Broad-Spectrum Efficacy of CEACAM6-Targeted Antibody–Drug Conjugate with BET Protein Degrader in Colorectal, Lung, and Breast Cancer Mouse Models Open Access

In the past few years, the high efficacy of antibody–drug conjugates (ADC) has been demonstrated in the treatment of solid tumors such as breast cancer and lung cancer. However, there are still many challenges for ADC use in some cancer types—especially in pancreatic cancer and colorectal cancer—leaving room for improvement with more appropriate targets and payloads. Many ADCs with various targets [EGFR, HER2, human epidermal growth factor receptor 3, trophoblast cell surface antigen 2 (TROP2), and carcinoembryonic antigen–related cell adhesion molecule 5 (CEACAM5)] and with various payloads [DXd, SN38, maytansine, and monomethyl auristatin E (MMAE)] have been examined in clinical trials for colorectal cancer. However, even with the use of the approved trastuzumab DXd (HER2-DXd) and sacituzumab govitecan (TROP2-SN38), the response rates in colorectal cancer have been relatively lower than those in the cancer types for which these drugs have been approved (1, 2). TROP2-SN38 is approved for metastatic or recurrent triple-negative breast cancer, and several anti-HER2 ADCs are approved for metastatic or recurrent HER2⁺ breast cancer. Recently, HER2-DXd was also approved as a pan-tumor therapy (3). In the case of hormone receptor–positive breast cancer, the results of recent phase III trials of TROP2-SN38 and datopotamab DXd have been promising (4, 5). In lung cancer, HER2-DXd is approved for HER2-mutant non–small cell lung cancer (NSCLC), but the target population accounts for only 1% to 4% of all NSCLCs. Both datopotamab DXd and TROP2-SN38 have demonstrated some efficacy in patients with NSCLC, but at present, a significant improvement in overall survival has not been observed (1, 6). TROP2-SN38 has also received accelerated approval for advanced urothelial cancer, but the TROPiCS-04 trial (NCT04527991) of TROP2-SN38 in locally advanced or metastatic urothelial cancer did not meet its primary endpoint of overall survival. Therefore, the current focus of late-stage ADC development for solid cancers is on well-known targets and payloads, but a novel ADC could emerge to address cancer types that are intrinsically resistant to the current ADCs or have become resistant to them.

Recently, we reported an ADC (84-EBET) for pancreatic cancer. This ADC combines the bromodomain and extra-terminal (BET) protein degrader EBET with an antibody (#84.7) against CEACAM6 (ref. 7). We confirmed that EBET degrades bromodomain-containing protein 2 (BRD2) and BRD4 (7), which are members of the BET family. CEACAM6 helps maintain epithelial structure, and its high expression is correlated with poor patient prognosis in various cancers (8). BRD4 is a pivotal molecule that has drawn great attention as an epigenetic regulator of cancer (9). BRD4 recognizes and binds to super-enhancers, thereby inducing the expression of oncogenes such as c-Myc, and is involved in cancer progression (10, 11). ADCs with degrader payloads are specifically designated as degrader–antibody conjugates (DAC), and BRD4 was one of the first DAC targets reported in the literature (12, 13). Multiple ADCs bearing chimeric BRD4 degrader payloads exhibit potent antigen-dependent antitumor efficacy in several mouse tumor models (14, 15). Therefore, DACs with BET degraders seem to be promising candidates. Our 84-EBET ADC not only kills pancreatic cancer cells, but also affects stromal cells lacking CEACAM6 expression through a bystander effect, resulting in the reversal of the inflammatory and immunosuppressive microenvironment of pancreatic cancer. In multiple pancreatic cancer mouse models, a single injection of 84-EBET has induced marked tumor regression, which is further enhanced by combination with standard chemotherapy and an immune checkpoint inhibitor.

Here, we evaluated the antitumor effect of 84-EBET in preclinical colorectal cancer, breast cancer, and lung cancer models. 84-EBET showed greater antitumor activity than ADCs with known payloads and standard chemotherapies, as well as synergy with PD-1 antibody (PD1-Ab) treatment in these cancer types. These findings might facilitate the development of a superior ADC therapy and provide hope to patients.

All protocols were approved by the Institutional Animal Care and Use Committee, and the experiments were carried out in accordance with the Animal Experimentation Regulations of Eisai Co., Ltd. (animal study protocols of Eisai v18, SBL038-158). All the experiments involving human samples were conducted in accordance with Japan’s Ethical Guidelines for Medical and Biological Research Involving Human Subjects and were approved by the ethics committee of Eisai Co., Ltd., and Osaka National Hospital (REP-2017-0166, REP-2018-0486, and #2007-0071). All patients provided written informed consent.

ER-001326054 (EBET-1055, tool compound; WO2024043319, Example 122; ref. 16), ER-001368316-000 (EBET-1197, tool compound; Supplementary Methods S1; Supplementary Fig. S1), ER-001388472-000 (EBET-1593, lead payload; WO2024043319, Example 144; ref. 16), EBET-1593 linker payload (WO2024043319, Example 269; ref. 16), DXd linker payload (WO2019044947, Example 14; ref. 17), and MMAE linker payload (WO2016054315, Example 29; ref. 18) were synthesized at Eisai Co., Ltd., in accordance with the procedures mentioned in patents. EBET-1197 is a structural analog of EBET-1055 in which a phthalimide glutarimide core with racemization risk was replaced by a phenyl dihydrouracil cereblon ligand (Supplementary Fig. S2A). EBET-1593 is a structural analog of EBET-1197 that was optimized for the ADC payload. We previously reported that there was little difference in the activities of EBET-1055, EBET-1197, and EBET-1593 in degrading BRD4-BD1 (7). The chemical materials were purchased from the following suppliers: oxaliplatin (OXA) from Sawai Pharmaceutical; 5-fluorouracil (5FU) from Kyowa Hakko Kirin; fulvestrant (FUL) from AstraZeneca; palbociclib (PAL) from Guiding Bio-Tech; paclitaxel (PTX) from FUJIFILM Wako Pure Chemical Corporation; carboplatin (CBDCA) from Bristol Myers Squibb; DXd from MedChemExpress; and MMAE and SN38 linker payloads from Levena Biopharma.

We prepared 84-EBET by using the following general procedure for cysteine-based conjugation (Supplementary Fig. S2B). To a solution of anti-CEACAM6_#84.7 antibody [15.0 mg; WO2024043319, SEQ ID NO: 6-17 (16)] in 25 mmol/L histidine (5.0 mg/mL, pH 6.5), a solution of Tris (2-carboxy-ethyl)-phosphin-HCl in 2 mmol/L EDTA–Dulbecco's phosphate buffered saline minus (600 µmol/L, 1.02 mL; 6.0 eq. per antibody) was added, and a microtube rotator was used to stir the solution overnight at room temperature. The reaction mixture was diluted with a solution of 1 mmol/L EDTA and 50% propylene glycol/Dulbecco's phosphate buffered saline minus (4.02 mL). A solution of linker payload in DMA (12 mmol/L, 102 μL; 12.0 eq. per antibody) was added, and a microtube rotator was used to stir the solution for 90 minutes at room temperature. The reaction mixture was evenly dispensed into an Amicon Ultra-15 centrifugal filter (30,000 MWCO, Millipore Corporation), and after dilution with a solution of 25 mmol/L histidine and 250 mmol/L sucrose (pH 5.5), the reaction mixture was concentrated to about 1/10^th volume by centrifugation (5,000 g for 11 minutes). After performing the same procedure two more times, the concentrated ADC solution was sterilized by using a syringe filter (MILLEX-GV, pore size: 0.22 μm, Millipore Corporation) to obtain 84-EBET ADC.

84-DXd, 84-SN38, and 84-MMAE were prepared by using the same procedure except for the amounts of Tris (2-carboxy-ethyl)-phosphin-HCl (2.5 eq. to 6.0 eq. per antibody) and linker payloads (8.0 eq. to 12 eq. per antibody; Supplementary Fig. S2B). Hen egg lysozyme 3 (HEL)–EBET, being a nontargeting control ADC, was prepared by using the same procedure and anti-HEL antibody.

The drug to antibody ratios (DAR) for 84-EBET and HEL-EBET (Supplementary Table S1) were determined by using liquid chromatography/mass spectrometry (LC/MS) and capillary electrophoresis–sodium dodecyl sulfate (c-SDS) analysis. The DAR for the inter- and intrachain isoforms was calculated by using LC/MS (Supplementary Fig. S3A), and the isoform abundance ratio was established by c-SDS analysis (Supplementary Fig. S3B). Each ADC was analyzed with a Sciex TripleTOF 6600+ LC/MS system, equipped with a MassPREP on-line desalting cartridge column (20 μm, 2.1 × 10 mm, Waters). The analytic conditions were as follows: detection mode: ESI(+), m/z = 500 to 4,500; column temperature: 80°C; detection wavelength: 280 nm; mobile phase A = 0.1 vol % – acetonitrile formate; mobile phase B = 0.1 vol % formic acid and distilled water; flow rate: 50 μL/minutes; injection volume: 1 to 2 μL; and gradient over 8 minutes from B, 10% (0–0.5 minutes), 10% to 90% (0.5–4 minutes), 90% (4–5 minutes), and 10% (5–8 minutes). c-SDS analysis was performed with a Sciex PA800+ Pharmaceutical Analysis System. The analytic conditions were as follows: separate voltage 15 kV for 35 minutes and detection wavelength 220 nm. The DAR calculation method was as follows: $DAR = RInter × {0 × (IDAR0/ITotal) + 1 × (IDAR1/ITotal) + 2 × (IDAR2/ITotal) + 3 × (IDAR3/ITotal) + 4 × (IDAR4/ITotal)}+ 2 × RIntra × {(0 × (NDAR0/NTotal) + 1 × (NDAR1/NTotal) + 2 × (NDAR2/NTotal)}; RInter = (HH + LHH + LHHL)/(H + LH + HH + LHH + LHHL)$ = the abundance ratio of interchain isoforms of total ADCs; R^Intra = (H + LH)/(H + LH + HH + LHH + LHHL) = the abundance ratio of intrachain isoforms of total ADCs; H = heavy chain area/450* determined by c-SDS; LH = LH chain area/(215** + 450) determined by c-SDS; HH = HH chain area/(450 + 450) determined by c-SDS; LHH = LHH chain area/(215 + 450 + 450) determined by c-SDS; LHHL = LHHL chain area/(215 + 450 + 450 + 215) determined by c-SDS; 450*: number of amino acids in heavy chain in common antibody; 215**: number of amino acids in light chain in common antibody; I^DAR0–4: antibody-derived DAR 0 to 4 ionic strengths obtained by MS; I^Total: antibody-derived total ionic strength obtained by MS; N^DAR0–2: half antibody–derived DAR 0 to 2 ionic strengths obtained by MS; and N^Total: half antibody–derived total ionic strength obtained by MS.

The DARs for 84-DXd, 84-SN38, and 84-MMAE (Supplementary Table S1) were determined by using hydrophobic interaction chromatography analysis. Each ADC was analyzed with an Agilent 1260 Infinity II HPLC system equipped with a TSKgel Butyl-NPR column (2.5 μm, 4.6 × 35 mm, Tosoh Bioscience). The elution conditions were as follows: mobile phase A = 1.5 mol/L ammonium sulfate and 25 mmol/L sodium phosphate aqueous solution (pH 6.95 ± 0.05); mobile phase B = 25 mmol/L sodium phosphate aqueous solution (pH 6.95 ± 0.05) and 2-propanol, 75/25, volume for volume; gradient over 45 minutes from B, 0% to 100% (0–30 minutes), 100% (30–35 minutes), 100% to 0% (35–35.5 minutes), and 0% (35.5–45 minutes); flow rate = 0.8 mL/minutes; column temperature: 25°C; and injection volume: 25 μg. DAR calculation was based on the area of absorbance (at a wavelength of 280 nm) of each separated DAR fraction (Supplementary Fig. S3C).

The CRC-1 and CRC-5 colorectal cancer patient–derived xenograft (PDX) models were obtained from Sofia Bio, and the BRE-704 breast cancer PDX model was obtained from Oncodesign.

NCI-H1573 (CRL-5877, RRID: CVCL_1478; Obtained in 2021), CT26 (CRL-2638, RRID: CVCL_7256; Obtained in 2016), 4T1 (CRL-2539, RRID: CVCL_0125; Obtained in 2016), and KLN-205 (CRL-1453, RRID: CVCL_3533; Obtained in 2018) cell lines were purchased from the ATCC and cultured at 37°C under 5% CO₂ in RPMI 1640 medium supplemented with 10% FBS (for CT26 and 4T1) or 5% FBS (for NCI-H1573) or in Eagle minimal essential medium supplemented with 10% FBS (for KLN-205). All cell lines were used for no more than 15 passages. All cell lines were confirmed to be Mycoplasma negative by using a MycoAlert Mycoplasma Detection Kit (Lonza, Cat. #LT07–318) on the following dates: NCI-H1573: June 2021; CT26, 4T1, and KLN-205: August 2024. All cell lines were authenticated by short tandem repeat profiling at AACR or Eisai Co., Ltd., by using GenePrint 24 System (Promega) in September 2024. CRC-5 cancer cells were isolated from CRC-5 subcutaneous tumors by using human CD326/EpCAM MicroBeads (Miltenyi Biotec, RRID: AB_2832928) in accordance with the manufacturer’s instructions. Isolated single cells were cultured in IntestiCult Organoid Growth Medium (STEMCELL Technologies) supplemented with 10 µmol/L Y-27632 (MedChemExpress). Mouse cancer-associated fibroblasts (CAF) were isolated from CRC-5 subcutaneous tumors by using a Mouse Tumor-Associated Fibroblast Isolation Kit (Miltenyi Biotec) in accordance with the manufacturer’s instructions. Isolated CAFs were cultured in mesenchymal stem cell growth medium (Lonza).

A lentiviral vector expressing human CEACAM6 was constructed by using the pLV[Exp]-Puro-EFS lentiviral vector (VectorBuilder). VSV-G–pseudotyped lentivirus particles were produced by using MISSION Lentiviral Packaging Mix (Sigma-Aldrich) and were used to infect target cells in accordance with the manufacturer’s instructions.

Murine Abcb1a and Abcb1b genes were simultaneously targeted in KLN-205 and CT26 cells by using the CRISPR/Cas9 ribonucleoprotein method. Recombinant Cas9 proteins (TrueCut Cas9 v2, Invitrogen or Alt-R S.p. HiFi Cas9 V3, Integrated DNA Technologies) and single-guide RNAs (sgRNA; TrueGuide gRNA, Invitrogen or Alt-R CRISPR sgRNA, Integrated DNA Technologies) were obtained from Thermo Fisher Scientific or Integrated DNA Technologies, Inc. The genomic target sequences for the sgRNAs were as follows: Abcb1a: 5′-TAA​GTG​GGA​GCG​CCA​CTC​CA-3′ and Abcb1b: 5′-TCC​AAA​CAC​CAG​CAT​CAA​GA-3′. CT26 and KLN-205 cells were electroporated with sgRNAs/Cas9 ribonucleoprotein complexes by using a NEPA21 Super Electroporator (NEPAGENE). The electroporated cells were subjected to rhodamine-123 staining (Sigma-Aldrich) followed by FACS to isolate rhodamine-positive P-glycoprotein–knockout cells. The isolated CT26 and KLN-205 cells were subsequently infected with human CEACAM6–expressing lentivirus. The P-glycoprotein–knockout 4T1 cells were established previously (19). These cells were infected with human CEACAM6–expressing lentivirus without rhodamine-positive sorting.

Body weight and tumor size were measured twice a week, and the formula V = (d² × D)/2 (in which d = minor tumor axis and D = major tumor axis) was used to determine the tumor volume (mm³) or the relative tumor volume compared with the initial tumor volume. Complete regression (100% reduction of tumor volume) was defined as a relative tumor volume of 0.01. For the subcutaneous colorectal cancer and breast cancer PDX models, the tumor tissues were cut into fragments of 3 × 3 × 3 mm and inoculated subcutaneously into the right flank of female NOD/SCID mice (Charles River Laboratories, RRID: IMSR_JAX:001303) through a trocar needle. For the subcutaneous lung cancer tumor model, 3.0 × 10⁶ H1573 cells were subcutaneously injected into the right flank of female nude mice (Charles River Laboratories, RRID: MGI:7707390). When the tumor volume reached approximately 100 mm³, mice were allocated randomly to each group. On day 0, 5FU (50 mg/kg) or OXA (5 mg/kg) was intraperitoneally injected; FUL (5 mg/head) was subcutaneously injected; PAL (100 mg/kg) was orally administered; or PTX (40 mg/kg), CBDCA (100 mg/kg), ADCs, and 84-naked antibody (84-naked-Ab) were injected into the tail vein. For the mouse syngeneic tumor models, 1 × 10⁶ CT26 cells or 1 × 10⁶ 4T1 cells were subcutaneously injected into the right flank of female BALB/c mice (Charles River Laboratories, RRID :MGI:6323059). For the KLN-205 model, 1.5 × 10⁶ cells were subcutaneously injected into the right flank of female DBA mice (Charles River Laboratories, RRID: IMSR_CRL:026). Mice were randomly assigned to each group when the tumor volume reached approximately 40 mm³ for CT26, 90 mm³ for 4T1, and 110 mm³ for KLN-205. In the CT26 model, anti–mouse CD8α antibody (Bio X Cell, Cat. #BE0061, clone #RMP1-14, RRID: AB_1125541, 200 μg/head) was intraperitoneally injected on days 0 and 7, anti–mouse PD-1 antibody (Bio X Cell, Cat. #BE0146, clone #2.43, RRID: AB_10949053, 200 μg/head) was intraperitoneally injected on days 1 and 8, and ADC was injected into the tail vein on day 1. In the 4T1 and KLN-205 models, anti–mouse PD-1 antibody (Bio X Cell, 200 μg/head) was intraperitoneally injected on days 0 and 7, and ADC was injected into the tail vein on day 0. A mouse PD-1 antibody was used because a human PD-1 antibody would have required engineered mouse models.

PDX tumors were extirpated from the mice and digested with a Tumor Dissociation Kit (Miltenyi Biotec Cat. #130-095-929, RRID: SCR_020276) and a gentleMACS Dissociator (Miltenyi Biotec, RRID: SCR_020271). Tumor fragments about 50 μm in diameter were separated by repeated low-speed centrifugation and embedded in 25 μL of medium (see below) with 10% Matrigel Growth Factor Reduced (Corning) on 384-well ultralow-attachment microplates (Corning) at a concentration of 150 fragments per well. After the solidification of the Matrigel for 2 hours at 37°C, fresh medium (25 μL) with compounds was added to each well, and the plates were further incubated at 37°C. After 5 days of culture, 25 μL of CellTiter-Glo 3D reagent (Promega) was added to each well and the luminescence was quantified with an EnVision microplate reader (PerkinElmer, RRID: SCR_018038). Depending on the model, the following kinds of culture medium were used: StemPro hESC SFM (Thermo Fisher Scientific) supplemented with 8 ng/mL basic fibroblast growth factor (Thermo Fisher Scientific), 1 µmol/L A83-01 (Wako), 10 µmol/L Y-27632 (MedChemExpress), 1 µmol/L CHIR-99021 (Cayman Chemical), and 1 µmol/L DMH-1 (Shanghai Haoyuan Chemexpress) for the BRE-704 model; IntestiCult Organoid Growth Medium (STEMCELL Technologies) was used for the CRC-1 and CRC-5 models.

H1573 cells were harvested, diluted in RPMI-1640 medium containing 5% FBS, and dispensed in 384-well plates at a concentration of 500 cells/well in 25 μL medium. After incubation of the cells at 37°C for 2 hours, fresh medium (25 μL) with compounds was added to each well and the plates were further incubated at 37°C. After the plates had been incubated for 5 days, 25 μL of CellTiter-Glo 2.0 reagent (Promega) was added to each well and the luminescence was quantified using an EnVision microplate reader (PerkinElmer).

Assays were performed at ChemPartner. After the cancer cells had been harvested, they were dispensed into 384-well plates with 40 μL medium/well in appropriate cell numbers for each cell line, ranging from 100 to 4,800 cells/well. After overnight incubation of the plates at 37°C, fresh medium (2 μL) with compounds was added to each well and the plates were further incubated at 37°C for 5 days. Serial dilutions were performed from the highest concentration of each compound: MMAE at 10 nmol/L, DXd at 300 nmol/L, and EBET at 100 nmol/L. After the 5-day culture, 25 μL of CellTiter-Glo 2.0 reagent (Promega) was added to each well and the luminescence was quantified using an EnVision microplate reader (PerkinElmer).

CRC-5 cancer cells (1 × 10⁵) were seeded on a collagen-I–coated six-well plate (Iwaki) in 2 mL IntestiCult Organoid Growth Medium (STEMCELL Technologies) supplemented with 10 µmol/L Y-27632 (MedChemExpress) per well. After incubation of the plates at 37°C for 3 days, the medium was removed and mouse CAFs (1 × 10⁵) were seeded on a six-well plate in 2 mL mesenchymal stem cell growth medium per well. The culture medium was replaced with 2 mL of fresh medium containing compounds 1 day after incubation at 37°C. After incubation at 37°C for an additional 2 days, the culture supernatants were collected and mouse IL-6 and mouse leukemia inhibitory factor (LIF) were quantified with Quantikine ELISA kits (IL-6: R&D Systems Cat. #M6000B, RRID: AB_2877063; LIF: R&D Systems Cat. #MLF00, RRID: AB_3224403) in accordance with the manufacturer’s instructions.

Seven days after drug administration, the mice were sacrificed and their tumors and the draining lymph nodes were collected. Tumors were dissociated by using a tumor dissociation kit (Miltenyi Biotec Cat. #130-095-929, RRID: SCR_020276) and a gentleMACS Dissociator (Miltenyi Biotec); CD45-positive cells were separated by using mouse tumor-infiltrating lymphocyte (CD45) microbeads (Miltenyi Biotec Cat. #130-110-618, RRID: AB_3224404) and an OctoMACS Separator (Miltenyi Biotec). The lymph nodes were ground, washed, and filtered and then the cells were collected. Before staining, tumor-infiltrating lymphocytes were cultured for 3 hours at 37°C in RPMI-1640 medium containing 10% FBS, 20 ng/mL phorbol 12-myristate 13-acetate (Sigma), 500 ng/mL ionomycin (Sigma), and the protein transport inhibitor BD GolgiPlug (BD Biosciences). For dead cell exclusion, 1.5 × 10⁶ cells were first incubated with 0.1 μmol/L cisplatin-198 isotope (Fluidigm) and then blocked with 1/20-diluted anti–mouse CD16/32 (BD Biosciences Cat. #553142, RRID: AB_394657) and stained with extracellular antibodies (Supplementary Table S2) in accordance with the manufacturers’ instructions. The cells were fixed with Maxpar Fix I Buffer (Fluidigm), permeabilized with Maxpar Perm-S Buffer (Fluidigm), and stained with an intracellular antibody. Next, the cells were fixed with 1.6% paraformaldehyde and labeled with Cell-ID Intercalator-Ir (Fluidigm). Finally, the cells were washed and suspended in Maxpar water (Fluidigm); 0.1× EQ Four Element Calibration Beads (Fluidigm) were then added and the suspension was mixed before data acquisition on a Helios CyTOF mass cytometry system (Fluidigm, RRID: SCR_019916). The measured data were analyzed by using the Cytobank Premium data analysis platform (Beckman Coulter, v10.1, RRID: SCR_014043).

For staining of xenograft tumors, 5-μm sections from formalin-fixed, paraffin-embedded tumor samples were deparaffinized, and antigen retrieval was performed with Dako Target Retrieval Solution, pH 6.0 (DAKO). Endogenous peroxidases were blocked with 3% H₂O₂. Primary antibodies used were as follows: BRD4 (Bethyl; A301-985A100, RRID: AB_2620184), phosphorylated STAT3 [(pSTAT3)^Y705; Abcam; ab76315, RRID: AB_1658549], phosphorylated small mothers against decapentaplegic 2 [(pSMAD2)^S465/S467; Thermo Fisher Scientific; 44-244G, RRID: AB_2533614], pSMAD3^S423/S425 (Abcam; ab52903, RRID: AB_882596), collagen I (Abcam; ab34710, RRID: AB_731684), CD8a (Abcam; ab217344, RRID: AB_2890649), CD4 (Abcam; ab183685, RRID: AB_2686917), and granzyme B (R&D Systems; AF1865, RRID: AB_2294988). For CEACAM6 staining of xenograft tumors, antigen retrieval was performed by using REAL Target Retrieval Solution, pH 9.0 (DAKO). Endogenous peroxidases were blocked with 3% H₂O₂, and the sections were incubated with #84.7 CEACAM6 antibody. Detection was performed by using a SignalStain DAB Substrate Kit (Cell Signaling Technology) or ImmPACT DAB EqV Substrate Kit (Vector Laboratories), and hematoxylin was used as a nuclear counterstain. For double staining of platelet-derived growth factor receptor α (PDGFRα) and α-smooth muscle actin (α-SMA) of xenograft tumors, antigen retrieval was performed with SignalStain Citrate Unmasking Solution (Cell Signaling Technology) and REAL Target Retrieval Solution, pH 6.0 (DAKO). Endogenous peroxidases were blocked with 3% H₂O₂. The primary antibodies used were PDGFRα (Cell Signaling Technology; 3174, RRID: AB_2162345) and α-SMA (Sigma-Aldrich; A5691, RRID: AB_476746), and the substrates used were a SignalStain DAB Substrate Kit (Cell Signaling Technology) and a Ferangi Blue Chromogen kit 2 (Biocare Medical). For CEACAM6 staining of clinical lung cancer, colorectal cancer, and breast cancer samples, all procedures were performed automatically in a BenchMark system (Ventana Medical Systems, RRID: SCR_025506). Sections (3 μm) from formalin-fixed, paraffin-embedded tumor samples were deparaffinized, and antigen retrieval was performed with CC2 cell-conditioning solution at 100°C for 64 minutes. Endogenous peroxidases were blocked with 3% H₂O₂, and the sections were incubated with the CEACAM6 antibody (MBL; D028-3/KOR, RRID: AB_592067). Detection was performed by using an iVIEW DAB Detection Kit (Ventana Medical Systems). Hematoxylin was used as a nuclear counterstain. For quantitative evaluations, whole digital slide images were obtained by using an Aperio AT2 slide scanner (Leica Biosystems, RRID: SCR_021256). Digital images of tumor tissues were annotated as tumor, stromal, and necrotic areas. The area of staining by each antibody was quantified by using HALO image analysis software (Indica Labs, v2.3.2089.69, RRID: SCR_018350).

RNA sequencing (RNA-seq) data from The Cancer Genome Atlas (TCGA, RRID: SCR_003193) database were downloaded from the UCSC Xena Data Hubs (https://xenabrowser.net/datapages/, RRID: SCR_018938). The gene expression data and consensus molecular subtyping calls for clinical colorectal cancer samples were obtained from Synapse (Synapse ID syn2623706, RRID: SCR_006307).

The activation levels of each pathway were evaluated by using a gene set variation analysis algorithm (20). The pathway data were obtained from either MSigDB v5.2 (RRID: SCR_016863) or Synapse (Synapse ID syn2623706).

For bulk RNA-seq of CRC-5 tumors, total RNA was extracted by using a Maxwell RSC simplyRNA Tissue Kit (Promega), and the cDNA library was prepared with an Agilent SureSelect Strand Specific RNA Library Prep Kit (Agilent). Reads (150 bp, paired-end) were sequenced on an Illumina NextSeq550 system (14 Gb per sample; Illumina, RRID: SCR_016381). Reads were separated by species of origin by using the human.GRCh38 and mouse.mm10 assemblies.

Expression data for mouse stromal genes with >1 transcripts per million were extracted from the bulk RNA-seq data of CRC-5 tumors. The pathways that were enriched in ADC-treated tumors were analyzed by using GSEA software (v4.1.0, Broad Institute, RRID: SCR_003199; refs. 21, 22).

No statistical method was used to predetermine sample size. Data collection and analysis were not performed with blinding to the conditions of the experiments. Sample sizes, statistical analyses, and P values are indicated in the figure legends. GraphPad Prism software (v9, RRID: SCR_002798) was used for statistical calculations.

The RNA-seq data generated for this study are available at the European Genome-Phenome Archive (https://ega-archive.org/) with study ID EGAS50000000598. RNA-seq data are not publicly available to maintain patient privacy. Therefore, to access next-generation sequencing data, please request access from our data access committee from the European Genome-Phenome Archive links above. The data access committee will send a consent form by e-mail upon request, and access will be granted immediately upon return by the requestor. Gene expression analyses were performed by using open-source GSEA software (Broad Institute, v4.1.0) and MSigDB (Broad Institute, v5.2).

We recently developed the CEACAM6 antibody #84.7 and the BET protein degrader EBET as a payload to establish an ADC for pancreatic cancer (84-EBET; ref. 7). To explore the further application of 84-EBET, we first examined the cell-killing effect of EBET on 40 different cancer types by using a panel of 408 cell lines. EBET-1197, one of the tool compounds, showed stronger cell-killing efficacy than MMAE and DXd—clinically approved payloads for solid cancers—with 100% of maximum growth inhibition of most cell lines (Supplementary Fig. S4). On the basis of these data, we expected that EBET would show broad-spectrum efficacy in various cancer types that was greater than that of known ADC payloads. We then performed IHC profiling with clinical samples and observed CEACAM6 expression in almost all colorectal cancer and lung cancer cases and nearly half of the breast cancer cases (Fig. 1A; Supplementary Table S3). CEACAM6 expression was higher in cancerous lesions than in normal tissues. We then examined the association among CEACAM6 expression, molecular subtypes, and histologic types in these cancers. In the colorectal cancer dataset of TCGA (23), CEACAM6 expression was significantly higher in the consensus molecular subtypes (CMS) 2, 3, and 4 [CMS2, CMS3, and CMS4 (24)] than in CMS1 (Fig. 1B and C). In the breast cancer dataset of TCGA, CEACAM6 expression was significantly higher in the luminal and HER2⁺ molecular subtypes than in the basal-like subtype (Fig. 1D and E). In the lung cancer dataset of TCGA, CEACAM6 expression was significantly higher in adenocarcinoma than in squamous cell carcinoma (Fig. 1F and G). Over the past decade, the incidence of CRC has increased in individuals under the age of 50 years, especially in developed countries. These patients with early-onset colorectal cancer (EOCRC) have a poorer prognosis than patients with late-onset colorectal cancer (LOCRC) despite a higher dose intensity of chemotherapy and fewer treatment adverse events (25). IHC analysis of another clinical sample set from EOCRC and LOCRC revealed that there was no difference in the intensity and frequency of CEACAM6 expression between EOCRC and LOCRC (Supplementary Table S4).

Because our colorectal cancer PDX models were too few for the application of hierarchical clustering–based classification, we estimated the CMS classifications from the activation levels of specific pathways evaluated by using a gene set enrichment analysis–based algorithm (Supplementary Fig. S5; ref. 24). Among 10 colorectal cancer PDX models, CRC-1 was likely to be classified as CMS2 owing to its higher WNT, MYC, and E2F targets and lower TGF-β activation and epithelial–mesenchymal transition, whereas CRC-5 was likely to be classified as CMS4 because of its lower WNT, MYC, and E2F targets and higher TGF-β activation. In addition, CMS4 cancers show a high level of stromal infiltration (24), and the stroma-rich histology of the CRC-5 tumor was consistent with that of CMS4 (Fig. 2A). As a breast cancer PDX model, we chose a BRE-704 luminal model resistant to CDK4/6 inhibitors and endocrine therapy (26). As we had no access to a lung cancer PDX model, we screened the Cancer Cell Line Encyclopedia database (27) by using CEACAM6 mRNA expression and histology data and selected the NCI-H1573 lung adenocarcinoma cell line.

All these models showed in vivo expression of the CEACAM6 protein (Fig. 2A). By using these representative models, we examined the efficacy of our CEACAM6-targeting ADC with EBET-1593 and 84-EBET or the efficacy of CEACAM6 ADCs with known payloads and standard chemotherapies. The antibody against HEL3 was also evaluated as a nontargeting control antibody. We found that 84-EBET showed growth inhibition effects superior to those of standard chemotherapies and CEACAM6 ADCs conjugated with common payloads (DXd, SN38, or MMAE) in PDX-derived organoid and cell line models (Fig. 2B). We also confirmed that 84-naked-Ab showed no growth inhibition effects in vitro or in vivo (Supplementary Fig. S6A–S6C). We then evaluated the antitumor effect of 84-EBET in the equivalent mouse models. Tumor regression was defined as a state of tumor volume below the baseline. In the CRC-1 PDX model, a single injection of 84-EBET maintained minimal tumor regression for 3 months at least, whereas standard chemotherapy (5FU + OXA) or HEL-EBET showed no sign of efficacy (Fig. 2C; Supplementary Fig. S7). There were no signs of tumor recurrence in the 3 mg/kg group of 84-EBET, and pathologic analysis revealed no viable cancer cells, confirming the presence of scar tissue. In the CRC-5 model, tumor regression was also induced by 84-EBET but not by standard chemotherapy. Similarly, in the breast cancer and lung cancer models, a single injection of 84-EBET induced substantial tumor regression, whereas key chemotherapeutic agents (CBDCA for lung cancer and PTX or FUL + PAL for breast cancer) were not as effective.

Because numerous studies have suggested that BET modulators possess anti-inflammatory or antifibrotic activity (28), we performed detailed mechanism-of-action analyses by using stroma-rich CRC-5 tumors. In the tumor microenvironment (TME), the composition of CAFs is related to the malignant phenotype of the tumor (29). Although the reported CAF subtypes differ by cancer types, they can be roughly categorized into two subtypes, namely myofibroblastic (myCAF) and nonmyofibroblastic CAFs. Nonmyofibroblastic CAFs often exhibit inflammatory traits and induce an immunosuppressive TME in pancreatic cancer, breast cancer, and colorectal cancer (29). We first confirmed that the coculture of mouse CAFs with CRC-5 cancer cells upregulated the secretion of the inflammatory cytokines IL-6 and LIF from CAFs, and treatment with EBET-1055 (one of the tool compounds) or 84-EBET canceled that upregulation (Supplementary Fig. S8). To clarify the stromal modulation by BET protein degradation in vivo, we stained BRD4 as a direct target of EBET; PDGFRα and pSTAT3 as inflammatory CAF (iCAF) markers; α-SMA and pSMAD2/3 as myCAF markers; and collagen I as a pan-CAF marker (30, 31) in CRC-5 tumors after treatment with 84-EBET, HEL-EBET, or 5FU + OXA. The percentage content of the BRD4 protein in both cancer and stromal cells in the tumors was significantly reduced by 84-EBET treatment, indicating that there was a bystander effect of the diffused payload on CEACAM6-negative stromal cells after targeted delivery (Fig. 3A and B). In the stroma, compared with the vehicle treatment, 84-EBET significantly reduced the percentage areas of PDGFRα⁺, pSTAT3⁺, α-SMA⁺, and pSMAD2/3⁺, but it did not affect the collagen I⁺ area (Fig. 3B). We then profiled the expression of mouse stromal genes in CRC-5 PDX tumors; 84-EBET treatment suppressed iCAF-, myCAF-, and pan-CAF–related gene expression and pathways (Fig. 4A and B; ref. 30).

We evaluated the combination of 84-EBET with PD1-Ab by using the human CEACAM6–expressing mouse syngeneic tumor models CT26 (colorectal cancer), 4T1 (breast cancer), and KLN-205 (lung cancer). Although PD1-Ab alone showed limited antitumor effect in any of the models, both single and combinational treatments with 84-EBET induced substantial tumor regression (Fig. 5A and B; Supplementary Fig. S9A and S9B). Notably, combinational treatment induced complete regression in most tumors (16 of 20). In the CT26 and 4T1 models, a combined effect was observed, with a significantly smaller tumor volume with PD1-Ab + 84-EBET than with 84-EBet alone (Supplementary Fig. S9A). IHC analysis revealed that both single and combinational treatments with 84-EBET significantly reduced the abundance of the BRD4 protein and PDGFRα⁺ iCAFs and significantly increased that of CD8⁺ T cells in CT26 tumors (Fig. 5C and D). To clarify the contribution of CD8⁺ T cells to the antitumor effect of 84-EBET, we also examined the depletion of CD8⁺ T cells by using CD8α antibody (32) in the CT26 model. Combination with CD8α antibody significantly attenuated the efficacy of 84-EBET 24 days after treatment in the CT26 tumor model (Supplementary Fig. S9A). Furthermore, mass cytometric analysis of immune cells in the tumors revealed that 84-EBET or 84-EBET + PD1-Ab significantly decreased the percentages of immunosuppressive cells—monocytic myeloid-derived suppressor cells and tumor-associated macrophages—and significantly increased the percentages of effector cells—CD8⁺ T cells and NK cells—compared with vehicle (Fig. 6A). In tumor-draining lymph nodes, the percentage of migratory dendritic cells increased significantly with 84-EBET or 84-EBET + PD1-Ab treatment compared with vehicle (Fig. 6B). Taken together, these findings demonstrate the potential efficacy of a CEACAM6–EBET ADC against a wide range of solid cancers through its direct anticancer activity and its reversal of immunosuppressive TME activity.

We proposed a novel ADC approach that not only kills target tumor cells but also modulates the surrounding TME components. We showed that the bystander effect was due to a mechanism in which 84-EBET was taken up by CEACAM6-positive cancer cells; the EBET payload then leaked from the dead cells and was taken up by the surrounding CEACAM6-negative CAFs (Figs. 3 and 4; Supplementary Fig. S8). One explanation for the strong bystander effect of 84-EBET is that our payload acts as a degrader and therefore continues to act catalytically in the cell that has taken up the payload and in the surrounding cells. Additionally, EBET is likely to have high membrane permeability owing to its high lipid solubility. We have previously reported that EBET exhibits a stronger bystander effect than other payloads (7). Another reason for 84-EBET being more potent than 84-MMAE and 84-DXd (Fig. 2B) could be that EBET showed a stronger efficacy than MMAE and DXd in the cancer cell panel (Supplementary Fig. S4).

Our in vivo studies of 84-EBET focused on comparing it with standard chemotherapies. We expect that EBET will be more effective than other payloads in mouse tumor models with high CEACAM6 expression, as we showed in vitro (Fig. 2B). Additionally, although we were unable to evaluate many models of colorectal cancer and breast cancer because of a lack of PDX resources, the superiority of 84-EBET over other payloads has been reported in pancreatic ductal adenocarcinoma PDX panels in vitro and in vivo (7). Although all tumor models induced regression following 84-EBET treatment, tumor regrowth eventually began in the H1573 model (Fig. 2C; Supplementary Fig. S7). Potential mechanisms for this could include genomic alterations that impair core components of the relevant E3 ligase complexes (33). Antigen loss is known to be a common mechanism of resistance to ADCs (34), and 84-EBET may pose a similar risk. To overcome these resistance mechanisms, we are currently investigating approaches to maximizing antitumor efficacy, such as optimal dosing regimens of 84-EBET and combination with PD1-Ab.

Although we cannot directly compare 84-EBET with previously reported DACs with BET protein degrader, the major differences between these and 84-EBET are in the antibody target and the conjugation method (13–15). CEACAM6 is highly expressed in cancer cells compared with normal cells, making it an optimal target for highly potent payloads such as EBET. We have confirmed by using an HEK293 overexpression system that our antibody #84.7 is specific for human CEACAM6 and does not bind to human CEACAM1 or CEACAM5 (WO2024043319; ref. 16). Furthermore, the disulfide rebridged conjugation method that we applied to generate 84-EBET has reportedly demonstrated improved pharmacokinetics, superior efficacy, and reduced toxicity in vivo compared with conventional heterogeneous ADCs (35–37).

Our findings suggest that 84-EBET inhibited the inflammatory phenotype of CAFs, reversed the immunosuppressive TME, and drove the cancer–immunity cycle. Especially in CMS4-type colorectal cancer tumors, a stroma-enriched inflamed TME is potentially linked to malignant phenotypes such as TGF-β activation, epithelial–mesenchymal transition, and immunosuppression (38, 39). In EOCRC, inflammatory bowel disease is one of the most important risk factors (40), and inflammation could be one of the key elements in its TME. CAFs with fibroblast-associated protein expression are more enriched in EOCRC tumors than in LOCRC tumors, and this abundance is associated with poorer prognosis in EOCRC (41). Our analysis of colorectal cancer PDX tumors revealed that 84-EBET treatment suppressed the expression of the gene encoding fibroblast-associated protein in the stroma (Fig. 4A). We therefore expect that this novel ADC, which can target the inflammatory and immunosuppressive stroma via a bystander effect, will be highly effective in these tumors.

The broad-spectrum effectiveness of 84-EBET in colorectal cancer, breast cancer, lung cancer, and pancreatic ductal adenocarcinoma (7) models, as well as in our cancer cell panel (Supplementary Fig. S4), is important in the treatment of refractory cancers. CRC-1 is a model that has undergone multiple chemotherapy regimens, whereas CRC-5 is of the CMS4 type, known for its worse prognosis (24). The effectiveness of 84-EBET in these models indicates the potential for treatment with 84-EBET. Furthermore, our previous report of high selectivity for cancer cells over normal cells suggests that our approach is relevant for patients. Although potent drugs are expected to be effective in late-line treatments, it is also necessary to ensure low toxicity to minimize side effects in these patients. The next step will be to conduct detailed toxicity studies in large animals, including nonhuman primates, to demonstrate the high safety profile of 84-EBET. We also intend to explore target treatment lines and patient stratification, focusing on the cancer types for which we have reported preclinical data.

M. Miyano reports a patent for WO2024043319 issued. K. Mori reports grants from Eisai Co., Ltd., during the conduct of the study. Y. Nakazawa reports other support from Eisai Co., Ltd., during the conduct of the study, as well as a patent for WO2024043319 pending. No disclosures were reported by the other authors.

H. Kogai: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, methodology, writing–original draft, writing–review and editing. S. Tsukamoto: Resources, data curation, investigation, writing–review and editing. M. Koga: Data curation, investigation, writing–review and editing. M. Miyano: Resources, data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. T. Akagi: Resources, supervision, investigation, methodology, writing–review and editing. A. Yamaguchi: Resources, data curation, writing–review and editing. K. Mori: Resources, investigation. K. Gotoh: Resources, supervision. Y. Nakazawa: Conceptualization, resources, data curation, supervision, investigation, methodology, writing–original draft, project administration, writing–review and editing.

We thank Y. Yabe and H. Umihara (Eisai Co., Ltd.) for preparation of the ADC; A. Akao for Quality check of ADC; P. Li (Eisai Co., Ltd.) for performing the mass cytometric analysis; R. Dairiki and M. Hamaguchi (Eisai Co., Ltd.) for performing the bioinformatics analysis; K. Sasai, S. Watanabe, and A. Yokoi (Eisai Co., Ltd.) for scientific discussions; and S. Nakamori and K. Morimoto (Osaka National Hospital) for providing clinical samples.