Understanding the Conformational Impact of Chemical Modifications

Mar 5, 2014 - Biopharmaceutical Analytical Sciences, Biopharm R&D, GlaxoSmithKline, 709 Swedeland Road, King of Prussia, Pennsylvania 19406,...
7 downloads 0 Views 3MB Size
Article pubs.acs.org/ac

Understanding the Conformational Impact of Chemical Modifications on Monoclonal Antibodies with Diverse Sequence Variation Using Hydrogen/Deuterium Exchange Mass Spectrometry and Structural Modeling Aming Zhang,* Ping Hu,† Paul MacGregor, Yu Xue, Haihong Fan, Peter Suchecki, Leonard Olszewski, and Aston Liu Biopharmaceutical Analytical Sciences, Biopharm R&D, GlaxoSmithKline, 709 Swedeland Road, King of Prussia, Pennsylvania 19406, United States S Supporting Information *

ABSTRACT: Chemical modifications can potentially induce conformational changes near the modification site and thereby impact the safety and efficacy of protein therapeutics. Hydrogen/deuterium exchange mass spectrometry (HDXMS) has emerged as a powerful analytical technique with high spatial resolution and sensitivity in detecting such local conformational changes. In this study, we utilized HDX-MS combined with structural modeling to examine the conformational impact on monoclonal antibodies (mAbs) caused by common chemical modifications including methionine (Met) oxidation, aspartic acid (Asp) isomerization, and asparagine (Asn) deamidation. Four mAbs with diverse sequences and glycosylation states were selected. The data suggested that the impact of Met oxidation was highly dependent on its location and glycosylation state. For mAbs with normal glycosylation in the Fc region, oxidation of the two conserved Met252 and Met428 (Kabat numbering) disrupted the interface interactions between the CH2 and CH3 domains, thus leading to a significant decrease in CH2 domain thermal stability as well as a slight increase in aggregation propensity. In contrast, Met oxidation in the variable region and CH3 domain had no detectable impact on mAb conformation. For aglycosylated mAb, Met oxidation could cause a more global conformational change to the whole CH2 domain, coincident with the larger decrease in thermal stability and significant increase in aggregation rate. Unlike Met oxidation, Asn deamidation and Asp isomerization mostly had very limited effects on mAb conformation, with the exception of succiminide intermediate formation which induced a measurable local conformational change to be more solvent protected. Structural modeling suggested that the succinimide intermediate was stabilized by adjacent aromatic amino acids through ring−ring stacking interactions.

M

A binding,7,8 FcRn binding,7 and circulation half-life9 as a result. In contrast, the antibody binding affinity to an array of Fcγ receptors was shown to be largely unaffected by Met oxidation with the exception of Fcγ RIIa in which case a slight decrease was observed.7 Though not reported for monoclonal antibody yet, oxidation of Met residue in protein therapeutics also has the potential to induce protein aggregation and immunogenicity in the case of recombinant human interferon β.10 In contrast, Asn deamidation and Asp isomerization can occur in many more locations across the mAb sequence.11 Besides the conserved hotspots in the constant regions,12,13 Asn deamidation and Asp isomerization have also been observed in many cases in the variable domains of IgG1 antibodies, particularly in the highly solvent exposed complementarity-

onoclonal antibodies (mAbs) have become a major class of protein therapeutics given their definitive mechanisms of action, high binding affinity, and specificity to selected targets.1,2 However, similar to other protein therapeutics, the inherent instability of polypeptides makes them susceptible to a variety of chemical modifications during the manufacturing, distribution, and long-term storage. Methionine (Met) oxidation, asparagine (Asn) deamidation, and aspartic acid (Asp) isomerization are three common chemical modifications that have been closely monitored during the antibody drug development.1,3,4 Their impacts on therapeutic mAb attributes such as the thermal stability, antigen binding, and Fc effector functions have also been extensively studied. Significant insight has been gained and reported in the literature. For instance, most IgG1- and IgG2-class mAbs have two conserved Met residues, Met252 and Met428 located in the CH2 and CH3 domain, respectively. The oxidation of these two Met residues has been reported to decrease mAb thermal stability,5,6 protein © 2014 American Chemical Society

Received: December 13, 2013 Accepted: March 5, 2014 Published: March 5, 2014 3468

dx.doi.org/10.1021/ac404130a | Anal. Chem. 2014, 86, 3468−3475

Analytical Chemistry

Article

determining regions (CDRs).14−16 In some situations, succinimide as an intermediate state can accumulate substantially in these chemical modification pathways.16,17 Because of the variety of potential sites for modifications and the possible succinimide intermediate accumulation, the impact of Asn deamidation and Asp isomerization on antibody is more molecule and location dependent and thus less predictable. For example, deamidation of heavy chain Asn55 in CDR214,16 and isomerization of Asp102 in CDR318 have been shown to significantly decrease the antigen binding affinity and therefore the potency of the mAb therapeutics. In contrast, isomerization and succinimide accumulation of Asp30 in the light chain CDR1 showed no observable impact on its antigen binding affinity.17 In the constant region, deamidation at the conserved hotspots of Asn389 and Asn434 (Kabat numbering) was shown to not affect the Fcγ receptors and FcRn binding.13 Despite the accumulating knowledge in the field, there is one fundamental question that remains unanswered, that is what the structural basis is for the various chemical modifications to induce distinct impacts on mAb attributes. In most of the studies mentioned above, the primary focus has been on establishing the correlation between the observed chemical modifications and the corresponding effects on mAb stability, aggregation rate, and biological functions.9,12,13 Only a very few studies attempted to assess the local or global conformational integrity after various chemical modifications to understand their impacts on mAb attributes from a high order structural perspective.5 In recent years, hydrogen/deuterium exchange mass spectrometry (HDX-MS) has been increasingly used in the biopharmaceutical industry to characterize protein high order structure,19,20 monitor conformational changes upon chemical modifications,5,21 and investigate the mechanisms of aggregation.22,23 In this study, we explored the use of the HDXMS technology together with structural modeling to study the conformational impact of three common chemical modifications on IgG1 monoclonal antibodies. The measured conformational changes were compared across four selected mAbs with diverse sequences to have a different number of modification sites, which allowed a more systematic evaluation of the correlation between the conformational impact and the modification locations and glycosylation states. Furthermore, the structural basis for the chemical modifications to affect important mAb attributes like thermal stability and aggregation behavior was elucidated.

samples were buffer exchanged to a common formulation buffer at pH 5.5. Hydrogen/Deuterium Exchange Mass Spectrometry (HDX-MS). HDX-MS analysis of control and chemically modified mAb samples were performed using a Waters HDX manager system coupled to a Synapt G2-S mass spectrometer. The data were analyzed using Waters DynamX software. Specifically, mAb samples were diluted 10-fold into D2O to initiate H/D exchange. The exchange reaction was maintained at room temperature for various durations at 10 s and 1, 5, 30, and 180 min. The reaction was then stopped by adding 4 times the volume of ice-cold quenching buffer (0.2 M glycine, 8 M Guanidine HCl, and 0.5 M TCEP, pH 2.7). The quenched mAb samples were further diluted 4-fold into 0.1% formic acid (pH 2.5, precooled at ice) and injected into a Waters HDX system. mAb sample was digested by flowing through an online immobilized pepsin column (AB applied science, 2.1 × 30 mm). The resulting peptide mixture was desalted on a Waters VanGuard precolumn (2.1 × 5 mm) for 2 min and then separated in a reverse phase UPLC column (Waters Acquity BEH300 column, 1.0 × 100 mm) at a flow rate of 40 μL/min. An acetonitrile (ACN) gradient from 15% to 40% over 8 min was used for peptide separation. The eluent was directly injected into a Synapt G2-S mass spectrometer running in the ESI positive mode. The data were acquired in full MS scan over m/z range of 200−2000 with lock mass spray correction using Glu-fibrinogen B peptide. The peptides resulting from online pepsin digestion were identified by running a separate experiment to collect tandem (MS/MS) mass spectrometry data from mAb samples without deuterium labeling. The MS/ MS data were analyzed using the Waters ProteinLynx Global Server (PLGS) and DynamX software to identify peptides with sufficient signal and confidence that could be reliably used for deuterium labeling analysis. The numbers of deuterium labeling into each peptide at different time points were calculated using the DynamX software and later were exported to Excel to make the HDX differential plots which compared the deuterium labeling for each peptide between the control and modified mAb states. mAb Structure Modeling Using MOE. Homology mAb structural models in this study were built using Molecular Operating Environment (MOE) antibody modeler.24,25 The MOE modeler employs a knowledge-based approach that utilizes a database of available antibody structures in GSK inhouse database depositing public antibody domain structures and GSK’s proprietary antibody structures. In Fab homology modeling, appropriate templates for the framework region and CDR loops were first found by searching against the antibody structure database. Then, the respective loop templates were grafted onto the designated light chain and heavy chain frameworks, followed by energy minimization in the transition area between CDRs and frameworks in order to relax strained geometries using AMBER99 force field. The structural model of mAbs after chemical modifications (oxidation, deamidation, isomerization, etc.) was created using MOE builder program with energy minimization. Loop conformations are generated using the conformation search module LowMode MD in MOE package.26



MATERIALS AND METHODS Materials. Four IgG1 mAbs designated as mAb1 (typical IgG1), mAb2-1M (IgG1 with one exposed Met for oxidation in the variable region), mAb3-2M (Fc-engineered IgG1, with 2 additional Met in CH3 domain), and mAb4-AG (Fc-engineered aglycosylated IgG1) were expressed and purified at GlaxoSmithKline (GSK). Deuterium oxide (D2O) was purchased from Sigma-Aldrich (99% deuterium). All other reagents were HPLC purity grade and purchased from Sigma-Aldrich unless otherwise specified. Sample Treatment. Met oxidation was carried out by mixing each mAb with hydrogen peroxide (H2O2) at a 1:10,000 protein/H2O2 molar ratio. The mixture was incubated at room temperature for 1 h. Asn deamidation and Asp isomerization were carried out by incubating the mAb samples at three different pH conditions (pH 4.0, 5.5, and 9.0) and 40 °C for 0, 3, 7, and 14 days. After H2O2 and pH incubations, all mAb



RESULTS AND DISCUSSION Met Oxidation and Its Impact on mAb Attributes. All four IgG1 mAbs selected in this study have two conserved Met residues at 252 and 428 (Kabat numbering) in the CH2 and

3469

dx.doi.org/10.1021/ac404130a | Anal. Chem. 2014, 86, 3468−3475

Analytical Chemistry

Article

Figure 1. HDX differential plots of mAb1 heavy chain (HC, Panel A) and light chain (LC, Panel B). The two protein states compared are Met oxidized vs control sample. Deuterium labeling was measured at 10 s (red), 1 min (orange), 5 min (cyan), 30 min (blue), and 180 min (green). Vertical sticks represent the total HDX differences of each peptide from five labeling time points. The gray and orange dotted lines represent the criteria for a significant HDX difference for one single labeling time point and for five labeling time points combined, respectively. The sequence regions with significant HDX difference are labeled accordingly.

points. The criteria used to determine significant HDX differences at a given confidence interval between two protein states were derived from statistical analysis of the HDX-MS method uncertainty as described previously.28 For our particular HDX-MS system, if a peptide meets the following two criteria simultaneously, it can be considered as being statistically significantly different with 99.7% confidence interval: (1) the measured individual HDX difference is greater than 0.35 Da for at least one labeling time point and (2) the total HDX difference summed from five time points is greater than 0.8 Da. In Figure 1, these two criteria are shown as horizontal dotted lines. The HDX differential plots in Figure 1 show most peptides from both heavy and light chains have deuterium labeling differences less than the predetermined significance level, suggesting no or minimal conformational changes detected upon Met oxidation, but meanwhile, six partially overlapping peptides were observed to have significantly increased deuterium labeling after Met oxidation. These peptides corresponded to two discrete sequence regions 246−256 and 324−353 in mAb1 heavy chain. It is interesting to note that region 246−256 is adjacent to Met257, one of the two Met residues that were nearly fully oxidized after H2O2 treatment. In addition, peptides from region 246−256 showed a much higher degree of increase in deuterium labeling compared to these from region 324−353. These observations suggested that oxidation of Met257 and Met433 could disrupt certain local conformations to have a more solvent exposed structure, and that such conformational impact appeared to be more

CH3 domain, respectively. mAb3-2M, which is Fc-engineered for enhanced therapeutic efficacy, has two additional Met residues at 358 and 397 in the CH3 domain. Upon H2O2 treatment, both conserved Met252 and Met428 in the Fc domain were highly oxidized (>98%, data shown in Supporting Information Table S-1) for all the four mAbs. This also held true for the two additional Met residues (Met358 and Met397) present in mAb3-2M CH3 domain. In contrast, Met residues in the Fab domain, including Met34, Met81, or equivalent positions in the heavy chain and Met4 in the light chain of all four mAbs, showed only less than 2% oxidation under the same oxidation condition (data not shown). This is consistent with previous studies suggesting the Fc domain Met residues were generally more prone to oxidation than those in the Fab domain.27 As a result, oxidation of Met residues in the Fab domain will not be the focus of this study due to their low levels. One exception, however, is Met50 in mAb2-1M CDR2, which had a comparable oxidation level as these in the Fc domain upon H2O2 treatment. The Impact of Met Oxidation on mAb1 Conformation. The conformational impact of Met oxidation was first examined for mAb1 using HDX-MS. Figure 1 shows the deuterium labeling differences between the control vs Met oxidized samples for mAb1 heavy and light chains. In these HDX differential plots, deuterium labeling into each peptide at various time intervals (10 s and 1, 5, 30, 180 min) was compared between the two states, and the differences were plotted in different color traces. The vertical sticks represent the total labeling difference of each peptide summed from five time 3470

dx.doi.org/10.1021/ac404130a | Anal. Chem. 2014, 86, 3468−3475

Analytical Chemistry

Article

suggested disruption of the electrostatic interactions between Lys253 and Glu385 resulting from Met oxidation. The collective effect of these predicted disruptions to the side chain interactions could lead to more solvent exposed conformations in the local interface regions, including the short α-helix of region 246−256 and the linker residues in region 324−353 as identified by HDX-MS. In a previous study, Houde et al. has demonstrated considerable conformational change with increased solvent accessibility in the same helix region in the CH2 domain upon Met oxidation for another IgG1 mAb.5 However, the author did not report a second region with minor conformational change that is equivalent to the region 324−353 found in this present study. It is possible that different thresholds were used to determine a significant HDX difference between the studies. Our current data suggest that Met oxidation can impact the local conformation not only near the modification site but also to certain distant residues linking the CH2 and CH3 domains, although to a much less degree. The Impact of Met Oxidation on mAb Conformation with Diverse Sequences. The conformational impact of Met oxidation was also analyzed with regard to the modification sites and the mAb glycosylation state by investigating four selected mAbs with different sequences. Relative to mAb1, mAb2-1M has one solvent exposed Met in the heavy chain CDR2 which is prone to oxidation. mAb3-2M has two additional Mets in the CH3 domain besides the conserved Met428. mAb4-AG has the same number of Met in the Fc domain as mAb1 but is aglycosylated via Asn to Ala mutation in the glycosylation motif. All these four mAbs were treated with H2O2 to increase the level of oxidation. Figure 3 shows the HDX comparison of the three mAbs between the control and Met oxidized samples. No significant HDX difference was seen in the light chain for any of the mAbs as in the case of mAb1 after Met oxidation; thus, the data was not shown. For the three glycosylated mAbs (shown in Figures 1 and 3A,B), it can be seen that the measured HDX difference patterns are all quite similar. A large increase in deuterium labeling after Met oxidation was mainly found in the 11 amino acid region adjacent to the oxidation site in the CH2 domain, but meanwhile, a much lower but statistically significant increase of deuterium labeling was also seen in the linking region between CH2 and CH3 domains. These data suggested that Met oxidation had very similar conformational impact across the three mAbs, though they had a varying number of Met residues at different locations subjected to oxidation. In contrast, the aglycosylated mAb4-AG showed a quite different HDX pattern compared to the other mAbs as demonstrated in Figure 3. Dramatic increases in deuterium labeling were seen not only in the local region adjacent to the Met oxidation site but also in many other regions across the CH2 domain. It indicated that Met oxidation could induce a more substantial conformational change to the CH2 domain when the antibody was aglycosylated. The different behaviors observed between glycosylated and aglycosylated mAbs were suggesting that the N-linked glycan was important in determining the impact of Met oxidation. In a similar study, Houde et al. investigated the impact of various chemical modifications including different glycosylation and Met oxidation on mAbs conformation and discovered that Met oxidation and deglycosylation affected mAb conformation in the same local region near the CH2 Met residue.5 The data presented here further suggested that the combined effect of Met oxidation and aglycosylation could

pronounced in the region adjacent to the Met257 oxidation site. Structural modeling was used as a complementary tool to evaluate the impact of Met257 and Met433 oxidation on mAb1 conformation. Figure 2A shows the Fc homology structure of

Figure 2. Modeling illustration of Met oxidation impact on mAb1 local conformation. (Panel A) mAb1 Fc homology structure. Regions 246− 256 and 324−353 are highlighted in red and pink, respectively, to represent different levels of conformational changes caused by Met oxidation. Zoomed-in view of local conformation at the CH2−CH3 interface is shown for native (Panel B) and Met oxidized (Panel C) mAb1. The distance between two sulfur atoms of Met257 and Met433 are labeled in green.

mAb1, with the two regions 246−256 and 324−353 identified by HDX analysis labeled in red and pink, respectively. In the 3D structure, both regions seem to be related to the CH2 and CH3 domain interface in some way. Region 246−256 contains a short α-helix located at the domain interface, while region 324−356 includes the residues linking CH2 and CH3 domains. In the zoomed view of Fc native structure (i.e., before Met oxidation, Figure 2B), Met257 and Met433, though located in two different domains, were packed in proximity with a distance of ∼4.25 Å between the two sulfur atoms at the CH2−CH3 domain interface. Certain hydrophobic interaction between the two Met residues was expected given the close distance. In addition, in the interface, Met433 was also believed to form hydrophobic interaction with Lys253, which further interacted with Glu385 in the CH3 domain via strong electrostatic interaction. Together, these interactions were considered to be critical in stabilizing not only the local conformations but also the CH2 and CH3 domain interface.8 Upon oxidation, the addition of oxygen atoms to both Met257 and Met433 could lead to some unfavorable repulsion interaction between the Met residues due to steric hindrance effect. As shown in Figure 2C, structural modeling of oxidized mAb1 suggested that both Met residues had new side chain orientations relative to their original positions to accommodate the oxygen atoms induced by oxidation. As a result, the sulfur distance between two Met residues increased to 6.26 Å, which was out of the effective hydrophobic interaction range. Therefore, as a direct impact of the oxidation, the interactions involving the Met residues were expected to be disrupted, including these between the two methonines and between Met433 and Lys253. In addition, the structural model also 3471

dx.doi.org/10.1021/ac404130a | Anal. Chem. 2014, 86, 3468−3475

Analytical Chemistry

Article

Figure 3. HDX analysis of conformational changes by Met oxidation for selected mAbs. In each panel, deuterium labeling between Met oxidized sample (top) vs control sample (bottom) are compared. The HDX labeling time is 10 s (red), 1 min (orange), 5 min (cyan), 30 min (blue), and 180 min (green). The gray and orange dotted lines represent the criteria for significant HDX difference for one single labeling time point and for five labeling time points combined, respectively.

induce a far more profound conformational impact that extended to the entire CH2 domain. It was likely because the removal of N-linked glycan reduced the CH2 domain conformational stability as suggested by Houde et al.5 and Zheng et al.,29 thus making the CH2 domain more vulnerable to structural perturbations upon Met oxidation. The Impact of Met Oxidation on mAb Thermal Stability and Aggregation Rate. Besides the conformational changes, Met oxidation can also impact antibody’s other attributes such as thermal stability and aggregation propensity. Figure 4 shows the changes in melting temperatures for four mAbs upon Met oxidation. A common trend was seen that the first unfolding transition shifted to a lower temperature, while the second, major unfolding transition remained almost identical after Met oxidation. The only exception seen was mAb2-1M which showed a slight decrease in the second transition temperature (Figure 4B). In the case of mAb3-2M, the DSC thermogram showed an additional unfolding transition between the first and the major transitions. This additional transition also had lower melting temperature after Met oxidation. The degree of decrease in melting temperature after Met oxidation was shown for four mAbs in Supporting Information Table S-2.

Figure 4. DSC analysis of thermal stability before (solid line) and after (dotted line) Met oxidation for four selected mAbs: mAb1 (Panel A); mAb2-1M (Panel B); mAb3-2M (Panel C); mAb4-AG (Panel D). 3472

dx.doi.org/10.1021/ac404130a | Anal. Chem. 2014, 86, 3468−3475

Analytical Chemistry

Article

Figure 5. HDX analysis of local conformational changes upon succinimide formation at Asp55 and Asp104 of mAb1. Relative deuterium labeling of Peptide AA50-60 and Peptide AA102-108 is shown Panels A and B, respectively. The error bars were calculated from HDX analysis in triplicates.

For antibody, the first thermal transition is typically attributed to CH2 domain unfolding, while the major transition can be due to the CH3 and Fab domains in cases where they are not resolved.30 Following this assignment, the data in Figure 4 suggested that Met oxidation mainly decreased the CH2 domain thermal stability, which was consistent with observations of major conformational changes only to the CH2 domain and the interface in HDX analysis. In addition, the aglycosylated mAb4-AG showed a larger degree of decrease in CH2 domain melting temperature compared to the other three mAbs (−12.4 vs −8.2 °C). Again, it agreed with the observation that mAb4-AG was subjected to more dramatic conformational changes in the CH2 domain relative to the other mAbs upon Met oxidation. In Figure 4, mAb2-1M also showed a small decrease of 1.7 °C in the second transition, probably due to the oxidation of Met50 in CDR2. However, no local conformational change was detected by HDX-MS. It was probably because the CDR regions were highly exposed in its native structure, and minor conformational disruptions if present by the local Met oxidation did not further increase its solvent accessibility (i.e., deuterium labeling).

Aggregation propensity of four mAbs upon Met oxidation was assessed by incubating the control and oxidized samples at 40 °C for up to 7 days. The aggregation level was measured by SEC and shown in Supporting Information Figure S-1. The oxidized samples of three glycoslyated mAbs had only a minor increase in aggregation ranging from 0.3% (mAb1) to 1.6% (mAb2-1M) when compared to the corresponding control samples after a 7 day incubation. It indicated that Met oxidation had limited impact on the aggregation behavior of mAbs with normal glycosylation profiles. In contrast, the aglycosylated mAb4-AG showed an increase of 71.2% in aggregation level after Met oxidation. The vast difference seen between glycosylated vs aglycosylated mAbs coincided with both the HDX-MS and DSC data which suggested that mAb4-AG had larger conformational disruption and thermal destabilization resulting from Met oxidation in the CH2 domain compared to the other mAbs. The consistence from three different measurements suggests some underlying correlations among mAb conformation, thermal stability, and aggregation. The conformational changes resulting from Met oxidation might be 3473

dx.doi.org/10.1021/ac404130a | Anal. Chem. 2014, 86, 3468−3475

Analytical Chemistry

Article

The impact of succinimide formation at Asp55 and Asp104 was also evaluated. In HDX-MS analysis, the formation of succinimide ring at Asp residue would lead to a mass decrease of 18.01 Da. Therefore, peptides containing these Asp residues from unmodified and modified mAb molecules can be unambiguously distinguished, which allows tracking of any potential conformational changes occurring to these local regions upon succinimide formation. Figure 5 shows the relative deuterium labeling of two peptides, AA50-60 and AA102-108 that contain the isomerization sites Asp55 and Asp104, respectively. The loss of the exchangeable amide proton to each Asp residue upon succinimide formation was taken into account when calculating the relative deuterium labeling. Both peptides showed significant decreases in deuterium labeling after succinimide formation, indicating the formation of less dynamic, more solvent protected local conformations at both sites. In addition, peptide AA102-108 was seen to have a much larger degree of decrease in deuterium labeling than peptide AA50-60, suggesting the Asp104 site underwent a more substantial local conformational change upon succinimide ring formation. Structural Modeling of Succinimide Formation at Asp50 and Asp104. Figure 6 predicted the local conformational

the molecular structural basis for the changes to thermal stability and aggregation. A decrease of mAb thermal stability upon Met oxidation has been reported in previous studies.5,6 Here, we showed the correlation of conformational changes to the degree of reductions in domain thermal stability caused by Met oxidation in several mAbs. Furthermore, the data in this study also suggested that a reduction in thermal stability arising from Met oxidation also likely resulted in an increased propensity for aggregation. The Impact of Asn Deamidation and Asp Isomerization on mAb Conformation. Asn deamidation and Asp isomerization are two other common chemical modifications. Their impacts on mAb conformation have not been well characterized due to the lack of a sensitive method to detect the likely minor conformational changes caused by these modifications. In this study, mAb1 was selected as the model antibody. It is because mAb1 also has two unique DG motifs (Asp55 and Asp104) in the flexible CDR regions besides the conserved NG motifs at Asn320 and Asn389 and DG motifs at Asp285 and Asp406 in the Fc constant region. The mAb sample was incubated at pH 4.0, 5.5, and 9.0 and 40 °C for up to 14 days before HDX conformational analysis. The modification levels to all the potential hotspots upon pH incubation were measured and partly shown in Supporting Information Table S-3. Only Asp at 55 and 104 and Asn at 389 showed a substantial increase in the level of isomerization and deamidation, respectively, after incubation. All the other Asn and Asp hotspots had no obvious increase in modifications and thus was not included in Supporting Information Table S-3. The major isomerization observed at Asp55 and Asp104 but not at Asp285 and Asp406 agreed well with previous studies suggesting the Asp residues in the flexible CDR regions were more susceptible to modification than those in the constant Fc domain.3,15 In addition to the isomerization end product (IsoAsp), both sites also showed substantial accumulations of succinimide intermediate (Su), which was favored at mild acidic conditions (pH 4.0 and 5.5).17 Interestingly, the data appeared to suggest that Asp104 had a higher tendency to accumulate succinimide intermediate than Asp55 as assessed by the ratio of succinimide to isoAsp at different incubation times, the potential cause of which will be discussed from the structural perspective in a later section. Unlike Asp isomerization, deamidation was found to mainly occur at Asn389 at pH 9 (Supporting Information Table S-3). The conformation of mAb1 sample after isomerization and deamidation modifications was analyzed, and the HDX-MS data were shown in Supporting Information Figures S-2 and S3, respectively. In both cases, no significant HDX difference was observed in either heavy or light chain. This may indicate that there were no measurable local conformational changes caused by Asp isomerization and/or Asn deamidation, which were consistent with previous observations.31 However, it should also be noted that the pH stressed samples selected for HDX analysis (i.e., 14 day incubation at pH 4.0 in Supporting Information Figure S-2 and 3 day at pH 9.0 to ensure minimal aggregation implication in Supporting Information Figure S-3) contained relatively low levels of modifications. The deamidation level at Asn389 was 30.9% while isomerizations at Asp55 and Asp104 were 14.1% and 11.2%, respectively. Therefore, it is possible that the lack of significant conformational difference as measured by HDX could be due to the low abundance of the modified species, in particular for the isomerized species.

Figure 6. Structural modeling of succinimide formation at degradation sites Asp55 (A) and Asp104 (B).

changes near the modification sites at Asp55 and Asp104 after succinimide formation using structural modeling. As stated earlier, both Asp55 and Asp104 are located in CDR loop regions of mAb1. In the native structure, these CDR loop regions usually have high conformational flexibility that is favorable for the succinimide ring structure formation during the isomerization pathway.3,15 As shown in Figure 6A, upon succinimide formation, the local conformational change at Asp55 was expected to be minor. Although there were slight changes in the side chain orientations of certain neighbor residues, the backbone folding was largely unaffected by succinimide formation. In contrast, the formation of succinimide ring structure at Asp104 induced a more pronounced local conformational change as shown in Figure 6B. It can be seen that the backbone of the whole CDR3 loop twisted relative to its native conformation. This backbone refolding brought two aromatic residues of the same loop, namely, Tyr103 and Tyr107, in proximity to the succinimide ring structure at Asp104 to form stable π−π interactions, which 3474

dx.doi.org/10.1021/ac404130a | Anal. Chem. 2014, 86, 3468−3475

Analytical Chemistry

Article

(5) Houde, D.; Peng, Y.; Berkowitz, S. A.; Engen, J. R. Mol. Cell. Proteomics 2010, 9, 1716−1728. (6) Liu, D.; Ren, D.; Huang, H.; Dankberg, J.; Rosenfeld, R.; Cocco, M. J.; Li, L.; Brems, D. N.; Remmele, R. L., Jr. Biochemistry 2008, 47, 5088−5100. (7) Bertolotti-Ciarlet, A.; Wang, W.; Lownes, R.; Pristatsky, P.; Fang, Y.; McKelvey, T.; Li, Y.; Li, Y.; Drummond, J.; Prueksaritanont, T.; Vlasak, J. Mol. Immunol. 2009, 46, 1878−1882. (8) Pan, H.; Chen, K.; Chu, L.; Kinderman, F.; Apostol, I.; Huang, G. Protein Sci. 2009, 18, 424−433. (9) Wang, W.; Vlasak, J.; Li, Y.; Pristatsky, P.; Fang, Y.; Pittman, T.; Roman, J.; Wang, Y.; Prueksaritanont, T.; Ionescu, R. Mol. Immunol. 2011, 48, 860−866. (10) van Beers, M. M.; Sauerborn, M.; Gilli, F.; Brinks, V.; Schellekens, H.; Jiskoot, W. Pharm. Res. 2011, 28, 2393−2402. (11) Du, Y.; Walsh, A.; Ehrick, R.; Xu, W.; May, K.; Liu, H. MAbs 2012, 4, 578−585. (12) Chelius, D.; Rehder, D. S.; Bondarenko, P. V. Anal. Chem. 2005, 77, 6004−6011. (13) Khawli, L. A.; Goswami, S.; Hutchinson, R.; Kwong, Z. W.; Yang, J.; Wang, X.; Yao, Z.; Sreedhara, A.; Cano, T.; Tesar, D.; Nijem, I.; Allison, D. E.; Wong, P. Y.; Kao, Y. H.; Quan, C.; Joshi, A.; Harris, R. J.; Motchnik, P. MAbs 2010, 2, 613−624. (14) Huang, L.; Lu, J.; Wroblewski, V. J.; Beals, J. M.; Riggin, R. M. Anal. Chem. 2005, 77, 1432−1439. (15) Vlasak, J.; Bussat, M. C.; Wang, S.; Wagner-Rousset, E.; Schaefer, M.; Klinguer-Hamour, C.; Kirchmeier, M.; Corvaia, N.; Ionescu, R.; Beck, A. Anal. Biochem. 2009, 392, 145−154. (16) Yan, B.; Steen, S.; Hambly, D.; Valliere-Douglass, J.; Vanden Bos, T.; Smallwood, S.; Yates, Z.; Arroll, T.; Han, Y.; Gadgil, H.; Latypov, R. F.; Wallace, A.; Lim, A.; Kleemann, G. R.; Wang, W.; Balland, A. J. Pharm. Sci. 2009, 98, 3509−3521. (17) Chu, G. C.; Chelius, D.; Xiao, G.; Khor, H. K.; Coulibaly, S.; Bondarenko, P. V. Pharm. Res. 2007, 24, 1145−1156. (18) Harris, R. J.; Kabakoff, B.; Macchi, F. D.; Shen, F. J.; Kwong, M.; Andya, J. D.; Shire, S. J.; Bjork, N.; Totpal, K.; Chen, A. B. J. Chromatogr., B: Biomed. Sci. Appl. 2001, 752, 233−245. (19) Houde, D.; Arndt, J.; Domeier, W.; Berkowitz, S.; Engen, J. R. Anal. Chem. 2009, 81, 5966. (20) Zhang, Z.; Zhang, A.; Xiao, G. Anal. Chem. 2012, 84, 4942− 4949. (21) Wei, H.; Ahn, J.; Yu, Y. Q.; Tymiak, A.; Engen, J. R.; Chen, G. J. Am. Soc. Mass. Spectrom. 2012, 23, 498−504. (22) Zhang, A.; Singh, S. K.; Shirts, M. R.; Kumar, S.; Fernandez, E. J. Pharm. Res. 2012, 29, 236−250. (23) Iacob, R. E.; Bou-Assaf, G. M.; Makowski, L.; Engen, J. R.; Berkowitz, S. A.; Houde, D. J. Pharm. Sci. 2013, 102, 4315−4329. (24) Almagro, J. C.; Beavers, M. P.; Hernandez-Guzman, F.; Maier, J.; Shaulsky, J.; Butenhof, K.; Labute, P.; Thorsteinson, N.; Kelly, K.; Teplyakov, A.; Luo, J.; Sweet, R.; Gilliland, G. L. Proteins 2011, 79, 3050−3066. (25) Long, W. F.; Labute, P. J. Comput.-Aided Mol. Des. 2010, 24, 907−916. (26) Labute, P. J. Chem. Inf. Model. 2010, 50, 792−800. (27) Chumsae, C.; Gaza-Bulseco, G.; Sun, J.; Liu, H. J. Chromatogr., B: Anal. Technol. Biomed. Life Sci. 2007, 850, 285−294. (28) Houde, D.; Berkowitz, S. A.; Engen, J. R. J. Pharm. Sci. 2011, 100, 2071−2086. (29) Zheng, K.; Bantog, C.; Bayer, R. MAbs 2011, 3, 568−576. (30) Ionescu, R. M.; Vlasak, J.; Price, C.; Kirchmeier, M. J. Pharm. Sci. 2008, 97, 1414−1426. (31) Tang, L.; Sundaram, S.; Zhang, J.; Carlson, P.; Matathia, A.; Parekh, B.; Zhou, Q.; Hsieh, M. C. MAbs 2013, 5, 114−125.

was expected to stabilize the succinimide intermediate. This explained why Asp104 had a higher tendency to accumulate succinimide intermediate than Asp55 (Supporting Information Table S-3). The distinct conformational changes predicted from structural modeling at two isomerization sites were also consistent with the HDX-MS observation which suggested a larger decrease of conformational dynamics in the local region of Asp104 compared to that Asp55 after succinimide formation.



CONCLUSIONS In this study, we have used HDX-MS and structural modeling to systematically examine the local conformational changes resulting from three common chemical modifications and correlated such conformational changes with measured impacts on thermal stability and aggregation of therapeutic antibody. Met oxidation was shown to affect the CH2 domain conformation to different degrees for mAbs with different Fc structures and glycosylation states. These conformational changes also showed good correlation with alterations in thermal stability and aggregation behavior upon Met oxidation for four selected mAbs analyzed. Asp isomerization and Asn deamidation were demonstrated to have limited impact on mAb local conformation in general. However, when there was significant accumulation of succinimide intermediate, these modifications could result in considerable local conformational changes. The impact of these modifications on the quality attributes of antibody-based protein therapeutics such as thermal stability, aggregation propensity, and biological activities will be the topic of future studies.



ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*Telephone: 610-270-5418. Fax: 610-270-7100. E-mail: aming. [email protected]. Present Address †

P.H.: Janssen Research & Development, Malvern, PA, 19355.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors would like to thank Dr. Shing Mai, Joshua Fuller, and Dr. Byron Dipaolo at GlaxoSmithKline Biopharm R&D for their support of the study. The authors also want to thank Dr. Narendra Bam for his continuing support and encouragement in exploring novel technologies for biopharmaceutical development.



REFERENCES

(1) Wang, W.; Singh, S.; Zeng, D. L.; King, K.; Nema, S. J. Pharm. Sci. 2007, 96, 1−26. (2) Beck, A.; Wurch, T.; Bailly, C.; Corvaia, N. Nat. Rev. Immunol. 2010, 10, 345−352. (3) Diepold, K.; Bomans, K.; Wiedmann, M.; Zimmermann, B.; Petzold, A.; Schlothauer, T.; Mueller, R.; Moritz, B.; Stracke, J. O.; Molhoj, M.; Reusch, D.; Bulau, P. PLoS One 2012, 7, No. e30295. (4) Wakankar, A. A.; Borchardt, R. T. J. Pharm. Sci. 2006, 95, 2321− 2336. 3475

dx.doi.org/10.1021/ac404130a | Anal. Chem. 2014, 86, 3468−3475