Design and Evaluation of ZD06519, a Novel Camptothecin Payload for Antibody Drug Conjugates

This article is featured in Selected Articles from This Issue, p. 593

Antibody drug conjugates (ADC) are an effective and promising class of cancer therapeutics. With 11 FDA-approved agents to date and over 170 ADCs currently in clinical development, this class of therapeutics is reshaping the treatment of cancer across tumor types (1). While emerging clinical data have shown that ADCs do not significantly increase the MTD compared with their unconjugated toxins, ADCs can demonstrate superior efficacy over related small-molecule chemotherapeutics when dosed at or near their MTD (2). In addition, the toxicity profile of ADCs may be altered compared with their related small molecules, though chemotherapy-like platform toxicities generally predominate.

In recent years, camptothecin-based topoisomerase I inhibitors have stood out as one of the most promising payload classes for ADCs. Camptothecin, a natural topoisomerase I inhibitor chemotherapeutic isolated in 1966 (3), exerts its action through binding to DNA–topoisomerase I complexes, and thus inhibiting DNA cutting, relaxing, and reannealing processes, ultimately leading to cell death (4). Despite extensive research on synthetic camptothecin small molecules spanning over six decades (5), their clinical progress has been hindered by challenges including toxicity (in particular, hematologic and gastrointestinal toxicities), low solubility, limited oral bioavailability, rapid in vivo clearance, and the reversible hydrolysis of the camptothecin lactone form at physiologic pH to a less active carboxylate form (6). Nevertheless, two camptothecin small molecules, topotecan and irinotecan (a prodrug of SN-38), are FDA approved for treating patients with ovarian, cervical, small cell lung, and colon cancers. Another small-molecule camptothecin drug, belotecan, is also approved in South Korea for the treatment of patients with ovarian cancer.

ADCs resolve many of the major challenges associated with camptothecin small molecules. The antibody component enhances the solubility and improves the half-life and the exposure of the conjugated payload. In addition, the trafficking of an ADC into endosome and lysosome compartments exposes the camptothecin payload to a low-pH environment (∼4.5–6; ref. 7), thereby shifting the equilibrium from the less active carboxylate form to the active lactone form (8). This is not the case for camptothecin small molecules, where after injection they rapidly establish an equilibrium favoring the less active form. Thus, ADC cellular uptake and trafficking provides a unique mechanism to convert the inactive camptothecin form to the active form, highlighting how the lactone equilibrium can be tuned on the basis of the delivery method.

Two FDA-approved ADCs, trastuzumab deruxtecan (T-DXd; refs. 9–11) and sacituzumab govitecan (SG; refs. 12, 13), employing camptothecin analogs as payloads (DXd for T-DXd and SN-38 for SG, respectively) have demonstrated effectiveness in treating patients with various solid tumors, including settings where camptothecin small molecules were less effective than standard of care (14). Despite their significant therapeutic benefit for patients, camptothecin ADCs exhibit severe payload-related toxicities. In addition to numerous other side effects, T-DXd is associated with nausea, neutropenia, anemia, and interstitial lung disease, while SG leads to gastrointestinal toxicities, neutropenia, and anemia (15, 16).

The clinical success of T-DXd and SG and the potential to address their limitations has sparked intense interest in the development of alternative camptothecin drug-linkers (Fig. 1). These new platforms employ either known camptothecin analogs (i.e., camptothecins previously in development as small molecules) conjugated using novel linker technologies or novel camptothecin analogs linked with well-established linker strategies (or, in some cases, a combination of both approaches). Critical to the design of a camptothecin ADC platform is minimizing aggregation, a major challenge that plagued early efforts to link known camptothecins. For example, low monomer content was observed during the attempts to conjugate exatecan (DX-8951) to trastuzumab using a standard para-aminobenzyl carbamate (PABC) group. To overcome the aggregation issue, exatecan was modified with a glycolate group to generate DXd (DX-8951 derivative; ref. 17). The primary alcohol moiety of DXd provided an optimal attachment point for the protease cleavable GGFG tetrapeptide linker via a novel self-immolative hemiaminal ether group (AM), which greatly reduced the aggregation compared with PABC-linked camptothecins. The linker was terminated with a thiol-reactive maleimidocaproyl (MC) unit to form the MC-GGFG-AM-DXd drug-linker (also known as deruxtecan). Conjugation of deruxtecan up to a drug-to-antibody ratio (DAR) of 8 yields ADCs with minimal aggregation. In an alternative approach, the SG hydrophilic linker CL2A (18) was developed to allow the conjugation of the known SN-38 payload as DAR∼8, also avoiding significant aggregation.

As a result of the extensive research on camptothecin drug-linkers and a growing interest in ADCs, at least 72 novel ADCs carrying various camptothecin payloads have entered clinical development as of January 1, 2024 (Fig. 1). These ADCs utilize more than 15 different payloads and more than 21 different linkers across 24 different targets. Among ADCs presently undergoing clinical development, the predominant (>60%) payloads are either exatecan or exatecan-derived compounds closely related to DXd (e.g., SHR9265, Ed-04, DDDXd, MH30010008).

Novel camptothecin payloads may differentiate from DXd, SN-38, and exatecan not only in terms of their compatibility with new linkers and conjugation strategies, but also in their intrinsic properties such as in vitro cytotoxicity, hydrophilicity, permeability, metabolic stability, pharmacokinetics, and susceptibility to multi-drug resistance efflux pumps.

Emerging clinical data suggest that ADC off-target toxicities and clinical MTDs are mainly related to their payloads (2), and ADC efficacy, especially for solid tumors, is likely driven by a complex combination of targeted payload delivery, free payload exposure, and tumor sensitivity to the ADC components (2, 19). ADC features (including conjugation and drug-linker designs) and target expression influence sites and rates of ADC disposition, and in turn, tumor, tissue, and systemic exposure to payload (20). Therefore, optimization of ADC payload properties is a key aspect to maximize clinical efficacy while balancing unwanted but inevitable toxicities.

Herein, we present the development of a novel camptothecin drug-linker platform tailored for ADC application, with the emphasis on optimizing the payload, the linker, and the resultant ADC properties in unison. In particular, the novel payload ZD06519 (FD1) was selected among approximately 100 camptothecins for its “moderate” potency (∼1 nmol/L), bystander killing, limited hydrophobicity, and prolonged plasma stability. The linker was chosen based on well-established technologies including thiol-maleimide conjugation and a cleavable peptide sequence combined with a self-immolative spacer for traceless payload release. Success criteria for the resultant ADCs included minimal aggregation, ADC pharmacokinetics and exposure similar to the unmodified parental antibody, efficacy in mouse models at a single ADC dose of <10 mg/kg, tolerability in mice and rats at ≥200 mg/kg and in non-human primates (NHP) at ≥30 mg/kg when the ADC is administered intravenously once every 3 weeks.

Payloads and drug-linkers were synthesized as described in the Supplementary Data.

Anti-HER2 antibody trastuzumab (Herceptin) was purchased from Roche. Anti-cMET antibody telisotuzumab was cloned using the published primary sequences (Patent US8545839B2) and expressed in CHO-K1 cells. Anti-FRα, anti-NaPi2b, and anti-GPC3 antibodies were expressed in house as described in WO2023/178452, PCT/CA2023/051385, and PCT/CA2023/051378, respectively. Antibodies in PBS were reduced using tris(2-carboxyethyl)phosphine (TCEP, 12 molar equivalents) for 3 hours at 37 °C in the presence of 1 mmol/L diethylenetriaminepentaacetic acid. Following reduction, excess TCEP was removed using 40 kDa Zeba Spin Desalting Columns (Thermo Fisher Scientific, catalog no. 87767), pre-equilibrated with 10 mmol/L sodium acetate buffer, pH 4.5. Reduced antibodies were then conjugated with 15 molar equivalents of drug-linker with up to 10% DMSO [volume for volume (v/v)] in the reaction. Each conjugation reaction was incubated at room temperature and protected from light for up to 2.5 hours prior to purification using 40 kDa Zeba Spin Desalting Columns (Thermo Fisher Scientific, catalog no. 87767), pre-equilibrated with 10 mmol/L sodium acetate with 9% (w/v) sucrose in water, pH 4.5. The DAR for each ADC was determined either by hydrophobic interaction chromatography-high-performance liquid chromatography (HIC-HPLC) using an Agilent 1290 Infinity II HPLC (Agilent Technologies) equipped with a photodiode array detector and a TSKgel Butyl-NPR HPLC Column, 3.5 cm × 4.6 mm, 2.5 μm particle size (Tosoh Bioscience, catalog no. 0014947) or by LC/MS using an Agilent 1290 Infinity II HPLC (Agilent Technologies) coupled with an Agilent 6545A Quadrupole Time of Flight (Q-TOF) Mass Spectrometer (Agilent Technologies). Additional details for ADC characterization are described in the Supplementary Data. The HIC-HPLC relative retention time (RRT) of each ADC was calculated by dividing the HIC-HPLC retention time of the DAR8 peak by the retention time of the unconjugated antibody.

The ability of camptothecin-free drugs and ADCs to inhibit proliferation of tumor cells in vitro was evaluated using the following cancer cell lines: SK-OV-3 (ovarian, ATCC, catalog no. HTB-77, RRID:CVCL_0532), JIMT-1 (breast, AddexBio, catalog no. C0006005, RRID:CVCL_2077), SK-BR-3 (breast, ATCC, catalog no. HTB-30, RRID:CVCL_0033), ZR-75-1 (breast, ATCC, catalog no. CRL-1500, RRID:CVCL_0588), Calu-3 (breast, ATCC, catalog no. HTB-55, RRID:CVCL_0609), and MDA-MB-468 (breast, AddexBio, catalog no. C0006003, RRID:CVCL_0419). SK-OV-3, SK-BR-3, ZR-75-1, and Calu-3 cell lines were obtained from ATCC (RRID:SCR_001672); JIMT-1 and MDA-MB-468 cell lines were obtained from AddexBio. All the cell lines were negative for Mycoplasma contamination prior to receipt and no additional Mycoplasma testing was performed. Cell lines were authenticated prior to receipt and no additional authentication was performed. After thawing, cells were cultured (per vendor instructions) for a maximum of approximately 12 weeks or approximately 30 passages. For cytotoxicity assay, cells (cultured per vendor instructions) were detached by incubation with TrypLE Express Enzyme (Thermo Fisher Scientific, catalog no. 12604013) and seeded in tissue culture (TC)-treated flat-bottom 384-well microtiter plates (Thermo Fisher Scientific, catalog no. 164610). Test articles were serially diluted in growth medium to create a 9-point dose–response curve, including growth medium-only blank controls, and added in technical duplicate to cell monolayers in the 384-well microtiter plates. Plates were incubated at 37°C with 5% CO₂ for 5 days. Cell viability was quantified by addition of CellTiter-Glo Reagent (Promega, catalog no. G7573) and incubated at room temperature for 30 minutes, then luminescence was measured with a microplate luminometer (BioTek Synergy H1 Microplate Reader). Percent cell growth inhibition for each test article at each concentration was calculated using the average luminescence signal using the corresponding untreated cells which were incubated with complete growth medium only, using the following formula for any given well: [1 − ([well relative light unit (RLU) value]/[average RLU values from untreated wells]) × 100]. Results from each independent experiment were plotted as percent cell growth inhibition on the y-axis and log of the test article concentration (nmol/L) on the x-axis using Prism 9.5.1 (GraphPad Software). IC₅₀ for each test article was calculated using a four-parameter logistic regression analysis with variable slope using Prism 9.5.1 (GraphPad Software).

The cytotoxicity of ADCs in three-dimensional (3D) spheroids was evaluated using the tumor cell line JIMT-1 grown in coculture (1:1) with human dermal fibroblast adult cell line HDFa (Thermo Fisher Scientific, catalog no. C0135C). Cells (cultured per vendor instructions) were detached by incubation with TrypLE Express Enzyme (Thermo Fisher Scientific, catalog no. 12604013) and seeded in Ultra-Low Attachment round bottom 384-well microtiter plates (Corning, catalog no. 3830), centrifuged, and incubated under standard culturing conditions to allow for spheroid formation and growth. Test articles were serially diluted in growth medium to create a 9-point dose–response curve, including growth medium-only blank controls, and added in technical duplicate to cell spheroids in the 384-well microtiter plates. Plates were incubated at 37°C with 5% CO₂ for 6 days. Cell viability was quantified by addition of CellTiter-Glo 3D Reagent (Promega, catalog no. G9683) and incubated at room temperature for 1 hour, then luminescence was measured with a microplate luminometer (BioTek Cytation5 Cell Imaging Multi-Mode Reader, Agilent Technologies). Percent cell growth inhibition for each test article at each concentration was calculated using the average luminescence signal using the corresponding untreated cells which were incubated with complete growth medium only, using the following formula for any given well: [1 − ([well RLU value]/[average RLU values from untreated wells]) × 100]. Results from each independent experiment were plotted as percent cell growth inhibition on the y-axis and log of the test article concentration (nmol/L) on the x-axis using Prism 9.5.1 (GraphPad Software). IC₅₀ for each test article was calculated using a four-parameter logistic regression analysis with variable slope using Prism 9.5.1 (GraphPad Software).

A total of 30,000 SK-BR-3 (HER2-high) and 10,000 MDA-MB-468 (HER2-negative) cells were seeded either as monocultures or cocultures in TC-treated 48-well microtiter plates (Greiner Bio-One, catalog no. 667180) and treated with 1.0 or 0.1 nmol/L ADCs diluted in growth medium. An untreated control, receiving growth medium only with no test article, was included. Cells were incubated with test articles at 37°C with 5% CO₂ for 4 days and detached by TrypLE Express Enzyme (Thermo Fisher Scientific, catalog no. 12604013). Cells were stained with the apoptotic dye YO-PRO-1 (Thermo Fisher Scientific, catalog no. Y3603) and an anti-HER2 antibody conjugated to AF647 (BioLegend, catalog no. 324412, RRID:AB_2262300). The stained cells were analyzed by flow cytometry on the BD LSRFortessa (BD Biosciences). For each sample, the main cell population was gated for single cells, then live cells were gated by exclusion of dead cells stained with YO-PRO-1 using the FITC channel. Finally, HER2-positive and negative cells were determined on the basis of the signals in the APC channel. For each test article, the percent cell viability relative to untreated cells was calculated by dividing the live single-cell count of treated cells by the live single-cell count of the untreated cells.

ADCs were diluted in mouse plasma (BioIVT, catalog no. MSE00PL38NCXNN) to 0.5 mg/mL and incubated in a 37°C water bath. Aliquots were taken out at 10 minutes, 1.5 hours, 8 hours, 24 hours, 72 hours, and 7 days and frozen at −80°C. Once all aliquots were collected, they were thawed and prepared for the affinity capture coupled to LC/MS analysis. Biotinylated goat anti-Human IgG F(ab')2 (Jackson ImmunoResearch, catalog no. 109-065-097, RRID:AB_2337629) was coupled to Streptavidin Mag Sepharose Beads (Cytiva, catalog no. 28-9857-99) for 30 minutes at room temperature before use. Mouse plasma samples containing approximately 2 μg ADC were diluted with PBS and treated with 0.25 μg of EndoS (recombinantly expressed in E. coli) for 1 hour at room temperature. Samples were incubated with the antibody-loaded beads for 1.5 hours at room temperature, washed with PBS pH 7.4, and reduced with 25 mmol/L dithiothrietol (DTT) for 1 hour at room temperature. ADC samples were then recovered by incubation with elution buffer (20% acetonitrile in water + 1% formic acid) for 1 hour at room temperature, and the supernatants containing the eluted ADC were collected and analyzed using an Agilent 1290 Infinity II HPLC (Agilent Technologies) coupled with an Agilent 6545A Q-TOF Mass Spectrometer (Agilent Technologies) equipped with a PLRP-S (1,000 Å, 8 μm, 50 × 2.1 mm) column (Agilent Technologies, catalog no. PL1912-1802) and a flow rate of 0.3 mL/minute. Mobile phase A was 0.1% formic acid, 0.025% trifluoroacetic acid, and 10% isopropanol in water (v/v) and mobile phase B was 0.1% formic acid and 10% isopropanol in acetonitrile (v/v). The gradient increased from 20% to 40% mobile phase B over 20 minutes. MassHunter (Agilent Technologies) qualitative analysis was used for deconvolution and data analysis.

For cell line–derived xenograft (CDX) and patient-derived xenograft (PDX) models, tumor cell suspensions or tumor fragments, respectively, were implanted subcutaneously in immunocompromised mice and upon tumor development were randomly assigned to groups and treated with a single intravenous dose of ADC. For the JIMT-1 breast cancer model (Crown Bioscience), 7–8 weeks old female CB17.SCID mice (Beijing HFK Bioscience Co.) were implanted with 5 × 10⁶ cells in 0.1 mL PBS and treated at a mean tumor volume (TV) of approximately 150 mm³ with 6 mice per group. For the OV-90 ovarian cancer model (Charles River Laboratories), 8-week-old female CB17.SCID mice (Charles River Laboratories) were implanted with 1 × 10⁷ cells in 0.1 mL PBS containing 50% Matrigel (BD Biosciences) and treated at a mean TV of approximately 125 mm³ with 7 mice per group. For pharmacokinetic analysis, serum was collected at 1, 24, 72, 168, 240, and 336 hours postdose. Total antibody concentrations were measured by sandwich ELISA as described below. For the HepG2 liver cancer model (Crown Bioscience), 8–10 weeks old female BALB/c nude mice (Beijing HFK Bioscience Co.) were implanted with 1 × 10⁷ cells in PBS containing 50% Matrigel and treated at a mean TV of approximately 150 mm³ with 6 mice per group. For the HT-29 colorectal cancer model (Crown Bioscience), 6–7 weeks old female BALB/c nude mice (Beijing HFK Bioscience Co.) were implanted with 3 × 10⁶ cells in 0.1 mL PBS and treated at a mean TV of approximately 150 mm³ with 5 mice per group. For the CTG-0958 ovarian cancer PDX model (Champions Oncology), 6–8 weeks old female Athymic Nude-Foxn1nu mice (Envigo), were implanted with tumor fragment harvested from stock animals and treated at a mean TV of approximately 225 mm³ with 3 mice per group. TV and animal body weight were monitored twice weekly throughout the duration of the study. Tumor diameters were measured with calipers and TVs were calculated according to the formula TV = width² × length × 0.5. Protocols and procedures were reviewed and approved by the Institutional Animal Care and Use Committee of the respective contract research organization prior to execution. During the studies, the care and use of animals were conducted in accordance with the regulations of the Association for Assessment and Accreditation of Laboratory Animal Care.

ADCs were administered to female BALB/c mice (6–8 weeks old, body weights ranging from 18.3 to 23.9 g at the initiation of dosing, sourced from The Jackson Laboratory) via 20 mL/kg intraperitoneal injections at dose levels of 60 and 200 mg/kg (6 animals/dose level). Mice were observed for up to 3 weeks following dosing. Mice were euthanized if body weight reduced by ≥20% from predose levels. At 1 week after dosing, 3 mice per group were subjected to scheduled necropsy, and selected organs were collected in 10% neutral buffered formalin for histologic processing and examination. Remaining 3 mice per group were observed for up to 3 weeks after dosing and euthanized later. For pharmacokinetic analysis, serum was collected for all mice at 24 hours and 7 days postdose.

ADCs were administered to female Sprague Dawley rats (9–10 weeks old, body weights ranging from 210 to 261 g at the initiation of dosing, sourced from Zhejiang Vital River Laboratory Animal Technology Co.) via slow intravenous injection over 10 minutes on day 1 and day 22 at the dose levels of 30, 60, and 200 mg/kg (6 animals/dose level). Bicinchoninic acid assay was used to analyze test article concentrations in all dose formulations. All the animals were evaluated for morbidity/mortality, clinical signs, body weight, food consumption, ophthalmic examinations, clinical pathology (hematology, serum chemistry, coagulation, and urinalyses), changes in organ weight, and macroscopic and microscopic changes in organs/tissues. Following first dose, blood samples were collected for toxicokinetic analysis of total antibody. Serum was collected at 2, 6, 12, 24, 96, 168, 336, and 504 hours postdose. Total antibody concentrations were measured by sandwich ELISA as described below. Scheduled complete necropsy was conducted at the end of dosing phase (3 animals on day 29) and recovery phase (3 animals on day 51).

Total antibody concentrations of ADCs were measured in murine or rat serum by sandwich ELISA. Nunc Maxisorp ELISA 384-well microplates (Thermo Fisher Scientific, catalog no. 464718) were coated with 2 μg/mL of anti-human IgG1 Fc capture antibody (Jackson ImmunoResearch Inc., catalog no. 709-005-098, RRID:AB_2340482) and incubated at 2°C–8°C overnight. The plates were washed three times with 100 μL/well of PBS pH 7.4, and prediluted serum samples (MRD 1:100) were added in duplicated and incubated at ambient temperature on plate shaker (700 RPM) for 1 hour. After plate washing, the bound test article was detected with 0.2 μg/mL horseradish peroxidase–conjugated anti-IgG1 Fab detection antibody (Jackson ImmunoResearch Inc., catalog no. 109-035-097, RRID:AB_2337585) for total IgG levels. After 1 hour incubation at ambient temperature on plate shaker (700 RPM) and plate washing, 3,3′,5,5′-tetramethylbenzidine was added to yield a measurable colorimetric change followed by addition of 1 mol/L HCl to stop the reaction. Absorbance at 450 nm was measured using a Synergy H1 Hybrid Multi-Mode Plate Reader (BioTek Instruments). Pharmacokinetic parameters were calculated from noncompartmental analysis using Phoenix WinNonlin software (Certara).

The sunburst diagram in Fig. 1A was generated using RawGraphs 2.0.1 (https://app.rawgraphs.io/). The data generated in this study are available upon request from the corresponding authors.

To design novel camptothecin analogs for conjugation, we leveraged six decades of camptothecin structure-activity relationship (SAR) data (Fig. 2). Camptothecin structural elements crucial for activity encompass the E-ring 20(S)-hydroxyl and lactone functionalities, the pyridone D-ring, and preserving the planarity of the ABCDE-ring system. The C-7 and C-9 positions can accommodate a variety of substituents, including additional fused rings as seen in exatecan and its derivatives, while substitutions within the C-12 to C-14 region significantly reduce potency. C-10 and C-11 can only accommodate small substituents, with 11-fluoro camptothecins displaying heightened potency, enabling the exploration of a more diverse array of substituents at C-7, C-9, and C-10 positions.

Guided by these insights, we focused on modifications at the C-7 (R¹) and C-10 (R³) positions of the camptothecin scaffold (Fig. 3A and B; see Supplementary Data for synthetic details), while keeping the 11-fluoro substituent constant and C-9 unsubstituted. Specifically, we sought to explore analogs with a hydrophilic functional group (e.g., amine, carbamate, urea, sulfonamide) bearing a linkable hydroxyl-group at the C-7 position with a methyl or methoxy group at C-10. In addition, we aimed to evaluate camptothecin analogs incorporating an amino group at C-10, providing an alternative position to install the linker. With these design criteria in mind, we prepared a library of approximately 100 camptothecin small molecules and evaluated them in vitro for their cytotoxicity against a panel of HER2-expressing cell lines, including SK-BR-3, Calu-3, SK-OV-3, and ZR-75-1 (Supplementary Table S1).

SAR analysis of these novel camptothecins and selected clinical small molecules (Fig. 3C), comparing their in vitro potency with their hydrophobic character, highlighted several trends. Notably, most compounds with a methyl group at C-10 position (Fig. 3C, orange) showed pIC₅₀ ≥ 8.0 (i.e., IC₅₀ < 10 nmol/L), with a large variability of hydrophobic character (clogD ranged from −1 to 3). Analogs with an amine at C-10 (Fig. 3C, green) were relatively more potent and more hydrophilic (i.e., with improved ligand-lipophilic efficiency) than C-10 methyl analogs. Conversely, camptothecins with a C-10 methoxy substituent (Fig. 3C, blue) were less potent without a significant improvement in hydrophilicity; therefore, none of the C-10 methoxy compounds were pursued further.

From this library, seven compounds (FD1–FD7, Fig. 3B) with either a methyl or an amino group at C-10, and encompassing a broad range of potency, hydrophilicity, and structural diversity at C-7 were selected for elaboration into drug-linkers (Fig. 3D, see Supplementary Data for synthetic details). For initial screening of these new camptothecin payloads as ADCs, we employed the clinically validated cleavable tetrapeptide sequence GGFG in combination with a polyethylene glycol (PEG)-containing maleimide, specifically the maleimido-triethyleneglycol (MT) propionyl moiety. PEG-containing maleimides have been employed with various ADCs across different payload classes and are known to help in compensating for suboptimal payload properties such as hydrophobicity and aggregation (21, 22).

The three C-10 methyl analogs bearing a carbamate (FD1), a urea (FD2), or a sulfonamide (FD3) at the C-7 position were linked through primary alcohols using a self-immolative aminomethylene spacer to obtain DL1–4. For analogs FD4, FD5, FD6, and FD7 the tetrapeptide cleavage sequence was directly attached to the C-10 amine via an enzymatically cleavable amide bond (23) to obtain DL9–14. Compound FD5, bearing both an alcohol group at C-7 and an amine at C-10 was linked two ways (direct peptide attachment at C-10 (DL11 and DL12) or a self-immolative aminomethylene group at the C-7 alcohol (DL5 and DL6)) allowing for direct comparison between the two linking strategies. In addition, DL7 and DL8 were prepared to allow direct comparison with ADCs carrying the clinically validated DXd payload with a matched maleimide anchor as well as an initial comparison of MC and MT anchors for this class of payloads.

Eight MT-containing drug-linkers (DL1, DL3, DL4, DL5, DL9, DL11, DL13, and DL14), representing seven unique payloads, along with DL7 (DXd benchmark) and DL6 (MC-containing FD5) were conjugated to the hinge disulfides of fully reduced trastuzumab (anti-HER2) to achieve a DAR of approximately 8. The resulting ADCs displayed minimal (<5%) aggregation and favorable hydrophilicity as assessed by HIC retention times relative to unconjugated trastuzumab (HIC-RRT, Table 1). Next, potency and antigen specificity of these ADCs were assessed in both a HER2-positive cell line (SK-BR-3) and a HER2-negative cell line (MDA-MB-468; Table 1; Fig. 4A). The three ADCs (T-DL1, T-DL3, T-DL4) carrying C-10-methyl payloads showed subnanomolar cytotoxicity in SK-BR-3 cells; however, T-DL4 displayed single digit nmol/L potency in HER2-negative MDA-MB-468 cells, suggesting a potential linker instability for this ADC. The four ADCs (T-DL9, T-DL11, T-DL13, T-DL14) with payloads bearing a C-10 amine showed markedly different results: ADCs carrying FD5 and FD6 were exceptionally potent in SK-BR-3 whereas ADCs with FD4 and FD7 were 100- to 1,000-fold less potent in the same cell line. This difference in ADC activity was unexpected given the relatively similar free drug potency and identical linker attachments. Notably, T-DL5, T-DL6, and T-DL11 showed equivalent activity regardless of linker attachment (C-7 or C-10) and anchor (MC vs. MT), highlighting how neither the linker position nor the anchor type influenced the in vitro cytotoxicity of these three different ADCs carrying the same FD5 payload.

Bystander activity is considered a key feature of T-DXd (24) which likely contributes to its activity observed in the clinic across patients with different HER2-expressing tumors, including patients with HER2-low, HER2-negative, or HER2-hetereogeneous tumors (9). ADCs bearing novel camptothecin payloads were compared with T-DXd in an in vitro bystander assay. Briefly, HER2-negative MDA-MB-468 cells were grown in either monoculture or in coculture with the HER2-high SK-BR-3 cells then treated with either 0.1 or 1 nmol/L of ADC (Fig. 4B and C). A known bystander inactive camptothecin ADC (T-ME-PEG2-GGFG-DXd2; ref. 25) was included as negative control. At 1 nmol/L concentration, ADCs bearing payloads FD1, FD5, and FD6 showed similar bystander activity to T-DXd. Conversely T-DL3 bearing FD2 showed lower bystander activity, and T-DL4 was active in monoculture, once again suggesting a possible linker instability. Interestingly, T-DL9 and T-DL14, which had the lowest potency in SK-BR-3, demonstrated the highest bystander activity. The superior bystander effect of these ADCs was particularly evident in the bystander assay at 0.1 nmol/L ADC concentration. However, excessive bystander activity could also be a double-edged sword, potentially leading to the rapid diffusion of a toxic payload into healthy tissues. Hence, when pursuing the most potent bystander payloads, it is imperative to assess the implications of high bystander activity in toxicology studies rather than relying on absolute values.

The stability of the nine ADCs was evaluated in mouse plasma for 7 days followed by immunoprecipitation-mass spectrometry analysis. No decomposition was observed for ADCs carrying payloads FD1, FD2, FD4, FD5, and FD7 linked either through the C-10 amine or the C-7 alcohol. Consistent with observations from the cytotoxicity and bystander assays, a loss of 393 m/z was observed for T-DL4 which corresponds to decomposition of the drug-linker to release an active small molecule (7,10-dimethyl-11-fluoro-camptothecin; Supplementary Fig. S1). In addition, rapid oxidation of the FD6 payload was observed for T-DL13 in plasma (Supplementary Fig. S1). These instabilities observed for both T-DL4 and T-DL13 underscore the significance of assessing in vitro plasma stability as a preliminary screening criterion for ADCs. Camptothecin small molecules and camptothecin-based ADCs are also known to be susceptible to photodegradation (26) necessitating careful handling to minimize light exposure. To assess photostability, the free drugs were incubated in pH 7.4 PBS at room temperature in either clear or amber vials. In the clear vials, partial or full degradation of all analogs was observed over time, with FD1, FD3, FD5, and FD6 showing 10%, 50%, 22%, and 100% loss of the parent compounds over 16 days, respectively (Supplementary Table S2). In contrast, all the free drugs were stable in the amber vials, except for FD6 which showed complete oxidation by day 16 (Supplementary Table S2). Moreover, the carbamate functionality in FD1 was stable in both acidic and basic conditions with no decomposition observed in 0.1 mol/L HCl (pH ∼1.0) or 0.1 mol/L Na₂CO₃ (pH ∼11.5) over 7 days (Supplementary Fig. S2).

To evaluate how different payload potency, bystander activity, and linker stability would affect in vivo efficacy, mice implanted with HER2-mid expressing JIMT-1 tumors were treated with a single 3 mg/kg ADC dose (Fig. 4D). The T-DM1–resistant JIMT-1 breast carcinoma cell line was chosen as a stringent in vivo CDX model to distinguish between test articles. Strikingly, sustained tumor regression was observed for mice treated with T-DL9 and T-DL14, which showed inferior in vitro two-dimensional (2D) cytotoxicity but superior bystander activity. ADCs that carry payloads FD1, FD3, FD5, and FD6, showed initial tumor regression followed by regrowth, comparable to the DXd-containing ADC T-DL7. All three ADCs carrying the FD5-payload demonstrated similar activity. ADC T-DL3, which carries the bystander inactive FD2, did not differentiate from vehicle control and demonstrated inferior pharmacokinetics versus the other ADCs in this model (Fig. 4E). Because of poor stability and/or lack of activity, ADCs with FD2, FD3, and FD6 were excluded from further evaluation.

For payload selection, the free drugs were screened using a panel of HER2-expressing cell lines in 2D culture. However, for the ADCs, minimal activity and/or non-sigmoidal curves were observed in most cell lines (Supplementary Fig. S3) when cultured in a 2D format, except for SK-BR-3. A lack of correlation from in vivo tumor growth inhibition efficacy (despite superimposable pharmacokinetics and linker stability for the majority of ADCs) and in vitro ranking of the ADCs in SK-BR-3 cultured in 2D, prompted us to investigate alternative assay formats that could better predict in vitro to in vivo translatability. It has been previously reported that a 3D spheroid assay may provide a more accurate representation of the in vivo tumor environment (27, 28). In addition, 3D assays have been used to test a variety of oncology drugs (29), including ADCs (30). Briefly, a 1:1 coculture of JIMT-1 cells and human dermal fibroblast were grown for 3 days in ultra-low attachment plates to generate tumor spheroids. The spheroids were then treated with a series of increasing concentrations of ADCs and incubated for 5 days to determine ADC IC₅₀ values, in a similar fashion to the well-established 2D cytotoxicity assays (Fig. 4F). Importantly, we observed a strong correlation (Fig. 4G) between the cytotoxicity of the ADCs evaluated in the spheroid assay (Fig. 4F) and the in vivo tumor growth inhibition results (Fig. 4D), highlighting the utility of the spheroid cytotoxicity assays for this class of payloads. In addition, this assay format enabled the measurement of IC₅₀ values for other cell lines (Supplementary Fig. S3) that were not suitable for conventional 2D cytotoxicity assays with camptothecin ADCs.

The encouraging results of the JIMT-1 CDX study with trastuzumab-ADCs prompted us to investigate the lead drug-linkers in the context of different antibodies targeting tumor-associated antigens (TAA) beyond HER2. It is known that trastuzumab-based ADCs tend to exhibit favorable behavior and can accommodate payloads with suboptimal properties. Therefore, testing novel drug-linkers with different antibodies allowed us to better select a lead payload. Conjugation of drug-linkers carrying payloads FD1, FD4, FD5, FD7, and DXd to an anti-folate receptor alpha (FRα) antibody produced ADCs anti-FRα-DL1, anti-FRα-DL9, anti-FRα-DL5, anti-FRα-DL11, anti-FRα-DL14, and anti-FRα-DL7 which were then evaluated for in vivo tolerability. In non–tumor-bearing BALB/c mice, the three ADCs carrying either FD1 or FD5 were well tolerated at both 60 and 200 mg/kg with minimal loss of body weight (Fig. 5A and B), similar to DXd conjugated to the same anti-FRα antibody (i.e., anti-FRα-DL7). However, mortality was observed in all mice treated with ADCs carrying FD4 or FD7 at both the 60 and 200 mg/kg doses demonstrating that the strong activity seen with the trastuzumab ADCs carrying these payloads comes at the cost of significant toxicity in rodents.

To compare in vitro ADME/DMPK properties of our lead payloads to the camptothecin analogs used in approved ADCs (DXd and SN-38) or in clinical trials (exatecan), FD1 and FD5 were tested for efflux pump susceptibility in a Caco-2 assay, stability in liver microsomes, and for plasma protein binding in different species (Supplementary Table S3).

In the Caco-2 assay, FD1, DXd, and SN-38 were identified as substrates for MDR1. FD1 and DXd were considered possible/marginal substrates for BCRP, due to the low efflux ratio in presence or absence of a BCRP inhibitor. In contrast, FD5 and exatecan were not substrates for either efflux pump. In the human liver microsome assay FD1, DXd, and exatecan underwent first-order metabolism with C-10 oxidation being the primary metabolite. DXd showed a further oxidation at the extra F-ring. Significantly more degradation was observed for exatecan, likely because of its benzylic amine. Minimal oxidation was observed for FD5 and importantly the C-7 alcohol was not a substrate for UDP-glucuronosyltransferases, a known source of gastrointestinal toxicity associated with the SN-38 prodrugs (e.g., irinotecan and SG; ref. 31). All compounds were highly plasma protein bound in humans and NHPs; however, FD5 was comparatively less bound in rodent species.

To identify a lead payload, drug-linkers with the MT propionyl anchor were used in the context of HER2 and FRα ADCs. For further platform development, MT was replaced with the clinically validated MC anchor. MC-linkers have been validated in five ADCs (trastuzumab deruxtecan, brentuximab vedotin, polatuzumab vedotin, enfortumab vedotin, tisotumab vedotin) approved by FDA for multiple indications. MC avoids highly stabilized linker-antibody conjugation which may result in the emergence of unexpected toxicities (19).

MC-containing DL2, DL6, DL8, DL10, and DL12 were used to generate corresponding FRα ADCs (anti-FRα-DL2, anti-FRα-DL6, and anti-FRα-DL12, along with the DXd-bearing anti-FRα-DL8 and the less tolerated anti-FRαa-DL10), which were evaluated in a two-dose tolerability study using Sprague Dawley rats (Fig. 5C–E). Anti-FRα-DL10 carrying FD4 was included to confirm that the lack of tolerability was consistent across species and was not significantly influenced by anchor selection (MT vs. MC). As seen in the mouse tolerability study, treatment with the FD4 ADC (anti-FRα-DL10) resulted in mortality of all rats following a single 60 or 200 mg/kg dose, and transient body weight loss following 30 mg/kg doses. Anti-FRα-DL2 demonstrated the greatest tolerability with no major macroscopic or microscopic findings at the 30, 60, and 200 mg/kg doses. Mortality was observed for 1 of the 6 animals at the 200 mg/kg dose for both ADCs carrying FD5 (anti-FRα-DL6 and anti-FRα-DL12) as well as transient bodyweight loss in all animals after a 200 mg/kg dose of anti-FRα-DL6. Importantly, pharmacokinetic analysis showed good dose proportionality for all ADCs across three dose levels and there were no discernible signs of payload-dependent ADC exposure (Supplementary Fig. S4). These observations reinforce the likelihood that the differences in tolerability can be attributed to the distinct properties of the payloads, rather than differences in ADC pharmacokinetics. On the basis of the ADC tolerability results and the overall small-molecule in vitro DMPK profile, FD1 (ZD06519) was selected as lead payload for the development of multiple ADC candidates.

To evaluate the breadth of activity of the ZD06519 (FD1) payload across different TAAs and tumor types, DL2 was conjugated to antibodies targeting FRα, NaPi2b, cMET, and GPC3. FRα is a clinically validated ADC target, with mirvetuximab soravtansine approved for treatment of patients with FRα-high ovarian cancer (32). NaPi2b is also overexpressed in ovarian and lung malignancies (33). In the OV-90 and CTG-0958 models, anti-FRα-DL2 and anti-NaPi2b-DL2 ADCs demonstrated sustained tumor growth inhibition following a single 6 mg/kg i.v. dose (Fig. 5F and G). cMET is a TAA highly expressed in non–small cell lung cancer, colorectal carcinoma, and other solid tumors and has been utilized as target for multiple ADC programs (34). In the cMET-expressing HT-29 colorectal cancer model, an anti-cMET-DL2 ADC showed strong antitumor activity following a single 6 mg/kg i.v. dose (Fig. 5H). Finally, elevated GPC3 expression occurs in hepatocellular carcinomas with limited expression found on healthy tissues (35). A single 8 mg/kg i.v. dose of anti-GPC3-DL2 ADC resulted in strong antitumor activity in the rapidly growing HepG2 model (Fig. 5I). The conjugation of ZD06519 (FD1) to antibodies targeting FRα, NaPi2b, cMET, and GPC3 yielded ADCs that were highly effective in preclinical tumor models, paving the way for potential therapeutic advancements in ovarian, lung, colorectal, and hepatocellular cancers. This diverse range of applications highlights the promise of ZD06519 (FD1) as a versatile platform for ADCs. In addition to evaluating efficacy in different tumor types, NHP toxicology studies have been performed with ZD06519 (FD1) ADCs. These studies have demonstrated highest non-severely toxic doses in NHPs for ZD06519 (FD1) ADCs up to 60 mg/kg as DAR8 and 120 mg/kg as DAR4 across multiple programs. The comprehensive preclinical development of novel ZD06519 (FD1) ADCs, including extensive in vitro and in vivo characterization, will be reported in future publications.

This work presents the development of ZD06519 (FD1), a promising camptothecin payload for ADCs, highlighting its favorable properties, efficacy, and tolerability in diverse preclinical in vitro and in vivo models. Camptothecin analogs have become the payload class of choice for ADCs entering clinical development, surpassing the number of new ADCs carrying microtubule inhibitors. Thus far, advanced clinical data are limited to a small number of topoisomerase I ADCs. In addition to the increasing number of drug candidates, multiple new drug-linker platforms have emerged with the aim of reducing hydrophobicity and ADC aggregation of known camptothecins, or to deliver novel camptothecin analogs (Fig. 1). Increasing payload potency and bystander activity holds the potential advantage of effectively targeting tumor cells with lower antigen expression. However, it is important to consider that more potent and more permeable payloads might also lead to increased toxicities, potentially requiring a reduction in the antibody dose. Higher antibody dose with a “moderately” potent payload could be a preferred strategy to overcome antigen sink effects and enhance tissue penetration (36). In addition to linker design and payload potency, intrinsic payload properties such as metabolic stability and susceptibility to transporters play a crucial role in ultimately determining the efficacy and safety profile for patients. New molecular entities can be selected to address these factors and optimize the overall therapeutic profile of ADCs based on the antibody, target, and indication. Striking the right balance between ADC efficacy and tolerability remains an ongoing challenge, making it a critical factor in designing a new ADC platform, ultimately enhancing the likelihood of clinical success.

Our study contributes to the evolving landscape of new ADC platforms by presenting the development and extensive characterization of a novel camptothecin, ZD06519 (FD1), designed specifically for application as an ADC payload. To identify this novel camptothecin lead payload with optimal properties, we leveraged insights gained from 60 years of camptothecin SAR data to generate a library of approximately 100 compounds featuring different substituents at the C-7 and C-10 positions. These new compounds were screened in a cytotoxicity assay against a panel of cell lines. Selected analogs spanning a range of potency and hydrophilicity were evaluated as ADC payloads using different linker strategies, while maintaining constant the tetrapeptide (GGFG) cleavable sequence in the linker, with the goal to select the most promising novel camptothecin payload. Two compounds (FD3 and FD6) were eliminated from consideration due to instability of their corresponding ADCs in mouse plasma, and further evaluation of FD2 was halted because of lack of in vitro bystander activity and limited in vivo efficacy in a JIMT-1 CDX study. ADCs with FD4 and FD7 demonstrated suboptimal cytotoxicity in 2D monoculture but exhibited higher bystander activity in the coculture assays, along with superior cytotoxicity in the in vitro spheroid assays. In the JIMT-1 CDX study in vivo, ADCs with FD4 and FD7 outperformed the other test articles; however, they featured significant toxicity in rodents. It is likely that the development of ADCs bearing FD4 or FD7 would result in a reduced antibody dose in humans, presenting potential hurdles in reaching sufficient target engagement and optimal tissue penetration for therapeutic efficacy. ADCs with FD5 demonstrated comparable efficacy in the JIMT-1 model to T-DL1 (FD1) but also increased toxicity in both Sprague Dawley rats and NHPs. The development of FD5 could face similar challenges, including the potential difficulty in achieving a significant antibody dose in humans.

Ultimately, FD1 (ZD06519) was selected as lead payload, on the basis of a favorable in vitro ADME/DMPK profile, in vivo efficacy, and the superior tolerability observed in rats and NHPs for its corresponding anti-FRα-DL2 ADC. Importantly, FD1 (ZD06519) has a unique structure and properties compared with other camptothecin payload used in ADCs currently in development, and it will likely differentiate in the clinic. This choice was further validated by the robust activity observed in multiple in vivo CDX and PDX models and the superior tolerability achieved in NHP tolerability studies with ZD06519 (FD1) ADCs across different TAAs. As ZD06519 (FD1) ADCs progress into clinical development, they hold the potential to establish ZD06519 (FD1) as an optimal ADC payload.

M.E. Petersen reports a patent for WO 2022/246576 pending, a patent for WO 2023/178452 pending, a patent for PCT/CA2023/051385 pending, and a patent for PCT/CA2023/051378 pending. M.G. Brant reports a patent for WO 2022/246576 pending, a patent for WO 2023/178452 pending, a patent for PCT/CA2023/051385 pending, and a patent for PCT/CA2023/051378 pending. M. Lasalle reports a patent for WO 2022/246576 pending, a patent for WO 2023/178452 pending, a patent for PCT/CA2023/051385 pending, and a patent for PCT/CA2023/051378 pending. S. Das reports a patent for WO 2022/246576 pending, a patent for WO 2023/178452 pending, a patent for PCT/CA2023/051385 pending, and a patent for PCT/CA2023/051378 pending. A.M.L. Wu reports a patent for PCT/CA2023/051378 pending. A. Hernandez Rojas reports a patent for WO 2023/178452 pending, a patent for PCT/CA2023/051385 pending, and a patent for PCT/CA2023/051378 pending. S O. Lawn reports a patent for WO 2023/178452 pending and a patent for PCT/CA2023/051385 pending. S.D. Barnscher reports a patent for WO 2022/246576 pending, a patent for WO 2023/178452 pending, a patent for PCT/CA2023/051385 pending, and a patent for PCT/CA2023/051378 pending. J.R. Rich reports a patent for WO 2022/246576 pending, a patent for WO 2023/178452 pending, a patent for PCT/CA2023/051385 pending, and a patent for PCT/CA2023/051378 pending. R. Colombo reports a patent for WO 2022/246576 pending, a patent for WO 2023/178452 pending, a patent for PCT/CA2023/051385 pending, and a patent for PCT/CA2023/051378 pending. No disclosures were reported by the other authors.

M.E. Petersen: Conceptualization, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. M.G. Brant: Conceptualization, investigation, visualization, methodology, writing–original draft, writing–review and editing. M. Lasalle: Conceptualization, investigation, methodology, writing–review and editing. S. Das: Investigation, methodology, writing–review and editing. R. Duan: Investigation, methodology, writing–review and editing. J. Wong: Investigation, methodology, writing–review and editing. T. Ding: Investigation. K.J. Wu: Investigation, writing–review and editing. D. Siddappa: Investigation, writing–review and editing. C. Fang: Investigation, writing–review and editing. W. Zhang: Investigation. A.M.L. Wu: Visualization, writing–review and editing. T. Hirkala-Schaefer: Investigation, methodology. G.A.E. Garnett: Investigation, methodology, writing–review and editing. V. Fung: Methodology, writing–review and editing. L. Yang: Methodology, writing–review and editing. A. Hernandez Rojas: Methodology, writing–review and editing. S.O. Lawn: Visualization, writing–review and editing. S.D. Barnscher: Conceptualization, supervision, project administration, writing–review and editing. J.R. Rich: Conceptualization, data curation, supervision, visualization, project administration, writing–review and editing. R. Colombo: Conceptualization, data curation, visualization, project administration, writing–review and editing.

The authors would like to acknowledge the following individuals for their contribution to this work: Kara White Moyes, Fariha Ahmed-Qadri, Madelyne Burcher, Samantha Michaels, Meredith Clark, Cathy Dang, Joel Smith, David Plotnik, Rupert Davies, Gerry Rowse, Hayato Tanaka, and Kari Frantzen.