EpCAM Signaling Promotes Tumor Progression and Protein Stability of PD-L1 through the EGFR Pathway

Epithelial cell adhesion molecule (EpCAM) is the most frequently expressed tumor-associated antigen; it is overexpressed in the majority of human adenocarcinomas and squamous cell carcinomas and is associated with poor prognosis in patients (1–4). EpCAM is sequentially processed by a disintegrin and metalloproteinase 17 (ADAM17), also called TNFα converting enzyme (TACE), and γ-secretase (5, 6). Processing by these enzymes, respectively, releases the extracellular domain (EpEX) and intracellular domain (EpICD). After release, EpICD interacts with four and a half LIM domains protein 2 (FHL2) and β-catenin to form a complex that translocates to the nucleus and interacts with Lef-1, which binds to DNA (5, 6). The EpICD complex promotes tumorigenesis of tumor-initiating cells through the upregulation of reprogramming genes and epithelial–mesenchymal transition (EMT; ref. 7). In addition, the increased release of EpEX enhances EpICD generation and upregulates the expression of reprogramming and EMT genes (5, 7–9). It was previously reported that EpEX could directly bind to EGFR and stimulate EGFR phosphorylation and its downstream signaling pathway (5, 8, 10). Moreover, EpEX-induced EGFR phosphorylation can activate ADAM17 and γ-secretase to further increase the shedding of EpEX and EpICD (5, 8, 10). However, the domain of EpEX that binds and activates EGFR has not been previously identified.

AKT is a primary mediator of EGFR signaling that promotes cell survival partially by inactivating proapoptotic proteins. One such proapoptotic protein that is inactivated by AKT phosphorylation is forkhead transcription factor O3a (FOXO3a; ref. 11). FOXO3a is also referred to as FKHRL-1 and is a member of the forkhead transcription factor family (12). Because genes activated by FOXO proteins generally function to limit cell growth and promote death, this family is thought of as tumor suppressors. When activated, FOXO3a accumulates in the nucleus, where it enhances the transcription of various genes involved in apoptosis and the cell-cycle control, such as BIM, FasL, and p21 (13). Furthermore, AKT inactivation is an essential step in anoikis, and FOXO3a regulation by the PI3K/AKT pathway may be essential for this inactivation (14, 15). Thus, AKT inactivation has been suggested to be an essential step in apoptosis (11).

In addition to its mediation of survival signals, EGFR activation is crucial for triggering immune escape (16). EGFR-activating mutations were reported to be associated with increased PD-L1 expression in mouse models of EGFR-driven lung cancer and bronchial epithelial cells with the expression of mutant EGFR (16). Moreover, EGFR inhibitors can reduce PD-L1 expression in non–small cell lung cancer (NSCLC) cell lines with activated EGFR, suggesting that EGFR signaling may trigger immune escape (16). Further investigations demonstrated that EGFR ligands, such as EGF, induce PD-L1 expression primarily at the level of posttranslational modifications. EGF was shown to stabilize PD-L1 by inducing PD-L1 glycosylation, which prevents GSK3-dependent proteasomal degradation of PD-L1 by β-TrCP (17). Moreover, the MAPK signaling pathway, another important axis of EGFR signaling, is associated with PD-L1 mRNA expression. It was reported that EGFR activation increases PD-L1 expression through p-ERK1/2/p-c-Jun (18–20). In addition to promoting PD-L1 transcription, MAPK signaling also stabilizes PD-L1 mRNA by attenuating TPP activity (21). Because EpCAM is an activator of EGFR signaling, EpCAM might promote escape from immune surveillance. However, EpCAM-mediated immunosuppression and its mechanisms are almost completely undescribed.

In this study, we report that the EGF-like domain I within EpEX binds to EGFR and activates both the AKT and ERK1/2 pathways. EpEX-induced AKT activation inhibits FOXO3a activity, which decreases high temperature requirement A2 (HtrA2) transcription. Moreover, a therapeutic antibody directed against EpCAM (EpAb2-6) promotes FOXO3a nuclear translocation, and increases HtrA2 expression. We also found that EpCAM inhibits antitumor immunity by activating the PD-1/PD-L1 pathway to suppress T-cell function. EpEX increases PD-L1 protein stability rather than mRNA level, while blockade of EGFR or MAPK signaling pathways attenuates EpEX-mediated stabilization of PD-L1 protein. Furthermore, treatment of EpAb2-6 markedly prolongs the survival of mice in metastasis and in orthotopic models of human colorectal cancer, and a combination of atezolizumab and EpAb2-6 shows enhanced therapeutic efficacy and high levels of tumor-localized CD8⁺ T cells in a peripheral blood mononuclear cell (PBMC) cell line–derived xenograft (CDX) mouse model. Our findings not only shed light on the molecular mechanisms underlying EpCAM signaling in cancer malignancy, but also suggest that therapeutic targeting of EpCAM may work well in combination with current immunotherapies.

Antibodies against human EpCAM and p84 were purchased from Abcam, and antibody against β-catenin was from Santa Cruz Biotechnology. Anti-α-tubulin antibody was from Sigma-Aldrich. Polyclonal antibodies detecting total ERK and Thr202/Tyr204-phosphorylated ERK, total AKT and Ser473-phosphorylated AKT, total FOXO3a, Ser253-phosphorylated FOXO3a, Ser318/321-phosphorylated FOXO3a, and The32-phosphorylated FOXO3a, and active (non-phospho Ser33/Ser37/Thr41) β-catenin, HtrA2, cytochrome c, COX IV, XIAP, and PD-L1 were from Cell Signaling Technology. U0126 (MEK inhibitor), wortmannin (PI3K inhibitor), and MG132 (proteasome inhibitor) were obtained from Selleck Chemicals.

Human lung adenocarcinoma cells (H441), lung carcinoma cells (H460), embryonic kidney cells (HEK293T), colorectal carcinoma cells (HCT116), and colorectal adenocarcinoma cells (SW620) were obtained from the ATCC. H441 and H460 cells were cultured in RPMI1640 Medium (Gibco), and HEK293T, HCT116, and SW620 cells were cultured in DMEM (Gibco). All cells were maintained in conditioned medium supplemented with 10% FBS (Gibco) and 100 μg/mL penicillin–streptomycin (Gibco) at 37°C in a humidified incubator with 5% CO₂.

Cells were lysed with RIPA buffer (20 mmol/L Tris-HCl, pH 7.4, 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L EGTA, 1% Nonidet P-40, 1% sodium deoxycholate, and 2.5 mmol/L sodium pyrophosphate) containing Protease Inhibitor (Roche) and Phosphatase Inhibitor (Roche) to extract whole-cell lysates. The lysates were then quantified using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). The lysates were mixed with 5× sample buffer (50 mmol/L Tris-HCl, pH 6.8, 2% SDS, β-mercaptoethanol, 0.1% bromophenol blue, and 10% glycerol), separated by SDS-PAGE, and transferred to polyvinylidene difluoride (PVDF) Membrane (Millipore). Nonspecific antibody-binding sites on the PVDF membrane were blocked with 3% BSA in TBST (TBS buffer containing 0.1% Tween 20) and membranes were incubated with indicated antibody overnight at 4°C, followed by incubation with horseradish peroxidase–conjugated secondary antibodies (Jackson ImmunoResearch Laboratories) at room temperature for 1 hour. The protein bands were subsequently visualized with Chemiluminescence Reagents (Millipore) and detected by UVP BioSpectrum 600 Imagining System (UVP). The protein expression was quantified by Gel-Pro Analyzer 3.1 (Media Cybernetics).

The DNA fragment encoding EpEX (amino acids 24–262 of EpCAM) was amplified by PCR using PfuTurbo DNA polymerase and cloned into pSecTag2 vector with C-terminal 6xHis tag to generate pSecTag2-EpEX-6xHis. The EpEX-6xHis fusion protein was produced with Expi293F Expression System (Thermo Fisher Scientific) and purified by Ni-Affinity Column (GE Healthcare).

HCT116 cells were harvested with 10 mmol/L EDTA in PBS, and incubated with EpEX-Fc for 1 hour at 4°C. After incubation, 2 mmol/L DTSSP (Thermo Fisher Scientific) was used as a cross-linker to stabilize the interaction between EpEX-Fc and EGFR. To stop the cross-linking reaction, Tris, pH 7.5, was added to a final concentration of 20 mmol/L. Membrane proteins were extracted using Mem-PER Eukaryotic Membrane Protein Extraction Reagent Kit (Thermo Fisher Scientific). Finally, the EpEX-Fc–EGFR complex was pulled down with Dynabeads Protein G (Invitrogen), and probed by Western blotting.

In its EpEX, EpCAM contains two EGF-like domains, at amino acids 27–59 (first EGF-like domain) and 66–135 (second EGF-like domain), and a cysteine-free motif (22). The EpCAM EGF-like domain deletion was generated using the standard QuikChange deletion mutation system with first forward mutagenic deletion primer (5′-GCAGCTCAGGAAGAATCAAAGCTGGCTGCC-3′) and first reverse mutagenic deletion primer (5′-GGCAGCCAGCTTTGATTCTTCCTGAGCTGC-3′); and second forward primer (5′-AAGCTGGCTGCCAAATCTGAGCGAGTGAGA-3′) and second reverse primer (5′-TCTCACTCGCTCAGATTTGGCAGCCAGCTT-3′). The PCR amplifications were performed using KAPA HiFi Hot Start DNA Polymerase (Kapa Biosystems), and products were treated with restriction enzyme, DpnI (Thermo Fisher Scientific), to digest methylated parental DNAs.

Cycloheximide, a protein synthesis inhibitor, was used to evaluate the stability of PD-L1. Cells were treated with cycloheximide for 0, 2, 4, or 6 hours. Proteins were extracted and Western blot analysis was performed to detect PD-L1 protein level.

Cells were lysed in lysis buffer (50 mmol/L Tris-HCl, pH 7.4, 150 mmol/L NaCl, and 1% NP-40) with Protease Inhibitors (Roche). For immunoprecipitation, cell lysates were incubated with antibodies for 6 hours at 4°C. Then, 20 μL Dynabeads Protein G was added and the mixture was incubated for 2 hours at 4°C to pull-down the antibody-bound protein. The immunoprecipitation samples were washed with PBS three times, denatured in sample buffer, and analyzed by Western blotting.

Generation of EpAb2-6 and control IgG was performed as described previously (23). The protocol was approved by the Committee on the Ethics of Animal Experiments of Academia Sinica [Nankang, Taipei, Taiwan, AS Institutional Animal Care and Use Committee (IACUC): 11-04-166].

All lentiviral short hairpin RNA (shRNA) constructs were purchased from the RNAi Core Facility at Academia Sinica (Nankang, Taipei, Taiwan). Lentiviral production, infection, and selection were performed according to protocols from the RNAi Consortium (Academia Sinica, Nankang, Taipei, Taiwan). To produce the lentivirus, HEK293T cells were transiently cotransfected with shRNA plasmid, packaging (pCMV-ΔR8.91) and envelope (pMD.G) expression plasmids, using PolyJet DNA Transfection Reagent (SignaGen Laboratories). The next day, the transfection solution was replaced with 1% BSA containing medium to improve virus yield. Two days after transfection, medium containing lentivirus was collected. Cells were cultured in lentivirus-containing medium, supplemented with 8 μg/mL polybrene, for another 48 hours. The transduced cells were selected with 2 μg/mL puromycin for 4 days and the knockdown efficiency was measured by Western blotting. The target sequence for human EpCAM–specific shRNA was shRNA 5′-GCAAATGGACACAAATTAC AA-3′. Luciferase shRNA (shLuc) was used as a negative control.

Total RNA extraction, first-strand cDNA synthesis, and SYBR Green–based real-time PCR were performed as described in the manufacturer's instructions. To extract total RNA, cells were lysed using TRIzol Reagent (Invitrogen), and proteins and phenol were removed from TRIzol using chloroform. After centrifugation, the top colorless layer was collected and mixed with isopropanol to precipitate RNA pellet. Then, the RNA pellet was washed with 70% ethanol, air-dried at room temperature, and dissolved in RNase-free water. For first-strand cDNA synthesis, 5 μg of total RNA was used for reverse transcription with oligo(dT) primer and SuperScriptIII Reverse Transcriptase (Invitrogen) at 50°C for 60 minutes. Target gene levels were evaluated by quantitative PCR (qPCR), using LightCycler 480 SYBR Green I Master Mix (Roche) and a LightCycler480 System (Roche). GAPDH mRNA expression was measured as endogenous housekeeping control to normalize all qPCR reactions. The qPCR reaction was as follows: 95°C for 5 minutes, followed by 40 cycles of denaturation at 95°C for 10 seconds, annealing at 60°C for 10 seconds, and extension at 72°C for 30 seconds. Final results were calculated from three independent experiments. Primer sequences used to detect the mRNA expression of genes of interest are listed in Supplementary Table S1.

For immunofluorescence, 2 × 10⁴ cells were seeded on 12-mm cover slips in 24-well culture plate, fixed with 2% paraformaldehyde, and incubated with 0.1% Triton X-100 to increase the permeability of cell membrane. Then, samples were blocked with 1% BSA for 1 hour at room temperature and stained with primary antibodies at 4°C overnight. Slides were washed and stained with FITC or Alexa568-conjugated secondary antibodies and DAPI for 40 minutes. Cover slips were then washed with PBS and mounted in Mounting Solution (Vector Laboratories). The slides were examined with a confocal microscope (Leica TCS-SP5).

HCT116 cells were stimulated with 20 μg/mL EpAb2-6 for 6 hours, and apoptosis arrays were performed according to the manufacturer's instructions (R&D System; ARY009). The membranes (arrays) were detected by the UVP BioSpectrum 600 Imagining System (UVP). The arrays were quantified by a Gel-Pro Analyzer 3.1 (Media Cybernetics, Inc.).

Cells were harvested, washed, and suspended in FACS buffer (PBS with 1% FBS), and 10⁵ cells were transferred to 96-well shrill-based plates. Cells were stained with different antibodies for 1 hour at 4°C, and then subsequently stained with indicated PE-conjugated secondary antibodies for 1 hour at 4°C. Then, cells were washed with FACS buffer twice and suspended in 400 μL FACS buffer. Florescence signals were analyzed using flow cytometry (BD FASCSC AntoII) and measured with FCS Express V3 software. The data were collected from three independent experiments.

In brief, HCT116 cells (1 × 10⁵) were treated with 20 μg/mL EpAb2-6 for 6 hours, followed by cross-linking fixation in 1% formaldehyde. Fixation was quenched by the addition of glycine to a final concentration of 200 mmol/L and fixed chromatin complexes were then sonicated to an average length of 250 bp using an MISONIX Sonicator 3000. The sonicated protein–DNA complexes were subjected to immunoprecipitation using 2 μg of antibodies against FOXO3a. The immunoprecipitated DNA was recovered by a PCR Purification Kit (Qiagen), and the amount of target DNA was detected by PCR using the primers complementary to the sequence listed in Supplementary Table S2.

The human HtrA2 proximal promoter fragments covering the regions −1628 to +86 and −700 to +86 (with the transcriptional start site denoted as +1) were PCR amplified and then cloned into the firefly luciferase reporter plasmid pGL4.18 Vector (Promega). To assess the effect of EpA2-6 on HtrA2 promoter activity, HCT116 and SW620 cells (1 × 10⁴/well), seeded onto 24-well dishes, were transiently transfected with HtrA2 promoter reporter constructs in combination with a plasmid expressing Renilla luciferase using the PolyJet Transfection Reagent (SignaGen Laboratories) for 24 hours, followed by 20 μg/mL EpAb2-6 for 6-hour stimulation. Cell lysates were prepared and subjected to the luciferase activity assay using the Dual-Luciferase Reporter Assay Kit (Promega). Firefly luciferase activity was normalized to Renilla luciferase activity, and final data were presented as the fold induction of luciferase activity compared with EpAb2-6-IgG controls. Data are expressed as means ± SD from three independent experiments. Primers used to amplify the human HtrA2 proximal promoter fragment are listed in Supplementary Table S3.

Cells (2 × 10⁵) were seeded onto 24-well dishes and treated with 20 μg/mL control IgG or EpAb2-6 for 6 hours. The apoptotic cell and mitochondrial membrane potentials were detected using FITC Annexin V Apoptosis Detection Kit (BD Biosciences) and MitoStatus Red (BD Biosciences), respectively. The apoptotic cell and mitochondrial membrane potentials were analyzed using a Flow Cytometer (Thermo Fisher Scientific). Each measurement was carried out at least three times to ensure reproducibility. The effect of gene knockdown on EpAb2-6–induced apoptosis is represented in Supplementary Table S4.

Human lung cancer tissue microarray was purchased from Super BioChip. After exhausting the endogenous hydroperoxidases with 3% hydrogen peroxide in methanol for 30 minutes, the tissue microarray was blocked with 1% BSA for 1 hour and exposed to anti-PD-L1 (Clone 28-2, Abcam) or EpAb3-5 (developed by our laboratory as described previously; ref. 23) at 4°C overnight. After washing with PBST, the tissue microarray was processed using the standard procedure of Super Sensitive IHC Detection System (Bio-Genex) and stained by DAB. Nuclei were stained with Mayer's hematoxylin solution (Wako).

For the experiments using PBMCs, blood samples were drawn from healthy donors and collected into 10 mL Vacutainer tubes containing the anticoagulant EDTA (BD Bioscience). After centrifugation at 1,500 rpm for 10 minutes, plasma was removed from the sample and mixed with an equal volume of PBS. The mixture was layered on Ficoll-Paque plus (Ficoll:blood = 1:2) and centrifuged at 1,500 rpm for 30 minutes. The PBMC layer (buffy coat) was collected and washed with PBS with 0.5% BSA and 2 mmol/L EDTA twice. To obtain purified CD3⁺ T cells, PBMCs were positively selected by anti-CD3 magnetic beads (MACS), according to the standard protocol. Isolated CD3⁺ T cells (1 × 10⁶) were cultured and activated by 25 μL anti-CD3/anti-CD28–coated Dynabeads (Invitrogen) in RPMI1640 medium supplemented with 10% FBS, 100 μg/mL penicillin–streptomycin, 12.5 ng/mL IL2 (Gibco), and 1 ng/mL IL15 (MACS) for 48 hours. The protocol was approved by the Institutional Review Board (IRB) of Academia Sinica (Nankang, Taipei, Taiwan, AS IRB: AS-IRB01-19049).

NOD/SCID mice were intravenously injected with HCT116-GFP cells; mice bearing circulating HCT116-GFP cells were intravenously treated with EpAb2-6 or an equivalent dosage of control IgG at 1 hour after cell injection (antibody was delivered at 20 mg/kg). Then, blood samples were obtained from the facial vein of mice and the fluorescence intensity of whole blood was quantified at the indicated timepoints. The fluorescence was measured with a Microplate Reader (Molecular Devices, SpectraMax M5) at an excitation wavelength of 355 nm and emission wavelength of 440 nm.

Subcutaneous studies were performed as reported previously (23). The protocol was approved by the Committee on the Ethics of Animal Experiments of Academia Sinica (Nankang, Taipei, Taiwan, AS IACUC: 11-04-166).

Orthotopic studies were performed as reported previously (24). Briefly, NSG mice (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) were used for orthotopic implantation of HCT116 cells, which were previously infected with Lenti-luc virus (lentivirus containing luciferase genes). The mice were anesthetized by intraperitoneal injection of avertin, 2,2,2-Tribromo-Ethanol (Sigma-Aldrich) at a dose of 250 mg/kg. Tumor development was monitored by bioluminescence imaging. For the orthotopic therapeutic study, tumor-bearing mice were treated with control IgG or EpAb2-6 (20 mg/kg). Tumor progression was monitored by quantification of bioluminescence. Mouse body weight and survival rate were measured. Animal care was carried out in accordance with the guidelines of Academia Sinica (Nankang, Taipei, Taiwan). The protocol was approved by the Committee on the Ethics of Animal Experiments of Academia Sinica (Nankang, Taipei, Taiwan, AS IACUC: 11-04-166).

The gene expression matrix was obtained from LUAD projects in The Cancer Genome Atlas. The 25% of samples with highest EpCAM expression were defined as the “EpCAM_High” group and the 25% samples with least EpCAM expression were defined as the “EpCAM_Low” group. T-cell activation and proliferation gene sets provided by gene set enrichment analysis (GSEA) website were then used to analyze the data.

All data are represented as means ± SD from at least three independent experiments. Significant differences from the respective control for each experimental condition were calculated using Student t test, unless otherwise specified. *, P < 0.05; **, P < 0.01; or ***, P < 0.001 are indicated as significant. Survival analysis was performed using a log-rank test. The correlation coefficient was calculated using Spearman analysis.

Previously, we reported that EpEX can bind to EGFR (10), and functions as a growth factor and induces EGFR phosphorylation (5, 10). To verify that EpEX interacts with EGFR, we used the cross-linker, DTSSP, to stabilize the EpEX–EGFR complex. The binding of EpEX to EGFR was confirmed by immunoprecipitation and Western blotting (Fig. 1A). To evaluate the direct binding of recombinant EpEX to the EpEX of EGFR (EGFR_ECD), we performed ELISA to probe the direct interaction between purified EpEX and EGFR_ECD protein (Fig. 1B).

To test whether membrane-bound EpCAM could bind to EGFR, we performed immunoprecipitation experiments using HEK293T cells with EpCAM-V5 and EGFR-Flag overexpression. The interaction between exogenous EpCAM and EGFR was detected by coimmunoprecipitation (Fig. 1C). In EpCAM EpEX, it contains two EGF-like domains, at amino acids 27–59 (first EGF-like domain) and 66–135 (second EGF-like domain), and a cysteine-free motif (22). To identify the specific region of EpEX that binds to EGFR, we constructed different EGF-like domain–deleted EpCAM mutants. Surprisingly, the EGF-like domain I–deleted EpCAM mutant (EpCAM_ΔEGFI) showed decreased binding to EGFR, while the EGF-like domain II–deleted EpCAM mutant (EpCAM_ΔEGFII) exhibited an enhanced interaction with EGFR (Fig. 1D). To further evaluate the binding of soluble EpEX to EGFR_ECD, HEK293T cells were cotransfected with different vectors expressing soluble EGFR_ECD-Flag or EpEX-Fc, and the protein complex was examined in culture media (Fig. 1E). Soluble EGF-like domain II–deleted EpEX-Fc (EpEX_ΔEGFII-Fc) showed significantly increased affinity, while EGF-like domain I–deleted EpEX-Fc (EpEX_ΔEGFI-Fc) exhibited decreased affinity for EGFR_ECD-Flag compared with EpEX-Fc (Fig. 1F). Overall, we found that the membrane-bound EpCAM and secreted EpEX both bind EGFR and deletion of the EGF domain I diminished this binding (Fig. 1C–F).

Similar results were observed using purified wild-type or mutant EpEX and EGFR_ECD recombinant proteins. The recombinant EpEX_ΔEGFII protein had a stronger binding affinity for EGFR_ECD than wild-type controls, and the EpEX_ΔEGFI protein lost its ability to bind EGFR_ECD (Fig. 1G). Moreover, the EpEX_ΔEGFII protein induced EGFR signaling like the wild-type control, but EpEX_ΔEGFI protein did not (Fig. 1H). These results suggest that the EGF-like domain I in EpEX is the major domain responsible for the EpEX–EGFR interaction and activation of EGFR signaling.

Because we found that EpEX activates EGFR through EpEX domain I, we further investigated whether EpEX-mediated EGFR activation is important for cancer progression. AKT is a major downstream effector of EGFR, which promotes cell survival partially by inactivating FOXO3a (12). Inhibition of AKT/FOXO3a signaling has been suggested an essential step in apoptosis (11) and could inhibit tumorsphere formation of the SKOV3 ovarian cancer cell line (25). To evaluate whether EpEX affects FOXO3a function, we examined FOXO3a phosphorylation and intracellular location. Immunoblotting analysis demonstrated that EpEX induced the phosphorylation of AKT and FOXO3a in a time-dependent manner (Fig. 2A), and EpEX-induced phosphorylation of FOXO3a was inhibited by AG1478 (EGFR inhibitor) or wortmannin (PI3K inhibitor; Fig. 2B). Moreover, after EpEX treatment, FOXO3a translocated from the nucleus to the cytoplasm (Fig. 2C). Because the EpEX acted through EGFR, we used EGF, a well-known ligand for EGFR, as an inducer. Similar to EpEX treatment, EGF induced nuclear exclusion of FOXO3a (Supplementary Fig. S1A–S1C). Therefore, we conclude that EpEX inhibits the FOXO3a activity through the EGFR–AKT pathway.

Previously, we had developed a neutralizing antibody, EpAb2-6, which targets EpEX and induces apoptosis (5, 23). EpAb2-6 was employed to block the EpEX. By blocking the function of EpEX, EpAb2-6 treatment is known to interrupt the EpEX/EGFR/ADAM17 axis, which represents a positive feedback loop promoting EpCAM cleavage and subsequently increases EpEX and EpICD production (5). Indeed, decreases in ADAM17 and γ-secretase activity and soluble EpEX release were observed after EpAb2-6 treatment (Supplementary Fig. S2A–S2C). Moreover, nuclear-activated β-catenin and downstream genes, including reprogramming genes and EMT-related genes, were also decreased in EpAb2-6–treated cells (Supplementary Fig. S2D and S2E). Next, we examined whether EpCAM contributes to anchorage-independent growth. For this purpose, a soft agar colony formation assay was performed with colon cancer cell lines. EpAb2-6–treated cells formed significantly smaller and fewer colonies compared with the control cells (Supplementary Fig. S2F). We further cultivated these attached colon cancer cells into tumorspheres and found that EpAb2-6 treatment inhibited tumorsphere formation (Supplementary Fig. S2G). These results suggest that targeting EpCAM with a specific antibody, EpAb2-6, is a potentially viable strategy to suppress colon cancer malignancy.

It has been previously reported that the nuclear accumulation of FOXO3a promotes cancer cell apoptosis after treatment with various chemotherapy drugs or EGFR inhibitors (26–28). Thus, we wanted to clarify whether inhibiting the function of EpEX facilitates nuclear accumulation of FOXO3a. Blocking the function of EpEX with EpAb2-6 treatment also decreased the phosphorylation of AKT and FOXO3a (Fig. 2D) and subsequently increased nuclear accumulation of FOXO3a, while diminishing the nuclear translocation of β-catenin (Fig. 2E). The downstream genes of FOXO3a (BIM, p21, and FASL) were also upregulated in HCT116 cells with EpAb2-6 treatment compared with IgG control (Fig. 2F). Using immunofluorescence staining, FOXO3a and β-catenin were detected in both the cytoplasm and nucleus of colon cancer cells. However, after EpAb2-6 treatment, FOXO3a was increased in the nucleus, but nuclear β-catenin was decreased both in vitro (Fig. 2G) and in vivo (Fig. 2H). Thus, inhibiting EpEX enhances the nuclear translocation of FOXO3a and consequently upregulates downstream genes, which are involved in promoting apoptosis.

To evaluate whether EpAb2-6 induces apoptosis via EGFR/AKT/FOXO3a, we used shRNA to knockdown EGFR, AKT, and FOXO3a expression in colon cancer cells and examined the effect of EpAb2-6 on cancer cell apoptosis. EGFR–, AKT–, and FOXO3a-knockdown cells were all less sensitive to EpAb2-6–induced apoptosis than control cells, and EpAb2-6–induced apoptosis was completely abrogated in EpCAM-knockout cells (Supplementary Fig. S3A–S3D). Together, our data show that EpEX induces phosphorylation of FOXO3a and inactivates its biological function; meanwhile, blocking the function of EpEX inhibits the phosphorylation of FOXO3a to promote nuclear accumulation of FOXO3a and activates expression of its downstream proapoptotic genes.

To gain further insight into the mechanisms of EpAb2-6–induced apoptosis, we compared the expression of 35 apoptosis-related signaling proteins in control IgG- and EpAb2-6–treated colon cancer cells. We found that among the 35 apoptosis-related proteins, two showed substantial increases in expression after EpAb2-6 treatment compared with the IgG control. These proteins were HtrA2 and X-linked inhibitor of apoptosis protein (XIAP; Fig. 3A).

Immunoblotting and qPCR experiments showed that EpAb2-6 treatment enhanced the protein and mRNA expression of HtrA2 in HCT116 cells (Fig. 3B). According to previous studies, HtrA2 can be released into the cytosol, where it contributes to apoptosis (29). As expected, we found that EpAb2-6 induced the release of cytochrome c and HtrA2 from mitochondria into the cytosol (Fig. 3C). Furthermore, EpAb2-6 treatment induced the dissipation of the mitochondrial membrane potential (Fig. 3D).

HtrA2 mRNA levels were increased after EpAb2-6 treatment, implying transcriptional regulation of HtrA2 gene was induced by EpAb2-6. To examine this hypothesis, HCT116 cells were transiently transfected with a reporter for the HtrA2 promoter; a region encompassing 1,704 bases upstream of the HtrA2 translational start site was used to drive the expression of the firefly luciferase gene (pGL4.18-HtrA2-1; Fig. 3E). EpAb2-6 treatment led to an approximate 4-fold induction in HtrA2 promoter activity, compared with the IgG control, indicating that the HtrA2 gene is indeed a transcriptional target of EpAb2-6–induced signaling (Fig. 3E). To further elucidate the molecular mechanism controlling HtrA2 gene transcription, the HtrA2 promoter sequence was analyzed using the PROMO virtual laboratory website. This analysis uncovered putative cis-acting response elements (−1283 to −1299) for FOXO transcription factors. Interestingly, to our knowledge, there is no previously published evidence showing that FOXO3a controls HtrA2 promoter activity. Thus, we examined whether HtrA2 promoter activity was regulated by FOXO3a by generating a reporter construct without the FOXO3a response element (pGL4.18-HtrA2-2), as depicted in Fig. 3E. Indeed, the EpAb2-6–induced increase in luciferase activity was abolished in the absence of a FOXO3a response element (Fig. 3E).

To further elucidate whether FOXO3a directly regulates the HtrA2 promoter, chromatin immunoprecipitation (ChIP) analysis was performed using primers flanking the putative FOXO3a response elements (−1283 to −1299). FOXO3a occupied the HtrA2 promoter after EpAb2-6 treatment, which was not observed in FOXO3a-knockdown cells (Fig. 3F). These results indicate that FOXO3a increased HtrA2 transcription through direct binding to the HtrA2 promoter. We next investigated whether HtrA2 plays an important role in EpAb2-6–induced cell death. In this experiment, we used shRNA to knockdown HtrA2 expression in colon cancer cells. In HtrA2-knockdown cells, EpAb2-6 treatment induced fewer apoptotic cells (Fig. 3G) and less disruption of mitochondrial membrane potential (Fig. 3H) than shLuc (control knockdown) cells. Hence, FOXO3a-mediated HtrA2 transcription appears to be involved in EpAb2-6–induced apoptosis.

EGFR activation was reportedly associated with increased PD-L1 expression in mouse models of EGFR-driven lung cancer and bronchial epithelial cells with mutant EGFR expression, suggesting that EGFR signaling is crucial for triggering immune escape (16). Further investigations then demonstrated that increased EGFR signaling upregulates PD-L1 expression by various mechanisms (17, 19, 21). Moreover, EpCAM plays a crucial role in EGFR activation (5, 8, 10). On the basis of these findings, we decided to investigate whether EpCAM regulates PD-L1 expression via EGFR activation. Because HCT116 cells expression of PD-L1 is extremely low, H441 cells with high expression of PD-L1 were substituted for HCT116 cells as an experimental model for further studies.

First, we wanted to clarify whether EpCAM is involved in escape from immune surveillance. Immunodeficient NSG mice carrying shLuc H441 xenografts in the left side and shEpCAM H441 xenografts in the right side were intravenously injected with PBMCs to reconstitute immune cells. EpCAM-knockdown cells exhibited decreased tumor weight, confirming that EpCAM promotes tumor progression (Fig. 4A and B). Interestingly, shLuc tumors in mice with PBMC injection were not different from shLuc tumors in mice without PBMCs. However, shEpCAM tumors with PBMCs were smaller than those without PBMCs, implying that EpCAM might inhibit antitumor immunity (Fig. 4A and B). By flow cytometry analysis, we found that CD8⁺ T cells were enriched in shEpCAM tumor tissue (Fig. 4C), and Western blot analysis demonstrated that PD-L1 expression was diminished in shEpCAM tumor tissue (Fig. 4D). These results indicate that EpCAM promotes PD-L1 expression and reduces tumor-associated CD8⁺ T cells in vivo.

To investigate the correlation between EpCAM and PD-L1 expression, we analyzed PD-L1 mRNA expression in two NSCLC cell lines, H441 (high EpCAM expression) and H460 (no EpCAM expression). Compared with H441 cells, H460 cells exhibited a higher level of PD-L1 mRNA, but a lower level of PD-L1 protein (Fig. 4E). Flow cytometry analysis of PD-L1 and EpCAM confirmed this finding (Fig. 4F). Because many previous studies had indicated that the protein stability of PD-L1 plays an important role in its expression (17, 30–32), and, on the basis of our results, we suspected that stability of PD-L1 might serve as a key regulator of protein level. Indeed, a cycloheximide chase assay indicated that the half-life of PD-L1 in H460 cells was reduced compared with that in H441 cells (Fig. 4G). These findings suggest that the low PD-L1 protein level in H460 cells results from reduced stability of PD-L1 protein.

To investigate whether EpCAM influences PD-L1 protein stability, we overexpressed EpCAM in H460 cells (low endogenous PD-L1 expression and no EpCAM expression). We found EpCAM expression increased PD-L1 protein level, but did not change its mRNA level (Supplementary Fig. S4A and S4B). A cycloheximide chase assay confirmed that PD-L1 protein half-life was increased by EpCAM overexpression (Supplementary Fig. S4C). In addition, knockdown of EpCAM in H441 cells (high endogenous EpCAM expression) decreased PD-L1 protein level and protein half-life, but did not change mRNA level (Fig. 4H and I). Consistent with this idea, H441 cells with cytomegalovirus promoter–driven overexpression of PD-L1-Myc, which lack endogenous regulatory elements, such as the PD-L1 promoter, 3′ untranslated region (UTR), and 5′UTR, also exhibited decreased PD-L1 protein when EpCAM was knocked down (Fig. 4J). Furthermore, to validate our findings in samples from human patients with cancer, PD-L1 and EpCAM protein levels were examined in both lung and colon tumor specimens. Similar to our results in lung cancer cell lines, the PD-L1 protein level was highly correlated with EpCAM expression in a lung tumor tissue array (Fig. 4K). On the other hand, most of the specimens in the colon tissue array showed only weak staining for PD-L1, which prevented us from making a clear conclusion, even though the signal for EpCAM was strong (Supplementary Fig. S5A and S5B). Furthermore, GSEA indicated the related genes of T-cell activation and proliferation enrichment in low EpCAM expression of lung cancer specimen (Fig. 4L). Accordingly, we concluded that EpCAM expression is positively correlated to PD-L1 protein stability.

Regulated intramembrane proteolysis triggers EpCAM-mediated signal transduction through the dual actions of EpEX shedding by ADAM17 and EpICD release by presenilin 2–containing γ-secretase complex (6). On the basis of the previous studies, EpEX is known to function as an EGF-like growth factor, triggering EGFR signaling (5, 8, 10). Moreover, activation of the EGFR signaling pathway can increase PD-L1 expression through various mechanisms (17, 19, 21). Thus, we suspected that EpEX might regulate PD-L1 expression through EGFR signaling. Moreover, EpICD associates with FHL2 and β-catenin in a complex that translocates to the nucleus and transcribes downstream genes. Thus, we tested whether EpEX or EpICD is sufficient to cause PD-L1 stabilization. To test whether endogenous EpEX or EpICD is necessary for PD-L1 protein stabilization, ADAM17 inhibitor (TAPI) and γ-secretase inhibitor (DAPT) were utilized to prevent the generation of endogenous EpEX and EpICD, respectively. Blocking the shedding of endogenous EpEX, but not EpICD, caused a reduction in PD-L1 protein level. Notably, no changes in PD-L1 mRNA levels were observed, suggesting that endogenous EpEX was crucial for PD-L1 protein stability (Fig. 5A). Moreover, our results show that PD-L1 protein level was increased in a dose- and time-dependent manner after H441 cells were treated with EpEX or EGF (Fig. 5B and C). PD-L1 mRNA levels did not increase after EpEX or EGF treatment (Fig. 5C). These results suggest that EpEX can increase PD-L1 protein level, but EpICD cannot.

IFNγ is known to induce PD-L1 expression through STAT family–mediated transcription (33). We next wondered whether endogenous EpEX and EpICD also participate in IFNγ-mediated PD-L1 upregulation. PD-L1 mRNA was upregulated in H441 cells treated with IFNγ and DMSO, IFNγ and TAPI, and IFNγ and DAPT (Supplementary Fig. S6A). Surprisingly, the PD-L1 protein level was not increased as much in the presence of TAPI as it was in the IFNγ and DMSO–treated cells (Supplementary Fig. S6B). In other words, despite the fact that PD-L1 mRNA expression was elevated by IFNγ stimulation, the protein level could not be fully upregulated without generation of EpEX. In addition, PD-L1 protein expression was not impaired in IFNγ and DAPT–treated cells (Supplementary Fig. S6B). While PD-L1 mRNA level was increased in all cells receiving IFNγ treatment, the protein level was only upregulated in those cells with IFNγ and EpEX or IFNγ and EGF treatment (Supplementary Fig. S6C and S6D). Thus, EpEX appears to be an essential factor for PD-L1 protein upregulation.

The EGFR signaling pathway is important for PD-L1 expression in cancer cells (16), and our previous study showed that EpEX can induce EGFR signaling (5). Therefore, we speculated that EpEX might act through EGFR signaling to prevent PD-L1 degradation. Indeed, shRNA knockdown of EGFR in H441 cells attenuated EpEX- or EGF-augmented PD-L1 levels (Fig. 5D). We next investigated which EGFR downstream signaling pathway was involved in EpEX- and EGF-mediated upregulation of PD-L1. Treatment with gefitinib abrogated both EpEX- and EGF-induced upregulation of PD-L1 protein level. Importantly, U0126, which is an MEK inhibitor, but not an AKT inhibitor, abolished EpEX-mediated upregulation of PD-L1 as well (Fig. 5E). These results imply that, like EGF, EpEX-mediated upregulation of PD-L1 requires signaling through the MAPK pathway.

Previous studies have shown that both ERK and AKT signaling could regulate PD-L1 expression through different mechanisms (17, 19, 21). Hence, we wanted to know which signaling pathway was more important for PD-L1 expression in our system. Inhibition of ERK signaling, but not AKT signaling, decreased PD-L1 expression in a time-dependent manner (Supplementary Fig. S7A). We then sought to verify that the MAPK pathway affects PD-L1 protein stability. In the presence of U0126 and cycloheximide, the turnover rate of PD-L1 was higher than it was in the DMSO control (Supplementary Fig. S7B). Different inhibitors were then used to block the MAPK pathway to prevent misinterpretation of effects from nonspecific targeting. Indeed, the different MEK inhibitors, trametinib and U0126, and ERK inhibitor, SCH772984, all decreased the PD-L1 protein level and increased polyubiquitination of PD-L1 (Supplementary Fig. S7C and S7D).

Previous studies have demonstrated that glycosylation of PD-L1 at N192, N200, and N219 antagonizes GSK3β binding, which normally induces phosphorylation-dependent proteasome degradation of PD-L1 by β-TrCP. As such, nonglycosylated PD-L1 is unstable compared with glycosylated PD-L1 (17). Consistent with this mechanism, PD-L1 was upregulated in GSK3β-knockdown cells (Supplementary Fig. S7E). However, EpEX and EGF could both induce PD-L1 upregulation in GSK3β-knockdown cells (Supplementary Fig. S7F), suggesting that GSK3β is dispensable for EpEX- or EGF-mediated PD-L1 stabilization.

We next tested whether EpAb2-6 could attenuate PD-L1 upregulation and EpEX production. After EpAb2-6 treatment, PD-L1 protein level was downregulated both in vitro and in vivo (Fig. 5F). Consistent with our other findings, EpAb2-6 also decreased PD-L1 protein stability and increased polyubiquitination of PD-L1 (Fig. 5G and H). Notably, EpAb2-6 treatment inhibited PD-L1 protein level in lung cancer cells and in cell lines derived from other EpCAM-positive cancer types, including breast cancer (BT474 and MCF7) and oral cancer (Cal27; Fig. 5I). Using an apoptosis assay, we further found that coculturing cancer cells with either EpAb2-6 or T cells could induce cancer cell apoptosis, and coculturing cancer cells in the presence of both EpAb2-6 and T cells was more effective in inducing apoptosis (Fig. 5J).

We next used an animal model of metastatic colon carcinoma to test whether EpAb2-6 treatment could increase the median overall survival of metastatic tumor-bearing mice. EpAb2-6 reduced the fluorescence intensity of mouse blood bearing circulating HCT116-GFP cells, suggesting that EpAb2-6 decreased the number of HCT116-GFP cells in mouse blood vessels in vivo (Fig. 6A). NOD/SCID mice were intravenously injected with SW620 cells and then intravenously treated with EpAb2-6, or an equivalent volume of control IgG, at 24 and 96 hours after cell injection (antibody was delivered at 20 mg/kg/dose for a total dose of 40 mg/kg). The median survival time of the EpAb2-6 treatment group was significantly increased compared with the control IgG group (Fig. 6B; P < 0.005 by the log-rank test).

Because the subcutaneous model may not accurately reproduce human colon cancer biology (34), an orthotopic mouse model of colorectal cancer was also established to study the efficacy of our therapeutic antibody in the colorectal tumor microenvironment. We investigated the antitumor potential of EpAb2-6 in an orthotopic model with HCT116-Luc tumors that stably express firefly luciferase. Before the first therapeutic injection (8 days after tumor cell implantation), growing orthotopic tumors were monitored by bioluminescence imaging (Fig. 6C). Mice were treated with either control IgG or EpAb2-6 (20 mg/kg) every 2 days for 16 days. Notably, the EpAb2-6–treated group showed metastasis at 55 days, while the control group did so at 45 days (Fig. 6C). Moreover, no significant changes in body weight were observed during the treatment period (Supplementary Fig. S8). At the end of the study, the median survival times for control IgG and EpAb2-6 treatment groups were 50 and 116.5 days, respectively (Fig. 6D). From these results, we conclude that EpAb2-6 treatment can prolong survival of tumor-bearing mice.

On the basis of our observations that EpAb2-6 treatment decreases PD-L1 expression both in vitro and in vivo, we further wanted to evaluate whether EpAb2-6 could improve the therapeutic efficacy of anti-PD-L1 treatment in vivo. Because EpAb2-6 is not equipped with cross-reaction with mouse EpCAM, using syngeneic mouse model fails to evaluate the therapeutic efficacy of EpAb2-6 for immunotherapy. As illustrated in Fig. 6E, H441 cells were subcutaneously injected into NSG mice to establish the PBMC-H441–xenografted mice model. Two weeks later, mice were intravenously injected with 10⁷ PBMCs and treated with atezolizumab, an anti-PD-L1 antibody, and EpAb2-6 twice weekly for 1 month. Treatment with atezolizumab or EpAb2-6 alone inhibited tumor growth in PBMC-H441 mice, and combination therapy produced a much better tumor inhibitory effect (Fig. 6F and G). At the end of treatment, tumor tissues were collected, and CD8⁺ T cells were analyzed by flow cytometry. The CD8⁺ T-cell population was increased in mice receiving combination therapy (Fig. 6H). Collectively, these data indicate that inhibition of EpCAM by EpAb2-6 may enhance the efficacy of anti-PD-L1 therapeutics in PBMC-H441–xenografted mice.

EpCAM-overexpressing carcinoma cells possess stem cell–like features, and their presence results in high rates of recurrence, metastasis, and drug resistance (35, 36). We have previously reported that the EpEX induces EGFR/ERK signaling, which activates ADAM17 and γ-secretase to stimulate shedding of EpEX and EpICD. Soluble EpEX subsequently activates EGFR to form a positive feedback loop, causing more cleavage of EpCAM. Moreover, the importance of EGFR signaling on PD-L1 expression has been conclusively shown, and clinical data demonstrate that PD-L1 expression in the tumor microenvironment is a determinant of responses to EGFR-targeted therapies. Although EpCAM expression correlates with cancer malignancy, the molecular mechanism by which EpCAM prevents apoptosis and suppresses immune response remains unclear.

In this study, we report that soluble EpEX and membrane-bound EpCAM can directly bind to EGFR through the EGF-like domain I in EpEX and induce EGFR signaling (Fig. 1). Similarly, recent studies have reported that EGF-like domains on growth factors could bind to members of the ErbB/HER family (37) and that membrane-bound tenascin-C, a matrix component containing EGF-like repeats, could directly bind to EGFR with very low affinity (38). In addition, because both EpEX and EpCAM triggered EGFR signaling, the signal may be transduced in either an autocrine or paracrine manner.

We have previously demonstrated that anti-EpCAM neutralizing antibody, EpAb2-6, could bind to positions Y95 and D96 of EpCAM to directly induce cancer cell apoptosis in vitro (23). Inhibition of EpCAM cleavage or EpEX binding to EGFR through EpAb2-6 might induce apoptosis both in vitro and in vivo to reduce survival of cancer stem cells. Notably, we also found that HtrA2 is a direct transcriptional target of FOXO3a (Fig. 3), based on ChIP evidence indicating that FOXO3a binds to the HtrA2 promoter upon EpAb2-6 stimulation (Fig. 3E and F). Interestingly, the expression of XIAP might promote resistance of colon cancer cells to EpAb2-6 (39), but the functional outcome of EpAb2-6 treatment was increased apoptosis of cancer cells. HtrA2 promotes or induces cell death through two different mechanisms. One is by directly binding to and inhibiting inhibitor of apoptosis proteins (IAP), an action that is accompanied by a significant increase in caspase activity. The other mechanism is through a relatively uncharacterized IAP inhibition–independent, caspase-independent, and HtrA2 serine protease activity–dependent mechanism (40).

Previous studies have shown that p53 induces the activation of HtrA2 after oncogenic Ras transformation (41). Oncogenic Ras increases cytoplasmic p53 accumulation, which promotes p38 MAPK translocation to mitochondria and phosphorylation of HtrA2 (41). The phosphorylated HtrA2 releases from mitochondria into the cytosol and cleaves F-actin to downregulate lamellipodia formation (41). Proapoptotic protein, Bax, increases endoplasmic reticulum Ca⁺²-ATPase inhibitor thapsigargin-induced HtrA2 release from mitochondria to induce apoptosis in HCT116 cells (42). However, whether EpAb2-6 regulation of HtrA2 expression involves p53 or Bax remains poorly understood. To the best of our knowledge, this is the first study to demonstrate that inhibition of EpCAM increases HtrA2 gene expression, mitochondria release, and the HtrA2-induced intrinsic pathway of apoptosis.

Among all antibody-based immune therapies, anti-PD-1/PD-L1 therapies have the most beneficial outcomes in the treatment of various malignancies. This fact underscores the importance of deeply understanding the processes that control PD-L1 expression. We found that EpCAM-mediated PD-L1 stabilization gives rise to escape from immune surveillance through the EpEX–EGFR–ERK signaling axis. According to these findings, EpCAM might be an excellent option for combination with anti-PD-1/PD-L1 therapy. Indeed, EpAb2-6 decreases PD-L1 protein level and improves the therapeutic efficacy of atezolizumab. Thus, our finding reveals a novel action of EpCAM in the regulation of PD-L1 protein stability and suggests a new strategy of EpCAM/PD-L1–targeted combination therapy. However, we did not test EpCAM-mediated PD-L1 stabilization or the correlation between EpCAM and PD-L1 in colon cancer, due to the low PD-L1 expression in both colon cancer cell lines and tumor specimens. In our model, we propose that EpCAM can stabilize PD-L1 protein, instead of increasing its transcriptional expression. Therefore, if expression of endogenous PD-L1 is already low in a tumor, more EpCAM expression would not increase its protein level to any major extent. Nevertheless, more colon cancer cell lines or tumor specimens should be examined to further test whether EpCAM-mediated PD-L1 stabilization and EpCAM/PD-L1–targeted combination therapy might be beneficial in colon cancer.

Activating EGFR mutations were reported to be associated with increased PD-L1 expression, and EGFR inhibitors can reduce PD-L1 expression in NSCLC cell lines, suggesting that EGFR signaling can trigger immune escape (16). Further investigations demonstrated that EGF induces PD-L1 expression at a posttranslational level, but does not affect PD-L1 mRNA expression. The authors also found that EGF stabilizes PD-L1 by inducing its glycosylation and inactivating GSK3β, which promotes degradation of nonglycosylated PD-L1 (17). However, we found that EpEX-mediated stabilization of glycosylated PD-L1 occurs through the MAPK pathway. This mechanism differs from GSK3β-mediated degradation of nonglycosylated PD-L1. Whether the MAPK pathway regulates stability of nonglycosylated PD-L1 needs further examination.

While the crystal structure of EpEX has been published previously (43), the crystal structure of the EpAb2-6–EpEX complex is worth exploring. A crystal structure for EpAb2-6–EpEX would reveal the detailed EpAb2-6 binding sites and conformational changes in EpEX that induce apoptosis. Previous studies have demonstrated that mAbs bind to receptors and induce conformational changes in the receptor, which modulates the expression, recycling, and function of the receptor as well as ligand affinity (44, 45). Previous studies have demonstrated that structure of EpEX forms a cis-dimer and endocytosis of EpCAM into acidic compartments dissociates the dimer to permit cleavage (43, 46, 47). EpAb2-6 might affect the conformation and the endocytosis of EpCAM to disrupt acidification, dissociation, and/or cleavage.

To the best of our knowledge, this is the first study to demonstrate that inhibition of EpCAM is correlated with an increase in HtrA2 gene expression. Moreover, this upregulation occurs through FOXO3a and induces apoptosis. EpEX increases PD-L1 protein stability and the combination immunotherapy of anti-EpCAM and anti-PD-L1 antibodies provides a novel strategy for cancer therapy (Fig. 7). Most importantly, our data indicate that the development of therapeutic antibodies targeting EpCAM may hold great potential in promoting immune surveillance and eradicating cancer stem cells in the cancer microenvironment.

No potential conflicts of interest were disclosed.

H.-N. Chen: Investigation, methodology, writing-original draft, writing-review and editing. K.-H. Liang: Investigation, methodology, writing-original draft, writing-review and editing. J.-K. Lai: Investigation, methodology. C.-H. Lan: Investigation, methodology. M.-Y. Liao: Investigation, methodology. S.-H. Hung: Investigation, methodology. Y.-T. Chuang: Investigation, methodology. K.-C. Chen: Investigation, methodology. W.W.-F. Tsuei: Investigation, methodology. H.-C. Wu: Conceptualization, data curation, supervision, funding acquisition, validation, writing-original draft, project administration, writing-review and editing.

We thank the Core Facility of the Institute of Cellular and Organismic Biology (Nankang, Taipei, Taiwan) and National RNAi Core Facility, Academia Sinica (Nankang, Taipei, Taiwan) for technical support. This research was supported by Academia Sinica (AS-SUMMIT-108) and the Ministry of Science and Technology (MOST-108-3114-Y-001-002 and MOST-108-2823-8-001-001 to H.-C. Wu).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.