636 BIOCHEMICAL AND SYNTHETIC POLYMER SEPARATIONS The porosity of an SEC column can be characterized by its phase ratio (V i /V 0 ), as in Figure 13.39a. Other factors equal, resolution increases for a larger phase ratio. Soft-gel SEC packings have high porosities with phase ratios of 1.5 to 2.4 [136], while high-performance SEC packings have more modest phase ratios of 0.5 to 1.5 [137]. However, the disadvantage of lower phase ratios for high-performance SEC columns is more than offset by their higher efficiencies and faster analysis times. As particle pore-diameter and pore-volume increase, the mechanical strength of the support decreases. 13.8.2.3 Particle Diameter As in interactive modes of chromatography, a reduction in particle diameter in SEC improves column efficiency. Column packings with particle diameters of 10 to 12 μm are available for less demanding applications, such as preparative separations, while SEC packings with particle diameters of 4 to 5 μm can be used for applica- tions demanding higher resolution (as for molecular-weight analysis or analytical separations of protein mixtures). 13.8.2.4 Increasing Resolution Resolution in SEC is controlled in two ways: (1) by selecting a column with a flatter slope of log M versus retention time (which effectively increases selectivity or values of α), and/or (2) by increasing the column plate number N. Higher values of N can be achieved by the use of smaller particle columns, lower flow rates, or an increase in column length. High-performance SEC columns are usually operated at flow rates of 1 mL/min or less for an 8-mm i.d. column. The small diffusion coefficients D m for proteins and other biopolymers (or synthetic polymers) means that N usually increases significantly when flow rate is reduced (Section 2.3.1). 13.8.3 Mobile Phases for Gel Filtration In contrast to interactive modes of chromatography, where the mobile phase is an active participant in the separation process, the mobile phase in SEC is simply a carrier that transports solute molecules through the column. The mobile phase is selected to maintain the solute in solution and in the appropriate conformation (e.g., native vs. denatured), to minimize column-solute interactions, and to maximize column lifetime. Thus the mobile phase may contain additives that suppress unde- sired interactions of the analyte with the support or the bonded stationary phase. These interactions may be electrostatic in nature, and in the case of silica-based columns, residual ionized silanols on the support often create a negative charge on the packing. For cationic solutes such as basic proteins, ionized silanols can result in a cation-exchange contribution to retention, so the solutes will elute later than pre- dicted by a purely SEC mechanism. For anionic solutes, such as acidic proteins and nucleic acids, ion exclusion can result, with solutes eluting earlier than predicted. In severe cases, solutes may elute after V M (ion exchange) or before V 0 (ion exclusion). A second type of non-ideal behavior in SEC is hydrophobic interaction, leading to increased solute retention. Undesired analyte-column interactions can often be minimized by adjusting the salt concentration: increasing ionic strength reduces electrostatic interactions, and 13.8 SIZE-EXCLUSION CHROMATOGRAPHY (SEC) 637 decreasing ionic strength reduces hydrophobic interactions. Thus an intermediate ionic strength will generally be required to avoid these non-ideal behaviors. A typical mobile phase for gel filtration is 100 mM potassium phosphate + 100 mM potassium chloride (pH-6.8). Hydrophobic interactions can also be reduced by adding a small amount (e.g., 5–10%) of an organic solvent such as methanol, ethanol, or glycerol. 13.8.4 Operational Considerations Once the appropriate column, mobile phase, and flow rate have been selected, successful separations by SEC may require the adjustment of sample weight and concentration. Mobile-phase additives such as surfactants and salts can be used to maintain analyte solubility, or to suppress undesired analyte-stationary phase interactions. Alternatively, such interactions can be exploited in order to achieve a desired separation (Section 13.8.4.4). 13.8.4.1 Column Capacity The loading capacity of SEC columns is relatively modest—compared to interactive modes of chromatograph—because high-molecular-weight sample solutions can be quite viscous. Viscous samples can result in undesired peak distortion and broad- ening for samples that are too concentrated—or for larger samples. Such samples therefore require either a smaller sample volume or a more dilute sample. A rule of thumb suggests that the sample-volume should be ≤ 2% of the column-volume for a sample molecular weight of 10,000 Da, a value that will be greater for lower molecular-weight samples, and smaller for higher molecular-weight samples. A typ- ical analytical SEC column with dimensions of 300 × 8.0-mm has V m = 10–11 mL, providing a maximum sample-injection volume of about 200 μL. Because sample size in SEC is limited mainly by sample volume and viscosity (which increases with sample concentration), the weight limit for a protein sample and a 300 × 8.0-mm column is then about 1 to 2 mg. For larger sample weights or volumes, resolution may be compromised. Sample capacity will scale in proportion to column volumes, for different column lengths and diameters. 13.8.4.2 Use of Denaturing Conditions For gel-filtration separation under nondenaturing conditions (as in Fig. 13.41), estimates of analyte molecular weight can be in error as a result of differences in molecular shape. Variations in molecular shape are less likely for denatured species, suggesting analysis and calibration with denaturing conditions whenever the shape of the native protein molecule is in question. The addition of a denaturant such as 4 to 6M guanidinium hydrochloride, 4 to 6M urea, or 0.1–1% sodium dodecylsulfate (SDS) to the mobile phase can be used to convert calibrants and analytes to random coil conformations. Denaturing conditions lower the fractionation range of a gel-filtration column, because of the increase in analyte hydrodynamic diameters [138]. Also surfactants, such as SDS, may bind strongly to the column and be difficult to remove; therefore it is advisable to dedicate the column to each application of this kind (or use an unretained denaturant such as urea or guanidine). High concentrations of chaotropic salts such as urea and guanidinium HCl in the mobile phase (for denaturing conditions) can compromise pump and injector seals, and should never be left standing in the HPLC system following use. 638 BIOCHEMICAL AND SYNTHETIC POLYMER SEPARATIONS 13.8.4.3 Column Calibration When installing a new gel-filtration column, the values of V 0 and V m should be determined using appropriate probes. The value of V 0 can be measured by injecting a large biopolymer whose molecular weight falls outside the exclusion limit of the column; high-M DNA (e.g., calf thymus DNA) is often used. The blue dextran used for measuring V 0 on soft gel columns can give erroneous values on some high-performance SEC columns because of hydrophobic binding with increased retention. The value of V m is determined using a very hydrophilic small molecule that can be detected by UV. Popular choices are cyanocobalamin (vitamin B 12 ), glycyl tyrosine, or p-aminobenzoic acid [137]. Non-ideal interactions can be characterized by using small-molecule probes [139, 140] that elute at V m . Cation-exchange interactions are indicated by retention times t R > V m for arginine or lysine. Ion exclusion is shown by early elution of citrate or glutamic acid (t R > V m ). Hydrophobic interactions can be detected by t R > V m for phenylethyl alcohol or benzyl alcohol as solute. Inasmuch as proteins can denature at higher temperatures, SEC retention can be temperature dependent. 13.8.4.4 Exploiting Non-ideal Interactions While non-ideal interactions can prevent accurate estimates of molecular weight, these interactions can also be used to advantage for the purpose of changing relative retention and improving resolution—especially for preparative separations. The above mentioned approaches for suppressing non-ideal interactions (Section 13.8.3; varying salt concentration, adding organic solvents, varying pH) also suggest means to enhance these interactions, depending on the interaction that is to be enhanced. 13.8.5 Advantages and Limitations of SEC Size-exclusion chromatography offers several advantages that make it a desirable technique for both preparative and analytical applications. First, separations are relatively fast: with a 300 × 8.0-mm analytical column operated at 1 mL/min, all analytes should elute in about 10 minutes. Second, because the stationary phase is designed to eliminate interactions with the sample, SEC columns usually exhibit excellent recovery of mass and biological activity. Third, all separations are performed under isocratic conditions, which generally favors convenience. There are also some limitations to SEC. First, the resolving power is quite modest, compared to interactive chromatography. The maximum number of baseline-resolved peaks in a separation is usually only 5 to 10, compared to several hundred for gradient RPC. For a gel-filtration column with a fractionation range from 10 to 500 kDa, this implies that two proteins can be resolved, if they differ in molecular weight by a factor of two. Thus analytical SEC is only useful for samples that contain a limited number of components with quite different molecular weights. The second limitation of SEC is its low volume- and mass-loading capacity. As a consequence of these two limitations, SEC is more likely to be used as a later step in a purification scheme. A third limitation of SEC is modest column lifetime, particularly for silica-based SEC columns. When operated with aqueous buffers at neutral pH, SEC column lifetime is typically shorter than that of a silica-based RPC column operated with aqueous-organic mobile phases. A final limitation of SEC is 13.8 SIZE-EXCLUSION CHROMATOGRAPHY (SEC) 639 the accuracy of molecular-weight estimates, which are usually limited to rough esti- mates of molecular weight. While SDS-PAGE and mass spectrometry provide more accurate values, the first technique is laborious and the second is expensive. Cou- pling an SEC column to a static laser light-scattering detector (in conjunction with a concentration-sensitive detector), however, can provide accurate molecular-weight values [141]. 13.8.6 Applications of SEC SEC can be used for separating and characterizing analytes based on molecular size, and as a preparative tool for recovering purified material from a mixture. 13.8.6.1 Analytical Applications Analytical applications of gel filtration include molecular-weight estimation, monitoring or characterizing protein folding and aggregation, and determining receptor-ligand interaction. As discussed above, the retention of biopolymers in SEC is governed by molecular size and shape. To obtain accurate estimates of molecular weight, the column must be calibrated with standards that possess the same shape as the analyte, or both analyte and calibrants must be converted to a random-coil (denatured) configuration—and maintained as such during the analysis. Gel filtration can also be used to characterize protein folding and aggregation. As an illustration of the former, Figure 13.42 summarizes several experiments that U N 6.2M 4.7M 4.4M 4.1M 3.9M 3.7M 0.0M 10 15 20 ( mL ) Figure 13.42 Use of SEC to monitor the folding states of lysozyme in the presence of increas- ing mobile phase concentrations of guanidine-HCl. U , unfolded (denatured) species; N,native species. Adapted from [142]. 640 BIOCHEMICAL AND SYNTHETIC POLYMER SEPARATIONS monitor the folding state of lysozyme in the presence of varying concentrations of guanidine-HCl [142] in the mobile phase. In the absence of guanidine (0.0M), the protein exists in the folded or native state (N) and has a retention volume of 15.5 mL. For a guanidine concentration of 6.2M, the protein is completely unfolded (U) or denatured; the unfolded (extended) protein molecule has a larger molecular size, and is therefore less retained (retention volume of 14.5 mL). For intermediate guanidine concentrations, the protein exists simultaneously in folded and unfolded states. An illustration of using SEC to monitor protein aggregation (Figure 13.43) is the determination of the aggregation state of human growth-hormone by the distribution of the protein among monomer, dimer, and oligomer [143]. Because the biological activity of a protein varies with its aggregation, measurements of aggregation are important for determining the quality of protein pharmaceuticals (as in this example). Gel filtration is also able to determine receptor-ligand (e.g., protein–drug) interactions using zonal chromatography, Hummel–Dreyer methodology, or frontal analysis [144]. In zonal chromatography a mixture of protein and ligand is applied to the column. The protein-ligand complex elutes first and is separated from the free ligand. Quantitative analysis of the two species allows calculation of an affinity constant. Zonal chromatography can be used if dissociation of the protein-ligand complex is slow relative to the chromatographic process. In the Hummel–Dreyer method [145] the mobile phase contains the ligand, and a small volume of protein is injected onto the column. The elution profile exhibits a leading peak representing the protein-ligand complex, followed by a negative peak representing ligand-depleted mobile phase. The advantage of the Hummel–Dreyer approach is that protein is always in equilibrium with free ligand. It also requires only a small amount of protein. The method is applicable to protein-ligand complexes with rapid association-dissociation kinetics. In frontal analysis a large volume of protein and ligand is injected onto the column. The elution profile exhibits plateaus representing free protein, the complex in equilibrium with dissociated components, and free ligand. Frontal analysis enables determination of binding ratios under conditions where the concentration of species is known and constant, but requires large amounts of analytes. 6 8 10 12 14 ( min ) Oligomer Dimer Monomer Figure 13.43 SEC separation of monomer, dimer, and oligomer forms of recombinant human growth hormone. Reprinted from [143] with permission. 13.9 LARGE-SCALE PURIFICATION OF LARGE BIOMOLECULES 641 13.8.6.2 Preparative Applications Speed and gentle elution conditions make gel filtration a convenient method for the rapid isolation of biopolymers, with high recovery of mass and biological activity. Although resolving power is modest compared to other chromatographic modes, gel filtration is useful for purifying a target species from higher and lower molecular-weight components. The aqueous buffers used as mobile phases usually are compatible with subsequent purification steps such as IEC and RPC. The inert nature of gel-filtration columns allows the use of mobile phases supplemented with additives such as surfactants and organic modifiers that may be necessary to maintain the solubility of hydrophobic species, for example, membrane proteins. Salts, buffering agents, and other small molecules will all elute in the total permeation volume in an SEC separation, well-resolved from biopolymers such as proteins and nucleic acids. Total run times can be limited to a few minutes, so gel filtration provides a quick means for desalting biological samples. For buffer exchange, the destination buffer is used as the mobile phase, and solutes elute in the new buffer resolved from the original sample buffer zone. Thus in a multi-step purification scheme, gel filtration can be used not only as a fractionation tool but as a link between otherwise incompatible chromatographic steps; for example, HIC and ion exchange. 13.9 LARGE-SCALE PURIFICATION OF LARGE BIOMOLECULES The general principles of preparative liquid chromatography (prep-LC) are dis- cussed in Chapter 15 and apply equally for the isolation or purification of large biomolecules. In this section, we will provide a brief background of the purification of biomacromolecules by prep-LC, as well as a representative example. 13.9.1 Background The chromatographic purification of peptides and proteins on a laboratory scale has been underway for the past five decades. The principal chromatographic modes were originally limited to anion exchange, cation exchange, and gel filtration. The use of prep-LC for biomolecules has since expanded to other separation modes, including their purification on a commercial scale (downstream processing); no other procedure is competitive for the separation of closely related proteins and other biomolecules. The sales of biomolecules that have been purified by prep-LC today account for billions of dollars per year, and the number of products and their dollar value continue to climb. A wide variety of column packings or ‘‘resins’’ are now available, including different modes, supports (matrices), particle sizes, and porosities. Automated systems with column diameters of a meter or more allow large-scale prep-LC to be carried out at any desired scale. Both analytical and preparative chromatography are based on similar principles, as discussed in Chapter 15. In this section we highlight some aspects of prep-LC that are relevant for large-scale separations of biomacromolecules. Several factors have contributed to the recent, rapid development of large-scale prep-LC for biomolecules. It is now recognized that protein molecules may undergo 642 BIOCHEMICAL AND SYNTHETIC POLYMER SEPARATIONS subtle modifications during processing, and these impurities must be removed from a final pharmaceutical product [146]. High-performance prep-LC is the only practical method for separating these closely related protein species on a large scale. An early example was the purification of insulin by Eli Lilly and Novo Nordisk, using large-scale gel filtration and ion exchange [147]. Subsequent improvements in both equipment and column packings have made this approach more versatile, more attractive, and more widely applicable [148]. Expanded-bed technology allowed the removal of particulates and soluble impurities in a single operation [149], and other separation modes also became available [150, 151]. Finally, the advent of recombinant DNA technology in the late 1970s made it possible to produce large amounts of human proteins in microbial or animal cells, resulting in high-value products—but with challenging purification needs. Recombinant human insulin (rh-insulin; also called ‘‘bacterial-derived human insulin’’) was the first product based on recombinant DNA technology to be approved by the FDA. The manufacturing process resulted in several closely related insulin derivatives as impurities that could be resolved by analytical RPC, but not by any of the other prep-LC methods existing at that time [152, 153]. A commercial-scale HPLC-process for the purification of insulin was subsequently developed, as dis- cussed in Section 13.9.2 below. In subsequent years, prep-LC in the RPC mode became the method of choice for purifying peptides and low-molecular-weight proteins [152, 154]. Other proteins (e.g., monoclonal antibodies) cannot be puri- fied by RPC because they do not tolerate organic solvents. However, alternative chromatographic modes that use aqueous mobile phases have been more suc- cessful (e.g., affinity, metal chelation, hydroxyapatite, and hydrophic interaction chromatography [151]). The purification of insulin by high-performance RPC served as an example for many other recombinant products, including growth hormone [155], erythro- poietin [156], hirudin [157], cytokines [158, 159], insulin-like growth factor 1 (IGF-1) [160], and granuloma colony-stimulating factor (G-CSF) [161]. It soon became apparent that preparative high-performance RPC is a powerful and versatile method for purifying peptides and small proteins; the value of products purified by high-performance RPC has been estimated to account for approximately a third of total biotech product sales [162]. 13.9.2 Production-Scale Purification of rh-Insulin The purification process for recombinant human insulin is a rare example for which the details of the separation and its development have been published [162]. It is therefore instructive to review this separation, various aspects of which can be organized as follows: • purification targets • stationary phase • packing the column • stability of the product and column • mobile phase • separation 13.9 LARGE-SCALE PURIFICATION OF LARGE BIOMOLECULES 643 • column regeneration • small-scale purification •scale-up • production scale purification • product analysis 13.9.2.1 Purification Targets The RPC purification goals were a purity > 97.5%, with yields of ≥ 75%. Two different microbial production processes were originally considered: (1) a two-chain process in which insulin A- and B-chains were produced in separate fermentations then combined, and (2) a single-chain process in which proinsulin was formed by fermentation, then enzymatically converted to insulin by protease treatment. A single purification process was designed to handle either feedstock, as long as insulin-like impurities did not exceed 20%; in each case IEC, SEC, and high-performance RPC were used. The RPC separation will be discussed further. The mobile phase was required to be compatible with insulin stability and other steps in the process. The allowed cost of this purification step per pound of product was determined, based on expected product sales in the ton-per-year range. 13.9.2.2 Stationary Phases The column packing was selected after screening products from five manufacturers. Different alkyl-chain lengths, particle diameters, and pore sizes were evaluated, using 150 × 9.4-mm columns. Best results were obtained with particle diameters ≤ 12 μm, pores of 12 to 15 nm, and C 8 or C 18 ligands; these packings proved to be less fragile and easier to pack than particles with pores ≥ 30 nm. The final packing (Zorbax™ Process Grade C8) was selected on the basis of the latter preferred properties, as well as availability of the packing in the required quantities, and a demonstration by the manufacturer of batch-to-batch reproducibility. 13.9.2.3 Packing the Column To achieve the required separation on a commercial scale, prep-LC columns were required with plate numbers N that were similar to values for analytical columns. Laboratory-scale columns were slurry-packed under high pressure. For larger columns containing 5 to 50 kg packing, axial-compression was used at a pressure of 750 psi. Values of N were measured by injecting small-molecule test samples; N was equal to 30,000 to 40,000 plates/m for the laboratory-scale columns, and 45,000 to 55,000 plates/m for the larger axial-compression columns. Axial-compression columns also maintained their performance for a longer time, by minimizing voids and channeling that normally occur as the packing deteriorates and/or settles. 13.9.2.4 Stability of the Product and Column Insulin purification is best carried out with a mobile-phase pH of 3.0 to 4.0, as insulin solubility decreases to a minimum at its isoelectric point of 5.4 [163]. Under acidic conditions insulin deamidates to form monodesamido (A-21) insulin [164]. This 644 BIOCHEMICAL AND SYNTHETIC POLYMER SEPARATIONS undesirable reaction was avoided, however, because the RPC separation required only a few hours, and the insulin product could be rapidly recovered from the mobile phase by crystallization as the zinc salt [163]. At pH-7 or above, the monodesamide (B-3) derivative forms, which is undesirable. The column packing was stable over the pH range 2.0 to 8.0. 13.9.2.5 Mobile-Phase Composition The separation of insulin and two impurities of interest are shown in Figure 13.44 for a mobile-phase pH that is either acidic (Fig.13.44a) or mildly alkaline (Fig.13.44b). (a) (b) 0 10 20 30 40 50 (min) 0 10 20 30 40 50 (min) 100 90 80 70 60 50 % B 40 30 20 10 0 100 90 80 70 60 50 %B 40 30 20 10 0 pH-2.1 pH-7.3 1 2,3 4 5 6 3 4,5 1 2 6 Figure 13.44 RPC separation of rh-insulin and insulin derivatives with (a)acidand(b) alkaline mobile phases. Sample: 1,7.5 μg rh-insulin and 1.3 μg of each insulin deriva- tive: 2, desamido A-21 insulin; 3, N-carbamoyl-Gly insulin; 4, N-formyl-Gly insulin; 5, N-carbamoyl-Phe insulin; 6, insulin dimers. Conditions: 250 × 3.5-mm Zorbax™ C8 col- umn; 35 ◦ C; 1.0-ml/min; gradients: (a) solvent-A is pH-2.1 phosphate buffer; solvent-B is 50% acetonitrile/solvent-A; (b) solvent-A is pH-7.3 phosphate buffer; solvent-B is 50% acetonitrile/solvent-A Adapted from [166]. 13.9 LARGE-SCALE PURIFICATION OF LARGE BIOMOLECULES 645 Better resolution was obtained under acidic conditions, and all of the impurities elute after insulin (very desirable!); for mildly alkaline conditions, the impurities elute both before and after insulin. Several acids were evaluated, including acetic, formic, propionic, and phosphoric acid. Acetic acid (0.25M) was selected, because high concentrations of insulin ( > 50 mg/mL) were found to dissolve in the monomeric form under these conditions. The tendency of insulin to aggregate restricted the choice of mobile-phase conditions because aggregation interfered with the separation. Several B-solvents were evaluated, including ethanol, isopropanol, acetonitrile, and acetone. Ethanol and isopropanol gave poor resolution and a lower recovery (50–60%), while the solubility of insulin in isopropanol was poor. Acetone gave low yields (<50%) due to insulin precipitation. Acetonitrile provided the best separation, highest yields (75–85%), and did not interfere with the zinc precipitation step. Acetonitrile was also available in bulk and could be recovered by distillation; for these reasons it was selected as B-solvent. 13.9.2.6 Separation Isocratic and gradient elution of insulin were compared. With gradient elution, the product could be eluted in less than one column volume, while isocratic elution required two or more column volumes. Gradient elution was selected because the smaller elution volume facilitated downstream processing. A gradient from 15 to 30% acetonitrile provided satisfactory purity and yield, with a minimal volume of mobile phase; step gradients with comparable performance could not be found. The saturation capacity of the column (Section 15.3.2.1) for insulin was approximately 85 mg insulin/mL packing. Insulin mass recovery exceeded 97%. 13.9.2.7 Column Regeneration For the process to be economical, a column must be usable over many cycles. Samples of cellular origin tend to foul chromatographic columns rapidly because of the presence of lipids, nucleic acids, cell wall and cell membrane fragments, complex carbohydrates, and other cellular components. Column fouling can be reduced by proper design of steps that precede RPC, including filtration, precipitation, and IEC. Following each insulin separation, it was found necessary to clean the column by a wash with 60% acetonitrile/buffer (50 mM ammonium phosphate; pH-7.4). 13.9.2.8 Small-Scale Purification A small-scale insulin purification procedure was developed on the basis of the experiments outlined above, including IEC before RPC and gel filtration after. The ion-exchange step removed most of the protein impurities while protecting the RPC column from potential column fouling. The role of the RPC separation column was as a ‘‘polishing’’ step for removal of species closely related to insulin. A series of scale-up runs were carried out next. Figure 13.45 illustrates the preparative RPC separation. A major product peak is seen, followed by impurities that elute after the main peak. The resolution of product from these impurities appears poor, but this is not the case. The analysis of fractions from the RPC separation showed that the center of the mainstream peak from 3.4 to 4.3 column volumes (C-Vs) was nearly pure insulin (98.7%) in 82% yield, while the side-stream fractions (3.3–3.4 . decreases. 13.8.2.3 Particle Diameter As in interactive modes of chromatography, a reduction in particle diameter in SEC improves column efficiency. Column packings with particle diameters of 10 to 12 μm. denaturant such as 4 to 6M guanidinium hydrochloride, 4 to 6M urea, or 0.1–1% sodium dodecylsulfate (SDS) to the mobile phase can be used to convert calibrants and analytes to random coil conformations limitations to SEC. First, the resolving power is quite modest, compared to interactive chromatography. The maximum number of baseline-resolved peaks in a separation is usually only 5 to 10, compared to several