KEYWORDS
Antibody, Disulfide bond, Reduction/Oxidation, Process development
1 INTRODUCTION
Antibodies are proteins produced by immune cells and help to defend the host when facing foreign invaders (Dowd, Halonen, & Maier, 2009). Monoclonal antibodies (mAbs) are antibodies with single specificity generated from plasma B cells in vitro (Nelson et al., 2000). Compared with small chemical molecules, mAbs showed advantages such as high selectivity and potency, which improve therapeutic efficiency and reduce the toxicity (Cui, Cui, Chen, Li, & Guan, 2017; Imai & Takaoka, 2006). Due to these advantages, commercial mAbs have been developed rapidly since U.S Food and Drug Administration (FDA) approved the first mAb product in 1986 (Ecker, Jones, & Levine, 2015). Based on Nature Reviews Drug Discovery reports (Hughes, 2010; Mullard, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2018, 2019, 2020), in the last ten years (2009 - 2019), 61 out of 95 FDA approved Biologics License Applications (BLAs) were mAb products or mAb related products, and the percentage of mAb (or mAb related) products in the BLAs increased over time (Figure S1).
Commercial mAbs are generally produced in mammalian cells such as Chinese hamster ovary (CHO) (Du et al., 2018; Kao, Hewitt, Trexler-Schmidt, & Laird, 2010; T. Wang, Liu, Cai, Huang, & Flynn, 2015). Ideally, the disulfide bonds of antibody should pair correctly before mAbs are secreted extracellularly. Correct disulfide bonds pairings are critical during the mAb development process: (1) to meet ICH Q6B specifications for biotechnological and biological products that the number and the position of disulfide bridges should be determined based on the gene sequence for the desired product (Lakbub, Shipman, & Desaire, 2018); (2) to ensure antibody drug therapy efficiency and eliminate the immunogenicity (Kao et al., 2010; Swope et al., 2020). It is also critically important for bi- and tri- specific antibody (BsAb and TsAb) stability improvement during the drug product development (Caravella & Lugovskoy, 2010; Rossi et al., 2006; Vaks, 2018). Many research studies have been conducted to analyze and understand the function of disulfide bond structure on antibody stability, and there are several reviews on this topic (Correia, 2010; H. Liu & May, 2012; Trivedi, Laurence, & Siahaan, 2009b). While developing the fundamental understanding of disulfide bond structure and function, great research efforts have been made to identify the root cause of disulfide bond reduction happened during mAb manufacturing process, and develop measures to minimize the reduction as well as to recover the reduced mAb during the downstream process (Du et al., 2018; Hutterer et al., 2013; Kao et al., 2010; Mun et al., 2015; Tan et al., 2020; Tang et al., 2020; Trexler-Schmidt et al., 2010). As far as the authors are aware of, most of these studies focused on certain downstream process steps instead of the whole downstream process, with relatively limited discussions on the effect of disulfide bond reduction on downstream processing. In reality, it is rather challenging to completely eliminate disulfide bond reduction in a single downstream process step. Therefore, to minimize disulfide bond reduction, understanding how to apply the mitigation strategies across multiple downstream process steps is necessary.
To accommodate the aforementioned need, this paper first reviews the root causes of disulfide bond reduction, and then discusses how existing mitigation strategies have been able to address this issue. By summarizing these mitigation strategies, we are able to provide a work chart to bridge disulfide bond reduction mitigation strategies during mAb downstream processing. A case study will then be presented as an example to illustrate how these approaches were applied to downstream manufacturing process. In addition, this paper also discusses the effects of disulfide bond reduction on downstream process, and the associated analytical methods for disulfide bond analysis, and future perspectives, based on our own experiences, in addressing the disulfide bond reduction challenges for mAbs and multi-specific antibodies, all of which provide a broader view of disulfide bond reduction challenge in mAbs and multi-specific antibodies downstream processing. This article provides a useful resource for people in biotech industry who are facing the challenge of disulfide bond reduction during the antibody process development.
2 DISULFIDE BONDS in Therapeutica Proteins
2.1 Disulfide bond structure in mAbs
A major advancement in antibody structure was revealed in the 1960s by Gerald M. Edelman and Rodney R. Porter, who were awarded 1970 Nobel Prize “for their discoveries concerning the chemical structure of antibodies” (Edelman & Gall, 1969; Edelman & Gally, 1962; Preud’homme, Petit, Barra, Morel, & Lelièvre, 2000; Raju, 1999). There are five subclasses of antibodies (immunoglobulins): IgA, IgD, IgE, IgG and IgM. Each subclass of antibodies are composed of one or several immunoglobulin (Ig) monomers. An Ig monomer has two identical heavy chains and two identical light chains. Figure 1A and 1B showed an example of the disulfide bonds presented in IgG monomers (Correia, 2010; H. Liu & May, 2012). Both heavy chains and light chains are composed of constant domain and variable domains, which are constructed from two beta-sheets (\(\beta\)-sheets). Disulfide bonds that connect the two\(\beta\)-sheets in a single domain are known as intra-chain disulfide bonds (Figure 1A)(W. Li et al., 2016). Disulfide bonds that connect two heavy chains or connect a light chain and a heavy chain are known as inter-chain disulfide bonds. In an IgG monomer (Figure 1B), there are twelve intra-chain disulfide bonds (one per domain), two inter-chain disulfide bonds between light chain and heavy chain, and two to eleven inter-chain disulfide bonds between two heavy chains. Researchers found that inter-chain disulfide bonds are more prone to degrade than intra-chain disulfide bonds (Kikuchi, Goto, & Hamaguchi, 1986; H. Liu, Chumsae, Gaza-Bulseco, Hurkmans, & Radziejewski, 2010). There are two possible reasons for the higher stability of intra-chain disulfide bond: (1) as shown in Figure 1A, intra-chain disulfide bond are buried inside the two \(\beta\)-sheets (Amzel & Poljak, 1979) and the accessible area of intra-chain disulfide bonds have been calculated to be zero (Kikuchi et al., 1986); (2) Molecule Dynamic (MD) simulation found that the atom distance of sulfur molecules for intra-chain disulfide bond is shorter than that for inter-chain disulfide bond, which may benefit the intra-chain disulfide bond higher stability than inter-chain disulfide bond (X. Wang, Kumar, & Singh, 2011).
2.2 Disulfide bond structure in BsAbs and TsAbs
BsAbs and TsAbs (also known as multi-specific antibodies) are artificially designed complex antibodies that are capable of binding two or more antigens (Brinkmann & Kontermann, 2017; Kontermann & Brinkmann, 2015; Runcie, Budman, John, & Seetharamu, 2018; Sedykh, Prinz, Buneva, & Nevinsky, 2018). They are thought to have improved therapy efficiency since they have the ability to bind to two or more different targets simultaneously (Runcie et al., 2018; Tustian, Endicott, Adams, Mattila, & Bak, 2016; Wu et al., 2020). Due to this advantage, multi-specific antibodies development has gained increasing attention in recent years. By 2019, there are more than 110 BsAbs reported under clinical development (Nie et al., 2020).
Disulfide bond structures also exist in BsAbs and TsAbs. Since there are more existing studies on BsAbs than TsAbs, here we use BsAbs as examples to illustrate disulfide bond function in multi-specific antibodies. Besides the similar functions that disulfide bonds exhibit in mAbs, disulfide bonds play a critical role in avoiding BsAb chain mispairing and keeping BsAb structure stable. Figures 1C-1E show three examples that illustrated disulfide bonds roles in two subclasses of BsAbs, immunoglobulin G (IgG)-like BsAb (having an Fc region, Figure 1C and Figure 1D) and small BsAb (lacking an Fc region, Figure 1E)(Elgundi, Reslan, Cruz, Sifniotis, & Kayser, 2017; Kontermann & Brinkmann, 2015).
First example shows how the disulfide bonds improve the stability of KIH IgG-like BsAb (Figure 1C). In an IgG-like BsAb design, one way to correctly pair the heavy chains from different antibodies is the “Knobs-into-holes” (KIH) approach: in one heavy chain where a small amino acid in the CH3 domain is replaced by a large amino acid (functioned as “knobs”) and in the other heavy chain where a large amino acid is replaced by a small amino acid (functioned as “holes”)(Brinkmann & Kontermann, 2017). The stability of this heterodimeric region is further improved by forming additional disulfide bonds in this region. Some research reported that inducing the disulfide bond to the KIH design can improve functional BsAb yield by 10%, and overall higher than 90% functional BsAb yield (Carter, 2001; Klein et al., 2012).
In the second example, disulfide bonds improve the rate and efficiency of correctly pairing heavy chain with light chain in IgG-like BsAb (Figure 1D). Besides forming heterodimeric heavy chain pair, two light chains also need to correctly pair with the corresponding heavy chains. One way to improve the correct light chain-heavy chain pairing is to introduce an artificial disulfide bond in one arm. Researchers mutated the pair of cysteines, forming a disulfide bond between heavy chain and light chain constant regions, to valines, and introduced a new pair of cysteines in different location in heavy chain and light chain constant regions to form an engineered disulfide bond. In this way, the mutated light chain can only pair with the mutated heavy chain and vice versa, the un-mutated light chain can only pair with the un-mutated heavy chain (Mazor et al., 2015).
Last example illustrates the disulfide bonds link the two heavy chain variable region in Dual-affinity Re-targeting (DART) small BsAb design (Figure 1E). DART BsAb is another type of the BsAb designs (Nie et al., 2020). As shown in Figure 1E, in DART design, light chain variable region from one antibody was linked with heavy chain variable region from the other antibody by small peptide, and the two heavy chain variable regions were linked by disulfide bond. The short linker sizes of DART design can improve small BsAb stability and reduce immunogenicity (Johnson et al., 2010).
3 IMPACT OF DISULFIDE REDUCTION ON DOWNSTREAM PROCESSING
As discussed above, correct disulfide bond formation is required for proper antibody folding and maintaining their bioactivity and stability (Lakbub et al., 2016). Researchers found that the unpaired free thiols could form incorrect disulfide bonds and result in covalent aggregation (Andya, Hsu, & Shire, 2015; Cromwell, Hilario, & Jacobson, 2006; Vázquez-Rey & Lang, 2011; W. Zhang & Czupryn, 2002). Protein aggregates in the final drug products could induce adverse immune responses in patients and cause immunogenicity (Moussa et al., 2016). The mechanisms of how protein aggregations cause immunogenicity is a separate but interesting research topic (Bessa et al., 2015; Moussa et al., 2016; Ratanji, Derrick, Dearman, & Kimber, 2014; Rosenberg, 2006) and out of the scope of this paper. Instead, to the best of our knowledge, there are limited studies reporting disulfide bond reduction impact on whole downstream processing performance. As such, this section we discuss disulfide bond reduction impact on downstream processing, including the challenge of identifying reduced mAbs and disulfide bond reduction effects on mAbs stability based on our own commercial antibody process development experiences.
3.1 Identification of reduced mAb during the downstream processing
Besides the incorrect disulfide bond formation, the unpaired free thiols may change the antibody’s surface charge distribution and hydrophobicity, which consequently will result in different chromatographic profiles during mAb process development. For example, in cation exchange chromatography - high performance liquid chromatography (CEX-HPLC) and anion exchange chromatography (AEX-HPLC) analysis, mAbs with and without free thiols showed separate peaks at different elution times (Chen, Nguyen, Jacobson, & Ouyang, 2009; Cheng et al., 2017; Pristatsky et al., 2009; T. Zhang et al., 2012). However, to the best of our knowledge, there are very few papers that reported the effects of disulfide bond reduction on the performance of the downstream processing, particularly whether the disulfide reduced mAbs can be identified during downstream processing. Here, we use three monoclonal antibodies (mAb1 (IgG1, pI 8.2), mAb2 (IgG4, pI 8.0), mAb3 (IgG4, pI 6.8)) purified in our in-house studies as examples to show the impact of disulfide bond reduction on individual downstream purification processes, including Protein A chromatography, low pH viral inactivation (VI) and ion exchange chromatography (IEX) respectively.
(1) Protein A chromatography . For Protein A chromatography separation, the affinity between mAb and Protein A ligand is primarily through the Fc region of the antibody. Since the high-order structure of the disulfide-reduced mAb is intact, the affinity between reduced mAb and Protein A resin would remain unchanged (Tan et al., 2020). As presented in Figure 2A, the Protein A chromatography profile of the low level disulfide reduced mAb3 sample (containing 89.5% intact mAb3) and the high level disulfide reduced mAb3 sample (containing 10.6% intact mAb3) were identical. Additionally, no noticeable process yield difference was observed between the reduced mAb and intact mAb on Protein A chromatography step. These results confirmed that disulfide reduction had no major impact on the Protein A chromatography step.
(2) Low pH viral inactivation (VI) . A previous study showed that high level of disulfide bond reduction may increase the aggregation level after low pH viral inactivation (Chung et al., 2017). However, a clear trend was not observed for disulfide reduction effects on protein aggregation based on mAb1 and mAb2 low pH viral inactivation results. As shown in Figure 3A, after low pH viral inactivation the high molecular weight (HMW) level increased by 0.6% and 0.5% for intact mAb1 pool and reduced mAb1 pool, respectively. In contrast, the HMW level decreased by 1.7% for the intact mAb2 pool verses a decrease of 0.3% for the reduced mAb2 pool. The lack of consistent trend for disulfide reduction impacts on aggregation at the low pH VI step indicates that there may be other more critical contributing factors to the aggregation formation, such as intrinsic molecular properties (e.g. mAb type and hydrophobicity).
(3) IEX . As discussed above, disulfide reduction may change mAb surface charge distribution, which subsequently lead to the appearance of separate peaks in CEX-HPLC and AEX-HPLC analysis. Here, we compared AEX profiles and CEX profiles respectively for two runs of mAb1 with different levels of disulfide reduction (Figure 2B and 2C). AEX step of mAb1 was operated in flow through (F/T) mode, and the results in Figure 2B showed that the overall patterns of AEX are similar between the low level disulfide reduced mAb sample (>95% intact mAb) and the high level disulfide reduced mAb sample (67% intact mAb). The differences in the F/T volumes during the loading step were due to differences in the initial loading sample volumes. Similarly for CEX step (bind elute mode, performed after AEX), disulfide reduction level did not show significant impact on the overall CEX profile (Figure 2C). However, it was noticed that the mAb1 sample having had a high level of disulfide reduction showed longer peak tailing than the ones that had a low level of disulfide reduction at the end of the AEX loading step (Figure 2B) and at the end of the CEX elution step (Figure 2C), respectively. The differences in the peak tailing for both AEX (F/T) and CEX (B/E) could be due to the difference of disulfide reduction levels. However, it is not clear whether this type of chromatographic difference generally exists in all types of mAbs. In fact, the difference in the chromatographic profiles between the intact mAb and reduced mAb is so subtle that it would be very challenging to identify the highly reduced mAb just based on chromatography profiles alone. Thus, fast but accurate analytical methods should be used in the downstream purification process for disulfide bond reduction monitoring and determination. These methods will be discussed in more detail in Section 4.
Besides comparing intact mAb and highly reduced mAb profiles at different downstream process unit operation step, it was also noticed that for the highly reduced sample of mAb1 and mAb2, the purity increased from 64% and 35% in the Protein A chromatography pools to 94% and 92.9% in IEX pools respectively (Figure 3B). The increase of mAb purity during downstream purification process may be due to removal of reduced mAbs species during downstream purification, and/or reoxidation of reduced mAbs during downstream purification. To confirm this, we did conduct a study that used the disulfide re-oxidation strategy to recover the previously reduced mAb in downstream process (Tan et al., 2020; Tang et al., 2020).
3.2 Disulfide bond reduction impact mAb stability in downstream processing
Disulfide bonds enhance protein stability (Bulleid & Ellgaard, 2011; Chung et al., 2017; Zhu, Dupureur, Zhang, & Tsai, 1995). It is reasonable to expect that improper disulfide bond formation and disulfide bond reduction would impact process performance, protein stability and biological functionality. (Chung et al., 2017; Fass, 2012; H. Liu et al., 2014; Manteca et al., 2017; Trivedi, Laurence, & Siahaan, 2009a; T. Wang et al., 2015). Disulfide reduction generates free sulfhydryl groups on the cysteine residue, potentially resulting in decreased thermal stability, pH stability, and potency as well as elevated aggregation (Harris, 2005; Huh, White, Brych, Franey, & Matsumura, 2013; Lacy, Baker, & Brigham-Burke, 2008). However, in our internal low pH viral inactivation (VI) study, we did not observe a general correlation between the aggregation level at the post-low pH VI step and the initial low molecular weight percentage resulting from disulfide bond reduction (as discussed in Section 3). It was suggested that the sensitivity of antibody aggregation to disulfide bond reduction may be dependent on the characteristics of the molecule (Manteca et al., 2017). In our study, we performed thermal and photostability studies using samples containing different levels of starting LMW species (resulting from disulfide reduction). As shown in Figure 4, light exposure led to significant increase of aggregate formation in comparison to the high temperature and room temperature exposures. Specifically, a higher aggregation rate was observed for the sample containing initial higher level of LMW species under the light exposure condition. In contrast, both initial high and low levels of LMW samples did not show significant difference of aggregation rate at room temperature. At high temperature (40 °C), the overall aggregation levels trended higher over time and the initial LMW level also influenced the aggregate formation rate, but at a much slower rate compared to the light exposure condition. The significant impact of the LMW level on the sample photostability highlighted the importance of controlling the disulfide bond reduction in the manufacturing process.
4 ROOT CAUSE ANALYSIS FOR DISULFIDE BOND REDUCTION
MAb disulfide bond reduction is essentially an oxidation-reduction (redox) reaction that involves redox enzyme. Glutathione and thioredoxin system (comprising thioredoxin, thioredoxin reductase and nicotinamide adenine dinucleotide phosphate (NADPH)) are the known enzymes, and enzyme systems, that contribute to disulfide reduction (Chakravarthi & Bulleid, 2004; Handlogten, Zhu, & Ahuja, 2017; Koterba, Borgschulte, & Laird, 2012; O’Mara et al., 2019). NADPH is generated from the pentose phosphate pathway and serves as an electron source in the disulfide bond reduction reaction (Arne Holmgren & Lu, 2010). Electrons first transfer from NADPH to thioredoxin reductase (TrxR) and reduce the TrxR disulfide bond, then move to the oxidized thioredoxin (Trx) to form the reduced thioredoxin, and finally reduce the mAb disulfide bond. Glutathione (GSH) catalyzes the disulfide bond reduction in a similar way as Trx system (Gilbert, 1995; Handlogten et al., 2017; Meister & Anderson, 1983) (Figure S2). Therefore, during manufacturing processing including both upstream and downstream, factors that promote these enzyme-catalyzed reactions can lead to disulfide bond reduction. With this root cause analysis, we categorized the factors that influence mAb disulfide bond reduction into the following types: