Rapid Assessment of Oxidation via Middle-Down LCMS Correlates with Methionine Side-Chain Solvent-Accessible Surface Area for 121 Clinical Stage Monoclonal Antibodies

Rong Yang, Tushar Jain, Heather Lynaugh, R. Paul Nobrega, Xiaojun Lu, Todd Boland, Irina Burnina, Tingwan Sun, Isabelle Caffry, Michael Brown, Xiaoyong Zhi, Asparouh Lilov, Yingda Xu 

mAbs 9, 646–653. DOI: 10.1080/19420862.2017.1290753 

Overview 

Methionine oxidation is an important chemical stability concern during therapeutic monoclonal antibody (mAb) development. Oxidation in the Fc domain can lower the binding affinity to Protein A and FcRn, reducing purification yields and accelerating in vivo clearance, while oxidation in the variable region can decrease antigen binding and increase heterogeneity. Lead mAb candidates are usually screened under forced oxidation conditions to identify these liabilities early in the discovery process. However, standard oxidation assessment techniques rely on peptide mapping coupled with mass spectrometry, which does not scale well to large candidate panels. This study developed a faster, middle-down LC-MS approach to screen oxidation susceptibility across 121 clinical-stage mAbs and compared the results to computational predictions based on protein sequence and modeling from crystal structures. 

Study rationale and objectives 

Susceptibility to methionine oxidation depends not only on its presence in the sequence, but the extent to which it is exposed within the folded structure of the mAb. This study aimed to test whether oxidation measured by a rapid LC-MS approach correlates with computational predictions of methionine side-chain solvent exposure, which would support using sequence-based modeling to predict oxidation risk before a candidate is produced. 

The objectives of the study were to: 

• Develop and apply a high-throughput middle-down LC-MS method to screen a large panel of clinical-stage mAbs for oxidation. 

• Compare experimental oxidation data with computational predictions based on solvent-accessible surface area (SASA). 

• Assess the reliability of upstream computational modeling for sequence-level oxidation risk assessment. 

Approach and techniques 

A set of 121 mAbs with variable regions matching clinical-stage therapeutics were expressed, purified, and treated with hydrogen peroxide to induce oxidation. Enzymatic cleavage and disulfide reduction were used to separate each antibody into three distinct segments (light chain, Fd, and Fc), allowing oxidation to be localized by LC-MS without full peptide mapping. Experimental oxidation levels were compared to computational predictions based on methionine SASA modeling.  

Major findings and impact 

• The middle-down LC-MS approach enabled rapid localization of oxidation to specific antibody segments across all 121 samples. 

• Experimental oxidation data correlated well with computational SASA-based predictions, and no false positives were observed (every methionine predicted to be exposed and oxidation-prone did in fact show oxidation. 

• Some false negatives occurred (oxidation observed where none was predicted). On further investigation, this was often traced to buried methionine residues whose reactivity was not well captured by static crystal structure conformations. 

• The results support using sequence-based computational screening to flag oxidation susceptibility before material is produced, at least as a first-pass filter. 

Implications for therapeutic antibody development 

This high-throughput screening method offers a way to flag oxidation-labile molecules earlier and faster than traditional peptide mapping, which is valuable during early discovery when many candidates require rapid chemical stability assessment. The validation of computational predictions adds a complementary layer of in silico screening that can be applied at scale before committing to physical production of each candidate.