Article Cite This: Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Metabolism of an Oxime-Linked Antibody Drug Conjugate, AGS62P1, and Characterization of Its Identified Metabolite Josh T. Snyder, Maria-Christina Malinao,† Julien Dugal-Tessier,Δ John E. Atkinson, Banmeet S. Anand, Akihiro Okada,¶ and Brian A. Mendelsohn*,# Agensys Inc. an Affiliate of Astellas Pharma Inc., 1800 Stewart Street, Santa Monica, California 90404, United States S Supporting Information *
ABSTRACT: AGS62P1 is an antibody drug conjugate (ADC) composed of a human IgG1κ monoclonal antibody against FLT3 (FMS-like tyrosine kinase 3) with a p-acetyl phenylalanine (pAF) residue inserted at position 124 of each heavy chain linked to the proprietary microtubule disrupting agent AGL-0182-30 via an alkoxyamine linker that forms an oxime upon conjugation to the antibody. AGS62P1 is currently in Phase I human clinical trials for acute myelogenous leukemia (AML). The identified primary metabolite of an oxime-linked ADC is presented for the first time. AGS62P1 metabolism was assessed in xenograft tumorbearing mice and rats treated with the ADC using liquid chromatography and mass spectrometry-based methods described herein. In this study, we identified the metabolite of AGS62P1 as pAF-AGL-0185-30, which contains a fragment resulting from the catabolism of the antibody component of the ADC and hydrolysis of the terminal amide portion of the linker-payload. We demonstrated that the metabolite of AGS62P1 is tolerated in rats above 1.5 mg/kg and above 0.334 mg/kg in cynomolgus monkeys when given as a single dose. Furthermore, we established in vitro that pAF-AGL-0185-30 does not significantly inhibit hERG or cytochrome P450 family enzymes (CYPs). KEYWORDS: antibody drug conjugate, auristatin, non-natural amino acid, metabolism, pharmacokinetics
1. INTRODUCTION Antibody drug conjugates have become a major class of anticancer therapeutics in recent years.1 FLT3 (FMS like tyrosine kinase-3) is a class III receptor tyrosine kinase (RTK) aberrantly expressed in precursor B-cell acute lymphogenous leukemia (ALL), nearly all acute myelogenous leukemia (AML) (over 90%) and a fraction of T-cell ALL.2 AGS62P1 is an ADC consisting of a human monoclonal IgG1 antibody conjugated to a potent cytotoxic tubulin inhibitor drug-linker payload, AGL0182-30. The payload is conjugated in a site-specific manner through oxime ligation to engineered non-natural amino acids, para-acetyl phenylalanine (pAF), expressed at position A121 by Kabat numbering3 on the heavy chains of the antibody resulting in a drug-to-antibody ratio (DAR) of approximately 2.4 To this point, most ADCs in the clinic have been made using heterogeneous conjugation methodologies in which drugs are attached to native antibody cysteine or lysine residues. In the past few years, a number of groups have developed methodologies to attach drugs to antibodies in a site-specific manner. Some of the methods include enzymatic ligation of payloads to tags engineered into the antibody, conjugation of drugs to engineered or chemically modified glycans, and molecular engineering to introduce cysteine or non-natural amino acids (such as pAF, pAMP, or azido-lysine) into the antibody backbone.5 These efforts have two main aims: to create a more homogeneous ADC and to enhance the stability of the linkage © XXXX American Chemical Society
of the drug to the antibody. There is evidence to suggest that both are means to improve the therapeutic index of ADCs.6 Site specific, oxime-linked ADCs show superior in vitro serum stability and preclinical toxicology profile in rats when compared with conventional stochastic cysteine-maleimide conjugated ADCs.5,7 They have also been shown to maintain their in vitro potency and in vivo efficacy against target expressing human tumor cell lines.8 When incubated in human serum in vitro, an anti-Her2 cysteine-maleimide ADC lost drug at a faster rate than two anti-Her2 oxime-linked ADCs.9 Pharmacokinetic analysis in rats corroborated this finding: animals administered anti-Her2 oxime-linked ADCs had greater total antibody and ADC exposure than the cysteine-maleimide control, indicating longer antibody half-life and a more stable drug-antibody linkage for the oxime ADC.9 Furthermore, the oxime-linked anti-Her2 ADCs were better tolerated in rats and had greater antitumor efficacy in xenograft-bearing mice.9 The drug-linker component of AGS62P1 is AGL-0182-30, which contains the microtubule disrupting agent AGD-0182 and a stable noncleavable linker (Figure 1A).10 AGD-0182 is a synthetic analogue of the naturally occurring tubulin-binding Received: March 2, 2018 Revised: April 18, 2018 Accepted: May 1, 2018
A
DOI: 10.1021/acs.molpharmaceut.8b00225 Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Article
Molecular Pharmaceutics
Figure 1. Structures of Payload, Drug Linker, Antibody and ADC. (A) Payload (AGD-0182) and Drug Linker (AGL-0182-30). (B) Antibody (AGS62P) with pAF residues shown. (C) AGS62P1, consisting of AGL-0182-30 conjugated to AGS62P pAF residues via an oxime.
Figure 2. Structure of AGL-0182-30 before and after conjugation. Targeted searches for metabolites resulting from cleavage at the sites indicated in the figure were conducted.
2.2. Conjugation of Antibodies. The AGS62P antibody in 50 mM sodium citrate containing 500 mM sodium chloride at pH 4.0 was conjugated by adding 10 mol equiv of the drug linker AGL-0182-30 dissolved in DMSO and an aqueous solution of acetic hydrazide (50 mM final). The conjugation reaction was allowed to proceed at 28 °C for 22 h with gentle mixing. Purified by tangential flow filtration, DAR determined by PLRP, purity by SEC. The final concentration of antibody during conjugation was 13.6 mg/mL, and the final DMSO concentration was 4.7% (v/v). 2.3. Identification of Metabolites in Tumors. CB17/ SCID mice were implanted subcutaneously with a suspension of human acute myeloid leukemia MOLM-13 cells (vendor: DSMZ). When subcutaneous tumors formed, ranging in size from 200−900 mm3, mice were administered a single intraveneous dose of ADC at 6 mg/kg of body weight or a vehicle control (20 mM histidine/5% trehalose pH 6.0). Tumors from mice were harvested 24 h postdose and snap frozen at −80 °C.
molecule Dolastatin 10. It has been modified to contain a monomethyl valine unit at the N-terminus, an alkyl azide moiety on the adjacent amino acid and a C-terminal phenyl carboxamide residue in comparison with Dolastatin 10.11 The drug linker is conjugated to the antibody at the para-acetyl phenylalanine unit via an oxime to the alkoxyamine functionality (Figure 1B,C). The expected mechanism for release of the AGL-0182-30 drug-linker from the AGS62P1 ADC is via proteolysis of the antibody in endosomal compartments, similar to the metabolism of mcMMAF ADCs, which yield cysteine-mcMMAF as the primary metabolite.12 Additional plausible cleavage sites of pAF-AGL-0182-30 adduct include the C terminal carboxamide hydrolysis to form the carboxylic acid, N−O cleavage of the alkoxyamine, and oxime hydrolysis (Figure 2).
2. EXPERIMENTAL SECTION 2.1. Generation of AGS62P. The AGS62P antibody was expressed as previously described and purified on a Protein A column.4 B
DOI: 10.1021/acs.molpharmaceut.8b00225 Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Article
Molecular Pharmaceutics Tumors were removed from −80 °C storage, isopropanol was added (3 mL of isopropanol per gram of tumor), and the tumors were allowed to thaw. Stainless steel beads were then added and the tumors were homogenized by mechanical disruption using a Bullet Blender (Next Advance, Inc.). The supernatant from each homogenized tumor was transferred to another tube and centrifuged to pellet the precipitated debris. A portion of the supernatant was diluted 50-fold with 20 mM Tris·HCl pH 7.4 (final isopropanol concentration 1.5% v/v). The Tris·HCl/isopropanol extract was loaded onto a column packed with an antidrug antibody immobilized on a Sepharose resin. After extensive washing to remove nonspecifically bound components, small molecule metabolites were eluted from the resin with 75 mM glycine·HCl pH 2.4 containing 15% DMSO. The eluted fraction was dried under vacuum and resuspended in 90/10 water/DMSO. Affinity-purified and crude isopropanol extracts were analyzed by high-resolution mass spectrometry using an Orbitrap Velos (Thermo Scientific) instrument to identify the metabolites. 2.4. Identification of Metabolites in Serum and Liver. Liver tissue and serum samples were obtained from rats dosed with AGS62P1 at 60 mg/kg weekly for 4 doses. Metabolites were extracted from liver tissue (harvested upon necropsy, 28 days after the initial dose) by adding 3 volumes of isopropanol per gram of liver tissue and homogenizing with stainless steel beads in a Bullet Blender. Homogenized samples were centrifuged at 16 000g to pellet precipitate and debris. The resulting supernatant was transferred to a fresh tube and dried under vacuum and resuspended in water/DMSO (90/10 v/v) for analysis. Metabolites were extracted from serum samples (harvested 24 h after the first dose) by adding 4 volumes of acetonitrile and vortex mixing. Precipitated protein and debris were precipitated by centrifugation at 16 000g. The resulting supernatant was dried under vacuum and resuspended in water/DMSO (90/10 v/v) for analysis. Metabolites were analyzed by LC/MS using an OrbiTrap Velos. Liquid chromatography and mass spectrometry was performed using an Orbitrap Velos (Thermo) and an Ultimate 3000 (Thermo). The column used was a 2.1 × 50 mm BEH C18 (Waters). The chromatographic separation was performed using a four-step gradient. Solvent A was a blend of 0.1% formic acid in water, while solvent B was 0.1% formic acid in acetonitrile. LC method: isocratic 0.00−1.5 min 5% B, gradient 1.5−5.0 min 5%−60% B, gradient 5.0−10.0 min 60%−100% B, isocratic 10.0−12.0 min at 100% B. The mass spectrometer was programmed to collect full scans over a 400 to 1800 m/z mass range at 100 K resolution throughout the entire chromatographic separation. Extracted ion chromatograms were collected using the Xcalibur software (Thermo). Areas under the curve were collected for the metabolite, and were used for relative quantitation within samples. The LOD for the pAF-AGL-0185-30 was 8 nM, and it was found to be 10× above LOD. 2.5. Identification of Metabolites in Cultured Human Hepatocytes. Cryopreserved human hepatocytes were purchased (XenoTech) and stored in the vapor phase of a liquid nitrogen tank. Cells were thawed and plated on Collagen I coated 12-well plates (Corning) using the hepatocyte supplier’s recommended culture media (supplemented DMEM) and grown to confluency. After a 2-day adaptation period, cells were treated with AGD-0182 (20 μM final concentration) or DMSO vehicle control (0.2% v/v final concentration). Following a 3-h treatment period cells and
culture supernatants were harvested and extracted for metabolite identification, as described in the following section. Culture supernatants were collected from cell culture plates and centrifuged at 12 000g. The upper portion of each resulting supernatant was transferred to a fresh tube and extracted with isopropanol (70% v/v) by vortex mixing. The resulting extract was centrifuged at 15 000g to pellet precipitate, dried under vacuum, and resuspended in water/DMSO (90/10 v/v) for analysis of metabolites by LC/MS using an OrbiTrap Velos. Cell monolayers were harvested by adding isopropanol (70% v/v in water) with disruption by pipetting. Extracts were prepared by vortex mixing, followed by centrifugation at 12 000g. The resulting supernatant was dried under vacuum, and resuspended in water/DMSO (90/10 v/v) for analysis of metabolites by LC/MS. 2.6. Cytochrome P450 Inhibition Assays. To evaluate pAF-AGL-0185-30 as a direct, time-dependent, and metabolism-dependent inhibitor of CYP activity, pooled human liver microsomes were incubated with marker substrates in the presence or absence of pAF-AGL-0185-30. To distinguish between metabolism-dependent and time-dependent inhibition, pAF-AGL-0185-30 was preincubated with human liver microsomes for 30 min with and without an NADPH-generating system, respectively, prior to the incubation with the marker substrate. Known direct and metabolism-dependent inhibitors of CYP enzymes were included as positive controls in all experiments. 2.7. Cytochrome P450 Induction Assays. Three cryopreserved preparations of cultured human hepatocytes from three separate livers were treated once daily for three consecutive days with dimethyl sulfoxide (DMSO, 0.1% v/v, vehicle control), flumazenil (25 μM, negative control), one of seven concentrations of pAF-AGL-0185-30 (0.1, 1, 3, 10, 30, 100, or 300 nM) or one of three known human CYP inducers, namely, omeprazole (50 μM), phenobarbital (750 μM), and rifampin (20 μM). After treatment, hepatocytes from the same treatment groups were harvested with Buffer RLT to isolate RNA, which was analyzed by qRT-PCR to assess the effect of pAF-AGL-0185-30 on CYP1A2, CYP2B6, and CYP3A4 mRNA levels. 2.8. P-gp Assays. pAF-AGL-0185-30 was evaluated for its ability to inhibit P-gp and as a substrate of P-gp using the positive control substrate, digoxin, which produces a signal sufficient for the detection of inhibition of the transporter.16,17 The bidirectional permeability of probe substrate (digoxin) across Caco-2 cells was measured to determine if pAF-AGL0185-30 is an inhibitor of P-gp. The bidirectional permeability of pAF-AGL-0185-30 (10 μM) across MDCKII-MDR1 cells was measured to determine if pAF-AGL-0185-30 is a substrate of P-gp. 2.9. hERG Inhibition Assays. The in vitro effects of pAFAGL-0185-30 were evaluated on the hERG (human ether-à-gogo-related gene) channel current (a surrogate for IKr, the rapidly activating delayed rectifier cardiac potassium current) at nearphysiological temperature. The concentration−response relationship of the effect of L-pAF-AGL-0185-30 on the hERG potassium channel current was evaluated in stably transfected human embryonic kidney cells (HEK293) that express the hERG gene. The effect of L-pAF-AGL-0185-30 on hERG potassium current amplitude was monitored using whole cell patch clamp technique. In brief, cells stably expressing hERG were held at −80 mV. Onset and steady-state inhibition of hERG potassium current due to L-pAF-AGL-0185-30 were C
DOI: 10.1021/acs.molpharmaceut.8b00225 Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Article
Molecular Pharmaceutics
Figure 3. Identification of small-molecule drug metabolites of AGS62P1, which gave AGL-0185-30 (top structure) and the payload AGD-0182 which gave AGD-0185 (bottom structure). High-resolution mass spectrum of the isotopic distribution of the metabolite, confirming the charge state. Observed masses for the parent ion and observed 13C isotopic variants were within 5 ppm of theoretical values. (A) Affinity-purified extracts of subcutaneously established MOLM-13 Xenograft Tumors in CB17/SCID Mice (24 h Post-Dose with AGS62P1). (B) Detection of pAF-AGL-018530 in crude extract of rat serum (24 h Post-Dose with AGS62P1). (C) Detection of pAF-AGL-0185-30 in rat liver extract (28 days after first dose with AGS62P1). (D) Detection of AGD-0185 in cultured human hepatocytes following incubation with AGD-0182.
measured using a pulse pattern with fixed amplitudes repeated at 5 s intervals. Each recording was ended with a final application of a supramaximal concentration of the reference substance (E-4031, 500 nM), to assess the contribution of endogenous currents. A range of concentrations up to 3uM LpAF-AGL-0185-30 were applied to at least three cells, and the effect on hERG potassium current amplitude was monitored. 2.10. In Vitro Cytotoxicity Assays. Karpas299 and Molm13 cell lines were seeded in 96-well plates at a density of 1500 cells per well. One day after seeding, the drugs pAF-AGL-018530 or AGD-0182 were added (0.0001 to 10 000 nM) and cultured for an additional 5 days. At the end of the culture period, cell viability was assessed by adding Presto Blue (Invitrogen) and measuring the fluorescent signal (Ex 540, Em 590) using a Biotek Synergy H4 plate reader.
3. RESULTS 3.1. Metabolism of AGS62P1 in Molm-13 Xenograft Tumors. Synthetic standards for pAF-AGL-0182-30, AGL0182-30, and the N−O cleavage product of AGL-0182-30 were analyzed by high-resolution mass spectrometry to serve as mass and retention time references to which peaks in tumor samples were compared to positively identify metabolites. In order to positively identify a peak seen in a tumor sample as a metabolite of AGS62P1, the peak had to match the retention time and mass (±5 ppm) of the synthetic metabolite standard, and the isotopic pattern of the peak had to be observed to confirm the peak’s charge state. Furthermore, the peak had to be absent from injection blanks and vehicle-treated tumor samples. On the basis of these criteria, a single metabolite of AGS62P1 was observed in tumors: pAF-AGL-0185-30 (Figure 3A). This metabolite was observed in crude isopropanol extracts and affinity-purified extracts of all four MOLM-13 D
DOI: 10.1021/acs.molpharmaceut.8b00225 Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Article
Molecular Pharmaceutics
below. Additionally, the known terminal amide hydrolysis of the payload which occurs in tumors was expected to be catalyzed in liver by esterases, and the free azide of the payload was reasoned to have the potential to be reduced by liver enzymes to yield the free amine.14 Results show that the only metabolism detected in human hepatocytes was hydrolysis of the terminal amide to yield AGD-0185 (Figure 3D). Thus, the payload portion of AGL-0182-30 (AGD-0182) appears to be metabolized by human hepatocytes similarly in xenograft tumors and rat liver. 3.4. In Vitro Cytotoxicity of Metabolite and Payload. The in vitro cytotoxicity of the metabolite was assessed in two relevant cell lines. The metabolite pAF-AGL-0185-30 was more than 100-fold less potent than the payload AGD-0182 in both cell lines (see Supporting Information SI-Figure 1 and SI-Table 1). Because the corresponding ADC AGS62P1 was potent in both of these cell lines (subnanomolar IC50 values; data not shown), it is possible that the inferior potency of pAF-AGL0185-30 is due to relatively lower membrane permeability.18,19 3.5. pAF-AGL-0185-30 Does Not Inhibit Cytochrome P450 Enzymes in Vitro. A study was conducted to evaluate the ability of pAF-AGL-0185-30 to inhibit, in vitro, the major CYP enzymes in human liver microsomes (namely, CYP1A2, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP3A4/ 5) with the aim of ascertaining the potential of pAF-AGL-018530 to inhibit the metabolism of concomitantly administered drugs. Under the experimental conditions examined, there was little or no direct, time-dependent (NADPH-independent) or metabolism-dependent (NADPH-dependent) inhibition of any CYP enzyme activity examined by pAF-AGL-0185-30 (Table 1). 3.6. pAF-AGL-0185-30 Does Not Induce Cytochrome P450 Enzymes in Cultured Hepatocytes. This study was conducted to investigate the effects of treating primary cultures of cryopreserved human hepatocytes with pAF-AGL-0185-30 on the expression of cytochrome P450 (CYP) enzymes. During the study, hepatocytes formed confluent monolayers with few intercellular spaces; they were cuboidal in shape with intact cell membranes and granular cytoplasm with one or two centrally located nuclei.
tumors treated with AGS62P1 (two harvested 1 day postdose and two harvested 3 days postdose) and was absent from vehicle-treated tumors and blank injections. The data does not rule out the presence of these metabolites, but pAF-AGL-018530 was observed at over 10× the LOD of 8 nM in tumors. 3.2. Metabolism of AGS62P1 in Rats. A follow-up study was done to compare the metabolism of AGS62P1 in rats, used for a preclinical toxicology assessment, to that detected in xenograft tumors. Based on the same criteria applied for metabolite identification described previously for tumor samples, a single metabolite of AGS62P1 was observed in rat serum: pAF-AGL-0185-30. This metabolite was observed in the extract of a rat serum sample collected 24 h after dosing with AGS62P1 (Figure 3B) and was absent from blank injections and serum samples taken prior to dosing with AGS62P1. At the conclusion of the study, following 4 weekly doses, livers were harvested and extracts prepared. Similarly, pAF-AGL-0185-30 was the only metabolite observed in rat liver (Figure 3C). 3.3. In Vitro Metabolism of AGD-0182 in Human Hepatocytes. To follow-up on the in vivo identification of metabolites of the ADC, the payload AGD-0182 was incubated with human hepatocytes in vitro to assess the potential for any secondary metabolism in patients. Due to the low abundance of pAF-AGL-0185-30 found in rat liver, it is possible that minor secondary metabolites may have been present but below the limits of detection. Based on the known metabolism of MMAE,13 several potential metabolites are shown in Figure 4
Figure 4. Potential hepatic metabolism of AGD-0182 payload.
Table 1. Summary of Results: In Vitro Evaluation of pAF-AGL-0185-30 as an Inhibitor of Human CYP Enzymes direct inhibition
time-dependent inhibition
metabolism-dependent inhibition
zero-min preincubation
30 min preincubation without NADPH
30 min preincubation with NADPH
enzyme
substrate
IC50 (nM)a
inhibition observed at 1000 nM (%)b
IC50 (nM)a
inhibition observed at 1000 nM (%)b
IC50 (nM)a
inhibition observed at 1000 nM (%)b
potential for metabolismdependent inhibitionc
CYP1A2 CYP2B6 CYP2C8 CYP2C9 CYP2C19 CYP2D6 CYP3A4/5 CYP3A4/5
phenacetin efavirenz amodiaquine diclofenac S-mephenytoin dextromethorphan midazolam testosterone
>1000 >1000 >1000 >1000 >1000 >1000 >1000 >1000
3.0 1.5 NAd NA 1.0 0.0 9.4 13
>1000 >1000 >1000 >1000 >1000 >1000 >1000 >1000
3.3 4.9 1.9 NA 9.0 4.1 10 4.5
>1000 >1000 >1000 >1000 >1000 >1000 >1000 >1000
6.8 1.1 NA 0.3 7.4 1.9 10 7.7
no no no no no no no no
a
Average data (i.e., percent of control activity) obtained from duplicate samples for each test article concentration were used to calculate IC50 values. Inhibition observed (%) is calculated with the following formula (results are rounded to two significant figures): inhibition observed (%) = 100% − percent solvent control. cMetabolism-dependent inhibition was determined by comparison of IC50 values both with and without preincubation and with and without NADPH-generating system present in the preincubation, by comparison of the observed inhibition (%) for all preincubation conditions and by visual inspection of the IC50 plots. dNA = not applicable. No value was obtained as the rates at the highest concentration of pAFAGL-0185-30 evaluated (1000 nM) were higher than the control rates. b
E
DOI: 10.1021/acs.molpharmaceut.8b00225 Mol. Pharmaceutics XXXX, XXX, XXX−XXX
Article
Molecular Pharmaceutics
in monkeys and represent a worst-case scenario where the metabolite is released in total upon dosing. We know from TK data of repeat dosing of AGS62P1 that the metabolite is slowly eliminated upon processing of the ADC, AGS62P1. The slow elimination of the metabolite is attributed to the continual processing of AGS62P1 to the metabolite and the antibody metabolite. The PK profile of the metabolite, pAF-AGL-0185-30 follows an absorption profile with a Tmax around 48−168 h post AGS62P1 administration (see Supporting Information, Figure SI-5) followed by elimination. Based on the TK data following dosing with the intact ADC in monkeys, the maximal concentrations of metabolite achieved are >5000 fold lower than the concentration of the ADC or total Ab on a molar basis. Therefore, the concentrations achieved in the single dose studies are large multiples of what would be expected to be achieved in a repeat dosing setting.
Treatment of cultured human hepatocytes with positive control CYP inducers caused anticipated and appropriate increases in CYP mRNA expression. Omeprazole (50 μM) caused increases ranging from 71.2- to 156-fold in CYP1A2 mRNA levels. Phenobarbital (750 μM) caused increases ranging from 10.3- to 15.3-fold change in CYP2B6 mRNA levels. Rifampin (20 μM) caused increases ranging from 15.4to 21.3-fold change in CYP3A4 mRNA levels (data not shown). Treatment of cultured human hepatocytes with up to 300 nM pAF-AGL-0185-30 had little or no effect on CYP1A2 mRNA levels (range of 0.626- to 1.82-fold change; Table SI-2), CYP2B6 mRNA levels (range of 0.671- to 1.34-fold) and CYP3A4 mRNA levels (range of 0.539- to 1.19-fold change). Under the conditions of this study, where the positive controls caused anticipated and appropriate increases in CYP enzyme expression, treatment of cultured human hepatocytes with pAF-AGL-0185-30 had little or no effect (i.e.,
