The goal of this work is to develop a downstream process for the purification of human growth hormone (hGH) and an hGH antagonist, named hGHG120R, from cultured recombinant mouse L cells grown in the presence of 3% – 5% Nu-Serum IV. The process we developed, capable of handling 10-15 liters of culture medium for each individual run, consists of centrifugation for removal of the culture medium, salt precipitation for reducing the medium volume, membrane ultrafiltration, size exclusion chromatography for purification, and a reversed phase high performance liquid chromatography (RP- HPLC) column with high resolution for final purification. Phase separation and buffer exchange are used to remove organic solvents. Then an ultrafiltration step is added to remove pyrogen. The last step is lyophilization. The products of hGH and hGHG120R produced by the process retain biological activity and can be pyrogen-free. In the process development, HPLC analytical methods are set up for evaluation and quality-control of the intermediate and finished products. A reasonable cost estimation for the downstream process is made based on the material cost, the labor cost, and equipment wear.
As a key purification step, RP-HPLC gradient elution is studied using a general rate model. A small analytical column is used to obtain a retention correlation and adsorption saturation capacity. Mass transfer parameters are estimated from empirical correlations in the literature. These data are used in the model to predict gradient elution profiles on the small column. For retention times of hGH and hGHG120R, simulated results are found to be in very good agreement with experimental results, with an average absolute relative error of 3% for retention time. Using the retention correlation and the adsorption saturation capacity from the small column, scale-up predictions of chromatograms at different gradient times, flow rates, and sample sizes on a larger preparative RP-HPLC column are made. Compared with experimental results, simulated retention times give an average relative error of 9.7%. The error came mainly from the slight different conditions of the same packing material in the two columns. This is verified by using a retention correlation obtained directly from the preparative column for simulations which give an average absolute relative error of 4.8%. Through the study of the effects of model parameters on the simulation, the retention correlation is found to be the most critical factor. The adsorption saturation capacity is not sensitive to the simulated retention times of the proteins.