T-m-Values and Unfolded Fraction Can Predict Aggregation Rates for Granulocyte Colony Stimulating Factor Variant Formulations but Not under Predominantly Native Conditions

Protein engineering and formulation optimization strategies can be taken to minimize protein aggregation in the biopharmaceutical industry. Short-term stability measures such as the midpoint transition temperature (T-m) for global unfolding provide convenient surrogates for longer-term (e.g., 2-year) degradation kinetics, with which to optimize formulations on practical time-scales. While successful in some cases, their limitations have not been fully evaluated or understood. T-m values are known to correlate with chemical degradation kinetics for wild-type granulocyte colony stimulating factor (GCSF) at at pH 4-5.5. However, we found previously that the T-m, of an antibody Fab fragment only correlated with its rate of monomer loss at temperatures close to the T-m. Here we evaluated T-m, the fraction of unfolded protein (f(T)) at temperature T, and two additional short-term stability measures, for their ability to predict the kinetics of monomer and bioactivity loss of wild-type GCSF and four variants, at 37 C, and in a wide range of formulations. The GCSF variants introduced one to three mutations, giving a range of conformational stabilities spanning 7.8 kcal molt. We determined the extent to which the formulation rank order differs across the variants when evaluated by each of the four shortterm stability measures. All correlations decreased as the difference in average T, between each pair of GCSF variants increased. The rank order of formulations determined by T-m was the best preserved, with R-2-values >0.7. T-m-values also provided a good predictor (R-2 = 0.73) of the aggregation rates, extending previous findings to include GCSF variant-formulation combinations. Further analysis revealed that GCSF aggregation rates at 37 degrees C were dependent on the fraction unfolded at 37 degrees C (fT37), but transitioned smoothly to a constant baseline rate of aggregation at f(T) 37 < 10(-3). A similar function was observed previously for A33 Fab formulated by pH, ionic strength, and temperature, without excipients. For GCSF, all combinations of variants and formulations fit onto a single curve, suggesting that even single mutations destabilized by up to 4.8 kcal morl, are insufficient to change significantly the baseline rate of aggregation under native conditions. The baseline rate of aggregation for GCSF under native conditions was 66-fold higher than that for A33 Fab, highlighting that they are a specific feature of each native protein structure, likely to be dependent on local surface properties and dynamics.