Large-scale process development for hydrophobic interaction chromatography. I: Gel selection and development of binding conditions

Abstract

In hydrophobic interaction chromatography, many common causes of scale-up failure are inadvertently built into the process during the development stage. As with other adsorptive chromatographic methods, one of the most prevalent deficiencies is inadequate characterization of product retention conditions. Part 1 of this four-part series addresses gel selection and the development of sample-application conditions to avoid these problems. The authors use monoclonal antibodies to illustrate key points. Introduction Hydrophobic interaction chromatography (HIC) has evolved into one of the most powerful methods in preparative biochemistry. Its speed, resolution, and capacity rival ion exchange chromatography; its selectivity is complementary to other popular preparative methods such as ion exchange and size exclusion chromatography; and its ability to clear endotoxins, nucleic acids, and viruses makes it an indispensable tool for the purification of therapeutic proteins (1–8). Nevertheless, developing preparative HIC methods can be challenging. In particular, selecting a gel and developing preparative sample application conditions involve considerations that are unique to HIC. Neglecting these considerations can lead to a range of scale-up problems—including severe losses and product denaturation—that sometimes cause process developers to unnecessarily abandon this valuable technique. However, a systematic approach can avoid these problems, allowing chromatographers to develop large-scale methods that provide the same level of performance and reproducibility as other chromatographic methods. Materials and Methods We obtained Source 15ETH, 15ISO, and 15PHE hydrophobic interaction prepacked columns and bulk media from Pharmacia Biotech AB (Uppsala, Sweden). Ligand structures are illustrated in the accompanying box. All three media are based on 15-μm dp monodisperse spheres with a pore-size distribution suitable for large proteins. The base matrices are composed of poly(styrene–divinylbenzene) coated with a hydrophilic polymer. The bed dimensions of the 1-mL prepacked columns are 30 mm x 6.4 mm. The bed dimensions of the 6-mL prepacked columns are 30 mm x 16 mm. Becton Dickinson Immunocytometry Systems (San Jose, California) provided monoclonal antibodies. We used purified and unpurified Large Scale Process Development for Hydrophobic Interaction Chromatography, Part 1: Gel Selection and Development of Binding Conditions Pete Gagnon,1 Eric Grund,2 and Torgny Lindback 2 This article originally appeared in BioPharm 8(3) 21–29 (1995) and is adapted with permission. Reprints of this and other BioPharm articles can be ordered through the Advanstar Reprint Office by calling Mary Clark at (541) 984–5226 1Validated Biosystems, Inc., 5800 North Kolb Road, Suite 5127, Tucson, AZ USA 85750 2Pharmacia Biotech AB, S-751 82 Uppsala, Sweden Hydrophobic Ligands The following hydrophobic ligands were used in this study: Source 15ETH R-O-CH2-CHOH-CH2-OH Source 15ISO R-O-CH2-CHOH-CH2-O-CH2-CHOH-CH2 -O-CH-(CH3)2 Source 15PHE R-O-CH2-CHOH-CH2-O-CH2-CHOH-CH2 -O-C6H5 ascites samples of a mouse IgG1 and a mouse IgM throughout this study. We purchased buffers and salts from Sigma Chemical Company (St. Louis, Missouri). All buffer components were American Chemical Society (ACS) grade or better. Process water was prepared using reverse osmosis and deionization. We filtered buffers through a 0.22-mm filter immediately after formulation and assigned five-day expirations. We determined the free-solution solubility of the antibodies by adding incremental concentrations of ammonium sulfate in 0.1 M sodium phosphate (pH 7.0) to purified 1-mg/mL antibody samples to create a range of final salt concentrations of 0.0–2.0 M ammonium sulfate. We incubated the samples for 1 hour at room temperature and then filtered them through a 0.22-mm filter into a twofold volume (double that of the sample) of 0.05 M sodium phosphate (pH 7.0). Then we measured each filtrate spectrophotometrically at 280 nm to determine the proportion of antibody remaining in the supernatant. To characterize precipitation time curves we prepared duplicate sets of test tubes in advance with 1 mL of 1.50 M ammonium sulfate and 0.05 M sodium phosphate (pH 7.0). We added 100 μL of purified antibody to the first tube, allowed it to incubate for 1 min, and then filtered it to remove any precipitate. The filtrate was collected into 2 mL of 0.05 M sodium phosphate (pH 7.0) to prevent additional precipitation. We repeated the experiment with 2, 3, 4, 5, 10, 15, 30, and 60-minute incubations and quantitated the diluted filtrates spectrophotometrically. We repeated the experiment with 3.0 M ammonium sulfate, collecting the filtrates into 4 mL of diluent. To obtain a relative expression of hydro-phobic interactions between the mouse IgG1 and the various HIC media under nonretaining buffer conditions, we measured the peak height of the unbound material as it passed through the column. This work was performed using 1-mL prepacked columns and 20-μL injections of purified protein at 1 mg/mL in 0.05 M sodium phosphate (pH 7.0). The columns were equilibrated with 10 column volumes of ammonium sulfate with concentration ranging incrementally from 0.0 to 2.0 M. We conducted all experiments using a linear flow rate of 940 cm/hour (5 mL/min). Using the 1-mL 15ISO columns, we determined the interfacial salt tolerance of purified proteins in a series of experiments in which we loaded protein through one line, binding buffer through another, and mixed the two streams before they reached the column. We loaded 1 mL of antibody at 5 mg/mL in 0.05 M sodium phosphate (pH 7.0) onto a 1-mL column at a mix ratio of 20% sample to 80% binding buffer (5-mL total sample-application volume). The ammonium sulfate concentration of the binding buffer varied in increments from 0.0 M to 2.5 M. In each iteration, the column was equilibrated to the mixed ammonium sulfate concentration of the intended sample stream by diluting the binding buffer on-line at a ratio of 20% 0.05 M sodium phosphate (pH 7.0) to 80% binding buffer. We determined the antibody capture efficiency for each experiment by measuring the amount of antibody recovered in the elution peak. All experiments were conducted at room temperature, except as noted. Immunoreactivity per milligram of antibody was determined by immunoassay. Results and Discussion Gel selection. Selecting the most appropriate HIC medium for purification of a particular product requires careful matching of the properties of the product with those of the gel. If the ligand is too weak or at too low a density, then binding will require an excessive amount of salt. Excess salt may be a mere logistical inconvenience in some cases, but if the concentration required to accomplish binding is higher than the level at which the protein precipitates in free solution, it can severely complicate development of preparative sample-application conditions. If the ligand binds sample proteins too strongly, it can cause on-column conformational rearrangements of labile proteins. These rearrangements are sometimes reversed spontaneously upon elution, but in other cases, the protein is denatured permanently (6–13). Denaturation problems are usually accompanied by poor mass recovery, which should consequently be interpreted as a warning. Figure 1 illustrates the results of an experiment designed to evaluate the appropriateness of 3 different supports for HIC purification of the mouse IgG1 and the mouse IgM monoclonal antibody. For the IgG1, we obtained mass 2