Breaking the Light and Heavy Chain Linkage of Human Immunoglobulin G1 (IgG1) by Radical Reactions

We report that the production of hydrogen peroxide by radical chain reductions of molecular oxygen into water in buffers leads to hinge degradation of a human IgG1 under thermal incubation conditions. The production of the hydrogen peroxide can be accelerated by superoxide dismutase or redox active metal ions or inhibited by free radical scavengers. The hydrogen peroxide production rate correlates well with the hinge cleavage. In addition to radical reaction mechanisms described previously, new degradation pathways and products were observed. These products were determined to be generated via radical reactions initiated by electron transfer and addition to the interchain disulfide bond between Cys215 of the light chain and Cys225 of the heavy chain. Decomposition of the resulting disulfide bond radical anion breaks the C–S bond at the side chain of Cys, converting it into dehydroalanine and generating a sulfur radical adduct at its counterpart. The hydrolysis of the unsaturated dehydropeptides removes Cys and yields an amide at the C terminus of the new fragment. Meanwhile, the competition between the carbonyl (-CαONH-) and the side chain of Cys allows an electron transfer to the α carbon, forming a new intermediate radical species (-·Cα(O−)NH-) at Cys225. Dissociative deamidation occurs along the N–Cα bond, resulting in backbone cleavage. Given that hydrogen peroxide is a commonly observed product of thermal stress and plays a role in mediating the unique degradation of an IgG1, strategies for improving stability of human antibody therapeutics are discussed.

Recombinant monoclonal antibodies (mAbs) represent one of the fastest growing segments of biotechnology. More than 20 mAbs have been approved by the United States Food and Drug Administration, and a volume of information is available in the research and development of therapeutic mAbs. Despite this progress, there are still unmet demands in improving the stability, efficacy, pharmacokinetic profile, and production yield of mAbs (1,2). New mechanisms that govern these characteristics, in particular new mechanisms underlying the instability of mAbs, remain to be elucidated. As one of the product quality criteria, the stability needs to be monitored in the entire development course because it can profoundly affect mAb properties relevant to their therapeutic application. One of the common observations in the stability testing program is the so-called "hinge cleavage," which generates a Fab domain and a partial IgG1 that is missing the Fab. These products contain ladder cleavage sites of the upper hinge residues at new termini of the products (3-7), which were found, interestingly, as the same as those observed in H 2 O 2 -mediated radical-induced hinge cleavage of an IgG1 (8). The observation of the same cleavage sites under different conditions (e.g. thermal incubation conditions versus cell culture production) implies that a common mechanism may exist to drive this unique degradation.
Two types of interchain disulfide bonds are present in the hinge region of an IgG1; one is in the core hinge connecting the heavy chains (the H-H bond), and the other connects the light and heavy chains (the H-L bond). H 2 O 2 -mediated radical reactions initiate by breaking the H-H bond by hydroxyl radical ( ⅐ OH) attack and the subsequent formation of radical species at one of the first hinge Cys 231 residues, which functions as the primary radical center in the hinge peptide of SC 225 DK-THTC 231 PPC (Fig. 1). Electron transfers from the radical center to the upper hinge lead to cleavage of the hinge at a oneradical cleavage per molecule basis to release the Fab domain.
Unlike the cell culture conditions that can generate H 2 O 2 by the activation of various cell surface receptors (9,10), there is no known source of the H 2 O 2 in buffers under thermal incubation conditions. This seems to cast a doubt if radical reaction mechanisms contribute significantly to thermal stress degradation. In addition, high oxygen tension is typically produced by catalytic decomposition of H 2 O 2 by redox metal ions in H 2 O 2 -mediated radical reactions either in the cell culture or in vitro reaction systems (11,12), whereas it remains unclear if such oxygen tensions exist under thermal incubation conditions. These factors are important because Davies and Dean (13) and Garrison (14) have demonstrated that radical reactions can take different pathways under different oxygen tensions. Thus, better understanding of these reactions could facilitate the development of efforts to improve the stability of an IgG1 under thermal stress conditions.
In this paper, we present experimental evidence that demonstrates that the thermal induced degradation of an IgG1 follows radical reaction mechanisms with new pathways, which generate products that were not determined previously. Our results first show the existence of detectable H 2 O 2 in buffer systems, which is a derivative product of superoxide radical (O 2 . ) under thermal incubation conditions. The O 2 . is the first intermediate product in a chain reaction of the reduction of molecular oxygen to water (Scheme 1) (15)(16)(17), and it can be converted to hydrogen peroxide (H 2 O 2 ) by redox active metal ions and/or superoxide dismutase (SOD) 2 introduced into buffer systems.
In addition to the previously observed Fab fragments (8,18), new fragments were observed (e.g. Glu 1 -Ser 224 and Glu 1 -Cys 225 of the heavy chain (HC) and Asp 1 -Glu 214 and Asp 1 -Cys 215 of the light chain (LC)). Among them, an additional sulfur group was attached to the side chain of the C-terminal Cys 215 LC and Cys 225 HC in the two fragments, respectively. In particular, a -CONH 2 rather than a -COOH at the C termini of the fragments of the Glu 1 -Cys 225 , Glu 1 -Ser 224 , and Asp 1 -Glu 214 was determined. These fragments were not the products of simple reduction of the interchain disulfide bond; rather, they were formed by radical reaction pathways that were not determined in mAbs previously. The results from analyzing the products allow us to propose new radical reactions that break the C-S bond of the side chain of Cys 225 HC or Cys 215 LC, yielding a sulfur radical adduct to the side chain of Cys 215 LC or Cys 225 HC and dehydroalanine in the counterpart. Subsequent radical reactions and hydrolysis reactions lead to cleavage of the peptide bond, yielding an amide at the new C terminus. Also, because H 2 O 2 is the inevitable by-product of radical reactions in buffer systems, related strategies of how to improve the stability of mAbs is discussed.

EXPERIMENTAL PROCEDURES
Material-The antibody used in this study is a recombinant fully human antibody of the IgG1 subclass. The molecule was expressed in Chinese hamster ovary (CHO) cells and chromatographically purified at Genentech, Inc. Examination of the primary sequence indicates that the antibody contains Glu at the N terminus of the HC, and an Asp at the N terminus of the LC.
Thermal Incubation of the IgG1-The IgG1 was buffer-exchanged into PBS (137 mM NaCl, 2.7 mM KCl, 10 mM phosphate, pH 7.4) and then incubated at 37°C, with or without 5 M Cu 2ϩ or Fe 3ϩ or SOD (200 units). Catalase (1000 milliunits; Sigma) was added to quench the reaction. In some cases, EDTA (50 mM) or 5,5-dimethyl-1-pyrroline N-oxide (100 mM) was added into the sample prior to the addition of Cu 2ϩ , Fe 3ϩ , or SOD.
Detection of H 2 O 2 Production-H 2 O 2 production was determined using the Amplex Red hydrogen peroxide/peroxidase assay kit (Invitrogen), as described previously (19,20). The buffer with or without Fe 3ϩ or SOD was incubated at 37°C and then mixed with Amplex Red working solution and incubated at room temperature, protected from light. The H 2 O 2 production was measured by fluorescence signals. Fluorescence emission excited at 545 nm was detected at 590 nm. All assays were performed in six replicates, and the results were reported as the average.

SEC Analysis and Purification of the Degradation Products-
The degraded products were separated by size exclusion chromatography (SEC) on TSK G3000SWxl dual columns, 7.8 ϫ 300 mm, at a flow rate of 0.5 ml/min. Eluting protein was monitored at 280 nm. The SEC running buffer contained 250 mM potassium phosphate, 200 mM potassium chloride, pH 6.2. The cleavage was measured by the relative percentage of the integrated peak area of partial molecules. SEC fractionation was carried out on an Agilent 1200 HPLC system (Agilent Technologies) equipped with a fraction collector. Purified fractions were pooled and concentrated by centrifugation in Millipore Centriprep YM-30 filter units with a 10,000 molecular weight cut-off (Millipore, Billerica, MA).

Reversed-phase Chromatography and Time-of-flight Mass Analysis (RP-HPLC-TOF/MS)-RP-HPLC-TOF/MS
was performed as described previously (7). Briefly, antibody samples were diluted to 1 mg/ml in 50 mM Tris-HCl (Sigma) at pH 8.0 with a final concentration of 3 M guanidine hydrochloride (Mallinckrodt). RP-HPLC was performed on an Agilent 1200 HPLC system. The mobile phase included water with 0.11% trifluoroacetic acid (TFA) as solvent A and acetonitrile (Burdick Jackson) with 0.09% TFA as solvent B. A Varian PLRP-S (Varian, Inc., Palo Alto, CA), 4.6 ϫ 50-mm, 8-m particle size, 1000-Å pore size column was used for the RP-HPLC-TOF/MS analysis. The column eluent was analyzed by UV detection at 215 nm and then directed in-line to a TOF mass spectrometer. The initial mobile phase was 25% solvent B for 5 min, and then a two-stage gradient was applied: 2% solvent B per min from 25 to 30% solvent B, followed by 0.3% solvent B per min from 30 to 42% solvent B. The separation was performed at 75°C at a flow rate of 0.5 ml/min. Electrospray ionization TOF/MS was performed on an Applied Biosystems QSTAR Elite XL mass spectrometer equipped with an Agilent 1200 HPLC system. The electrospray ionization mass spectra were analyzed using BioAnalyst protein deconvolution software (Applied Biosystems).
Protease Digestion and Peptide Maps-The IgG1 was denatured in the presence of 4 M guanidine hydrochloride for 5 min at 75°C. Prior to digestion, samples were buffer-exchanged into 50 mM Tris-HCl, pH 8.0, using Bio-Spin 6 columns (Bio-Rad) according to the manufacturer's instructions. Recombinant sequencing grade Asp-N or Lys-C (Roche Applied Science) was added to samples at an enzyme/protein ratio of 1:10 (w/w), and the samples were digested at 37°C overnight. Analytical peptide maps consisted of loading 50 g of the digest onto a Phenomenex Jupiter Proteo C18 column, 2.0 ϫ 250 mm, heated at 55°C. The separation was performed by gradient elution on an Agilent HP 1200 HPLC system. The column was held at the initial condition of 0.5% solvent A (0.11% trifluoroacetic acid in water) at a flow rate of 0.3 ml/min for 5 min, and then the digest was eluted with a linear gradient to 60% solvent B (0.09% trifluoroacetic acid in acetonitrile) over 160 min. The peptides FIGURE 1. Schematic illustration of the interchain disulfide bond between the heavy and light chain of an IgG1. The hinge cleavage sites that were reported previously (3)(4)(5)(6)(7)(8) are indicated by the arrows. The interchain disulfide bond is connected by Cys 215 LC and Cys 225 HC. Cys 231 HC is the primary radical site for hydroxyl radical attack. The amino acid sequence suggests that the Lys-C and Asp-N proteases are appropriate for peptide mapping (see "Results" for details). SCHEME 1. Reduction of molecular oxygen to water.

Breaking the Light and Heavy Chain Linkage of Human IGG1
were identified by data-dependent tandem MS fragmentation using a Thermo LTQ Orbitrap mass spectrometer.

Detection of H 2 O 2 as the Product of O 2 Reduction in the Buffer
System-Molecular oxygen (O 2 ) has two unpaired electrons, one on each oxygen atom. With the electron spin restriction, the univalent reduction of O 2 to O 2 . is a facile process (reaction 1 in Scheme 2). In fact, the reduction of O 2 into more reactive species like H 2 O 2 and ⅐ OH has been determined in a biological system (16,17,21). However, there is no report showing the presence of the O 2 . and its derivatives in a buffer system, probably due to the lack of an easy and direct assay to measure the O 2 . in a quantitative way.
We found that the chain reaction of the reductions of the O 2 also takes place in buffer solutions. Evidence in support of this reaction was obtained from the measurement of H 2 O 2 production using a method as described previously (19,20). The rate of the H 2 O 2 production is slow, but it can be accelerated by SOD and/or Fe 3ϩ to a level that can be easily detected (19,20) (Fig.  2A). The production of H 2 O 2 in PBS buffer is very limited because no evidence for a continuing accumulation of the H 2 O 2 was obtained during a 3-week incubation. In addition, the level of the H 2 O 2 production varies, depending on the buffer age. Similar observations were obtained from other buffers, such as Tris-HCl and acetate (not shown). We found that redox metal ions (e.g. Cu 2ϩ or Fe 3ϩ ) are required because EDTA blocks the H 2 O 2 production ( Fig. 2A) . /H 2 O 2 to the hinge cleavage of an IgG1 was obtained from analysis of an IgG1 incubated in phosphate-buffered saline at 37°C, using the same conditions as above. The products of the hinge cleavage were measured by SEC. As shown in Fig. 2B, conditions under which more H 2 O 2 is produced resulted in more cleavage products. Incubation of the IgG1 in PBS for 15 days induced a 3.7% level of hinge cleavage, whereas the addition of 5 M FeCl 3 into the reaction resulted in an increase of cleavage to 9.7%. A similar increase in cleavage was observed with the addition of CuSO 4 (not shown). In addition, SOD was also capable of accelerating the hinge cleavage. Incubation of the IgG1 with both Fe 3ϩ and SOD resulted in a cleavage level of ϳ17%, and inclusion of Cu 2ϩ further increased the cleavage to ϳ32%. However, EDTA was found capable of inhibiting Ͼ95% of the metal ion-induced cleavage, implying an involvement of the transitional metal ions in the reaction. In addition, all cleavage reactions could be stopped by catalase or blocked in the presence of 5,5-dimethyl-1-pyrroline N-oxide, a spin radical trap. Collectively, these results suggest an involvement of H 2 O 2 in the reactions and that the radical reaction mechanism may be responsible for the hinge cleavage, as described previously (8). An IgG molecule is itself capable of generating H 2 O 2 (19,20), which may lead to radical reactions that result in hinge cleavage. However, the self-produced H 2 O 2 has been determined to generate at the interface of the heavy and light chains with a SCHEME 2. Reactions to produce hydroxyl radical.  (19,20). B, hinge cleavage of an IgG1 under the same incubation conditions as in A. The cleavage was shown as a relative percentage of the peak area of the products to the whole molecule at 37°C for 15 days in the PBS buffer. The cleavage can be accelerated by the addition of Fe 3ϩ (5 M), or SOD (200 units) but inhibited by 5,5dimethyl-1-pyrroline N-oxide (10 mM) or EDTA (20 mM). Results represent the mean of six replicates. relative steady production rate (19,(22)(23)(24). Thus, the fact that redox metal ions and/or SOD mediate the H 2 O 2 production and the hinge cleavage excluded the possibility that the selfproduced H 2 O 2 was the main force to drive the hinge cleavage.
Analysis of the Cleaved Fab by RP-HPLC-TOF/MS-Although the radical mechanism may be responsible for the hinge cleavage, it remains to be determined if the thermal induced degradation generates the same products as those produced by H 2 O 2 mediated reactions. Garrison (14), Shimazu and Tappel (25), and Buxton et al. (26) had demonstrated that the products of radical reactions could be quite different under different reaction conditions. Characterization of the products would help to address these questions. To this end, cleavage products of the partial IgG1 and Fab were purified by SEC (Fig. 3A), and the Fab was first analyzed by RP-HPLC-TOF/MS under nonreducing conditions. Fig. 3B shows the RP-HPLC profile with three major products labeled by peaks 1, 2, and 3. TOF/MS analysis revealed that the peak 1 eluting at 16.5 min derived from the LC, which was characterized by multiple peaks with masses of 23 Da, respectively (Fig. 4B), which correspond to the fragment of Glu 1 -Ser 224 , the LC, and the LC with ϩ16 Da, ϩ32 Da, and 48 Da adducts. The absence of the ϩ64 and ϩ80 Da adducts under the reducing conditions suggests that these two adducts were disulfide bond-related, implying that the ϩ32 Da adduct observed under non-reducing conditions may be the result of gaining a sulfur group that was attached to the side chain of Cys. The peptide mapping confirmed this conclusion (see below). It has been known that atmospheric oxygen is capable of oxidizing the free -SH group of Cys (27,28). Earlier studies indicated that Cys-SH is heat-labile and can be oxidized into Cys-SOH and Cys-SO 2 H; these two products have been found to be stable and can be converted into Cys-SO 3 H as the ultimate product (29,30). It is possible that the observed adducts of ϩ16, ϩ32, and ϩ48 Da derive from oxidation during RP-HPLC-TOF/MS or sample processing, and these adducts were added on to the extra Sulfur group on the LC.
Analyzing the Fab portion of the H (Fd) under non-reducing conditions revealed very similar results. The Fd species eluting at 23.3 min was a mixture of two components with masses of 23,375.7 and 23,557.4 Da (Fig. 4C), corresponding to Glu 1 -Ser 224 and Glu 1 -Cys 225 with a ϩ78 Da adduct, respectively. Peptide mapping analysis indicated that the ϩ78 Da extra mass was attributable to an additional SO 3 group at the side chain and a -CONH 2 at the C terminus of Cys 225 (see below).
Unlike peaks 1 and 2, the Fab peak 3 eluting at 24 min in the RP-HPLC profile showed a ladder of cleavage sites with the upper hinge residues of Cys 225 -Asp 226 -Lys 227 -Thr 228 -His 229 at the C-terminal ends (Fig. 4D). The UV signal-based peak integration revealed a 1:1 ratio for the peak 3 (Fab) to the sum of the peaks 1 (LC) and 2 (Fd), suggesting that half of the products were generated by the expected breakage of the H-L bond under the thermal incubation conditions.   Table 1.

Breaking the Light and Heavy Chain Linkage of Human IGG1
JULY 15, 2011 • VOLUME 286 • NUMBER 28

JOURNAL OF BIOLOGICAL CHEMISTRY 24677
ucts should be an amide (-NH 2 ) rather than -COOH according to radical reaction mechanisms (8,18). To this end, Lys-C and Asp-N peptide mapping were conducted under non-reducing conditions to monitor the C-terminal peptides at the LC and Fd. As expected, no adducts of ϩ32 or ϩ48 Da were detected at residues except for the C-terminal Cys from the LC or Fd. As shown in Fig. 5A, three forms of the C-terminal peptide of the LC elute at ϳ21, 22, and 23 min in the Lys-C map, which correspond to SFNRGEC 215 with a ϩ48 Da adduct at 21 min, the same peptide with a ϩ32 Da adduct at 22 min, and the peptide SFNRGE 214 at 23 min, respectively. The observation of the series of b ions (b4, b5, and b6) and y ions (y4 and y5) provides the necessary data to assign the location of the ϩ48 Da adduct at the side chain of Cys 215 , as shown in Fig. 6A. Similarly, the ϩ32 Da was assigned to the side chain of Cys 215 by the b ions (b2, b4, and b6) and y ions (y2 and y5) (Fig. 6B). Combining the results from RP-HPLC-TOF/MS, the ϩ48 and ϩ32 Da adducts could be explained by a -SOH and -S group at the side chain of Cys 215 . Fig. 6C shows the MS/MS spectrum of the peptide SFNRGE 214 , a truncated form of SFNRGEC 215 . The observed y ions (y3, y4, and y5) clearly indicate a CONH 2 group, rather than COOH, at the C terminus. Among these three forms of the  Fig. 3A are presented in A, C, and D, respectively. B, the deconvoluted mass spectrum of the reduced LC. Each major peak is labeled with the observed mass, and corresponding fragments are summarized in Table 1. C-terminal peptide of the LC, it appears that the truncated form accounts for ϳ25%, whereas the full-length modified forms account for ϳ75% of the C-terminal peptide ( Table 2). In contrast to the LC, the C-terminal peptide of Fd contains more of the truncated form DKKVEPKS, approximately ϳ75%, which eluted at 17.8 min in the Asp-N map, as shown in Fig. 5B. The minor form of the C-terminal peptide DKKVEPKSC, eluting at 20.7 min, accounted for ϳ25%, with a ϩ79 Da adduct. The MS/MS spectrum provides clear evidence that the ϩ79 Da adduct (-SO 3 ) is located at Cys 225 . The most dominant ion of m/z 557.33 was observed in the full spectrum, which corresponds to the [MH-SO 3 ] 2ϩ form. In addition, the observed b ions (b5 and b7) and y ions (y2, y3, and y7 ions and y4-SO 3 , y6-SO 3 , and y7-SO 3 ) indicate that the -SO 3 group is located at the side chain of Cys 225 with a -CONH 2 at the C terminus (Fig.  6D). Fig. 6E shows the MS/MS spectrum of the truncated peptide DKKVEPKS, and a series of y ions (y3 and y5-y7) and b ions (b4 -b7) clearly indicate an amide at the C terminus. Table 2 summarizes the quantitative information of each form of the C-terminal peptides. The fact that more truncated forms of the L corresponded to more Fd fragments with an extra -S at Cys 225 , or vice versa, suggests that these products were generated through a loss or gain of the sulfur group at the side chain between these two Cys residues. In other words, a fragment without Cys in the C terminus could be the result of the gain of a sulfur group at the side chain of Cys in the counterpart. The truncated fragment may be explained by hydrolysis reactions at the dehydroalanine formed by the loss of the sulfur group at Cys as described previously (14). Finding a mixture of -SOH and -S adducts at the side chain of Cys 215 implied that these products were probably generated by limited oxidation under low oxygen tension. Collectively, these observations indicated that half of the products are the disulfide-bonded population with the ladder cleavages of the C-terminal residues in the Fd, as described previously under high oxygen tension (7), and the other half are non-disulfide-bonded LC and Fd fragments that were generated by the unique breakage of the H-L bond under low oxygen tension.

TABLE 1 The average masses of the Fab fragments analyzed by RP-HPLC-TOF/MS
Analysis of the Partial IgG1 and Main Peak Purified from SEC Isolation by RP-HPLC-TOF/MS-The H 2 O 2 -mediated radical reactions under high oxygen tension also release the LC species by the reduction of the H-L bond (8). If such reduction takes place under the thermal incubation conditions, a released LC from the partial molecule or the main peak would be expected. To this end, the partial IgG1 and main peak were analyzed under non-reducing conditions by RP-HPLC-TOF/MS. As shown in Fig. 7A, the non-disulfide-bonded LC eluted at 16 min in the RP-HPLC profile. The level of the LC varied in the samples; the control sample (non-treated sample) showed a level of 1.6%, whereas the main peak and the partial IgG1 showed an increase to ϳ2.3 and 2.5%, respectively. TOF/MS analysis indicated molecular masses of 23,360.6 and 23,392.0 Da for the control sample, corresponding to a reduced form of the LC and the LC with a ϩ32 Da adduct (Fig. 7B). The main peak revealed three species with masses of 23,256.9, 23,360.6, and 23,391.7 Da, corresponding to Asp 1 -Ser 214 , a reduced form of the L, and the reduced LC with sulfur adduct, respectively (Fig. 7C). However, the partial IgG1 showed multiple species that are similar to the LC of the cleaved Fab, with the main component as the reduced form of the LC (Fig. 7D). The additional species in the partial IgG1 and main peak that are similar to those in the cleaved Fab suggested that the thermal incubation-induced radical breakage of the H-L linkage could occur without any cleavage of the backbone.

Radical Reaction Mechanism for the Thermal Induced Hinge
Degradation-Many different physiological and environmental processes can lead to the formation of H 2 O 2 and O 2 . . H 2 O 2 has been found responsible for degradation of many proteins through radical reaction mechanisms (31)(32)(33). However, the O 2 . is a relatively unreactive species whose major type of reaction is reduction (i.e. donation of an electron) (21,34). Estimated in vivo concentrations of H 2 O 2 and O 2 . have been reported as 10 Ϫ7 to 10 Ϫ9 M and 10 Ϫ11 M, respectively (35)(36)(37), whereas these two species do not interact to produce hydroxyl radical ( ⅐ OH) (38). In the chain reaction of the reduction of oxygen into water, the overall reaction rate is proportional to collision frequency, so O 2 . and H 2 O 2 fluxes depend directly upon the ambient concentration of oxygen (15)(16)(17)39). Thus, low oxygen tension may allow thermal incubationinduced radical reactions to take place via different routes and generate products of a different nature from those generated under high oxygen tension. Like all chemical reactions that generate a product, the accumulation of the product (e.g. H 2 O 2 ) results in inhibition of its further production. Even in the presence of a catalyst (e.g. SOD or redox metal ion), the production of H 2 O 2 would be limited unless there is a continuous supply of oxygen and sequential consumption of the H 2 O 2 . In Reactions 1-5 of Scheme 2, the regeneration of oxygen and H 2 O 2 enables the degradation of an IgG1 hinge, which in turn consumes the H 2 O 2 and ⅐ OH. This may explain why an accumulation of the cleavage products was observed, although the buffer itself has a limited capacity to produce H 2 O 2 .
New Pathways for the Breakage of the Light and Heavy Chain Linkage-Observing the L and Fd with an extra sulfur at Cys 225 or Cys 215 and fragments without these two Cys residues sheds light on new radical reaction pathways for the thermal incubation-induced degradation of the heavy-light chain linkage. It is reasonable to believe that the unique degradation of the H-L linkage also initiates by breaking the first hinge disulfide bond, and radical formation in one of the first hinge Cys and electron transfer leads to localization of the electron to the H-L bond. Although long range electron transfer has been observed in some proteins, none has shown an electron transfer capable of bypassing a disulfide bond (40 -42). Thus, these fragments must be generated by new pathways via breakage of the H-L bond linkage and cleavage of the backbone. Based on the previous work described by others (13,14,43,44), reaction of cystine (a disulfide bond) with an electron is known to be extremely rapid, with the reaction rate constant of 1.6 ϫ 10 10 M Ϫ1 s Ϫ1 , faster than with ⅐ OH at 2.1 ϫ 10 9 M Ϫ1 s Ϫ1 (13,26). Similar to other proteins described previously (45,46), an electron with elevated energy transferred from Cys 231 to the H-L of an IgG1 results in reduction of the H-L bond under high oxygen tension (18). However, under low oxygen tension, electron addition to a residue leads to the formation of a radical anion that can undergo a number of reactions depending on the conditions and substrate (47,48). Although the chemical consequences of e Ϫ attachment to the disulfide bonds of proteins are not fully understood (14), it has been demonstrated that electron addition to the thiol groups of Cys and Met leads to efficient cleavage of the C-S bond with elimination of the -SH group to give a carbon-centered radical and a thiol anion (13,26,49). Such reactions are slow, with low yields of minor products arising from cleavage of the C-S-C link, but produce significant amounts of cleavage products under low oxygen tensions (13).
Based on the results presented in this study, we propose the following pathways for the degradation of the H-L linkage in an  (Ref). Partial, the partial IgG1; Main, the main peak purified by SEC. The LC was eluted at ϳ16 min, as indicated by an arrow. B, the deconvoluted spectrum of the LC from the control. C, the deconvoluted spectrum of the LC from the partial IgG1. D, the deconvoluted spectrum of the LC from the main peak.

TABLE 2
The C-terminal peptides in the L and Fd determined by Lys-C and Asp-N peptide maps The estimation of abundance for each peptide was based on the individual integrated percent peak area relative to the total integrated peak area of the C-terminal peptide derivatives from the LC or Fd. ND, because the amino acid sequence SFNRGEC is the C-terminal peptide of the LC, no -CONH 2 was determined. IgG1. In the first step, electron addition to the H-L bond results in the formation of a disulfide bond radical anion (Mechanism 1, reaction 1). Decomposition of the anion results in breakage of the C-S bond of the side chain of Cys 225 HC or Cys 215 LC that is in the H-L bond (Mechanism 1, reactions 2a-2d). With the low oxygen tension, some sulfur groups are oxidized, and the rest remain in a reduced form. In the counterpart, the Cys that lost the sulfur group forms a carbon-centered radical that is the same species as a result of hydrogen abstraction from the side chain of Ala, forming unstable dehydroalanine. Hydrolysis of the dehydropeptide yields an amide at the N-1 position of the new C terminus (Mechanism 1, reaction 3). Meanwhile, as demonstrated by Garrison (14), the addition of e Ϫ to the carbonyls (-C ␣ ONH-) of peptides in oxygen free solutions occurs in competition with e Ϫ addition to the disulfide bond (14); such competition results in the formation of a new intermediate radical species, ⅐ C ␣ (OH)NH-, at Cys (Mechanism 1, reaction 4). Dissociative deamidation of this new radical species occurs along the N-C ␣ bond that cleaves the peptide bond to generate a -CONH 2 at new C terminus (Mechanism 1, reaction 5). Such mechanisms have been described by others (25, 26, 49 -52).

Peptide
It should be pointed out that reaction 2d (Mechanism 1) only releases the LC without any cleavage of the HC because no radical remains at the side chain of Cys 225 after decomposition of the radical anion. This reaction could also rationalize the observation of the LC and LC fragment in the partial IgG1 and main peak as shown in Fig. 7. Based on reactions 2a-2c (Mechanism 1), a theoretical ratio of 2:1 for the fragments of Glu 1 -Ser 224 to Glu 1 -Cys 225 in the Fd, and Asp 1 -Glu 215 to Asp 1 -Cys 214 in the LC would be expected, which is not in good agreement with the observed ratio of 3:1. A possible explanation could be that, as indicated by Garrison (14), an electron radical addition to ␥-carbon sites on the side chain of certain residues could occur and also lead to the formation of the dehydropeptide. In such a case, the side chain of Cys 225 HC would be the target for the electron addition, and the formed dehydroalanine would be hydrolyzed by reaction 3 (Mechanism 1), leading to backbone cleavage and producing a new amide at the N-1 position.
As indicated by Davies and Delsignore (34), radical attack results in gross distortions of secondary and tertiary structure of a protein, and such changes had been observed in an IgG1 under high oxygen tension (18). It is reasonable to believe that radical reactions under the thermal incubation conditions also alter local conformation or conformational dynamics of an IgG1. Although the distance between Cys 231 and the H-L bond of ϳ16 -17 Å is greater than a typical electron tunneling distance of 14 Å (18), the new local conformation or conformational dynamic in the upper hinge region could bring the H-L bond more proximal to Cys 231 , which would allow interaction between the electron and the H-L bond or electron transfer from the transient radical center His 229 to the H-L bond (18). Collectively, our results demonstrated that the thermal incubation-induced degradation of the IgG1 hinge follows radical reaction mechanisms with a combination of pathways taking place under high and low oxygen tension that results in different products.

Implication of H 2 O 2 and O 2 . in the Development of mAb
Therapeutics-The thermal incubation-induced hinge degradation presents great challenges for the stability programs in the development of mAb therapeutics. Given the highly conserved hinge sequence in IgG1 molecules, radical reaction mechanisms that may compromise the stability, safety, and efficacy of an IgG1 need to be better understood and controlled appropriately. Because O 2 . and H 2 O 2 species are the inevitable by-products of the reduction chain reaction of oxygen into water, new strategies are needed to minimize the impact of these species on the stability of an IgG1. Preventing metal ions from redox cycling is one mechanism to inhibit the production of hydrogen peroxide in the formulation development and during storage; this can be done by using reagents that have the least content of redox metal ions or by using metal-chelating reagents. Alternatively, substitution of the "hot spot" residues in the upper hinge is an attractive means of resisting radicalinduced fragmentation based on our previous substitution results (18). Further investigation of the influence of the upper hinge residues and evaluation of some promising mutants could enrich our understanding of mAb degradation and provide new insights into how to engineer a new generation of therapeutic IgG1 that is capable of resisting such degradation around the H-L bond.