RESUMO
High-throughput screening (HTS) of chromatography resins for identifying optimal protein purification conditions is becoming an integral part of industrial process development. In this work, ceramic hydroxyapatite (cHA) chromatography of 15 humanized monoclonal antibodies (mAbs) was examined by HTS. MAb binding, as quantified by partition coefficient (K(p)), was measured under 92 combinations of sodium chloride, phosphate, and pH. Binding varied inversely with these variables for all mAbs tested. However, the magnitudes of binding among mAbs under identical conditions varied significantly, showing a >1.5 log range in K(p). Analysis of variance (ANOVA) techniques were used to describe the binding of each mAb as a function of the three screen variables. Linear models relating log K(p) to the pH, log[sodium chloride], and log[phosphate] fit the data for each antibody with 93-96% accuracy. From these models, characteristic charge values for the cation exchange and metal coordination components of the multi-modal mAb/cHA interaction varied twofold across the mAbs, reflecting inherent variability in the number of contacts between a particular mAb and the cHA surface. Furthermore, we reduced the number of test conditions required from 92 to 8 while maintaining an accurate representation of the full binding response surface. This eight-point modeling method accurately predicted the binding behavior of mAbs as well as mAb aggregates, a common impurity in crude mAb preparations. Using this eight-point modeling method, binding and selectivity information for mAb and aggregate can be obtained from less than two milligrams of protein, making the method attractive for early manufacturability assessments.