Chapter 7 Innovative separation methods in bioprocessing J. A. ASENJO, Biochemical Engineering Laboratory, Department of Food Science and Technology, The University of Reading, Reading RG6 6AP, UK and J. B. CHAUDHURI, School of Chemical Engineering, University of Bath, Bath BA2 7AY, UK 7.1 INTRODUCTION Discoveries and achievements in modern biology and recombinant DNA technology in the last few years have resulted in the development of a number of new therapeutics for human use such as insulin, human growth hormone (hGH), tissue plasminogen activator (tPA) for cardiac disease, erythropoietin (EPO) and hepatitis B vaccine and thus the possibility of their industrial large-scale production. This poses a tremendous challenge for the chemical and biochemical engineer in terms of developing efficient separation processes for these new proteins. As they are intended for human use the levels of purity required are of the order of 99.9% or 99.98% or even higher (depending on dosage) and they have to be separated from a very large number of contaminants, other proteins, nucleic acids, polysaccharides and many other components present in the cell culture or cell lysate used to manufacture these proteins. Competitive advantage in production depends not only on innovations in molecular biology and other areas of basic biological sciences but also on innovation and optimisation of separation and downstream proc- esses. The main issues important for the development of novel separation techniques to give improved resolution, simplicity, speed, ease of scale-up and possibly continuous operation are presented and discussed. The assessment of the state of the art as well as promising future developments concentrate on the separation and purification of proteins from complex mixtures. The present trend to develop techniques that exploit fundamental physicochemical principles more efficiently is emphasised. This includes the analysis of the physicochemical properties of proteins such as PI, charge as a function of pH, biological affinity (including metal ion and dye affinity), hydrophobicity and size and its 180 relation to efficiency in a bioseparation. Some properties (e.g. charge and affinity) can show extremely high resolution in purification operations, whereas others (e.g. molecular weight) show much lower resolution. 7.2 SYSTEM CHARACTERISTICS 7.2.1 Physicochemical basis for separation operations Development of new and efficient separation processes will be based on more effectively exploiting differences in the actual physicochemical properties of the product such as surface charge/titration curve, surface hydrophobicity, molecular weight, biospecificity towards certain ligands (e.g. metal ions, dyes), PI and stability, compared to those of the contaminant components in the crude broth. The main physicochemical factors involved in the development of separation processes are shown in Table 7.1 (Asenjo, 1993). J. A. Asenjo and J. B. Chaudhuri Table 7.1. Physicochemical basis for the development of separation processes Physicochemical basis Separation process Charge Ion-exchange chromatography Electrodialysis Aqueous two-phase partitioning Reverse micelle extraction Hydrophobicity Hydrophobic interaction chromatography Reversed phase chromatography Precipitation Aqueous two-phase partitioning Specific binding Affinity chromatography Size Gel filtration Ultrafiltration Dialysis Electric mobility Electrophoresis Isoelectric point Chromatofocusing Isoelectric focusing Sedimentation rate Centrifugation Surface activity Adsorption Solubility Solid-liquid extraction Foam fractionation Supercritical fluid extraction (From Asenjo, 1993) Innovative separation methods in bioprocessing 18 1 7.2.2 Kinetics and mass transfer The physical behaviour of the system has an effect on the development of novel separation processes. Processes can be divided into equilibrium and rate processes. In equilibrium processes selective separation depends on the attainment of a favourable equilibrium state. This, for example, includes liquid-liquid extraction and ion exchange chromatography, Rate processes, on the other hand, separate different proteins on the basis of their response to an imposed field (such as an electric field). Mobility and other similar properties determine the selectivity of this type of operation; a successful process is one in which the proteins have markedly different mobilities (e.g. electrophoresis). In a number of protein separation processes the residence time in the reactor is insufficient for equilibrium to be achieved and the kinetics of adsorption play an important role for example in affinity chromatography ana in the CARE (continuous adsorption recycle extraction) process. New developments in materials have recently shown dramatic advances in overcoming mass transfer limitations in processes such as perfusion and membrane chromatography and adsorption resulting in extremely fast separations, Some recent examples of novel techniques, which exploit the principles discussed above and provide useful analyses for optimal design of operations, include expanded bed (fluidised bed) adsorption of proteins, which allows direct broth extraction; cross-flow electrofiltration of disrupted microbial cells and for improved ultrafiltration of proteins; mathematical modelling of partitioning and phase behaviour in liquid-liquid extraction; mathematical modelling of chromatographic columns; perfusion and membrane chromatography; and advanced reversed phase chromatography using HPLC. The potential for scale-up of many of these systems is analysed and discussed. 7.3 LIQUID-LIQUID EXTRACTION: INTRODUCTION Liquid-liquid extraction as a technology has been used in the antibiotics industry for several decades and it is now beginning to be recognised as a potentially useful separation step in protein recovery and separation, particularly because it can readily be scaled-up and can, if necessary, be operated on a continuous basis. The physicochemical factors of the protein that determine partitioning are also starting to be understood. It is a reasonably high-capacity process and can offer good selectivity for the desired protein product. However, poor solubility of the large protein molecules in typical organic solvents restricts the range of solvents available for use in such a separation process. Two classes of solvents that appear to offer advantages for protein recovery for protein separations are aqueous polymer/salt (in some cases also polymer/polymer) systems and reverse micellar solutions. In both cases two phases are formed and the separation exploits the difference in partitioning of the proteins in the feed and extraction phases. In the aqueous polymer/salt separation systems the partitioning of the protein occurs between two immiscible aqueous phases; one rich in a polymer (usually polyethylene glycol, PEG) and the other in a salt (e.g. phosphate or sulphate). These systems show a non-denaturing solvent environment, small interfacial resistance to mass transfer, relatively high protein capacity and high selectivity. On the other hand reverse micelles exploit the solubilising properties of surfactants that can aggregate in organic solvents to form so-called inverted or reverse micelles. These aggregates consist of a polar core of 182 water and the solubilised protein stabilised by a surfactant shell layer. For protein extraction, one phase is the aqueous feed solution, the other the reversed micellar phase that acts as the extractant. They have several of the advantages quoted for aqueous two- phase systems. The suitability of using foam separation as well as gas aphrons as novel separation techniques for proteins are presently under investigation. 7.3.1 Aqueous two-phase separation Partitioning in two aqueous phases can be used for the separation of proteins from cell debris as well as for purification from other proteins. Partitioning can be done in a single step or as a multistage process. Differences in partition coefficients, however, between the different proteins can be high, hence one step tends to be sufficient (usually one for extraction and one for elution or back-extraction). The use of affinity partitioning can greatly enhance the specificity of the extraction. A typical process for extraction of a protein into the top PEG phase in a first stage and the back extraction into a bottom salt phase (e.g. phosphate or sulphate) in a second ‘back extraction’ step from a cell homogenate that includes recycle of the PEG phase is shown in Fig. 7.1 (Hustedt et al., 1985). J. A. Asenjo and J. B. Chaudhuri .................................................................... Fig. 7.1. Scheme of enzyme purification by liquid-liquid extraction. The cells are disrupted by wet milling, and after passing through a heat exchanger, PEG and salts are added into the process stream of broken cells. After mixing and obtaining of equilibrium the phase system is separated, the outflowing bottom phase is going to waste. The product-containing PEG-rich top phase goes to a second mixer after addition of more salt to the process stream. The product is recovered in the resulting bottom phase while the concentrated PEG solution (upper phase) goes to waste or is recycled. (From Hustedt et al., 1985) Innovative separation methods in bioprocessing 183 Most soluble and particulate material partitions to the lower, more polar (e.g. salt) phase and the protein of interest partitions to the top less polar phase, usually PEG. Separation of actual proteins in such systems is based on manipulating the partition coefficient (K) by altering parameters such as average molecular weight of the polymer, type of phase forming salt used for the heavy phase, the types of ions included in the system and ionic strength of added salts (e.g. NaCI) (Schmidt et al., 1994). Figure 7.2 shows that the partition coefficient of a-amylase is a strong function of the presence of NaCl in a PEG/sulphate system. For extraction of the a-amylase from its contaminants a high concentration of NaCl is used in the first extraction stage, whereas a low concentra- tion of NaCl in the back-extraction stage will allow the recovery of a-amylase into the bottom sulphate phase as shown in Fig. 7.1. 4- 2- -4 I I I I I 0 2 4 6 a 10 Concentration of NaCl (Yo w/W) Fig. 7.2. Partition behaviour of a-amylase (log KJ and contaminant protein (log K,) from industrial supernatant from E. subtilis fermentation in PEG 4000/Sulphate systems as a function of added NaCl concentration at pH 7 and a phase volume ratio of 1. The partition coefficient (K) is defined as the concentration of a particular protein in the lighter phase divided by the concentration in the heavier phase. The main factors that determine partition depend on the type of system used: (1) Hydrophobicity. Differences in the surface hydrophobicity between proteins are exploited when partitioning them in PEG/salt two-phase systems. Typical systems that exploit a protein’s hydrophobicity are PEG/phosphate and PEG/sulphate with addition of a high concentration of NaCl (e.g. 10%). Size-dependent partition. Molecular size of the proteins or surface area of the particles to be partitioned is the dominating factor. It has been shown that for PEG/Dextran systems a protein’s molecular weight is inversely proportional to its partition coefficient. Electrochemical. Electrical potential between the phases is used to separate molecules or particles according to their charge. This is demonstrated in (2) (3) 1.0 Y g 0.5 - 0 - - I I 0.4 0.2 0.0 E . 9 v) -0.2 0, 0 - -0.4 -0.6 -0.8 - - I - - - - - I I -4 - Innovative separation methods in bioprocessing 187 Organic phase 4V& .t nv - Reverse 4 ' L micelle ?fiC%4k/ h 3 nc L.UVVaUL. V UdU UU!iiCUY u Protein a 0 a Aqueous phase Fig. 7.6. Protein partition into reverse micelles. There are two techniques for transferring proteins into the micellar phase. The most widely used method involves extraction of the protein with a biphasic liquid system, i.e. liquid-liquid extraction. One phase is the aqueous solution of the protein, and the other the organic micellar solution, usually in equal volume. By gently shaking the two phases, the protein partitions from the aqueous into the micellar phase. In the second method, solid state extraction of the protein, the protein powder is suspended in the micellar phase and gently stirred. The protein solubilised in the reverse micellar solution can be transferred back into an aqueous solution, by contacting the micellar solution with an aqueous solution containing a high concentration of a particular salt (KC1, CaC12), which has the capability to ex- change with the protein in the micelles. The basic idea is that the process of protein extraction by reverse micelles can be made specific (Le. tailored to a specific protein) and efficient (Le. high extraction yield) by controlling the micellar parameters such as the water content, the type and concentration of surfactant, the type and concentration of salt, and the pH. Leser et al. (1986) examined the transfer of ribonuclease-A, lysozyme, trypsin and pepsin, monitoring the protein concentration and the concentration of water found in the organic phase. It was observed that the transfer of water is generally moderate (below 4%), whereas, under certain conditions, the protein is quantitatively transferred. This fact demonstrated that the transfer of the protein into the micellar phase is not a passive process, i.e. is not simply due to the fact that water is transferred and with it the protein. The conclusion was that there is a thermodynamic driving force for the hydrophilic protein to leave the aqueous environment and to transfer into the reverse micelles. In other words, it seems that under certain conditions the protein-reverse micelle complex is energetically favoured above the free protein and empty reverse micelles. Interactions can be electrostatic, when surfactants with charged head groups are used, or hydrophobic with the surfactant interface or the apolar solvent. 188 J. A. Asenjo and J. B. Chaudhuri The fact that electrostatic interactions play an important role in the distribution of proteins over reverse micellar and aqueous phase is shown by the dependence of the aqueous phase pH and ionic strength. The pH of the solution will affect the solubilisation characteristics of a protein prima- rily in the way in which it modifies the charge distribution over the protein surface. With increasing pH the protein becomes less positively charged until it reaches its isoelectric point (PI). At pHs above the PI the protein will take on a net negative charge. If electro- static interactions play a significant role in the solubilisation process, partition with anionic surfactants should be possible only at pHs below the PI of the protein, where the protein is positively charged and electrostatic attractions between the protein and the surfactant head groups are favourable. At pHs above the PI, electrostatic repulsions would inhibit protein solubilisation. Goklen and Hatton (1987) have presented results on the effect of pH on solubilisation of cytochrome-c, lysozyme, and ribonuclease-A, in AOT/isooctane reverse micelle solu- tions. The results were presented as the percentage of the protein transferred from a 1 mg/ml aqueous protein solution to an equal volume of isooctane containing 50 mM of the anionic surfactant AOT. A summary of their results is presented in Table 7.3. Table 7.3. Effect of pH on solubilisation Protein PI pH range of maximum solubilisation cy tochrome-c 10.6 5-10 ribonuclease-A 7.8 1-7 lysozyme 11.1 6-1 1 As anticipated, only at pHs lower than the pl was there any appreciable solubilisation of a given protein, while above the PI the solubilisation appears to be totally suppressed. However, at extremes of pH there is a drop in the degree of solubilisation of the proteins due to protein denaturation, observed as precipitate formation at the interface (Chaudhuri et al., 1993). Luisi et al. (1979) used the quaternary ammonium salt methyl-trioctylammonium chloride (TOMAC) for the transfer of a-chymotrypsin from water to cyclohexane. It was found that the pH had to be reduced to values significantly below the PI (PI = 8.3) for there to be any appreciable solubilisation. The solubilisation occurred only over a very narrow pH range before decreasing rapidly again with further decreases in the pH of the aqueous feed phase, accompanied by precipitation at the interface. Similar results have been obtained by Dekker et al. (1986) for the enzyme a-amylase. Significant solubilisation of the enzyme was observed over a narrow pH range in the vicinity of 10-10.5 (PI = 5.1). In this pH range, all basic residues will be deprotonated and the only charged residues being the carboxyl groups bearing a negative charge. Innovative separation methods in bioprocessing 189 These results suggest that for higher molecular weight proteins a precise match be- tween the surface charge densities of the protein and the reverse micelle is needed for the transfer to occur. The effect of the ionic strength of the aqueous phase is primarily to mediate the electrostatic interactions between the protein surface and the surfactant headgroups. AS a result of Debye screening, increases in the ionic strength of the protein feed solution can be expected to reduce the interaction between the protein and surfactant headgroups, hence decreasing the solubilisation of the protein. This has been studied by Goklen and Hatton (1987). The transfer to the reverse micellar phase decreases for all proteins stud- ied (cytochrome-c, lysozyme and ribonuclease-A) at increasing ionic strength, but the point where this decrease starts depends on each particular protein. Such studies have shown that protein transfer is influenced by electrostatic interactions between the protein and surfactant headgroup, and that selectivity with respect to other proteins can be con- trolled by manipulation of pH and ionic strength (Goklen and Hatton, 1987). Other recent fundamental research in this field has focused on aspects of the kinetics of protein partition and the structure of the microemulsions (Fletcher et al., 1987; Luisi et al., 1988). Other work has indicated the considerable potential for the use of reverse micelles in selective protein purification (Dekker et al., 1989). Woll et ul. (1989) have shown that affinity ligands can be accommodated within the micelles to enhance selectivity. A study on the fractionation of intracellular proteins from Bakers’ yeast (Chaudhuri, 1991) proved to be complex, with problems of low protein yield, interfacial precipitation and denaturation, which were not evident in the studies with model solutions. As a result of disrupting the yeast cells, the intracellular contents (sugars, lipids, nucleic acid etc.) will be present as ‘impurities’. It is not known if these non-protein components contributed to the observed precipitation and denaturation, or if they can partition into the microemulsion, thereby reducing protein capacity and selectivity. There has been some activity in this field: for example, Schomaecker et al. (1988) have observed the inactivation of a-chymotrypsin in an AOT/n-heptane system. How- ever, they concluded that activity was lost as a result of enzyme autolysis as well as by interaction with the microemulsion system. Finally, the use of reverse micelles to isolate and refold pure denatured proteins has been studied (Hagen et al., 1990a) as shown in Fig. 7.7. The results reported 100% renaturation of denatured ribonuclease-A refolded in the reverse micelles. This method proved unsuitable for the refolding of interferon-y because its hydrophobic nature causes it to aggregate during the extraction process (Hagen et al., 1990b). There remain several unanswered questions regarding the use of reverse micelles, for example, how applicable is this method to other proteins, how will protein from inclusion bodies behave in this system, and is any activity lost during the extraction process? 7.3.3 Perfluorocarbon affinity separations Affinity chromatography exploits the natural, biospecific interactions that occur between biological molecules. These interactions are very specific and because of this affinity separation processes are very high resolution methods for the purification of proteins. However, conventional affinity chromatography has its drawbacks. Namely the gel 190 J. A. Asenjo and J. B. Chaudhuri Inclusion body F Refolded Solubilised protein j. 4 u k protein 4hOc-O 4' k Reverse 7 n x micelle Fig 7 7 Schematic diagram of protein refolding in reverse micelles matrix used as the support for the affinity ligand is not very stable towards extremes of pH which are found during sterilisation, and fouling of the columns may occur unless all particulate material is removed prior to application on the column. This requires the use of a solid-liquid separation operation increasing the process costs. A development in affinity purification is the use of perfluorocarbons as supports for affinity chromatography. Perfluorocarbons are synthetic molecules consisting of only carbon and fluorine. These compounds are chemically and biologically inert and are insoluble in both organic and aqueous solutions. However, the extreme hydrophobicity of perfluorocarbons would lead to protein denaturation on contact. The perfluorocarbon can be wetted in the presence of fluorosurfactants which adsorb to the perfluorocarbon By attaching a triazine dye molecule to the fluorosurfactant an affinity ligand IS constructed. The perfluorocarbon emulsion is formed by homogenising the perfluorocarbon in the presence of the surfactant. The resulting emulsion droplets have diameters ranging from 10 to 37 pm (McCreath et al., 1992). The droplet size may be manipulated through the emulsification conditions. Following emulsification the surfactant is cross-linked using glutaraldehyde in the presence of HCl. The reduction of ligand leakage may be achieved by derivatising poly(viny1 alcohol) PVA with fluoroalkyl groups and then using this to coat the perfluorocarbon surface. Perfluorocarbon emulsions have been utilised in an expanded bed configuration. The perfluorocarbon emulsion must be used in an expanded bed to get round the problems of droplet compression which would occur in a fixed bed column. The additional advantage of this method is that the expanded bed arrangement allows any solid particles to flow around the suspended emulsion droplets, thereby giving the potential for direct product removal from fermentation or cell culture broths. The emulsion droplets are allowed to settle in a column and are fluidised by the upward flow of buffer through the bed. At the maximum flow rate this resulted in an expanded bed twice the height of the settled bed (McCreath et al., 1992). By the use of the ligand CI Reactive Blue 4 such a system has been used to adsorb human serum albumin (HSA) from plasma (McCreath et al., 1992). The HSA was recovered at 87% yield and 91% purity. The purification factor was 1.44. Innovative separation methods in bioprocessing 19 1 More recently this technique has been exploited in a continuous reactor for protein purification. Current chromatographic practice results in batch protein purification with the product being eluted and recovered at one stage in the process. Continuous separation would enable the protein to be recovered continuously, and would be amenable to scale up more easily than conventional chromatography. The emulsion reactor is based on liquid-liquid contact between the perfluorocarbon emulsion and the protein solution. The high density of the perfluorocarbon (1.8-2.1 g/ml, Stewart et al., 1992) results in fast settling of the emulsion and aqueous phases and thus is suitable for a liquid-liquid extraction process. The protein is adsorbed onto the affinity perfluorocarbon emulsion which separates from the depleted aqueous phase under gravity. The loaded perfluorocarbon emulsion is eluted with a buffer to recover the protein and then re-equilibrated for further use. Continuous protein purification is carried out using a four- chambered mixer-settler type configuration known as a perfluorocarbon emulsion reactor for continuous affinity separations (PERCAS) (McCreath et al., 1993). Each chamber is identical and consists of a mixing zone agitated by a turbine. The perfluorocarbon- aqueous mixture passes over a weir and into a settling chamber, where the perfluorocarbon emulsion settles to the bottom with the aqueous phase on top. Either phase can then be pumped out either into the top of the next chamber or to waste. In the first chamber the adsorption of the protein onto the emulsion takes place. After settling the loaded emulsion is pumped into the second mixing chamber, where it is washed to remove any trapped contaminants. The washed emulsion is pumped into the third chamber, where the protein is eluted and subsequently recovered. The depleted emulsion into the first chamber to adsorb more protein. This process was tested using the adsorption of HSA from plasma using the dye ligand CI Reactive Blue 2 attached to the perfluorocarbon emulsion described above (McCreath et al., 1993). Protein binding was found to fit the Langmuir isotherm. The continuous protein separation was controlled by varying the aqueous flow rates; the emulsion flow rate was kept constant. The total protein recovery was 89% with HSA recovery at 81%. The overall process yield was 71% with the HSA recovered at 91% purity (purification factor of 1.52). The perfluorocarbon emulsions behave as normal chromatographic materials when operated in a fluidised bed. The protein adsorption capacities are comparable to conven- tional matrices. Advantages of these materials are the fast adsorption and desorption which arise as the droplets are non-porous. The advantages of the PERCAS system are the continuous nature of the operation and the relative simplicity of the system - a mixer- settler combined with gravity settling. 7.3.4 Liquid membrane separations Liquid membrane extraction is a relatively new separation technology which has signifi- cant potential for the selective separation and concentration of low molecular weight chemicals produced by fermentation and used in the food-processing industries. Separa- tion is achieved by the transport of the solute from a feed phase across a film of organic solvent into a stripping phase. Examples of products successfully extracted using liquid membranes include organic acids such as citric acid (Boey et al., 1987), lactic acid is F""@ in20 3 fL\l?fih &aT[>her, Jyl!qwq i> is >T,3&,& 2nd ;rc.-q,i>iha..l&, ax,d s!qex, pup@ 192 (Chaudhuri and Pyle, 1990), and amino acids such as L-phenylalanine (Itoh et al., 1990). Currently, the recovery and purification processes for these species involve several steps specific to individual manufacturers. There is, therefore, scope for the application of liquid membranes which are generally one-step processes, and can simultaneously sepa- rate and concentrate the solute. A schematic diagram of a liquid membrane process is shown in Fig. 7.8. The liquid membrane consists of the organic solvent which separates the two aqueous phases (the feed and stripping phases), and contains a camer species to enhance both selectivity and rates of extraction. Aliphatic diluents are generally preferred as the solvent because of their lower solubility in water. In an ideal situation the solvent should have no solubility in water to ensure that there is no aqueous phase contamination by trace organics. There are two main configurations by which liquid membrane extraction can be exploited, as discussed in the next two paragraphs. A supported liquid membrane (SLM, Fig. 7.9) can be achieved by impregnating a porous solid film with an organic solvent, which is held in place by capillary forces that exist within the pores In order for the membrane pores to be effectively wetted, the surface tension of the solvent must be less than the critical surface tension of the mem- brane polymer. The membrane separates an aqueous phase, initially containing the re- quired species, from an aqueous phase into which the solute is extracted, the stripping phase. The solid supports used are generally microporous polymeric films, e.g. Feed Liquid Stripping phase membrane phase phase J. A. Asenjo and J. B. Chaudhuri Solute - - transfer Fig 7 8 Schematic diagram of liquid membrane processes Solid membrane Stripping phase External phase + Solute flux Fig 7 9 Schematic diagram of supported liquid membrane (SLM) Innovative separation methods in bioprocessing 193 polypropylene, polysulphone, or other hydrophobic materials. Typical dimensions are a membrane thickness of 25-50 pm, with pore sizes of 0.02-1.0 pm. An emulsion liquid membrane (ELM, Fig. 7.10) is formed by creating, under high shear, a dispersion of the stripping phase within the organic solvent which forms a non- porous film around the stripping phase droplets. The emulsion thus formed (stabilised by a surfactant) is dispersed into the feed phase containing the solute, which is then trans- ported into the stripping phase. Depending on the dispersion conditions the globule diameter is 1-2 mm and the internal phase droplets are micron sized. The two aqueous phases cannot physically contact each other and the solute is transported into the internal phase droplets by diffusion through the stabilised solvent film. The use of a chemical reagent in the stripping phase, which reacts with the extracted solute, prevents the solute from diffusing back across the membrane phase. This strategy allows the removal of virtually all of the solute from the feed solution, which makes emulsion liquid membrane extraction very attractive for the recovery of solutes formed in low concentration. After the extraction step the solute is recovered by allowing the emulsion and feed phases to separate, by settling under gravity, then removing the emulsion and breaking it to release the separated solute from the membrane phase components. Electrostatic splitting is generally used for de-emulsification as the membrane components can be recycled for further use. ELM systems give rise to very fast extraction kinetics and allow the use of conven- tional liquid-liquid extraction equipment; they are also prone to emulsion swelling, which gives rise to dilution and instability problems, The necessity to make and break an emulsion does not arise with the SLM system; however, this configuration has slower kinetics, and loss of the membrane phase may occur. In summary, liquid membrane processes offer high separation factors, low capital and operating costs, a lower solvent inventory than solvent extraction, ease of scale-up and the possibility of continuous operation. The key to selectivity in liquid membrane extraction is the use of a carrier species incorporated in the organic solvent to increase the solute solubility: by introducing a 0 External w< 0 c> 0 0 Membrane phase 0 **@ 0"yp Internal phase 0 droplets 0 Fig. 7.10. Schematic diagram of emulsion liquid membrane (ELM). 194 ‘carrier’ molecule into the membrane phase, the solute solubility is increased by the reversible formation of a membrane-soluble carrier-solute complex. This results in faster mass transfer rates, and selectivity is introduced into the extraction as the carrier-solute reaction can be selective. This is known as facilitated transport. The use of a carrier enhances selectivity by the formation of a reversible complex between the carrier and the solute, which is only soluble in the organic solvent. This is particularly effective for the recovery of charged solutes which may be poorly soluble in the organic solvent. Many of the carriers so far employed in liquid membrane processes are extractants used in conventional liquid-liquid extraction, e.g. secondary and tertiary amines, and phosphorus-containing extractants. There have been several applications of liquid membrane extraction to biotechnologi- cal separations. The recovery of citric acid has been studied by Boey et al. (1987). The liquid membrane consisted of Alamine 336 and Span 80 dissolved in Shellsol A. Sodium carbonate was used as the internal phase reagent. This work looked at the batch extrac- tion of both model and real fermentation broths. The results show that very fast extraction of citric acid can be achieved: over 80% of a 5% (w/v) citric acid solution was removed in under 5 min. Significant emulsion swelling was also observed in this study; the volume of the internal phase was more than doubled. The recovery of citric acid in supported liquid membranes has also been reported (Sirman et al., 1990). Lactic acid extraction has also been achieved using emulsion liquid membranes (Chaudhuri and Pyle, 1992a, b). The tertiary amine Alamine 336 was used as the carrier species. Extraction yields were found to be low when recovery was attempted from real fermentation broth. This was a result of competition for the carrier by ‘impurities’ in the broth. In this study significant swelling was observed; however, it was also possible to eliminate the effects of swelling by adjustment of the initial osmotic pressure difference. Scholler et al. (1993) have found that recovery of lactic acid from homogeneous ‘model’ solutions resulted in fast extraction with significant product concentration, whilst extractions attempted from Lactobacillus delbreuckii fermentation broth exhibited poor extraction kinetics and yields. This was believed to be a result of low carrier selectivity for the solute of interest and competition by other compounds present. Work on the extraction of phenylalanine using emulsion systems has been carried out (Itoh et al., 1990). In common with other studies, work has focused on the recovery of the solute from a ‘model’ fermentation broth. In this study the carrier used was Aliquat 336 (tricaprylmethylammonium chloride), in Solvent 1 OON, stabilised by the surfactant Paranox 100. In common with other work, significant emulsion swelling was observed. It was found that swelling increased with the concentration of the internal phase reagent. The nucleotides adenosine and deoxyadenosine have been transported through a chloroform liquid membrane using lipophilic camers. Adenosine mono- and diphosphate (AMP and ADP) have also been transported through chloroform using a lipophilic diammonium salt of diazobicyclooctane (Pellegrino and Noble, 1990). Chiral separation, or the resolution of optical isomers has been attempted for the transport of amino acids with 4-28% enantioselectivity. Chiral transport of sodium mandelate has been used on the optically active carrier N-( 1 -napthyl)methyl- 7a 1- methylbenzylamine dissolved in chloroform. The choice of anion was found to influence chiral selectivity. Separation of racemic N-(3,5-dinitrobenzoyl) 7a 1-amino acid J. A. Asenjo and J. B. Chaudhuri Innovative separation methods in bioprocessing 195 derivatives was achieved using (S)-N-( 1 -napthyl)-leucine octadecyl ester as the carrier in dodecane (Pellegrino and Noble, 1990). Racemic amine salts have been separated into their optical isomers by complexation with an optically active macrocyclic ether and transported from one aqueous phase, through chloroform, into a second aqueous phase. The application of liquid membranes as a separation technique has mainly been in the field of hydrometallurgy, but more recently in biotechnology. Some of these applications have been described above. It is likely that more applications will become apparent through advances in carrier chemistry. 7.4 SOLID-BASED SEPARATIONS Various new and unconventional technologies for biological molecules are presently being developed. These include systems that facilitate the handling of r.aterials by the use of membrane, fluidised bed and novel chromatographic matrix technologies, improving separation specificity and efficiency by using metal, dye and other ligands and also by developing techniques used very efficiently at the analytical scale such as isoelectric focusing now at preparative scale. 7.4.1 Adsorption systems: expanded bed adsorption Expanded bed adsorption or fluidised bed adsorption is a new technique that in one step accomplishes removal of whole cells and cell debris, concentration and initial purification of the target protein. It is based on the design of an adsorbent whose density enables the formation of a stable expanded or fluidised bed (two- or threefold expansion) that allows the establishment of a plug-flow concentration profile across the column. Binding capacities for proteins are similar to those obtained in fixed bed adsorption and linear velocities can be higher than 300 cm/h (Janson and Arve, 1993). Pharmacia has recently started to commercialise expanded bed adsorption under the trade name Streamline. This operation allows the fusing of two or three operations, namely clarification, concentration and capture, into one operation. At the outset, the adsorbent is sedimented in the column. The bed is then expanded by pumping buffer upward through the column. Once the bed is expanded, the buffer is replaced with crude feed which contains the products, cells, cell debris, contaminants and other particulates. After the product is adsorbed, loosely bound material is flushed with a buffer wash and the direction of the liquid flow is reversed with the change to elution buffer and the product is eluted as in packed bed chromatography. If the product is extracellular the process stream is the crude fermentation broth and if it is intracellular the stream is the crude homogenate. The adsorption behaviour of a fluidised bed has been investigated in detail (Draeger and Chase, 1990). A frontal analysis during adsorption to investigate the shape of the breakthrough curve is shown in Fig. 7.1 1. The shapes of the two curves for the adsorption of BSA onto Q-Sepharose FF are very similar, suggesting that the fluidised bed system is behaving in much the same way as the fixed bed system. The apparent maximum adsorp- tion capacity calculated from both runs was 80 g/l, which is consistent with the values from the batch isotherm experiments. A theoretical model for predicting adsorption per- formance in a fixed bed was used to fit theoretical curves to the data from both a fluidised or expanded bed and fixed bed systems (Draeger and Chase, 1990). The model 196 J. A. Asenjo and J. B. Chaudhuri predicted the performance of the system very accurately, indicating again the similarity between the fluidised and fixed bed systems. Fig. 7.11. Fixed and fluidised bed adsorption of BSA to Q-Sepharose fast flow in 0.01 M Tris buffer pH 7.0 (0 fixed bed, o fluidised bed). 7.4.2 Continuous adsorption recycle extraction Continuous adsorption recycle extraction was recently adapted for biotechnological ap- plications as a downstream process for efficient separation of proteins from crude feedstocks to be carried out continuously (Pungor et at., 1987; Gordon et at., 1990). Modem biotechnology involves relatively small-scale processes that favour batch proce- dures, usually following the concept that a batch is well defined (from beginning to end) and that this will pose fewer problems with regulatory approval for therapeutic use. However, it is well known that in the chemical and biochemical process industries batch- to-batch variations can be substantial. It is also well known that a continuous process can operate under steady-state conditions minimising batch-to-batch variations. In addition, a continuous process is much more appropriate for implementing a control strategy in order to maintain constant conditions more readily (Rodrigues et at., 1992). Continuous adsorption recycle extraction, also called CARE, is based on the continu- ous protein adsorption to solid phase supports and features two well-mixed reactors with solids recycle, as shown in Fig. 7.12. The adsorbing stage takes place in the first reactor (adsorber) where the liquid feed is contacted with the adsorbent; desorption of the protein to be purified is obtained in the second reactor (desorber) by maintaining appropriate conditions and by the action of a suitable eluting solvent. The adsorbent beads are recycled to the adsorber reactor while the product is continuously removed. The system' s performance is determined by the nature of the adsorbent and the feed material, flow rates and reactor volume and the conditions of the adsorption and de sorption stages. An accurate mathematical model has recently been developed which can be used for investi- gating optimisation criteria, but also is important for system design and process control (Rodrigues et at., 1992). Innovative separation methods in bioprocessing 197 Eluent , F;ue, Desorber Adsorber recycle Waste Product Fig. 7.12. Continuous adsorption recycle extraction (CAFE) process (see text for key to symbols). During operation, the feed composition and flow rate may experience uncontrolled variations which may upset the steady state. This is also true for adsorbent binding capacity and even recycle flow rates and composition. Hence, appropriate control schemes are needed to meet product specifications and achieve operational stability in the face of potential external and internal disturbances. As a first approximation, the two reactors (adsorber and desorber) are assumed to be perfectly mixed (Pungor et al., 1987). The adsorption process is regarded as a reversible second-order reaction, whereas the desorption stage is modelled as a first-order irrevers- ible reaction scheme (Rodrigues et al., 1992): kl k2 adsorption: A+B + AB k3 desorption: AB --+ A + B where A, B and AB are the target protein, the adsorbent and the adsorbed protein, respectively. The rate constants, k,-,, represent not only the intrinsic adsorption and desorption kinetics, but also include contributions from both external and internal mass transfer resistances (Sherwood et al., 1975). Thus, these effects are lumped into a single coefficient which can be determined experimentally (Chase, 1984). This approach makes the mathematical formulation more tractable at the expense of a less rigorous physical description. On the other hand, experimental evidence suggests that in such a system external mass transfer limitations can be neglected without loss of accuracy (Pungor et al., 1987) and, in the case of high molecular weight proteins, these are mostly adsorbed in active sites located at or near the gel’s surface since molecules attached to those sites block the way to diffusion into the gel. Any thermal effect associated with adsorption and desorption is neglected and the operation is carried out under isothermal conditions. The ordinary differential equations system can be easily solved using the Runge-Kutta fourth- order algorithm. The steady-state solution is obtained by setting the accumulation terms (Le. time derivatives) equal to zero, giving rise to a set of non-linear algebraic equations which can be solved analytically. The steady-state model provides a valuable tool in the formulation of control schemes. Table 7.4 shows the parametric sensitivity of steady state outputs, when each input varies 20% around the baseline value. The sensitivity index S was calculated as the ratio between the percentage variation of the output and the percentage variation of the input, 198 J. A. Asenjo and J. B. Chaudhuri Table 7.4. Steady state sensitivity analysis (sensitivity index, S) output Input Cl c2 41 42 C2/CO F1 0.7 0.3 0.3 0.3 0.2 4m -0.7 0.6 0.6 0.6 0.1 kl -0.3 0.2 0.2 0.2 0.2 k2 0.3 -0.2 -0.2 -0.2 -0.2 k3 0.0 0.0 0.0 0.8 0.0 CO 1.2 0.4 0.4 0.4 0.4 F2 -0.1 -1 .o -0.1 -0.1 -0.6 FR -0.6 0.5 -0.3 0.7 0.4 q, = bound protein concentration in adsorber q2 = bound protein concentration in desorber qm = maximum adsorption capacity of gel All other variables and parameters are defined in Fig. 7.13 and in text. and represents the corresponding steady state gain. It can be seen that all outputs are strongly affected by the feed protein concentration (CO), the gel recycle rate (FR), the gel maximum adsorption capacity (qm) and, to a lesser extent, by the feed flowrate (F1). The eluent flowrate (F2) has a strong influence on the product stream concentration (C2) and the protein recovery ratio (C2/C0). The kinetic parameters (kl-3) do not seem to have a strong influence on performance under the baseline conditions; in the case of k3, 42 is the only output showing some sensitivity. Figure 7.13 shows the time response of protein concentration in the product stream after step changes in feed, eluent and gel recycle flow rates. - O 1.2 G 8 1.0 cu 2 0.8 G 20% increase in: I 0.6 2 2 0.4 - a, FR .e 0.2 00 10 20 30 40 .E 0 0 cu c m c a, Time (hours) Fig. 7.13. Open-loop response of dimensionless production concentration to a 20% step change in input flow rates (Rodrigues er al., 1992). F1 = feed flowrate; F2 = eluent flowrate; FR = recycle flowrate; CO, C1, C2 = protein concentration in streams shown. Innovative separation methods in bioprocessing 199 In conclusion, a simple and effective process control scheme to regulate the continuous adsorption recycle extraction of proteins has been investigated. A dynamic mathematical model, based on unsteady mass balances, was derived and used in the study of the steady state and dynamic behaviour of the system, and to assist in the synthesis of appropriate control schemes. Sensitivity analysis showed that performance was strongly affected by the feed flow rate and feed protein concentration. The eluent flow rate was shown to be the most suitable manipulated variable. The protein concentration in the product stream was successfully controlled using a conventional PI (proportional integral) feedback controller, when measurement time delays are not significant (less than 50% of sampling interval). Model-based feedforward compensation further improved performance in those cases where the main disturbances can be measured on-line in the feed stream. The control philosophy developed can be extended to other separation processes such as continuous two-phase aqueous separations featuring continuous liquid-liquid extrac- tion (Asenjo et al., 1991; Hustedt et al., 1988). 7.4.3 Membrane chromatography Chromatography is a well-used and well-understood method of purifying proteins. How- ever, limitations in this process arise from mass transfer effects that occur in the beads. As a result of the relatively large diameter and porosity of the beads, diffusion times into the beads may be long, which results in broad peaks. In order to shorten the process time new efficient adsorbents are sought. One option was to reduce the bead size to minimise the effects of diffusive transport into the pores. This option gives rise to high pressure drops across the bed. A recent development is that of perfusion chromatography in which the adsorbents have many larger pores allowing greater flow velocities without compromising the resolution (Afeyan and Regnier, 1990). An alternative approach is to use membrane matrices as adsorbents. These have favourable characteristics such as high surface area, high solute throughput, mechanical strength, a highly porous structure and low pressure drops (Fig. 7.14). A potential advantage of membrane chromatography is its use in unclarified broth streams. There are also various process configurations which may be exploited. Membranes that are going to be used as adsorbents must have a high hydrophilicity, low non-specific protein adsorption, a uniform structure, a narrow pore size distribution, chemical and mechanical resistance and the ability to allow ligand coupling (Briefs and Kula, 1992). Ligand coupling may limit the modes of adsorption and types of interaction that may be suitable for membrane chromatography. Dye affinity ligands have been successfully coupled directly to the surface of nylon membranes; however, it was not possible to create an anion exchange membrane by this method and the use of a dextran spacer was required (Briefs and Kula, 1992). Affinity membranes have been created by the coupling of triazine dyes to the mem- brane surface. These dyes have been successfully exploited as agents that mimic enzyme cofactors and other prosthetic groups, and thus may be used as affinity ligands. The technique of dye-ligand chromatography has been developed as a high-resolution purifi- cation technique. The membranes that have been used are nylon-based (Ulitpore, Loprodyne, Immunodyne), and the dyes have been coupled to the membranes either directly or using 200 J. A. Asenjo and J. B. Chaudhuri Flow distributor Membrane stack Gaskets Flow distributor Fig. 7.14. Membrane chromatography system (Briefs and Kula, 1992). 1,6-diaminohexane or polyethylenimine spacers (Champluvier and Kula, 1990). These membranes are flat circular discs used in the laboratory for sample preparation. Affinity membranes have been tested by batch adsorption and by membrane chromatography. In batch adsorption experiments pieces of dye-membrane are shaken with protein solutions in a test tube. In the membrane chromatography mode the dye- membrane discs were encased in a filter holder and protein solutions and buffers intro- duced with a syringe. Alternatively, up to 10 filter housings were stacked and connected to an FPLC apparatus (Pharmacia, Sweden) (Champluvier and Kula, 1991). Experiments were carried out where the membrane-ligand system was compared with a bead-ligand system (Sepharose CL-4B) (Champluvier and Kula, 1990). Adsorption of malate dehydrogenase into Cibacron blue dye was performed. The loading of the protein was similar whatever the carrier. However, higher desorption of the protein was obtained with the beads. Further studies with the proteins adenylate kinase and glucose-6- phosphate dehydrogenase and beads of Sepharose CL4B with yellow and blue dyes were carried out. The volumetric capacity of the membranes was similar to that of the gel beads. However, desorption was poor with respect to the Sepharose. The effect of the spacer was found to be very important. The dye concentration attached to the membrane was found to be five to ten times greater when polyethylenimine spacers were used as compared to directly binding the ligand to the membrane (Champluvier and Kula, 1991). Similarly, adsorption of lysozyme was found to be significantly higher onto dye- membranes rather than the non-specific binding to the undyed membrane. Affinity membrane chromatography has been tested with the purification of glucose-6- phosphate dehydrogenase (G6PDH) from Saccharomyces cerevisiae as a case study (Champluvier and Kula, 1992). The ligand used was Cibacron blue F3GA bound to a Sartorius membrane with a thickness of 210,um and pore size of 0.45 gm. The conclusions of this study were that the particles should be removed prior to adsorption to Innovative separation methods in bioprocessing 201 assure fast operation and long-term performance. Adsorption of dilute enzyme solution was found to be very fast and results from the convective flow through the adsorbent and the fewer short diffusion pathways. Less than 25% of the bulk protein was adsorbed during the loading step, showing good selectivity. Desorption with KC1 and PEG was slow and gave a trailing peak. This was improved dramatically by eluting with the cofactor NADP. Scale-up of this process to a membrane with a cross-sectional area 40 times greater was achieved. The eluted protein concentrations were similar as were the purification factors and yields. Scale-up may also be performed by stacking membranes, provided unacceptable pressure drops do not occur at the desired flow rates. Other enzymes that have been purified by affinity membrane chromatography are pyruvate decarboxylase (PDC) from Zyrnornonas rnobilis and formate dehydrogenase (FDH) from Candida bodinii (Briefs and Kula, 1992). Both enzymes are amenable to purification by dye-ligand chromatography; PDC was recovered using Procion yellow HE-4R and FDH by Procion red HE-3R. PDC was purified with an anionic membrane system and two gel matrices; DEAE-Sephacel and Mono-Q. It was found that the quality of resolution obtained by the membrane was better than the DEAE-Sephacel and nearly as good as the Mono-Q column. The cycle times for purification were shortest with the membrane, 1 minute as compared with 21 minutes for the Mono-Q column and 420 minutes for the DEAE-Sephacel system. As part of this study membrane chromatography was mathematically modelled. It was found that the kinetic effects leading to peak broadening can be neglected at flow rates less than 5 cm/min for affinity adsorption, and less than 20 cm/min for ion exchange. Secondly, the size of the pores has only a minor effect on resolution. In conclusion it is evident that the high flow rate and high efficiency of capture characteristics of membrane chromatography are well suited for processing large volumes of dilute protein solutions. Choice of membrane will be very important. It was found that the available area for binding was not dramatically reduced when membranes with larger pore sizes were used. Pores with diameters of 1.2-3.0 pm are suitable for low pressure, high capacity and good resolution. 7.4.4 Chromatographic and adsorption materials Perj5usion chromatography represents an interesting approach to overcoming speed and throughput limitations in chromatography caused by diffusive mass transport. The pack- ing used for perfusion chromatography contains two classes of pores: throughpores (see Fig. 7.15) to allow convective flow through the particles and smaller, difSuusive pores lining the throughpores to provide high adsorption surface area. Conventional materials contain only diffusive pores and thus solute transport to the binding sites on the medium is limited by the slow diffusion process. In perfusive media, when the mobile phase reaches a sufficiently high velocity, con- vection in the throughpores takes place. In this regime, the combined rapid convective transport and ultra short path length in the diffusive pores make resolution and dynamic loading capacity virtually independent of flow rate with linear velocities, ca. ten times or even higher than with conventional materials. Inverted matrix chromatography uses a resin matrix where the resin is the continuous phase in the matrix and the void spaces are distributed. This is in contrast to the normal 202 J. A. Asenjo and J. B. Chaudhuri Fig. 7.15. Perfusion chromatography. resins which consist of discrete particles. Very low pressure drops are possible and their use in direct broth extraction is presently being investigated (Howell et al., 1993). Affinity chromatography has been used for a number of years both at the laboratory as well as at the large scale. Most early industrial processes for therapeutic proteins included one or more affinity steps. This corresponds to a very high-resolution purification even in the absence of physicochemical knowledge about the main contaminants (see Table 7.1). These, however, relied in many cases on very large ligand molecules (e& antibodies) and were usually exorbitantly expensive at either the large scale or even the laboratory scale. At present there is a tendency to use less expensive, more ‘generic’ ligands. Important advances in this area have been obtained in the last few years which include the synthesis and use of specific dyes and also of metal ions. This technique is also known as immobi- lised metal affinity chromatography (IMAC). 7.5 OTHER DEVELOPMENTS 7.5.1 Electrically enhanced separations Electrophoresis is capable of resolving biological molecules on the basis of differences in their molecular weights, isoelectric points and mobilities in an electric field with a very high resolution. At the laboratory scale this includes some of the most powerful tech- niques available for the purification of biologically active molecules. Isoelectric focusing (IEF) and recycle isoelectric focusing (RIEF) are used in the laboratory as efficient preparative techniques to fractionate complex protein mixtures. The main advantage is the very high resolution; however, removal of ampholites and the low protein solubility at its isoelectric point are disadvantages for large-scale use. Other techniques that have been at least partially scaled up are recycle continuous flow electrophoresis (RCFE) and coun- teracting chromatographic electrophoresis (CACE) (Asenjo, 1990). As electrophoresis is seldom employed at the commercial scale primarily due to the convective mixing problems caused by ohmic heating it has recently been shown that an aqueous two-phase system can be used as a medium for electrophoretic transport and that the liquid-liquid interphase provides stability against convection and facilitates product recovery. The method combines the scalability of aqueous two-phase systems with the selectivity of electrophoresis. Electrically enhanced membrane processes are presently being investigated as poten- tially useful techniques both for resolution of protein mixtures and for the separation of disrupted microbial cells (Brors et al., 1993; White et al., 1993). Innovative separation methods in bioprocessing 203 7.5.2 Genetic approaches to protein purification Recombinant DNA techniques have also allowed the development of highly specific separations. This has been carried out by adding a specific tail to the target protein by genetic engineering techniques (e.g. the zz domain of protein A) which binds a specific ligand (e.g. IgG or a more inexpensive ligand such as metal ion). After purification of the protein the tail is removed usually by enzymatic cleavage. 7.5.3 Purification of intracellular proteins A large number of proteins synthesised in Escherichia coli and yeast are intracellular. These include intracellular protein particles such as recombinant vaccines. Hence the first step in their separation consists of their extraction from the contaminant cell material. Techniques for selectively carrying out the solubilisation or extraction of the protein are presently being developed (Fig. 7.16). These include the use of solvents (e.g. toluene), chelating agents (e.g. EDTA), detergents (e.g. Triton X-100) and chaotropic agents (guanidine and urea). Also, the use of a pure lytic glucanase to selectively release recombinant 60 nm protein particles (virus-like particles or VLPs) from yeast has recently been reported (Asenjo et al., 1993). The VLPs have been used in trials to manufacture an AIDS vaccine already for 3 years. When using the crude lytic complex which in addition to glucanase contained some lytic protease, the protease components of the complex were found to degrade the VLPs. The purified glucanase enzymes from these complexes produced cell lysis without degradation of the VLPs. The Oerskovia lytic glucanase enzyme released the recombinant protein particles selectively as it only produced ca. 17% cell lysis compared to the use of the crude lytic enzyme preparation (with lytic protease). This selectivity, which results in the release of the recombinant particles with only a fraction of contaminating proteins, represents an improvement over presently used mechanical or enzymatic cell disruption processes. This pure lytic glucanase is presently being cloned in E. coli and Bacillus (Ferrer et al., 1993) strains used for commercial production of extracellular enzymes. The large-scale availability of ) / 1A u R Mechanically disrupted cell Permeabilized cell Fig 7 16 Conceptual comparison of mechanical disruption and permeabilisation for obtaining intracellular product release 204 an inexpensive lytic p (1-3) glucanase will find use in improved processes for selective recovery of intracellular proteins from yeast. J. A. Asenjo and J. B. Chaudhuri REFERENCES Afeyan, N. B. and Regnier, F. (1990) Biotechnol, 7,203. Asenjo, J. A. (1990) Separation Processes in Biotechnology, Marcel Dekker, New York. Asenjo, J. A. (1993) New separation processes in biotechnology, Proc. 6th European Cong. Biotechnol., Alberghina, L., Frontali, L. and Sensi, P. (eds.), Elsevier, Amsterdam. Asenjo, J. A., Franco, T., Andrews, A. T. and Andrews, B. A. (1991) In Biologicals from Recombinant Microorganisms and Animal Cells - Production and Recovery, White, M. D., Reuveny, S. and Shafferman, A. (eds.), VCH, Weinheim, pp. 439-454. Asenjo, J. A., Ventom, A. M., Huang, R. B. and Andrews, B. A. (1993) Biotechnology, 11,214-217. Asenjo, J. A., Schmidt, A. S., Hachem, F. and Andrews, B. A. (1994) J. Chromat., 668, 47-54. Boey, S. C., Garcia del Cerro, M. C. and Pyle, D. L. (1987) Chern. Eng. Res. Des., 65, 218-223. Briefs, K.-G. and Kula, M.-R. (1992) Chem. Eng. Sci., 47 (l), 141-149. Brors, A., Kroner, K. H. and Deckwer, W. D. (1993) Cross-flow electrofiltration of disrupted microbial cells, Proc. 6th European Cong. Biotechnol., Alberghina, L., Frontali, L. and Sensi, P. (eds.), Elsevier, Amsterdam. Champluvier, B. and Kula, M.-R. (1990) In Separations for Biotechnology 2, Pyle, D. L. (ed.), Elsevier, Amsterdam, pp. 295-304. Champluvier, B. and Kula, M.-R. (1991) J. Chromat., 539 (2), 315-325. Champluvier, B. and Kula, M.-R. (1992) Biotech. Bioeng., 40, 33-40. Chase, H. A. (1984) J. Chromat., 297, 179. Chaudhuri, J. B. (1991) 2nd Interlaken Conference on Advances in Purification of Re- Chaudhuri, J. B. and Pyle, D. L. (1990) In Separationsfor Biotechnology 2, Pyle, D. L. Chaudhuri, J. B. and Pyle, D. L. (1992a) Chem. Eng. Sci., 47 (l), 47-48. Chaudhuri, J. B. and Pyle, D. L. (1992b) Chem. Eng. Sci., 47 (l), 49-56. Chaudhuri, J. B., Spirovska, G. and Mall, S. (1993) Unpublished results. Dekker, M., Van’t Riet, K., Weijers, S. R., Baltussen, J. W. A., Laane, C. and Dekker, M., Van’t Riet, K., Bijsterbosch, B. H., Wolbert, R. B. G. and Hilhorst, R. Draeger, N. M. and Chase, H. A. (1990) In Separationsfor Biotechnology 2, Pyle, D. L. (ed.), Elsevier, Amsterdam. Ferrer, P., Escuredo, K., Mir, N., Diers, I., Savva, D. and Asenjo, J. A. (1993) An Integrated Approach for the Cloning, Overexpression and Use of Lytic Glucanases for the Selective Protein Release from Yeast Cells, Book of Abstracts, 6th Europ. Cong. Biotechnol., Vol. 111, pp. WE016. combinant Proteins, Interlaken, Switzerland, March. (ed.), Elsevier Applied Science, London, pp. 112-121. Bijsterbosch, B. H. (1986) Chem. Eng. J., 33, B27-B33. (1989) AlChE J., 35,321-324. Innovative separation methods in bioprocessing 205 Fletcher, P. D. I., Howe, A. M. and Robinson, B. H. (1987) J. Chem. SOC. Faraday Trans. Goklen, K. E. and Hatton, T. A. (1987) Sep. Sci. Technol., 22, 831-841. Gordon, N. F., Tsujimura, H. and Cooney, C. L. (1990) Bioseparation, 1,9-21. Hachem, F. (1992) Hydrophobic behaviour of five model proteins in chromatography, precipitation and aqueous two-phase systems. PhD Thesis, University of Reading. Hachem, F., Andrews, B. A. and Asenjo, J. A. (1994) Paper in preparation. Hagen, A. J., Hatton, T. A. and Wang, D. I. C. (1990a) Biotechnol. Bioeng., 35, 955-965. Hagen, A. J., Hatton, T. A. and Wang, D. I. C. (1990b) Biotechnol. Bioeng., 35,966-975. Howell, J. A,, Ming, F., Gollan, K. L., Noel, R. J. and McKee, C. (1993) Inverted matrix chromatography - A possibility for direct broth extraction? Book of Abstracts, 6th Europ. Cong. Biotechnol., Vol. 111, pp. WE021. Hustedt, H., Kroner, K. H., Menge, U. and Kula, M. R. (1985) Protein recovery using two-phase systems, Trends Biotechnol., 3, 139-144. Hustedt, H., Kroner, K. H. and Papamichael, N. (1988) Process Biochem., 23, 129-137. Itoh, H., Thien, M. P., Hatton, T. A. and Wang, D. I. C. (1990) Biotechnol. Bioeng., 35, Janson, J. C. and Arve, B. (1993) Optimisation of expanded bed and fixed bed adsorption of proteins, Proc. 6th European Cong. Biotechnol., Alberghina, L., Frontali, L. and Sensi, P. (eds.), Elsevier, Amsterdam. Leser, M. E., Wei, G., Luisi, P. L. and Maestro, M. (1986) Biochem. Biophys. Res. Comm., 135 (2), 629-635. Luisi, P. L., Bonner, F. J., Pellegrini, A., Wifet, P. and Wolf, R. (1979) Helv. Chim. Acta., 62, 740-753. Luisi, P. L., Giomini, M., Pileni, M. P. and Robinson, B. H. (1988) Biochim. Biophys. Acta., 947, 209-246. McCreath, G. E., Chase, H. A., Purvis, D. R. and Lowe, C. R. (1992) J. Chromatog., 597, McCreath, G. E., Chase, H. A., Purvis, D. R. and Lowe, C. R. (1993) J. Chromatog., 629, Pellegrino, J. J. and Noble, R. D. (1990) TIBTECH, 8,216-224. Pungor, E., Afeyan, N. B., Gordon, N. F. and Cooney, C. L. (1987) Biotechnology, 5, Rodrigues, M. I., Zaror, C. A., Maugeri, F. and Asenjo, J. A. (1992) Chem. Eng. Sci., 47, Schmidt, A. S. (1994) PhD Thesis, University of Reading. Schmidt, A. S., Ventom, A. M. and Asenjo, J. A. (1993) Partition and purification of a- Schomacker, R., Robinson, B. H. and Fletcher, P. D. I. (1988) J. Chem. SOC. Faraday Scholler, C., Chaudhuri, J. B. and Pyle, D. L. (1993) Biotechnol. Bioeng., in press. Sherwood, T. K., Pigford, R. L. and Wilcke, C. R. (1975) Mass Trunsfer, McGraw-Hill, Sirman, T., Pyle, D. L. and Grandison, A. S. (1990) In Separations for biotechnology, I, 83,985-1006. 853-860. 189-196. 201-2 13. 604. 263-269. amylase in aqueous two-phase systems, Enz. Microb. Technol., in press. Trans. I, 84,4203421 2. New York. Vol. 2, Pyle, D. L. (ed.), Elsevier Applied Science, London, pp. 245-254. 206 Stewart, D. J., Purvis, D. R., Pitts, J. M. and Lowe, C. R. (1992) J. Chromatog., 623, 1- 14. White, M. D., Siegel, M. H., Glick, B. R. and Robinson, C. W. (1993) Fractionation of crude protein mixtures using alternating electric field electro-ultrafiltration, Book of Abstracts, 6th Europ. Cong. Biotechnol., Vol. 111, p . WE282. J. A. Asenjo and J. B. Chaudhuri Woll, J. M., Hatton, T. A. and Yarmush, M. L. (1989) Biotechnol. Prog., 5 (2), 57-62. Wuenschell, G. E., Naranjo, E. and Arnold, F. (1990) Bioproc. Engin., 5, 199-202.