room temperature with gentle shaking. The membrane was incubated with specific primary antibodies overnight at 4 °C. The next day, after rinsing with TBST three times, the membrane was probed with species-specific IRDye 800CW-conjugated secondaries (IRDye 800CW Goat anti-Rabbit/Mouse, LI-COR) for 2 h and finally developed by LI-COR Odyssey CLx. Protein quantification was performed via β-tubulin normalization. The primary antibodies used were listed below: GJB6 (Invitrogen, Carlsbad, CA, USA), GRHL3 (Aviva Systems Biology, San Diego, CA, USA), AKT (Proteintech, Wuhan, China), p-AKT (Cell Signaling Technology, Danvers, MA, USA), PTEN (Cell Signaling Technology, Danvers, MA, USA), HA-tag (Proteintech, Wuhan, China), DYKDDDDK tag (Proteintech, Wuhan, China), β-tubulin (Proteintech, Wuhan, China). Measurement of calcium concentrations (Ca2+) Cells were seeded in the 15 mm glass-bottomed dishes for confocal imaging, followed by 24 h of spreading. Following medium removal and Hank’s balanced salt solution (HBSS, Solarbio, Beijing, China) washing, cells were loaded with 5 μM Fura-2 AM (Dojindo Laboratories, Kumamoto, Japan) at 37 °C for 30 min under light-protected conditions. Immediately afterwards, the molecular probe was removed, and cells were washed gently with HBSS solution 2–3 times, at last kept at 37 °C and 5% CO2 to fully release residual intracellular dye. Fluorescence signals were captured using a Zeiss LSM 880 confocal system, with excitation at 340/380 nm and emission at 510 nm. Dual-luciferase reporter assay In brief, cells were plated into 24-well plates for 24 h. Subsequently, transient co-transfection was conducted with firefly and Renilla luciferase reporter plasmids by Lipofectamine 2000 (Invitrogen, California, USA). Forty-eight hours later, the luciferase signals were determined based on the instructional manual of the dual-luciferase assay kit (Transgene, Beijing, China). Chromatin immunoprecipitation (ChIP) assay The ChIP assay was conducted using the SimpleChIP Enzymatic Chromatin IP Kit (Cell Signaling Technology, Boston, USA) according to the instruction manual. Briefly, cultured cells were crosslinked with formaldehyde. Next, cell disruption was achieved by ultrasonication to obtain chromatin fragmentation. Immediately afterward, GRHL3 antibody (Aviva Systems Biology, California, USA) or control IgG (Cell Signaling Technology, Boston, USA) was added to the chromatin fragments for immunoprecipitation at 4 °C overnight. DNA-protein-antibody complexes were captured by incubating the supernatants with magnetic beads the next day. The precipitated ChIP DNA fragments were de-crosslinked, purified and subjected to PCR and qPCR. The primers for ChIP-PCR/qPCR were shown in Table S1. Animal experiments All animal experiments were complied with ARRIVE guidelines and proceeded under the direction of the Institutional Animal Care and Use Committee (IACUC) of Shanxi Medical University (Approval No. SYDL2024003). Female BALB/c-Nude mice (Strain No. D000521) were purchased from GemPharmatech (Nanjing, China) and acclimated for a week before the experiments. Mice were group-housed (2–5 per cage) under specific-pathogen-free (SPF) conditions. After 2% isoflurane anesthesia (reflex/respiratory rate monitoring), mice were euthanized via cervical dislocation. To confirm the tumorigenicity of GJB6 in vivo, 3 × 106 KYSE30 cells (either control or endogenous GJB6-knockdown) were subcutaneously injected into twelve nude mice (5 weeks old, n = 6 per group). Tumor sizes were monitored by a digital vernier caliper once every three days. The tumor volume was converted by the formula V = (W2 × L)/2. After 28 days, mice were euthanized under anesthesia, and tumors were dissected and weighed. To evaluate the anti-tumor efficacy of AKT inhibitors, twenty-five five-week-old nude mice were subcutaneously injected with KYSE30 cells (3 × 106 per mouse) with stable knockdown of GJB6 or corresponding control cells. When the tumor reached a volume of nearly 50 mm3, mice with GJB6 knockdown cells were randomly divided into four groups, and treatment commenced (n = 5 per group). The animals were dosed via intraperitoneal injection with vehicle, MK-2206 or capivasertib (50 mg/kg QD; 1 day on, 2 days off) QD = once daily. Following six injections, mice were euthanized and the tissues were carefully removed for further processing. To assess AKT inhibitor-based anti-metastatic therapy, 7 × 10⁵ KYSE30 cells bearing stable GJB6 knockdown or control cells were intravenously injected into twenty-five five-week-old nude mice. One week later, vehicle, MK-2206, or capivasertib was administered intraperitoneally at a dose of 50 mg/kg every three days (n = 5 per group). Body weights were monitored during the course of treatment. Mice were sacrificed five weeks after the initial treatment for tissue harvesting and lung metastasis quantification. All the collected tissues in animal experiments were paraffin-embedded, sectioned, and subsequently stained with hematoxylin and eosin (H&E), or processed for immunohistochemistry (IHC) in line with the regular procedures. Immunohistochemistry (IHC) analysis Paraffin-embedded xenotransplanted tumors were serially sectioned at 4 μm thickness and then processed for immunohistochemical staining. Concisely, the slices were dewaxed with xylene and washed with ethanol. To block endogenous peroxidase activity, 0.3% H2O2 was applied to incubate the slices for 30 min. After the antigen retrieval in citrate buffer pH 6.0 for 15 min and rinsing three times with PBS, sections were incubated with the first antibodies overnight at 4 °C [GJB6 (Invitrogen, California, USA), GRHL3 (Aviva Systems Biology, California, USA), Ki-67 (Abcam, Cambridge, UK), p-AKT (Cell Signaling Technology, Boston, USA)]. The next day, the slides were treated with DAB plus kit (MaiXin, Fuzhou, China) for color development. Digital images were scanned via Aperio Scan Scope (Aperio Technology Inc, USA) and analyzed by K-viewer software subsequently. Data mining Gene expression levels and clinical data of 179 ESCC patients with tumor specimens and matched adjacent normal tissues were downloaded from the Gene Expression Omnibus (GEO) dataset (GSE53625)56. Proteomics data of GJB6 and GRHL3 were acquired from the iProX database (IPX0002501000)57. TCGA-derived gene expression profiles for GJB6 and GRHL3 across four squamous cell carcinomas (ESCC, LUSC, HNSC, CESC) were retrieved from UCSC Xena browser (https://xenabrowser.net/heatmap/)58. Closely-related GJB6 genes were investigated using LinkedOmics (https://www.linkedomics.org/login.php), cBioPortal (https://www.cbioportal.org/), and STRING tool (https://cn.string-db.org/)59,60. The data for single-cell RNA sequencing were retrieved from GSE188900 to detect the expression of GJB6 in ESCC tissues61. The DoubletFinder package, Seurat package, and Harmony package were employed for data processing to obtain 49,651 cells across eight samples, including one normal tissue sample. The TCGA-ESCC data were stratified into two groups: low (below median) GJB6 expression group and high (above median) GJB6 expression group, and the low group was set as the reference. DESeq2 package was used to identify DEGs according to the parameters of |fold change | ≥ 2 and adjusted P value < 0.05. Statistical analysis In the research, all experiments were repeated in at least three separate trials and three biological replicates were performed for each treatment. Data were presented with mean ± SD and processed by SPSS 22.0 software (IBM Inc, Armonk, New York, USA) or GraphPad Prism 8.0 software (La Jolla, CA, USA). Student’s t-test was applied to comparisons between two groups, whereas one-way ANOVA test was employed for comparisons of ≥ 3 groups. Gene-clinical feature correlations in ESCC were statistically examined by chi-square analysis. Kaplan-Meier survival analyses were performed by Log-rank test. Cox regression (univariate and multivariate) was applied to identify the impacts of variables on overall survival (OS). Pearson correlation analysis by linear regression was calculated to evaluate gene expression relationships. Statistical significance was defined as P < 0.05. Supplementary information Supplementary information Supplementary figure legend Supplementary information Supplementary figure legend