FZ-AD005, a Novel DLL3-Targeted Antibody–Drug Conjugate with Topoisomerase I Inhibitor, Shows Potent Antitumor Activity in Preclinical Models

Small cell lung cancer (SCLC) is the most malignant type of lung cancer, accounting for approximately 15% to 20% of newly diagnosed lung cancers, with a 5-year survival rate less than 7% (1, 2). SCLC can be classified as limited stage or extensive stage. Approximately 70% of patients with SCLC present with extensive-stage disease at diagnosis with a median survival of approximately 10 months (3). For several decades, the standard first-line chemotherapy for SCLC has been etoposide combined with platinum. Despite its high response rate, most patients relapse within 6 months.

Recently, the combination of anti-PD1/PDL1 antibodies with chemotherapy has significantly improved survival of patients with SCLC (4–7), which makes it the recommended first-line regimen (8), however, most patients experience relapse within 10 months. Standard second-line therapy with topotecan has minimal clinical benefit, with a response rate of approximately 22% and an overall survival of approximately 8 months (9, 10). Lurbinectedin has been approved as second-line treatment for patients with SCLC with platinum-sensitive recurrence as it showed a response rate of 45%, a median duration of response of 6.2 months, and an overall survival of approximately 11.9 months (11). Only modest improvements in survival can be expected in SCLC with lurbinectedin, and there remains an urgent need for new treatments to improve outcomes for patients with SCLC.

Delta-like ligand 3 (DLL3), an atypical ligand of the Notch receptor family, is rarely expressed in normal tissues but is specifically expressed in neuroendocrine tumors such as SCLC (12). Approximately 70% of patients with SCLC showed high levels of DLL3 expression in tumor tissues (13). DLL3 has emerged as a therapeutic target in SCLC. Rovalpituzumab tesirine (Rova-T) is the first antibody–drug conjugate (ADC) against DLL3, which is composed of a humanized monoclonal antibody, rovalpituzumab, linked to a DNA-alkylating agent pyrrolobenzodiazepine (PBD) via a protease-cleavable linker. On average, each antibody was conjugated to two PBD molecules (14). Rova-T was administered at a dose of 0.3 mg/kg with longer dosing intervals (every 6 weeks) compared with other available ADCs due to the severe toxicities of Rova-T, such as pleural effusion, pericardial effusion, photosensitivity reactions, thrombocytopenia, and delayed systemic toxicity, those toxicities may be attributed to the payload of PBD (15). Despite its initial promise, Rova-T did not show a survival benefit in NCT03033511 (16) and NCT03061812 (17) phase III clinical studies. Therefore, the development of Rova-T was terminated.

Although Rova-T exhibited significant toxicity and low tolerability, it showed the potential of targeted therapy for SCLC in achieving disease control (disease control rate of Rova-T: 69.6%; ref. 13). DLL3 remains a target of interest for the detection and treatment of SCLC and other strategies are being developed to target DLL3 (18–20). In this study, we report the development of FZ-AD005, a novel DLL3-targeting ADC, with a topoisomerase I inhibitor, DXd. In contrast to Rova-T, FZ-AD005 employs DXd, a relatively less-toxic molecule, and achieves a high drug-to-antibody ratio (DAR) of 8.0. Strategic design aims to facilitate the delivery of more toxin molecules specifically to tumor cells. Preclinical studies showed that FZ-AD005 has potent in vitro and in vivo antitumor activities and exhibited a stable and acceptable PK profile, with a good safety profile in animals.

The parental antibody for FZ-AD005 was the FZ-A038 mAb. The sequences coding the light chain and heavy chain of FZ-A038 were cloned in an expression vector that was transfected into CHO K1 cells, followed by monoclonal cell line selection and antibody generation. Antibodies were purified by Protein A affinity chromatography (Cytiva, MabSelect SuRe LX), low pH inactivation, anion exchange chromatography (Nanomicrotech, NanoGel-50Q HC), and cation exchange chromatography (Nanomicrotech, NanoGel-50SP HP), followed by virus nanofiltration (Virosart Max and Virosart HF, Sartorius). The full sequences of FZ-A038 were provided in the Supplementary Material. A linker–payload was synthesized as described in Example 10 of patent WO2020259258A1 (21), and the structure of the linker–payload was LE14 in the above potent.

FZ-AD005 was prepared from fully reduced FZ-A038 mAb using excess tris (2-carboxyethyl) phosphine hydrochloride. The linker–payload was added to the reduced solvent at a 9 mol/L equivalent linker–payload/mAb ratio and incubated for 1 hour to produce the ADC intermediate. Subsequently, the ADC intermediate was purified by ultrafiltration/diafiltration to obtain FZ-AD005 with DAR 8. The isotype control ADC was produced in the same manner as FZ-AD005, with a comparable DAR. The SEC and DAR analysis methods are detailed in the “Supplementary Materials and Methods.”

Cell lines NCI-H2227 (HITES), NCI-H82 (RPMI 1640 + 10% FBS), DMS53 (Waymouth MB752/1 + 10% FBS), DMS79 (RPMI 1640 + 10% FBS), NCI-H1105 (HITES), NCI-H1963 (RPMI 1640 + 10% FBS), NCI-H660 (HITES + 10% FBS), NCI-H1092 (HITES), NCI-H889 (RPMI 1640 + 10% FBS), and HEK293 (DMEM + 10% FBS) were purchased from the ATCC. Ramos cell lines (RPMI 1640 + 10% FBS) were purchased from the National Infrastructure of Cell Line Resource. All cell lines were authenticated by short tandem repeat DNA profiling and Mycoplasma test was performed by the 4′6-diamidino-2-phenylindole DNA fluorescence staining method. After resuscitation, cell lines were passaged and cultured at 37°C with a 5% CO₂ atmosphere in appropriate media listed above. Cell lines were passaged for no more than 20 generations.

HEK293 cells expressing human DLL3 (HEK293-hDLL3) was generated by DLL3 plasmid transfection with Lipofectamine 3000 (Invitrogen) as transfection reagent. Positive clone cells were obtained by screening with 200 μg/mL hygromycin B. Monoclonal cell lines that stably expressed HEK293-DLL3 were obtained by limiting dilution analysis.

PDX tumors were added to a mixture of EnzymeH, EnzymeA, and EnzymeB tissue dissociation enzymes, disassembled using a gentleMACS Octo tissue processor, and filtered through a 70-µm filter membrane into a single-cell suspension. The prepared single-cell suspensions were then incubated with fluorescent antibody at 4°C for 30 minutes, protected from light, washed, and labeled. Finally, the labeled cells were subjected to flow cytometric analysis using BD LSRFortessa.

DLL3 extracelluar domains from human, rat, mouse and rhesus macaque (ECD; ACROBiosystems, Cat. # DL3-H52H4, DL3-R52H3, DL3-M52H9, DL3-R52H4) were coated onto the MAXISORP 96-well plates (Thermo Scientific, Cat. # 446469) overnight. After blocking with BSA, the plates were incubated with serially diluted FZ-AD005 (0.016–1,000 ng/mL) for 1 hour. Then the plates were washed and incubated with goat anti-human IgG (Fc)-HRP (Thermo, 31413). Tetramethylbenzidine was used as the chromogenic reagent, and the reaction was terminated with phosphoric acid. The plates were read at 450 nm using SpectraMax M2e (Molecular Devices). Human DLL1 and DLL4 (ECD; ACROBiosystems, Cat. # DL1-H52H8, DL1-H5227) were coated onto the plates overnight at 1 μg/mL. Serial dilutions of wild-type FZ-A038 were added to the plates. The experimental steps for the ELISA were consistent with those described above.

The antibody-dependent cell-mediated cytotoxicity (ADCC), antibody-dependent cell-mediated phagocytosis (ADCP), and complement-dependent cytotoxicity (CDC) of FZ-AD005 were measured. See “Supplementary Materials and Methods” for further information.

GatorPlus (Gator Bio) was used for the human DLL3 (ACROBiosystems, DL3-H52), FcγRI (Rhingen, QRE-133), FcγRIIa (Rhingen, QRE-135-A), FcγIIIa (Rhingen, QRE-137-A), and FcRn (ACROBiosystems, FCM H82W7) binding kinetics assays. Fab binding assay was first performed by capturing 1 μg/mL of FZ-AD038 or FZ-AD005 using a HFC biosensor (Gator, 160003). The biosensors were submerged in wells containing serially diluted human DLL3 (100∼0.78 nmol/L) to determine the binding kinetics. FcγRI, FcγRIIa, and FcγIIIa receptors were immobilized using a His biosensor (Gator, 160009). The FcRn receptor was immobilized using a streptavidin biosensor (Gator, 160002). After immobilizing, biosensors were submerged in wells containing serial dilutions of FZ-A038 and FZ-AD005. The analysis results were obtained using GatorPlus software.

FZ-AD005 and DS-8201 was diluted to 100 μg/mL with human plasma, and aliquots of 250 μL were incubated at 37°C. At specific times (days 1, 3, 5, and 7), samples were removed from the climate chamber and stored at −80°C until sample analysis. Measurements were performed by Waters ACQUITY UPLC H-Class Bio System with FLR Detector and Titanium flow cell. A ZORBAX Eclipse Plus C18 column (4.6 × 2.1 mm, 3.5 μm) was used with a 20 minutes LC gradient. The mobile phase A was composed of MilliQ-grade water with 0.1% formic acid, and mobile phase B consisted of 0.1% formic acid in MilliQ-grade acetonitrile. The starting gradient conditions were 90% solvent A and 10% solvent B, with a steep ramp to 70% solvent B by 10 minutes. Then 100% solvent B was held for 3.9 minutes. The percentage of drug release was calculated as the ratio of released DXd, determined by UPLC, to the hypothetical total amount of DXd conjugated to FZ-AD005.

The internalization of FZ-AD005 was evaluated by flow cytometry, and intracellular localization was observed by laser scanning confocal microscopy. See “Supplementary Materials and Methods” for further information.

For this assay, 4.5 × 10⁶ DLL3-transfected HEK293 cells were seeded in T25 cell culture flasks. After overnight culture, the cells were treated with 200 pmol FZ-AD005, and the culture medium was collected after 2, 4, 6, 24 and 48 hours of treatment. The collection of supernatant was diluted and sedimented with internal standard of 30 nmol/L Exatecan mesilate which was prepared in a mixture of perchloric acid:methanol:water at the ratio of 1:20:20. Then the standard curve with 20 nmol/L exatecan mesilate was achieved in solution of perchloric acid:methanol:water at the ratio of 1:20:40. The quantification of extracellular and intracellular DXd was analyzed by LC/MS-MS (Waters, Vion IMS QTof) with MRM mode. The solvent system for LC was comprised with liquid phase method: mobile phase A: acetonitrile + 0.1% formic acid; phase B: H₂O + 0.1% formic acid.

To measure the bystander killing effect of FZ-A038 and FZ-AD005, NCI-H82 and Ramos cell lines were used as DLL3-positive and -negative cells, respectively. NCI-H82 and Ramos cells were diluted with 1 μg/mL FZ-A038 and FZ-AD005 at 3.5 × 10⁵ cells/mL. For the coculture study, Ramos cells were plated in the bottom chamber of 24-well cell culture plates (Corning, catalog No. 3472-Clear), whereas NCI-H82 cells were plated in the upper Transwell permeable inserts, with a membrane pore size of 0.3 μm. In this model, any potential bystander factor can diffuse through the permeable membrane. After 144 hours of incubation with ADCs or medium control, the cell inserts in the upper and lower wells were counted using a Countess automated cell counter (Invitrogen, Thermo Fisher Scientific, USA).

Cells were seeded in a 96-well plate at the following concentrations: NCI-H1105 (6,000 cells per well), DMS 53 (1,000 cells per well), DMS 79 (5,000 cells per well), and NCI-H2227 (4,000 cells per well). After overnight incubation, the diluted substances were added and then cell viability was evaluated after 144 hours using the CellTiter-Glo Luminescent Cell Viability Assay (Promega Corp. Madison, WI, USA) according to the manufacturer’s instructions.

The sensitivity of SCLC cell lines to DXd, topotecan, and cisplatin was evaluated. See “Supplementary Materials and Methods” for further information.

All procedures related to animal handling, care, and treatment in the studies were performed according to the guidelines approved by the Institutional Animal Care and Use Committee (IACUC) of Lanly Bioscience (Suzhou, China), Crown Bioscience (Suzhou, China), and TriApex Laboratories (Nanjing, China), following the guidance of the Association for Assessment and Accreditation of Laboratory Animal Care.

SCLC NCI-H889 and human neuroendocrine prostate cancer (NEPC) NCI-H660 models were established by subcutaneously injecting 10 × 10⁶ cells suspended in Matrigel/PBS (1:1) into female and male NOD/SCID mice, respectively. SCLC NCI-H82 model was established by subcutaneously injecting 5 × 10⁶ cells suspended in Matrigel/PBS (1:1) into female BALB/c nude mice.

Group assignment was carried out when the tumor volume reached approximately 140 to 170 mm³. Tumor-bearing mice were randomly divided into vehicle control, FZ-AD005 treatment, or other test groups with eight mice in each group. In the NCI-H889 and NCI-H660 models, tumor-bearing mice were treated with FZ-AD005 or relevant controls intravenously as a single dose on the day of grouping. In the NCI-H82 model, tumor-bearing mice were treated with FZ-AD005 or relevant controls intravenously weekly on the day of grouping for a total of two doses. Tumor volume was measured twice a week, and when the average tumor volume of the vehicle control group reached approximately 2,000 mm³, the antitumor activity was assessed. Tumor growth inhibition (TGI) was calculated according to the following formula: 100 × [1−(average tumor volume of the treatment group)/(average tumor volume of the vehicle control group)].

The SCLC patient-derived xenograft (PDX) models LU-5236, LU-5184, LU-5217, LU-5242, LU-5250, and LU-5171 were provided by Crownbio. Group assignment was performed when the tumor volume reached approximately 130 to 220 mm³. Tumor-bearing mice were intravenously injected with FZ-AD005 or 5% glucose on the grouping day. The frequencies of tumor volume measurements, experimental endpoints, and evaluation of antitumor activity in the PDX model were consistent with those in the cell line–derived xenograft (CDX) models.

FZ-AD005 was administered intravenously to cynomolgus monkeys once at 1, 3, and 10 mg/kg. Serum concentrations of FZ-AD005, total antibody (drug-conjugated and -unconjugated antibody), and whole-blood concentrations of DXd were measured up to 672 hours postdose. FZ-AD005 and total antibody were determined by ELISA, and DXd was determined by LC/MS-MS. Methods are detailed in the “Supplementary Materials and Methods”. Below the lower limit of quantification (BLQ) for FZ-AD005 and total antibody: <100 ng/mL; BLQ for DXd: <0.0500 ng/mL. Pharmacokinetic parameters were calculated by noncompartmental analysis using Phoenix WinNonlin 8.3.3.33.

FZ-AD005 was intravenously administered once a week (four times in total) to SD rats, the non–cross-reactive species, or once every 3 weeks (five times in total) to cynomolgus monkeys, the cross-reactive species. Clinical observation and detection, clinical pathology, immunogenicity analysis, and pathologic examination (end of the administration and recovery periods) were monitored throughout the study. Necropsies were performed on rats and monkeys in the main group 1 week after the last dose. At the end of the recovery phase, the remaining animals were necropsied and assessed for the reversibility of the toxicity changes in rats and monkeys.

Statistical analysis in CDX and PDX studies were performed using SPSS 18.0 and R 3.6.0, respectively. The independent samples t test was used to compare the treatment groups and the vehicle group in the CDX model studies, and P < 0.05 was considered to be significant. In the PDX model study, a two-way analysis of variance with the Bartlett test was used to compare the treatment and vehicle groups, and statistical significance was set at P < 0.05. Statistical analysis of the toxicity studies was performed using Provantis 10.5.0, which has built-in SAS, SPSS 22.0, and Microsoft Excel.

The data generated in this study are available in the manuscript and Supplementary Materials.

Surface expression of DLL3 is evident in a variety of human Lung carcinoma cell lines, including NCI-H2227, NCI-H82, DMS53, DMS79, NCI-H1105, NCI-H1963, NCI-H889, and NCI-H1092 cell lines, as well as DLL3-transfected HEK293 cells (Supplementary Fig. S1). The relative MFI of NCI-H2227, NCI-H82, NCI-H209, DMS53, DMS79, NCI-H1105, NCI-H1963, NCI-H889, NCI-H1092 and DLL3 transfected HEK-293 were 5.9, 2.1, 3.4, 3.0, 3.0, 3.1, 1.9, 2.5 and 3.3, respectively. These results indicated that DLL3 was positively expressed in these cell lines while with generally low expression levels. This observation was consistent with Giffin findings that a maximum of approximately 3200 DLL3 molecules were present on the surface of the cell line (19). Six tumor tissues from PDXs were examined for DLL3 expression. In LU-5184, LU-5236, and LU-5217 PDX tumors, flow cytometric analysis showed a comparatively high positive expression of DLL3, whereas LU-5242, LU-5171, and LU-5250 PDX tumors showed a low positive expression of DLL3 (Fig. 1).

Female BALB/c mice were subcutaneously immunized with DLL3 (ACRO, DL3-H5255), and splenocytes of mice were fused with SP2/0 cells. 1A5, a mouse anti–human DLL3 antibody with a characteristic of rapid internalization was obtained by immunization and hybridoma screening. The variable region of 1A5 was fused to the constant region of human IgG1 to form a chimeric antibody. The Fc region of this chimeric antibody was introduced with Fc mutations (L234F/L235E/P331S; Kabat; ref. 22) to silence the Fc effector function and to form the FZ-A038 mAb. The DNA topoisomerase I inhibitor DXd was conjugated to FZ-A038 via a novel cathepsin B–cleavable dipeptide valine–alanine (Val–Ala) based linker to generate FZ-AD005, and the structure of FZ-AD005 is shown in Fig. 2A. Using RPLC, the DAR of FZ-AD005 was determined to be 8.0 (Supplementary Fig. S2). In addition, the SEC purity was 98.8% confirmed by HPLC (Supplementary Fig. S3). Then the in vitro stability of FZ-AD005 in human plasma was evaluated. The release rate of DXd from FZ-AD005 after 7 days of incubation was approximately 0.7%, whereas the release rate of DXd from DS-8201 is 1.6% (Fig. 2B). These results demonstrated that FZ-AD005 remained stable under physiologically relevant conditions.

Both FZ-AD005 and FZ-A038 mAbs exhibited high affinity binding to human DLL3 with KD of 58.3 and 65.9 pmol/L, respectively (Fig. 2C; Supplementary Table S1). During antibody discovery, wild-type FZ-A038 showed specific binding to human DLL3 but not to DLL1 and DLL4 (Fig. 2D). Species cross-reactivity revealed that FZ-AD005 exhibited similar affinities for human, mouse, and monkey DLL3 but not for rat DLL3 protein (Fig. 2E). These results suggest that the cynomolgus monkey is an appropriate species for nonclinical pharmacokinetic and toxicity studies of FZ-AD005.

For ADCs, FcγR-mediated effector functions may contribute to target-independent uptake and toxicity in normal cells (23). Notably, ADCs can induce off-target toxicity by binding to FcγR on target-negative cells (24). For instance, severe thrombocytopenia occurred in T-DM1 treatment that is mediated by the antibody binding to FcγRIIa (25, 26). To mitigate these potential risks, Fc-silencing technology was introduced into FZ-AD005 to eliminate FcγRs binding. Binding analysis using biolayer interferometry indicated that FZ-AD005 showed a lower or negligible affinity for FcγRI, FcγRIIa, and FcγIIIa receptors, with KD values of 550 nmol/L, 13.29 µmol/L, and 6.1 µmol/L, respectively (Supplementary Table S2). Notably, FZ-AD005 bound to FcRn with a KD value of 29 nmol/L (Supplementary Fig. S4A–S4D; Supplementary Table S2). These results were consistent with the binding profiles of Fc-silencing antibodies (22, 27). As measured by the PBMC assay, neither FZ-AD005 nor FZ-A038 elicited robust ADCC, CDC, or ADCP activity, even at an antibody concentration up to 10 µg/mL (Fig. 3).

After binding to tumor surface antigens, ADC is internalized into the lysosome and releases the drug. The free drug reaches the cytosol and binds to its target (microtubules or DNA) triggering cell death (28–30). The rapid internalization of FZ-AD005 was observed within half an hour after the ADC binding to the cell surface DLL3 in HEK293-DLL3 cells (Supplementary Fig. S5). Intracellular trafficking of FZ-AD005 in NCI-H82 cells were studied by confocal microscopy. Colocalization between FZ-AD005 and lysosomal was observed within half an hour of cell internalization (Fig. 4A). In addition, the release of DXd from FZ-AD005 in transfected HEK293 cells was detected by LC/MS. Extracellular DXd release was observed after treatment with 200 pmol of FZ-AD005 in HEK293-DLL3 cells for 2 to 48 hours. Notably, the quantity of DXd released into the culture medium displayed a time-dependent trend over the 48-hour observation period (Fig. 4B).

Bystander cytotoxic effects contribute to a more comprehensive eradication of heterogeneous tumor tissues by ADCs (31). The bystander effect of DXd-based ADCs such as DS-8201 has been demonstrated previously (31, 32). Whether FZ-AD005 exhibits a similar bystander effect remains unclear. To explore this, Dual-chamber Transwell dishes were used to test whether ADC-treated DLL3-positive NCI-H82 cells induced bystander cell death in DLL3-negative Ramos cells. Analysis of NCI-H82/Ramos cocultures revealed that treatment with 1 μg/mL FZ-AD005 effectively eliminated both the cell populations. Although treatment with 1 μg/mL FZ-A038 mAb and IgG-ADC did not show any noticeable cytotoxic effects (Fig. 4C). These results indicated that FZ-AD005 had an obvious bystander effect.

The in vitro cytotoxicity of FZ-AD005 was assessed against various SCLC cell lines with DXd, FZ-A038 and IgG-ADC as comparisons. Cytotoxicity was evaluated by IC₅₀ values after 7 days of incubation. As shown in Fig. 5, FZ-AD005 exhibited potent cytotoxicity against DMS53, DMS79, NCI-H1105, and NCI-H2227, with IC₅₀ values ranging from 0.152 to 4.295 nmol/L. In contrast, FZ-A038 showed no cytotoxicity, and IgG-ADC showed no specific cytotoxicity. It suggested that the antitumor cytotoxic effect of FZ-AD005 was due to the selective delivery of DXd to tumor cells through FZ-A038. All antigen-positive cell lines were sensitive to DXd, with picomolar IC₅₀ values.

Moreover, the sensitivity of SCLC cell lines to DXd, topotecan, and cisplatin was evaluated. Data revealed that DXd outperformed topotecan in all SCLC cell lines, particularly in platinum-sensitive cell lines, such as NCI-H209, NCI-H526, and DMS153. DXd also exhibited good cytotoxicity against the platinum-resistant cell lines (Supplementary Fig. S6). These results support the potential of DXd for the treatment of patients with SCLC.

The above data indicated that FZ-AD005 was able to induce potent cytotoxic effects in vitro. To examine whether these characteristics can translate into antitumor efficacy in vivo, we evaluated FZ-AD005 in two SCLC CDX models (NCI-H889 and NCI-H82) and one NEPC CDX model (NCI-H660), as well as one SCLC PDX model (LU-5236). In all tested models, FZ-AD005 demonstrated substantial antitumor activity. For NCI-H82, treatment with doses of 3 and 5 mg/kg led to tumor growth inhibition (TGI) of 94.19% and 97.13%, respectively. Similarly, in NCI-H889, doses of 2.5, 5, and 10 mg/kg resulted in TGIs of 86.21%, 94.53%, and 95.54%, respectively. Surprisingly, in the NCI-H660 model, all dose groups exhibited complete tumor growth inhibition compared with the vehicle control group. In the LU-5326 PDX model, FZ-AD005 exhibited significant inhibition with TGI of 98.51% (1.5 mg/kg), 99.30% (3 mg/kg), and 100.00% (6 mg/kg) on day 21 (Fig. 6). In addition, FZ-AD005’s antitumor activity was notably superior to that of lurbinectedin in the NCI-H82 model. Lurbinectedin achieved a TGI of only 23.48% (at 0.18 mg/kg). To substantiate the tumor-targeting specificity of FZ-A038, the NCI-H889 model was employed. FZ-AD005 inhibited tumor growth in a dose-dependent manner. In contrast, 10 mg/kg of FZ-A038, IgG-ADC, and the free DXd equivalent of the ADC dose failed to suppress tumor growth. No obvious body weight loss was observed in the mice treated with FZ-AD005 in any of the models. These results suggested that FZ-AD005 had therapeutic potential for treating DLL3-expressing SCLC.

FZ-AD005 was administered intravenously at 1, 3, and 10 mg/kg to cynomolgus monkeys. Serum concentration of FZ-AD005 and total antibody, and whole-blood concentration of DXd versus time after treatment with each dose of FZ-AD005 are shown in Fig. 7. Pharmacokinetic analysis of serum concentration revealed that the terminal elimination half-life (t_1/2) of FZ-AD005 increased with increasing doses. The clearance of FZ-AD005 was 0.968 ± 0.189, 1.23 ± 0.412, and 1.04 ± 0.249 mL/hour/kg at doses of 1, 3, and 10 mg/kg, respectively. The AUC_0–t of FZ-AD005 was dose-proportional and showed an approximately linear pharmacokinetic profile. Low concentrations of free DXd were detected at the limited points for each dose, which suggests that the FZ-AD005 linker is stable in vivo. The pharmacokinetic parameters of FZ-AD005, total antibody, and DXd at each dose are shown in Supplementary Table S3.

To assess the general toxicity profile of FZ-AD005, repeated-dose toxicity studies were conducted in cynomolgus monkeys as a cross-reactive species, and in rats as a non-cross-reactive species. FZ-AD005 was well tolerated in both species, with no fatal cases or life-threatening toxicities at any dose level throughout the study period. A summary of repeated-dose toxicity studies is provided in Supplementary Table S4.

The toxicity of FZ-AD005 was evaluated in a repeat-dose study in rats. FZ-AD005 was administered at doses of 10, 30, and 100 mg/kg once a week for four doses, with a 6-week recovery period. The main toxic target organs were the thymus, spleen, testes, epididymis, male mammary glands, and the skin. However, after the recovery period, signs or trends of organ recovery were noted. Cynomolgus monkeys received five doses of FZ-AD005 intravenously every 3 weeks. The primary toxic reactions observed in monkeys included vomiting and decreased food intake; however, no severe changes were observed in gastrointestinal tissues. Additionally, a decrease in the reticulocyte count was observed only in the 30 mg/kg group. The thymus exhibited toxicity with reduced size, organ weight, and lymphocyte counts; however, these effects tended to recover after the recovery period. Both species showed no significant abnormalities in food consumption, body weight, cardiovascular function, neurologic properties, or respiratory functions. Based on these findings, the highest nonseverely toxic dose was determined to be 100 mg/kg in rats and 30 mg/kg in monkeys.

In this study, we report the development of a DLL3-specific ADC, the FZ-AD005. FZ-AD005 utilizes a novel and stable linker–payload based on the Val–Ala dipeptide linker and DXd (Fig. 2A). FZ-AD005 was designed with a DAR of eight to deliver a greater quantity of toxin to tumor cells and to maximize the therapeutic window. As expected, FZ-AD005 was stable in human plasma in vitro, showed excellent antitumor activities against DLL3-positive tumors in preclinical models, and was well tolerated in monkeys.

FZ-AD005 effectively inhibited the growth of DLL3-expressing cancer cells in vitro (Fig. 5) and exhibited potent antitumor activity in vivo (Fig. 6), even when DLL3 expression levels were relatively low (Supplementary Fig. S1), which may be explained by the rapid internalization of DLL3 in tumor cells. This result is similar to the observation from Saunders study, in which modest expression of DLL3 permits ADC-mediated cytotoxicity in vitro and in vivo (14). Moreover, in our study, all SCLC cells tested exhibited greater sensitivity to DXd than to topotecan, irrespective of their platinum sensitivity (Supplementary Fig. S5). This underscores the substantial potential of DXd in the treatment of SCLC. It is important to note that DLL3 antigen expression was required for cells to be sensitive to FZ-AD005; however, the sensitivity of tumor cells to ADCs seems to be more closely related to the toxin activity (33). For instance, DS-8201 demonstrated excellent antitumor activity in breast cancers with low levels of Her2 expression (34).

The Val–Ala linker of typical ADCs connects with payloads via p-aminobenzyl (PAB) as the spacer, a methylsulfonyl modification was applied on the PAB group to enhance the plasma stability of FZ-AD005. FZ-AD005 exhibited excellent plasma stability in vitro compared with DS-8201 and precise DXd release upon internalization into the lysosomes (Figs. 2B and 4B). In vivo, minimal differences were observed between the ADC and total antibody pharmacokinetic profiles, further demonstrating the stability of this innovative linker–payload (Fig. 7). Moreover, low plasma concentrations of free DXd resulted in minimal systemic exposure, with an AUC_0–t of approximately 18.6 ± 8.82 (hour × ng/mL) at 10 mg/kg (Supplementary Table S3). The reduced exposure may account for the reduced toxicity of FZ-AD005 in monkeys. Linker plays a critical role in influencing ADC stability, pharmacokinetics, efficacy, and toxicity profiles (35, 36). Stabilized linkers may mitigate off-target toxicity associated with linker cleavage: a problem often encountered in conventional ADCs. For instance, DS-8201 demonstrated higher stability and lower free drug release than conventional ADCs (37, 38) and exhibited a compelling safety profile in clinical applications (39, 40).

Fc-silencing technology was introduced in FZ-AD005 to eliminate binding to FcγRs (Supplementary Fig. S4A–S4C; Supplementary Table S2), as most dose-limiting toxicities are thought to be caused by irrelevant uptake-mediated off-target toxicity in normal cells (41–43). The Fc-silencing mutation was also applied in durvalumab, an anti-PDL1 antibody, and clinical data of durvalumab demonstrated that Fc-silencing mutation do not induce severe immunogenicity which may affect efficacy and safety of the drug. In vitro studies showed that FZ-AD005 exhibited negligible ADCC, CDC, and ADCP effects (Fig. 3). The absence of severe hematologic and gastrointestinal toxicity observed in monkeys and rats may be attributed to the Fc-silencing mutations in FZ-AD005. Furthermore, if the binding of Fc with FcRn is removed, the ADC will be released rapidly, resulting in low tolerance (43). FZ-AD005 retained its binding affinity to FcRn (Supplementary Fig. S4D; Supplementary Table S2) to ensure that it will not be quickly removed (reflected by the half-life; Fig. 7), which not only assures efficacy but also enhances the tolerability of the ADC (43). Nonetheless, further research is essential to confirm this association and gain a comprehensive understanding of the underlying mechanisms. Moreover, the antibody of FZ-AD005, FZ-A038 mAb, is a chimeric antibody, and preparation of ADCs with chimeric antibodies is also seen in Adcetris, the immunogenicity of which is overall controllable. No serious immunogenic reactions of FZ-AD005 were observed in repeated toxicity studies in SD rats and cynomolgus monkeys. However, there may still be some immunogenic risks in clinical studies, and it is worth further attention and prevention and management in subsequent clinical studies.

Severe pulmonary toxicity was observed in monkeys treated with DS-1062, DS-7300, and DS-8201, whereas no pulmonary toxicity was observed with DXd (40, 44–46). Furthermore, clinical studies have reported serious and potentially life-threatening adverse events related to interstitial lung disease in association with DS-8201a (47). These severe pulmonary toxicities may be attributed to the expression of the targets of these compounds in normal lung tissue. Interestingly, no similar pulmonary toxicity was observed with FZ-AD005 in our study. This observation may be due to little DLL3 expression in normal lung tissues. FZ-AD005 seemed to have a better safety profile compared with previously reported DXd-based ADC drugs (40, 44, 45, 48). Further clinical studies are required to verify this favorable safety profile.

In summary, preclinical studies demonstrated the excellent antitumor activities of FZ-AD005 both in vitro and in vivo with an acceptable safety profile in monkeys. These promising results indicated that FZ-AD005 has the potential to be a valuable therapeutic option for patients with SCLC. The first-in-human clinical study has been approved by the NMPA, and the clinical trial is expected to be initiated in April 2024.

Q. Guo reports a patent for WO2020259258 A1 issued and licensed and a patent for WO2023006084 A1 issued. B. Gao reports a patent for WO2020259258A1 issued and licensed and a patent for WO2023006084 A1 issued. Y. Zhang reports a patent for WO2020259258A1 issued and licensed. T. Yang reports a patent for WO2020259258A1 issued and licensed and a patent for WO2023006084A1 issued. No disclosures were reported by the other authors.

Q. Guo: Conceptualization, formal analysis, supervision, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. B. Gao: Data curation, formal analysis, supervision, investigation, methodology, writing–original draft. R. Song: Data curation, formal analysis, supervision, investigation, methodology, writing–original draft. W. Li: Data curation, formal analysis, supervision, investigation, methodology, writing–original draft. S. Zhu: Conceptualization, formal analysis, supervision, investigation, methodology, visualization. Q. Xie: Data curation, formal analysis, supervision, investigation, methodology, writing–original draft. S. Lou: Data curation, investigation, writing–review and editing. L. Wang: Data curation, investigation, and methodology. J. Shen: Data curation, investigation, methodology. T. Zhao: Data curation, investigation, methodology. Y. Zhang: Data curation, investigation, methodology. J. Wu: Data curation, supervision, investigation, methodology. W. Lu: Conceptualization, resources, supervision, investigation, methodology, writing–review T. Yang: Conceptualization, resources, supervision, investigation, methodology, writing–review and editing.

Preclinical studies were funded by Shanghai Fudan-Zhangjiang Bio-Pharmaceutical Co., Ltd. The authors would like to thank all the researchers in Fudan-Zhangjiang and the external collaborators who provided their expertise and assistance in this study.