We identified the presence of free thiols and total free thiol content variability in a recombinant monovalent one-armed IgG1 antibody (mAb1) expressed in E. coli. Observed free thiol content variability of mAb1 DS materials was due to the variable levels of species with predominantly two unpaired intra-chain cysteine residues in the CH2 domain. This distribution is different compared to distribution of free thiols among IgG domains of several CHO-expressed recombinant IgGs, where the CH3 domain had the highest level of free thiols followed by CH1, CH2 and the variable domain in the heavy chain . It is also distinct from reported more unique cases of free thiols corresponding to intra-chain cysteine residues in the heavy chain variable domain [19, 20]. The intra-chain location of free thiols in the CH2 domain (where disulfide bond would normally be buried between two anti-parallel β-sheets and would not be solvent exposed) suggested decreased reactivity . This is consistent with our observations that Fc free thiols were not accessible to thiolreactive Ellman’s reagent under nondenaturing conditions (levels were below limit of quantitation, ~ 0.1 mole free thiols/mole protein, of the assay), were not susceptible to in vitro reoxidation, and remained stable during long-term storage of mAb1 drug product.
Experiments to identify the process step(s) responsible for free thiol formation determined that production culture and downstream purification did not contribute to variability of the free thiol levels and that one or more of the harvest steps was a culprit. We demonstrated that the longer hold time and increased temperature of E. coli homogenate can lead to the increase in total free thiol content of mAb1. The location of these induced free thiols predominantly in the CH2 domain, matched that of DS materials, thus corroborating that free thiol variability of DS materials was likely introduced when the product was in the homogenate. Additional information of high practical importance was obtained when we determined that flocculation, which removes insoluble cell debris and host cell proteins, completely blocked free thiol formation. This indicated that more stringent process control of the homogenate (hold time and temperature) prior to flocculant addition should be able to minimize mAb1 free thiol content, as well as decrease its variability. Indeed, subsequently designed and implemented process enhancements controlling those parameters resulted in low free thiol content of DS materials and ensured improved control over the consistency of free thiol levels in mAb1 produced at manufacturing-scale.
Mechanistically, the ability of the homogenate to induce Fc free thiols suggested that free thiols in DS materials were not the result of incomplete disulfide bond formation during assembly of mAb1, but rather that they were caused by the subsequent reduction of disulfide bonds. A dialysis experiment indicated that the mAb1 disulfide bond reduction in homogenate was enzyme mediated, and thus genetic studies were designed to identify enzyme(s) that reduce mAb1 disulfide bonds in the homogenate. Thioredoxin and glutaredoxin pathways are two well-known pathways responsible for the reduction of disulfide bonds in the cytoplasm of wild-type E. coli . Inhibition of free thiol formation by the ebselen, a known competitive inhibitor of thioredoxin reductase (trxB gene product) , initially suggested involvement of thioredoxin pathway. However, single mutants of thioredoxin pathway (∆trxB, ∆trxA) did not show a decrease in the rate of free thiol formation, indicating thioredoxins may not be involved. Similarly, the glutaredoxin pathway mutant, ∆gor, also did not result in a decrease in the rate of free thiol formation. Modest effect of the double mutant (∆trxA ∆gor) was likely due to the low mAb1 concentration in homogenate, resulting in the lower rate of the enzymatic reaction. This apparent lack of impact from both reducing pathways seemed surprising; however, it appeared to be consistent with the observed selectivity of homogenate reduction for the intra-chain disulfide bonds rather than inter-chain bonds (hinge region and heavy chain-light chain disulfides). E. coli and mammalian thioredoxins (structurally similar to bacterial thioredoxins), can reduce more easily accessible inter-chain disulfide bonds in IgGs and not intra-chain disulfides [10, 11]. Similarly, glutaredoxins, the glutaredoxin pathway’s terminal enzymes also belonging to thioredoxin fold class, have been implicated in reduction of inter-chain disulfide bonds of IgGs expressed in CHO cells .
The apparent lack of involvement of thioredoxin and glutaredoxin pathways shifted our focus to the disulfide bond isomerase C (DsbC), one of the two periplasmic enzymes over-expressed to promote correct folding and disulfide bond formation in mAb1. DsbC is soluble homodimeric protein with two C-terminal catalytic domains containing the thioredoxin fold . DsbC can function as disulfide bond isomerase responsible for rearranging incorrectly formed (non-native) disulfide bonds [4, 23–25] introduced by DsbA, which catalyzes rapid formation of disulfide bonds in newly synthesized proteins [3, 26]. Over-expression of DsbC is required for proper folding of proteins with multiple non-consecutive disulfide bonds  and it can improve yields of recombinant proteins containing multiple disulfide bonds . Interestingly, rate of free thiol formation was significantly reduced in the homogenate derived from the ∆dsbC mutant fermentation. We were also able to demonstrate that recombinant DsbC spiked into homogenate derived from the blank host fermentation caused an increase in mAb1 free thiols compared to control (where some level of free thiol induction was observed, possibly due to expression of chromosomal dsbC). In addition, HPLC pattern of DsbC-induced free thiol species was consistent with that of DS materials. Altogether this was intriguing, since it indicated that DsbC is involved in homogenate’s ability to reduce disulfide bonds in mAb1.
In E. coli cells, compartmentalization ensures separation of the reducing environment of cytoplasm from the oxidative environment of periplasm required for the proper formation of disulfide bonds . It was clear that in intact cells, DsbC effectively engages mAb1 as substrate in the oxidative protein folding process occurring in the periplasmic space, since without over-expression of DsbC, very low titers of mAb1 were obtained. Lack of increase in free thiols during the whole cell broth hold indicated that already folded mAb1 (residing in the periplasm) is not prone to reduction, suggesting an inability to effectively engage the reducing enzyme activity observed in the homogenate. On the other hand, homogenization destroys cell compartmentalization, which results in non-native redox environment maintaining released components of both cytoplasm and periplasm under reducing conditions. Under such conditions, DsbC might be able to re-engage with mAb1 leading to the selective Fc disulfide bond reduction. Besides acting as disulfide bond isomerase, DsbC can also function as a disulfide reductase, where the reaction results in the reduction of the substrate and the oxidation of DsbC  instead of disulfide bond reshuffling. Increasing evidence appears to suggest that this mode of action, resulting in reduced substrate which can be then reoxidized by oxidants like DsbA, might be more relevant than true isomerase activity of DsbC . Thus, reductase activity of DsbC could be causing disulfide bond reduction of mAb1 and relatively more reducing potential of homogenate (compared to the periplasm) might be preventing formation of those disulfide bonds.
Homogenate-induced reduction is remarkably selective towards the disulfide bond buried within CH2 domain and this poses a question how DsbC can re-engage with folded mAb1 and achieve such site-specific reduction. The answer could be a chaperone activity that DsbC also exhibits, and which is thought to be important for its function as an isomerase , since incorrectly formed disulfide bonds often might be buried within misfolded protein. DsbC can interact with folded proteins like for example AraF, where DsbC was shown to reduce poorly accessible single cysteine residue buried in the cleft of the protein . In IgGs, the CH2 domains are the only unpaired domains. However, they contain conserved N-glycosylation site (Asn-297), occupied by glycans which extend into the Fc cavity and can make contacts with the protein surface [32, 33], thus making it potentially less accessible. On the other hand, mAb1 lacks Fc glycans and its Fc cavity surface (with exposed hydrophobic patches) may be more accessible to DsbC (or other chaperones), thus potentially allowing it to re-engage as chaperone and reductase. This would be consistent with our observation that presence of Fc glycans in mAb2 prevented, while their removal or trimming conferred susceptibility to free thiol formation during homogenate hold.
There might be additional structural basis for the selectivity towards the CH2 domain. In E. coli periplasm, DsbC is maintained exclusively in catalytically active reduced state by inner membrane protein DsbD [34–38]. Interestingly, DsbD belongs to a small subset of E. coli proteins containing immunoglobulin-like domains . DsbD-α, a periplasmic N-terminal domain of DsbD with thiol oxidoreductase activity, forms a complex with DsbC and reduces it. This DsbD-α domain has classical c-type Ig fold structure [40, 41], therefore, maintenance of the normal function of DsbC in periplasm relies on the recognition and interaction with the Ig fold-based protein domain. Since, IgG CH2 domain shares structural similarity with DsbD-α due to presence of c-type Ig fold (Additional file 1: Fig. S1), we hypothesize that in homogenate, re-engagement of DsbC with the CH2 domain of mAb1 could be facilitated by recognition of the Ig fold structure and chaperone-substrate like interactions due to lack of Fc glycans. This could explain selective disulfide bond reduction predominantly in the CH2 domain of mAb1 and absence of significant free thiol formation in the Fab. Further work is needed to examine this hypothesis and to show direct interaction of DsbC with IgG1 Fc.