CDCP1-targeting ADC outperforms standard therapies in Ras-mutant pancreatic cancer

Abstract

RAS mutations are found in 10%–30% of various cancers and in up to 90% of pancreatic cancers, where they are associated with aggressive phenotypes, poor prognosis, and reduced overall survival. CUB domain containing protein 1 (CDCP1), a transcriptional target of activated RAS, is implicated in these cancers irrespective of the specific RAS mutation. Given the limited effectiveness of small-molecule inhibitors against mutant Ras-driven cancers, we developed a CDCP1-targeting antibody-drug conjugate (ADC). In this study, we demonstrate that CDCP1 overexpression significantly correlates with RAS mutations in pancreatic cancer. We generated and characterized a CDCP1-specific monoclonal antibody, 2G10, and conjugated it to the topoisomerase II inhibitor, PNU159682, to produce 2G10-PNU159682. The anti-tumor activity of this ADC was evaluated in vitro and in vivo using pancreatic cancer cell lines. 2G10-PNU159682 exhibited superior efficacy compared to MRTX1133 and sotorasib in G12D- and G12C-mutant cell lines. In a mouse xenograft model, 2G10-PNU159682 demonstrated robust anti-tumor activity against RAS-mutant pancreatic cancers, outperforming gemcitabine and FOLFIRINOX and achieving complete tumor remission for up to 100 days—even following relapse after standard chemotherapy. These findings support the potential of 2G10-PNU159682 as a promising therapeutic candidate for the treatment of Ras-mutant cancers.

Graphical abstract

Park and colleagues report that a CDCP1-targeting antibody-drug conjugate achieves durable tumor remission in Ras-mutant pancreatic cancer models, surpassing current therapies. This work highlights CDCP1 as a promising therapeutic target and suggests a new strategy to overcome resistance in Ras-driven malignancies.

Introduction

RAS proteins, GTPase family members, are encoded by HRAS, KRAS, and NRAS, which exhibit significant sequence homology and activate signaling networks regulating cell proliferation, differentiation, and survival.^1,2 In addition, RAS is frequently mutated in cancers, including pancreatic, colorectal, lung, and small intestinal cancers, with an overall frequency of 10%–30%. Mutant RAS proteins constitutively activate downstream signals, transforming normal cells into tumorigenic cells.^3,4 Gain-of-function missense mutations in RAS predominantly cluster at three hotspots codons 12, 13, and 61,⁵ resulting in enhanced GTP binding due to rapid nucleotide exchange and/or impaired GAP binding.⁶ Notably, patients with cancer harboring mutant RAS exhibit an aggressive phenotype, worse prognosis, and shorter overall survival (OS) compared to those with wild-type RAS.⁷

Direct targeting of RAS-mutant proteins was once considered impossible due to the lack of drug-binding pockets on their surface. Nonetheless, significant efforts have been made to develop inhibitors targeting RAS mutant signaling, including upstream molecules, direct targeting of RAS mutant proteins, and downstream effectors.⁸ AMG510, which targets the KRAS^G12C mutant, exhibited a significant objective response in advanced NSCLC and was approved for KRAS^G12C NSCLC patients by the US Food and Drug Administration.^9,10,11 Although AMG510 was the first approved targeted therapeutic for tumors with KRAS mutations, it is not effective in patients with other Ras mutations and has a narrow therapeutic scope, since the KRAS^G12C mutation, for instance, accounts for <2% of pancreatic cancer.^12,13 Therefore, there remains an unmet need to treat patients harboring RAS mutations.

CUB domain containing protein 1 (CDCP1), also known as CD318, SIMA135, gp140, and TRASK, is a type I transmembrane glycoprotein composed of 836 amino acids. It includes a signal peptide, three CUB domains, a transmembrane domain, and an intracellular domain with five tyrosine phosphorylation sites.¹⁴ The intracellular domain of CDCP1 can be phosphorylated by Src family kinases, including SRC, YES1, and FYN,^15,16,17 promoting interaction with PRKCD and leading to anoikis resistance.^16,18 Moreover, CDCP1 forms a heterodimer with HER2, promoting HER2-driven tumorigenesis and inducing trastuzumab resistance through SRC-kinase-mediated survival signaling in breast cancer.¹⁹ CDCP1 is overexpressed in various malignancies, including pancreatic, colorectal, lung, ovarian, prostate, renal, and breast cancers, and is associated with tumor progression, drug resistance, and poor survival prognosis.^{16,19,20,21,22,23,24,25,26,27,28,29} CDCP1 is initially synthesized as a full-length form and trafficked to the cell membrane. In certain cancer types, it is cleaved by extracellular proteases, such as matriptase and plasmin, into two fragments: the CUB1 domain and a truncated form (between 63 and 75 kDa) comprising the CUB2 and CUB3 domains. According to previous studies,^27,30 even after cleavage, the CUB1 fragment remains associated with the truncated CDCP1 molecule via non-covalent interactions, thereby preserving its structural and functional integrity. Recently, CDCP1 was shown to interact with CD6 on T and natural killer (NK) cells, and disrupting the CD6-CD318 interaction with an anti-CD6 antibody improved mouse survival in xenograft models of human breast and prostate cancer, suggesting CDCP1 may function as an immune checkpoint.^31,32

Interestingly, Martinko et al.³³ found that introduction of KRAS^G12V and activation of the MAPK pathway led to increased surface expression of CDCP1 in cancer cells. It has been shown that CDCP1 expression is upregulated regardless of the mutation type. In addition, the introduction of RAS mutations in NSCLC cell lines significantly increases CDCP1 expression.³⁴ Furthermore, EGFR activation by EGF or the formation of heterodimers between CDCP1 and EGFR in breast cancer activates RAS-mediated downstream signaling, which inhibits the proteasome-mediated degradation of CDCP1 and increases CDCP1 expression at the transcriptional level, resulting in the increased expression of CDCP1 on the surface of cancer cells.³⁵ These results suggest a potential vicious cycle between RAS activation and CDCP1 expression, which may contribute to enhanced tumor aggressiveness. Although the overall frequency of Ras mutations in pan-cancer is approximately 10%–30%, it rises dramatically to ∼90% in pancreatic cancer,^5,36 suggesting that this cancer type may be effectively treated using antibody-based therapeutics targeting CDCP1 in an RAS mutation-agnostic manner, without the need for mutation-specific small-molecule inhibitors.

Although a variety of therapeutic options are available for many other cancer types, treatment choices for pancreatic ductal adenocarcinoma (PDAC) remain limited to regimens such as FOLFIRINOX (FFX), 5-fluorouracil (5-FU), or gemcitabine. The clear unmet medical need to improve OS in PDAC underscores the urgency for novel therapeutic approaches. Accordingly, we prioritized the evaluation of our investigational CDCP1-targeting therapy in PDAC, with the strategic aim of broadening its clinical application to additional indications in the future. In this study, we generated and characterized a monoclonal antibody targeting CDCP1 and developed antibody-drug conjugates (ADCs) with various payloads. We then treated CDCP1-positive RAS mutant PDAC with these ADCs and investigated their therapeutic potential.

Results

CDCP1 expression in cancer

Previous studies have shown that CDCP1 is highly expressed in various cancers.^{21,24,25,26,27} Therefore, we analyzed the mRNA expression across multiple cancer types—including pancreatic, breast, ovarian, colon, and prostate cancers—using data from the TCGA and GTEx databases and compared tumor samples to their corresponding normal tissues. As shown by in silico analysis (Figure 1A), CDCP1 transcript levels are significantly elevated in PDAC, breast cancer, ovarian cancer, and colon cancer compared to adjacent non-tumor tissues, but not in prostate cancer. The in vitro transcript analysis using various cancer cell lines, including PDAC, breast cancer, ovarian cancer, colon cancer, and prostate cancer, clearly showed that most cancer cell lines express high levels of CDCP1, except for HL-60, a CDCP1-negative cell line (Figure S1). Western blot analysis further supported the correlation between mRNA and protein levels, although the ratio between full-length and cleaved CDCP1 differed (Figure 1B).

Figure 1.

CDCP1 is highly overexpressed in KRAS-mutant pancreatic cancer

(A) The mRNA levels of CDCP1 in different types of human cancer (TCGA) and normal tissue (GTEx and TCGA) were analyzed in silico. ∗∗∗ indicates a significant difference compared to the respective normal tissue. (B) CDCP1 protein levels were analyzed by western blot. HL-60, an acute myeloid leukemia cell line, was used as a CDCP1-negative cell line. The arrowhead indicates a truncated CDCP1 protein. Alpha-tubulin was used as a loading control. Abbreviations: PDAC, pancreatic adenocarcinoma; BRCA, breast cancer; EOC, epithelial ovarian cancer; COAD, colon adenocarcinoma; PRAD, prostate adenocarcinoma. (C) In silico comparison of CDCP1 mRNA levels in KRAS wild-type and mutant type pancreatic cancer (n = 179, TCGA). Seven primary tumor cases and one metastatic case were excluded from analysis due to no information of KRAS. ∗∗∗ vs. wild-type KRAS cancer. (D) Correlation between mutant KRAS (mKRAS) and CDCP1 was examined in silico using TCGA dataset and plotted using Pearson or Spearman correlation coefficient with corresponding p values. (E) Kaplan-Meier survival curve based on in silico analysis of TCGA datasets, comparing OS between high and low CDCP1 expression, and between wild-type and mutation KRAS status in patients with PDAC. High CDCP1 expression or KRAS mutation was associated with worse OS. (F) Representative images of IHC staining of CDCP1 using PDX of PDAC patients. Intensity scores were determined as 0 (negative), 1 (weak intensity), 2 (moderate intensity), or 3 (strong intensity). (G) Populations were grouped by intensity score and plotted with IHC score. ∗∗∗ vs. their respective control group with intensity score 1 or 2. ∗∗∗p < 0.001.

KRAS mutations induce constitutive activation of downstream signaling, leading to increased CDCP1 expression.³³ Although several CDCP1-high cancer cell lines, such as PC-3, do not contain any mutated KRAS, and on the other hand, MDA-MB-231 and HCT-116 p53+/+ harbor KRAS mutations (Figures S1 and 1B), PDAC was chosen as the primary model for this study due to the significantly higher prevalence of KRAS mutations in pancreatic cancer (∼90%) compared to breast (∼4%), ovarian (∼14%), colorectal (∼44%), and prostate (∼8%) cancers.^1,3,4 This distinct mutational profile underscores the suitability of PDAC for investigating CDCP1 regulation in a mutant-KRAS-driven context. Consistently, ectopic expression of mutant RAS in PDAC cells led to CDCP1 overexpression, confirming a functional relationship between KRAS mutation and CDCP1 expression in vitro (Figure S2). Given that CDCP1 is a downstream target gene of activated RAS,^30,33,37 we sought to further investigate whether KRAS mutations are correlated with CDCP1 overexpression in patient-derived tissues as well.

To that end, in silico comparison of transcript levels revealed that CDCP1 expression was significantly higher in PDAC with KRAS mutations compared to those with wild-type KRAS (Figure 1C), thereby further highlighting the crucial functional interplay between mutant KRAS and CDCP1. The G12D, G12V, and G12R mutations in KRAS are known to account for over 60% of KRAS mutations in pancreatic cancer,¹³ which is consistent with our analysis of the TCGA dataset (Figure S3A). Interestingly, although CDCP1 expression did not significantly differ among the various RAS mutation subtypes (Figure S3B), it showed a significant positive correlation with mutant KRAS expression in the in silico analysis (Figure 1D), further supporting the notion that activated KRAS upregulates CDCP1 expression.

Additionally, in silico survival probability analysis revealed that high expression of either KRAS or CDCP1 was significantly associated with poor prognosis (Figure 1E). Immunohistochemical (IHC) analysis using a tumor microarray derived from patient-derived xenograft (PDX) samples of pancreatic cancer revealed that 92% (46/50) of cases harboring RAS mutations were positive for CDCP1 (Figure 1F; Table S2). Moreover, 67% (42/63) of CDCP1-positive samples exhibited strong staining intensity (3+) (Figure 1G; Table S2).

Generation and characterization of 2G10 targeting CDCP1

To treat RAS-mutant cancer, we used an ADC targeting CDCP1. We generated a monoclonal antibody targeting CDCP1 and investigated whether 2G10 could specifically bind to CDCP1. Surface plasmon resonance (SPR) analysis showed that 2G10 binds to human and monkey CDCP1 with affinities of 1.7 × 10⁻¹⁰ M (ka = 2.5 × 10⁵, kd = 4.16 × 10⁻⁵) and 4.9 × 10⁻¹⁰ M (ka = 7.81 × 10⁵, kd = 3.92 × 10⁻⁵), respectively (Figure 2A), but not to mouse CDCP1 (results not shown). FACS analysis using various cancer cell lines showed that 2G10 specifically binds to CDCP1-positive cells, but not to CDCP1-negative cells (Figures 2B and S4A). In addition, knockdown analysis using CDCP1 small interfering RNA (siRNA) further supports that 2G10 specifically binds to CDCP1 (Figures 2C, S4B and S4C). Off-target binding analysis of 2G10 using a ProtoArray chip containing over 21,000 full-length human recombinant proteins revealed weak binding to leucine zipper and ICAT-homologous-domain-containing proteins, although the interactions were not significant (data not shown). Furthermore, FACS analysis after transfection of various human CDCP1 deletion mutants into COS-7 cells, a CDCP1 negative cell line, and ELISA showed that 2G10 bound to amino acids 30–341, including the CUB1 domain (Figure 2D).

Figure 2.

Generation and characterization of 2G10 antibody targeting CDCP1

(A) SPR analysis of 2G10 antibody. CDCP1 proteins from human and monkey were immobilized onto the chips as described in the methods. The 2G10 antibody was injected as an analyte in a dose-dependent manner. The K_D value was evaluated using the Scrubber2 software and Biacore T200 Evaluation software v.3.2. (B and C) Determination of the specificity of 2G10 antibody to CDCP1. (B) CDCP1-positive or -negative cells were incubated with 2G10 antibody, followed by FACS analysis. (C) The indicated cells were transfected with control or CDCP1 siRNA for 48 h. The binding of 2G10 antibody to the indicated cells was analyzed by FACS. The images shown are representative of the siRNA experiments from three times independent replicates. (D) Identification of the CDCP1 binding domain of 2G10 antibody. A schematic diagram shows the construction of CDCP1 deletion mutants (upper). The indicated CDCP1 deletion constructs were cloned into the pCMV-3Tag-3a vector containing FLAG. Plasmids were transfected into COS-7 cells, and their expression was confirmed by western blotting using an anti-FLAG antibody (middle left). 2G10 antibody (0.5 μg/mL) was incubated with COS-7 cells expressing CDCP1 constructs, and the binding of 2G10 antibody was analyzed by FACS (middle right). ∗∗ vs. NT. The results represent means ± SEM of at least three independent experiments. NT stands for non-transfected. The protein including the CUB1 domain (F30—S341 aa, 50 ng/well) was coated to 96-well plates, and the binding of 2G10 antibody was examined at the indicated concentrations (bottom). Normal mouse IgG was used as a negative control. ∗∗p < 0.01.

Next, we investigated whether 2G10 antibody affects CDCP1 protein expression. As shown in Figure S5A, treatment with 2G10 antibody led to a time-dependent decrease in CDCP1 protein levels. CDCP1 mRNA expression remained unchanged following treatment with the 2G10 antibody (Figure S5B), suggesting that the observed reduction in CDCP1 protein levels is mediated through a post-translational mechanism. To assess the feasibility of using 2G10 as an ADC, we evaluated its internalization efficiency in a pancreatic cancer cell line. Confocal microscopy analysis revealed that the 2G10 antibody was internalized and localized to lysosomes within 30 min of treatment (Figure 3A). FACS analysis further confirmed that the 2G10 antibody was internalized by over 70% of pancreatic cancer cells (Figure 3B), supporting its potential utility as an ADC.

Figure 3.

Internalization of the 2G10 antibody in pancreatic cancer cells

(A) Cells were seeded onto microscope cover glasses in a 12-well culture plates: PANC-1 (1 × 10⁵ cells), AsPC-1 (2 × 10⁵ cells), BxPC-3 (2 × 10⁵ cells), and MIA PaCa-2 (0.5 × 10⁵ cells). After blocking with a blocker for 30 min, cells were treated for 1 h at 4°C with 2G10 antibody (10 μg/mL) conjugated with the anti-mouse IgG-conjugated Alexa 488 (10 μg/mL). Lysosomes were stained with anti-LAMP-1 for 1 h, and the nuclei were counterstained with DAPI (100 nM) at room temperature. The fluorescent images were captured using a confocal microscope at 60× magnification. The experiments were repeated independently at least three times. (B) Pancreatic cancer cells were pre-incubated with CHX (75 μg/mL) and blocked with Fc blocker for 10 min to inhibit the Fc receptor-mediated internalization. The cells were then incubated in the presence or absence of 2G10 antibody (100 ng/mL) at 4°C or 37°C for 2 h and followed by flow cytometry analysis. The fluorescent signal of the 2G10 antibody/CDCP1 complex on the cell surface decreased after incubation at 37°C.

2G10-ADC exhibits anti-tumor efficacy in vitro and in vivo

Further, to find the optimal payload to treat pancreatic cancer, we investigated and compared the in silico expression of genes involved in cell division (TUBB3), transcription (POLR2A), DNA replication and recombination (DNA topoisomerase I, TOP1; DNA topoisomerase II alpha, TOP2A; DNA topoisomerase II beta, TOP2B), DNA repair (ERCC6 and ERCC8), and drug resistance (ABCB1, ABCG2, ABCC1, ABCC2, ABCC3, ABCC4, and ABCC5). The expression of the examined genes was significantly upregulated in cancer tissues compared to that in normal tissues (Figure 4A). Moreover, in silico correlation analysis revealed that ABCC1, ABCG2, TUBB3, TOP2A, and TOP1 were significantly correlated with the expression of both CDCP1 and KRAS. Additionally, TOP1 and TOP2A showed weak but positive correlations with several drug-resistance-associated genes, including ABCC1, ABCC4, ABCB1, ABCG2, and TUBB3 (Figure 4B). These results suggest that inhibitors targeting microtubules, TOP1 or TOP2A that can be effluxed by ABCC1, ABCC4, ABCB1, or ABCG2, should be excluded from consideration. Since TUBB3, TOP1, and TOP2A are highly correlated with KRAS and CDCP1 expression, we examined the cytotoxic activity of linker-payloads using MMAE, αAmanitin, Duocarmycin SA, or PNU159682. Half maximal inhibitory concentration (IC₅₀) values ranged from 43.51 to 3,438.9 nM for MMAE, from 13.67 to 218.5 nM for Duocarmycin SA, and from 1.57 to 54.79 nM for PNU159682, across CDCP1-positive and -negative cancer cell lines (Figure S6; Table S3). Interestingly, αAmanitin did not exhibit an IC₅₀ value up to 2.7 μM. We then generated ADCs using the aforementioned payloads described above (Figure S7A) and compared their binding affinities to monkey and human CDCP1. As shown in Figure S7B, ELISA revealed no significant differences in binding affinity between the unconjugated antibody and its corresponding ADCs. The in vitro cytotoxicity assay demonstrated that 2G10-PNU159682 and 2G10-Duocarmycin SA exhibited potent anti-cancer activity, showing over 3,000-fold selectivity for CDCP1-positive cancer cell lines compared to CDCP1-negative cell lines. 2G10-MMAE exhibited a narrow therapeutic index between CDCP1-positive and -negative cancer cell lines. In contrast, 2G10-αAmanitin showed no significant cytotoxicity in either cell type (Figure 4C; Table S4).

Figure 4.

CDCP1 is highly correlated with expression levels of MDRs and topoisomerases in PDAC

(A) The mRNA levels of various target genes, including those involved in cell division (TUBB3), transcription (POLR2A), DNA replication and recombination (TOP1, TOP2A, and TOP2B), DNA repair (ERCC6 and ERCC8), and genes involved in drug resistance (ABCB1, ABCG2, ABCC1, ABCC2, ABCC3, ABCC4, and ABCC5) were analyzed in silico using data from TCGA and GTEx databases. ∗∗∗ vs. their corresponding normal tissue. (B) In silico correlation analysis of CDCP1 and payload target genes or MDR gene expression levels in TCGA database. (C) IC₅₀ values were determined in CDCP1-positive cell lines (PANC-1, AsPC-1, BxPC-3, MIA PaCa-2, and MDA-MB-468) and CDCP1-negative cell lines (CT26 and MCF-7) with the unconjugated 2G10 antibody and various 2G10-ADCs conjugated with MMAE, αAmanitin, Duocarmycin SA, or PNU159682. Cells were treated with serially diluted concentrations of the indicated test articles for 3 days. The cells were stained with Hoechst 33342 (3.3 μM), at 37°C for 60 min and analyzed using a Celigo Imaging Cytometer. The results represent the mean ± SEM of at least three independent experiments. ∗∗∗p < 0.001.

Most payloads currently in preclinical or clinical development for targeting PDAC are microtubule or TOP1 inhibitors (Table S5). In addition, treatment regimens for PDAC patients commonly include FFX or combinations of gemcitabine with irinotecan, capecitabine, or erlotinib.^38,39,40,41 Therefore, we explored the therapeutic feasibility of using payloads with mechanisms distinct from those of agents previously developed or currently under investigation. DNA topoisomerase II (TOP2) expression is known to be upregulated by activated RAS signaling.^42,43 As shown in Figures 4A and 4B, TOP2A expression was significantly elevated in pancreatic tumors compared to adjacent non-tumor tissues and was strongly associated with reduced OS and disease-free survival.⁴⁴ Since PDAC exhibits high expression levels of multi-drug resistance (MDR) proteins that efflux MMAE from cancer cells, and TOP1 inhibitors are already in clinical use, we selected 2G10-PNU159682 as a candidate due to its superior in vitro cytotoxicity compared to 2G10-Duocarmycin SA. An ADC was generated with an average drug-antibody ratio (DAR) of 4.95 (Figure S8A). As shown in Figure 5A, SDS-PAGE analysis revealed that conjugation of PNU159682 to the 2G10 antibody induced a slight size shift. FACS analysis showed no difference in cell binding between the 2G10 antibody and 2G10-PNU159682 (Figure 5B). ELISA also exhibited no difference in the binding affinities of the 2G10 antibody and 2G10-PNU159682 to monkey and human CDCP1 (Figure S8B). Cell-cycle analysis and apoptosis assays demonstrated that 2G10-PNU159682 induced S phase arrest and activation of caspase 3/7 in CDCP1-positive PANC-1 cells, but not in MCF-7 cells, a CDCP1-negative cell, indicating that 2G10-PNU159682 induces CDCP1-dependent apoptosis (Figures 5C and 5D). An in vitro cytotoxicity assay using 2G10-PNU159682, irinotecan, gemcitabine, MRTX1133, or sotorasib was performed to compare the IC₅₀ values. 2G10-PNU159682 exhibited potent anti-tumor activity with an IC₅₀ value ranging from 3.23 to 40.32 × 10⁻¹² M, but not in MCF-7, CDCP1-negative cell line, supporting that 2G10-PNU159682 induces CDCP1-dependent cytotoxicity. Gemcitabine and irinotecan exhibited cytotoxicity at single-digit micromolar concentrations, respectively. However, their cytotoxic activity was independent of CDCP1 expression (Figure 5E; Table S6). Sotorasib, a KRAS^G12C-selective inhibitor, and MRTX1133, a KRAS^G12D-selective inhibitor, exhibited target-dependent cytotoxicity in MIA PaCa-2 (KRAS^G12C) and AsPC-1 (KRAS^G12D) cells, respectively, but with marginal activity (Figure 5E; Table S6).

Figure 5.

Generation and characterization of 2G10-PNU159682

(A) The unconjugated antibody and 2G10-PNU159682 were compared using non-reducing and reducing SDS-PAGE. R and C stand for reduced antibody and conjugated antibody with PNU159682, respectively. (B) The binding affinity of naked 2G10 antibody and 2G10-PNU159682 to pancreatic cancer cells was compared using flow cytometry. (C) PANC-1 or MCF-7 cells were treated with vehicle, isotype control antibody, 2G10 antibody (0.1 μg/mL), isotype-PNU159682 (0.1 μg/mL), or 2G10-PNU159682 (0.1 μg/mL) for 24, 36, and 48 h. Cells were then fixed and stained with propidium iodide, followed by cell-cycle analysis using a Celigo Imaging Cytometer (∗, ∗∗, and ∗∗∗ vs. their respective corresponding vehicle; #, ##, ### vs. isotype-PNU159682). 2G10-PNU159682 increased the S phase cell population in PANC-1. MCF-7 cells were used as the CDCP1-negative cells. The results represent mean ± SD from at least three independent experiments. (D) For apoptosis assay, PANC-1 or MCF-7 cells were seeded into 96-well plates and incubated with vehicle, isotype control antibody (0.1 μg/mL), 2G10 antibody (0.1 μg/mL), isotype-PNU159682 (0.1 μg/mL), or 2G10-PNU159682 (0.1 μg/mL) for 36 h. Cells were stained with caspase 3/7 reagent (2 μM) and Hoechst 33342 (10 μM) and analyzed using a Celigo Imaging Cytometer (∗∗∗, ### vs. their respective corresponding control, vehicle or isotype-PNU159682). MCF-7 cells were used as the CDCP1-negative cells. The results have been represented as mean ± SD from at least three independent experiments. (E) IC₅₀ values of the indicated treatments in pancreatic cancer cells or breast cancer cells were evaluated. Cells were treated with serially diluted concentrations of the indicated test articles for 3 days. The cells were stained with Hoechst 33342 (3.3 μM) at 37°C for 60 min and analyzed using a Celigo Imaging Cytometer. The results represent mean ± SEM of at least three independent experiments. ∗, #p < 0.05; ∗∗, ##p < 0.01; ∗∗∗, ###p < 0.001.

Next, the in vivo efficacy of 2G10-PNU159682 was examined using mouse models xenografted with CDCP1-positive cancer cell lines, PANC-1 (KRAS^G12D) and MIA PcCa-2 (KRAS^G12C), as well as the CDCP1-negative cancer cell line HL-60. In both PANC-1 and MIA PaCa-2 xenograft models, 2G10-PNU159682 suppressed tumor growth in a dose-dependent manner without affecting body weight (Figures 6A, 6B and S9). While treatment with 2G10-PNU159682 at 0.1 mg/kg and 0.2 mg/kg in PANC-1 xenografts induced tumor stasis for approximately 60 days followed by regrowth, administration at 0.5 mg/kg resulted in complete remission lasting up to 80 days. In MIA PaCa-2 xenograft model, all treated doses, including 0.1 mg/kg, 0.2 mg/kg, and 0.5 mg/kg, induced complete remission lasting up to 50 days. Interestingly, the 2G10 antibody alone exhibited significant tumor growth inhibition in PANC-1 xenografts and partial suppression in MIA PaCa-2 tumors (Figures 6A and 6B). No significant antitumor effect was observed in the HL-60 xenograft model (Figures 6C and S9), supporting the conclusion that 2G10-PNU159682 specifically suppresses tumor growth in CDCP1-positive models, irrespective of RAS mutation status. Currently, various chemotherapeutic regimens are employed for the treatment of PDAC, including gemcitabine (GEM), GEM in combination with nab-paclitaxel, FFX, NALIRIFOX (liposomal irinotecan with FOLFOX), and capecitabine.^45,46,47 To evaluate the combinatorial effects of 2G10-PNU159682 with standard chemotherapy, we compared the therapeutic efficacy of low-dose 2G10-PNU159682 combined with either GEM or FFX against monotherapy with GEM or FFX. As shown in Figures 6D–6I, while monotherapy with GEM or FFX partially suppressed tumor growth in both PANC-1 and AsPC-1 models, their combinations with low-dose 2G10-PNU159682 resulted in complete remission until day 70 in PANC-1 cells and tumor stasis until day 40 in AsPC-1, followed by regrowth. Notably, although low-dose 2G10-PNU159682 monotherapy induced tumor stasis for 60 days in the PANC-1 model (Figure 6A), its combination with GEM or FFX extended complete remission up to 70 days, followed by tumor regrowth (Figures 6E, 6H and S10). In addition, cessation of GEM or FFX induced rapid tumor regrowth. Surprisingly, the administration of 2G10-PNU159682 (0.5 mg/kg), even after tumor regrowth due to cessation of GEM or FFX, induced complete remission lasting up to day 100 (Figures 6E, 6F, 6H, 6I, and S10), suggesting that 2G10-PNU159682 might be competitive not only as a first-line treatment but also for patients with recurrence after first-line or second-line treatment.

Figure 6.

2G10-PNU159682 suppresses PDAC tumor growth both as a monotherapy and in combination with standard chemotherapy

(A–C) Mice with established tumors were randomized into treatment groups (n = 5–7) when tumors volume reached 150–250 mm³. The mice were intravenously administered the indicated test articles on days 0, 7, and 14. Isotype control antibody, isotype-PNU159682, and 2G10 antibody were administered at 0.5 mg/kg. PANC-1- or MIA PaCa-2-implanted groups were used as CDCP1-positive cancer model. HL-60-implanted groups were used as a CDCP1-negative control. Green arrows indicate the administration of test articles. ∗, ∗∗, ∗∗∗ vs. vehicle; #, ## vs. isotype-PNU159682. (D–I) When tumors volume reached ∼170 mm³, mice were randomized into treatment groups: vehicle (n = 10), GEM or FFX (n = 10), and combination of GEM or FFX with 2G10-PNU159682 (n = 10). (D) Drug administration schedules are depicted in schematic diagrams. (E and F) When the volume of tumors reached ∼200 mm³, mice were randomized into different treatment groups: vehicle (n = 10), GEM (n = 10), and a combination of GEM and 2G10-PNU159682 (n = 10). GEM (50 mg/kg, light blue arrow, i.p.) was administered on days 0, 3, 7, and 10, and 2G10-PNU159682 (0.1 mg/kg, blue arrow, i.v.) was administered on days 0, 7, and 14. Upon tumor relapse (volume 300–400 mm³) after GEM cessation, mice were re-randomized into vehicle and 2G10-PNU159682. 2G10-PNU159682 (0.5 mg/kg) was intravenously administered at the indicated days (deep blue arrowhead). ∗, ∗∗∗ vs. vehicle; ##, ### vs. GEM. (G) Schematic diagram for drug administration. (H and I) When the volume of tumors reached ∼200 mm³, mice were randomized into vehicle (n = 10), FFX (n = 10), and combination of FFX and 2G10-PNU159682 (n = 10). FFX (leucovorin at 50 mg/kg; 5-FU at 25 mg/kg; irinotecan at 25 mg/kg; oxaliplatin at 2.5 mg/kg, light purple arrow, i.p.) was administered on days 0 and 7, and 2G10-PNU159682 (0.1 mg/kg, purple arrow, i.v.) was administered on days 0, 7, and 14. Upon tumor relapse (volume 300–400 mm³) after FFX cessation, mice were re-randomized into vehicle and 2G10-PNU159682. 2G10-PNU159682 (0.5 mg/kg) was intravenously administered at the indicated days (deep purple arrowhead). i.p and i.v. stand for intraperitoneal injection and intravenous injection, respectively. ∗∗∗ vs. vehicle; #, ## vs. FFX. ∗, #p < 0.05; ∗∗, ##p < 0.01; ∗∗∗, ###p < 0.001.

Discussion

In this study, we demonstrated that CDCP1 overexpression positively correlates with Ras mutations in various cancers. Interestingly, some cancers exhibit CDCP1 overexpression regardless of Ras mutation status. Although there was no significant difference in CDCP1 expression compared to normal tissue in prostate cancer (Figure 1A), CDCP1 was found to be overexpressed in the prostate cancer cell lines PC-3 and LNCaP (Figure 1C), both of which are deficient in the tumor suppressor PTEN (Figure S11). Recent studies have reported that loss of PTEN in prostate cancer is strongly correlated with CDCP1 overexpression,^21,48 suggesting that CDCP1-targeting ADCs could be applicable to cancers harboring PTEN loss or RAS mutations. Although the possibility that KRAS mutations contribute to CDCP1 overexpression in other cancer types cannot be excluded, our findings highlight the genetic and clinical relevance of this axis specifically in PDAC. Further investigation across additional tumor types may provide broader insights into this regulatory relationship.

Since irinotecan, an SOC drug for pancreatic cancer, is a TOP1 inhibitor, we reasoned that utilizing a payload from the same class may not be a strategically suitable approach. Additionally, since ADC targeting CDCP1 can be applied to breast cancer and colon cancer, we considered that a TOP1 inhibitor might not be the best option for breast cancer and colon cancer treatments, as Enhertu (trastuzumab-deruxtecan), containing a TOP1 inhibitor, is already used in breast cancer,^49,50 and Trodelvy (sacituzumab-govitecan) which also contains a TOP1 inhibitor, is also being used to treat breast cancer.⁵¹ Notably, Trodelvy has demonstrated no therapeutic efficacy in TROP2-positive colon cancer,^52,53 suggesting that SN-38, a TOP1 inhibitor, is ineffective in suppressing colon cancer growth. Furthermore, one of the SOCs in pancreatic cancer is a combination of GEM with paclitaxel, a microtubule inhibitor. Therefore, we anticipated that selecting a payload other than microtubule or TOP1 inhibitors could enhance therapeutic efficacy while broadening treatment options and potential indications. Based on our screening of various payloads, we identified PNU159682, a TOP2 inhibitor, as the most promising candidate. Subsequent studies confirmed that 2G10-PNU159682 effectively suppressed tumor growth (Figures 5E and 6).

The critical requirements for ADC development include high binding affinity to the target antigen and efficient internalization into cancer cells after antibody-antigen complex formation. The 2G10 antibody used in this study exhibited K_D of 1.7 × 10⁻¹⁰ M and internalization efficiency over 70% in pancreatic cancer cells (Figures 2A and 3B), supporting its potential as an ADC. For an ADC to exert cytotoxic activity, the ADC-antigen complex should be trafficked to lysosome, where proteolytic degradation releases the cytotoxic payload to induce apoptosis. Although internalization of the antibody was confirmed, further evidence is needed to demonstrate lysosomal targeting of the ADC-antigen complex. This can be validated by co-staining with a lysosome marker and observing degradation of the target antigen. Interestingly, while treatment with cycloheximide (CHX) indicated that the half-life of CDCP1 protein in PANC-1 was between 4 and 8 h (Figure S12), treatment with the 2G10 antibody induced scarce expression of CDCP1 within 3 h (Figure S5), suggesting that the 2G10 antibody-CDCP1 complex is rapidly internalized and targeted to lysosomes for degradation. These results, along with the data shown in Figure 3, further support the suitability of the 2G10 antibody for ADC development.

Solid tumors are often exposed to hypoxia because of their high proliferation rates and active metabolism.⁵⁴ Hypoxia induces CDCP1 expression and tyrosine phosphorylation by hypoxia-inducible factor (HIF)-2 in clear cell renal cell carcinoma (ccRCC), correlating with poor OS.²⁸ Von Hippel-Lindau tumor suppressor (VHL) induces proteasome-dependent degradation of HIF under normoxia.⁵⁵ However, hypoxia stabilizes HIF, leading to the induction of various target genes, including vascular endothelial growth factor and CDCP1. Given that loss of function of the tumor suppressor gene VHL—a hallmark of ccRCC—results in constitutive HIF activation,⁵⁶ the overexpression of CDCP1 in this cancer type is not surprising. Additionally, the exposure of MDA-MB-468 and BxPC-3 cells to CoCl₂, a hypoxia mimetic agent, increased CDCP1 expression (Figure S13), further supporting that CDCP1 is a target gene of HIF. These results suggest that ADC targeting CDCP1 can be used to treat VHL-mutant renal cancer.

One of the advantages of ADCs in cancer therapy is the targeted delivery of cytotoxic payload without adverse effects on normal tissue. However, normal tissue expression of ADC target proteins can lead to amplified adverse effects. IHC staining of CDCP1 showed restricted expression in normal tissues and cytoplasmic localization, except in epithelial cells, which showed cytoplasmic membrane localization.⁵⁷ Additionally, there was no cross-reactivity with bone marrow, reducing concerns about myelosuppression in clinical trials. The 2G10 antibody used in this study also exhibited CDCP1-specific binding as described in Figures 2 and S4. Therefore, CDCP1-targeting ADCs are expected to achieve a favorable therapeutic index, defined as the ratio between the dose inducing significant toxicity and the dose required to produce a therapeutic effect.

Activation of RAS signaling increases the TOP2 expression via MEK,^42,58 suggesting that cancers harboring RAS mutations exhibit high TOP2 expression. Notably, the expression of TOP2 is significantly increased in pancreatic cancer harboring RAS mutations in 90% of cases, which correlates with tumor metastasis and poor prognosis.^59,60 Interestingly, human tumor cell lines with activated RAS oncogenes are more sensitive to TOP2 inhibitors than are wild-type RAS cells.⁶¹ Furthermore, since pancreatic cancer exhibits overexpression of various MDR genes and PNU159682, a TOP2 inhibitor, is not a substrate of MDR proteins, it seems to be a reasonable payload to treat RAS-mutant cancer.⁶² There may be concern about systemic adverse events if CDCP1 is expressed in endothelial cells. RT-qPCR and western blot analyses showed that CDCP1 was not expressed in endothelial cells (Figure S14), suggesting that ADCs targeting CDCP1 are not involved in endothelial-cell-mediated systemic adverse events, at least in part.

While sotorasib exhibited a two-digit nanomolar IC₅₀ value in MIA PaCa-2 cells (G12C mutant), it showed a micromolar IC₅₀ value in BxPC-3 cells, which are known to express wild-type KRAS (Figure 5E; Table S6). Interestingly, sotorasib did not exhibit significant IC₅₀ values in PANC-1 and AsPC-1 cells, which are G12D mutants. In contrast, MRTX1133 exhibited a single-digit nanomolar IC₅₀ value in PANC-1 and AsPC-1 cells, but only a marginal IC₅₀ value in MIA PaCa-2 and BxPC-3 cells (Figure 5E; Table S1 and S6). Surprisingly, 2G10-PNU159682 exhibited a picomolar IC₅₀ value in all tested cancer cells, irrespective of the Ras mutation, and showed more than a 2,500-fold difference in cytotoxicity in CDCP1-positive cancer cells compared to CDCP1-negative cancer cells. Additionally, 2G10-PNU159682 exhibited superior efficacy in tumor regression compared to GEM or FFX and achieved complete remission, even in relapsed tumors after cessation of GEM or FFX (Figure 6). Although body weight remained stable, comprehensive evaluations beyond tumor suppression, including clinical observations and organ-specific toxicity assessments, were not performed. For the successful preclinical and clinical development of this potential ADC therapeutic, it is imperative to conduct thorough safety assessments encompassing serum biochemistry, evaluation of myelosuppression, and detailed histopathological examination of major organs. Taken together, these findings suggest that 2G10-PNU159682 could be further developed as a therapeutic agent against Ras mutations.

Materials and methods

Cell culture

Twenty-five different types of cell lines were used in this study (Table S1). HL-60, AsPC-1, BxPC-3, COS-7, MDA-MB-231, MDA-MB-453, MDA-MB-468, DU4475, SK-OV-3, OV-90, UWB1.289, SW626, SW480, SW620, HCT116, p53 wild-type, PC-3, PANC-1, CT26, DU145, LNCaP, and MJ cells were purchased from the American Type Culture Collection (Manassas, VA, USA). MIA PaCa-2 and MCF-7 were purchased from the Korean Cell Line Bank. Human endothelial primary cells (EPC and HUVEC) were kindly provided by Dr. Wonhee Suh (University of Chung-Ang, Korea) and cultured with EGM-2 Endothelial Cell Growth Medium-2 BulletKit (Lonza, Basel, Switzerland). HL-60, AsPC-1, BxPC-3, DU4475, SK-OV-3, UWB1.289, MCF-7, CT26, and LNCaP were cultured in Rosewell Park Memorial Institute-1640 medium (Cytiva, Amersham, UK) supplemented with 10% fetal bovine serum (FBS, Cytiva) and 1% penicillin/streptomycin (P/S, Cytiva). PANC-1, COS-7, MDA-MB-231, MDA-MB-453, MDA-MB-468, SW626, SW620, SW480, DU145, HCT116 p53 wild-type, and PC-3 cells were maintained in DMEM (Cytiva) with high glucose, 10% FBS, and 1% P/S. MIA PaCa-2 cells were cultured in DMEM supplemented with 10% FBS, 0.125% heat-inactivated horse serum (Sigma-Aldrich, St. Louis, MO, USA), and 1% P/S. OV-90 cells were cultured in a 1:1 (v/v) mixture of MCDB-105 (Sigma-Aldrich) and Medium 199 (Cytiva) with 15% FBS and 1% P/S. The MJ cells were maintained in Iscove’s Modified Dulbecco’s medium (Welgene, Gyeongsan, Korea) containing 20% FBS. All cells were incubated at 37°C in a humidified atmosphere incubator (Sanyo, Osaka, Japan) containing 95% air and 5% CO₂.

Antibody generation

All animal experiments were approved by the Institutional Animal Care and Use Committee of Ajou University (IACUC 2016-0044). To produce a monoclonal antibody against human CDCP1, eight-week BALB/c (Orient Bio, Seongnam, Korea) were subcutaneously or intraperitoneally immunized with 30 μg of rhesus CDCP1 recombinant protein (Sino Biological, Beijing, China). After confirming anti-CDCP1 antibody production by ELISA, hybridoma was generated by fusing F0 cells (ATCC) with splenocytes using a 50% (w/v) polyethylene glycol solution (PEG, Sigma-Aldrich). Positive clones were selected with HAT Media Supplement Hybri-Max (Sigma-Aldrich) and subsequently rescued with HT Media Supplement Hybri-Max (Sigma-Aldrich). Positive clones were selected by ELISA, monoclonal cells were generated by limiting dilution, and mouse antibodies were purified using Protein A/G Plus agarose chromatography (Santa Cruz Biotechnology, Dallas, TX, USA). Antibody isotyping was performed using a Monoclonal Isotyping Kit I-HRP/ABTS (Thermo Fisher Scientific, Waltham, MA, USA).

Specificity assessment and domain mapping analysis

An siRNA sequence targeting CDCP1 was synthesized as a duplex with sense and antisense oligomers (Bioneer, Daejeon, Korea), whereas a negative control siRNA was also obtained from Bioneer. The siRNA sequence was as follows: #1, 5′-CAUCGAGUC UGUGUUUGAGGGUGAA-3′; #2, 5′-CACAGCUUCUGGGUCAACAUCUCUA-3′; #3, 5′- CCGCUG UGGGUAUCUACAAUGACAA-3′. PANC-1 (2 × 10⁵ cells), AsPC-1 (3 × 10⁵ cells), BxPC-3 (3 × 10⁵ cells), and MIA PaCa-2 (1 × 10⁵ cells) were seeded into a 6-well plate and transfected with 20 nM siRNA for 48 h using Lipofectamine RNAi MAX (Invitrogen, Waltham, USA) according to manufacturer’s protocol. Knockdown of CDCP1 expression was confirmed by immunoblotting, RT-qPCR, and FACS analysis. For domain mapping analysis, full-length CDCP1 (M1-E835), ΔCUB 1-Flag (M1-A29, R368-E835), or Δ CUB 1, Δ CUB 2-Flag (M1-A29, G545-E835) was cloned into pCMV-3Tag-3a vector. COS-7, a CDCP1-negative cell line, was seeded at a density of 5 × 10⁵ cells in a 60 mm² culture dish. The cells were transfected with full-length or mutant CDCP1 using polyethyleneimine (Polysciences, Warrington, PA, USA) and incubated for 48 h. CDCP1 expression was confirmed using immunoblotting, RT-qPCR, and FACS analysis.

Surface plasmon resonance

SPR analysis was performed using a PEG chip on the SR7500DC system (Reichert Technologies, Depew, MA, USA) or a CM5 chip on a Biacore T200 (Cytiva). Ligands were immobilized on the SPR sensor chip surface using 10 mM sodium acetate (pH 5.0). The blocking step was carried out with 1 M ethanolamine (pH 8.5). Analytes were then introduced at various concentrations to interact with the immobilized ligands. All experiments were conducted with the running buffer (PBS, pH 7.4) and the regeneration buffer (40 mM NaOH) at a flow rate of 30 μL/min. Data were analyzed using a 1:1 binding model, with curve fitting performed using the Scrubber 2 software (Reichert Technologies) or Biacore T200 Evaluation software v.3.2 (Cytiva). The binding affinity constant (K_D = kd/ka) was calculated from the association (ka) and dissociation (kd) rate constants obtained from the affinity curves.

Flow cytometry analysis

To determine CDCP1 expression on the cell surface, cells at 80% confluency were washed with Dulbecco's phosphate buffered saline (DPBS) (Lonza) and then detached using Gibco enzyme-free cell dissociation buffer (Thermo Fisher Scientific). The cells were blocked at 4°C with DPBS containing 5% BSA and human BD Fc block (BD Biosciences). Cells were stained with primary antibodies, including 2G10 or isotype control antibody at 100 ng/mL or 10 μg/mL for 1 h. Cells were washed twice with chilled washing buffer by centrifuging at 200 ×g at 4°C for 3 minutes and then incubated with 2 μg/mL of goat anti-mouse immunoglobulin G (IgG)-conjugated Alexa 488 (Thermo Fisher Scientific) at 4°C for 1 h. Fluorescence signals were detected by FACS (CyFlow Cube 6, PARTEC, Münster, Germany; CytoFLEX, Beckman Coulter, Brea, CA, USA) with a 488-nm blue laser and analyzed using an FCS Express 6 flow research edition (De Novo Software, Pasadena, CA, USA) software.

Immunohistochemistry

All tumor tissues used in this study were collected from anonymous analysis by Champions Oncology (Hackensack, NJ, USA). Tumor microarray tissue obtained from PDX was used to stain CDCP1 expressed on the cell membrane, depending on the KRAS mutation status. The antibody used for CDCP1 staining was purchased from Abcam (#ab252947, Cambridge, UK), and the isotype control antibody was purchased from CST (#3900). Staining was performed according to the internal protocol of Champions Oncology. Scoring was conducted by a pathologist based on the nuclear, cytoplasmic, or membranous staining of individual cells and was evaluated as follows: 0, no staining; 1, faint to weak; 2, moderate; and 3, strong.

Internalization assay

Cell-surface binding and antibody internalization were evaluated using CytoFLEX (Beckman Coulter). The cells were prepared and detached as described above. Blocking was performed with human BD Fc block to prevent off-target mediated internalization and 75 μg/mL of CHX to inhibit de novo protein synthesis at room temperature for 10 min. The 2G10 antibody (100 ng/mL) was applied to cells to detect CDCP1 on the cell surface and incubated at 4°C for 1 h. After washing twice, cells were resuspended in chilled or pre-warmed serum-free medium containing 1% BSA and 75 μg/mL CHX and then seeded onto 60 mm² culture dish. Each sample was incubated at 4°C or 37°C for 2 h. Subsequent steps were performed as described above.

An additional internalization assay was performed using confocal microscopy. Pancreatic cancer cells were seeded on coverslips and treated with human BD Fc block at normal culture conditions to block FcRn on the cell surface. The 2G10 antibody (10 μg/mL) was conjugated with anti-mouse IgG-conjugated Alexa 488 (10 μg/mL, Thermo Fisher Scientific) at 4°C for 30 min. Cells were stained with 2G10-Alexa 488 at 4°C for 1 h, washed with chilled PBS three times, and then incubated in warm complete media in a time-dependent manner. Cells were fixed with 4% paraformaldehyde for 15 min and permeabilized with PBS containing 0.1% (v/v) Triton X-100 for 15 min. Cells were blocked with blocking buffer (PBS containing 3% BSA) and stained with BD Pharmingen APC Mouse Anti-Human CD107a (LAMP-1) and 4′, 6-diamidino-2-phenylindole (DAPI, 100 nM, Sigma-Aldrich) at room temperature for 1 h. After washing four times, the immobilized microscope coverslip was mounted using a Permanent Aqueous Mounting Medium (Biomeda Corp., Foster City, CA, USA). Images were captured with a confocal microscope (A1R HD25 N-SIM S, Nikon, Tokyo, Japan) and obtained using NIS-Elements Viewer 4.50 software (Nikon).

Generation and characterization of ADC

Isotype control and 2G10 antibodies were dialyzed with conjugation buffer (10 mM Sodium phosphate, pH 7.0, and 150 mM NaCl) in Cellu·Sep regenerated cellulose T1 tubular membranes with 3,500 Da MWCO (Membrane Filtration Products Inc., Seguin, USA). Linker-payload complexes were dissolved in an organic solvent according to the manufacturer’s instructions. An aqueous solution of tris (2-carboxyethyl) phosphine hydrochloride (TCEP, Sigma-Aldrich, three equivalents per equivalent of antibody) was added to the antibody solution. The mixture was incubated for 2 h at 40°C. The linker-payload complex (six equivalents per equivalent of antibody) was then added to this solution, finalizing the mixture at a 15% DMSO concentration. The resulting mixture was incubated with slow, continuous rocking for 1 h. Conjugation was quenched by adding an excess of cysteine (20 equivalents per equivalent of antibody). The solution was spun down at 25,000 ×g at 4°C for 20 min to remove any aggregates. The final product was dialyzed in storage buffer (10 mM sodium phosphate, pH 6.8, 150 mM NaCl, 10% glycerol, and 0.05% Tween-80).

Cytotoxicity assay

Cells were seeded in a 96-black well cell culture plate and cultured in a humidified CO₂ incubator for 12 h: PANC-1 (5 × 10³ cells), AsPC-1 (5 × 10³ cells), BxPC-3 (5 × 10³ cells), MIA PaCa-2 (1 × 10³ cells), MDA-MB-468 (5 × 10³ cells), MCF-7 (2 × 10³ cells), CT26 (1 × 10³ cells), and HUVEC (1.5 × 10³ cells). The cells were then treated with the indicated concentrations of 2G10, linker-payload complexes, 2G10-ADCs, or chemotherapeutic drugs and incubated for 3 days. Cells were stained with Hoechst 33342 (Thermo Fisher Scientific) at a concentration of 3.3 μM for 1 h. Cells were analyzed using a Celigo Imaging Cytometer (Nexcelom BioScience, Lawrence, MA, USA). IC₅₀ were determined using Prism 5 (GraphPad Software, San Diego, CA, USA).

Apoptosis assay and cell-cycle assay

For the apoptosis assay, PANC-1 (6 × 10³ cells) or MCF-7 (5 × 10³ cells) cells were seeded into a 96-black well plate and allowed to attach overnight. Cells were then incubated with vehicle, 2G10 antibody (0.1 μg/mL), isotype control antibody (0.1 μg/mL), isotype antibody-PNU159682 (0.1 μg/mL), or 2G10 antibody-PNU159682 (0.1 μg/mL) for 36 h. Cells were stained with a Live Caspase 3/7 Detection kit (2 μM, Nexcelom Bioscience) to measure caspase3/7 activity and Hoechst 33342 (8 μM) for 1 h, and then analyzed using Celigo Imaging Cytometer.

For the cell-cycle assay, PANC-1 (6 × 10³ cells) and MCF-7 (2 × 10³ cells) cells were seeded into a 96-black well plate and incubated overnight. Antibody or ADC was treated at 0.1 μg/mL and incubated for 24, 36, and 48 h. Then, cells were fixed with chilled 80% ethanol at 4°C for 2 h and washed twice with chilled PBS. Cells were stained with the propidium iodide solution (50 μg/mL, with 0.1 mg/mL RNase A and 0.05% Triton X-100) diluted in chilled PBS for 1 h. The cells were documented using Celigo Imaging Cytometer and analyzed using FCS Express 6 flow cytometry software.

In vivo study

All animal experiments performed in this study were approved by the Institutional Animal Care and Use Committee of Ajou University (IACUC 2022-0037). Five-week-old female C.B-17 severe combined immunodeficiency mice were purchased from Janvier Labs (CB17-SCID, Le Genest St-Isle, France) and maintained under specific pathogen-free conditions on a 12-h light-dark cycle with food and water ad libitum. For the xenograft model, PANC-1 (3 × 10⁵ cells), MIA PaCa-2 (5 × 10⁵ cells), and HL-60 (1 × 10⁷ cells) were mixed with an equal volume of Corning Matrigel Basement Membrane Matrix (Sigma-Aldrich) and subcutaneously injected into the right flank of mice at anesthetized with isoflurane (Hana Pharm Co., Seoul, Korea). When the average tumor size reached 150–250 mm³, the mice were randomized into treatment groups for the efficacy study. Vehicle, control IgG, 2G10, or ADCs diluted in DPBS were intravenously administered to the mice once a week, three times, at the indicated concentrations. Chemotherapeutic drugs were administered intraperitoneally twice a week (GEM) or once a week (FFX) for 14 days at the indicated concentrations. Tumor size was measured twice a week using a Vernier caliper, and body weight was recorded separately until the end of the experiment. Tumor volume was calculated using the formula: volume = (4/3) × π × (length/2) × (width/2) × (depth/2). The mice were euthanized when the tumor volume exceeded 2,000 mm³ by CO₂ inhalation. The tumor tissues were removed and weighed. Tumor growth inhibitor (TGI) was calculated using the formula: TGI = 1 × (relative tumor volume (RTV) in the treated group) ÷ (RTV in the control group) × 100 (%). RTV = (tumor volume on the measured day) ÷ (tumor volume at day 0).

For the combination study of 2G10-PNU159682 with chemotherapeutic drugs, PANC-1 (3 × 10⁵ cells) and AsPC-1 (5 × 10⁵ cells) were subcutaneously injected as described above. When the tumor size reached around 170 mm³, the mice were randomized into each group: (Ⅰ) vehicle; (Ⅱ) 50 mg/kg of GEM; (Ⅲ) FFX (leucovorin at 50 mg/kg; 5-FU at 25 mg/kg; irinotecan at 25 mg/kg; oxaliplatin at 2.5 mg/kg); (Ⅳ) GEM combined with 2G10-PNU159682; (Ⅴ) FFX combined with 2G10-PNU159682. Chemotherapeutic drugs were dissolved according to the manufacturer’s instructions and diluted in DPBS to their representative concentrations before administration. To examine the combined efficacy of 2G10-PNU159682 and chemotherapy drug, both 0.1 mg/kg of 2G10-PNU159682 and GEM or FFX at the indicated concentrations were administered according to the schedule. 2G10-PNU159682 was intravenously administered. Initially, leucovorin and oxaliplatin were administered in combination. The mouse rested for at least 30 min, and then, a mixture of 5-FU and irinotecan was injected.

To examine the effect of 2G10-PNU159682 on the refractory or relapse model after chemotherapy, chemotherapeutic drugs were initially administered after randomization. GEM was administered by intraperitoneal injection twice a week for two cycles (on days 1–3 and days 7–10). FFX was administered intraperitoneally once a week for two cycles. When the tumor regrew and reached 300–400 mm³, the mice treated with chemotherapeutic drugs were randomized into two groups: one group received 0.5 mg/kg 2G10-PNU159682 and the remaining other received DPBS as a control. Tumor growth and body weight were monitored twice weekly.

Analysis of RNA-sequencing data

RNA sequencing data of tumor tissues from The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/) and normal tissues from the Genotype-Tissue Expression database (GTEx, https://gtexportal.org/home/) were downloaded from the UCSC Xena website (https://xenabrower.net). The mRNA expression patterns were compared between tumor and normal tissues using normalized values, specifically transcripts per million (TPM), as calculated by RSEM. To analyze the OS rate, the relationship between KRAS mutation status and CDCP1 expression in TCGA pancreatic cancer was assessed by dividing the data into high and low CDCP1 expression groups, as well as KRAS mutation and wild-type groups, using the median value of each gene. The OS of the patients was analyzed using Kaplan-Meier curves, and the significance of prognostic differences between groups was evaluated with the log rank test. All graphs were plotted using Prism 5.

The correlation between genes expressed in pancreatic cancer was confirmed using expression profiles (HTSeq-Counts) consisting of 327 datasets for normal pancreatic tissue from GTEx and 182 datasets from TCGA (4 datasets for normal pancreatic tissue and 178 datasets for pancreatic primary tumors). For visualization, the read count was normalized using the DESeq2 algorithm using the DESeq2 package in R 4.3.1, and correlations were calculated using Spearman’s correlations. All scripts are available for download from Bioconductor (https://bioconductor.org/packages/) and GitHub (https://github.com).

Statistical analysis

All statistical analyses were performed using GraphPad Prism software (version 5.0) or R software v. 4.3.1. Data are presented as mean and standard deviation (SD) or standard error of the mean (SEM). Statistical significance was determined using one-way ANOVA, two-way ANOVA, or t test for nonparametric data using the Mann-Whitney test (t test) with a two-tailed p value or Kruskal-Wallis (ANOVA) test, followed by Dunn’s multiple comparison test. Results were considered significant at p < 0.05.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Acknowledgments

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (No. 2023R1A2C1004335) and Novelty Nobility Inc. (No. S-2021-C2040-00001). We would like to thank Dr. Wonhee Suh from Chung-Ang University for providing the primary HUVEC cell.

Author contributions

Y.J.U., methodology, resources, validation, formal analysis, investigation, data curation, visualization, and writing—original draft; H.-D.N., resources; J.G.C., investigation and formal analysis; H.-J.K., J.-O.K., and T.M.W., formal analysis; S.G.P., conceptualization, project administration, funding acquisition, supervision, resources, writing the original draft, writing the review, and editing. All co-authors have given their consent to the submitted version of the manuscript for publication.

Declaration of interests

S.G.P. is a founder and CEO of the Novelty Nobility Inc. J.G.C., H.-J.K., J.-O.K., and T.M.W. are employees of the Novelty Nobility.