Preclinical Antitumor Activity of a Novel Anti–c-KIT Antibody–Drug Conjugate against Mutant and Wild-type c-KIT–Positive Solid Tumors Free

The c-KIT (CD117) receptor binds the ligand stem cell factor (SCF), resulting in receptor homodimerization and activation of its tyrosine kinase activity along with cKIT internalization (1). Activation of c-KIT regulates cellular functions, including apoptosis, differentiation, and proliferation (2, 3). Gain-of-function mutations resulting in constitutive c-KIT activation play a central pathogenic role in gastrointestinal stromal tumors (GIST), seminomas, subsets of melanoma, acute myelogenous leukemia (AML), and acute systemic mastocytosis (ASM; refs. 4, 5). Approximately 95% of GISTs express c-KIT, with the majority (75%–90%) harboring activating mutations (6), leading to uncontrolled proliferation and resistance to apoptosis. Imatinib, a small-molecule inhibitor of c-KIT, significantly reduces proliferation of GIST cell lines, supporting the importance of c-KIT as a driver of proliferation in subsets of GIST. The small-molecule inhibitors exhibit preferences for selective c-KIT mutations (e.g., imatinib is effective in GIST patients harboring exon 11 mutations), while being less effective against secondary c-KIT mutations (leading to resistance in ∼50% of patients within 2 years) and remain mostly inactive against diseases with wild-type c-KIT (7). Beyond GIST, wild-type c-KIT is overexpressed in small cell lung cancer (SCLC), AML, non–small cell lung cancer (NSCLC), melanoma, adenoid cystic carcinoma, and Merkel cell carcinoma (8).

Antibody–drug conjugates (ADC) leverage the specificity of monoclonal antibodies to deliver highly potent cytotoxic agents to antigen-positive tumor cells. Here, we describe a novel c-KIT targeting ADC, LOP628, consisting of the cytotoxic maytansinoid, DM1, covalently linked via N-succinimidyl-4-(N-maleimidomethyl)cyclohexane-1-carboxylate (SMCC) to lysine residues of the antibody LMJ729, which is the same linker-payload utilized by the HER2-targeting ADC, Kadcyla (Trastuzumab-DM1; refs. 9, 10). Upon binding to c-KIT, LOP628 becomes internalized and lysosomally processed to yield the active catabolite Lys-N^ϵ-MCC-DM1, resulting in cell-cycle arrest (11, 12) and potent activity against mutant and wild-type c-KIT–positive tumor models. This demonstration of compelling activity in models of SCLC, AML, and GIST (with single and double c-KIT mutations) suggests that LOP628 has the potential for treating patients with these diseases.

The murine anti–c-KIT antibody 9P3, humanized IgG1/κ anti–c-KIT antibody LMJ729 (derived from 9P3), and the isotype control human IgG1/κ antibody (IgG1) were generated at Novartis. The anti–c-KIT and the IgG1 isotype control antibodies were directly conjugated to allophycocyanin (APC). The APC–fluorochrome-conjugated anti-human IgG antibody was from Southern Biotechnology, # 9042-11. Conjugation of antibodies to DM1 was performed using the cross-linking agent N-succinimidyl-4-(N-maleimidomethyl)cyclohexane-1-carboxylate (SMCC) at ImmunoGen, Inc. Membrane-permeable maytansine (L-DM1-Me) was kindly provided by ImmunoGen, Inc.

Cell lines were purchased from ATCC [NCI-H526 (CRL-5811), NCI-H1048 (CRL-5853), MDA-MB-468 (HTS-132), MDA-MB-453 (HTB-131), KU812 (CRL-2099), Kasumi-6 (CRL-2775), Kasumi-1 (CRL-2724), NCI-H2170 (CRL-5928)], or DSMZ [Mo7e (ACC-104)]. NCI-H1048 cells were obtained from the Broad Institute–Novartis Cancer Cell Line Encyclopedia collection. The GIST-T1 line with a heterozygous KIT 52-bp deletion in exon 11 was provided by Dr. Taguchi from Kochi University, Japan (13), while GIST 882 with a homozygous missense mutation in KIT exon 13 (14) and GIST430 with a heterozygous KIT exon deletion and a heterozygous KIT exon 13 V654A substitution (15) were obtained from Dr. Fletcher (Brigham and Women's Hospital, Boston, MA). Cell lines were authenticated by single-nucleotide polymorphism (SNP) fingerprinting (Sequenom) as previously described (16), and KIT mutations were confirmed with RNA sequencing and deemed mycoplasma free (Mycoplasma PCR assay, IDEXX).

Cell lines used for internalization studies were suspended in SF-RPMI (serum-free RPMI, 0.02% BSA, 1% PSG). Antibodies were used at 1 μg/mL in SF-RPMI containing 75 μg/mL cycloheximide (Sigma-Aldrich) and mixed with cells for 45 minutes at 4°C. Cells were washed and samples resuspended in 100 μL of SF-RPMI with 75 μg/mL cycloheximide. One set of plates was incubated at 4°C, while the other at 37°C for 30 minutes, 2, or 4 hours. Cells were incubated with an APC-conjugated anti-human IgG for 45 minutes. For plates with SCF, the ligand was added at a final concentration of 10 ng/mL for 5 minutes at room temperature prior to adding antibodies and transferring plates to 4°C.

KU812 cells were incubated with 1 μg/mL PE-labeled LMJ729 and an antibody against a lysosomal marker, LAMP1 (APC-conjugated CD107, BD Biosciences, 560664) for 30 minutes at 4°C. Cells were washed to remove unbound antibodies. c-KIT protein levels, internalization, and colocalization with CD107 were monitored for up to 40 minutes at 37°C using the Amnis ImageStreamX instrument.

GIST-T1 or NCI-H526 cells were seeded in growth media (DMEM, 10% FBS or RPMI, 10% FBS, respectively) overnight, then treated with 100 mmol/L cycloheximide (Sigma) in methionine-free medium (GIBCO: DMEM, 21013-024; RPMI, A14517-01). Additionally, cells were either treated with 5 μg/mL ADC (LOP628), 10 ng/mL rh-SCF (R&D Systems, 255-SC), or both for 1, 4, or 6 hours at 37°C, at which time protein was isolated, denatured, and 5 μg loaded on a NuPAGE 4% to 12% Bis–Tris gel (Life Technologies). After protein transfer, membranes were blocked in TBST-5% milk and probed with an anti–c-KIT antibody (Cell Signaling Technology, 3074) or GAPDH antibody (Cell Signaling Technology, 3683). The secondary antibody (goat-anti rabbit-HRP 1:30,000; Santa Cruz) was used.

After seeding cells overnight in standard media, Maytansine (L-DM1-Me), LOP628 or a nontargeting ADC control (IgG-ADC) was added. In certain experiments, LMJ729 or IgG1 was included. The starting concentration of LOP628 and IgG-ADC was 68 nmol/L. A 1:5 dilution was performed for 9 concentration points. After 5 days, viability was assessed using CellTiter Glo. Luminescent counts from untreated cells (100% viability) were used to normalize treated samples. The concentrations of treatment resulting in 50% of cells remaining were calculated using nonlinear regression analysis (GraphPad Prism 6).

Human and cynomolgus monkey CD34⁺ bone marrow mononuclear cells were evaluated for c-KIT expression by flow cytometry using anti–c-KIT antibody 104D2 (BD Biosciences, 563856). Cells were suspended in FACS buffer (PBS, 2% FBS, 0.1% sodium azide) and incubated with 104D2 at 10 μg/mL for 60 minutes. Samples were assayed using FACS (BD FACSCanto II) with a microsphere bead set (Quantum Simply Cellular anti-Mouse IgG kit; Bangs Labs) to determine the specific antibody binding capacity (sABC) of each antibody.

c-KIT expression in formalin-fixed, paraffin-embedded (FFPE) tumor xenografts or human patient samples (Novartis tissue archive) was evaluated by immunostaining (Dako, A4502) including heat and standard exposure to Ventana Cell Conditioning #1 antigen retrieval reagent. The primary antibody was diluted to 14 μg/mL and incubated for 60 minutes at room temperature. Subsequently, incubation with Ventana UltraMap prediluted HRP-conjugated anti-rabbit antibody (760-4315) was performed for 16 minutes.

An H-score of c-KIT immunostaining was generated to reflect the expression and heterogeneity levels in tumors, using the formula: [(% of 1+ × 1) + (% of 2+ × 2) + (% of 3+ × 3)], where 1+ designates weak, 2+ designates moderate, 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.

Female SCID-beige mice (Harlan Laboratories) used for tumor xenograft studies were handled in accordance with Novartis Institutional Animal Care and Use Committee regulations and the ILAR Guide for the Care and Use of Laboratory Animals in an AAALAC accredited facility.

GIST-T1 cells were grown as described above. GIST430 cells were grown in Iscove's modified DMEM (Cellgro), 15% FBS (Omega Scientific Inc.), and 2 mmol/L l-glutamine (Cellgro). NCI-H1048 and NCI-H2170 cells were grown in RPMI-1640 medium (ATCC), 10% FBS, and 2 mmol/L l-glutamine. Cells were subcutaneously implanted in 50% Matrigel (BD Biosciences).

In the GIST T1 model, LOP628, its vehicle (10 mmol/L Tris–hydrochloric acid, 80 mmol/L sodium chloride, 3.5% sucrose, 0.01% Tween 20 pH7.5), or isotype control IgG-ADC were administered as a single intravenous (i.v.) injection to mice with tumors collected at specified time points after dosing (Fig. 3A and B) and processed to generate FFPE blocks.

The IHC protocol for assessing phospho-histone H3-positive nuclei included heat and standard exposure to Ventana Cell Conditioning #1 antigen retrieval reagent. The primary antibody (Cell Signaling Technology, 9701) was diluted 1:50 and incubated for 60 minutes. Subsequently, incubation with Jackson ImmunoResearch Laboratories goat anti-rabbit biotinylated secondary antibody (111-065-144) was performed. For quantification of positive cells, tumor section images were quantified using the Aperio Nuclear (version 9) image analysis algorithm on the Aperio ScanScope XT system (Leica Biosystems), which included ImageScope version 11.1.2.760 software. Three tumors per time point were assessed.

In vivo efficacy studies were initiated when tumors were ∼200 mm³. LOP628, IgG-ADC, or LMJ729 were administered i.v. Orally administered imatinib (Glivec) was formulated in 5% dextrose in water. Tumors were measured twice weekly, and volumes were calculated as (length × width²)/2. Data are expressed as the percent tumor growth inhibition (TGI) in 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-groups comparisons of final measurements were performed using ANOVA and a post hoc test (Sigma Plot, Systat Software Inc.)

The toxicity of LOP628 was evaluated in non-GLP and GLP studies in cynomolgus monkeys (Charles River Laboratories) with intravenous administration once every 3 weeks. In the dose range finding study, doses were 3, 10, and 30 mg/kg for a 6-week duration and 6-week recovery period. This study included a nontargeting IgG-ADC at 30 mg/kg. LOP628 doses for the GLP study were 3, 8, and 20 mg/kg for approximately a 3-month duration. In addition to assessment of several standard safety parameters, cardiovascular telemetry was incorporated using jacketed telemetry devices.

A single-dose nonterminal cynomolgus monkey study was conducted to investigate previously observed infusion reactions as well as the potential for mast cell degranulation. LOP628 was administered to 3 monkeys intravenously at 0.3, 1, 8, and 20 mg/kg. Animals were monitored for serum tryptase, cytokines, and complement analysis.

c-KIT is overexpressed in numerous tumor types, such as melanoma, SCLC, AML, and GIST (Supplementary Fig. S1). While considered essential for many processes, such as the function of the interstitial cells of Cajal, in hematopoiesis and melanogenesis (17–20), no evidence exists that suggests tumors overexpressing wild-type c-KIT are dependent on c-KIT/SCF signaling for survival. Tumors harboring c-KIT–activating mutations signal independently of SCF. Due to this, we investigated the activity of both ligand-blocking (antagonist with strong signaling inhibition) and non–ligand-blocking anti–c-KIT ADCs (weak signaling inhibition, possibly via steric hindrance; Supplementary Fig. S2). Antibodies and ADCs in this publication and their properties are listed in Table 1. The ADCs generated were conjugated via ImmunoGen, Inc. Technology to –SMCC-DM1, the noncleavable linker and maytansine-derived toxin.

To understand differences between the activities of ligand-blocking versus non-ligand-blocking anti–c-KIT antibodies, internalization and degradation of c-KIT upon treatment with the different antibody classes were evaluated. Internalization (measured as the loss of surface c-KIT by flow cytometry) upon binding of c-KIT antibodies was interrogated in lines expressing mutant c-KIT (GIST-T1, exon 11 mutation) or wild-type c-KIT (M07e). In GIST-T1, there was no significant difference in internalization kinetics between LMJ451 (ligand-blocking antibody) and DYL884 or LMJ729 (non–ligand-blocking antibodies) within a 4-hour time course (Fig. 1A). However, the internalization rate mediated by LMJ451 appeared diminished relative to DYL884 in the M07e cell line, and this difference was enhanced in the presence of SCF within a 2-hour time course (Fig. 1A). Coincubation of SCF with DYL884 accelerated internalization, confirming DYL884 did not block SCF binding to c-KIT. In contrast, the addition of SCF did not impact internalization upon binding of LMJ451, due to the ability of LMJ451 to inhibit binding of SCF to c-KIT. Once internalized, the c-KIT/antibody complex rapidly trafficked to the lysosome (Fig. 1B), as seen in the costaining of the complex with the lysosomal marker CD107/LAMP1 in the KU812 line.

c-KIT is efficiently degraded both in the wild-type and mutant forms, as shown in Fig. 1C. Inhibition of new protein synthesis with cycloheximide treatment allows for the monitoring of degradation of the steady-state c-KIT pool. Treatment with the isotype ADC was ineffective, suggesting the impact on c-KIT degradation is dependent on c-KIT ADC binding. The degradation of mutant c-KIT, detected in the GIST-T1 line, appears unaffected by SCF treatment or by the addition of a ligand-blocking ADC (GZQ167). In contrast, incubation with the non–ligand-blocking ADC (LOP628) accelerated the degradation of c-KIT and cotreatment of SCF with LOP628 further enhanced the degradation rate. A similar finding was observed in the wild-type c-KIT line, NCI-H526, as the addition of LOP628, SCF, or LOP628 with SCF augmented the c-KIT degradation rate, with the combination being the most efficient. The addition of GZQ167 exhibited no impact on its own, but when combined with SCF, the presence of the ligand-blocking antibody ameliorated the SCF effects. These results suggest that treatment with a non–ligand-blocking ADC may enhance degradation of the c-KIT/ADC complex similar to that of SCF treatment, resulting in efficient trafficking to the lysosome and payload release, allowing for multiple rounds of ADC internalization. In contrast, a ligand-blocking ADC would utilize the intrinsic turnover rate of c-KIT and decrease the degradation rate mediated by the addition of SCF to wild-type c-KIT. This could cause less efficient release of the active catabolite and decreased activity.

We next profiled the activity of the ligand-blocking and non–ligand-blocking ADCs in cell lines representing c-KIT expressing indications. Several ADCs from each class were tested in GIST (GIST882, GIST430, and GIST-T1), AML (Kasumi-1 and Kasumi-6), and SCLC (NCI-H526 and NCI-H1048) lines. A c-KIT–negative line, MDA-MB-453, was included as a negative control. The c-KIT ADCs from both classes demonstrated similar antiproliferative activity with EC₅₀ values in the subnanomolar range, which represented a 3- to >1,000-fold difference in activity over the negative cell line. Similar activity was not surprising, as these cell lines are not dependent on SCF for growth (Supplementary Table S1).

To extend our comparison of ligand-blocking and non–ligand-blocking c-KIT ADCs in vivo, an efficacy study (single 0.625 mg/kg dose) in the GIST-T1 xenograft model in mice was performed. The non–ligand-blocking LOP628 trended toward increased efficacy compared with the ligand-blocking candidates and was the only treatment statistically different from the isotype control IgG-ADC (Fig. 2A). The observation of efficacy at such a low dose was consistent with the in vitro potency and may relate to c-KIT's rapid internalization and lysosomal processing.

Given the role of c-KIT in hematopoiesis (3), there was concern that a ligand-blocking antibody may induce severe bone marrow toxicity based on published data describing hematopoietic failure with conditional deletion of c-KIT in adult mice (21). To assess the tolerability, we evaluated two mouse cross-reactive ADCs, LMJ451 (ligand-blocker) and LPG167 (non–ligand-blocker). After a single 10 mg/kg i.v. administration, LPG167 was well tolerated, while the mice treated with LMJ451 showed severe weight loss ∼13 days after treatment (Fig. 2B). Bone marrow histology revealed severe hypocellularity due to loss of hematopoietic cells of all lineages (Fig. 2C) as compared with that of an untreated mouse (Fig. 2D). These findings, in addition to in vivo screening of non–ligand-blocking ADCs and the in vitro data (Fig. 1), led to the selection of the non–ligand-blocking antibody LMJ729, which when conjugated to SMCC-DM1 is designated LOP628.

LOP628 demonstrates similar binding affinities to human and cynomolgus c-KIT extracellular domains, at 6.9 and 5.1 nmol/L, respectively. Examples of in vitro studies performed to address the activity of LOP628 are shown in Fig. 3A. Cell lines devoid of c-KIT (MDA-MB-468), expressing wild-type c-KIT (NCI-H526) or a single mutation in c-KIT (GIST-T1) were tested for inhibition of proliferation mediated by LMJ729 or LOP628. The robust antiproliferative activity of LOP628 appears to be DM1 driven as LMJ729 exhibited no activity on NCI-H526 or GIST-T1 lines. LOP628 activity appears target dependent as there is no specific activity on the c-KIT–negative line. To determine if there might be a correlation between c-KIT surface expression and sensitivity to an ADC, a cell line panel was treated with the conjugated mouse hybridoma antibody, 9P3-SMCC-DM1, from which humanized LMJ729 was derived. When cell lines were sensitive to the maytansinoid payload (IC₅₀ < 1 nmol/L), a minimum of 20,000 receptors appeared to enrich for sensitivity to a c-KIT ADC in vitro, yielding low single-digit nanomolar IC₅₀ values (Supplementary Fig. S3).

A study was conducted to assess the degree and duration of LOP628 treatment to induce mitotic arrest by measuring the accumulation of cells positive for the mitotic marker phosphorylated Histone H3 (pHH3) conferred by the maytansinoid toxin in the mutant c-KIT–expressing GIST T1 model in mice. LOP628 yielded a time-dependent increase in the percentage of pHH3-positive cells, with a peak of 30% at 48 hours after dose relative to the isotype IgG-ADC or untreated controls and decreasing thereafter (Fig. 3B and C). These data suggest that the high initial exposure over the first several days drives a significant portion of the mitotic arrest and demonstrate that LOP628 is capable of eliciting a robust PD response in tumor xenografts, consistent with the mechanism of the toxin.

To elucidate if the ability of LOP628 to mediate mitotic arrest translates into in vivo efficacy, studies were conducted using GIST (high mutant c-KIT levels) and SCLC (wild-type moderate c-KIT level) models (Supplementary Fig. S4).

LOP628 efficacy in the GIST-T1 (exon 11 mutation; imatinib-sensitive) and GIST430 (exon 11, 13 mutations; imatinib-insensitive) xenograft models is shown in Fig. 3D and E. In the GIST-T1 study, a single 2.5 mg/kg LOP628 dose induced tumor stasis ∼60 days, with subsequent regrowth. Imatinib alone regressed tumors, but tumors grew back upon treatment cessation. In contrast, the single 2.5 mg/kg LOP628 administration combined with imatinib (80 mg/kg b.i.d. for 50 days) resulted in complete and durable regressions with no tumor regrowth 130 days after imatinib treatment ended.

To achieve a similar efficacy level as in GIST-T1, 10 mg/kg LOP628 was required in the GIST430 model, demonstrating the differential sensitivities among models. There was 41% and 81% TGI (day 28) as compared with the vehicle control for the 5 and 10 mg/kg LOP628 doses, respectively. The efficacy of 10 mg/kg LOP628 was superior to imatinib administered at its highest tolerated dose, which induced 53% TGI. Imatinib was reduced from 100 to 80 mg/kg due to weight loss at the higher dose level (Fig. 3E). LOP628 activity in the NCI-H1048 SCLC xenograft model (Fig. 3F), with lower and more heterogeneous c-KIT levels (Supplementary Fig, S4), required 5 mg/kg to induce tumor stasis, while 15% tumor regression was observed at 10 mg/kg (day 27). LOP628 was ineffective in the NCI-H2170 model expressing very low c-KIT levels (Supplementary Fig. S5).

H-score analyses of c-KIT expression by IHC in tumor xenograft models and tissue biopsies from SCLC, melanoma, and GIST patients demonstrate that xenograft models sensitive to LOP628 express c-KIT at levels observed in cancer patients (Fig. 4). The efficacy data suggest that focusing on cancers with an H-score of 160 or greater, corresponding to ≥50% of cells exhibiting 2+ to 3+ staining intensity, may enrich for patients more likely to benefit from LOP628 treatment (Fig. 4B); however, due to the limited number of c-KIT–positive models in these indications, the data set is small for making efficacy predictions.

c-KIT is expressed on mast cells and a subset of bone marrow hematopoietic stem cells (22). Evaluation of c-KIT levels on bone marrow mononuclear cells (IHC and FACS) and mast cells (IHC) from human and cynomolgus monkey confirmed similar expression levels (Supplementary Table S2) and the use of cynomolgus monkey as a relevant species for a safety assessment of LOP628.

To understand the contribution of c-KIT–mediated toxicity to the safety profile of LOP628, a dose range finding study in cynomolgus monkeys was conducted, where both LOP628 and an isotype control IgG ADC were evaluated at a matched dose level of 30 mg/kg. Animals were dosed every 3 weeks and were either necropsied 1 week after last dose administration or after a 6-week recovery period. Overall, the toxicity profiles were similar across both targeting and nontargeting ADCs. There were limited clinical signs with the exception of a single animal dosed at 30 mg/kg LOP628. After the first administration, this animal exhibited lethargy and hypothermia among other effects which resolved within 48 hours. The most prominent changes at all dose levels were noted in the bone marrow and consisted of transient, dose-dependent decreases in red cell mass (hemoglobin, hematocrit, red blood cells) with compensatory reticulocyte response and decreased neutrophil counts (Fig. 5B, panels B–E). Hematology changes were of greater magnitude with LOP628 as compared with the isotype control, most likely due to the contribution of target-mediated toxicity in the bone marrow. The nadir for neutropenia generally occurred 3 weeks after dose administration (days 35 and 56), which was immediately before the next dose administration followed by recovery during the week after the next dose administration (Fig. 5B, panel C). Increased liver enzymes were noted with both ADCs (Fig. 5B, panel A) to a similar severity and most commonly noted preclinically as an off-target effect related to the payload, DM1. Similarly, microscopic changes 1-week after last dose administration in the bone marrow consisted of a mild decrease in bone marrow cellularity at 30 mg/kg LOP628 compared with vehicle (Fig. 5A). A dose level of 30 mg/kg LOP628 exceeded the maximum tolerated dose due to progressively declining hemoglobin concentrations.

In a subsequent phase I-enabling Good Laboratory Practices (GLP) cynomolgus monkey study, LOP628 was dosed at 3, 8, and 20 mg/kg. LOP628-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. Additional clinical monitoring implemented in the GLP study indicated that infusion reactions were noted in animals within 5 minutes of dosing and were prevalent at 8 and 20 mg/kg LOP628. These clinical observations included decreased blood pressure, increased heart rate, excessive scratching and facial reddening and resolved within 20 minutes after dose administration. Upon repeat dosing on day 22, clinical responses were limited to emesis, hypoactivity, and cardiovascular changes and were noted most prominently at 3 and 8 mg/kg LOP628, but resolved within 2 hours after dose administration. These observations were consistent with the onset of anti-drug antibody (ADA) formation in all study animals starting on day 21. Cardiovascular effects related to administration of LOP628 on day 1 included transient decreases in blood pressure with compensatory increases in heart rate in all animals. Subsequently, following repeated LOP628 administration, cardiovascular effects were most notable at 3 and 8 mg/kg dose levels. Onset of cardiovascular changes correlated with changes in clinical signs. Similar to the dose range finding monkey study, the most apparent clinical pathology changes were limited to bone marrow changes in a dose-responsive manner. At the highest non-severely toxic dose (HNSTD) of 8 mg/kg LOP628, the changes in parameters measuring bone marrow function were transient and minimal. Dose levels of 3 and 8 mg/kg were considered well tolerated in monkeys based on the intended end-stage cancer patient population and resulted in minimal and transient changes that were consistent with payload effects of DM1 and the pharmacodynamic effects of LOP628.

Coincident with ADA formation, systemic exposure to LOP628 was impacted with subsequent fast clearance of the ADC. However, the pharmacodynamic effect of neutropenia remained after each subsequent dose administration regardless of ADA formation (data not shown).

To understand the infusion reaction mechanism observed after LOP628 dosing in the GLP monkey study (cardiovascular changes, scratching, emesis, and hypoactivity), a subsequent investigative single dose monkey study was conducted. Assessment of serum histamine and tryptase were included at several collection time points (up to 24 hours post dose) to determine if infusion reactions were related to mast cell degranulation. Based on these analyses, LOP628 did not cause elevations in serum tryptase and histamine above baseline values (Fig. 5C). Therefore, it was concluded that infusion reactions observed in monkeys were likely not due to mast cell degranulation.

For patients with advanced cancers expressing either wild-type or mutant c-KIT, treatment options are limited, which is notably the case in recurrent c-KIT⁺ GIST after the approved c-KIT/RTK-targeting agents such as imatinib, sunitinib, and regorafenib have failed, often because of the emergence of secondary mutations (23, 24). This is primarily due to the heterogeneity of c-KIT resistance mutations such that even the newer targeting agents with the potential to inhibit a broader range of mutations, like the activation loop inhibitor BLU-285 or switch control inhibitor DCC-2618, may still not block the full spectrum of driver mutations (25, 26). An alternative approach was to implement HSP90 inhibitors to reduce wild-type and c-KIT–mutant protein levels in GIST, but this failed to demonstrate significant clinical benefit (27).

Thus, we explored a novel angle to target c-KIT–expressing cancers, whereby an ADC binding to this receptor could deliver a potent toxin directly to the cancer, regardless of c-KIT mutational status provided the extracellular domain was unperturbed. This would broaden activity to those cancers overexpressing wild-type c-KIT, where the target may not be a driver, such as in SCLC, because the ADC activity is driven primarily by the DM1 cytotoxin. In selecting c-KIT as a target, we purposefully took advantage of its rapid internalization kinetics and route of lysosomal-mediated degradation to effectively deliver the ADC for processing to its active catabolite, Lys-SMCC-DM1 (data not shown).

However a concern was that a ligand-blocking antibody with strong antagonism used for the ADC would be highly hematotoxic (as observed with LMJ451 in mice) and may confer little to no efficacy benefit in cancers expressing c-KIT mutants or in those expressing wild-type c-KIT where it may not be a driver. Thus we focused on developing a c-KIT ADC with no or minimal ligand-blocking activity.

Therapeutics utilizing the ADC modality such as Kadcyla (Trastuzumab-DM1) and Adcetris have produced impressive clinical responses while minimizing systemic toxicity. LOP628, a highly selective and potent ADC, was developed with the aim of targeting cancers overexpressing c-KIT. Our data provides evidence that LOP628 selectively binds c-KIT expressing cell lines, becomes rapidly internalized and is efficiently processed to release sufficient active catabolite to potently induce a cytotoxic response in cell viability assays. In vitro profiling of LOP628 activity in a cell line panel revealed effective killing of c-KIT-positive tumor cells representing GIST, AML and SCLC. In vivo, LOP628 induced mitotic arrest and was efficacious in multiple xenograft models expressing wild-type c-KIT (SCLC: NCI-H1048, NCI-H526) and mutant c-KIT (GIST-T1, GIST430), with varying sensitivities that did not correlate with mutational status of the target, because the two most sensitive models tested were the GIST-T1 and NCI-H526, which regressed tumors at 2.5 mg/kg (NCI-H526 graph not shown), considered a low dose for an ADC conjugated to –SMCC-DM1. Some other models required a higher 10 mg/kg dose to induce significant efficacy, in line with preclinical observations with Kadcyla (28). As a whole, robust antitumor responses were observed in tumor xenograft models representing c-KIT expression in greater than 50% of the tumor at a 2+/3+ staining intensity. LOP628 was not effective against the NCI-H2170 NSCLC xenograft model (Supplementary Figs. S4 and S5), which expresses little to no c-KIT, but exhibited sensitivity to the maytansine payload, thereby supporting the ADC specificity. The possibility to combine LOP628 with SOC or targeted therapeutics was promising, as demonstrated by a single dose of LOP628 with 50 days of imatinib resulting in GIST-T1 tumor regression for the duration of the study (180 days). The data presented here showcase the cytotoxicity of LOP628 on c-KIT overexpressing tumors (regardless of mutational status).

Being that LOP628 is not rodent cross-reactive, safety was assessed in the cynomolgus monkey, where it was well tolerated up to 8 mg/kg with primary on-target effects noted in the bone marrow due to c-KIT expression on hematopoietic cells and off-target (payload) effects in the liver. The therapeutic index was predicted to be narrow based on those less sensitive preclinical models requiring 10 mg/kg for efficacy. Bone marrow effects have been noted with other DM1 ADCs both preclinically and clinically (10); however, the kinetics and severity of neutropenia in the monkey as compared with previous studies with DM1 ADCs supports contribution of both targeted c-KIT and off-target DM1-mediated bone marrow toxicity. The bone marrow effects were transient, and 3 and 8 mg/kg were considered to be well tolerated in monkeys. At the HNSTD in monkeys (8 mg/kg), which is commonly used to establish first-in-human (FIH) dose estimation for endstage cancer patients, there was a 6-fold safety margin at the clinical starting dose of 0.3 mg/kg LOP628. Clinically, infusion reactions were noted in monkeys after each dose administration. Due to the presence of c-KIT on mast cells, biomarkers associated with mast cell degranulation (histamine and tryptase) were assessed as part of the investigative monkey study and were not elevated above baseline levels. Therefore, clinical reactions noted in monkeys were likely not related to mast cell degranulation. The information from the GLP monkey study (HNSTD) was used to support the first-in-human (FIH) dose estimation for the LOP628 phase I clinical trial.

The monkey studies did not predict clinical acute hypersensitivity reactions noted at the clinical starting doses of 0.15 and 0.3 mg/kg, presumably due to mast cell degranulation. This led to cessation of the phase I clinical trial. The clinical findings and mechanism behind this hypersensitivity reaction are explored elsewhere (29). These data highlight the potential benefit of targeting c-KIT via an ADC approach, but also the challenges of translating preclinical safety findings to the clinic.

N. Pryer, W. Kluwe, and W.R. Sellers hold ownership interest (including patents) in Novartis Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T. Abrams, A. Connor, C. Fanton, S.B. Cohen, T. Huber, K. Miller, K. Krauser, F. Galimi, K. Mansfield, N. Pryer, S.A. Ettenberg, W.R. Sellers, E. Lees, P. Kwon, S.C. Schleyer

Development of methodology: T. Abrams, A. Connor, C. Fanton, T. Huber, K. Miller, X. Niu, S. Harris, M. Ghoddusi, K. Mansfield

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Abrams, C. Fanton, E.E. Hong, S. Harris, K. Krauser, Z. Wang, M. Ghoddusi, W. Kluwe, S.C. Schleyer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Abrams, A. Connor, C. Fanton, K. Miller, X. Niu, S. Harris, D. Walker, K. Krauser, F. Galimi, Z. Wang, M. Ghoddusi, K. Mansfield, S.T. Lee-Hoeflich, J. Holash, W. Kluwe, S.A. Ettenberg, E. Lees, P. Kwon, S.C. Schleyer

Writing, review, and/or revision of the manuscript: T. Abrams, A. Connor, S. Harris, D. Walker, K. Mansfield, S.T. Lee-Hoeflich, W. Kluwe, W.R. Sellers, E. Lees, S.C. Schleyer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Abrams, X. Niu, J. Kline, M. Ison-Dugenny, D. Walker, Z. Wang

Study supervision: T. Abrams, A. Connor, K. Krauser, N. Pryer, W.R. Sellers, J.A. Abraham, S.C. Schleyer

The authors thank Dr. T. Tagashi for providing the GIST-T1 cell line and Dr. Jonathan Fletcher for the GIST882 and GIST430 cell lines.

All authors worked at Novartis Pharmaceuticals while this work was conducted.

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.