1588

Bioconjugate Chem. 2010, 21, 15881595

Physicochemical Stability of the Antibody-Drug Conjugate Trastuzumab-DM1: Changes due to Modication and Conjugation Processes
Aditya A. Wakankar,*, Maria B. Feeney, Javier Rivera, Yan Chen,| Michael Kim,| Vikas K. Sharma, and Y. John Wang
Late Stage Pharmaceutical and Processing Development, Early Stage Pharmaceutical Development, and Protein and Analytical Chemistry, Genentech Inc., 1 DNA Way, South San Francisco, California 94080, and Department of Pharmaceutical Chemistry, The University of Kansas, 2095 Constant Avenue, Lawrence, Kansas 66047. Received October 5, 2009;
Revised Manuscript Received July 19, 2010

In the manufacture of the antibody-drug conjugate Trastuzumab-DM1 (T-DM1), the lysine residues on the antibody trastuzumab (Tmab) are modied to form the intermediate Tmab-MCC (T-MCC) and then conjugated with the drug DM1. Our goal is to understand the effects of modication and conjugation steps on the physicochemical stability of the antibody. The structural stability of Tmab relative to its modied and conjugated forms was assessed, employing thermally induced stress conditions to formulations containing Tmab, T-MCC, and T-DM1. DSC, SEC, CE-SDS, and LC-MS were used to study the stability of Tmab, T-MCC, and T-DM1 to thermal stress. The DSC thermograms show a decrease in melting temperature for the CH2 transition, in the order Tmab > T-MCC > T-DM1. As per SEC analysis, a signicant increase in level of aggregation was detected in T-MCC (32%) and T-DM1 (5%) after 14 days at 40 C. Tmab did not show signicant aggregate formation. CE-SDS and LC-MS data demonstrate that the aggregation in the case of T-MCC is largely covalent and involves mechanisms other than formation of intermolecular disulde cross-links. The aggregation observed for T-MCC was signicantly inhibited upon addition of amino acids with nucleophilic side chains containing thiol (Cys) and hydroxyl moieties (Ser, Tyr). The covalent aggregation observed for T-MCC and the ability of nucleophilic amino acids, particularly Cys, to inhibit it indicate that the maleimide moiety in the MCC linker may react to form intermolecular covalent cross-links between T-MCC molecules, possibly through a Michael addition mechanism. In addition, DSC results demonstrate that the conjugation of the drug moiety DM1 to Tmab results in destabilization of the CH2 domain of the antibody.

INTRODUCTION
Antibody-drug conjugates, or immunoconjugates, are a class of chemotherapeutic agents that harness the site specicity of monoclonal antibodies to achieve targeted delivery of cytotoxic drug moieties (1, 2). On the basis of the targeted delivery of these cytotoxins, these antibody-drug conjugates selectively eliminate the cancer cells that express the surface antigen (3-6). Antibody-maytansine conjugates are a subclass of these immunoconjugates that provide for selective delivery of the cytotoxin maytansine to the tumor cells. Maytansines, rst isolated from the East African shrub Maytenus serrata by Kupchan and co-workers, are members of the ansamycin group of natural products (7, 8). DM1, a semisynthetic ansamacrolide derived from the naturally occurring compound maytansine, exhibits its cytotoxicity by inhibiting microtubule polymerization, thereby blocking cell division (9, 10). Several antibodymaytansinoid conjugates involving the maytansine derivative, DM1, have been evaluated in clinical trials. Examples of these include bivatuzumab mertansine (7), huN901-DM1, BB-10901, MLN-DM1 (3, 7), and DM1-huJ591 (11, 12), which are evaluated in clinical trials to treat different forms of cancers, including small-cell lung cancer and prostate cancer.
* wakankar.aditya@gene.com. Late Stage Pharmaceutical and Processing Development, Genentech Inc. The University of Kansas. Early Stage Pharmaceutical Development, Genentech Inc. | Protein and Analytical Chemistry, Genentech Inc.

Trastuzumab-DM1, or T-DM1, consists of the therapeutic anti-HER2 monoclonal antibody trastuzumab (Herceptin) covalently linked to DM1 via a linker moiety to achieve targeted delivery of the cytotoxic agent to tumor cells overexpressing HER2 (3, 10, 13, 14). This antibody-drug conjugate is currently being evaluated in clinical trials for treatment of HER2 refractory metastatic breast cancer (13, 15). Scheme 1 illustrates the synthesis involved in the manufacture of the antibody-drug conjugate T-DM1. The manufacture of T-DM1 consists of a two-step process in which the lysine residues on the antibody trastuzumab (Tmab) are rst modied using the linker species SMCC to form the intermediate TmabMCC (T-MCC). In the second step, T-MCC is reacted with the drug DM1 to obtain the antibody-drug conjugate T-DM1. This conjugation scheme results in a heterogeneous distribution having an average of 3.5 DM1 molecules per molecule of the antibody Tmab (16). The heterogeneity of T-DM1 arises due to the potential number of lysines (total 88) on Tmab that could be linked to DM1, as well as the variability in the particular sites of conjugation. Previous studies with antibody-drug conjugates have provided insight into the distribution, load, and location of conjugation sites on these antibodies (9, 12, 17-19). However, the effect of these modication and conjugation reactions on the physicochemical stability of the antibody has not been addressed. As shown in the synthetic scheme for T-DM1, the manufacturing process results in neutralization of the surface charge on the lysine residues that are modied and conjugated, as well as addition of a hydrophobic DM1 moiety to the surface of the antibody molecule. We hypothesize that the addition of

10.1021/bc900434c 2010 American Chemical Society Published on Web 08/10/2010

Physicochemical Stability of Trastuzumab-DM1

Scheme 1. Process Flow for Production of T-DM1 from Tmab, via the Key Intermediate T-MCC

Bioconjugate Chem., Vol. 21, No. 9, 2010 1589

the linker and drug moieties during processing, as well as the hydrophobicity of the drug DM1 itself, could lead to structural changes on the antibody molecule that affect its physicochemical properties. In our studies, we employed thermal stress using a differential scanning calorimetry (DSC) temperature scan to assess the changes in structural stability of Tmab due to modication and conjugation reactions. The thermal stability of Tmab has been investigated previously using DSC (20). In this study, the stability of Tmab (an IgG1) and its modied (TMCC) and conjugated forms (T-DM1) to thermally induced stress was analyzed using DSC, SEC, CE-SDS, and LC-MS. This information could be important for selecting appropriate processing, storage, and handling conditions for critical intermediates during manufacture of antibody-drug conjugates such as T-DM1.

MATERIALS AND METHODS

Materials. Trastuzumab (Tmab), a humanized IgG1 monoclonal antibody, was manufactured at Genentech, South San Francisco. The modied (T-MCC) and conjugated (T-DM1) forms of Tmab were obtained from an early developmental manufacturing run. The amino acids used in the blocking studies had their N-termini capped as an N-acetyl derivative. The amino acids used were obtained from different sources: cysteine and serine from Sigma Aldrich (St. Louis, MO), histidine from VWR (West Chester, PA), and lysine and tyrosine from Fisher Scientic. (Pittsburgh, PA). Formulation Preparation. Tmab, T-MCC, and T-DM1 solutions were dialyzed into 10 mM sodium succinate buffer (pH 5) and then concentrated using tangential ow ltration (TFF). After measuring the protein concentration of this TFF pool by UV absorption, 10 mM sodium succinate buffer (pH 5) was added to adjust to a target protein concentration of 24

mg/mL. To ve parts of this solution was added one part of a conditioning buffer, containing 10 mM sodium succinate, 36% (w/v) trehalose dihydrate, and 0.12% (v/v) polysorbate 20 (pH 5). The nal formulation thus consists of 20 mg/mL antibody in 10 mM sodium succinate buffer (pH 5), containing 6% w/v trehalose dihydrate and 0.02% w/v polysorbate 20. The nal formulation was ltered through a 0.2 m lter prior to using it for our stability investigations. Accelerated Stability Studies. The formulations were prepared as aliquots of either 200 or 60 L per vial, depending on the study, in 3 mL Foma Vitrum glass vials, covered with 13 mm Daikyo D777-1 Fluoro Tec coated stoppers, and then sealed with Paralm to prevent evaporation of solution during incubation. The glass vials and rubber stoppers were autoclaved prior to use. Duplicate vials were prepared for each sample time point and then placed in an incubator room held at 40 C. The vials were sampled at each predetermined time point and then stored frozen at -70 C prior to analyses. Formulation with Selected Amino Acids. Using a stock solution of amino acids in deionized water, each amino acid was added to a formulation to achieve an amino acid concentration of 3 mM in the nal protein formulations. This concentration corresponds to a 5-fold molar excess of the amino acid over the linker moiety, MCC. The concentration of the linker moiety was calculated based on the average antibody/linker ratio of 4.5, determined using LC-MS analysis (internal document) of the development lot. In separate sample sets, N-Ac-Cys was tested using a 2-fold molar excess of this amino acid, and N-Ac-Ser was tested using a 20-fold molar excess over the linker moiety. Accelerated stability studies of the formulation samples spiked with amino acids were conducted as described under the section accelerated stability studies. Differential Scanning Calorimetry (DSC). Thermal analysis of Tmab, T-MCC, and T-DM1 was performed using highthroughput VP Capillary DSC (MicroCal; Northampton, MA). Formulations of Tmab, T-MCC, and T-DM1 were diluted to give a nal protein concentration of 1 mg/mL by adding 475 L formulation buffer to 25 L of sample in a 96-well plate, alternating with wells containing 500 L formulation buffer. The instrument scanned each sample-buffer pair over the temperature range 15-95 C at a rate of 1 C/min. In the case of experiments to evaluate the reversibility of the melting transitions, samples were scanned twice at the rate of 1 C/min in the temperature range 15-70 C. Data analysis was performed using Origin software (OriginLab; Northampton, MA). Size-Exclusion Chromatography (SEC). TSKgel G3000SWXL 7.8 300 mm size-exclusion column from Tosoh Biosciences (So. San Francisco, CA) was connected to an Agilent 1100 series HPLC system (Agilent, Santa Clara, CA). The mobile phase consisted of 0.2 M potassium phosphate buffer (pH 6.8), 0.2 M potassium chloride, and 15% isopropyl alcohol and was pumped at a ow rate of 0.5 mL/min. 5 L of sample volume was injected for analysis. The column was maintained at 25 C using a built-in thermostat. Detection was performed by monitoring UV absorbance at 280 nm. On-Chip Capillary Gel Electrophoresis (Nonreduced Conditions). Agilent 2100 Bioanalyzer and Agilent Protein 230 Kit were purchased from Agilent. Formulations of Tmab, T-MCC, and T-DM1 were diluted with water to a nal protein concentration of 1 mg/mL. Gel-dye solution, destaining solution, sample buffer, and ladder kit reagents were prepared as per the protocol (internal document). To 24 L of sample buffer was added 4 L of sample and 2 L of denaturing solution. Samples and ladder were heated to 70 C for 5 min in a temperature-controlled water bath, then cooled and diluted with 60 L water. The chip was primed as instructed by the manufacturer, and 6 L of each sample was loaded onto a well

1590 Bioconjugate Chem., Vol. 21, No. 9, 2010

Wakankar et al.

on the chip. The chip was placed into the Bioanalyzer, and electrophoresis with uorescence detection was performed using the manufacturers preprogrammed assay parameters in the software. Peaks were manually integrated, and relative peak areas were used to calculate percentages for aggregate, monomer, and fragment. On-Chip Capillary Gel Electrophoresis (Reduced Conditions). The on-chip capillary gel electrophoresis was conducted using the same instrument as described in the protocol above. The sample preparation involved addition of 4 L of sample (protein concentration 1 mg/mL), 2 L of denaturing solution, and 4 L of -mercaptoethanol (BME) to 20 L of 0.5% SDS solution. The sample treatment, priming, and analysis was identical to that described in the previous section. LC/MS Analysis of Deglycosylated and Reduced TMCC Samples. A PE-Sciex QSTAR Pulsar mass spectrometer (Applied Biosystems, Foster City, CA) was used for proteinmass measurement. Samples were analyzed in positive ion mode. The capillary and cone voltages were optimized for maximum sensitivity. The optimized instrument conditions included applying an electrospray voltage of 5200 V, a rst and second declustering potential (DP) of 60 V (DP1) and 15 V (DP2), and a focusing potential of 200 V. MS data were acquired and deconvoluted using Analyst QS 1.1 software. The mass spectrometer was calibrated using an aqueous solution containing 2 mg/mL NaI, 0.05 mg/mL CsI, and 30% acetonitrile for acquiring mass ranges of 650 to 2000 m/z for protein mass measurement. Deglycosylation was performed by incubating 200 g of T-MCC samples with 2 L of PNGase F (New England Biolabs, Ipswich, MA) at 37 C overnight. Deglycosylated T-MCC samples were reduced with 20 mg/mL tris(2-carboxy-ethyl) phosphine (TCEP) at 37 C for 60 min. Approximately 15 g reduced, deglycosylated sample was loaded onto a PLRP-S 2.1 150 mm, reverse-phase HPLC column (Varian Inc., Palo Alto, CA) equilibrated at 78 C and coupled with QSTAR mass spectrometer. The column was eluted with a ow rate of 0.25 mL/min using a gradient of acetonitrile in water as follows: 15% acetonitrile in water for 3 min, then 15-47% acetonitrile in 37 min, 47-95% in 0.5 min, equilibration at 95% for 2 min, and back to 15% acetonitrile in 0.5 min and equilibration at 15% for 7.5 min. The HPLC ow was diverted to waste during the rst 4.5 min of the gradient to avoid salt interference.

Figure 1. Comparison of the DSC thermograms of Tmab, T-MCC, and T-DM1 ([antibody] ) 1 mg/mL) in 10 mM sodium succinate, 6% w/v trehalose dihydrate, and 0.02% w/v polysorbate 20. Inset in the gure depicts reversibility of the rst (CH2) transition for Tmab, T-MCC, and T-DM1.

RESULTS
DSC Analysis. The thermal stability of Tmab, T-MCC, and T-DM1 was assessed using DSC (Figure 1). Tmab and its modied and conjugated forms were analyzed in solutions containing the same formulation components with the antibody concentration at 1 mg/mL. The DSC thermograms for these three molecules show two major transitions. The second DSC transition did show differences in melting temperatures (tm) for the three forms investigated. A slight decrease (0.8 C) in tm was observed following modication and conjugation of Tmab, giving the following order for tm values: Tmab > T-MCC > T-DM1. The rst transition showed a signicantly greater difference in the thermal stability of Tmab and its modied and conjugated forms. The rst transition for Tmab occurred at 68.2 C, whereas the same transition for T-MCC and T-DM1 occurred at 66.2 and 63.8 C, respectively. The reversibility of the rst transition was also evaluated for Tmab and its modied and conjugated forms. This calorimetric reversibility typically indicates that the transition from the native to the unfolded state is reversible and that the unfolded state does not undergo any subsequent reactions such as aggregation. The implication of reversibility in DSC is that the unfolded state will return to the native state upon cooling (20). For testing

Figure 2. SEC chromatograms showing elution proles of Tmab, T-MCC, and T-DM1 after 7 days of storage at 40 C. Each individual SEC elution prole shows the main peak (Peak 1), the dimer peak (Peak 2), and the high molecular weight aggregate peak (Peak 2).

reversibility, the samples were rescanned following the initial sample scan in temperature range 15-70 C. The trace of the DSC thermograms of the initial sample scan and the subsequent rescan were assessed for superimposability, as shown in the inset on Figure 1. On the basis of our results, the rst DSC transition for Tmab and T-DM1 was reversible. However, in the case of T-MCC, the rst transition was observed to be partially reversible. Effect of Modication and Conjugation on Aggregation Behavior. SEC Analysis. The aggregation behavior of Tmab and its modied (T-MCC) and conjugated (T-DM1) forms was assessed using size exclusion chromatography (SEC). The changes in the SEC prole following storage for 7 days at 40 C are shown in Figure 2. Each chromatographic prole shows monomer species (Peak 1), dimer species (Peak 2), and highermolecular-weight aggregate species (Peak 2). No signicant level of fragmentation was detected under conditions of the assay. The aggregation kinetics for Tmab, T-MCC, and T-DM1 were followed for up to 70 days at 40 C using the SEC assay (Figure 3). T-MCC showed an increase of up to 40% aggregate

Physicochemical Stability of Trastuzumab-DM1

Bioconjugate Chem., Vol. 21, No. 9, 2010 1591

Figure 3. Time course displaying the formation of the aggregates, as per SEC analysis, in T-MCC (b), T-DM1 (2), and Tmab (9) samples incubated at 40 C over 70 day time period.

formation at the end of the 70day incubation period, whereas Tmab and T-DM1 showed 0% and 11% increase, respectively. Also, the aggregation kinetics for T-MCC exhibited a biphasic increase: a fast initial aggregation rate followed by a much slower rate of aggregation. The majority of the aggregate formation occurred during the rst 14 days of storage at 40 C. On-Chip Nonreduced and Reduced CE-SDS Analysis. The aggregate formation in Tmab, T-MCC, and T-DM1 samples after storage at 40 C for 14 days was further analyzed using nonreduced and reduced CE-SDS. The nonreduced CE-SDS prole of T-MCC showed the presence of covalent aggregates corresponding to molecular weights of 260, 330, and 390 kDa (Figure 4a). The nonreduced CE-SDS prole of T-DM1 upon incubation was similar to T-MCC. However, the total peak area contribution from the covalent aggregates in T-DM1 was 3% compared to 39% observed in the case of T-MCC. Aggregates were not detected in Tmab sample incubated at 40 C (data not shown).

To assess whether the covalent aggregation observed in T-MCC was due to formation of interchain disuldes, on-chip CE-SDS was performed under reducing conditions. As shown in Figure 4b, the reduced CE-SDS prole of T-MCC shows two main peaks corresponding to the heavy chain (molecular weight 60 kDa) and light chain (molecular weight 20 kDa). These peaks were the only two peaks observed in the reduced CE-SDS prole of Tmab (data not shown). However, in case of T-MCC in addition to these two main peaks, several other peaks corresponding to molecular weights of 85, 120, 140, and 160 kDa were observed. On the basis of the calculated molecular weights, these additional peaks corresponded to fractions such as HL, HH, HHL, and HHLL. These peaks contributed to 45% of the total peak area. The detection of these peaks suggests that, in the case of T-MCC, the formation of interchain cross-links involves mechanisms other than disulde bond formation. The reduced CE-SDS prole of T-DM1 sample was similar to T-MCC; however, the peaks corresponding to interchain cross-links contributed to less than 10% of the total peak area (data not shown). The reduced T-MCC samples were also analyzed using LCMS. The most intense ions in the deconvoluted mass spectra of the reduced T-MCC samples correspond to light chain and heavy chain related fragments as expected (data not shown). As shown in Figure 5, minor species with m/z of 72 822 and 73 041 Da were also observed. The theoretical and experimental molecular weights for cross-linked species comprising light chain and heavy chain linked through one and two MCC linkers are shown in Table 1. Our results show agreement between the experimental molecular weights and the theoretical molecular weights. The total extents of aggregation, both covalent and noncovalent, for Tmab, T-MCC, and T-DM1 were compared after 14 days of storage at 40 C using results obtained from SEC and nonreduced CE-SDS analysis (Figure 6). For T-MCC, a majority of the aggregation observed was covalent in nature. However, for T-DM1, the majority of the aggregation observed was noncovalent in nature (Figure 6). No signicant level of aggregation, either covalent or noncovalent, was observed for Tmab under the conditions of incubation. A comparison of the results for T-DM1 with T-MCC demonstrates that the extent

Figure 4. Electropherograms showing the elution prole of T-MCC upon incubation for up to 14 days at 40 C. Panel A is the elution prole of T-MCC (0 and 14 days) analyzed using on-chip CE-SDS under nonreduced conditions, whereas Panel B is the elution prole of T-MCC analyzed by on-chip CE-SDS under reduced conditions. The labels on Panel B correspond to the heavy chain (H) and light chain (L) moieties and also the combinations of these H and L moieties that were observed.

1592 Bioconjugate Chem., Vol. 21, No. 9, 2010

Wakankar et al.

Figure 5. Deconvoluted mass graph of T-MCC showing cross-linked heavy chain-light chain (HL) with one linker and two linkers.
Table 1. Masses of Reduced Fragments, Heavy Chain (H) and Light Chain (L), and Cross-Linked Heavy Chain-Light Chain (HL) Observed from LC-MS along with Their Respective Theoretical Masses L theoretical experimental 23443 23443 H 49157 49156 HL+1 Linker 72819 72822 HL+2 Linkers 73038 73041

Figure 7. Effect of addition of N-acetylated amino acids Cys (b), Ser (2), Tyr (1), Lys ([), and His (+) on aggregate formation observed in the control (9) formulation. The control formulation consisted of 20 mg/mL T-MCC, 10 mM succinate (pH 5.0), 6% w/v trehalose dihydrate, and 0.02% w/v polysorbate 20. The concentration of the N-acetyl amino acids was maintained at 3 mM, and the different formulation samples were incubated at 40 C for up to 11 days.

of aggregation is substantially inhibited when T-MCC is conjugated with the drug DM1 (Figure 6). Effect of Blocking Agents on Aggregation of T-MCC. On the basis of this result, it was hypothesized that maleimide species in T-MCC could be responsible for covalent aggregation. Our goal was to assess whether the aggregation behavior of T-MCC could be inhibited using agents that react with the free maleimidyl species. N-acetyl derivatives of amino acids such as Cys, Lys, His, Ser, and Thr, each with nucleophilic sidechains, were added to the formulation of T-MCC to investigate their ability to react with the maleimide moiety of the T-MCC species. Tmab has very low levels of free Cys residues 0.2 mol/mol of Tmab (internal document), as these residues are involved in formation of both inter- and intrachain disulde linkages. On the other hand, nucleophilic residues such as Lys, His, Ser, and Thr are abundant in Tmab and could also potentially interact with the maleimide moiety of the linker.

T-MCC and its formulations with the different amino acid residues were incubated for a period of 11 days at 40 C. The samples were then assessed using SEC analysis. Our results shown in Figure 7 demonstrate 25% increase in aggregation in the control sample of T-MCC, with no amino acids added, following storage for up to 10 days at 40 C. No signicant level of aggregate formation was detected in formulations containing 3 mM Cys as an additive. Also, formulations containing 3 mM Ser or Tyr greatly reduced the extent of aggregation observed relative to the control. The total amount of aggregation observed for T-MCC in the presence of 3 mM His and Lys was comparable to the control. To determine if lower amounts of Cys could result in a similar inhibitory effect on T-MCC aggregation, a formulation containing a 2-fold excess of Cys over the MCC moiety in T-MCC was prepared. This formulation was stored at 40 C and the stability assessed over a 10 day period. On the basis of our results, the inhibitory effect on T-MCC was comparable in formulations containing a 2-fold and a 5-fold molar excess of Cys. Also, formulations containing a 20-fold molar excess of Ser showed one-third the amount of aggregation observed in solutions containing a 5-fold excess (data not shown). To further conrm the results obtained using SEC analysis, the T-MCC formulation samples containing no added amino acid (control), Cys, Ser, and Tyr incubated at 40 C for 11 days were reduced and analyzed using LC-MS assay (Figure 8). Figure 8 shows the relative peak areas of the cross-linked HL species for the control and samples containing the different amino acid additives. Both the addition of Ser and Tyr leads to reduction in the HL peak area, whereas no HL species was observed in incubated samples containing Cys.

DISCUSSION
In the manufacturing scheme for T-DM1, the lysine residues on the antibody Tmab are rst modied via reaction with SMCC to obtain the intermediate T-MCC (Scheme 1). This intermediate consists of lysine residues on the antibody covalently bonded to the linker moiety MCC through an amide bond. In the subsequent conjugation reaction, the free thiol of the drug moiety DM1 reacts with the maleimide of the intermediate T-MCC via a Michael addition mechanism to produce the antibody-drug conjugate T-DM1. This process, a modication of the surface lysine residues, results in neutralization of the positive charge

Figure 6. Comparison of the aggregate formation in formulation samples of Tmab, T-MCC, and T-DM1 incubated for 14 days in 40 C based on SEC (black bars) and nonreduced CE-SDS data (white bars).

Physicochemical Stability of Trastuzumab-DM1

Bioconjugate Chem., Vol. 21, No. 9, 2010 1593

Figure 8. Differences in levels of cross-linked species present in thermally stressed samples of T-MCC and T-MCC treated with N-acetylated amino acids Ser (NAS), Tyr (NAY), and Cys (NAC). Relative levels of cross-linking were quantied from deconvoluted mass spectra.

on these residues as well as addition of a large (molecular weight 959 Da) hydrophobic moiety on the antibody surface. The conjugated antibody T-DM1 that results from this processing is a heterogeneous species, with regard to both the number of lysine residues that are conjugated as well as the location of these conjugation sites. The average drug to antibody ratio of T-DM1 is 3.5 having anywhere from 1 to 7 lysine residues modied. Our studies assess the impact of these modication and conjugation reactions on the physicochemical attributes of the antibody. To assess the structural stability of Tmab and its modied and conjugated forms, the three antibody forms were subjected to temperature-induced unfolding using DSC. The DSC prole for Tmab, the parent antibody, shows two transitions: the rst transition at 68.2 C and the second transition at 82.2 C (Figure 1). Using DSC, Ionescu et al. studied the temperature-induced unfolding of three humanized IgG1 molecules and their Fab and Fc fragments (20). Their studies also included an investigation of the effect of variable domains present in IgG1s on the structural stability of the antibody. On the basis of the measured enthalpy of unfolding, it was proposed that, if the variable domain sequences in the IgG1 do not affect the thermal stability of the Fab fragment, then the rst transition in the DSC thermogram represents the melting of the conserved heavy chain 2 (CH2) domain and the second transition represents melting of the CH3 domain and the Fab fragment. Tmab was one of three IgG1s investigated in their research study, and it was demonstrated that the rst transition, which has a lower experimental enthalpy, represents the unfolding of the CH2 domain, while the second transition, with a higher experimental enthalpy, represents melting of the CH3 domain and the Fab region. Our results for Tmab were consistent with their observations. The DSC proles of T-MCC and T-DM1, similar to Tmab, also showed two transitions (Figure 1). A comparison of the DSC proles of the three antibody forms showed a decrease in tm for both the rst and second transitions following modication and conjugation processes with the following order: Tmab > T-MCC > T-DM1. The decrease in tm observed for the rst transition of Tmab following modication and conjugation processes was 4.4 C, whereas the decrease in tm observed for the second transition was only 0.8 C. Also, the CH2 domain unfolding was observed to be a partially reversible transition in case of T-MCC. These results suggest that the modication and conjugation of Tmab has a signicantly greater impact on the thermal stability of the CH2 domain than on the rest of the antibody.

Comparison to huN901-DM1, another humanized IgG1 antibody-maytansinoid conjugate that utilizes a similar linker chemistry to covalently attach the drug species DM1 to lysine residues on the antibody surface, provides some insight to the results presented here for T-DM1. The structural characterization of huN901-DM1 was performed by Wang et al. using mass spectrometry (9). Their studies identied 40 different conjugation sites on the IgG1 antibody, out of a total of 86 lysine residues that could possibly be covalently attached to the DM1 species. A majority of these conjugation sites were observed to be in the constant heavy chain and light chain regions of the antibody. Their investigations demonstrated that the drug molecules can distribute over 47% of the available lysine residues, with the abundance of drug molecules at each particular site being relatively low. Out of the 40 conjugation sites that were identied, 12 were identied in the CH2 domain. The X-ray crystal structure of the CH2 domain of an IgG1 molecule (PDB ID: 1L6X) structure reveals that the Lys residues in the CH2 domain of the IgG1 show a high degree of solvent accessibility (structure not shown). The atomic displacement parameter, the B values, for the Lys side chains were computed based on the X-ray crystal structure of the IgG1. The B values were used to determine the local exibilities of the lysine side chains. The values for the lysine side chain residues indicate that these lysine residues located in the CH2 domain are associated with more exible regions of the IgG1 molecule (9). Both the surface accessibility and the exibility are factors that could lead to greater reactivity of the CH2 Lys residues. The decrease in the structural stability observed, as per DSC, for the CH2 domain of Tmab after modication and conjugation could be attributed to the density of Lys residues present in the CH2 region as well as their propensity to react with the linker and drug species. To further assess the biophysical stability of Tmab, T-MCC, and T-DM1, the three antibody forms were incubated at a stress condition of 40 C over approximately 70 days (Figure 3). The major degradants identied following temperature stress at 40 C using SEC were the dimer species and some highermolecular-weight aggregate species (Figure 2). With regard to formation of these degradants, the following order of stability to temperature stress was observed among the three antibody forms investigated: Tmab > T-DM1 > T-MCC. Recall that the structural stability predicted as per DSC was as follows: Tmab > T-MCC > T-DM1. Thus, the order of susceptibility of the three antibody forms to aggregate formation is not consistent with the order of thermal stability predicted as per DSC. Also, the aggregation behavior of T-MCC demonstrated a biphasic prole with a faster aggregation rate observed during the initial time period of 14 days (Figure 3), followed by a much slower aggregation rate upon prolonged storage. The unique aggregation behavior observed for T-MCC may be indicative of mechanisms other than the typical hydrophobic interactions that play a role in its degradation. To further investigate the aggregation phenomenon observed in T-MCC, formulation samples subjected to temperature stress at 40 C were analyzed using on-chip CE-SDS under nonreduced and reduced conditions (Figure 4a,b), as were samples of Tmab and T-DM1 exposed to the same thermal stress (electropherograms not shown). Under nonreduced conditions, several peaks corresponding to covalent aggregates were observed in the electrophoretic prole of T-MCC (Figure 4a). The observed molecular weights of these aggregates suggest presence of a covalent dimer species (molecular weight 330 kDa), as well as some other aggregates involving monomer and dimer species. A comparison of the aggregate levels in SEC and nonreduced CE-SDS indicated that covalent aggregation contributed to 84% of the total aggregation in the case of

1594 Bioconjugate Chem., Vol. 21, No. 9, 2010

Wakankar et al.

T-MCC and 30% of the total aggregation observed in the case of T-DM1 (Figure 6). The free thiol content of Tmab is fairly low, experimentally determined to be 0.2 mol/mol of antibody (internal document). As a result, the extent of aggregation observed cannot be explained solely on the basis of formation of intermolecular disulde cross-links. However, the processing of Tmab to T-MCC could probably generate free thiols within the antibody via potential reduction of Cys residues. To investigate the mechanism of covalent aggregation in case of T-MCC, on-chip CE-SDS was performed under reducing conditions. As shown in Figure 4b, for T-MCC a number of peaks in addition to those corresponding to heavy chain and light chain were observed. On the basis of the calculated molecular weights, these additional peaks corresponded to species such as HL, HH, HHL, and HHLL. If the covalent aggregates were primarily mediated through disulde bonds, then the predominant species in the reduced CE-SDS is expected to be the H-chain and the L-Chain. The presence of various cross-linked species (HL, HH, HHL, and HHLL) suggests that other forms of covalent interactions are involved. Furthermore, it is possible that the cross-linking could be either intramolecular (i.e., within an antibody molecule) or intermolecular (i.e., involving different antibody molecules). It was noted that during incubation the percent increase in covalent aggregates for T-MCC, measured as per nonreduced CE-SDS assay, was comparable to the percent increase in cross-linked species, determined as per reduced CE-SDS assay (data not shown). This suggests that the formation of covalent aggregates in T-MCC samples results predominantly from formation of intermolecular nondisulde mediated covalent cross-linking between the different H and L chains. Mass spectrometric analysis of reduced T-MCC samples was performed to conrm the presence of the cross-linked species observed in the reduced CE-SDS assay. The most intense ions in the deconvoluted mass spectra of the reduced T-MCC samples correspond to light chain and heavy chain related fragments with minor species observed at m/z of 72 822 and 73 041 Da (Figure 5). The molecular weights of these species are in agreement with cross-linking between light chain and heavy chain through one and two MCC linkers, respectively (Table 1). However, other interchain cross-linked species such as HH, HHL, and HHLL were not observed. This is possibly the result of factors including poor ionization efciency and poor transmission of higher mass ions through the mass spectrometer. A comparison of the aggregation behavior of T-MCC and T-DM1 shows that T-DM1 has an increased stability to aggregation compared to T-MCC. As mentioned earlier, the aggregation of T-MCC is predominantly covalent in nature. The chemical difference between T-MCC and T-DM1 is the presence of free maleimide moieties on the T-MCC surface. T-DM1 is the product of the conjugation reaction, during which the thiol group of DM1 undergoes nucleophilic Michael addition to the double bond of the maleimidyl moiety of T-MCC. Other nucleophilic groups, such as amino groups and hydroxyl groups that are present abundantly on amino acid side chains of Tmab, can also react with the maleimide moiety of T-MCC to form interchain cross-links. To investigate whether the formation of these intermolecular cross-links between nucleophilic amino acid side chains of T-MCC and its maleimide moiety was responsible for the covalent aggregation observed, the N-acetylated amino acids Cys, Lys, His, Ser, and Tyr were added to T-MCC formulations at concentration of 3 mM. After 11 days under thermal stress conditions at 40 C, SEC results (Figure 7) demonstrate that the aggregation observed for T-MCC was inhibited completely following addition of 3 mM Cys to the formulation. At 1.2 mM

concentration of Cys, this represents a 2-fold molar excess over free maleimide moieties; a similar effect on inhibition of T-MCC aggregation was observed. Addition of amino acids Ser and Tyr also resulted in a signicant (8%) decrease in aggregation. In the case of Ser, it was noted that this inhibitory effect was concentration-dependent with a 4-fold increase in Ser concentration leading to 3-fold increase in inhibition. The inhibitory effect of Cys, Ser, and Tyr was also conrmed using LC-MS analysis of the reduced T-MCC formulation samples containing the different amino acids. Figure 8 shows the relative response of the cross-linked HL species in these samples. It was observed that reaction of T-MCC with Cys resulted in a very stable product where little to no cross-linked species were present after 11 days under thermal stress conditions at 40 C. Samples stressed in the presence of Ser and Tyr amino acids showed lower amounts of the cross-linked HL species compared with the control. This inhibition in the cross-linked HL species could be due to nucleophilic addition of Cys, Ser, and Tyr amino acids to the maleimide moiety that is present on T-MCC. This reaction would block formation of both intermolecular and intramolecular cross-links between the different heavy chain and light chains. In conclusion, the modication and conjugation processes involved in synthesis of the antibody-drug conjugate T-DM1 destabilize the CH2 domain of the antibody. Wang et al. demonstrated that the lysine residues in this domain, due to their greater exibility and solvent accessibility, are more susceptible to conjugation with the drug species DM1 than are those in other parts of the molecule (9). The increased susceptibility of the CH2 Lys residues to conjugation leads to disruption in the structural stability of this domain. In addition, T-MCC, an intermediate in the processing of T-DM1, is highly susceptible to covalent aggregation. This covalent aggregation could be due to formation of intermolecular cross-links involving the maleimidyl moiety and the side chains of nucleophilic amino acid such as Cys, Ser, or Tyr. The conjugation of the drug DM1 to T-MCC greatly increases the stability of the molecule to aggregation behavior.

ACKNOWLEDGMENT
The authors thank Drs. Oscar Salas-Solano and Fred Jacobson for their technical support and advice. We also thank Dr. Charlie Eigenbrot for providing us the X-ray crystal structure information on the IgG1 CH2 domain.

LITERATURE CITED
(1) Chari, R. V. (1998) Targeted delivery of chemotherapeutics: tumor-activated prodrug therapy. AdV. Drug DeliVery ReV. 31, 89104. (2) Chari, R. V. (2008) Targeted cancer therapy: conferring specicity to cytotoxic drugs. Acc. Chem. Res. 41, 98107. (3) Erickson, H. K., Park, P. U., Widdison, W. C., Kovtun, Y. V., Garrett, L. M., Hoffman, K., Lutz, R. J., Goldmacher, V. S., and Blattler, W. A. (2006) Antibody-maytansinoid conjugates are activated in targeted cancer cells by lysosomal degradation and linker-dependent intracellular processing. Cancer Res. 66, 44264433. (4) Payne, G. (2003) Progress in immunoconjugate cancer therapeutics. Cancer Cell 3, 207212. (5) Wu, A. M., and Senter, P. D. (2005) Arming antibodies: prospects and challenges for immunoconjugates. Nat. Biotechnol. 23, 11371146. (6) Lambert, J. M. (2005) Drug-conjugated monoclonal antibodies for the treatment of cancer. Curr. Opin. Pharmacol. 5, 543 549. (7) Cassady, J. M., Chan, K. K., Floss, H. G., and Leistner, E. (2004) Recent developments in the maytansinoid antitumor agents. Chem. Pharm. Bull. (Tokyo) 52, 126.

Physicochemical Stability of Trastuzumab-DM1 (8) Kupchan, S. M., Komoda, Y., Court, W. A., Thomas, G. J., Smith, R. M., Karim, A., Gilmore, C. J., Haltiwanger, R. C., and Bryan, R. F. (1972) Maytansine, a novel antileukemic ansa macrolide from Maytenus ovatus. J. Am. Chem. Soc. 94, 1354 1356. (9) Wang, L., Amphlett, G., Blattler, W. A., Lambert, J. M., and Zhang, W. (2005) Structural characterization of the maytansinoidmonoclonal antibody immunoconjugate, huN901-DM1, by mass spectrometry. Protein Sci. 14, 24362446. (10) Ranson, M., and Sliwkowski, M. X. (2002) Perspectives on anti-HER monoclonal antibodies. Oncology 63 (Suppl 1), 17 24. (11) Ross, S., Spencer, S. D., Holcomb, I., Tan, C., Hongo, J., Devaux, B., Rangell, L., Keller, G. A., Schow, P., Steeves, R. M., Lutz, R. J., Frantz, G., Hillan, K., Peale, F., Tobin, P., Eberhard, D., Rubin, M. A., Lasky, L. A., and Koeppen, H. (2002) Prostate stem cell antigen as therapy target: tissue expression and in vivo efcacy of an immunoconjugate. Cancer Res. 62, 25462553. (12) Lu, S. X., Takach, E. J., Solomon, M., Zhu, Q., Law, S. J., and Hsieh, F. Y. (2005) Mass spectral analyses of labile DOTANHS and heterogeneity determination of DOTA or DM1 conjugated anti-PSMA antibody for prostate cancer therapy. J. Pharm. Sci. 94, 788797. (13) Beeram, M., Krop, I., Modi, S., Tolcher, A., Rabbee, N., Girish, S., Tibbitts, J., Holden, S., Lutzker, S., and Burris, H. (2007) A phase I study of trastuzumab-MCC-DM1(T-DM1), a rst-in-class HER2 antibody-drug conjugate (ADC), in patients (pts) with HER2+ metastatic breast cancer (BC). J. Clin. Oncol., ASCO Annual Meeting Proceedings 25. (14) Lewis-Phillips, G. D., Li, G., Dugger, D. L., Crocker, L. M., Parsons, K. L., Mai, E., Blattler, W. A., Lambert, J. M., Chari, R. V., Lutz, R. J., Wong, W. L., Jacobson, F. S., Koeppen, H.,

Bioconjugate Chem., Vol. 21, No. 9, 2010 1595 Schwall, R. H., Kenkare-Mitra, S. R., Spencer, S. D., and Sliwkowski, M. X. (2008) Targeting HER2-positive breast cancer with trastuzumab-DM1, an antibody-cytotoxic drug conjugate. Cancer Res. 68, 92809290. (15) Morris, P. G., and Fornier, M. N. (2009) Novel anti-tubulin cytotoxic agents for breast cancer. Exp. ReV. Anticancer Ther. 9, 175185. (16) Jacobson, F., Wang, Y., Wakankar, A., Rivera, J., Lee, C., Bechtel, C., and Chen, Y. (2009) Characterization and comparability of Trastuzumab-DM1, WCBP 2009 Abstracts P-127-M. (17) Adamczyk, M., Gebler, J., Shreder, K., and Wu, J. (2000) Region-selective labeling of antibodies as determined by electrospray ionization-mass spectrometry (ESI-MS). Bioconjugate Chem. 11, 557563. (18) Siegel, M. M., Tabei, K., Kunz, A., Hollander, I. J., Hamann, R. R., Bell, D. H., Berkenkamp, S., and Hillenkamp, F. (1997) Calicheamicin derivatives conjugated to monoclonal antibodies: determination of loading values and distributions by infrared and UV matrix-assisted laser desorption/ionization mass spectrometry and electrospray ionization mass spectrometry. Anal. Chem. 69, 27162726. (19) Lazar, A. C., Wang, L., Blattler, W. A., Amphlett, G., Lambert, J. M., and Zhang, W. (2005) Analysis of the composition of immunoconjugates using size-exclusion chromatography coupled to mass spectrometry. Rapid Commun. Mass Spectrom. 19, 18061814. (20) Ionescu, R. M., Vlasak, J., Price, C., and Kirchmeier, M. (2008) Contribution of variable domains to the stability of humanized IgG1 monoclonal antibodies. J. Pharm. Sci. 97, 14141426.
BC900434C