The binding capacity data from the 96-well plate experiments were then used to predict yield and purity of the monomer in
the flow-through fraction. This is possible using the assumption that the monomer plate binding capacity equals the dynamic
binding capacity in a column, which is a good approximation for longer residence times. Purity and yield can then be calculated
using the following equations:
in which Vload is volume of sample loaded onto a column, C is concentration, CV is column volume, and SBC is the binding capacity found in the plates.
A column prediction for the strong anion exchanger at a simulated load of 150 g/L is shown in Figure 2, which also shows the
strong anion exchanger's predictions. The yield increased with decreased pH while the purity increased with increased pH.
In this case, the purity was considered the most important factor for the flow-through step. The highest purity was identified
by following the "ridge" from pH 8.0 without NaCl to pH 9.2 with approximately 50 mM NaCl.
Predictions for yield and purity at different sample loads also were performed for the multimodal medium in the 96-well plate
format and the prediction at the most favorable conditions was compared with the column data. Based on the previous data,
a column prediction of purity and yield at various sample loads could be made to find the best conditions for the flow-through
step. Figure 3 shows the data for a load of 122 g/L. The yield increased with decreased pH while the purity increased with
increased pH. In this case, the purity was considered to be the most important factor for the flow-through step, with the
highest purity prediction highlighted with the red box.
These results were verified using a 1-mL prepacked column (Figure 4, HiTrap, GE Healthcare). The column was equilibrated with
25 mM sodium phosphate at pH 7.5. The sample (desalted MabSelect Sure eluate at approximately 7 mg/mL) was loaded at 10 min
residence time. Experiments were performed with two sample loads, 130 and 260 g/L, and the corresponding prediction calculations
were made for these two loads.
The predicted yield and purity at both sample loads (130 and 260 g/L) were compared to the experimental values obtained using
both 96-well plate and 1-mL HiTrap columns. The results correlated well between columns and plates (Figure 4), demonstrating
that the experimental conditions in 96-well plates can be successfully scaled up to the column format. For both formats the
yield increased with increased load, but the purity showed the opposite trend. With purity being the most important factor,
the lower sample load was of greater interest because it provided 98% (96-well) and 97% (column) purity. However, in both
cases the yields were significantly below the target 85%.