H1975, A549, H1299, H2030, H2009, HPAC and HPAF-II cells were purchased from ATCC. H1975, A549, H1299, H2030, and H2009 cells were maintained in RPMI1640 (Gibco) supplemented with 10% v/v FBS (Gibco) and 1% penicillin-streptomycin (P/S, Gibco). HPAC cells were cultured in F12K (ATCC) with 10% FBS and 1% P/S. HPAF-II cells were maintained in EMEM (ATCC) with 10% FBS and 1% P/S. H1975-CEACAM5, H2009-CEACAM5, A549-CEACAM5, and H1299-CEACAM5, stably expressing CEACAM5, were generated by stable infection with lentivirus from the CEACAM5 lentiviral plasmid (Origene) using a commonly used protocol (36). Stably transfected cells were selected in RPMI1640 supplemented with 10% FBS, 1% P/S and 1 μg/ml (for A549-CEACAM5) or 2 μg/ml (for H1975-CEACAM5, H2009-CEACAM5, and H1299-CEACAM5) puromycin (Gibco). The cell surface CEACAM5 expression of the stably transfected cells was then assessed with PE-conjugated anti-CEACAM5 IgG1 (Miltenyi Biotec, 130-114-217) by flow cytometry. Anti-CEACAM5 CAR-T cells were generated as previously described (35) and expanded in the T cell media (RPMI1640 supplemented with extra 2mM glutamine, 10% human serum, and 1% P/S) in the presence of hIL-2 (fed every 2 days, 50 IU/ml, Miltenyi Biotec).

Human tumor and normal multiple frozen tissue arrays were purchased from Fisher scientific (50-180-886). 14 tumors and 14 correspondent normal tissues (brain, breast, colon, muscle, kidney, liver, lung, pancreas, prostate, skin, small intestine, stomach, ovary, and uterus) were mounted on a positively charged glass slide. For immunohistochemistry, the tissue slide was blocked with 10% normal horse serum for 1 h at 25°C and incubated with mouse anti-human CEACAM5 antibody (Novus biologicals, NB11058734, 1:100) for overnight at 4°C in humidified chamber. The tissue slide was washed with PBS, incubated with biotin-conjugated anti-mouse antibody (Invitrogen, B2763, 1:200) for 1 h at 25°C, and washed again with PBS. Slide was incubated with ImmunoCruz ABC staining (Santacruz, sc-516216) by the manufacturer instructions. Stains were then visualized using DAB peroxidase substrate (Santacruz, sc-249982). The positive pixel areas of CEACAM5 staining of the entire tissue were quantified using Image J software and the total CEACAM5 area of tumor tissue normalized to the total CEACAM5 pixel area of the corresponding normal tissue.

The cell killing activity of DM4 or the ADC SAR408701 analog was measured by CellTiter-Glo Luminescent cell viability assay kit (Promega, G7571). Cells (2.5×10³ cells/well in white 96-well plate) were cultured for 12 h prior to treatment with the indicated doses of DM4 or ADC for 96 h at 37°C. Normalized % ATP values were calculated by normalizing luminescence values for buffer (DPBS or DMSO)-treated cells. The LDH-Glo cytotoxicity assay kit (Promega, J2381) was used to measure cell viability in treatment of anti-CEACAM5 CAR-T cells. Control T or CAR-T cells as effector cells were incubated with target cells (5×10³ cells/well in 96-well plate) at the indicated E:T ratio for 24 h at 37°C. Controls conducted for the calculation of percent cytotoxicity were included according to the manufacturer’s instructions.

All studies were approved by the University of Pittsburgh institutional Animal Care and Use Committee. A549 cells (7×10⁶/mice), A549-CEACAM5 cells (7×10⁶/mice) or H1975-CEACAM5 cells (5×10⁶/mice) resuspended in 200 μl of DPBS were subcutaneously injected into the right flank of female NSG mice (6-8 weeks old, The Jackson Laboratory). When the tumor volume reached approximately 150 mm³, mice were intravenously treated with ADC SAR408701 analog (10 mg/kg or 5 mg/kg), Control T (5x10⁶/mice) or anti-CEACAM5 CAR-T cells (5×10⁶ or 2×10⁶/mice) every 4 days, two times, via tail vein. Tumor volume was measured by two-dimensional measurements with a caliper and calculated according to the formula V=0.5 × length × (width)². Tumor growth inhibition (TGI) by anti-CEACAM5 ADC or CAR-T compared to that by vehicle or control T was determined on the last day of the study according to the formula: TGI (%) = [1-( V f treated - V i treated )/( V f control - V i control )] ×100, where V_f is the final mean tumor volume in the treated group (ADC or CAR-T cells), and V_i is the initial mean tumor volume in the control group (vehicle or control T cells). Animals were euthanized when the tumor volume reached >1-1.5 cm³.

Statistical analyses were performed using GraphPad Prism software (GraphPad, Inc.). Data are presented as the mean ±SD for representative data from three independent experiments. One-way analysis of variance (ANOVA) with Tukey’s post hoc test was used to evaluate the significance of differences. Survival curve was represented as Kaplan-Meier plots, with statistical significance determined by log-rank (Mantel-Cox) tests. P values less than 0.05 were considered statistically significant. P values less than 0.05, 0.01, 0.001, and 0.0001 are indicated as *, **, ***, and ****, in the respective figure.

To verify protein expression of CEACAM5 in human tissues, we performed IHC analysis with 14 different human tumor and 14 correspondent normal tissues (brain, breast, colon, muscle, kidney, liver, lung, pancreas, prostate, skin, small intestine, stomach, ovary, and uterus) on the tissue microarray. Lung tumor and pancreas tumor highly expressed CEACAM5 compared to normal lung and pancreas tissues ( Figure 1A ). In contrast, CEACAM5 expression was detected in normal tissue as well as tumor tissue from the colon, and no CEACAM5 was observed in normal and tumor of other tissues, as reported previously (16, 37). Recent data indicated that prostate cancer subtypes are differentiated as prostate adenocarcinoma (PrAd) and neuroendocrine prostate cancer (NEPC) (37). CEACAM5 is especially prominent as a therapeutic target in NEPC (35). In accordance with CEACAM5 overexpression in lung tumor tissues, the anti-CEACAM5 antibody drug-conjugate (ADC) SAR408701 is being evaluated in non-small cell lung cancer (NSCLC) with different interventions in clinical trials (16, 20).

DM4 is a potent cytotoxic agent derived from maytansine that blocks tubulin polymerization and is used as a payload for ADC SAR408701. We determined the sensitivity of NSCLC cells (NSCLCs) and PDAC cells (PDACs) to DM4. Three NSCLCs, A549, H2030, and H1299 cells, responded to treatment with DM4, but two NSCLCs, H1975 and H2009 cells, showed resistance to DM4 ( Figure 1B ). The two tested PDACs, HPAC and HPAF-II cells, were sensitive to DM4 in vitro ( Figure 1B ). Vecchione, L. et al. reported that BRAF(V600E) is a predictive biomarker of the DM4 response in colon cancer PDX models (26). To identify whether the DM4 sensitivity would be predicted by oncogene mutation status, we examined oncogene mutations through Cancer Cell Line Encyclopedia (CCLE) database (38). No correlation was found between the observed DM4 sensitivity and oncogene(s) mutation status ( Supplementary Table S1 ).

To test in vitro cytotoxicity of CAR-T cells and ADC targeting CEACAM5, we successfully produced an analog of ADC SAR408701 derived from Sanofi (16). The ADC SAR408701 analog, produced by NJ Biopharmaceuticals, has a DAR of 3.7 as determined by reverse-phase (RP) chromatography coupled with mass spectrum (MS) analysis, and exhibits a homogenous folding as tested by size-exclusion chromatography (SEC) ( Supplementary Figure S1 ). The CEACAM5 binding specificity and affinity of in-house developed ADC SAR408701 analog were also confirmed (data not shown). The CEACAM5 surface expression was first screened in 7 lung cancer cell lines, 4 pancreatic cancer cell lines, and 3 other cancer cell lines ( Supplementary Figure S2 ). Unfortunately, total 7 lung cancer cell lines exhibited no CEACAM5 surface expression, so we generated CEACAM5-stably expressing DM4-sensitive (DM4^S) NSCLCs, H1299-CEACAM5 and A549- CEACAM5. CEACAM5 expression was confirmed by flow cytometry ( Figure 2A ). ADC SAR408701 analog was tested in dose-response cell viability assays using four DM4^S CEACAM5-positive cell lines (H1299-CEACAM5, A549-CEACAM5, HPAC, and HPAF-II) and two DM4^S CEACAM5-negative cell lines (H1299 and A549). Treatment with ADC SAR408701 analog reduced cell viability in a dose-dependent manner with all CEACAM5-positive cells ( Figures 2B, D ). However, the ADC SAR408701 analog also showed a killing activity in DM4^S CEACAM5-negative H1299 and A549 cell lines ( Figure 2B ) which may be attributed to the payload-induced toxicity independent of direct antigen-mediated internalization (39, 40). We next assessed the cytotoxicity of our anti-CEACAM5 CAR-T cells for DM4^S NSCLCs and PDACs in vitro. Our anti-CEACAM5 CAR-T cells were previously generated and evaluated in prostate cancer cells in vitro and in vivo (35). Anti-CEACAM5 CAR-T cells showed a potent cytotoxicity against CEACAM5-positive NSCLCs (H1299-CEACAM5 and A549-CEACAM5) ( Figure 2C ) and PDACs (HPAC and HPAF-II) ( Figure 2E ). Contrary to the ADC SAR408701 analog, CAR-T cells did not exhibit non-specific toxicity in CEACAM5-negative NSCLCs (H1299 and A549) ( Figure 2C ).

Second, we evaluated the effects of anti-CEACAM5 ADC SAR408701 analog and CAR-T cells on DM4-resistant (DM4^R) NSCLCs (H1975 and H2009). CEACAM5-expressing NSCLCs - H1975-CEACAM5 and H2009-CEACAM5, were constructed ( Figure 3A ). The DM4^R NSCLCs expressing or not expressing CEACAM5 showed a mild concentration-dependent response to treatment with the ADC SAR408701 analog (measured maximum of 20% cytotoxicity) ( Figure 3B ). In contrast, the anti-CEACAM5 CAR-T cells demonstrated a strong E:T ratio-dependent cytotoxicity for cells expressing CEACAM5 and no response to cells not expressing CEACAM5 ( Figure 3C ).

To further examine the therapeutic potentials of our CAR-T cells in vivo, we utilized mouse xenograft tumor models of CEACAM5-expressing DM4^S (A549-CEACAM5), DM4^R (H1975-CEACAM5) and CEACAM5-negative DM4^S (A549) NSCLCs. Anti-CEACAM5 CAR-T cells and ADC SAR408701 were administered as Figure 4A . In mice bearing DM4^S A549-CEACAM5, ADC SAR408701 analog treatment significantly reduced tumor growth and improved mouse survival in dose-dependent manner. CAR-T cells also showed a dose-dependent tumor growth inhibition ( Figure 4B ) and extended survival in CEACAM5-positive DM4^S A549-CEACAM5 ( Figure 4C ). However, a different result was observed in the DM4^R H1975-CEACAM5 tumors, where potent anti-tumor activity was observed for CAR-T cells-treated group, but not for the ADC SAR408701 analog treated group ( Figures 4D, E ). We sought further insights into the tumor growth inhibition activity of anti-CEACAM5 CAR-T and ADC SAR408701 analog by examining their activity in a CEACAM5-negative DM4^S A549 tumor model, because the ADC SAR408701 analog showed a CEACAM5-independent killing activity in vitro cell-based system. The high dose (10 mg/kg) of ADC SAR408701 analog suppressed A549 tumor growth (51% TGI) even in the absence of expressed CEACAM5, but anti-CEACAM5 CAR-T displayed no tumor growth inhibition ( Figures 4F, G ). The mouse body weight, monitored as an indicator of drug toxicity, was similar compared with vehicle group ( Supplementary Figure S3 ). These results indicate that our CAR-T cells therapy is effective and safe against NSCLCs and can be an alternative treatment strategy in ADC-non-responsive NSCLCs.