Multi-stage laccase extraction and separation using aqueous two-phase systems: Experiment and model

Process Biochemistry 49 (2014) 1020–1031

Contents lists available at ScienceDirect

Process Biochemistry journal homepage: www.elsevier.com/locate/procbio

Multi-stage laccase extraction and separation using aqueous two-phase systems: Experiment and model Axel Prinz a , Katharina Koch a , Andrzej Górak a,b , Tim Zeiner a,∗ a TU Dortmund University, Department of Biochemical and Chemical Engineering, Laboratory of Fluid Separations, Emil-Figge-Straße 70, D-44227 Dortmund, Germany b Lodz University of Technology, Faculty of Process and Environmental Engineering, Department of Environmental Engineering, Wólcza˜ nska 213, 90-924 Lódz, Poland

a r t i c l e

i n f o

Article history: Received 24 December 2013 Received in revised form 14 February 2014 Accepted 7 March 2014 Available online 21 March 2014 Keywords: Aqueous two-phase extraction Laccase Multi-stage experiments Equilibrium-stage-modeling Scale-up

a b s t r a c t This work presents results of experimental and model investigation of continuous multi-stage enzyme extraction using aqueous two-phase systems for the first time. The aqueous two-phase system comprised polyethylene glycol 3000 and phosphate with additional sodium chloride buffered to pH 7. Two different laccases served as model enzymes. One of the laccases was directly taken from fungal culture supernatant, while the other laccase was solubilized lyophilisate. The modeling is based on an equilibrium stage approach. Equilibrium data were taken from single-stage experiments and approximated by different correlation equations. The model describes densities, phase equilibrium, enzyme activity partitioning between the phases. Moreover it allows to consider activity changes due to the aqueous two-phase system. Eight multi-stage mixer-settler experiments under varying operation conditions were performed to validate the proposed model; whereas the total throughput of all multi-stage extraction experiments was about 350 g h−1 . The average relative deviation of modeled activities from experimentally measured activities was 23%. Therefore, the model is able to calculate the behavior of the phases as well as the partitioning of the two enzymes between the two phases for a multi-stage process based on single-stage data. © 2014 Elsevier Ltd. All rights reserved.

1. Introduction In the pharmaceutical as well as in the chemical industry biotechnological products are gaining more and more importance. Every year new enzymes and bio-based drugs such as monoclonal antibodies (mAbs) are discovered, developed and produced. Increasing capacities and titers in biotechnological production processes pose a challenge to the existing downstream processes of biotechnological products in general [1]. The challenge can be met by either improving existing processes or establishing new, innovative ones. Aqueous two-phase extraction (ATPE) belongs to the latter category. This process bases on aqueous-two phase systems (ATPS), which can be formed by the solution of two hydrophilic, but incompatible components in water [2]. Examples of such ATPS are the aqueous solutions of two polymers (e.g. polyethyleneglycol (PEG)-dextran) or the aqueous solution of a polymer and a salt (e.g. PEG – phosphate). Fig. 1 shows a rather rough and general scheme of the alternatives and steps in downstream processing of

∗ Corresponding author. Tel.: +49 2317552670. E-mail address: [email protected] (T. Zeiner). http://dx.doi.org/10.1016/j.procbio.2014.03.011 1359-5113/© 2014 Elsevier Ltd. All rights reserved.

biotechnological products and how the ATPE can replace existing process options. In general, process capacity decreases, while the purity increases. Due to its high capacity and reasonable purity gain, aqueous two-phase extraction is considered as an option in early downstream processing. In case of reduced purity demands, e.g. for some industrial enzymes, the downstream process may stop at an earlier stage with less purity. It is well known that multi-stage purification can lead to improved product recovery and/or purity. Already at the end of the 1970s, the group of Kula [3–5] has reported enzyme isolation and separation using ATPS in a centrifugal extractor comprising multiple separation stages. More recently, Rosa et al. have investigated multi-stage antibody purification in ATPS experimentally in test tubes [6], a mixer-settler apparatus [7] and a packed extraction column [8]. In spite of extensive knowledge of fundamentals of ATPS and their applicability in purification and separation processes, reports on industrial application are scarce. Available publications report about strong influences of parameters, such as components forming the phase system, ionic strength, molar mass of polymers, concentrations, temperature, additives and the product itself, on extraction performance [9,10,2]. The multiplicity of parameters

A. Prinz et al. / Process Biochemistry 49 (2014) 1020–1031

1021

Fig. 1. General scheme of the purification of biotechnological products.

causes a high experimental effort for developing ATPE processes. Process modeling contributes to reducing experimental effort and enables an efficient process design- and variant evaluation. Mistry et al. [11,12], Ahmad et al. [13,14] and Samatou et al. [15] have proposed process models using ATPS. Mistry’s and Ahmad’s models lack experimental validation and Samatou validates only the partitioning of a mAbs and not for an enzyme which brings several new challenges along. The scope of this work was to model a multi-stage aqueous two-phase enzyme extraction process based on single-stage data and to validate it experimentally. Pleurotus sapidus culture supernatant containing laccase was spiked with laccase from Trametes versicolor. This spiked supernatant containing two different laccases was proceed in an ATPS consisting of

PEG3000 and phosphate in order to show the separation potential of this technology. Moreover this ATPS was chosen for the extraction of the used enzymes, as the concentration of PEG is comparable low for the formation of two phases and so these systems have a lower viscosity. As the phases have to be pumped from one stage to another, it is an important property of an ATPS used for multistage extraction. Sodium chloride, as a partitioning influencing additive, was included in the modeling framework and the experiments. In [18] single-stage separation of the laccases was investigated thoroughly resulting in clearance factors of 5.23 for laccase from P. sapidus and 6.45 from T. versicolor serving as a bench mark for the evaluation of the multi-stage separation.

1022

A. Prinz et al. / Process Biochemistry 49 (2014) 1020–1031

2. Materials and methods In this work, two different sources of laccases were used. One laccase was the cell supernatant (CS) of a submerged culture of Pleurotus sapidus (P.s.). The fermentation was performed according to the procedure published by Linke et al. [16]. After the fermentation, the CS was frozen at −20 ◦ C in aliquots of 1 L. For extraction experiments, aliquots were defrosted and filtered (5–13 ␮m). The laccase activity of the defrosted and filtered sample was around 1100 U/L and the protein concentration about 120 mg/L. As the second laccase from Trametes versicolor (T.v.) was chosen. This laccase is in form of a white powder and was purchased by Sigma–Aldrich (cat.# 38429; ≥0.5 U/mg). For the experiments, the powder was solubilized in water at room temperature and afterwards filtered (5–13 ␮m). In the course of the separation experiments with mixtures of both laccases (P.s. + T.v.), the enzyme activity of laccase from T.v. was adjusted to the laccase activity in the CS by adding the appropriate amount of laccase powder to the culture supernatant. For all experiments and dilutions, water purified by a Milli-Q Synthesis A10 system (Millipore, conductivity K ≤ 0.06 ␮S/cm) was used. The activity assay was done using 2,2 -azino-bis-(3-ethylbenzothiazoline-6-sulphonic acid) (ABTS) with a purity of ≥98% which was purchased from Sigma–Aldrich (cat.-# A1888). Sodium chloride (NaCl) with a purity of ≥99.5% was acquired also from Sigma–Aldrich (cat.-# 71379). Polyethylene glycol with a molecular weight of 3000 g/mol Ph. Eur. grade (PEG3000) was purchased from Merck (cat.-# 34042000). Di-potassium hydrogen phosphate trihydrate (≥99%) was bought from AppliChem (cat.-# A4001). Sodium dihydrogen phosphate dihydrate Ph. Eur. grade (99.8%) was acquired from VWR Prolabo (cat.-# 28014.360). For the experiments a phosphate stock solution of 0.29 g/g PO4 3− and a PEG3000 stock solution with a mass fraction of 0.50 g/g were used; whereas the pH of the phosphate solution was adjusted to pH 7 by varying the amount of sodium and di-potassium phosphate. The mass ratio of sodium dihydrogen phosphate dihydrate and di-potassium hydrogen phosphate trihydrate was about 0.5. The pH was finally adjusted by adding the corresponding salt. Sodium/di-potassium phosphate buffer solution with a pH of 7 will be referred to as “phosphate” and considers only the mass of the anion. In case of experiments with sodium chloride, 0.025 g/g of the salt was added to the stock solutions, CS and enzyme solutions as well as to the water used for adjusting the mixing points.

The laccase activity act was measured using the 2,2 -azino-bis-(3ethylbenzothiazoline-6-sulphonic acid) (ABTS) assay according to Majcherczyk et al. [17] using a Multiskan FC well plate reader and calculated using:

  U L

=

DF · Vt · E t · Vs · d · ε420

2.3. Measurement of phase forming components PEG3000 mass fraction (wtPEG ) was determined by high performance liquid chromatography (HPLC). The procedure has been described in detail in [18]. Sample standard deviation, dilution error and standard error of the calibration and were used for Gaussian error propagation. The maximum and average absolute errors of the PEG measurement were 0.0128 g/g and 0.0038 g/g, respectively. The phosphate (wtPO4 ) and chloride mass fractions were determined by ion chromatography. In detail this procedure has been described in [18], whereas the maximum and average errors of the measurement were 0.0102 g/g and 0.0066 g/g for phosphate and 0.0085 g/g and 0.0054 g/g for sodium chloride, respectively. Errors are indicated in diagrams using error bars. 2.4. Electrophoresis For the qualitative analysis and differentiation between laccases of P.s. and T.v., native isoelectric focusing (IEF) was applied. The gels (SERVAGelTM IEF 3-10 for IEF (cat.-# 43240)) were purchased from SERVA, and the electrophoresis was performed as described by the manual. The electrophoresis was done in a Mighty Small II SE260B chamber from Hoefer® , and a SERVA BluePowerTM 3000 served as power supply; whereas the samples were pretreated using centrifugal filters (VWR Centrifugal Filter PES 10K) with a molecular weight cut-off of 10 kDa in a MiniSpin Plus Centrifuge from Eppendorf at 12,100 times the gravity in order to remove phase forming components such as electrolytes, which interfere with electrophoresis. Staining was achieved by incubating in 5 mM ABTS-buffer for up to 10 min. Afterwards, the gels were scanned instantly for documentation. 2.5. Multi-stage mixer-settler experiments

2.1. Measurement of laccase activity

act

by Smith et al. [19] with bovine serum albumin as standard as described in [18]. Using this method the protein concentration of the CS was estimated as about 120 mg/L. Considering the dilution (up to 3-fold) in the extraction process and additionally the further dilution due to the multi-stage operation, further reducing the total protein concentration, protein specific enzyme purity (U/mg) is neglected and only volumetric laccase activity (U/L) is considered.

(1)

The procedure is described in detail in [18]. In Eq. (1), the dilution factor (DF), total volume Vt , change of absorption E t, sample volume Vs , sample height in the wellplate cavity d and the extinction coefficient at 420 nm ε420 = 0.04321 L/(␮mol cm) are applied for the calculation of laccase activity. 2.2. Measurement of total protein concentration The total protein concentration was measured on a Multiskan FC well plate reader from Thermo Scientific using the bicinchoninic acid (BCA) assay introduced

For investigating continuous multi-stage extraction, experiments in a mixersettler miniplant were conducted. In addition, these experiments were used for the validation of the multi-stage process model. The experiments were performed in a mixer-settler-unit designed by Bayer Technology Services GmbH, partly produced and assembled by Normag and shown in Fig. 2. The unit comprised 15 stages and each stage consisted of a pump-mixer and a double-jacketed settler made of glass, but for the experiments only three stages were used. The pump-mixer (65 mL volume) dispersed the phases entering from the adjacent stages and pumped them into the settler (DN30, 135 mL) of the stage. Every part of equipment in contact with media is made of glass or polytetrafluoroethylene (PTFE). In the settlers, phases separated. In Fig. 3, dispersed feed from the pump-mixer enters the settler on the right side. After entering, the dispersion settled and the interphase formed. The simultaneous outlet flow of top and bottom phase was realized as shown in detail in the figure on the right side. Only bottom phase can enter the cylinder in the center as long as the interphase is above the lower end of the outer cylinder. The bottom phase

Fig. 2. Mixer-settler-unit (with courtesy of Normag).

A. Prinz et al. / Process Biochemistry 49 (2014) 1020–1031

1023

Fig. 3. Scheme of the outlet flow from the mixer settler.

leaves the settler through a tube with an oval hole in it. The height of the tube and hereby the height of the hole can be adjusted using a screw at the top. The top phase leaves the settler over a tubing shown in the top left side in the figure. Hereby, the tubing operates as a weir. The difference of height between top phase surface outside the cylinder and bottom phase surface in the cylinder depends on the height of the interphase as well as the densities of the phases. The adjustable height of the bottom phase outlet enables a simultaneous flow of top and bottom phase from the settler to accelerate the achievement of hydrodynamic steady-state. The settlers and feed streams were tempered by two cryostats (F33-EH, JULABO). For the settlers this was done using the double jackets and for the feed streams, using heat exchangers made of glass. The individual mixer-settler units were connected in a counter-current way. The pump-mixers were run by a transmission, which operated at about 1000 rpm. Since the pump-mixer convey only the liquid, which has entered them, into the subsequent settler, total mass and volume flows in the mixer-settler miniplant depended only on the convective inlet streams on the first and last stage. Four membrane pumps for the inlet mass flows (Ritmo® 05/120, Fink Chem + Tec) were controlled by scales (XS160001M, METTLER TOLEDO) on which the PEG- and phosphate-stock solutions as well as two water tanks were placed. Each of these two inlet pumps (master) controlled another pump (slave), which pumped either water or enzyme solution into the continuous extraction process.

The ratio of stock solution to water or supernatant was chosen in such a way that the feed mass fractions of PEG- and phosphate feed lied on extensions of the working tie line. This operation minimized the mass fraction changes of PFCs over the stages. The ratio between the amount pumped by master and slave was entered into the master pump. The global mixing point of phase forming components in the process could be calculated from the mass flows of the PEG- and phosphate stock solutions as well as the additional water streams to dilute the top and bottom feed streams. This point enables a comparison of different experiments. Hereby, the theoretical mass fractions of PEG and phosphate were calculated based on all four entering streams constituting the global mixing point. The outlet streams (TP-n and BP-1) were recorded using scales (XS160001M, METTLER TOLEDO). All mass flows entering and leaving the extraction process were monitored and recorded using a process control system (LabVision® , HiTEC ZANG). The top phase stream was fed at stage 1, while the bottom phase entered the extraction process at stage n (see Fig. 4). The continuous extraction experiment was started without enzyme solution, but only water instead. First, mixers and settlers were filled with equilibrated phase system using a phase ratio of one. Then the pump-mixers were started to form dispersion from the two phases in the mixer. After that, the membrane pumps were started to impose convective mass flow on the extraction process.

Fig. 4. Experimental setup for multi-stage mixer-settler miniplant experiments.

1024

A. Prinz et al. / Process Biochemistry 49 (2014) 1020–1031

Table 1 Extraction parameters for multi-stage extraction experiments. EXP

˙ PEG [g h−1 ] m

˙ H2 O(PEG) [g h−1 ] m

˙ PO4 [g h−1 ] m

˙ H2 O(PO4 ) [g h−1 ] m

Feed

Inlet

NaCl [g/g]

PO4 [g/g]

PEG [g/g]

1 2 3 4 5 6 7 8

103 127 126 126 126 127 118 118

63 62 62 65 63 63 63 63

59 62 62 63 63 63 58 58

112 116 118 124 120 124 117 116

P.s. + T.v. P.s. + T.v. P.s. + T.v. P.s. + T.v. T.v. P.s. T.v. P.s.

TP BP TP BP BP BP BP BP

0.000 0.000 0.025 0.025 0.025 0.025 0.000 0.000

0.052 0.050 0.048 0.048 0.049 0.049 0.049 0.049

0.151 0.171 0.162 0.158 0.159 0.158 0.163 0.163

At this point, the interphases in the settlers fluctuated and had to be controlled manually using the vertically adjustable height of the bottom phase outlet in the settlers (see Fig. 3). Once a hydrodynamic stable condition was reached, a tank with enzyme solution replaced one of the water tanks on either side and hereby the extraction process was started. Samples were taken at every stage from the stage-exiting streams. Every sample was diluted to 10 wt.% of water to avoid phase separation due to temperature shifts. Immediately afterwards, activity assays were performed. Fig. 4 depicts a flow sheet of the n stage extraction process. In the course of these experiments, feed stage for the culture supernatant or enzyme solution (1st and 3rd), type of feed (enzyme solution with T. versicolor laccase, culture supernatant with P. sapidus laccase and culture supernatant with P. sapidus laccase spiked with T. versicolor laccase) and sodium chloride mass fraction (0 and 0.025 g/g) were varied. Table 1 displays the conditions of the extraction experiments. In case the feed ˙ H2 O(PEG) ) was substituted with CS and in case was the top phase, the water stream (m ˙ H2 O(PO4 ) ) was substituted with CS. it was the bottom phase the water stream (m Since the number of parameters to be varied in an extraction process applying ATPS is very large, it is limited to feed position and sodium chloride mass fraction. 2.6. Calculation of key figures A very important parameter to evaluate the characteristics of an extraction in general is the partitioning coefficient between the two liquid phases. In enzyme extraction, the activity partitioning coefficient Kact,k is defined according to Eq. (2): Kact,k

actTP,k = actBP,k

(2)

The index k refers to the source of laccase (P.s. or T.v.), respectively. In order to evaluate the experimental results, the recovery Yp,i,k (see Eq. (3)) was calculated using the volume Vp or volume flow Vp . Hereby, p represents the phase (TP or BP) the laccase activity was measured in and ini represents the initial values prior to the extraction: VP,i · actk YP,k [%] = × 100 Vini · actini,k

(3)

For mass and component balances, the total recovery is physically fixed to 100%. In contrast, this in not necessarily true for enzymatic activity since it depends on the physiological environment and can therefore be inhibited or activated. Therefore, balances with more than 100% recovery indicate an activation of the enzymatic activity, or the separation of inhibitors. Laccase purity Xij (Eq. (4)) compares laccase activity of one source i to the activity of both sources together in phase j. In contrast to activity recovery this key figure is restricted to 100%: Xij [%] =

actij actij,tot

× 100

(4)

In order to assess high-resolution polishing of proteins separation, Asenjo proposes the clearance factor CFij [20]. This factor (Eq. (5)) is applied for the assessment of the separation of one enzyme j. CFij =

Xout · (1 − Xin ) Xin · (1 − Xout )

(5)

Since the ABTS assay is not able to distinguish between laccases from P. sapidus and T. versicolor, superposition of single-stage experiments [18] was chosen as a method to calculate laccase purity Xij . In case of multi-stage experiments, modeling enables a calculation of the clearance factors, because the partitioning of each laccase is calculated individually.

3. Modeling For the process simulation of extraction processes, the modeling of the ATPS is essential. In literature there are several thermodynamic models proposed to compute ATPS [21]. In this work the partitioning of enzymes on the two phases is investigated by their

enzyme activity. But this activity distribution cannot be modeled by thermodynamic models as the exact mass fraction of the enzymes is not known. Moreover the used ATPS consisting of PEG, phosphate, sodium chloride and water includes six species. The modeling of such ATPS is quite challenging and needs a lot of numerical effort. This numerical effort would increase disproportionally the simulation time. For this reason the ATPS is modeled using a correlation. The general principle of this modeling approach was already proposed by Mistry et al. in [11], but it was limited to two stages and the separation of proteins, which can be modeled using a mass balance. We extend it to enzyme separation using a deactivation and an activation factor because of the components of ATPS. Moreover multi-stage operation in a different aqueous two-phase system containing buffered phosphate and PEG3000 can now be modeled. The correlation for the binodal curve (BC) used in this work calculates the mass fraction of the PEG of the binodal curve as follows [11]: wtPEG,BC = A1 wtPO4 · e(A2 ·wtPO4 +A3 wtNaCl ) + (A4 − wtPO4 ) ·e(A5 wtPO4 +A6 wtNaCl )

(6)

where the parameter Ai are fitted on experimental data. Using a correlation the tie line cannot be calculated directly with the binodal curve, but it has also to be correlated. The correlation for the tie lines (TL) is a linear equation depending on the phosphate concentration: wtPEG,TL = Bi (wtPO4 − wtmix,PO4 ) + wtmix,PEG

(7)

where the subscript mix means the mixing point on the tie line and the factors Bi are calculated as: Bi = Ba + Bb · wtNaCl + Bc · TLL

(8)

Ba , Bb and Bc are coefficients for the correlation of Bi . TLL is the tie line length, which is defined as:



TLL =

(wtTP,PEG − wtBP,PEG )2 + (wtTP,PO4 − wtBP,PO4 )2

(9)

For the correlation of the activity partitioning coefficient an equation was found by a non-linear optimization method. Depending on the thermodynamic phase equilibrium the partitioning of both laccases can be calculated as: Kact,k = (P1,k · wtmix,NaCl + P2,k · TLL)actmix,k (P3,k ·wtmix,NaCl +P4,k ·TLL) k = P.s., T.v

(10)

whereas the constants Pi,k are fitted on experimental data for the partitioning of laccase from P. sapidus and laccase from T. versicolor. Based on the thermodynamic phase behavior an equilibrium stage model [22] can be developed. The model proceeds in two sequential steps. In the first step, the equilibrium of the ATPS is modeled. The four components considered as PFCs are water, PEG, phosphate and sodium chloride, whereas phosphate and sodium chloride is considered as undissociated. In preliminary experiments, it was shown that the in the investigated concentration range the enzymes and impurities have no significant impact on the ATPS. During this step equilibrium conditions as well as mass

A. Prinz et al. / Process Biochemistry 49 (2014) 1020–1031

and component balances are used to calculate mass flows and mass fractions in each stage of the extraction process. In general, the outlet mass fraction of the PFCs and enzyme activities depend on the mixture (mix) entering the stage, not accounting for the occurrence of two phases. In the second step, the mass fractions and mass flows of the PFC serve as a model input variable for the calculation of the phase equilibrium of the two laccases’ partitioning. The general principle of this modeling approach was already proposed by Mistry et al. in [11], but it was limited to two stages and the separation of proteins. This work extends it to enzyme separation and multi-stage operation in a different aqueous two-phase system containing buffered phosphate and PEG3000. 3.1. Modeling of PFC Mass balances (1 per stage; massbal ) and component balances (3 per stage; compbal ) as well as equilibrium conditions (3 per stage; compeq ) and the lever rule (1 per stage; LR) [23] formulate the system of equations describing the extraction system and resulting in all the stage exiting mass fractions. In the model, the following indices are used in the equations: stage (i; 1 → number of stages), component (j; 1 → m), phase (p; TP and BP) and type of laccase (k; P.s. and T.v.). Input variables for this first step of the model are ˙ TP,0,j and m ˙ BP,n+1,j , whereas the entering component mass flows m all the other component mass flows and mass fractions of PFCs on each stage are calculated by the model. Since the number of equations increases linearly with the number of stages, the model is not limited by number of stages, but only by computing capacity. In this first step of the model only the components PEG3000, phosphate, sodium chloride, and total mass are included. For every stage the total mass balance has to be solved: ˙ BP,i+1 − m ˙ TP,i − mBP,i = 0 ˙ TP,i−1 + m m

(11)

Eq. (11) can be explained by looking at Fig. 5. The component balances for each stage and component are defined as: ˙ BP,i+1 − wtBP,j,i wtj,TP,i−1 · mTP,i−1 + wtBP,j,i+1 · m ˙ BP,i − wtTP,j,i · m ˙ TP,i = 0 ·m

(12)

The mixing point for the calculation of the tie line of stage i can then be calculated as follows: wtmix,i,j =

˙ TP,i−1 + wtBP,i+1,j · m ˙ BP,i+1 wtTP,i−1,j · m ˙ BP,i+1 ˙ TP,i−1 + m m

(13)

Preliminary experiments have shown that the chloride ions partition evenly, if the sodium chloride mass fraction does not exceed 0.04 g/g. Therefore, it was assumed that sodium chloride partitions evenly and the sodium chloride mass fractions are the same in both phases on every stage. So the tie line can be calculated using Eqs. (7)–(9). The composition of each phase can be calculated by the intersections of the tie line with the binodal curve (Eq. (6)). So the composition of in and outlet streams of stage i are known.

3.2. Correlation of density Since the enzyme activity actp,i,k is an enzyme specific entity  with the unit U/L, it is necessary to calculate volume flows Vp,i using correlations for the densities  of the equilibrated and entering phases. In this correlation, sodium chloride as well as the major PFCs in the phase were used, because the densities of the phases showed a linear dependency on sodium chloride and main phase forming component. These correlations were fitted from singlestage data [18] and resulted in the following linear Eq. (14) for four different types of streams in the model: P = D1 + D2 · wtNaCl + D3 · wtPEG/PO4

(14)

D1 , D2 and D3 are coefficients to correlate the densities. Using the total mass flows and the density correlations, it is possible to calculate volume flows and use them for modeling of the activities in the ATPS. 3.3. Modeling of laccase partitioning After the calculation of the volume flows using the mass flows and densities, the calculation of the laccase partitioning can be conducted. In analogy to the calculation of ATPS, enzyme activity on each stage and in each phase was calculated using activity balances and activity partitioning equilibria. The product of enzyme activity  [L/h] is defined as the enzyme actp,i,k [U/L] and volume flow Vp,i flow Z˙ p,i,k [U/h]. Input variables for this second step of the modeling are the entering activity streams Z˙ TP,0,k and Z˙ BP,n+1,k . Single-stage data was taken from experiments already published [18]. In analogy to the component balances, the general balance equation is aligned as follows: Z˙ TP,i−1.k + Z˙ BP,i+1.k − Z˙ TP,i,k − Z˙ BP,i,k = 0

(15)

A previous work has shown the activity in- and decreasing influences of phosphate, PEG, and sodium chloride [18]. Stage equilibrium modeling relies on thermodynamic phase equilibria as well as on mass and component balances, but enzyme activity changes depending on the physiological environment and cannot be simply balanced. Therefore, measures had to be taken to enable a more accurate modeling of enzyme activity. It was newly found that the inclusion of an activity in- and decreasing factor on the feed stage based on initial single-stage experiments leading to a certain activity yield Yk,p corrects the activity balances and enables adequate modeling. For the bottom phase of extractions of T. versicolor and both phases of extractions of P. sapidus the following equation is applicable: Yk,j,p = Ak,p · wtj,p + Bk,p · wtNaCl,p + Ck,p

(16)

For the activity yield in the top phase of extractions of T.v., a correlation was found and embedded into the model as follows: YT.v.,TP =

AT.v.,TP (BT.v.,TP · wtTP,PEG + wtTP,NaCl )

(17)

The activity in- or decrease Ytot on the feed stages is equal to the sum of YTP,k and YBP,k . The differences between both approaches of the correlation for the activity yield of laccase of P. sapidus and T. versicolor will be briefly discussed in Section 4.3. Using the entering top and bottom phase activities are corrected to compensate for the activity in- and decreasing effect of the phases: Z˙ TP,o,k = Ytot,0,k · V˙ 0,TP · act0,k

Fig. 5. In and outlet flows of stage i.

1025

and

Z˙ BP,n+1,k = Ytot,n+1,k · V˙ n+1,BP · actn+1,k

(18)

1026

A. Prinz et al. / Process Biochemistry 49 (2014) 1020–1031

The equilibrium condition for the partitioning has to be fulfilled as well: Ki,k (wtmix,i,NaCl , actmix,i,k, TLLi ) −

actTP,i,k =0 actBP,i,k

(19)

For the process simulation of a countercurrent n-stage mixersettler extraction unit, these equations have to be solved for all stages simultaneously. The results of these simulations are shown in the next chapters. 4. Results and discussion 4.1. Modeling results of PFC In Table 2 the parameters for the correlation of equilibrium compositions of the PFCs on each stage using the BC (Eq. (6)) and tie line (Eqs. (7)–(9)) are displayed. Fig. 6 displays parity plots for experimental (exp) and correlated (corr) PFC equilibria as well as densities discussed later. Low mass fractions of PEG and high phosphate mass fractions correspond to the bottom phase and vice versa for the top phase. From these figures it can be concluded that the equilibrium modeling works in good agreement of both phases and all PFC. The proposed model is valid in the ranges from 0 to 0.4 g/g, 0 to

Table 2 Parameters for binodal curve (Eq. (4)) and slope of the tie line (Eqs. (5)–(7)). Parameter

Value

Parameter

Value

A1 A2 A3 A4 A5 A6

54.05 −63.70 −16.70 0.6073 −153.0 7.654

Ba Bb Bc

−4.136 6.621 3.037

0.15 g/g and 0 to 0.025 g/g for PEG3000, phosphate, and sodium chloride, respectively. Experimental errors are always within deviation range of 0.01 g/g for PEG and phosphate and even 0.003 g/g for sodium chloride. 4.2. Correlation results of densities The correlations proposed in Eq. (14) were used to fit the densities of the top phase TP , bottom phase BP , PEG-containing feed phase PEG-FEED and phosphate containing feed phase PO4 -FEED . In Fig. 6d, a parity plot of the correlated (corr) and experimentally (exp) determined densities are shown. Here, the maximum deviation is about 2%, but 97% of the correlated densities are a lot more precise (<0.86%). Table 2 displays the parameters fitted to the

Fig. 6. Parity plot of experimental and modeled PEG (a, ), phosphate (b, 䊉) and sodium chloride (c, ) mass fraction and densities (d) of TP (+) and BP (×) with deviation lines marked in the figures. Errors are indicated using error bars.

A. Prinz et al. / Process Biochemistry 49 (2014) 1020–1031 Table 3 Parameters and range of validity for density correlations. Parameter D1 D2 D3 R2 x [g/g] y [g/g] xlow [g/g] xup [g/g] ylow [g/g] yup [g/g]

1027

Table 4 Parameters and range of validity for partitioning coefficient Kk correlations.

TP [kg m−3 ]

BP [kg m−3 ]

PEG-FEED [kg m−3 ]

PO4 -FEED [kg m−3 ]

1079 157.6 5.768 0.8134 wtNaCl wtPEG 0.0 0.025 0.195 0.298

1019 355.6 1131 0.8438 wtNaCl wtPO4 0.0 0.025 0.085 0.115

998.5 178.8 695.2 0.9998 wtPEG wtNaCl 0 0.5 0 0.025

991.3 1550 869.8 0.9955 wtPO4 wtNaCl 0 0.29 0 0.025

experimental data and the accuracy of the fit. Corresponding to the experimental interval Table 3 displays, in what range the density correlations are valid.

4.3. Modeling results of laccase partitioning In Fig. 7, parity plots of the correlation equations for the partitioning of laccase from P.s. and T.v. are displayed. Using Eq. (19) it is possible to correlate most of the partitioning coefficients for laccases from both fungi within a relative deviation range of ±20%. The largest deviation occurs for partitioning coefficients below 0.1. In this case, already most of the enzyme activity is present in the extract phase, while in the raffinate phase activities reach the magnitude of the analytical errors. From a modeling perspective, it is of major importance to have good agreement for partitioning coefficients around one, because at extreme activity partitioning coefficients (e.g. 0.1 > Kact > 10) the absolute change in activity in the product phase is rather small. Therefore, high accuracies of the activity partitioning coefficient are not crucial, in cases of partitioning coefficients below 0.1 and above 10. In addition, analytical errors get an increasing impact on the accuracy of the measured activity partitioning coefficient and make the values for Kact less reliable. In Table 4, parameters for the correlation (Eq. (15)) are displayed. The correlation is valid in ATPS mixtures for activities from 5 to 500 [U/L], for TLL from 0.2 to 0.32 [g/g] and for sodium chloride mass fractions up to 0.025 [g/g].

Parameter

KP.s.

KT.v.

P1,k P2,k P3,k P4,k

−36 6.9 −8.5 0.81

7.8 1.6 −16.2 −0.616

Table 5 Parameters for the influence of PFC on the activity yield in the phases. Parameter

Ak,p

Bk,p

Ck,p

R2

YP.s.,TP YP.s.,BP YT.v.,TP YT.v.,BP

1.944 −5.586 0.0007711 0.3219

−24.35 10.97 0.01365 −0.831

0.4886 0.7136 – 1.051

0.931 0.8752 0.8723 0.0188

In order to deal with the challenge of balancing enzyme activities, parameters for activity in- and decrease were fitted according to Eqs. (17) and (18). These parameters are displayed in Table 5. The coefficient of determination is good for the first three correlation equations. The parameter Bk,p for the activity recovery of laccase from P. sapidus is smaller than zero in the top phase and larger than zero in the bottom phase. Since this is the coefficient of the sodium chloride concentration in the ATPS, it correlates with an increasing activity recovery in the bottom and decreasing activity recovery in the top phase with increasing sodium chloride content. Ak,p is the coefficient for the main phase forming component in the respective phase (PEG in the top and PO4 in the bottom phase). This result indicates that increasing tie line length increase the activity recovery in the top phase and decrease the recovery in the bottom phase and corresponds to previously published results where increasing activity partitioning coefficients of laccase from P. sapidus were observed for longer tie lines [18]. In contrast, the coefficient of determination for the correlation of the activity recovery in the bottom phase of an extraction with T.v. is very low. This is the case, because the activity in- or decrease of laccase from T.v. in the bottom phase is almost negligible and shows only very little dependency on the given variables. Most of the enzyme activity is already in the phase regardless of the factors varied. Looking at the parameters AT.v.,BP and BT.v.,BP in Eq. (17), this limited influence of the PFC and sodium chloride on the activity yield is apparent. Still,

Fig. 7. Parity plots for partitioning coefficients of P.s. (a; ) and T.v. (b; ) with relative deviation lines at 20%. Errors are indicated using error bars.

1028

A. Prinz et al. / Process Biochemistry 49 (2014) 1020–1031

Fig. 8. Test for steady-state conditions on stages 1 (), 2 (䊉) and 3 () in the top phase and on stages 1 ( ), 2 ( ) and 3 ( ) the bottom phase for PEG (a) and phosphate (b) mass fraction and laccase activity (c) over time. The arrow indicates the feed stage. Errors are indicated using error bars.

for a general application and extension of the model the yield here was treated as a function of the PFC and sodium chloride.

4.4. Establishing steady-state in mixer-settler unit As in this work the steady-run of the mixer-settler is investigated this state has to be established. In order to show a change of activity and mass fraction of PFC samples were taken in the course of an experiment. In Fig. 8 the enzyme activity as well as the mass fractions of PEG and phosphate are plotted over time. At 0 h, water was exchanged by CS. The mass fractions of the PFC show that the phase system was already in steady-state before CS entered the system and in addition the CS had no influence on the phase system since mass fractions are constant over time. Only the phosphate mass fractions in samples of the bottom phase of stage 2 at the beginning and of stage 3 after about 8 h vary. Since the remaining mass fractions are constant these are considered as outliers. The slight deviations of activities in Fig. 8c show that after about 5 h the activities in the different phases vary only within the experimental error. Therefore, it was assumed that after 7 h of operation, steady-state is reached. Together with the total throughput of about 300 mL/h this correlates with a residence time of <3 h and therefore 3.5–4 exchanges of the total volume in the multi-stage mixer settler unit. In every following diagram, experimental points in steady-state after at least 7 h are shown.

4.5. Separation of two laccases As shown in Table 1, eight different experiments were done in order to validate the proposed multi-stage model for ATPE. In the course of these experiments, feed stage (TP or BP) and sodium chloride mass fraction were varied and their influence on the partitioning of T.v. laccase and P.s. laccase and a mixture of both enzymes was investigated. Now, three of the eight experiments are described and their results discussed in detail.

4.5.1. Laccase separation with top phase feed In Fig. 9, the experimental and modeling results of a multi-stage extractive separation of laccase from P.s. and laccase from T.v. (EXP 1) is shown. The culture supernatant was fed into the extraction process together with the PEG stock solution on stage 1. In Fig. 9a, the mass fractions of PFC over all stages are shown. Here the black circle (䊉) in the miscibility gap on the tie lines indicates the calculated, global mixing point from the streams that enter the extraction process; whereas the empty squares ( ) indicate modeled PFC mass fractions exiting the stages as well as the calculated mixing points on the stages calculated. The tie lines are not overlaying each other perfectly, but the TLL increases from stages 1 to 3. This is the case, because the connecting line of top and bottom phase feed is not perfectly aligned with the slope of the corresponding tie line, which is a little steeper. Still, the change of mass fraction over thestages is rather small. Taking the PEG and

Fig. 9. Experimental and modeling results for the separation of P. sapidus and T. versicolor laccases fed in the top phase in a PEG3000-phosphate ATPS at pH 7 and 25 ◦ C without sodium chloride (EXP1): (a) mass fraction of PFCs on feed stage (䊉), stage 1 (mod - - ; exp — ), stage 2 (mod - - ; exp — ) and stage 3 (mod - - ; exp — ); ; exp ). Errors are indicated using error bars. (b) laccase activity in the feed () and top (mod - -; exp ) and bottom phase (mod

A. Prinz et al. / Process Biochemistry 49 (2014) 1020–1031

1029

Fig. 10. Experimental and modeling results for purification of P. sapidus and T. versicolor laccases fed in the bottom phase in a PEG3000-phosphate ATPS at pH 7 and 25 ◦ C without sodium chloride (EXP2): (a) mass fraction of PFCs on feed stage (䊉), stage 1 (mod - - ; exp — ), stage 2 (mod - - ; exp — ) and stage 3 (mod - - ; exp — ); ; exp ). Errors are indicated using error bars. (b) laccase activity in the feed () and top (mod - -; exp ) and bottom phase (mod

phosphate mass fractions of the experiment as input variables for the model, the exact same trend is observed proofing the correlation of the PFC. The model also calculates the increase in TLL from stage 1 to 3. On the right side of Fig. 9, the total laccase activity in top and bottom phase over the three stages is plotted. Since the CS enters the extraction process in the top phase, it is depicted on the left side of stage 1 (see Fig. 9b). The total enzyme activity in the bottom phase decreases strongly, while the decrease of total activity in the top phase is less steep. Only on stage 1, the activity in the bottom phase is higher than in the too phase, but on stages 2 and 3 it is the other way around. From previous investigations [18] it is known that in ATPS under these conditions without sodium chloride the laccases from P.s. and T.v. partition into opposite phases. Therefore, it is presumed that on stage 1 P.s. is mainly in the top phase and is passed on into stage 2. Laccase from T.v. partitions into the bottom phase and instantly leaves the extraction process on that stage in BP1. The experimental setup suggests a good separation of laccase from P.s. of laccase from T.v. in the top phase outlet stream (TP3), because the top phase containing mainly laccase from P.s. was washed thrice with bottom phase. This suggestion is confirmed looking at the results from IEF in Fig. 12 (TP-feed 0.0 g/g NaCl): the feed (d) shows activity of both laccases (b and c). Comparing TP3 (e) with BP1 (f), there is no T.v. activity, while BP1 still shows some P.s. activity. In the model, P.s. and T.v. activity are added and plotted as dashed line. Obviously, the modeled activity partitioning excellently corresponds to the experimental results. 4.5.2. Laccase separation with bottom phase feed In Fig. 10, the results of the second experiment and modeling (EXP 2) are shown. In contrast to the previous experiment, the feed changed from top to bottom phase, while the remaining parameters are the same (Fig. 10a). Differing from this experiment, the extension of the working tie line and the line connecting top and bottom feed are aligned better. Therefore, it is almost impossible to distinguish between the tie lines of the different stages. Here, the model gives the same results. Again, model and experiment show good agreement. In Fig. 10b, the total activities in the phases over all stages are shown. Since the CS is fed over the BP inlet, the initial activity is on the right side of stage 3. The total activity in the top phase decreases from stage 3 to 1, while the activity in the bottom phase decreases slightly from stage 3 to 2 and then increases again. This is probably the case due to volume reduction of the bottom phase in the course of the experiment. A total CS volume flow of 116 mL/h entered the extraction, while only 91 mL/h of bottom phase leaves the

extraction on stage 1. This mode of operation should enable a separation of laccase from T.v. from the mixture of T.v. and P.s. Looking at the second row (TP-feed 0.0 g/g NaCl) in Fig. 12 this assumption is confirmed. The bottom phase outlet BP1 (f) shows only activity of laccase from T.v. while in outlet of the top phase TP3 (e) still contains a mixture of both laccases. Very similar to previous example with top phase feed the modeling results agree with the experimentally measured values qualitatively as well as quantitatively. 4.5.3. Laccase purification with bottom phase feed In Fig. 11 the experimental and modeling results of a feed in the bottom phase and a sodium chloride mass fraction of 2.5 wt.% is (EXP 4) shown. Again, the feed was composed of a mixture of laccase from T.v. and P.s. Comparing the mass fractions of PFC in Fig. 9a with Figs. 10a and 11a, one observes the miscibility gap widening influence of sodium chloride in the experiment as well as in the model. In this experiment, the feed mass fraction of the PFC on ordinate and abscissa coincides with the extension of the tie line. Therefore, there is almost no change in PFC mass fraction over the three stages and data points for the experimental and modeling are overlaying each other. In terms of activity partitioning, the model displays the activity in the top phase on every stage in good agreement with experimental data. Regarding the laccase activities in the bottom phases, the trend is met qualitatively including the minimal activity on stage 2, but the model generally underestimates the activities in the phase by up to 20% on stage 1. The reason for this deviation can be explained with the addition of errors from phase modeling, the modeling of activity partitioning and the modeling of activity increasing effects. Nevertheless, the partitioning of both laccases into the bottom phase enables a partitioning of both enzymes in one phase increasing the total recovery. For the first two cases (EXP 1 & 2) without sodium chloride the activity recoveries in the bottom phase were 54% (modeled 68%) and 46% (model 54%), respectively. The addition of sodium chloride to the extraction process resulted in an increased activity recovery of 75% (model 64%) in the bottom phase. Although the actual modeled activity recoveries do not match perfectly with the experimental activity recoveries, the trends match acceptable. A possible explanation for this deviation was already discussed above. In spite of the deviation, the modeling is successful considering the single-stage data basis for the multi-stage modeling. In Fig. 12, the results for the IEF with ABTS staining for the previously discussed experiments are displayed. Every electrophoresis

1030

A. Prinz et al. / Process Biochemistry 49 (2014) 1020–1031

Fig. 11. Experimental and modeling results for purification of P. sapidus and T. versicolor laccases fed in the bottom phase in a PEG3000-phosphate ATPS at pH 7 and 25 ◦ C with 0.025 g/g of sodium chloride (EXP4): (a) mass fraction of PFCs on feed stage (䊉), stage 1 (mod - - ; exp — ), stage 2 (mod - - ; exp — ) and stage 3 (mod - - ; exp ; exp ). The arrow indicates the feed stage. Errors are indicated using — ); (b) laccase activity in the feed () and top (mod - -; exp ) and bottom phase (mod error bars.

Fig. 12. IEF results from multi-stage extraction experiments (EXP 1, 2, 4) with ABTS activity staining: (a) P.s. + T.v., (b) P.s., (c) T.v., (d) Feed, (e) TP3 and (f) BP1.

was conducted in the same way with the samples in the same cavities with the two exceptions. In EXP 2(d) no sample and in EXP 4(b) a mixture of P.s. + T.v. was pipetted into the cavities. The first served as an additional blank and the second to proof experimental reproducibility. In EXP 1, laccase from P.s. (TP3) was separated from T.v. by adding the feed with the top phase. The raffinate stream (BP1) contains mainly laccase from T.v. but also some laccase from P.s. By changing the feed from top to bottom phase (EXP 2), laccase from T.v. (BP1) was separated from P.s. while a mixture of mostly laccase from P.s. and less amounts of laccase from T.v. remain in the

raffinate (TP3). The addition of 2.5 wt.% of sodium chloride shifts the activity of P.s. from the top phase into the bottom phase (EXP 4). In this case, there is only little laccase activity in the top phase. Table 6 shows the deviations (absolute in U/L and relative in %) of modeling results from the eight experiments. In the first row, the initial enzyme activities and their experimental errors are shown. The largest deviation of experiment to the model in absolute values occurred in stage 1 of EXP 4 in the bottom phase where the modeled activity is 431 U/L (20%) below the experimentally determined activity. Since the feed was fed into the bottom phase and this error is on stage 1, it is possible that deviations of PFC partitioning coefficients and in addition analytical errors have led to this result. In some other cases (e.g. EXP 3 stage 3: 53% in TP and 104% in BP. EXP 5 stage 2: 58% in TP), the relative deviations are high, but in those cases, the total activity was very low leading to high relative deviations. Overall, it can be stated that there are deviations between the modeled and experimental results, but considering the simple data basis of single-stage experiments and the simplicity of the model, modeling result agree qualitatively very good with the experiment in all the cases and quantitatively in most of the cases. The average deviation of every experimentally measured activity is 98 U/L and 23%. Comparing these values with the average feed activity of 2193 U/L having and analytical error of 133 U/L the model is considered valid. A comparison of EXP 1 & 2 (Figs. 9 and 10) shows, which influence the feed stage has on the separation process. If the feed is placed on stage 1 (PEG inlet), laccase from P.s. is separated from

Table 6 Deviations of experimental and modeled extraction of P.s. and T.v. using ATPS. Stage

Phase

TP 1 BP TP 2 BP TP 3 BP

EXP Feed [U/L] Error [U/L] Dev. [U/L] Dev. [%] Dev. [U/L] Dev. [%]

1 2871 197 −57 −9% 90 9%

Dev. [U/L] Dev. [%] Dev. [U/L] Dev. [%]

−35 −7% −29 −10%

Dev. [U/L] Dev. [%] Dev. [U/L] Dev. [%]

−51 −10% 4 3%

2 3462 200

3 3032 182

4 2872 162

5 1356 83

6 1048 73

7 1750 101

8 1154 62

94 35% −200 −14%

68 50% −431 −20%

−22 −50% −261 −17%

41 43% −48 −24%

49 18% 14 1%

56 35% 211 25%

−120 −26% −24 −1%

86 50% 79 38%

74 26% −297 −16%

−27 −58% −258 −20%

75 40% 35 14%

−12 −5% 237 17%

69 42% 164 24%

−76 −6% −21 −1%

50 53% 78 104%

−11 −2% −270 −13%

−19 −47% −212 −17%

3 1% −24 −5%

−21 −8% 284 22%

44 19% 148 24%

−44 −12% 82 4%

A. Prinz et al. / Process Biochemistry 49 (2014) 1020–1031 Table 7 Comparison of experimental single and modeled multi-stage separation for mixtures of laccase from P. sapidus and from T. versicolor without sodium chloride. Feed Single-stage Multi-stage with top phase feed Multi-stage with bottom phase feed

2g 62 g h−1 124 g h−1

CFTP,P.s.

CFBP,T.v.

3.34 64.3 4.2

6.35 6.00 282.0

the enzyme mixture. In case the feed is on stage 3 (phosphate inlet), the situation changes. Now, laccase from T.v. is separated from the enzyme mixture. The addition of sodium chloride pushed laccase from P.s. into the bottom phase enabling an additional mode of operation. In case the goal of the separation task is to recover two similar enzymes in one phase the addition of sodium chloride maximizes the total recovery. Comparing now the experimental with the modeling results based on single-stage data, it is apparent that all these conclusions could already be drawn from the model. In Table 7, results from single- and multi-stage extraction for the separation of laccase of P. sapidus from laccase of T. versicolor are listed. The extraction was successfully scaled-up from 2 g in batch mode to a continuous feed of up to 124 g h−1 . In addition, this table shows that the multi-stage operation improves clearance factors by one order of magnitude. Depending, on which stage the feed enters the extraction process, either laccase from P. sapidus or laccase from T. versicolor is clarified better. With these results it was shown how a multi-stage extraction process was used to increase the clearance factors for either laccase from P. sapidus or T. versicolor in a PEG3000-phosphate ATPS at 25 ◦ C and pH 7. In addition, the extraction process was scaled-up from lab- to miniplant scale. Although the clearance factors for the multi-stage experiments are modeled, because it is not possible to differentiate between the two laccases quantitatively, the results reflect the experiment. This assumption is supported by the results from the IEF, because the change of clearance factors correlated with the staining results. For high clearance factors, only one band was visible (see Fig. 12). 5. Conclusions In this work an equilibrium stage model for enzyme purification using aqueous two-phase extraction was used for the phase system (PEG3000, phosphate and sodium chloride). In contrast to earlier process model [11], it was extended to varying slopes of the tie lines as well as multi-stage extraction and applied on enzyme separations including the activity in- and decreasing effects of the phase system on the enzymes. For the modeling of the equilibrium of PFC and enzymes different correlation equations were proposed and parameters fitted. After that, eight continuous experiments in a three stage mixer-settler unit in mini-plant scale were performed and used to validate the model successfully. Experiments and model results show that laccases from two different fungi can be separated using multi-stage ATPS and the mode of operation (feed position and sodium chloride content) enables either a selective separation of laccase from either P. sapidus or T. versicolor. In addition, it is shown that sodium chloride can be used to maximize the recovery of both enzymes in the bottom phase. This work shows that multi-stage aqueous two-phase extraction for the separation of enzymes can be modeled based on single-stage data and that the separation of the enzymes improves by multi-stage operation. For the future, the experimental basis should be extended to different aqueous two-phase systems and different target products. The modeling approach should be extended in such a way that the phase equilibrium is based on thermodynamic modeling instead of correlations. In addition, the thermodynamic modeling should

1031

be extended on the partitioning of proteins and enzymes. Furthermore, the mass transfer phenomena should be included in order apply the model to a process in an extraction column. Acknowledgements The authors gratefully acknowledge funding of the research leading to these results from the Ministry of Innovation, Science and Research of North Rhine-Westphalia in the frame of CLIBGraduate Cluster Industrial Biotechnology, contract no: 314-108 001 08. The applied mixer-settler unit was funded by the “Deutsche Forschungsgemeinschaft” and the Ministry for Innovation, Science, Research and Technology of the State of North Rhine-Westphalia (MIWF) (GZ: INST 212/2 73-1FUGG). References [1] Cramer SM, Holstein MA. Downstream bioprocessing: recent advances and future promise: open issue 1/1. Curr Opin Chem Eng 2011;1(1):27–37. [2] Albertsson P. Partition of cell particles and macromolecules. 3rd ed. New York: John Wiley & Sons; 1986. [3] Kroner KH, Hustedt H, Granda S, Kula MR. Technical aspects of separation using aqueous two-phase systems in enzyme isolation processes. Biotechnol Bioeng 1978;20(12):1967–88. [4] Kula MR. Extraction and purification of enzymes using aqueous two-phase systems. Appl Biochem Bioeng 1979;2:71–95. [5] Hustedt H, Kroner KH, Menge U, Kula M. Protein recovery using two-phase systems. Trends Biotechnol 1985;3(6):139–44. [6] Rosa PAJ, Azevedo AM, Sommerfeld S, Mutter M, Aires-Barros MR, Bäcker W. Application of aqueous two-phase systems to antibody purification: a multistage approach. J Biotechnol 2009;139(4):306–13. [7] Rosa PAJ, Azevedo AM, Sommerfeld S, Mutter M, Bäcker W, Aires-Barros MR. Continuous purification of antibodies from cell culture supernatant with aqueous two-phase systems: from concept to process. Biotechnol J 2013;8(3):352–62. [8] Rosa PAJ, Azevedo AM, Sommerfeld S, Bäcker W, Aires-Barros MR. Continuous aqueous two-phase extraction of human antibodies using a packed column. J Chromatogr B: Anal Technol Biomed Life Sci 2012;880(1):148–56. [9] Zaslavsky BY. Aqueous two-phase partitioning: physical chemistry and bioanalytical applications. New York, NY: Dekker; 1995. [10] Walter H, Fisher D, Brooks DE. Partitioning in aqueous two-phase systems: theory, methods, uses, and applications to biotechnology. Orlando: Academic Press; 1985. [11] Mistry SL, Asenjo JA, Zaror CA. Mathematical modelling and simulation of aqueous two-phase continuous protein extraction. Bioseparation 1992;3(6):343–58. [12] Mistry SL, Kaul A, Merchuk JC, Asenjo JA. Mathematical modelling and computer simulation of aqueous two-phase continuous protein extraction. J Chromatogr A 1996;741(2):151–63. [13] Ahmad MM, Hauan S, Przybycien TM. Flowsheet simulation of aqueous twophase extraction systems for protein purification. J Chem Technol Biotechnol 2010;85(12):1575–87. [14] Ahmad MM, Przybycien TM. Towards optimal aqueous two-phase extraction system flowsheets for protein purification. J Chem Technol Biotechnol 2013;88(1):62–71. [15] Samatou JA, Wentink AE, Alexandra J, Rosa P, Margarida Azevedo A, Raquel Aires Barros M, et al. Modeling of counter current monoclonal antibody extraction using aqueous two-phase systems. Comput Aided Chem Eng 2007;24:935–40. [16] Linke D, Bouws H, Peters T, Nimtz M, Berger RG, Zorn H. Laccases of Pleurotus sapidus: characterization and cloning. J Agric Food Chem 2005;53(24):9498–505. [17] Majcherczyk A, Johannes C, Hüttermann A. Oxidation of aromatic alcohols by laccase from Trametes versicolor mediated by the 2,2 -azino-bis-(3ethylbenzothiazoline-6-sulphonic acid) cation radical and dication. Appl Microbiol Biotechnol 1999;51(2):267–76. [18] Prinz A, Hönig J, Schüttmann I, Zorn H, Zeiner T. Separation and purification of laccases from two different fungi using aqueous two-phase extraction. Process Biochem 2014;49:335–46. [19] Smith PK, Krohn RI, Hermanson GT, Mallia AK, Gartner FH, Provenzano MD. Measurement of protein using bicinchoninic acid. Anal Biochem 1985;150(1):76–85. [20] Asenjo JA. Separation processes in biotechnology. In: Courtney MW, editor. Bioprocess technology. New York, NY: Dekker; 1990. [21] Cabezas H. Theory of phase formation in aqueous two-phase systems. J Chromatogr B: Biomed Sci Appl 1996;680(1):3–30. [22] Khoury FM. Multistage separation processes. 3rd ed. Boca Raton: CRC Press; 2005. [23] Othmer DF, Tobias PE. Liquid–liquid extraction data – toluene and acetaldehyde systems. Ind Eng Chem Res 1942;34(6):690–2.