In vivo Drug-Linker Stability of an Anti-CD30 Dipeptide-Linked Auristatin Immunoconjugate

Effective antibody-drug conjugates (ADC) combine high drug-linker stability in circulation and efficient intratumoral release of drug. Conjugation of monomethyl auristatin E (MMAE) to the anti-CD30 monoclonal antibody (mAb), cAC10, produced a selective and potent ADC against CD30⁺ anaplastic large cell lymphoma and Hodgkin's disease models. This ADC, cAC10-valine-citrulline-MMAE, uses a protease-sensitive dipeptide linker designed to release MMAE by lysosomal cathepsin B in target cells but maintain a stable linkage and attenuate drug potency in circulation. To evaluate ADC stability in vivo, we developed methods for measuring drug/mAb ratios at progressive times in plasma from ADC-treated mice and nonhuman primates. Anti-idiotype mAb permitted the capture and quantitation of mAb cAC10, whereas antidrug mAb and MMAE-conjugated horseradish peroxidase reporter provided quantitative detection of conjugated drug following its in vitro release by cathepsin B. These data were validated by an alternative ELISA using anti-idiotype and anti-MMAE mAbs for capture and detection, respectively. Both methods differentiated ADC with variable levels of drug loading and were subsequently applied to stability studies in severe combined immunodeficient mice and cynomolgus monkeys. Evaluation of ADC from mouse circulation showed the linker half-life to be ∼144 hours (6.0 days), significantly greater than that reported for disulfide- or hydrazone-linked ADCs in mice or human trials. In cynomolgus monkey, the apparent linker half-life was ∼230 hours (9.6 days), suggesting that the drug-linker will be highly stable in humans. These data represent the longest reported drug-linker half-life to date and provide the basis for the pronounced specificity and antitumor activity of cAC10-valine-citrulline-MMAE.

Antibody-targeted cancer therapies provide focused treatment to the tumor cell while reducing untoward toxicity to normal cells and tissues (1, 2). The efficacy of an antibody-drug conjugate (ADC) depends on coupling a potent cytotoxic agent to a tumor-selective internalizing monoclonal antibody (mAb) through a conditionally labile linker. The linker must show a high degree of stability in circulation and exhibit efficient release of active drug within the tumor cell to provide targeted antitumor activity. Common linkers used in ADCs for cancer therapy include acid labile hydrazones and disulfides that are cleaved by biological reducing agents. Mylotarg, a clinically approved anti-CD33 mAb conjugated to calicheamicin, illustrates the utility of hydrazone linkers for attachment and delivery of drug (3, 4). Following internalization, the hydrazone is believed to be hydrolyzed within the acidic environment of target cell endosomes and lysosomes to release active drug (4). SGN-15 (cBR96-doxorubicin), an anti-Lewis^Y mAb-doxorubicin immunoconjugate currently in phase II clinical trials, is also designed to release active drug on hydrazone hydrolysis within these acidic compartments (5, 6). Other ADCs in development, including conjugates of the maytansine derivative, DM1 (7), the DNA alkylating agent, CC-1065 (8), and a second-generation taxane (9), are conjugated to their respective mAbs by disulfide linkers specifically designed to release drug on reduction by intracellular thiols.

We have described previously an ADC composed of a chimeric anti-CD30 mAb (cAC10) conjugated to the antimitotic agent, monomethyl auristatin E (MMAE), through a novel, protease-sensitive valine-citrulline dipeptide linker (10, 11). Other ADCs using the same drug-linker are directed against E-selectin and EphB2 for the targeted treatment of prostate and colorectal cancer, respectively (12, 13). The valine-citrulline dipeptide linker was designed for maximum serum stability and for efficient drug release by human cathepsin B (14, 15). Thus, cAC10-valine-citrulline-MMAE (cAC10-Val-Cit-MMAE) incubated with purified human lysosomal cathepsin B leads to rapid hydrolysis of the linker and generation of the free cytotoxic agent, MMAE (10). Similar results were obtained using crude lysosomal extracts or cathepsin B to hydrolyze the valine-citrulline linker of a doxorubicin immunoconjugate, demonstrating the versatility of the dipeptide for the conjugation and release of multiple classes of drug (16). Ex vivo studies with cAC10-Val-Cit-MMAE showed that the ADC is highly stable in plasma. In human or mouse plasma, respectively, only 2% or 5% of the conjugated drug was released after 10 days at 37°C (11). Consistent with these observations, in vitro cytotoxicity assays with cAC10-Val-Cit-MMAE showed up to 4 logs of selectivity toward antigen-positive cells following 96 hours of continuous exposure (11). Collectively, the dipeptide-linked MMAE conjugates showed efficient release of drug in the presence of lysosomal proteases and exceptional drug and drug-linker stability in plasma and in cell culture.

To assess ADC stability in preclinical studies, we developed ELISA-based assays for quantifying both mAb and mAb-associated MMAE from in vivo samples. These methods were evaluated in vitro using cAC10 conjugates with two, four, and eight Val-Cit-MMAE drug-linkers per antibody (E2, E4, and E8, respectively) and then applied to samples from ADC-treated naive severe combined immunodeficient (SCID) mice and cynomolgus monkeys. The data show differences in circulating ADC stability compared with prior evaluations in plasma yet indicate that the Val-Cit-MMAE drug-linker is significantly more stable in vivo than disulfide and acid-labile hydrazone linkers reported previously.

Chemicals, Reagents, and Enzymes. Cathepsin B (human liver) was from EMD Biosciences (San Diego, CA). Horseradish peroxidase (HRP) was from Worthington Biochemical Corp. (Lakewood, NJ). Biotin (long arm) NHS was from Vector Laboratories (Burlingame, CA). Mouse anti-human IgG-HRP (clone JDC-10) and goat anti-human κ IgG-HRP were from Southern Biotechnology Associates, Inc. (Birmingham, AL). F(ab′)2 goat anti-mouse IgG-HRP (F_cγ fragment specific) and goat anti-human IgG-HRP (F_cγ fragment specific) were from Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA). MMAE and maleimidocaproyl-valine-citrulline-MMAE (mc-Val-Cit-MMAE) were prepared as described previously (10).

Synthesis of maleimidocaproyl-MMAE (mc-MMAE) required the addition of maleimidocaproic acid to a solution of MMAE in dichloromethane followed by the addition of diethyl cyanophosphonate and diisopropylethylamine. The product was collected and isolated by flash chromatography on silica gel.

Synthesis of maleimidocaproyl-valine-citrulline acid was initiated by mixing Fmoc-Val-Cit (16) in anhydrous dimethylformamide with an equal volume of diethylamine at ambient temperature for 2 hours. After concentration, the solution was added dropwise into diethyl ether. The resulting white precipitate was collected and dissolved in DMSO followed by the additionof maleimidocaproyl-OSu and diisopropylethylamine. Product was purified using C₁₂ reverse-phase preparative high-performance liquid chromatography.

Production of Anti-Idiotype cAC10 mAb. Female BALB/c mice (Harlan, Indianapolis, IN) were immunized with mAb cAC10 for the purpose of generating anti-idiotype cAC10 mAbs. Single-cell suspensions of spleen and lymph node cells from immunized mice were fused to X63-Ag8.653 myeloma cells as described previously (17, 18). Hybridomas were selected and maintained in complete Iscove's Modified Dulbecco's Medium supplemented with 10% (v/v) fetal bovine I serum, 5% (v/v) hybridoma cloning factor, and 1 × HT.

Hybridomas were screened for specific reactivity against the F(ab) fragment of AC10 mAb by standard ELISA procedures (19). Supernatants that were positive for AC10 F(ab) (A₄₅₀ > 0.500) were rescreened against an irrelevant murine F(ab) fragment control. Selected hybridomas were expanded and cloned by limiting dilution and anti-idiotypic selectivity of the clonal mAb was shown by testing against normal mouse IgG, irrelevant chimeric IgG, and by specific blockage of murine AC10 mAb binding to CD30⁺ cells (data not shown).

Production of mAbs Recognizing MMAE. Keyhole limpet hemocyanin was thiolated with 2-iminothiolane and conjugated to mc-Val-Cit-MMAE to generate the KLH-Val-Cit-MMAE immunogen. Briefly, a thiolation reaction mixture containing 8 mg keyhole limpet hemocyanin and 2.7 mmol/L 2-iminothiolane in 50 mmol/L sodium borate (pH 8.0) was incubated for 1 hour at room temperature before PD-10 (Sephadex G-25 medium) chromatography (Amersham Biosciences, Piscataway, NJ). Thiolated keyhole limpet hemocyanin was mixed with 20 molar equivalents of mc-Val-Cit-MMAE in the presence of 20% acetonitrile. After 2 hours on ice, the reaction was quenched with 20 molar equivalents of cysteine per mole of drug-linker, and the conjugate was repurified by PD-10 chromatography and filter sterilized.

Female BALB/c mice were immunized with the KLH-Val-Cit-MMAE immunogen for the purpose of generating mAbs specific to MMAE and related drug derivatives. Hybridoma fusions were generated and maintained by standard methods (17, 18).

Hybridoma supernatants were screened for reactivity to MMAE conjugates of the chimeric mAb, cBR96, using a standard ELISA. Hybridomas positive for conjugated MMAE (A₄₅₀ > 0.500) were expanded and cloned. Anti-MMAE mAbs were also assessed for binding to unconjugated MMAE by preincubation with free drug and then screened for blockage of mAb binding to MMAE-conjugated cBR96 by ELISA.

Antibody Purification. Anti-idiotype cAC10 mAb (clone 30.16) and anti-MMAE mAbs (clones SG2.15, SG3.33, and SG3.218) were purified from clarified hybridoma supernatants by recombinant protein G-Sepharose chromatography (Pierce Biotechnology, Inc., Rockford, IL). mAb purity was analyzed by SDS-PAGE and size exclusion chromatography. Concentration of mAbs was determined by A₂₈₀/ε = 1.4 for a 1 mg/mL solution.

Preparation of cAC10-Val-Cit-MMAE ADCs and Standards. The cAC10-Val-Cit-MMAE (E8) conjugate was prepared as described previously (10, 11). To prepare cAC10 conjugates with two and four drug-linkers per antibody (E2 and E4, respectively), cAC10 was partially reduced with DTT and conjugated to the mc-Val-Cit-MMAE drug-linker. Pure E2 and E4 were resolved from the conjugation reaction mixture by hydrophobic interaction chromatography, and the drug/mAb ratio was determined as described previously by Hamblett et al. (20). To generate the cAC10-Val-Cit-OH standard, the maleimidocaproyl-valine-citrulline acid linker was conjugated to fully reduced cAC10 using the method described for the E8 ADC (10, 11).

Animal Studies. All animal studies were carried out in accordance with Animal Care and Use Committee guidelines. Sterile, purified E4 was given as a single i.v. dose (10 mg/kg) to 10 female SCID mice (20-25 g, Harlan). At specific times thereafter (4 hours, 1 day, 3 days, 6 days, and 9 days), two animals were anesthetized with diethyl ether, and terminal blood samples were drawn from the inferior vena cava. Animals used for saphenous blood draws at 10 minutes postinjection were subsequently sacrificed at the 9-day time point. Blood was collected into heparin-coated tubes and centrifuged (12,500 × g, 5 minutes) to isolate plasma, and the samples stored at −20°C until analysis.

Sterile E8 was given as a single i.v. dose (0.3 or 3.0 mg/kg) to three purpose-bred cynomolgus monkeys (Charles River Laboratories, Reno, NV). At specific times thereafter (10 minutes, 4 hours, 1 day, 2 days, 3 days, and 8 days), blood samples were collected into heparin-coated tubes, and plasma was isolated and stored as described above.

ELISA for the Detection and Quantitation of cAC10 mAb. The concentration of the cAC10 mAb component of ADC in plasma from animal studies was determined using a direct ELISA as described previously (20). The same ELISA format was also used to quantitate the cAC10 mAb component of ADC standards (E2, E4, or E8) and experimental samples after in vitro treatment of ADC with cathepsin B.

Preparation of HRP-MMAE Conjugate. HRP was thiolated with 2-iminothiolane and conjugated to mc-MMAE to generate the HRP-MMAE reporter enzyme-drug conjugate. Briefly, a thiolation reaction mixture containing 0.2 mmol/L HRP (8 mg/mL) and 50 mmol/L 2-iminothiolane in 25 mmol/L sodium borate decahydrate (Na₂B₄O₇·10H₂O) buffer (pH 9.0) was incubated for 1 hour at 37°C. Unreacted 2-iminothiolane was removed by passage through a PD-10 desalting column equilibrated in PBS (pH 7.4). Peak fractions were pooled and mc-MMAE was coupled to thiolated HRP (HRP-SH) at a molar ratio of 3:1. The final conjugation reaction mixture contained 80 μmol/L HRP-SH (3.2 mg/mL) in sodium borate buffer [50 mmol/L H₃BO₃, 50 mmol/L NaCl (pH 8.0); 80% v/v] and 240 μmol/L mc-MMAE in ice-cold CH₃CN (20% v/v). After 30 minutes on ice, the reaction was terminated with a 20-fold molar excess of free cysteine (4.8 mmol/L) before PD-10 chromatography. Peak fractions containing HRP-MMAE (exchanged into PBS) were pooled and evaluated for extent of modification using the thiol-reactive dye, Alexa Fluor 594 C₅ maleimide (Molecular Probes, Inc., Eugene, OR).

Biotinylated Antibody. Biotin (long arm) N-hydroxysuccinimide ester was incubated with an anti-MMAE mAb (clone SG3.33) at a molar ratio of 20:1 to generate SG3.33-biotin. Briefly, a biotinylation reaction mixture containing 19 μmol/L mAb SG3.33 (4.2 mg/mL) and 380 μmol/L biotin (long arm) N-hydroxysuccinimide ester in sodium borate buffer was incubated for 1 hour at 37°C. The reaction mixture was applied to a PD-10 column equilibrated in PBS, and peak fractions containing mAb SG3.33-biotin were pooled. The avidin-HABA reagent (Pierce Biotechnology) was used to calculate a conjugation ratio of 2.9 moles of biotin per mole of mAb SG3.33.

Drug Release by In vitro Cathepsin B Treatment of ADC-Containing Plasma Samples. All E2, E4, and E8 ADC standards and experimental plasma samples were normalized to 20 μg/mL cAC10 mAb component of ADC into plasma (normal mouse, cynomolgus monkey, or untreated SCID mouse) before further dilution into the cathepsin B reaction mixture. Cathepsin B (50 units/mL) was preincubated in activation buffer containing 50 mmol/L sodium acetate (pH 5.0), 2 mmol/L DTT, and 25% (v/v) glycerol for 15 minutes at 37°C and then added as a 20% (v/v) addition to initiate each reaction. Final in vitro cathepsin B reaction mixtures contained 50 mmol/L sodium acetate (pH 5.0), 2 mmol/L DTT, 10 units/mL activated cathepsin B, 5% (v/v) glycerol, and 0.5 μg/mL cAC10 mAb in a 2.5% (v/v) plasma background. Incubation occurred for 6 hours at 37°C. Reactions were terminated by addition of E64 cysteine protease inhibitor (50 μmol/L, Sigma-Aldrich, St. Louis, MO).

Free MMAE Competition ELISA. MMAE released from ADC following in vitro incubation with cathepsin B was quantified using a MMAE competition ELISA. Microtiter plates (96-well) were coated with an anti-MMAE mAb (clone SG3.218) at a concentration of 1 μg/mL in 100 mmol/L sodium carbonate-bicarbonate buffer (pH 9.6) and blocked with PBS containing 0.1% Tween 20 (PBST) and 1% bovine serum albumin (BSA). Cathepsin B–treated ADC reference standards and experimental samples and free MMAE quantitation standards were prepared by titration into diluent buffer [50 mmol/L sodium acetate (pH 5.0), 2 mmol/L DTT, 5% (v/v) glycerol, and 2.5% (v/v) plasma] and mixed 1:1 (v/v) with a fixed concentration of HRP-MMAE (4 ng/mL in PBS-1% BSA). HRP-MMAE/diluent controls and PBS-1% BSA/diluent blanks were prepared similarly to obtain absorbance values corresponding to saturating HRP-MMAE and nonspecific background levels, respectively. ADC reference standards, experimental samples, and free MMAE standards were subsequently added to wells in triplicate. HRP-MMAE/diluent controls and PBS-1% BSA/diluent blanks were included in replicates of eight for each sample set. Competition reactions were done for 1 hour at room temperature. Wells were washed and TMB substrate was added (100 μL per well). After 30 minutes at room temperature, reactions were terminated by addition of 1 N H₂SO₄ (50 μL per well), and absorbance values were measured at 450 nm. Absorbance values obtained for free MMAE standards were subjected to a four-variable curve fit (GraphPad Prism version 4.01). Acceptance criteria for dilutions of ADC reference standards and experimental samples were restricted to between 30% and 70% HRP-MMAE binding relative to the HRP-MMAE/diluent controls.

ADC ELISA for the Detection of ADC. Plasma samples from ADC-treated SCID mice were analyzed for mAb-conjugated MMAE using a sandwich ELISA. Microtiter plates were coated at a concentration of 1 μg/mL anti-idiotype cAC10 mAb (clone 30.16) diluted into PBS (100 μL per well). After quantifying the cAC10 mAb component of ADC, all plasma samples were normalized to 100 ng/mL cAC10 mAb, serially diluted into PBST-1% BSA, and added to wells in duplicate. Identical dilution series of E4 and E2 standard controls were similarly prepared and assayed in parallel with the experimental samples. After 1 hour at room temperature, wells were washed and anti-MMAE mAb SG3.33-biotin (1 μg/mL in PBST-1% BSA) was added and incubated for 1 hour at room temperature. After washing, streptavidin-HRP (diluted into PBST-1% BSA) was added and incubated for 1 hour at room temperature, and detection was done using TMB substrate. Absorbance was measured at 450 nm against a reference wavelength of 630 nm.

Western Immunoblot Analysis. Plasma from ADC-treated SCID mice was analyzed for conjugated MMAE and cAC10-derived human γ and κ chains. Briefly, samples were normalized for cAC10 mAb component of ADC, reduced, and subjected to 4% to 12% Bis-Tris SDS-PAGE. Protein was transferred to polyvinylidene difluoride membranes and blocked with 2% nonfat dry milk in PBST. Individual membranes were probed with antibodies directed against conjugated MMAE [anti-MMAE mAb (clone SG2.15)], human γ chain [HRP-conjugated goat anti-human IgG (F_cγ fragment specific)], or human κ chain (HRP-conjugated goat anti-human κ IgG). All HRP-conjugated antibodies were diluted into PBST-1% BSA and incubated for 1 hour at room temperature. The anti-MMAE mAb was diluted into PBST-1% BSA, incubated overnight at 4°C, and probed with F(ab′)2 goat anti-mouse IgG (F_cγ)-HRP. HRP-conjugated antibody-antigen complexes were detected using the SuperSignal West Pico Chemiluminescent kit (Pierce Biotechnology).

Preparation of cAC10-Val-Cit-MMAE with eight drugs per mAb (E8) was described previously (10). Conjugates of cAC10-Val-Cit-MMAE with fewer drugs per mAb (E2 or E4) were produced by partial reduction of cAC10 interchain disulfide bonds with DTT followed by conjugation to mc-Val-Cit-MMAE. The thiol-reactive, maleimido-containing drug-linker used for the preparation of cAC10 ADCs contains a protease-sensitive valine-citrulline dipeptide sequence designed for optimal stability in human plasma and efficient cleavage by human cathepsin B (14, 15). Once internalized into the tumor cell, vesicular protease activity can potentially metabolize both mAb and linkage to release active drug. Protease-mediated hydrolysis at the COOH-terminal side of the dipeptide sequence is followed by a spontaneous [1,6]-fragmentation of the adjacent p-aminobenzylcarbamate spacer to generate the free drug, MMAE, and a valine-citrulline-OH linker reaction product that remains covalently attached to cAC10 (Fig. 1).

The strategy used for preparing partially loaded ADCs produced a heterogeneous mixture of cAC10 with two, four, or six drugs per mAb (E2, E4, or E6, respectively) and included minor quantities of unmodified cAC10 (E0) and fully loaded E8. Figure 2A shows the cAC10-Val-Cit-MMAE_mix separated into five major peaks. The extent of conjugation for each species was determined by A₂₄₈/A₂₈₀ ratio, as drug-linker attachment results in greater absorbance at 248 nm (λ_max for drug) relative to 280 nm (λ_max for cAC10; Fig. 2A , inset). Homogeneous conjugates with two or four drugs per mAb (E2 and E4) were subsequently separated and purified from the cAC10-Val-Cit-MMAE_mix by hydrophobic interaction chromatography.

Using an anti-idiotype cAC10 mAb to affinity capture ADC in samples derived from in vitro and in vivo studies, a cAC10 mAb ELISA was developed to detect and quantify the mAb component of the ADC. A titration series of purified E2, E4, and E8 compared with that of parental cAC10 showed that conjugation of up to eight drug-linkers per mAb had minimal effect on capture or detection of cAC10 (Fig. 2B). This is an important criterion for the accurate measurement of mAb and ADC derivatives in biological samples where the quantity of ADC and the amount of intact Val-Cit-MMAE drug-linker associated with an ADC population is unknown.

The pharmacokinetic evaluation of cAC10 ADCs required the ability to quantitate both the mAb component and the number of drugs remaining stably attached to the ADC at specific times post–ADC administration. The amount of total mAb (intact ADC, partially loaded, and unconjugated) in plasma was first quantified by the cAC10 mAb ELISA. The amount of conjugated MMAE associated with ADC was subsequently determined by in vitro incubation of plasma samples with cathepsin B and detection of the released drug by competition ELISA using a HRP-MMAE reporter enzyme-drug conjugate and an anti-MMAE mAb. The linker used to couple MMAE to HRP lacks the protease-sensitive dipeptide sequence and p-aminobenzylcarbamate spacer and remains stably associated with HRP in the presence of cathepsin B (data not shown).

An anti-MMAE mAb (clone SG3.218) was established as the most sensitive reagent for the detection of free MMAE in the competition ELISA. To determine the sensitivity of the assay using mAb SG3.218 for the capture of free and HRP-conjugated MMAE, dilutions of free drug were prepared and mixed with a defined quantity of HRP-MMAE conjugate. After incubation on mAb SG3.218-coated plates, the percentage of HRP-MMAE detected for each MMAE dilution point relative to the noncompeted diluent control was determined (Fig. 3). The binding of the HRP-MMAE reporter was effectively competed by free drug in a dose-dependent manner, and saturating levels of MMAE (≥25 nmol/L or 18 ng/mL) eliminated binding of HRP-MMAE. The EC₅₀ of the assay was 2.4 nmol/L (∼1.7 ng/mL) MMAE, and the acceptance criteria for the detection of free MMAE was restricted to between 30% and 70% HRP-MMAE binding or 0.9 to 4.1 nmol/L (∼0.6–2.9 ng/mL) MMAE. To evaluate the effect of the biological matrix on MMAE detection by this assay, drug diluent was supplemented with increasing concentrations of plasma from normal mouse or cynomolgus monkey. Final background concentrations of up to 25% (v/v) plasma from either species did not significantly perturb drug quantitation (data not shown).

To derive the average drug/mAb ratio of an ADC in plasma, experimental samples and ADC standards were normalized to 500 ng/mL mAb and treated with cathepsin B to facilitate the release and quantitation of MMAE using the competition ELISA described above. The presence of endogenous cathepsin B inhibitors in plasma required a dilution to 2.5% (v/v) plasma to achieve >95% hydrolysis of the dipeptide linker as determined by reverse-phase high-performance liquid chromatography analysis of released MMAE (data not shown).

The MMAE competition ELISA was first evaluated using progressively loaded cAC10-Val-Cit-MMAE to determine if the assay could discriminate between conjugates with two, four, or eight drugs per mAb. Samples of E2, E4, and E8 were prepared in normal mouse or cynomolgus monkey plasma and treated with cathepsin B to release drug, and the concentration of MMAE and cAC10-Val-Cit-OH reaction products were quantified in the respective assays. Experimentally determined drug/mAb ratios in both plasma backgrounds (Table 1) approximated the known ratios for each ADC, establishing the utility of combining data from the cAC10 mAb ELISA and MMAE competition ELISA to determine the average drug/mAb ratio of ADC in plasma samples from treated animals.

Prior studies with cAC10-Val-Cit-MMAE (E8) in SCID mice showed antitumor efficacy at doses of <1 mg/kg, and the maximum tolerated dose was >40 mg/kg (11). Further improvements to this therapeutic window were established using E4, which displayed a significantly increased circulating half-life compared with E8 and comparable efficacy and an MTD that was better than twice that of E8 (20). To evaluate the pharmacokinetics of the ADC and the in vivo stability of the Val-Cit-MMAE drug-linker, E4 was given i.v. to SCID mice at a single dose of 10 mg/kg. Animals were bled at progressive times, and the concentration of total cAC10 mAb component of ADC in plasma was determined (Fig. 4A). The in vivo stability of the drug-linker expressed as drug/mAb ratio (Fig. 4B) was also measured as a function of time. The drug/mAb ratio obtained for an E4 reference standard (4.8 drug/mAb) in SCID mouse plasma was in close agreement with that determined previously in normal mouse plasma (4.5 drug/mAb; Table 1), whereas experimental samples exhibited a time-dependent loss of MMAE ranging from 4.1 drug/mAb on day 1 to 1.7 drug/mAb on day 9. Log transformation of the drug/mAb ratio and fitting a linear curve to the data in Fig. 4B showed that the half-life of the drug-linker was ∼6.0 days using a noncompartmental model (PK Solutions 2.0, Summit Research Services, Montrose, CO). A parallel study with E8 in SCID mice showed a similar rate of drug loss, indicating Val-Cit-MMAE drug-linker stability is independent of the total number of drugs loaded per mAb (data not shown).

A limited stability study of E8 was also done in cynomolgus monkey. The mAb component of the ADC at each time point was determined (Fig. 4C), and plasma samples were assayed to derive the drug/mAb ratios of the ADC over time (Fig. 4D). As with SCID mice treated with E4 and E8, a time-dependent loss of MMAE from E8 was detected in cynomolgus monkey, although at a slower rate, indicating that the ADC is more stable in circulation in nonhuman primates. The average extrapolated half-life of the drug-linker for all animals in this experiment was estimated to be ∼9.5 days after fitting a linear curve to the data.

An ADC ELISA provided a qualitative alternative measurement of the amount of intact drug-linker stably associated with the ADC in plasma samples from treated SCID mice. Using an anti-idiotype cAC10 mAb to affinity capture E4, conjugated drug was detected with biotinylated anti-MMAE mAb (clone SG3.33) and streptavidin-HRP. A decrease in the signal intensity of an experimental sample relative to an E4 standard control indicated loss of conjugated MMAE. After normalizing for cAC10 mAb content, serial dilutions of plasma samples ranging from 100 to 0.4 ng/mL ADC were assayed in parallel with E4 and E2 reference standards. The E2 standard titration curve was included to define the point at which 50% of the conjugated Val-Cit-MMAE were no longer detected and estimate the half-life of the drug-linker associated with E4. Titration of plasma samples taken at 10 minutes and 4 hours postinjection yielded curves overlying that of the E4 standard and suggested that the drug-linker is largely intact 4 hours postinjection (Fig. 5A). Titration curves of samples taken at 1 and 3 days postinjection showed time-dependent shifts toward the E2 standard, suggesting that loss of drug from the ADC occurred (Fig. 5B). Titration curves of samples taken from mice 6 and 9 days postinjection approached or progressed beyond the curve for the E2 standard (Fig. 5C), suggesting that the half-life of the drug-linker in rodent circulation is ∼6 days. To ensure that shifts in signal intensity were related to drug loss and not to a decrease in total immobilized cAC10, all samples from 10 minutes to 9 days were subjected to ELISA analysis for quantifying total cAC10 mAb. Overlaying the dose-response curves (Fig. 5D) show that comparable levels of the cAC10 ADC were immobilized for each experimental sample. Samples obtained from mice treated with E8 were similarly analyzed and displayed an identical time-dependent shift in the signal intensity of the experimental samples relative to E8 and E4 reference standards (data not shown). Taken together, the results of the drug/mAb ratio assay (Fig. 4B) and the ADC ELISA (Fig. 5A-D) strongly suggest the in vivo half-life of the conjugated Val-Cit-MMAE drug-linker in tumor-free SCID mice is ∼6 days.

Biochemical characterization of Val-Cit-MMAE ADCs has shown that appended drugs are distributed to sulfhydryl groups normally comprising the IgG interchain disulfide bonds (10). Thus, for maximally loaded E8, three Val-Cit-MMAE drug-linkers are appended to each IgG heavy chain (H chain) and one to each light chain (L chain). To examine drug cleavage from L and H chains of E4 given to naive SCID mice, plasma samples were subjected to Western analysis using an anti-MMAE mAb (clone SG2.15). Figure 6A (lane C) shows MMAE-conjugated H and L chains of the E4 reference standard control. Based on high-performance liquid chromatography analysis of reduced samples, 30% of the purified E4 population are conjugated at the sites of the heavy-heavy interchain disulfides, whereas 60% of the population are conjugated at the sites of the heavy-light interchain disulfides (data not shown). This heterogeneity is detected as two distinct H-chain bands corresponding to either one or two appended drug-linkers, respectively (Fig. 6A,, H1 or H2). A single band corresponding to L chain with one drug-linker is also detected (Fig. 6A,, L1). Experimental samples revealed a distinct, time-dependent decrease in H-chain conjugate signal intensity, whereas levels of L-chain MMAE conjugate seem to remain relatively constant (Fig. 6A), suggesting that loss of MMAE in vivo is associated primarily with release from the H chain. Loss of drug signal was not due to mAb degradation, as sample probed in parallel with antibodies specific to human γ chain (Fig. 6B) and human κ chain (Fig. 6C) showed discrete H and L chains with evolution of a lower molecular weight species with increased time in circulation, likely reflecting H and L chains with reduced or no drug appended. The combined signal intensity of H or L chain does not change significantly across the time course, indicating that equal transfer of both H and L chains occurred for all samples, that mAb integrity is largely preserved following 9 days in circulation, and that loss of signal in Fig. 6A is due to loss of drug from the mAb.

Appropriate choice of mAb, drug potency, and conjugation methodology are key elements in developing optimized ADCs for the targeted treatment of cancer. Stable conjugation of highly potent drugs to tumor-selective mAbs permits selective targeting of chemotherapeutic agents that would otherwise be too toxic for general administration. In circulation, premature drug release could result in untoward toxicities and a reduction in the therapeutic efficacy of the ADC. This latter scenario is compounded by a disparity between the extended circulating half-life of the mAb relative to a reduced drug-linker half-life. Because circulating half-lives of mAbs can exceed 2 weeks (21), an unstable drug-linker could cloak tumor with unmodified mAb, which devoid of cytotoxic activity may compete with or prevent subsequent ADC binding.

We have described previously the highly effective antitumor agent, cAC10-Val-Cit-MMAE (E8 and E4), for targeting CD30-positive malignancies (11, 20). Saturation binding (11) and competition binding²

studies showed that conjugation of up to eight Val-Cit-MMAE drug-linkers per mAb (E8) did not alter the antigen binding characteristics but instead influenced the circulating mAb half-life of the cAC10 mAb component. Whereas the half-life of E8 was significantly decreased relative to that of the unconjugated mAb, reducing the drug/mAb ratio to 4:1 (E4) yielded an ADC half-life comparable with parental cAC10. More importantly, this increased circulating half-life of E4 was combined with an in vivo antitumor efficacy similar to that of E8, further increasing the therapeutic window of the ADC (20). The dipeptide linker used to prepare MMAE-conjugated cAC10 was designed for optimal stability in human plasma and efficient cleavage and release of multiple drug types by the human lysosomal protease, cathepsin B (14, 15). Conjugation of MMAE attenuated its cytotoxic effect by ∼1,000-fold outside of nontarget cells, yet cleavage of the protease-sensitive dipeptide linker within CD30⁺ tumor cells restored full drug potency (11). Subcellular fractionation of CD30⁺ cells treated with cAC10-Val-Cit-MMAE showed MMAE release coincident with lysosomal fractions as evaluated by Western analysis and treatment of cells with inhibitors of cathepsin activity diminished ADC cytotoxicity.³ Comparisons with hydrazone-linked conjugates of cAC10 underscore the importance of the dipeptide-linked MMAE conjugation strategy (10). Cytotoxicity assays showed that peptide-linked MMAE conjugates remained immunologically specific under continuous exposure (96 hours), whereas hydrazone-linked conjugates exhibited time-dependent, increased toxicity toward antigen-negative cells due to hydrolytic liberation of drug. In ex vivo stability studies, the half-life of cAC10-Val-Cit-MMAE in mouse and human plasma was projected to be 30 and 230 days, respectively, vastly superior to similar conjugates that employ hydrazone linkers.

In vitro incubation in serum or plasma, however, does not fully predict for ADC stability in the metabolic complexity of circulation in vivo. Evaluation of ADC stability in samples taken from ADC-treated mice and cynomolgus monkey presents multiple challenges for quantifying both the carrier mAb and its drug component. These include accurate determination of mAb concentration independent of drug loading, efficient release of covalently attached drug, and subsequent quantitation of free drug in biological samples to derive drug/mAb ratios. As drug released from ADC in circulation seems to be cleared much faster than the ADC itself, the free drug was undetectable (data not shown). Using the assays described here, the in vivo half-life of the dipeptide drug-linker associated with E4 in naive SCID mice was estimated to be 144 hours (6.0 days). No significant difference was observed between E4 and E8, indicating that drug-linker stability is independent of the extent of drug loading. These methods measure the average loss of drug from an ADC population and do not discriminate between distributive and cooperative drug loss. Although it is possible that the drug is cooperatively stripped from some antibodies before others, Western analysis shows a preferential drug loss from the H chain compared with L chain, suggesting a noncooperative processing. The difference between these results and prior ex vivo plasma stability studies (10, 11) suggest that ADCs are further destabilized by interactions with cells or tissues. ELISA (Fig. 5D) and Western immunoblot analyses (Fig. 6) show that drug is cleaved from the ADC, leaving the mAb and its constituent H and L chains intact. As cAC10 does not cross-react with murine CD30,⁴

increased drug release in vivo is potentially the result of antigen-independent mechanisms, such as F_c receptor–mediated internalization and recycling (22).

Clearly, the relationship between the half-life of the mAb and that of the ADC in circulation is an important determinant in overall ADC potency and efficacy in humans. However, linker stability may not be as problematic when the half-life is more closely matched to that of the circulating mAb. Shown here, the apparent half-life of the E4 valine-citrulline linker in mice is ∼6 days, whereas that of the circulating cAC10 mAb component was reported previously to be 14.0 days (20). The ADC was significantly more stable in cynomolgus monkey compared with mice. The 9.5-day estimate for linker half-life in nonhuman primate suggests that (a) a comparable level of stability will be observed in humans and (b) that this will be more closely aligned with the half-life of the mAb.

In comparison, acid-labile hydrazone and disulfide linkers, which use the low pH environment of lysosomes or high intracellular thiol concentrations to elicit efficient drug release, are relatively unstable in circulation. For example, pharmacokinetic studies of the disulfide-linked maytansine derivative DM1 to mAb huC242 showed the elimination half-life of the mAb to be 99.8 hours, whereas that of conjugated DM1 was 23.9 hours (7, 23). The hydrazone-linked cBR96-doxorubicin showed a doxorubicin half-life of 43 hours, whereas that of the carrier mAb was 300 hours (6). Collectively, these studies show that the in vivo half-life of a hydrazone linker is <2 days, whereas that of a disulfide linker is ∼1 day (6, 23, 24).

Increased linker stability can dramatically increase the time that tumors remain exposed to drug. Assuming that linear pharmacokinetics for all three types of drug-linker apply, the ratio of the relative area under the curve can be calculated for a set of hypothetical ADCs with identical mAb half-lives and estimated drug-linker half-lives of 1.0, 1.8, and 6.0 days (for disulfide-linker, hydrazone-linker, and dipeptide-linker, respectively) over a 2-week period. For a single equivalent dose of each ADC, these calculations yield a 3.3- or 6.0-fold increase in the area under the curve for a dipeptide-linked ADC compared with a hydrazone-linked or disulfide-linked ADC, respectively. To illustrate the significance of this difference, 50% of dipeptide-linked drug remain associated with the ADC, whereas in the corresponding period hydrazone- or disulfide-linked ADCs will contain only 9.9% and 1.6% of the original conjugated drug, respectively. On this basis, the enhanced stability of the dipeptide linker likely contributes to the exceptionally high therapeutic window of mAb-Val-Cit-MMAE conjugates (10, 11, 20) and should prove to be a key factor in the efficacy of ADCs for treating human cancers.

We thank Drs. Dennis Miller and Kevin Hamblett for their assistance with pharmacokinetic evaluations and discussion of data and Damon Meyer, Starr Bernhardt, and Brian Mendelsohn for expert technical assistance.