Process Development and Structural ... - ACS Publications

700 Chesterfield Parkway West, Chesterfield, Missouri, 63017. Page 1 of 33. ACS Paragon Plus Environment. Organic Process Research & Development. 1. 2...
17 downloads 0 Views 3MB Size
Article Cite This: Org. Process Res. Dev. 2018, 22, 286−295

pubs.acs.org/OPRD

Process Development and Structural Characterization of an AntiNotch 3 Antibody−Drug Conjugate Durgesh V. Nadkarni,*,† Qingping Jiang,‡ Olga Friese,§ Nataliya Bazhina,† He Meng,†,⊥ Jianxin Guo,∥ Robert Kutlik,§ and Jeffry Borgmeyer§ †

Bioprocess Research and Development and ‡Analytical Research and Development, Pfizer Inc., 401 North Middletown Road, Pearl River, New York 10965, United States § Analytical Research and Development and ∥Pharmaceutical Research and Development, Pfizer Inc., 700 Chesterfield Parkway West, Chesterfield, Missouri 63017, United States S Supporting Information *

ABSTRACT: The development of a process for the preparation of a conventional anti-Notch 3 antibody−drug conjugate (ADC) is described. The initial reaction conditions used for the conjugation of an auristatin payload to an anti-Notch 3 monoclonal antibody led to the formation of an ADC mixture with a significant level of aggregates. Further process optimization studies resulted in the identification of reaction conditions for formation of the conjugate with a low level of aggregates. The temperature of the antibody reduction step was found to have an impact on the formation of aggregates in the ADC mixture. Differences in the antibody reduction temperatures also caused changes in the distribution of conjugated payload on the ADC species. Stability studies of anti-Notch 3 ADCs prepared by two processes differing in the antibody reduction temperature showed subtle differences in their aggregation propensities. The aggregates produced in the crude ADC reaction mixture could be separated from the desired monomer on the hydroxyapatite column under mild conditions without significantly impacting the average drug loading of the purified ADC.



INTRODUCTION

it is important to either minimize their formation during conjugation or develop efficient purification processes for their removal from the conjugation mixture. Because of promising early preclinical results, a significant quantity of 1 was required for phase I clinical trials. The process that was used for the preparation of small quantities of preclinical ADC supplies was not suitable for its preparation on larger manufacturing scale. In this paper we report our efforts to develop a process for the preparation of 1 that was amenable to scale-up and detailed structural characterization to enable process development.

Antibody−drug conjugates (ADCs) are a novel class of targeted anticancer agents that are composed of a monoclonal antibody that is covalently conjugated to a small-molecule cytotoxic drug. The conjugates are typically formed by conjugating amino acids such as lysine, glutamine, or cysteine on the antibody. Several ADCs with a range of conjugation chemistries are currently undergoing evaluation in various phases of clinical trials. To date, one cysteine-conjugated ADC, brentuximab vedotin (Adcetris), and three lysine-conjugated ADCs, ado-trastuzumab emtansine Kadcyla), gemtuzumab ozogamicin (Mylotarg), and inotuzumab ozogamicin (Besponsa), have received regulatory approval in some countries.1−3 The novel anti-Notch 3 ADC 1 (Figure 1) is targeted for triple-negative breast cancer and ovarian cancers.4 The ADC is composed of an immunoglobulin G1 (IgG1) monoclonal antibody that is conjugated via interchain cysteines, produced after partial reduction of interchain disulfides, to an auristatin-containing linker-payload 2.5 ADC 1 is a mixture of cysteine-conjugated species with drug loadings potentially ranging from 0−8 species (Figure 1). The conjugation of linker-payloads to antibodies can lead to the formation of undesired aggregates in the conjugation mixture.6 Several factors such as the amino acid sequence and posttranslational modifications of the antibody, the hydrophobicity of the linker-payload, the conjugation conditions, the site of conjugation, the drug loading, and the ADC formulation can impact the formation of aggregates in the conjugation reaction.7 The conventional cysteine conjugation of auristatin linkerpayload 2 to an IgG1 antibody leads to the formation of aggregates under thermally stressed conditions.8 Because of possible immunogenicity issues associated with the aggregates, © 2018 American Chemical Society



RESULTS AND DISCUSSION The process used for the preparation of 1 in early discovery research during screening studies is shown in Figure 2.4 Because of the significantly larger material needs to support further clinical trials, this process was not suitable for the preparation of 1 for scale-up and further product development. For early nonclinical studies, we initially prepared 1 using similar tris(2-carboxyethyl)phosphine (TCEP) reduction− conjugation chemistry but eliminated the size-exclusion chromatography (SEC) step. The IgG1 antibody was partially reduced at 37 °C with 2.2 equiv of TCEP. The partially reduced antibody was mixed with a dimethyl sulfoxide (DMSO) solution of linker-payload 2 at 25 °C. The conjugation mixture was purified by ultrafiltration−diafiltration (UF−DF) to perform buffer exchange and to remove organic solvent and Received: October 21, 2017 Published: January 30, 2018 286

DOI: 10.1021/acs.oprd.7b00337 Org. Process Res. Dev. 2018, 22, 286−295

Organic Process Research & Development

Article

Figure 1. Chemical structures of ADC 1 and linker-payload 2.

Figure 3. Impact of quenching agents on the clearance profile of free drug-related species during purification of 1 by UF−DF.

Figure 2. Procedure for the preparation of 1 in early discovery research.

Minimizing Aggregate Formation by Optimization of the Conjugation Conditions. We sought to evaluate whether aggregate formation in the conjugation reaction could be minimized by optimization of the conjugation conditions. During initial experiments it was observed that the temperature of the conjugation step did not impact aggregate formation and that aggregate formation occurred only after the conjugation with 2. While complete absence of agitation during the TCEP reduction step led to an increase in the level of aggregation, the rate of agitation of the mixture during the TCEP reduction step as well as during the conjugation step did not significantly impact the level of aggregation. A design of experiments (DoE) study was carried out to evaluate the conjugation parameters that were suspected to impact the level of aggregation in the conjugation mixture. Three parameters, namely, the concentration of the antibody in the conjugation mixture during the TCEP reduction step, the concentration of DMSO in the conjugation mixture, and the temperature during reduction of the antibody by TCEP were evaluated (Table 1; details are provided in the Supporting Information). The conjugation reactions were executed by partial reduction of the antibody solution at various concentrations with 2.2 equiv of TCEP solution at defined temperatures. Each mixture was stirred at the specified temperature for 90 min. The solution was then cooled to 25 °C. A solution of 2 in DMSO was added to the

residual free drug-related species. Although most of the residual solvent was removed by the UF−DF process, the free drugrelated species could not be removed completely from the isolated conjugate 1. Attempts to improve clearance of free drug-related species by evaluating several membranes and DF conditions proved ineffective. This perhaps was due to the hydrophobic nature of linker-payload 2. In order to render 2 more hydrophilic, several thiol-based quenching agents were screened for postconjugation derivatization of 2 by Michael addition on the maleimide ring. Quenching of the conjugation mixture with L-cysteine resulted in efficient removal of the cysteine adduct through the UF−DF process (Figure 3).9 The conjugate material 1 isolated using the process described above contained ∼4−5% aggregates as determined by SEC assay. Experiments designed to probe the origin of aggregate formation showed that the aggregate formation occurred only after the conjugation with payload 2 and not during the TCEP reduction step. No further change in the aggregation was observed during the quenching and UF−DF steps. Though the level of aggregate content in the isolated 1 was acceptable for preclinical testing, further reduction in the aggregate level in the final isolated conjugate material was desirable for clinical trial evaluation. 287

DOI: 10.1021/acs.oprd.7b00337 Org. Process Res. Dev. 2018, 22, 286−295

Organic Process Research & Development

Article

species were further characterized by native size-exclusion chromatography coupled to mass spectrometry (SEC/MS) (see the Supporting Information). Comparability Assessment of ADC 1 Prepared by Processes A and B. The ADC 1 products prepared by both processes A and B from above experiments were structurally characterized further by mass spectroscopy and other biophysical techniques to confirm their structural identity and assess their comparability. Organic Size-Exclusion Chromatography−Mass Spectrometry. Accurate molecular masses of individually loaded species from samples of 1 prepared by processes A and B were determined under denaturing conditions by organic SEC/MS.10 The interchain disulfide bonds of 1 were partially reduced prior to conjugation with 2. As a result, many conjugated isoforms of 1 were detected in the dissociated form in this assay, as the secondary structure is only maintained under these conditions by those interchain disulfide bonds that remain intact. Two peaks were observed in the organic SEC chromatogram (Figure 6). From the mass analysis results, the earlier-eluting peak (peak 1) contains isoforms that are larger than the light (L) chain, including various conjugated species of heavy (H) and light chains (e.g., HHL, HL, HH, etc.). The most abundant species detected by organic SEC/ESI-MS was composed of disulfidelinked L and H chains with two linker-payload 2 moieties, which is labeled as HL + 2 × 2 (Figure 7). Additional minor and trace-level species such H + 3 × 2, HH + 2 × 2, HHL + 1 × 2, and mAb + 4 × 2 were also observed and are consistent with the interchain disulfide bond conjugation chemistry. For the sample of ADC 1 prepared by process A, the levels of H + 3 × 2 and HH + 2 × 2 species are higher because of the increased relative level of species formed by conjugation to the cysteines involved in disulfide bonds between H and L. The later-eluting

Table 1. Experimental Design, Process Parameters, and Ranges Used in the DoE Study process parameter

low

medium

high

antibody concentration in the conjugation mixture (mg/mL) concentration of DMSO in the conjugation mixture (% v/v) temperature of the TCEP reduction step (°C)

3.8

16.7

29.6

5

15

25

5

22.5

40

conjugation mixture, and the mixture was stirred at 25 °C for 60 min. At the end of the conjugation, an aliquot of the conjugation mixture was evaluated by SEC and hydrophobic interaction chromatography (HIC) assays for aggregate content and drug-to-antibody ratio (DAR) of the crude conjugate. The process parameters of the antibody concentration and the temperature of TCEP reduction were found to have maximum impact on the aggregate content of the conjugate mixture. The level of aggregates varied within a narrow range (1.0−5.4%) for all of the experimental conditions within the DoE model. Lower aggregate levels were observed at low reduction temperature and low antibody concentration. To further confirm the impact of the reduction temperature on aggregate levels, conjugation experiments were carried out at two different reduction temperatures (5 and 37 °C) at the same antibody and DMSO concentrations. The conjugate material 1 prepared from an experiment with a TCEP reduction temperature of 5 °C (process A) contained fewer aggregates (1%) compared with the aggregate level (6%) in the ADC 1 obtained from an experiment where the reduction temperature was 37 °C (process B) (Figure 4). The drug loadings (DAR) for the conjugate materials prepared by both processes A and B were ∼4 by the HIC assay (Figure 5). The identities of individual species in the HIC assay were confirmed by their separation and isolation by analytical HIC. The individual ADC

Figure 4. Impact of the DMSO and antibody concentrations on aggregation at different temperatures, depicted in contour plots. 288

DOI: 10.1021/acs.oprd.7b00337 Org. Process Res. Dev. 2018, 22, 286−295

Organic Process Research & Development

Article

Figure 7. Zero-charge deconvoluted mass spectra of peak 1 in ADC 1 conjugated by (top) process A and (bottom) process B and analyzed by organic SEC/MS.

Figure 5. (a, b) SEC chromatograms of 1 from (a) process A and (b) process B. (c, d) HIC chromatograms of 1 from (c) process A (DAR ∼ 3.7) and (d) process B (DAR ∼ 4.2).

Figure 8. Zero-charge deconvoluted mass spectra of peak 2 in ADC 1 conjugated by (top) process A and (bottom) process B and analyzed by organic SEC/MS.

established by alkylated nonreduced Lys-C peptide mapping by liquid chromatography−mass spectrometry (LC−MS).11 There are a total of 32 cysteine residues in each ADC sample. All of the intrachain cysteine residues are expected to be present as disulfide bonds. There are two intrachain disulfide bonds in each light chain and four in each heavy chain. Interchain cysteines are present as both intact disulfide bonds and as partially reduced ones that are conjugated to 2. There is one interchain disulfide bond between each light and heavy chain

Figure 6. UV280 nm profile of ADC 1 using organic SEC/UV/MS.

peak (peak 2) in the SEC chromatogram contains one moiety of 2 conjugated to the interchain cysteine residue of the L chain involved in the disulfide bond to the H chain (Figure 8). Nonreduced Peptide Mapping by Liquid Chromatography−Mass Spectrometry of ADC 1. The disulfide connectivity in the ADC samples prepared by processes A and B was 289

DOI: 10.1021/acs.oprd.7b00337 Org. Process Res. Dev. 2018, 22, 286−295

Organic Process Research & Development

Article

Figure 9. Extracted mass chromatograms for disulfide-bonded peptides of ADC 1 prepared by (top) process A and (bottom) process B.

Figure 10. Extracted mass chromatograms for conjugated peptides of 1 prepared by (top) process A and (bottom) process B.

and two interchain disulfides between the heavy chains in the hinge region. All of the intrachain and interchain predicted disulfide bonds were detected as expected. Interchain cysteines were also detected as conjugated to 2. The nonreduced peptide mapping results for 1 prepared by both processes demonstrate that the disulfide linkages are consistent with the predicted structure. The similarity of the profiles can be seen from the extracted mass chromatograms for the expected disulfidebonded peptides produced by Lys-C digestion of process A and process B samples of ADC 1 (Figure 9).

The extracted mass chromatograms for conjugated peptides of 1 prepared by processes A and B are shown in Figure 10. The L15 peptide containing Cys214 and the H12 peptide containing Cys200 were detected with one conjugated payload 2. The hinge-region H13 and H13H14 peptides containing residues Cys226 and Cys229 were observed as disulfide-bonded using both Cys226 and Cys229 (Figure 9) and as conjugated with one molecule of 2 to one of the cysteine residues and disulfide-linked via the other cysteine residue with L15 peptide (Figure 10). Additionally, the H13 and H13H14 peptides were 290

DOI: 10.1021/acs.oprd.7b00337 Org. Process Res. Dev. 2018, 22, 286−295

Organic Process Research & Development

Article

observed with two moieties of 2 conjugated to both hingeregion cysteines. However, the extracted mass chromatograms for the conjugated peptides show that the ADC material produced by process A has a lower level of hinge-region conjugation (H13 + 2 × 2 and H13H14 + 2 × 2 peptides) and higher levels of L15- and H12-conjugated peptides compared with the levels of the same conjugated peptides observed for 1 prepared by process B (Figure 10). The difference in the levels of hinge-region conjugation observed for the ADC materials prepared by the two processes was also confirmed by nonreduced capillary gel electrophoresis (CGE) and reduced reversed-phase HPLC (RP-HPLC) (see the Supporting Information). Trace levels of mispaired interchain cysteine-containing peptides were observed for L15 and H12. These include mispairing to the nearby cysteine-containing peptides such as L14 for the L15 peptide and H6, H7, and H15 for the H12 peptide. In addition, trace levels of mispairing of the hingeregion peptide H13H14 conjugated to one molecule of 2 were observed for both the L15 and H12 peptides. The levels of mispairing of the hinge-region peptide H13H14 is higher for the sample of ADC 1 produced by process B. The nonbonded (free thiol) cysteine-containing peptides were also observed at very low levels. Stability Comparison of Samples of 1 Prepared by Processes A and B by Differential Scanning Calorimetry. The stabilities of the ADC samples prepared by processes A and B were evaluated by differential scanning calorimetry (DSC) (Figure 11).12 In the thermogram of the ADC sample

Samples from the solutions of ADC 1 were prepared in a BioRad multiplate PCR plate. The sample plates were subjected to thermal stress from 5 to 95 °C at increments of 1 °C. At each temperature, the plate was equilibrated for 1 min prior to measurement. SYPRO Orange binds to hydrophobic residues or patches of a protein, and spectral changes reflect changes in the environment around the probe. In aqueous solution, the fluorescence emission intensity from the probe is negligible. However, in the presence of an unfolded protein, the fluorescence intensity increases significantly because of the increased exposure to hydrophobic regions to which the probe can bind.13 During the SYPRO Orange fluorescence study, two transitions were observed for both samples of ADC with similar Tm1 at 47 °C, indicating that thermal unfolding of the CH2 domain occurs at similar temperatures. However, Tm2 for ADC 1 prepared by process A (62 °C) is approximately 2 °C higher than that prepared by the process B. The data from this study under heating conditions showed that the ADC 1 prepared by process A was less prone to unfolding than the sample prepared by process B (Figure 12).

Figure 12. SYPRO Orange fluorescence intensity profiles as functions of temperature.

DCVJ is one of a class of viscosity-sensitive fluorescent dyes commonly called molecular rotors. When DCVJ binds to aggregates, the dye becomes partially immobilized, accompanied by an increase in quantum yield.14,15 Studies have shown that DCVJ responds to the early stages of aggregation and thus has a strong preference for oligomeric aggregates. A decrease in fluorescence intensity upon heating to 40 °C was observed for all of the solutions (Figure 13), including the DCVJ buffer solution (not plotted in the figure), indicating that the decrease could be attributed to the inherent property of the fluorophor.16 The transitions detected by DCVJ are at 53 °C for 1 prepared by process A and 50 °C for 1 prepared by process B, indicating that ADC 1 produced by process A is less prone to aggregation. Reduction in the Aggregate Level by Chromatographic Purification of 1 Prepared by Process B. As described above, the level of aggregates formed during the conjugation reaction could be minimized by the conjugation protocol used for process A (5 °C TCEP reduction). However, this protocol also led to changes in the drug distribution profile of ADC 1 compared with the conjugate mixture produced by process B (37 °C TCEP reduction) (Figure 10). Since in vivo properties of 1 prepared by process B were more thoroughly studied during early development work, we decided to develop

Figure 11. DSC thermograms of samples of ADC 1 from processes A and B.

prepared by process B, an additional peak was observed at approximately 62 °C that was not present in the similar trace for the ADC sample prepared by process A. This indicates potentially slightly lower conformational stability of ADC 1 prepared by process B compared with the sample prepared by process A. Assessment of Stability of 1 Prepared by Processes A and B by Extrinsic Fluorescence Spectroscopy. The stability of 1 prepared by processes A and B was further evaluated by extrinsic fluorescence spectroscopy. The fluorescent dyes SYPRO Orange and 4-(dicyanovinyl)julolidine (DCVJ) were used in differential scanning fluorimetry (DSF) experiments to monitor thermal unfolding and formation of aggregates.13 291

DOI: 10.1021/acs.oprd.7b00337 Org. Process Res. Dev. 2018, 22, 286−295

Organic Process Research & Development

Article

Figure 13. DCVJ fluorescence intensity profiles as functions of temperature.

a chromatographic purification procedure for the removal of aggregates from 1. Several resins were screened for the selective separation of aggregates from monomeric 1. Hydroxyapatite resin yielded the best results, leading to the removal of a major portion of the aggregates without significantly impacting the average DAR of the conjugate mixture. The predominantly monomeric ADC material was obtained by elution with phosphate buffer at pH 7 using a linear gradient (Figure 14). The solvent DMSO and the residual free drug species were separated from the ADC in the flow-through fraction during this purification.

Figure 15. Aggregate characterizations by (a) SEC assay, (b) reduced RP-HPLC assay, and (c) SEC/MS assay.

and heavy chains without covalent interchain disulfide linkages separated on the SEC column into two main peaks (peaks 1 and 2). From the results of the mass analyses, peak 1 contained ADC species with various combinations of heavy and light chains (e.g., HHL, HL, HH, etc.) attached to payload 2. The most abundant species detected was composed of the heavy chain with three molecules of 2. Peak 2 contained one molecule of payload 2 conjugated to the light chain. These data from RPHPLC and organic SEC/MS together showed that the majority of isolated aggregates were non-covalent in nature and primarily composed of 8-loaded ADC species. Comparison of the analytical HIC chromatograms of the purified ADC 1 and isolated aggregates (Figure 16) did not show the presence of monomeric 8-loaded species in the sample of ADC 1. Evidently either a majority of the 8-loaded species formed in the reaction mixture during conjugation led to the formation of aggregates or the monomeric 8-loaded ADC species coelute with the aggregates in the HIC chromatogram.

Figure 14. Purification profile of 1 on the hydroxyapatite resin column with phosphate buffer as the eluent (UV detection at 280 nm).

The solution of purified conjugate was concentrated and diafiltered in 20 mM histidine buffer (pH 5.8) using a regenerated cellulose membrane cassette. The isolated ADC material contained primarily monomeric ADC species (∼98%) with the average DAR ∼ 3.6−4.0. Characterization of Aggregates Produced during Preparation of 1. The aggregates isolated during purification of 1 by column chromatography were further characterized by SEC, RP-HPLC, and SEC/MS assays (Figure 15). The SEC results confirmed that the isolated aggregate material consisted of 68% aggregates and 32% monomer species (Figure 15a). The average loading of the aggregate sample was determined to be ∼7 by the RP-HPLC assay (Figure 15b) and SEC/MS assay (Figure 15c). The SEC/MS assay was performed under denaturing conditions where the ADC species of the light



CONCLUSION A novel anti-Notch 3 cysteine-conjugated ADC indicated for triple-negative breast cancer and ovarian cancers is currently in 292

DOI: 10.1021/acs.oprd.7b00337 Org. Process Res. Dev. 2018, 22, 286−295

Organic Process Research & Development

Article

Purification of 1 Prepared by Process B Using Column Chromatography. The ADC mixture prepared by process B was purified by column chromatography over hydroxyapatite resin (type 1, 40 μm). The following purification parameters were used for separation of aggregates from the monomer species: mobile phase A, 5 mM sodium phosphate (pH 7); mobile phase B, 200 mM sodium phosphate (pH 7); buffer gradient = 10−70% mobile phase B over 20 column volumes; protein loading = 15 mg/mL; flow rate = 1 mL/min; column bed height = 22 cm. Most of the desired monomer product eluted between the beginning of the peak and the next 6.9 column volumes, affording a yield of 85−88% (based on the protein yield) with an average DAR of 3.6 (HIC assay) and an aggregate level of ∼1% (SEC assay). Evaluation of Different Quenching Reagents. The crude mixture of ADC 1 was prepared by process B as described above except that the quenching step was omitted. The reaction mixture was split into four equal-sized portions. The reaction mixtures were quenched with appropriate quenching reagents: L-cysteine, N-acetylcysteine, and 2(dimethylamino)ethanethiol. Each mixture was stirred for 30 min at 25 °C. No quenching reagent was added to one of the portions of the reaction mixture. All four portions of the reaction mixture were purified by a UF−DF process using regenerated cellulose membrane (30 kDa) and histidine buffer as DF buffer at a protein concentration of 40 mg/mL and a feed flux of 500 LMH. General Procedure for the Execution of DoE. The pH of a stock solution of anti-Notch 3 antibody was adjusted to 6.5−7 with a histidine buffer solution. The solution of the antibody (100 mg, 1 equiv) was adjusted to the desired concentration with the histidine buffer. The temperature of the antibody solution was set to the predetermined temperature based on the requirement of each experiment. The aqueous solution of TCEP (2.2 equiv) was added to the antibody solution with stirring. The mixture was stirred with a magnetic stir bar for 90 min and then cooled to 25 °C, and a solution of linker-payload 2 in DMSO (containing the required amount of DMSO for each experiment) was added. The resulting mixture was stirred at 25 °C for 1 h. L-Cysteine (4 equiv) was added, and the mixture was stirred for 30 min. An aliquot from each reaction mixture was taken for evaluation by analytical HIC and SEC assays. Determination of Average Drug Loading (DAR) and Drug Distribution by Analytical HIC. The analytical HIC method was used to determine the drug loading and drug load distribution of the ADC. The reference material (25 μL) and test samples were diluted to 2 mg/mL with diluent (halfstrength mobile phase A) and injected onto a Thermo Propac HIC10 column (4.6 mm × 10 cm, 5 μm). Mobile phase A was 1.5 M ammonium sulfate and 50 mM potassium phosphate dibasic (pH 7), and mobile phase B was 10% isopropyl alcohol and 50 mM potassium phosphate dibasic (pH 7). The samples were eluted at a flow rate of 0.8 mL/min and a column temperature of 30 °C. Species with different DARs were separated using a salt gradient and detected by UV absorbance at 280 nm. The average DAR and distribution profile were determined by the peak area percentage of each species. Determination of Aggregate Content by Analytical SEC. The SEC method was used to determine the product purity and aggregate content. The test samples were diluted with the mobile phase (20 mM sodium phosphate and 400 mM

Figure 16. Analytical HIC chromatograms of 1 purified by column chromatography (blue) and isolated aggregates (black).

phase I clinical trials. The conjugation process for the preparation of this ADC led to the formation of higher aggregate levels in the conjugation reaction mixture. The aggregates were formed in the conjugate mixture only after conjugation with the payload 2 and not during the TCEP reduction step. The formation of aggregates is likely due to a combination of several factors such as the nature of the antibody, the hydrophobicity of the payload, the average drug loading (DAR) on the conjugate, and the sites of drug conjugation. Optimization studies on the reduction/conjugation steps led to the identification of conjugation conditions that resulted in the formation of the ADC mixture with a low level of aggregates. At lower TCEP reduction temperature (0− 5 °C) followed by the conjugation, ADC 1 was obtained with a lower aggregate level. Differences in the TCEP reduction temperature also caused changes in the distribution of ADC species. At a higher temperature (∼37 °C) for the TCEP reduction step, the resulting ADC contained more species conjugated at the hinge region of the antibody. Stability studies by DSC and extrinsic florescence spectroscopy indicated that 1 prepared by process B may be more prone to aggregation at higher temperature compared with the ADC material produced by process A, though at ambient temperature no discernible difference in stability was observed. Separation of aggregates in the crude mixture of 1 (prepared by process B) was accomplished by purification on a hydroxyapatite column under mild conditions without significantly impacting the average DAR of the purified ADC. In addition to aggregates, solvent DMSO and residual free drug species were successfully removed under these purification conditions.



EXPERIMENTAL SECTION General Procedures for the Preparation of ADC 1 by Processes A and B. Process A. The pH of a stock solution of anti-Notch 3 antibody was adjusted to 6.5−7 with a histidine buffer solution. The solution of the antibody was cooled to 0−5 °C. An aqueous solution of TCEP (2.4 equiv/1 equiv of antibody) was added to the antibody solution. The resulting mixture was stirred for 90 min at 0−5 °C. The solution was then cooled to ambient temperature. To this antibody solution was added a solution of 2 (6 equiv) in DMSO (11% v/v). The reaction mixture was stirred for 1 h. The mixture was quenched by addition of a solution of L-cysteine. The resulting unpurified product consisted of an ADC with an average DAR of ∼4 as determined by analytical HIC assay and with aggregate levels of 1−2% as determined by SEC assay. Process B. ADC 1 was prepared in the same manner as in process A except that the temperature of the TCEP reduction step was 37 °C. 293

DOI: 10.1021/acs.oprd.7b00337 Org. Process Res. Dev. 2018, 22, 286−295

Organic Process Research & Development

Article

emission was collected using a ROX filter (600−630 nm). DCVJ stock solutions (4 mM) were prepared in DMSO. Prior to use, the stock solutions were diluted 200-fold to 20 μM in the assay samples. After excitation at 450−490 nm, the fluorescence emission was collected using a FAM filter (515− 530 nm). The ADC solutions were prepared in a BioRad multiplate PCR plate. Each cell was filled with 25 μL of sample solution. The plate was sealed with optically clear adhesive PCR Microseal “B” film. The fluorescence was monitored using a Bio-Rad CFX real-time system with a C1000 thermal cycler. The above plate, plate sealer, and plate reader were purchased from BioRad Laboratories (Hercules, CA).The sample plates were subject to thermal stress from 5 to 95 °C at increments of 1 °C. At each temperature, the plate was equilibrated for 1 min prior to measurement. The fluorescence data and the second derivatives obtained from the built-in software were exported as a CSV file into Microsoft Excel (Microsoft Corporation, Redmond, WA) for further data analysis. Stability Assessment of 1 by Differential Scanning Calorimetry. Samples were diluted with degassed buffer to a final protein concentration of 0.5 mg/mL for all runs. The same degassed buffer was also used as a reference in the reference cell. Samples were injected into a MicroCal VP capillary differential scanning calorimeter (Malvern Instruments, Worcestershire, UK). DSC scans were performed from 25 to 95 °C at a rate of 60 °C/h. The data were corrected for buffer, and the raw signal was converted to molar heat capacity with protein concentrations and theoretical molecular weight.

sodium chloride, pH 7.2) at a concentration of 2 mg/mL and injected onto the YMC-Pack Diol-200 (300 mm × 8 mm i.d.) column at 0.75 mL/min for 20 min. The aggregate and monomer species in the crude ADC preparation were reported as the percentages of the total area for all protein-related peaks. Determination of Average Drug Loading (DAR) by Reduced RP-HPLC. The sample was denatured in 4 M GdnHCl, 50 mM Tris (pH 7.8), and 20 mM DTT at 37 °C for about 30 min. The reduced samples (10 μL) were separated and monitored at 214 nm using an Agilent HPLC with an Agilent Zorbax 300SB-CN column (150 mm × 4.6 mm, 3.5 μm) at a column temperature of 75 °C. The samples were eluted using mobile phase A (0.1% TFA in water) and mobile phase B (80% acetonitrile, 20% isopropanol, and 0.1% TFA) at a flow rate of 0.75 mL/min. Characterization of ADC 1 by LC−MS and Peptide Mapping. Intact Mass Analysis by Organic SEC/MS. The ADC 1 samples and HIC fractions were analyzed by organic SEC interfaced with mass spectrometry. Samples were transferred into HPLC vials and placed in a 6 °C autosampler. Approximately 20 μg of sample was applied to a Waters Acquity UPLC BEH200 SEC column (4.6 mm × 150 mm, 1.7 μm) at a column temperature of 25 °C. The samples were eluted at a flow rate of 0.2 mL/min using an isocratic separation of 70% water and 30% acetonitrile with 0.1% trifluoroacetic acid (TFA) and monitored at 280 nm with a tunable UV detector. The Waters Acquity UPLC was coupled to a Waters Xevo Q-Tof mass spectrometer (Waters, Milford, MA) that was calibrated by infusion of 2 mg/mL NaI solution at 5 μL/min. LC−MS of Nonreduced and Alkylated Lys-C Peptides. For the nonreduced and alkylated sample preparation, 100 μg of protein sample was denatured by addition of 100 μL of the buffer containing 7 M Gnd-HCl, 0.1 M Tris (pH 7.5), and 80 μL of water. The alkylation step was performed at room temperature for 1 h by addition of 3 μL of 50 mM iodoacetic acid (IAA). The Lys-C digestion reaction was prepared by addition of 10 μL of 1 mg/mL (10 μg) of endoproteinase LysC (Wako Chemicals) [enzyme:substrate = 1:10 w/w] and then incubating at 37 °C for 5 h. Prior to analyses, samples were diluted with 90 μL of 0.1% TFA in water for a final concentration of 0.33 mg/mL. The digested sample was applied to a Vydac 214TP C4 column (5 μm particle size, 300 Å pore size, 2.1 mm × 250 mm) set at 60 °C and eluted using an Agilent 1200 HPLC. A linear gradient from 0 to 45% mobile phase B (0.1% TFA in acetonitrile) in mobile phase A (0.1% TFA in H2O) over 116 min at a flow rate of 0.2 mL/min was used to elute the peptides. Peptides were monitored at 214 nm. The Agilent 1200 HPLC was coupled to a Thermo LTQ Orbitrap XL mass spectrometer (Thermo Scientific, San Jose, CA) that utilized an internal lock mass ion of hexakis(1H,1H,3Hperfluoropropoxy)phosphazene at m/z 922.009798 for [M + H]+ for dynamic calibration. The observed multiply charged peptides were converted to neutral molecular masses using the Xtract algorithm of Xcalibur software. Neutral, monoisotopic masses were reported for all of the peptides observed in the LC−MS data. Stability Evaluation of 1 by Extrinsic Fluorescence Spectroscopy. SYPRO Orange and DCVJ were used to monitor thermal unfolding and formation of aggregates. SYPRO Orange dye was supplied in 5000-fold concentrated stock solutions. The dye was diluted 2000-fold in the assay samples. With excitation at 560−590 nm, the fluorescence



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.oprd.7b00337. Details of experimental design and results from DoE evaluation, analysis of variance (ANOVA) summary for the level of aggregates, data for the characterization of individual ADC species by native SEC/MS, and characterization data for ADC 1 by nonreduced CGE and reduced RP-HPLC (PDF)



AUTHOR INFORMATION

Corresponding Author

*Phone: 845-602-4505. Fax: 845-474-3359. E-mail: durgesh. nadkarni@pfizer.com. ORCID

Durgesh V. Nadkarni: 0000-0003-0733-4380 Present Address ⊥

H.M.: Sanofi Genzyme, 1 The Mountain Road, Framingham, Massachusetts 01701, United States. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors thank Manoj Charati, Nadira Prashad, Frank Loganzo, Elwira Muszynska, Kiran Khandke, and Ken Geles for providing details of early development work and Heyi Li, Leo Letendre, Bo Arve, Aparna Deora, Margaret Ruesch, and Brad Evans for supporting this work and providing valuable comments and suggestions. 294

DOI: 10.1021/acs.oprd.7b00337 Org. Process Res. Dev. 2018, 22, 286−295

Organic Process Research & Development



Article

ABBREVIATIONS ADC, antibody−drug conjugate; CGE, capillary gel electrophoresis; LC−MS, liquid chromatography−mass spectroscopy; MW, molecular weight; SEC, size-exclusion chromatography; TFA, trifluoroacetic acid; TCEP, tris(2-carboxyethyl)phosphine; UF−DF, ultrafiltration−diafiltration



REFERENCES

(1) Gerber, H. P.; Senter, P. D.; Grewal, I. S. mAbs 2009, 1, 247− 253. (2) Senter, P. D. Curr. Opin. Chem. Biol. 2009, 13, 235−244. (3) Antibody−Drug Conjugates: Fundamentals, Drug Development, and Clinical Outcomes to Target Cancer; Olivier, K. J., Hurvitz, S. A., Eds.; John Wiley and Sons: Hoboken, NJ, 2016. (4) Geles, K. G.; Gao, Y.; Spara, P.; Tchitstiakova, L. G.; Zhou, S. B. (Pfizer Inc., New York, NY). U.S. Patent 0127211A1, 2014. (5) Maderna, A.; Leverett, C. A. Recent advances in the development of new auristatins. Mol. Pharmaceutics 2015, 12, 1798−1812. (6) Hollander, I.; Kunz, A.; Hamann, P. R. Bioconjugate Chem. 2008, 19, 358−361. (7) (a) Li, W.; Prabakaran, P.; Chen, W.; Zhu, Z.; Feng, Y.; Dimitrov, D. S. Antibodies 2016, 5, 19. (b) Adem, Y. T.; Schwarz, K. A.; Duenas, E.; Patapoff, T. W.; Galush, W. J.; Esue, O. Bioconjugate Chem. 2014, 25, 656−664. (8) Guo, J.; Kumar, S.; Chipley, M.; Marcq, O.; Gupta, D.; Jin, Z.; Tomar, D. S.; Swabowski, C.; Smith, J.; Starkey, J. A.; Singh, S. K. Bioconjugate Chem. 2016, 27, 604−615. (9) Doronina, S. O.; Toki, B. E.; Torgov, M. Y.; Mendelsohn, B. A.; Cerveny, C. G.; Chace, D. F.; DeBlanc, R. L.; Gearing, R. P.; Bovee, T. D.; Siegall, C. B.; Francisco, J. A.; Wahl, A. F.; Meyer, D. L.; Senter, P. D. Nat. Biotechnol. 2003, 21, 778−784. (10) Huang, R. Y.; Chen, G. Drug Discovery Today 2016, 21, 850− 855. (11) Wakankar, A.; Chen, Y.; Gokarn, Y.; Jacobson, F. S. mAbs 2011, 3, 161−172. (12) Ross, P. L.; Wolfe, J. L. J. Pharm. Sci. 2016, 105, 391−397. (13) He, F.; Hogan, S.; Latypov, R. F.; Narhi, L. O.; Razinkov, V. I. J. Pharm. Sci. 2010, 99, 1707−1720. (14) Kung, C. E.; Reed, J. K. Biochemistry 1989, 28, 6678−6686. (15) Lindgren, M.; Sorgjerd, K.; Hammarstrom, P. Biophys. J. 2005, 88, 4200−4212. (16) Ablinger, E.; Leitgeb, S.; Zimmer, A. Int. J. Pharm. 2013, 441, 255−260.

295

DOI: 10.1021/acs.oprd.7b00337 Org. Process Res. Dev. 2018, 22, 286−295