Immunoglobulin binding proteins (IBPs) are broadly used as reagents for the purification and detection of antibodies. Among the IBPs, the most widely used are Protein-A and Protein-G. The C2 domain of Protein-G from Streptococcus is a multi-specific protein domain; it possesses a high affinity (KD ~ 10 nM) for the Fc region of the IgG, but a much lower affinity (KD ~ low εM) for the constant domain of the antibody fragment (Fab), which limits some of its applications. Here, we describe the engineering of the Protein-G interface using phage display to create Protein-G-A1, a variant with 8 point mutations and an approximately 100-fold improved affinity over the parent domain for the 4D5 Fab scaffold. Protein-G-A1 is capable of robust binding to Fab fragments for numerous applications. Furthermore, we isolated a variant with pH-dependent affinity, demonstrating a 10,000-fold change in affinity from pH 7 to 4. Additional rational mutagenesis endowed Protein-G with significantly enhanced stability in basic conditions relative to the parent domain while maintaining high affinity to the Fab. This property is particularly useful to regenerate Protein-G affinity columns. Lastly, the affinity-matured Protein-G-A1 variant was tethered together to produce dimers capable of providing multivalent affinity enhancement to a low affinity antibody fragment-antigen interaction. Engineered Protein-G variants should find widespread application in the use of Fab-based affinity reagents.

Engineering monovalent Fab fragments into bivalent formats like IgGs or F(ab') 2 can lead to aggregation presumably because of nonspecific off-target interactions that induce aggregation. In an effort to further understand the molecular determinants of nonspecific interactions for engineered antibodies and natively folded proteins in general, we focused on a synthetic Fab with low nanomolar affinity to histone chaperone Anti-silencing factor 1 (Asf1) that demonstrates off-target binding through low solubility (~5 mg/mL) in the multivalent F(ab') 2 state. Here, we generated phage display-based shotgun scanning libraries to introduce aspartate as a negative design element into the antibody paratope. The antibody-combining site was amenable to aspartate substitution at numerous positions within the antigen binding loops and one variant, Tyr Asp, possessed high solubility (>100 mg/ml). Furthermore, the mutations decreased nonspecific interactions measured by column interaction chromatography and ELISA in the multivalent antibody format while maintaining high affinity to the antigen. Structural determination of the antibody-antigen complex revealed that the aspartate-permissive residues formed a polar ring around the structural and functional paratope, recapitulating the canonical feature of naturally occurring protein-protein interactions. This observation may inform future strategies for the design and engineering of molecular recognition.

Human lymphocytes exposed to Aggregatibacter actinomycetemcomitans (Aa) cytolethal distending toxin (Cdt) undergo cell cycle arrest and apoptosis. In previous studies, we demonstrated that the active Cdt subunit, CdtB, is a potent phosphatidylinositol (PI) 3,4,5-triphosphate phosphatase. Moreover, AaCdt-treated cells exhibit evidence of PI-3-kinase (PI-3K) signaling blockade characterized by reduced levels of PIP3, pAkt, and pGSK3β. We have also demonstrated that PI-3K blockade is a requisite of AaCdt-induced toxicity in lymphocytes. In this study, we extended our observations to include assessment of Cdts from Haemophilus ducreyi (HdCdt) and Campylobacter jejuni (CjCdt). We now report that the CdtB subunit from HdCdt and CjCdt, similar to that of AaCdt, exhibit potent PIP3 phosphatase activity and that Jurkat cells treated with these Cdts exhibit PI-3K signaling blockade: reduced levels of pAkt and pGSK3β. Since non-phosphorylated GSK3β is the active form of this kinase, we compared Cdts for dependence on GSK3β activity. Two GSK3β inhibitors were employed, LY2090314 and CHIR99021; both inhibitors blocked the ability of Cdts to induce cell cycle arrest. We have previously demonstrated that AaCdt induces increases in the CDK inhibitor, p21CIP1/WAF1, and, further, that this was a requisite for toxin-induced cell death via apoptosis. We now demonstrate that HdCdt and CjCdt also share this requirement. It is also noteworthy that p21CIP1/WAF1 was not involved in the ability of the three Cdts to induce cell cycle arrest. Finally, we demonstrate that, like AaCdt, HdCdt is dependent upon the host cell protein, cellugyrin, for its toxicity (and presumably internalization of CdtB); CjCdt was not dependent upon this protein. The implications of these findings as they relate to Cdt’s molecular mode of action are discussed.