444b Approaches to Achieving High Productivity In Size Exclusion SMB Chromatography Processes

Shawn D. Feist, Inorganic Chemistry and Catalysis, The Dow Chemical Company, 1776 Building, Michigan Operations, Midland, MI 48667, Yogesh Hasabnis, Engineering and Process Sciences, The Dow Chemical Company, Tower C, Panchsil Technical Park, Pune, India, Bruce W. Pynnonen, Dow Water Solutions, The Dow Chemical Company, Larkin Laboratory, 1691 N. Swede Rd., Midland, MI 48674, and Timothy Frank, Engineering & Process Sciences Laboratory, Process Separations Discipline, The Dow Chemical Company, 1319 Building, Midland, MI 48674.

This presentation will describe the generation and modeling of high-productivity SMB miniplant data for a model system involving separation of aqueous bovine serum albumin (BSA, 66 kDa) from equine heart myoglobin (EHM, 17 kDa). The separation was accomplished using a 4-zone, 12-column SMB miniplant packed with cellulose-based porous beads. High purities and recoveries were achieved for each protein, yielding up to 99.5+% recovery of 99.9% pure BSA in the raffinate stream (protein-only basis) and 99.9% recovery of 97.5% pure EHM in the extract. The material balance accountabilities for the proteins were within 100 ± 5%. Productivities as high as 1.7 grams total product (isolated BSA + isolated EHM) produced per hour per liter of media were obtained using total protein feed concentrations up to 30 g/L. This high-productivity performance was achieved at operating pressures within industrially-viable ranges (less than 120 psig or about 8 bar per column) by using relatively large diameter beads (100 – 250 micron) and tall columns. The SMB process was simulated using a rate-based model developed in Athena Visual Studio®. A value for the intra-particle mass transfer coefficient was obtained directly from a previously published study of this system (Houwing et al., AIChE J. 2003, 49(5), 1158-1167). Linear adsorption isotherms were used to represent the equilibrium, and appropriate values for the isotherm constants were determined by regression of miniplant operating data. The resulting process simulation provided an excellent representation of the miniplant's performance over a wide range of protein concentrations and operating conditions. Strategies used in this work to achieve high-productivity operation will be described.