Introduction

The Carcinoembryonic Antigen Cell Adhesion Molecule (CEACAM) family, consisting of membrane-linked glycoproteins, plays a vital role in regulating numerous cellular processes, including cell adhesion, proliferation, differentiation, survival, and interactions with pathogens and hosts through signal transduction mechanisms that activate kinases, induce cytoskeletal rearrangements, and stimulate gene expression1. Among these family members, CEACAM5 (CEA) and CEACAM6, located on human chromosome 19, are particularly noteworthy for their involvement in cancer progression within the aerodigestive tract2. Initially identified as highly expressed in colon cancers and the fetal digestive tract, CEACAM5 demonstrates distinct tissue expression patterns compared to CEACAM6, which shares structural similarities3,4,5. Notably, elevated serum CEACAM5 levels following surgical resection in colon cancer patients strongly correlate with recurrence or distant metastasis, prompting the routine use of serum CEACAM5 immunoassays in postoperative colon cancer surveillance6.

In lung adenocarcinoma, activating mutations of the Epidermal Growth Factor Receptor (EGFR) serve as effective therapeutic targets and predictive biomarkers for EGFR tyrosine kinase inhibitors (TKIs)7,8,9,10,11. However, resistance to EGFR-TKI treatments eventually develops in patients12,13,14,15,16. Given its noninvasive nature, CEACAM5 in the bloodstream is a monitoring marker during EGFR-TKI targeted therapy or chemotherapy in lung adenocarcinoma patients17,18,19. Several studies have linked serum CEACAM5 levels with EGFR-TKI treatment outcomes, although some cases indicate that CEACAM5 expression in the blood does not necessarily reflect disease progression during EGFR-TKI treatment20,21,22,23,24.

CEACAM6 has been identified as a key player in cancer progression and therapy response. CEACAM6 interacts with human epidermal growth factor receptor 2 (HER2) and has been shown to influence the therapeutic efficacy of trastuzumab, a HER2 inhibitor commonly used in breast cancer treatment25. Beyond breast cancer, CEACAM6 has been associated with increased cell proliferation and migration in renal and gallbladder cancers, primarily through the activation of ERK and AKT signaling pathways26,27. In lung cancer, CEACAM6 expression has been linked to enhanced cellular proliferation28, and its elevated levels in EGFR mutation-positive lung adenocarcinomas are predictive of improved progression-free survival with EGFR tyrosine kinase inhibitor (EGFR-TKI) therapy29. However, the precise role of CEACAM6 in EGFR-mutant lung cancer remains unclear.

Transforming growth factor-beta (TGF-β), a multifunctional cytokine implicated in cell differentiation and epigenetic regulation, is known to induce epithelial-mesenchymal transition (EMT) and promote resistance to EGFR-TKIs in EGFR-mutant lung cancers30,31. Despite these findings, the interplay between TGF-β signaling, epigenetic modifications, CEACAM6 expression, and EGFR-TKI resistance in lung cancer has yet to be fully understood. This study aims to elucidate the potential crosstalk between these factors to better comprehend the mechanisms underlying therapy resistance in EGFR-mutant lung cancer.

Our study observed higher CEACAM6 expression levels compared to CEACAM5 in lung adenocarcinoma. Moreover, we discovered an interaction between CEACAM6 and EGFR, which stabilized EGFR expression and facilitated growth in lung cancer cells with EGFR mutations. Through a comprehensive analysis of CEACAM6 expression before and after EGFR-TKI resistance, we uncovered its potential as a biomarker for discerning EGFR dependency in lung cancer cells.

Results

High expression of CEACAM6 in lung adenocarcinoma

We investigated the differential expression of CEACAM6 across various subtypes of non-small cell lung cancer (NSCLC) to assess its potential as a biomarker. Gene expression profiling analysis revealed that CEACAM6 expression levels were significantly higher in lung adenocarcinoma compared to small cell lung cancer and squamous cell carcinoma (Fig. 1A). Simultaneously, we confirmed elevated expression of CEACAM5 in adenocarcinoma compared to other lung cancer subtypes (Supplementary Fig. 1A). Notably, CEACAM6 expression surpassed that of CEACAM5 in lung adenocarcinoma (Supplementary Fig. 1B). Further analysis showed that CEACAM6 expression levels were markedly elevated in lung adenocarcinoma compared to normal tissues (Fig. 1B). This trend remained consistent when analyzing paired non-tumorous and tumorous tissues, further highlighting the higher expression of CEACAM6 in tumor tissue (Fig. 1C, left). The tumor-to-normal (TN) ratio was calculated, with an average value greater than 2, emphasizing the elevated expression of CEACAM6 in tumorous tissue compared to normal tissue in adenocarcinoma (Fig. 1C, right). Immunohistochemistry staining of paired tumorous and non-tumorous tissues corroborated these findings, illustrating higher CEACAM6 expression in tumorous tissue (Fig. 1D). Immunohistochemistry analysis of a lung cancer tissue array confirmed that CEACAM6 was more abundantly expressed in lung adenocarcinoma compared to squamous cell carcinoma and other subtypes of lung cancer (Fig. 1E). Additionally, RNA-seq analysis was conducted to assess the CEACAM family gene expression in non-small cell lung cancer and colorectal cancers. CEACAM6 emerged as the most abundantly expressed gene in lung adenocarcinoma and squamous cell carcinoma, whereas CEACAM5 showed significant expression in colon and rectal cancer (Fig. 1F and Supplementary Fig. 2A-C). Moreover, we compared CEACAM5 and CEACAM6 expression in lung adenocarcinoma and colorectal cancer with that in breast, ovarian, pancreas, and bladder cancers, where CEACAM5 has been utilized as a biomarker. Our analysis indicated that CEACAM6 displayed higher expression levels than CEACAM5 in most cancers, except for colon and rectal cancers (Supplementary Fig. 2D). In summary, our data underscore the high expression of CEACAM6 in lung adenocarcinoma, suggesting its potential utility as a biomarker in this context.

Fig. 1: CEACAM6 expression in lung cancer.
figure 1

A Gene expression profiling analysis assessing CEACAM6 expression in small cell lung cancer (SCLC, n = 20), lung squamous cell carcinoma (SQC, n = 13), and lung adenocarcinoma (ADC, n = 39) from the GSE29016 database. Statistical significance indicated as *P < 0.05; ***P < 0.001; NS, non-significant. B Gene expression profiling analysis evaluating CEACAM6 expression in non-tumorous (normal, n = 19) versus tumorous (tumor, n = 20) tissues from primary human lung adenocarcinoma from the GSE2514 database. **P < 0.01. C Gene expression profiling analysis showing CEACAM6 expression (left) in paired non-tumorous versus tumorous tissues (n = 83) and analyzing tumor versus normal (T/N) ratio (right) from primary human lung adenocarcinoma from the GSE75037 database. D Immunohistochemical staining assessing CEACAM6 expression in a pair of cancerous (upper) and non-cancerous (lower) lung tissues. E Immunohistochemical staining evaluating CEACAM6 expression in different lung cancer subtypes from CC5 tissue array, with representative images of lung adenocarcinoma (ADC) versus squamous cell carcinoma (SQC) and quantitative H-score values (right). F RNA-seq analysis assessing CEACAM family gene expression in lung adenocarcinoma (n = 517) from the TCGA-LUAD database. Expression levels of CEACAM6 were compared to those of other CEACAM family genes using one-way ANOVA. ***P < 0.001.

Full size image

CEACAM6 expression influences the growth of lung adenocarcinoma

CEACAM6 is highly expressed in the lung adenocarcinoma cell line HCC827, characterized by the EGFR mutant DelE746-A750 (Fig. 2A). To explore the role of CEACAM6 in lung cancer, we silenced CEACAM6 in HCC827 cells, validated by qRT-PCR and immunoblotting analysis (Fig. 2A). Annexin V assay demonstrated that CEACAM6 knockdown induced apoptosis (Fig. 2B). Furthermore, to understand the downstream pathways affected by CEACAM6 knockdown, we conducted RNA-sequencing analysis on CEACAM6-silenced and control HCC827 cells. Gene Set Enrichment Analysis (GSEA) analysis of the RNA-seq data revealed downregulation of proliferation-related pathways in CEACAM6-high (top 100 cases) versus CEACAM6-low (lowest 100 cases) expression in adenocarcinomas from TCGA-LUAD (Fig. 2C and Supplementary Fig. 3A). Given the known impact of phosphorylated ERK1/2 (p-ERK1/2) and phosphorylated AKT (p-AKT) levels on lung cancer cell growth, we evaluated their expression in CEACAM6-silenced HCC827 cells. Immunoblotting confirmed that p-ERK1/2, rather than p-AKT, was attenuated in CEACAM6-silenced cells compared to control HCC827 cells (Fig. 2D). Reverse phase protein array (RPPA) assay from TCGA also demonstrated a positive correlation between p-ERK1/2 levels and CEACAM6 expression (Fig. 2E). Additionally, the epithelial cell proliferation signature was significantly higher in CEACAM6-high tumors compared to CEACAM6-low tumors from the TCGA-LUAD cohort (Fig. 2F). These findings underscore the impact of CEACAM6 expression on the growth of lung adenocarcinoma.

Fig. 2: CEACAM6 influence on proliferation in lung cancer cells.
figure 2

A qRT-PCR (upper) and immunoblotting (lower) analysis of CEACAM6 expression in HCC827 cells transduced with the lentiviral vector encoding shCEACAM6 or scramble control. ***P < 0.001. B Annexin V assay measuring apoptotic cells in HCC827 cells transduced with shCEACAM6 or scramble control. **P < 0.01. C GSEA from RNA-seq data identifying differential gene expression involved in the regulation of epithelial-cell-proliferation pathway (left) and epithelial-cell-proliferation pathway (right) in CEACAM6-high (top 100 cases) versus CEACAM6-low (lowest 100 cases) expression in adenocarcinomas from TCGA-LUAD. D Immunoblotting analysis assessing p-ERK1/2 and p-AKT expression in HCC827 cells transduced with shCEACAM6 or scramble control. Quantification was performed using ImageJ software. E Dot-plot analysis to assess the correlation between p-ERK1/2 (pT202Y204) and CEACAM6 expression by RPPA profiling and RNA-seq from the TCGA-LUAD database (n = 357). F Correlation analysis evaluating the epithelial cell proliferation signature levels in CEACAM6-high (top 50 cases) versus CEACAM6-low (lowest 50 cases) expression in adenocarcinomas from TCGA-LUAD. The epithelial cell proliferation signature was derived from Gene Ontology63.

Full size image

Loss of CEACAM6 expression and decreased ERK1/2 signaling in lung adenocarcinoma cells under EGFR-TKI selection

To investigate the impact of EGFR-TKI resistance on CEACAM6 expression and EGFR dependency in lung adenocarcinoma, we generated EGFR-TKI-resistant lung cancer cells (HCC827GR) by subjecting HCC827 cells to incremental concentrations of gefitinib over more than six months. Clonogenic assays confirmed that HCC827GR cells exhibited higher resistance to gefitinib compared to the parental HCC827 cells (Fig. 3A). Immunoblotting analysis revealed reduced p-ERK 1/2 levels in HCC827GR cells (Fig. 3B). Further analysis using qRT-PCR showed decreased CEACAM6 expression in HCC827GR cells (Fig. 3C). The loss of CEACAM6 protein in HCC827GR cells post-EGFR-TKI selection was confirmed by ELISA and immunoblotting assays, contrasting with its high expression in the parental HCC827 cells (Fig. 3D, E). GSEA analysis of gene expression profiling assays indicated upregulation of cell-cycle-related pathways and ERK1/2 cascade signaling in HCC827 cells compared to HCC827GR cells (Fig. 3F). These findings suggest a potential involvement of CEACAM6 expression in the EGFR-ERK1/2 signaling pathway in lung adenocarcinoma and imply that the loss of CEACAM6 expression may contribute to the development of resistance to EGFR-TKIs.

Fig. 3: Deficient CEACAM6 expression and ERK1/2 signaling in EGFR-TKI resistant lung cancer cells.
figure 3

A Clonogenic analysis depicting the growth of HCC827 and HCC827GR cells in the presence of increasing concentrations of gefitinib for 14 days. HCC827GR cells were derived from a pool of HCC827 cells that survived incremental concentrations of gefitinib selection over 6 months. B Immunoblotting analysis assessing p-ERK1/2 and p-AKT expression in HCC827 versus HCC827GR cells. p-ERK1/2 and p-AKT expression were quantified using ImageJ software and normalized with GAPDH. C qRT-PCR analysis of CEACAM6 expression in HCC827 versus HCC827GR cells, represented as the cycle-threshold (Ct) values (left) and relative expression folds (right). ***P < 0.001. D ELISA measuring CEACAM5 and CEACAM6 expression in the conditioned medium from HCC827 and HCC827GR cells. E Immunoblotting analysis of CEACAM6 expression in HCC827 versus HCC827GR cells. F GSEA from RNA-seq data identifying differential gene expressions involved in epithelial cell proliferation (left) and ERK1/2 cascade (right) pathways between HCC827 and HCC827GR cells.

Full size image

CEACAM6 interacts with EGFR and prolongs its ability in EGFR-mutated lung adenocarcinoma

We investigated whether CEACAM6 expression correlates with the oncogenic driver EGFR status in lung adenocarcinoma. Gene expression profiling revealed that CEACAM6 expression, but not CEACAM5, is more highly expressed in EGFR-mutated lung adenocarcinomas than the wild-type ones (Fig. 4A, B, and Supplementary Fig. 4A). RPPA assays and RNA-seq data further supported a positive correlation between phosphorylated EGFR (p-EGFR) levels and CEACAM6 expression, suggesting CEACAM6-high lung adenocarcinoma tumors may harbor high EGFR activity (Fig. 4C). Immunoblotting assays showed that higher expression levels of both CEACAM6 and EGFR in HCC827 cells compared to their EGFR-TKI resistant counterpart HCC827GR prompted us to explore their potential interaction (Fig. 4D). Co-immunoprecipitation assays using antibodies against CEACAM6 were conducted in HCC827 and HCC827GR cells to test for endogenous interaction with EGFR. The results demonstrated an interaction between CEACAM6 and EGFR in HCC827 cells. In contrast, no co-immunoprecipitated EGFR was detected in HCC827GR cells (Fig. 4D). To explore the functional implications of this interaction, we conducted cycloheximide chase assays to assess the stability of endogenous EGFR in HCC827 and HCC827GR cells. These assays indicated that both total EGFR and p-EGFR were more stable in CEACAM6-positive HCC827 cells compared to CEACAM6-negative HCC827GR cells (Fig. 4E). Further experiments involved overexpressing CEACAM6 in HCC827GR cells confirmed by immunoblotting (Fig. 4F). Cycloheximide chase assays demonstrated that CEACAM6 overexpression stabilized total EGFR and p-EGFR in HCC827GR cells (Fig. 4G). To investigate potential degradation pathways of EGFR in HCC827GR cells, we blocked autophagy with chloroquine and inhibited the proteasome with MG115. It is that chloroquine treatment, but not MG115 treatment, prolonged EGFR stability, suggesting the involvement of autophagy-mediated protein degradation in regulating EGFR stability (Supplementary Fig. 4B). To understand the impact of CEACAM6 on EGFR-TKI resistance, we manipulated CEACAM6 levels by knocking it down in HCC827 cells and overexpressing it in HCC827GR cells. Clonogenic assays revealed that CEACAM6 knockdown improved EGFR-TKI resistance in HCC827 cells (Fig. 4H). Conversely, overexpression of CEACAM6 re-sensitized HCC827GR cells to gefitinib (Fig. 4I). These findings suggest that CEACAM6 interacts with EGFR, promotes EGFR stability, and plays a critical role in EGFR-TKI resistance in lung adenocarcinoma.

Fig. 4: CEACAM6 expression enhances EGFR stability in lung adenocarcinoma.
figure 4

A RNA-seq analysis evaluating the correlation of CEACAM5 (left) and CEACAM6 (right) expression with EGFR status [wild-type (n = 193) versus mutant (n = 20) EGFR] in lung cancer cells from the CCLE database. *P < 0.05. B RNA-seq analysis assessing the correlation of CEACAM5 (left) and CEACAM6 (right) expression with EGFR status [wild-type (n = 444) versus mutant (n = 73) EGFR] in lung adenocarcinoma (n = 517) from the TCGA-LUAD database. C Dot-plot analysis assessing the correlation between p-EGFR (pY1068) and CEACAM6 by RPPA profiling and RNA-seq, respectively, from the TCGA-LUAD database (n = 357). D Co-immunoprecipitation analysis using an anti-CEACAM6 antibody to assess the interaction of CEACAM6 with EGFR in HCC827 and HCC827GR cells. E Immunoblotting analysis assessing EGFR and p-EGFR expression in HCC827 versus HCC827GR cells treated with cycloheximide (CHX, 50 ng/ml) for the indicated periods. F Immunoblotting analysis assessing CEACAM6 expression in HCC827GR cells transduced with the lentiviral vector encoding Flag-CEACAM6 or control vector. G Immunoblotting analysis assessing EGFR and p-EGFR expression in HCC827GR cells transduced with Flag-CEACAM6 or control vector followed by treatment with cycloheximide (CHX, 50 ng/ml) for the indicated periods. H Clonogenic assays of HCC827 cells transduced shCEACAM6 or scramble control in the presence of gefitinib (100 nM) for 14 days. The mean colony area (%) is presented relative to shCEACAM6-transduced cells treated with gefitinib (set as 100%), with error bars representing the standard deviation. *P < 0.05. I Clonogenic assays in HCC827GR cells transduced with Flag-CEACAM6 or control vector in the presence or absence of gefitinib (1 μM) for 14 days. The mean colony area (%) is presented relative to control or CEACAM6-overexpressing cells in the absence of gefitinib (set as 100%), with error bars representing the standard deviation. **P < 0.01.

Full size image

The loss of CEACAM6 in lung cancer is associated with EMT and EGFR-TKI resistance

To investigate the role of CEACAM6 in EMT-related EGFR-TKI resistance, we analyzed phenotypical changes and EMT marker gene expression in HCC827 and HCC827GR cells. Morphological examination revealed a spindle-like morphology in HCC827GR cells, indicative of an EMT phenotype, whereas HCC827 cells exhibited an epithelial-like appearance (Fig. 5A). Additionally, GSEA showed an upregulation of the EMT pathway in CEACAM6-silenced HCC827 cells compared to control cells (Fig. 5B). Unsupervised classification of gene expression in lung adenocarcinoma demonstrated that CEACAM6 expression clustered with epithelial markers EPCAM and CDH1, distinct from mesenchymal markers TWIST1, SNAI1/2, and ZEB1/2 (Fig. 5C). qRT-PCR analysis further revealed higher expression of mesenchymal markers VIM, AXL, and SNAI2 in HCC827GR cells, whereas epithelial markers CDH1 and EPCAM were more prevalent in HCC827 cells (Fig. 5D). Immunoblotting analysis also demonstrated that epithelial marker CDH1 was downregulated and mesenchymal marker AXL was upregulated in HCC827GR compared to HCC827 cells (Fig. 5E). Selection of H1975 cells with incremental osimertinib concentrations received resistant cells H1975OR. In H1975OR cells, CEACAM6 expression decreased while mesenchymal marker VIM increased and CDH1 decreased (Supplementary Fig. 5A-C). Similar patterns were observed in EGFR-TKI-resistant HCC827ER clones (HCC827ER and HCC827ER2) and PC9GR clones (PC9GR1 and PC9GR3), selected by treatment with erlotinib and gefitinib, respectively, showing lower CEACAM6 levels along with increased EMT marker gene expression (Supplementary Fig. 6A, B). Gene expression profiling from primary lung adenocarcinoma tumor MGH119 cells and the gefitinib selected counterpart MGH119GR revealed lower CEACAM6 levels and higher EMT marker gene expression in lung cancer cells after EGFR-TKI selection (Supplementary Fig. 6C). Following erlotinib selection in xenograft tumors of HCC827 cells, CEACAM6 expression decreased while mesenchymal marker VIM increased (Fig. 5F), indicating an impact of EGFR-TKI treatment on CEACAM6 and EMT marker gene expression. Knockdown of CEACAM6 in HCC827 cells induced EMT, as evidenced by decreased CDH1 and increased SNAI2 and AXL expression (Fig. 5G). These findings underscore the critical role of CEACAM6 in maintaining an epithelial phenotype and sensitivity to EGFR-TKI treatment.

Fig. 5: Reversed association of CEACAM6 expression with EMT signature in lung adenocarcinoma.
figure 5

A Phase contrast imaging analysis showing the distinct morphologies between HCC827 and HCC827GR cells. B GSEA from RNA-seq data identifying differential gene expression in EMT pathway in CEACAM6-silenced HCC827 cells versus scramble control cells. C Hierarchical clustering analysis depicting CEACAM5, CEACAM6, and EMT marker gene expression in lung adenocarcinoma from the TCGA-LUAD database. D qRT-PCR analysis assessing CEACAM6, epithelial (CDH1 and EPCAM), and mesenchymal (VIM, SNAI2, and AXL) marker gene expression in HCC827 versus HCC827GR cells. **P < 0.01; ***P < 0.001. E Immunoblotting analysis assessing CDH1 and AXL expression in HCC827 and HCC827GR. F Gene expression profiling analysis evaluating CEACAM6, CDH1, VIM, and AXL expression in the erlotinib-resistant tumors versus the parental xenograft tumors of HCC827 cells from the Bivona database. G qRT-PCR analysis assessing CEACAM6 and epithelial (CDH1) and mesenchymal (SNAI2 and AXL) marker gene expression in HCC827 transduced with shCEACAM6 and scramble control. ***P < 0.001.

Full size image

CEACAM6 expression is influenced by TGF-β stimulation

We explored the impact of TGF-β stimulation on CEACAM6 and EMT marker gene expression to understand its role in inducing EMT-related EGFR-TKI resistance. qRT-PCR analysis demonstrated that TGF-β stimulation led to the upregulation of VIM and the suppression of CDH1, indicating TGF-β-induced EMT in HCC827 cells (Fig. 6A, left). Furthermore, TGF-β stimulation also decreased CEACAM6 expression in HCC827 cells (Fig. 6A, right). The immunoblotting analysis also showed that TGF-β stimulation induced EMT in HCC827 (Fig. 6B). Additionally, TGF-β stimulation increased EGFR-TKI resistance in HCC827 cells (Fig. 6C). Morphological changes in HCC827 cells treated with TGF-β further supported the induction of EMT by TGF-β stimulation (Fig. 6D). Similarly, qRT-PCR and immunoblotting analyses in A549 cells showed that TGF-β induced EMT and downregulated CEACAM6 expression (Figs. 6E, F, G, and Supplementary Fig. 7A). ELISA analysis confirmed the downregulation of CEACAM6 protein expression following TGF-β stimulation (Supplementary Fig. 7B). To investigate the involvement of the SMAD-dependent pathway in TGF-β-induced CEACAM6 downregulation, SMAD4 was silenced in A549 cells, followed by TGF-β stimulation. qRT-PCR analysis revealed that silencing SMAD4 partially reversed TGF-β-mediated suppression of CEACAM6 (Fig. 6H). These results suggest that the SMAD-dependent pathway regulates CEACAM6 expression in response to TGF-β stimulation.

Fig. 6: TGF-β signaling regulates CEACAM6 expression and EGFR-TKI resistance.
figure 6

A qRT-PCR analysis evaluating CDH1, VIM, and AXL expression (left) and CEACAM6 expression (right) in HCC827 cells treated with or without TGF-β (1 ng/mL) for 3 weeks. B Immunoblotting analysis assessing CDH1 and AXL expression in HCC827 and HCC827GR. C Clonogenic analysis assessing the growth of TGF-β-treated and parental HCC827 cells in the presence of increasing concentrations of gefitinib for 14 days. ***P < 0.001. D Phase contrast imaging analysis showing the distinctive morphologies between TGF-β-treated HCC827 and parental cells. E qRT-PCR analysis evaluating CDH1, VIM, and AXL expression in A549 treated with or without TGF-β (1 ng/mL) for 3 weeks. **P < 0.01. F Immunoblotting analysis assessing CDH1 and AXL expressions in A549 treated with or without TGF-β (1 ng/mL) for 3 weeks. G qRT-PCR analysis assessing CEACAM6 expression in A549 treated with or without TGF-β (1 ng/mL) for 3 weeks. **P < 0.01. H qRT-PCR analysis evaluating SMAD4 expression (left) in A549 transduced with the lentiviral vector encoding shSMAD4 or scramble control. Additionally, qRT-PCR analysis assessing CEACAM6 expression (right) in A549 transduced with shSMAD4 or scramble control, followed by treatment with TGF-β (1 ng/mL) for 3 days. ***P < 0.001.

Full size image

CEACAM6 expression is subject to epigenetic regulation

Given that the epigenetic modification of H3K27ac, which signifies active enhancers, is tightly regulated by cytokine stimulation, we explored whether epigenetic modifications influence CEACAM6 expression under EGFR-TKI selection or TGF-β stimulation. ChIP-seq analysis illustrated a robust H3K27ac signal at the CEACAM6 locus in EGFR-TKI-sensitive HCC827 and PC9 lung cancer cells (Fig. 7A). Further ChIP-qPCR analysis showed that EGFR-TKI-sensitive HCC827 cells exhibited a stronger H3K27ac signal at the CEACAM6 locus compared to EGFR-TKI-resistant HCC827GR cells (Fig. 7B, left). Conversely, TGF-β-treated A549 cells displayed a reduced H3K27ac signal at the CEACAM6 locus compared to untreated cells (Fig. 7B, right). To investigate the impact of epigenetic regulation on CEACAM6 expression, we treated HCC827 cells with the HDAC1/2 inhibitor romidepsin. Interestingly, CEACAM6 expression decreased following romidepsin treatment in HCC827 cells (Fig. 7C, left), concomitant with increased EGFR-TKI resistance (Fig. 7C, right). Romidepsin treatment also induced EMT in HCC827 cells (Fig. 7D). Additionally, we assessed the expression of the TGF-β family and found that romidepsin treatment led to upregulation of TGF-β family members (Fig. 7E). Furthermore, treatment of romidepsin-pretreated HCC827 cells with the TGF-β receptor inhibitor SB-431542 rescued CEACAM6 expression (Fig. 7F). These findings collectively suggest that epigenetic mechanisms, including H3K27ac modification and HDAC activity, play a crucial role in regulating CEACAM6 expression and TGF-β signaling, thereby influencing EGFR-TKI sensitivity and EMT dynamics in lung cancer cells.

Fig. 7: Epigenetic control of CEACAM6 expression in lung cancer cells.
figure 7

A ChIP-seq analysis from the SRX1528527 and SRX2636288 database assessing the H3K27ac signal around the CEACAM6 loci in EGFR-mutant HCC827 and PC9 lung cancer cells. B ChIP-seq analysis evaluating the H3K27ac signal at indicated regions of the CEACAM6 locus in (left) HCC827 versus HCC827GR and (right) A549 treated with or without TGF-β (1 ng/mL) for 3 weeks. *P < 0.05; **P < 0.01; ***P < 0.001; NS, non-significant. C qRT-PCR analysis (left) evaluating CEACAM6 expression in HCC827 cells treated with or without HDAC1/2 inhibitor romidepsin (HDACi, 1 nM) for 3 weeks. Clonogenic assay (right) assessing the growth of romidepsin-treated HCC827 and parental cells in the presence of gefitinib (100 nM) for 14 days. D qRT-PCR analysis assessing CDH1 and VIM expression in HCC827 cells treated with or without romidepsin (HDACi, 1 nM) for 3 weeks. E qRT-PCR analysis evaluating TGF-β family member and receptor gene expression in HCC827 cells treated with or without romidepsin (HDACi, 1 nM) for 3 weeks (left). F qRT-PCR analysis assessing CEACAM6, CDH1, and VIM expression in HCC827 cells pretreated with romidepsin (HDACi, 1 nM) for 3 weeks, followed by further treatment with or without TGF-β receptor inhibitor SB-431542 (SB, 5 μM) for 3 weeks.

Full size image

Discussion

The CEACAM family of proteins initially recognized for their adhesive properties have emerged as significant players in cancer signaling pathways and potential biomarkers for cancer monitoring. In our study, we observed heightened levels of CEACAM6 compared to CEACAM5 in lung adenocarcinoma. Notably, CEACAM6 exhibited a direct interaction with EGFR, and its expression correlated with the EGFR mutation status in lung cancer. Moreover, our investigation unveiled that CEACAM6 expression extended the stability of EGFR in lung cancer cells. Upon silencing CEACAM6, we observed diminished ERK phosphorylation and increased cell death, highlighting the pivotal role of CEACAM6 signaling in EGFR-mutated lung cancer cells. During EGFR-TKI treatment, we observed a loss of CEACAM6 expression, coinciding with a loss of EGFR expression and EGFR-TKI sensitivity in lung cancer cells. Furthermore, our study identified TGF-β and epigenetic mechanisms as critical regulators of CEACAM6 expression and EGFR-TKI sensitivity. These findings collectively suggest that CEACAM6 plays a regulatory role in the EGFR signaling pathway and holds promise as a potential biomarker for monitoring activating EGFR-dependency during EGFR-TKI treatment in lung cancer (Fig. 8).

Fig. 8: Schematic illustration of CEACAM6-mediated EGFR signaling in lung adenocarcinoma.
figure 8

CEACAM6 interacts with EGFR, enhancing EGFR stability and maintaining epithelial morphology, thereby promoting lung cancer cell growth. TGF-β stimulation and epigenetic regulation downregulate CEACAM6, shifting toward EGFR-independent signaling and consequent EGFR-TKI resistance characterized by elevated EMT features.

Full size image

CEACAM5 is expressed in variant cancers, especially those from the gastrointestinal tract, and exhibits high specificity in colon cancer compared with other biomarkers32,33,34,35,36,37. Reports indicate that the CEACAM5 level in lung adenocarcinoma is higher than in squamous cell carcinoma38. We observed lung adenocarcinoma exhibits higher CEACAM5 and CEACAM6 levels than lung squamous cell carcinoma and small cell lung cancer. Similarly, CEACAM5 is the most abundant member of the CEA family in colon and rectal cancers, whereas CEACAM6 is the most enriched in lung adenocarcinoma and squamous cell carcinoma. In addition to colorectal and lung cancers, CEACAM5 has been used as a biomarker in pancreas, breast, bladder, and ovarian cancer39,40,41,42. Here, we observed that except for colorectal cancers, CEACAM6 levels are higher than CEACAM5 in cancer of the pancreas, breast, bladder, ovarian, and lung (adenocarcinoma and squamous cell carcinoma). The notion that while CEACAM5 is suitable as a biomarker in colorectal cancer, CEACAM6 is a potential biomarker in lung adenocarcinoma is supported by our findings.

HER2, belonging to the EGFR family, is a significant driver of tumor development for a subset of breast cancer. In breast cancer, high CEACAM6 expression is linked with HER2-enriched and luminal subtypes compared with the basal-like one43. CEACAM6 interacts with HER2 in breast cancer cells, and CEACAM6 knockdown decreases the inhibitory effects of trastuzumab, a humanized anti-HER2 antibody, in the trastuzumab-sensitive cells but not in the trastuzumab-resistant ones25. CEACAM6 has interacted with the wild-type EGFR in oral cancer upon ligand activation44. Here, we showed that the expression of CEACAM6, excluding CEACAM5, is associated with EGFR mutation status in lung cancer. We observed that CEACAM6 interacts with EGFR. Additionally, we discovered that CEACAM6 expression prolongs the half-life of mutant EGFR. To verify the effect of prolonging the half-life of EGFR, we check p-ERK and p-AKT, the downstream pathway of EGFR signaling. We found that the knockdown of CEACAM6 attenuated p-ERK instead of p-AKT. The results correspond to the treatment of EGFR-TKIs45. These findings indicate that CEACAM6 stabilizes the EGFR, thus maintaining oncogenic EGFR signaling in EGFR-mutated lung cancer.

Although most lung cancer containing activating EGFR mutants respond well to EGFR-TKIs, drug resistance inevitably occurs11,46,47,48,49,50,51. Resistance mechanisms to EGFR-TKIs can be credited to EGFR-dependent and EGFR-independent mechanisms. EGFR gatekeeper mutations, such as T790M and C797S, contribute to the EGFR-dependent TKI resistance mechanism52,53. Conversely, the transition to EGFR-independent signaling pathways, such as AXL upregulation, also plays a role in TKI resistance54,55. Our study investigated EGFR-TKI resistance mechanisms, particularly in CEACAM6 expression. Our findings revealed a notable downregulation of CEACAM6 expression in both in vitro EGFR-TKI-resistant cells and in vivo tumor models. Notably, the EGFR-resistant HCC827GR cells lacked the typical EGFR T790M or C797S mutations but showed reduced levels of phosphorylated EGFR and increased AXL expression compared to the TKI-sensitive parental cells. This reduction suggested that HCC827GR cells utilized an EGFR-independent mechanism, such as AXL upregulation, to resist TKIs. Interestingly, the decrease in CEACAM6 expression was inversely correlated with EGFR stability; reintroducing CEACAM6 expression led to increased EGFR stability and enhanced sensitivity to EGFR-TKIs. These results support the hypothesis that CEACAM6 plays a crucial role in EGFR signaling pathways. Given that CEACAM6 interacts with active EGFR and stabilizes its activity, the observed downregulation of CEACAM6 expression in EGFR-TKI-resistant patients upon recurrence suggests a potential predictive marker for tumors employing EGFR-independent TKI resistance mechanisms. This insight could inform personalized treatment strategies, where monitoring CEACAM6 levels could help identify patients likely to benefit from alternative therapeutic approaches targeting EGFR-independent pathways.

EMT has been identified as a crucial mechanism of acquired resistance to EGFR-TKIs, mainly belonging to non-EGFR gatekeeper mutation-dependent pathways13,51,56,57. Our study builds on this understanding by investigating the role of CEACAM6 in this process. We found that CEACAM6 expression is inversely associated with EMT in lung adenocarcinoma tumors and that it can be downregulated through a TGF-β-induced, SMAD-dependent pathway. CEACAM6 appears to interact with EGFR to enhance its stability. Thus, TGF-β-induced downregulation of CEACAM6 suggests that CEACAM6 could act as a regulatory switch, modulating the dependence on oncogenic EGFR signaling. Our data demonstrate that silencing CEACAM6 promotes both EGFR-TKI resistance and EMT, which supports the hypothesis that CEACAM6 mediates EMT-related EGFR-TKI resistance. Furthermore, we uncovered heightened H3K27ac epigenetic modifications at the CEACAM6 locus in EGFR-TKI-sensitive cells compared to their resistant counterparts. TGF-β stimulation was observed to decrease CEACAM6 expression and reduce H3K27ac levels at the CEACAM6 locus, contributing to EGFR-TKI resistance in lung cancer cells. These findings underscore the substantial impact of cancer epigenetics on CEACAM6 expression and the development of TGF-β-mediated EGFR-TKI resistance. Additionally, the use of the HDAC1/2 inhibitor romidepsin in EGFR-TKI-sensitive lung cancer cells resulted not only in decreased CEACAM6 expression but also in the induction of TGF-β receptor and ligand expression. Although HDAC inhibitors are typically known to promote MET through transcriptional activation58,59, in this study, romidepsin treatment resulted in a significant increase in TGFBR2 expression (more than 20-fold), which strongly activated TGF-β signaling. As a result, CEACAM6 expression was suppressed, leading to the induction of EMT. These findings highlight epigenetic regulation and TGF-β signaling’s intertwined nature in modulating CEACAM6 levels. Furthermore, our study revealed that treatment with SB-431542, a TGF-β type I receptor inhibitor, effectively restored CEACAM6 expression and reversed EMT, providing further evidence of the intricate interplay between TGF-β signaling and epigenetic control of CEACAM6.

In summary, our study underscores the pivotal role of CEACAM6 in modulating EGFR signaling and its impact on EGFR-TKI resistance. We discovered that CEACAM6 enhances EGFR signaling by interacting directly with EGFR. Conversely, factors like TGF-β stimulation or epigenetic regulation decrease CEACAM6 expression, prompting a switch to EGFR-independent pathways and contributing to EGFR-TKI resistance in lung cancer cells. These findings emphasize the potential of CEACAM6 as a biomarker for identifying and understanding oncogenic EGFR-driven lung cancer. This understanding opens avenues for targeted therapeutic strategies to overcome EGFR-TKI resistance by focusing on CEACAM6 and its regulatory mechanisms within the EGFR signaling network.

Methods

Cell culture

HCC827 (RRID: CVCL_2063) cells were provided by Dr. Jeff Wang (Development Center for Biotechnology, Taipei, Taiwan). HCC827GR cells were developed from HCC827 (RRID: CVCL_2063) cells in our laboratory by treatment with incremental concentrations of gefitinib stepwise16. All cell lines were authenticated by STR profiling within the three years (Supplementary Table 1) and cultured in RPMI-1640 medium containing L-glutamine (4 mM), sodium pyruvate (1 mM), HEPES (10 mM), and FBS (10%). Cells were checked routinely and found to be free of contamination by Mycoplasma.

Chemicals and reagents

Gefitinib was obtained from Cayman Chem (Ann Arbor, MI). Recombinant human TGF-β and CEACAM6-Flag plasmid (HG10823-NF) were acquired from Sino Biological (Beijing, China). shRNA clones (shCEACAM6, TRCN0000062299; shSMAD4, TRCN0000040030) were ordered from the National RNAi Core Facility, Academia Sinica, Taiwan. RPMI-1640 and FBS were purchased from GIBCO-BRL (Grand Island, NY). Romidepsin (FK 228) was purchased from Med Chem Express (USA). TGF-β inhibitor (SB-431542) was obtained from Med Chem Express (USA).

Clonogenic assay

Cells were plated in 6-well cell culture plates (1000 cells/well) and incubated for 2 weeks at 37 °C with 5% CO2. Then, the medium was aspirated, and colonies were subjected to crystal violet staining (Thermo Fisher Scientific, Waltham, CA), followed by ImageJ software quantification.

Quantitative real-time PCR (qRT-PCR)

Relative gene expression was assessed using the 2–ΔΔCT method with specific primers and TaqMan probes or probes from the Universal Probe Library (Roche Diagnostics, Indianapolis, IN) in the StepOneTM Real-Time PCR system (Applied Biosystems Inc., Foster City, CA). Total RNA was extracted using REzol reagent (Protech Technology Enterprise CO., Ltd., Taiwan) according to the manufacturer’s protocol. 1 μg of total RNA was transferred to cDNA by PrimeScript RT-PCR Kit (TaKaRa, RR014B). Reaction protocols started with a 3 min 95 °C denaturation step, followed by 40 cycles consisting of 5 sec of denaturation at 95 °C, 20 sec of annealing at 60 °C, and 20 sec of extension at 72 °C. 18S rRNA was taken as a reference transcript. Primer sequences and probes used in qPCR are listed in Supplementary Table 2.

Co-immunoprecipitation (Co-IP) assay

The cells were lysed with a non-denaturing lysis buffer (20 mM Tris- HCl, pH 8.0, 137 mM NaCl, 1% Nonidet P-40, 2 mM EDTA) containing protease inhibitor and phosphatase inhibitor (Roche). 500 μg of whole cell lysates in 200 μL were incubated with 2 μL of anti-FLAG antibody (Abcam, ab205606) at 4 °C overnight on a rotator at a low speed. 40 μL of protein A/G sepharose beads (BioVision, Milpitas, CA) were added to the lysate-antibody mixture and incubated for 3 h at 4 °C on a rotator at a low speed. The immunocomplexes were washed with non-denaturing lysis buffer three times, resuspended in 40 μL of 2xSDS-sample buffer, and boiled for 10 minutes. 30 μL of final supernatants were loaded and separated by 10% SDS-PAGE.

Chromatin immunoprecipitation-qPCR (ChIP-qPCR)

5 ‘106 cells were collected for each sample. The cells were treated with 1% formaldehyde, and glycine was added (0.125 M). The cells were washed with an ice-cold phosphate-buffered saline solution, lysed in IP buffer (50 mM NaCl, 50 mM Tris-HCl, pH 7.5, 5 mM EDTA, 0.5% Nonidet P-40, 1% Triton X-100) and sonicated using Bioruptor (Diagenode, Denville, NJ). The sonicated lysate was precleared with protein A/G agarose beads (BioVision, Milpitas, CA). 2 μg of anti-H3K27ac antibody (Genetex, GTX60815) or normal mouse IgG (Genetex, GTX35009) was added to the precleared lysates and incubated overnight at 4 °C. Protein A/G agarose beads were added for immunoprecipitation and washed with IP buffer three times. 100 μL of 10% Chelex 100 slurry (Bio-Rad Laboratories, Hercules, CA) was added to each sample and boiled for 10 min. The slurry was centrifuged, and DNA from the supernatant was collected and analyzed by qPCR. Primer sequences and probes used in ChIP-qPCR are listed in Supplementary Table 3.

Immunoblotting

The cell lysate was harvested with RIPA buffer containing protease and phosphatase inhibitors (Roche). The concentration of protein was measured by the Bradford protein assay (Bio-Rad). Total protein lysate was loaded on a 10% SDS-PAGE and then transferred to the PVDF membrane. The membrane was incubated with primary antibodies at 4 °C overnight. The secondary antibody HRP-conjugated anti-rabbit was used according to the manufacturer’s recommendation (GeneTex, GTX213110-01). The protein band was visualized using ECL and imaged with the LAS-3000 system (FujiFilm, Tokyo, Japan). Primary antibodies included anti-CEACAM6 (Abcam, ab78029), anti-EGFR (4267S, Cell Signaling), anti-phospho-EGFR (Cell Signaling, 3777S), anti-phospho-ERK1/2 (Cell Signaling, 4370S), anti-phospho-AKT (Cell Signaling, 4060S), anti-CDH1 (GeneTex, GTX100443), anti-AXL (Cell Signaling, 8661S), anti-VIM (GeneTex, gtx100619), and anti-GAPDH (GeneTex, GTX100118).

Immunohistochemistry (IHC)

The IHC procedure was performed as previously described60. Antigen retrieval was performed by incubating the slide with 10 mM citrate buffer (pH 6.0) in an autoclave for 30 min at 121 °C. The primary antibody in IHC staining was the anti-CEACAM6 antibody (Abcam, ab275022). The chromogen reaction was performed by applying a diaminobenzidine tetrahydrochloride (DAB) solution (Dako, Santa Clara, CA). IHC images were quantified by H-score, calculated by intensity and percentage of positive cells. H-score = [(0 x % negative cells) + (1 x % weak positive cells) + (2 x % moderate positive cells) + (3 x % strong positive cells).

Enzyme-linked immunosorbent assay (ELISA)

5 × 105 cells were plated into the 6-cm plate and lysed with RIPA buffer after 72 hr of incubation. Levels of CEACAM5 and CEACAM6 in cell lysates were evaluated by the enzyme-linked immunosorbent assay with CEACAM5 (General Biological, 4CEA3) and CEACAM6 (Sino Biological, KIT10823) kits.

Tissue samples and public Domain Data Analysis

The human lung adenocarcinoma sample was collected from surgery with the approval of the Joint Institutional Review Board of the Tri-Service General Hospital (Approval number: 1-107-05-102), and written informed consent was obtained from the participant prior to sample collection, in accordance with the Declaration of Helsinki. We also procured the CC5 tissue array from SuperBioChips Laboratories. The intravenous xenograft human lung tumor was obtained as previously described60. The gene expression profiling dataset for CEACAM6 knockdown in HCC827 cells is publicly available in the Gene Expression Omnibus (GEO) under accession number GSE289000. The public gene expression profiling datasets used in this study were analyzed as described previously61. The sources of these gene expression profiling data sets are listed in Supplementary Table 4.

Statistical analysis

Three duplicates were performed in all experiments. Results were presented as the mean with the standard error of the mean. Data were analyzed using the Student’s t-test or Mann-Whitney U analysis. A clustering analysis was performed using Heatmapper62. p values < 0.05 was taken as a statistically significant difference. All statistical analyses were performed using Prism 5 software, version 5.01 (GraphPad Software).