PCA062, a P-cadherin Targeting Antibody–Drug Conjugate, Displays Potent Antitumor Activity Against P-cadherin–expressing Malignancies

This article is featured in Highlights of This Issue, p. 1221

Classical cadherins are single span transmembrane glycoproteins that mediate calcium-dependent cell-to-cell contacts in adherens-type junctions of the epithelium (1, 2). P-cadherin (placental cadherin; CDH3) is a calcium-dependent adhesion molecule and member of the classical cadherin superfamily. The extracellular domain of P-cadherin mediates both cis and trans interactions between multiple P-cadherin molecules or between P-cadherin and other cell surface proteins such as CDCP1/Trask (3), while the intracellular tail of P-cadherin links it to proteins such as p120 catenin and other cytoskeletal elements which contribute to cellular architecture and polarity (4, 5). P-cadherin is believed to play a role in ductal mammary branching during development; however, in contrast to the observations made pertaining to E- and N-cadherin deficient mice, genetic knockout of the CDH3 locus is not associated with severe adverse phenotypes, suggesting that P-cadherin's function in mature organisms is either nonessential or redundant (6). As opposed to other cadherin family members, such as E-cadherin, P-cadherin is known to be overexpressed in a number of malignant cancers, such as those arising from breast, lung, bladder, esophagus, stomach, endometrium, and colon (5, 7–10) while its expression is restricted in normal tissues (11). P-cadherin mRNA and protein expression are known to be upregulated in cancers through different mechanisms, such as the inactivation of the tumor suppressor BRCA1, activation of the transcription factor C-EBPβ, CDH3 promoter hypomethylation, and treatment with the antiestrogen agent fulvestrant (12–15). In alveolar rhabdomyosarcoma, the chimeric oncogenic transcription factors PAX3-FOXOA1 and PAX7-FOXOA1 (resulting from chromosomal translocations) directly induce P-cadherin expression, resulting in increased tumor aggressiveness (16).

In multiple cancers, overexpression of P-cadherin is significantly correlated with lower overall and progression-free survival (15, 17–19). Within breast cancer, molecular epidemiologic data indicate that P-cadherin expression is most frequently associated with the basal-like breast cancer molecular subtype and the triple-negative breast cancer (TNBC) histologic subtype (ER⁻/PR⁻/Her2⁻; refs. 20, 21). Emerging biological evidence also suggests that P-cadherin may mediate cancer stem cell properties in models of basal-like breast cancer (22). Through FACS and siRNA knockdown approaches, Vieira and colleagues demonstrated that P-cadherin expression was correlated with an increased cellular capacity for forming mammospheres, resisting radiation-induced cell death and growth in three-dimensional culture systems in vitro (22). In addition, cells expressing high levels of P-cadherin displayed greater tumorigenic properties when implanted in immunocompromised mice compared with cells expressing low levels of P-cadherin (22). On the basis of its function as a cellular adhesion molecule, P-cadherin is also believed to promote tumor cell motility, invasiveness, and metastasis in several cancer types, including breast and ovarian (23–25). These data suggest that targeting P-cadherin–expressing cell populations in the context of heterogeneous tumors has the potential to offer clinical benefit.

On the basis of the favorable expression profile of P-cadherin, including limited normal tissue expression and broad tumor cell overexpression, the antibody–drug conjugate PCA062 was developed to deliver a potent cytotoxic drug selectively to cancer cells. PCA062 consists of a fully human P-cadherin specific antibody coupled to the maytansine-derived potent microtubule-inhibiting agent DM1 via the noncleavable thioether linkage SMCC [succinimidyl trans-4-(maleimidylmethyl)cyclohexane-1-carboxylate]. Clinical proof of concept of this linker-payload chemistry in breast cancer was demonstrated by trastuzumab emtansine, an antibody–drug conjugate (ADC) targeting the HER2 oncoprotein (26). Although several approaches have been used to target P-cadherin clinically including function blocking mAbs, radioimmunoconjugates, and T-cell engaging CD3-bispecific antibodies, PCA062 is the first P-cadherin targeting ADC developed to treat patients with TNBC, as well as other tumor types overexpressing P-cadherin (27–29).

Human IgG1/k anti–P-cadherin antibody CQY684 [denoted as NVP169N31Q in Table 1; ref. 30], the isotype control human IgG1/k antibody directed against a nonhuman protein (huIgG1) and rodent cross-reactive anti–P-cadherin antibody 3D21 (CH3D21) were generated at Novartis. Human anti–P-cadherin antibody CQY684 and huIgG1 isotype control antibody were directly conjugated to APC (Allophycocyanin) using the Lightning-Link APC kit from Innova Biosciences (catalog No. 705-0010). Conjugation of CQY684 and huIgG1 to the DM1 maytansinoid was performed using the cross-linking agent N-succinimidyl-4-(N-maleimidomethyl) cyclohexane-1-carboxylate (SMCC) or N-succinimidyl-4-(2-pyridyldithio) butanoate (SPDB) at ImmunoGen, Inc., as described previously (31). Cell lines were obtained from Novartis' CCLE collection authenticated by SNP analysis and were tested to be free of Mycoplasma and murine viruses (RADIL, University of Missouri, Columbia, MO). Source and culture conditions for cell lines used in this article were published earlier (32).

A co-complex of human P-Cadherin EC1_EC2 (amino acids 108–324) bound to a Fab fragment of CQY684 was crystallized and diffraction data were collected at the PXII beamline of the Swiss Light Source (Paul Scherrer Institute, Switzerland), with a Pilatus pixel detector and processed with XDS (33). For details on data processing and modeling, refer to the Supplementary Materials and Methods.

A total of 1 to 5 × 10⁵ cells were suspended in FACS buffer (PBS + 2% FBS + 0.1% sodium azide) and incubated CQY684-APC or huIgG1-APC at a concentration of 10 μg/mL for 45 to 60 minutes on ice. Samples were washed with buffer and assayed on a FACS machine (BD FACSCanto II) in conjunction with a microsphere bead set (Quantum Simply Cellular anti-Human IgG kit; Bangs Labs catalog No. 816B) to determine the specific antibody binding capacity of each antibody. In experiments examining the binding of PCA062 or IgG1-SMCC-DM1, antibody binding was detected by the addition of fluorochrome-conjugated anti-human antibody (goat anti-human APC, Southern Biotechnology, catalog No. 9042-11) after the first binding step.

Cell internalization of IgG antibodies by target-mediated endocytosis was assessed by either confocal microscopy or by flow-based cell surface fluorescence quenching (34). Cells were grown in chamber slides (Lab-Tek II -CC²) to about 75% confluence. PCA062 was added to wells at a concentration of 5 μg/mL in serum-free media with protease inhibitors and incubated for 1 hour on ice. After washing to remove excess PCA062, cells were transferred to 4°C or 37°C for 3, 8, and 24 hours, then washed and fixed with 4% methanol-free formaldehyde for 15 minutes. ADC was detected with goat anti-human IgG Alexa Fluor 488 (Invitrogen). LMP, a lysosomal marker, was detected with biotinylated mouse anti-human CD107a (BD Biosciences).

Cytotoxicity of free DM1 (L-DM1-Me), naked anti–P-cadherin antibody CQY684, PCA062 or isotype control ADC (huIgG1-SMCC-DM1) were profiled. Cells were seeded in black-walled clear bottom 96-well plates at a density of 2,500 cells per well in standard growth media and incubated at 37°C overnight. The next day, free maytansine (L-DM1-Me), anti–P-cadherin antibody CQY684, PCA062 or a nontargeting ADC control (IgG1-SMCC-DM1) were added to each well. After 5 days in a tissue culture incubator, cell viability was assessed using CellTiter-Glo and a luminescence counter. Luminescent counts from wells containing untreated cells (100% viability) were used to normalize treated samples. The concentrations of treatment required to inhibit 50% of cell growth or survival (IC₅₀) were calculated using nonlinear regression analysis (GraphPad Prism 6 software). Each cell line was evaluated at least three times and representative IC₅₀ values are shown.

A total of 5,000 target cells were seeded in 96-well flat bottom plates (Corning No. 3603). Test and control antibodies or ADCs were added to wells for 20 minutes at 37°C, followed by fresh human natural killer effector cells at an effector to target cell ratio of 5:1. After a 48-hour incubation period, media were dumped from plates and wells washed with PBS to remove nonadherent effector cells. CellTiter-Glo was utilized to assess cell viability.

P-cadherin expression in formalin-fixed, paraffin-embedded (FFPE) human tumor samples (Novartis tissue archive) was evaluated by IHC, using a mouse monoclonal anti-human P-cadherin antibody obtained from BD Transduction Lab (clone 56/P-cadherin, catalog No. 610228, lot no. 09934). Sensitivity and specificity of the antibody was established using a panel of cell lines with known P-cadherin expression levels. IHC was performed using a Ventana Discovery XT autostainer. Sections were deparaffinized, treated with the Ventana Cell Conditioning No. 1 (CCIS) antigen retrieval reagent, and then incubated for 60 minutes at room temperature in the primary antibody at a concentration of 10 μg/mL. This was followed by incubation with a biotinylated goat anti-mouse (Jackson Laboratories, catalog No. 115-066-072, lot no. 63620) used at working concentration of 1:250 was performed. Detection was performed using DAB Map Kit (Ventana Medical, catalog No. 760-124), A nonspecific mouse IgG antibody (Vector Laboratories, catalog No. I-2000, lot no. V0610) was used as a control to ensure that staining observed with the P-cadherin antibody was specific. All stained slides were digitized using an Aperio Scanscope whole slide scanner at 20× magnification. No further image modification was performed. A semiquantitative H-score of P-cadherin immunostaining was generated to reflect the expression and heterogeneity levels in tumors, using the formula: H-score = [(% of 1+ × 1) + (% of 2+ × 2) + (% of 3+ × 3)], where 1+ designates weak staining, 2+ designates moderate staining, and 3+ designates strong staining, resulting in a range of 0 to 300. An H-score >150 correlates to ≥ 50% of the cells exhibiting 2+ to 3+ staining intensity.

To assess exposure to the maytansinoid payload and its proximal mechanistic effect in tumor xenografts, we performed IHC with anti-DM1 and anti-phospho-histone H3 antibodies, respectively. DM1 IHC was conducted with a FITC-labeled mouse monoclonal anti-DM1 antibody (IMGNCAA-162; ref. 35). Conjugation with FITC was performed to avoid mouse-on-mouse background, using NHS_Fluorescein (170 nmol in DMSO; Thermo Fisher Scientific No. 46410). DM1 IHC was conducted on Ventana Discovery Ultra platform. Sections (4 μm) were deparaffinized, treated with the Cell Conditioning No. 1 antigen retrieval solution (Ventana, catalog No. 950-124) for 32 minutes (CC1M) with heat (95°C), and then at 37°C incubated for 60 minutes in the primary antibody at a concentration of 5 μg/mL, followed by incubation with rabbit anti-FITC (Thermo Fisher Scientific Molecular Probe, catalog No. A-889) at 1:200 dilution (Dako Diluent, catalog No. S0809) for 32 minutes, and then with DISC. OmniMap anti-Rabbit HRP RUO (Ventana Medical, catalog No. 760-4311) for 16 minutes. Detection was performed using DISC. ChromoMap DAB RUO (Ventana Medical, catalog No. 760-159), followed by counterstaining (4 minutes, respectively) with Hematoxylin RUO (Ventana, catalog No. 760-2021) and then Bluing Reagent RUO (Ventana, catalog No. 760-2037).

pHH3 IHC was conducted on Ventana Discovery XT platform. Sections (4 μm) were deparaffinized, treated with the Cell Conditioning No. 1 antigen retrieval solution (Ventana, catalog No. 950-124) for 64 minutes (CCIS) with heat (95°C), and then at 37°C incubated for 60 minutes in Phospho-Histone H3 (Ser10) antibody (Cell Signaling Technology, catalog No. 9701) at 1:100 dilution (Dako Diluent, catalog No. S0809), then incubated with DISC. OmniMap anti-Rabbit HRP RUO (Ventana Medical, catalog No. 760-4311) for 4 minutes. Detection was performed using DISC. ChromoMap DAB RUO (Ventana Medical, catalog No. 760-159), followed by counterstaining (4 minutes, respectively) with Hematoxylin RUO (Ventana, catalog No. 760-2021) and then Bluing Reagent RUO (Ventana, catalog No. 760-2037).

All animal studies were performed under approval by the Novartis Institutes for BioMedical Research Institutional Animal Care and Use Committee and in compliance with the Guide for the Care and Use of Laboratory Animals. For toxicology studies, all in-life procedures were conducted in compliance with the Animal Welfare Act, the Guide for the Care and Use of Laboratory Animals, and the Office of Laboratory Animal Welfare.

HCC70 and HCC1954 cells were grown in RPMI1640 medium (Corning Cellgro, catalog No. 10-040) containing 1% L-glutamine (Cellgro, catalog No. 25-005-CI) plus 10% FBS (Omega Scientific Inc., catalog No. FB-09), SCABER was grown in EMEM (ATCC, catalog No. 30-2003) containing 1% l-glutamine (Cellgro, catalog No. 25-005-CI) plus 10% FBS (Omega Scientific Inc., catalog No. FB-09) and BICR6 was grown in DMEM (ATCC, catalog No. 30-2002) containing 1% l-glutamine (Cellgro, catalog No. 25-005-CI) plus 10% FBS (Omega Scientific Inc., catalog No. FB-09) and 0.4 μg/mL hydrocortisone at 37°C in a humidified atmosphere containing 5% carbon dioxide. Cells were suspended in a cold Hank's Balanced Salt Solution (catalog No. 20-021, Corning Cellgro) at a concentration of 50 million cells/mL (containing 50% Matrigel, BD Biosciences, catalog No. 354234) for subcutaneous implantation into female C.B-17 SCID mice (HCC70) or SCID-beige mice (other cell lines) purchased from Envigo. PCA062 efficacy was also assessed in two internally developed non–small cell lung cancer patient-derived xenograft (PDX) models (HLUX1482 and HLUX1323) and an esophageal carcinoma ES2267 PDX model conducted at Crown Biosciences (China).

PCA062, its vehicle (20 mmol/L histidine, 8.22% sucrose, 0.02% PS-20, pH 5.6) or a nontargeting IgG1 isotype control huIgG1 ADC with the same linker and payload were administered as a single intravenous injection in 8 mL/kg to mice with HCC70 tumors approximately 200 mm³ in volume. Tumor were collected at specified timepoints after dosing as described in Fig. 4A, placed in 10% neutral buffered formalin overnight at room temperature, transferred to 70% ethanol and processed for paraffin embedding that were immunostained to detect DM1 and phospho-histone H3, as described above.

Antitumor efficacy studies were initiated when tumors were approximately 150 mm³. For the efficacy comparison of the rat-cross-reactive 3D21-SPDB-DM4 and 3D21-SMCC-DM1 ADC and their huIgG1 isotype control ADCs were administered intravenously in 8 mg/mL. For other efficacy studies, PCA062, its vehicle or huIgG1 isotype control ADC were administered at 10 mg/mL. The tumors were measured twice weekly. Tumor volume was calculated as (length × width²)/2. Data are expressed as the percent inhibition in tumor volume of the treated group from initial divided by the change in tumor volume of the vehicle control group from initial (%T/C)-100%. Between-group comparisons for final tumor measurements were performed using the ANOVA and a post hoc test using SigmaPlot (Systat Software Inc.).

Rodent safety studies was performed in IGS Wistar Hannover rats, obtained from Charles River Laboratories. Animals were randomly assigned to nine groups and received an intravenous bolus of vehicle control (10 mmol/L succinate/250 mmol/L glycine/0.01% Tween-20/0.5% sucrose pH5.5), CH3D21 or 3207 conjugated to DM1 (1 CH3D21-SMCC-DM1, 3207-SMCC-DM1) at 10 or 45 mg/kg, or CH3D21 or 3207 conjugated to DM4 (CH3D21-SPDB-DM4, 3207-SPDB-DM4) at 5 or 25 mg/kg. Compounds CH3D21 and 3207 were formulated in 10 mmol/L Succinate/250 mmol/L glycine/0.01% Tween-20/0.5% sucrose pH5.5. Clinical observations and body weights were recorded daily and at the time of euthanasia. Toxicokinetic samples were taken on day 1 at 1-hour postdose and day 2 at 3, 7, 14, and 21. Clinical pathology samples were collected prior to euthanasia on day 7 and 21 for hematology and serum chemistries. Necropsy and tissue collection was performed on day 7 and 21 and samples were fixed in 10% neutral buffered formalin for routine processing to paraffin. In a second study, IGS Wistar Hannover rats were administered a single intravenous dose of CH3D1-SMCC-DM1 or 3207-SMCC-DM1 at 60 mg/kg and similarly evaluated on day 7 and 21. All animal procedures followed the NIH Guide for the Care and Use of Laboratory Animals and were conducted under an Institutional Animal Care and Use Committee–approved protocol.

A safety study using male and female Cynomolgus monkeys (Macaca fascicularis; Asian mainland origin) was conducted following intravenous administration of PCA062 as a solution in vehicle (20 mmol/L Histidine, 240 mmol/L sucrose, 0.02% polysorbate 20, pH 5.3). At the initiation of dosing, the animals were approximately 2 to 4 years of age and weighed 2.1 to 5.1 kg. The dosing solutions were administered intravenously as a bolus to four groups of monkeys (3 or 5/sex/group) at doses of 0, 2, 6, and 15 mg/kg (four total doses administered 2 weeks apart). Clinical signs (predose, approximately 1 and 4 hours postdose on dosing days, at least once daily on all other days), body weights and food consumption determinations were performed. In-life evaluation included serial clinical chemistry, hematology, and coagulation profiles as well as urinalysis. Ophthalmoscopic examinations were performed pretest, after the first and fourth doses, and four times during the recovery phase. Electrocardiographic assessments were performed once pretest, the day after the second and fourth doses and twice during the recovery phase. Approximately 1 week after the fourth dose, all main study animals (three/sex/group) were necropsied and the remaining monkeys (two/sex) from the control, 6 and 15 mg/kg dose groups were placed on recovery for 8 weeks. Tissues were collected at necropsy in 10% neutral buffered formalin and microscopic examinations were conducted on gross lesions and a protocol-directed list of organs and tissues. In addition to routine sampling, skin was examined from the following anatomic locations: axilla, eyelid with conjunctiva, interscapular region, lower inner arm, palm, thorax, abdomen, back, hind leg and neck. Specific environmental enrichment was provided.

The CRISPR library was constructed as described previously (Liu and colleagues, 2019).

For the screen, PCA062-sensitive line HCC1954 and PCA062-resistant line KYSE510 were plated in CellStack culture chambers (Corning) 24 hours prior to transduction. Cells were then infected with single-guide RNA (sgRNA) pools at a representation of 1,000 cells per sgRNA at an multiplicity of infection of 0.4. Cells were selected for 2 days in the presence of 2 μg/mL puromycin after which the cells were checked by flow cytometry to ensure >90% of the remaining cells contained the sgRNA vector. The pools were split into two arms maintaining 1,000 × representation and treated with PCA061 at 10 nm (for KYSE510) or 1 nmol/L (for HCC1954) and DM1 at 0.1 nmol/L for both cell lines. After 7 (KYSE510) or 10 (HCC1954) days of compound treatment, maintaining 1,000 × representation throughout, genomic DNA was extracted using the QIAamp DNA Blood Maxi Kit (Qiagen) according to the manufacturer's protocol. DNA sequencing and data analysis methods described in Supplementary Materials and Methods.

The mRNA levels of CDH3, the gene encoding P-cadherin, were evaluated in human normal tissue and malignant samples using publicly available datasets from The Cancer Genome Atlas and Gene-Tissue-Expression (36, 37). CDH3 mRNA is frequently elevated in a variety of solid tumors including bladder, breast, and squamous carcinomas of the cervix, esophagus, head/neck, and lung (Fig. 1A). Low-level CDH3 mRNA is found in normal esophagus, skin, and reproductive tissues (Fig. 1A).

Frequent expression of high surface P-cadherin protein was further validated by IHC on a set of FFPE tumor blocks for TNBC, esophageal squamous cell carcinoma (ESCC) and head and neck squamous cell carcinoma (HNSCC). In this analysis, 37% (7/19) of TNBC, 68% of (15/22) ESCC, and of 30% (6/20) HNSCC tumors displayed greater than 2+ P-cadherin staining (Fig. 1B). Examples of IHC staining intensity of 1+, 2+, and 3+ can be found in Supplementary Fig. S1.

PCA062 is an ADC formed by an anti–P-cadherin humanized immunoglobulin (IgG)1/k antibody conjugated to a maytansine-derived cytotoxic payload (DM1) via a noncleavable SMCC linker (Fig. 1C). The antibody component, CQY684 was isolated from a Morphosys Human combinatorial Antibody Library using phage display technology (38). As shown by X-ray structure determination (vide infra), CQY684 selectively binds to the first N-terminal cadherin-repeat domain of human P-cadherin with no detectable cross-reactivity to E or N cadherins (Fig. 2). FACS experiments were performed in HCC1954, a P-cadherin positive human TNBC cell line, to evaluate whether cellular binding was similar between unconjugated CQY684 and PCA062 (Fig. 1D). CQY684 retained binding affinity to live cells after conjugation of the SMCC-DM1 linker payload (Fig. 1D). The K_D of binding between CQY684 and PCA062 to human P-cadherin was determined by surface plasmon resonance to be 33 and 27 nmol/L, respectively. As such, CQY684 was amenable for use in live cell binding experiments as a proxy for PCA062.

For ADCs utilizing the SMCC-DM1 linker-payload format, cellular internalization and trafficking to the lysosomal compartment is a known prerequisite for the processing and release of active drug catabolites (39). Thus, a fluorescence microscopic analysis of CQY684 internalization and lysosomal colocalization was performed in HCC1954 cells. Internalization of Alexa Fluor 488–conjugated CQY684 was visible after 3 hours of assay initiation and by 8 hours, CQY684 accumulated within lysosomes (Fig. 1E).

The ability of PCA062 to inhibit cell proliferation and survival was assessed using an in vitro proliferation assay. In these experiments, free maytansine (L-DM1-Me) was used to illustrate maximum potential cytotoxicity of the DM1 payload in the cell lines and a nontargeting isotype control ADC (IgG1-SMCC-DM1) was used to assess non–P-cadherin specific cytotoxic activity of the ADC. When compared with IgG1-SMCC-DM1, PCA062 demonstrated potent cell killing against HCC1954 cells while the anti–P-cadherin antibody CQY684 (the antibody component of PCA062) showed no effect on HCC1954 cell proliferation, indicating that the cytotoxic effect of PCA062 was due to payload (Fig. 1F). PCA062 showed specific cytotoxic activities toward cell lines that express P-cadherin (e.g., HCC70 and HCC1954), while demonstrating no significant activity against P-cadherin–negative cells which were sensitive to the cytotoxic activity of free DM1 (HT29; Fig. 1G). On the basis of these data, the in vitro selectivity and potency of PCA062 was found to be contingent upon tumor cell expression of P-cadherin. Representative IC₅₀ values of PCA062 in a panel of cell lines tested are summarized in Supplementary Table 1. Interestingly, no specific activity of PCA062 was seen in HCC1806 or A431 cells, despite the presence of P-cadherin expression and sensitivity to free DM1. Thus, other factors besides P-cadherin expression may contribute to PCA062 sensitivity. Both PCA062 and CQY684 demonstrated an ability to mediate P-cadherin specific antibody-dependent cell-mediated cytotoxicity (ADCC) in vitro (Supplementary Fig. S2).

We determined the crystal structure of the first two cadherin-repeat domains of human P-cadherin at 1.4 Å resolution and then solved the crystal structures of the CQY684 Fab in the free and antigen-bound states (Supplementary Table S1). Our structure of human P-cadherin is highly similar to that of monomeric P-cadherin published by Kudo and colleagues (ref. 40; PDB entry 4ZMY) and shows a crystal packing driven by cis homotypic interactions (41), with Trp109 of the N-terminal adhesion arm bound to its own acceptor pocket. The X-ray analysis of the Fab complex shows that CQY684/PCA062 binds to the N-terminal cadherin-repeat (EC1) domain, on the opposite side of the domain with respect to the region involved in trans interactions (Fig. 2A). The two copies of the CQY684 Fab complex in the asymmetric unit of the crystal reveal virtually identical sets of epitope residues except for minor variations due to different crystalline environments. Four complementarity-determining regions (CDRs) contribute to antigen binding (H-CDR2, H-CDR3, L-CDR1, and L-CDR3), with approximately 1,500 Ų of combined buried surface upon complex formation. The CQY684 epitope on human P-cadherin is conformational and made of two noncontiguous sequences, entirely comprised within the EC1 domain and encompassing residues 123 through 127, and residues 151 through 177 (Supplementary Fig. S3). Among those, residues 124 and 125, and residues 151 through 172 are contributing to most intermolecular contacts (Supplementary Fig. S3). The epitope does not include any of the calcium sites or any residue involved in calcium coordination. Among all epitope residues, Glu155 stands out as a likely hotspot of the binding interface, in view of its many important interactions with CQY684 (Fig. 2B). This residue is located within a nonconserved insertion which is part of the unusual “quasi–β-helix” motif first described by Shapiro and colleagues (42), and only found in human cadherins 1 to 4, as shown by a multiple sequence alignment (Fig. 2C). The lack of conservation of Glu155 in human cadherins 1, 2, and 4 is in line with the high selectivity of CQY684/PCA062 toward human P-cadherin (aka cadherin-3) compared with other human cadherins.

Structural overlays of the CQY684 Fab complex with free P-cadherin and free Fab do not reveal any large structural changes, but only local conformational changes affecting a few epitope and paratope residues, notably Glu155 of P-cadherin and the antibody L-CDR3 and H-CDR2 loops. Hence, the association of CQY684 and P-cadherin largely conforms a rigid lock-and-key fit mechanism, typical of high-affinity antibodies (43).

Together with two other epitope residues (Asp151 and Lys162), Glu155 is part of the cis dimerization interface. Therefore, PCA062 directly competes with cis interactions between P-cadherin molecules. In contrast, PCA062 binding does not interfere with trans interactions, as shown by structural overlays with strand-swapped human P-cadherin dimers (ref. 40; Fig. 2D). The distance between two Fabs bound to trans interacting P-cadherin molecules, however, is too large (∼145 Å) to be spanned by the linker region of a single IgG molecule. Therefore, simultaneous binding to strand-swapped P-cadherin dimers can only happen when the two CQY684 Fabs are provided by distinct IgG molecules. This is probably an important aspect of the CQY684/PCA062 epitope and mode of action, as simultaneous binding of one IgG molecule (bivalent binding) to two trans interacting P-cadherin molecules from apposing cells would block the dissociation of the P-cadherin dimer and thus the desired cellular internalization of the antibody–drug conjugate.

The rodent cross-reactive CQY684 surrogate, CH3D21, was conjugated via the cleavable linker SPDB to DM4 or to the noncleavable linker SMCC to DM1 (28) for an efficacy comparison in the HCC70 human TNBC xenograft model in mice. The cleavable ADC exhibited greater potency, with long-lasting tumor regression after a single 5 mg/kg dose (Fig. 3B), while the noncleavable ADC was less effective at 10 mg/kg, resulting in shorter-lived efficacy (Fig. 3A). Bystander killing induced by the membrane permeable payload released from the cleavable ADC was likely the main contributor to this greater efficacy.

However, in a rat tolerability study, the cleavable ADC induced strong skin toxicity starting on day 8 after a single intravenous dose of 25 mg/kg, while the noncleavable ADC was better tolerated, with skin toxicity at 60 mg/kg. An example of microscopic skin toxicity as compared with skin in an untreated rat is shown in Fig. 3C. P-cadherin expression level and pattern is comparable between human and rat skin (Fig. 3D), further raising the concern of exaggerated skin toxicity in humans when using the cleavable linker. Taken together, these data supported the selection of a noncleavable ADC, from which PCA062 was selected as a clinical candidate due to potency and good developability properties.

To assess the ability of PCA062 to induce mitotic arrest relative to DM1 levels in tumor, a single PCA062 dose of either 3 or 10 mg/kg i.v. was administered to mice with HCC70-engrafted tumors. Dose-dependent tumor DM1 exposure peaked on day 1 and continued to decrease afterward (Fig. 4A). DM1 signal remained detectable after 7 days (Fig. 4A). The accumulation of cells positive for phosphorylated Histone H3 (pHH3), a mitotic marker, was delayed compared with tumor DM1 exposure, peaking on day 3 and maintained through day 7 postdose (Fig. 4A). These data suggest that the high initial exposure over the first several days drives much of the mitotic arrest, consistent with the mechanism of action of the maytansine-derived payload.

PCA062 efficacy in the P-cadherin–expressing HCC70 TNBC, BICR6 HNSCC and PDX ES2267 esophageal carcinoma xenograft models is shown in Fig. 4B–D. In the HCC70 study, a single PCA062 treatment demonstrated dose-dependent efficacy across the range of 2.5, 5, and 10 mg/kg, with %ΔT/ΔC values of 48% (2.5 mg/kg), 1% (5 mg/kg), and 64% tumor regression at 10 mg/kg, with durable regression out to 86 days posttreatment. After dosing on days 0 and 14, PCA062 activity in the BICR6 and ES2267 xenograft models induced stasis and regression, respectively.

In Fig. 4E, the efficacy of PCA062 is summarized across a panel of cell line–derived and PDX xenograft models, relative to the H-score of P-cadherin expression in these tumors. The efficacy data suggest that focusing on cancers with an H-score of approximately 150 or greater, corresponding to ≥50% of cells exhibiting 2+ to 3+ staining intensity, may enrich for patients more likely to benefit from PCA062 treatment. However, high P-cadherin expression alone is not sufficient to predict response (Fig. 4E). This suggests additional factors such as payload concentration in the cell, sensitivity to payload, or target biology may impact ADC in vivo efficacy.

Lack of PCA062 cytotoxicity in some cell lines expressing high surface levels of P-cadherin both in vitro (Table 1) and in vivo (Fig. 4E) suggest that although high P-cadherin expression is required for PCA062 activity, it is not the only determinant. ADC internalization is a key determinant of ADC efficacy (39). We compared PCA062 internalization rates in three sensitive (COLO-680N, HCC1954, HCC70) versus three insensitive (A431, BICR22, KYSE510) cell lines. All selected cell lines expressed comparable surface level P-cadherin (Supplementary Fig. S4) and have similar cellular sensitivity to free DM1 (Fig. 5A). Unlike sensitive lines where PCA062 has a lower IC₅₀ than its huIgG isotype control, the resistant cell lines have overlapping response curves for PCA062 and huIgG, indicating little or no cytotoxicity mediated by anti–P-cadherin (Fig. 5A). The PCA062 internalization rate was measured by the uptake of a fluorescent dye-labeled anti–P-cadherin antibody CQY684, the Ab portion of PCA062, in both PCA062-sensitive and PCA062-resistant lines. PCA062-resistant lines show slower CQY684 internalization as compared with PCA062-sensitive lines, indicating that a defect in ADC internalization may contribute to PCA062 insensitivity (Fig. 5B).

To explore additional resistance mechanisms to PCA062 as well as to find critical components for PCA062 internalization, a genome-wide CRISPR screen was performed in PCA062-sensitive HCC1954 and in PCA062-resistant KYSE510 cell line in the presence of PCA062 or naked DM1 to identify genes that when knocked out may specifically modulate PCA062 sensitivity. CDH3 was the strongest rescue hit in HCC1954 cells, consistent with the expectation that P-cadherin expression is necessary for PCA062-mediated cytotoxicity (Fig. 5C). The lysosomal transporter SLC46A3 and SAGA transcription complex components were strong rescue hits for PCA062 in PCA062-sensitive HCC1954 cells (Fig. 5C). The multidrug-resistant gene ABCC1 was a strong hit for PCA062 sensitization in PCA062-resistant KYSE510 cells (Fig. 5D). No strong hits linked to endocytosis pathway were identified in these two screens.

Both ABCC1 and SLC46A3 have been shown to affect ADC cellular efficacy (44–47). To evaluate the predictive value of ABCC1 and SLC46A3 for PCA062 sensitivity, we compared the mRNA level of these two genes in PCA062-sensitive and PCA062-insensitive lines. Forty-seven cancer cell lines of TNBC, HNSCC, and esophageal origin were profiled for PCA062 and isotype control sensitivity. All these cell lines express relatively high P-cadherin mRNA levels [transcripts per million (TPM) > 75]. Cell lines with IC50_PCA062<IC50_{isotype_control} are classified as PCA062-sensitive lines (Supplementary Table S2). Mean ABCC1 expression is significantly higher in PCA062-insensitive lines while mean SLC46A3 expression is significantly higher in PCA062-sensitive lines (Fig. 5E).

Safety studies to support first-in-human clinical trials were conducted in cynomolgus monkeys, for which the binding affinity of PCA062 to P-cadherin is similar to that of the human. In the pivotal 8-week cynomolgus monkey study, PCA062 was administered intravenously at 2, 6, and 15 mg/kg every 2 weeks for four dose cycles. PCA062-related changes were dose dependent and consisted of effects in the bone marrow and serum chemistry changes in the liver of similar magnitude to the dose range finding study. The most significant changes in clinical pathology were nontarget-mediated and dose-dependent decreases in absolute platelet counts in animals dosed at 6 and 15 mg/kg. In general, the decreases were most pronounced at 3 days postdose and showed partial to complete recovery by 10 days postdosing. The principal targets of toxicity for PCA062 included the skin, eye (cornea and pigmented retinal epithelium), liver, lymphoid organs (spleen), and bone marrow. A spectrum of clinical signs affecting the skin were observed in animals dosed at ≥6 mg/kg at the injection site and over the body in a dose-related fashion in terms of incidence, onset and severity including bruising, hyperpigmentation, crusts and blister formation. Microscopically, these findings correlated with acanthosis, degeneration, hyperkeratosis, inflammation, increased mitotic figures edema and pigment deposits. The aforementioned toxicities related to PCA062 in monkeys were completely reversible or showed trends toward recovery based on incidence and/or severity 8 weeks after cessation of dosing. The highest nonseverely toxic dose (HNSTD) was determined to be 15 mg/kg when given as four biweekly doses. Consistent with maximal expression of P-cadherin in epithelium of various tissues, including but not limited to the skin, cornea, and retina, PCA062 caused cutaneous, corneal and retinal target-related toxicity.

Therapeutics utilizing the ADC modality have recently proven to be highly effective at producing strong antitumor responses in solid and liquid tumors while minimizing systemic toxicity in patients (48, 49). Although trastuzumab emtansine is the only DM1-based ADC approved to date the recent approval of belantamab mefadotin, which also uses a noncleavable, nonpermeable payload, supports that this payload class may be appropriate when paired with certain antibody targets. Overall, the diversity of payloads used in approved ADCs targeting a range of antigens including sacituzumab govitecan, trastuzumab deruxtecan, and enfortumab vedotin, supports the need to carefully match payload potency, release, and permeability characteristics with the biology and expression profile of the individual antigen in question.

In our development of a P-cadherin–targeted ADC, we employed early assessments of cleavable and noncleavable linker conjugated rodent cross-reactive ADCs to identify a candidate with the greatest potential for overall therapeutic benefit. While the cleavable permeable SPDB-DM4 ADC demonstrated improved efficacy, it also induced strong skin toxicity in the rat. Because P-cadherin expression was similar in the rat and human skin, the risk of exacerbating skin toxicity led to the decision to only develop a noncleavable ADC. Although ADCs utilizing cleavable-permeable linker payloads, such as SPDB-DM4, are known in the field to drive both increased efficacy and toxicity, our comparative studies allowed for the selection of a candidate best matched to the expression profile of P-cadherin with the goal of maximizing therapeutic index. From this effort, PCA062 was identified as a highly selective and potent noncleavable ADC with the aim of targeting tumor types overexpressing the P-cadherin cell surface glycoprotein. In vitro, PCA062 was demonstrated to selectively bind P-cadherin–expressing cell lines, to rapidly internalize and traffic to lysosomes, and to release sufficient activated payload to potently induce a cytotoxic response in cell viability assays. In vitro profiling of PCA062 activity in a cell line panel indicated that this ADC effectively targets and kills P-cadherin–positive tumor cells representing breast, head and neck, and bladder carcinomas. PCA062 induced potent ADCC in vitro; thus, this may have the potential to further contribute to efficacy in the clinic.

The in vitro profile of PCA062 was corroborated by the determination of its binding epitope and mode of action, using X-ray analyses. In line with its cross-reactivity profile, the PCA062 epitope includes structural features unique to P-cadherin, notably Glu155, a nonconserved amino acid and hotspot of the binding interface, located within the “quasi–β-helix” insertion only present in type I human cadherins. Moreover, the PCA062 epitope is independent of glycosylation and calcium-binding state of the target antigen. Although PCA062 is not expected to disrupt intercellular adhesion, as trans homotypic interactions are not affected by PCA062 binding, cellular internalization is not compromised by bivalent binding to strand-swapped P-cadherin dimers engaged in intercellular, trans interactions. As such, bivalent binding is precluded by the location of the epitope and the topology of the antibody–antigen complex.

In vivo, PCA062 administration to mice with HCC70 xenograft tumors resulted in an induction of mitotic arrest that corresponded with DM1 levels in the tumors assessed by IHC compared with the nontargeting isotype control ADC. This level of tumor penetrance and retention of PCA062 reflected the high systemic stability of the therapeutic. Antitumor efficacy of PCA062 was observed in multiple P-cadherin–positive xenograft models. In the HCC70 TNBC model, a single 2.5 mg/kg dose induced tumor stasis, while 10 mg/kg induced tumor regression. In an effort to build an understanding of the potential for PCA062 efficacy in multiple tumor types, several models with varying levels of P-cadherin expression were profiled in efficacy studies. Marked efficacy responses (tumor stasis or regression) were observed in P-cadherin–expressing tumor models representing basal-like breast cancer, HNSCC, esophageal, and bladder cancer with an H-score of approximately 150. PCA062 was inactive in xenografts that had low or no P-cadherin (H-score < 10), supporting a requirement for a higher expression level to confer efficacy.

Although a high P-cadherin cutoff may enrich for responders, it was not sufficient to predict antitumor activity, suggesting additional factors such as sensitivity to payload or target biology may impact ADC efficacy. Despite good target expression in both primary lung PDX models (HLUX1323 and HLUX1481; Fig. 4E), PCA062 was only effective in one (HLUX 1481; Fig 4E). When PCA062 activity was examined in 47 cell lines expressing relatively high P-cadherin mRNA (TPM > 75), only a small fraction of these (12/47) showed specific sensitivity to PCA062. Some of the PCA062-resistant cell lines have comparable cell surface PCA062 expression and sensitivity to naked payload (Fig. 5A). These findings suggest that other factors, such as receptor internalization, lysosomal processing, and cellular payload concentration may also contribute. A slower internalization rate for PCA062 was observed in some of the PCA062-resistant lines (Fig. 5B).

A genome-wide CRISPR screen revealed potential cellular modulators of PCA062 efficacy, such as the drug efflux pump ABCC1 (also known as MRP1), lysosomal transporter solute carrier family 46 member 3 (SLC46A3), and the SAGA transcription complex (Fig. 5C and D). Induction of ABCC1 as well as loss of SLC46A3 was observed in HER2-amplified breast cancer cell lines with acquired resistance to T-DM1 (47, 50). SLC46A3 was also identified as a key lysosomal membrane transporter to transport catabolites of ADCs bearing the payloads DM1 or PBD with noncleavable linkers across lysosomal membrane into the cytosol (45, 46). Increased expression of ABCC1 and decreased expression of SLC46A3 could lead to an overall decrease in intracellular PCA062 catabolite concentration and render cells more resistant to PCA062. Low expression of ABCC1 and high expression of SLC46A3 significantly associated with cell line sensitivity to PCA062 suggesting these genes could be potential biomarkers for PCA062 sensitivity (Fig. 5E). It is unclear how loss of SAGA complex components may lead to resistance to PCA062. Because the SAGA complex plays a key role in transcription regulation (51, 52), it is possible that loss of SAGA activity may perturb expression of P-cadherin or other cellular components critical to mediate PCA062 efficacy. Examining such findings in the clinic may allow for a more refined selection of patients with a greater chance of responding to these types of ADCs beyond target expression.

Repeat administration of PCA062 in primates was tolerated up to 15 mg/kg biweekly. Overall, the toxicity profile was consistent with expression of P-cadherin in the skin and esophageal epithelium, and with known target-independent toxicities of the maytansinoid ADC platform. On the basis of the encouraging preclinical efficacy and tolerability, PCA062 was evaluated in a phase I trial (NCT02375958). In patients, PCA062 was sufficiently tolerated to escalate to 3.6 mg/kg every 2 weeks, to the point where payload-related toxicity limited further escalation. During dose escalation, there were limited efficacy signals leading to the discontinuation of the trial prior to the expansion phase. The clinical findings are described elsewhere (submitted). Overall, these findings demonstrate the significant challenges to translate preclinical findings into meaningful clinical benefit for patients with cancer with the ADC modality.

Q. Sheng reports other support from Novartis Oncology outside the submitted work. J.A. D'Alessio reports personal fees from The Novartis Institutes for Biomedical Research outside the submitted work. D.L. Menezes reports other support from Novartis and BMS outside the submitted work; in addition, D.L. Menezes has a patent for ADCs pending. J.-M. Rondeau reports a patent for US 10626172 issued. F.C. Geyer reports other support from Novartis Institutes for Biomedical Research during the conduct of the study and other support from Novartis Institutes for Biomedical Research outside the submitted work. J. Gu reports other support from Novartis Institute of Biomedical Research during the conduct of the study and other support from Novartis Institute of Biomedical Research outside the submitted work. M.E. McLaughlin reports other support from Novartis outside the submitted work. T. Huber reports a patent for WO2016075670 pending and reports employment with Novartis at the time of the study. K.G. Rendahl reports a patent for patent number 10626172 issued, a patent for publication number 20190194315 pending, a patent for publication number 20190119375 pending, a patent for patent number 10005836 issued, and a patent for publication number 20160137730 pending. N.K. Pryer was a full-time employee of Novartis Institutes for BioMedical Research at the time the work was conducted. T.J. Abrams is a full-time employee and shareholder at Novartis. No disclosures were reported by the other authors.

Q. Sheng: Conceptualization, data curation, formal analysis, supervision, investigation, writing–original draft. J.A. D'Alessio: Conceptualization, data curation, formal analysis, supervision, investigation, writing–original draft. D.L. Menezes: Conceptualization, data curation, formal analysis, methodology, writing–original draft. C. Karim: Conceptualization, data curation, formal analysis, methodology, writing–original draft. Y. Tang: Data curation, methodology. A. Tam: Data curation, methodology. S. Clark: Data curation, methodology. C. Ying: Data curation, methodology. A. Connor: Conceptualization, data curation, methodology. K.G. Mansfield: Formal analysis, writing–original draft. J.-M. Rondeau: Data curation, formal analysis, methodology. M. Ghoddusi: Conceptualization, data curation, methodology. F.C. Geyer: Conceptualization, data curation, supervision, methodology. J. Gu: Data curation, methodology. M.E. McLaughlin: Conceptualization, supervision. R. Newcombe: Resources, data curation, methodology. G. Elliott: Data curation, formal analysis, methodology. W.R. Tschantz: Resources, data curation, methodology. S. Lehmann: Data curation, methodology. C.P. Fanton: Conceptualization, data curation. K. Miller: Conceptualization, data curation, formal analysis, methodology. T. Huber: Conceptualization, investigation, methodology. K.G. Rendahl: Conceptualization, data curation, formal analysis, methodology. U. Jeffry: Conceptualization, data curation, methodology. N.K. Pryer: Conceptualization, formal analysis. E. Lees: Conceptualization, supervision, visualization. P. Kwon: Conceptualization, project administration. J.A. Abraham: Conceptualization, formal analysis, project administration. J.S. Damiano: Conceptualization, data curation, formal analysis, supervision, methodology. T.J. Abrams: Conceptualization, data curation, formal analysis, supervision, investigation, writing–original draft.

We would like to offer special thanks to Zhen Wang for her IHC contributions to the preclinical research supporting the development of PCA062, who, although no longer with us, continues to remind us by her dedication in the search for new cancer therapeutics. We thank Janina Gasser for the expression of the CQY684 Fab and Aurelie Winterhalter for the cloning and expression of human P-cadherin EC1_EC2.

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