Discontinuous and continuous purification of single-chain antibody fragments using immobilized metal ion affinity chromatography

Journal of Biotechnology 163 (2013) 233–242

Contents lists available at SciVerse ScienceDirect

Journal of Biotechnology journal homepage: www.elsevier.com/locate/jbiotec

Discontinuous and continuous purification of single-chain antibody fragments using immobilized metal ion affinity chromatography Carlos Andrés Martínez Cristancho a , Florian David b , Ezequiel Franco-Lara b , Andreas Seidel-Morgenstern a,c,∗ a

Max Planck Institute for Dynamics of Complex Technical Systems, Sandtorstr. 1, 39106 Magdeburg, Germany Institute of Biochemical Engineering, Technische Universität Braunschweig, Gaussstr. 17, 38106 Braunschweig, Germany c Institute of Process Engineering, Otto-von-Guericke University, P.O. Box 4120, 39106 Magdeburg, Germany b

a r t i c l e

i n f o

Article history: Received 15 March 2012 Received in revised form 28 August 2012 Accepted 30 August 2012 Available online 14 September 2012 Keywords: Single-chain Fragment variable antibody Immobilized metal ion affinity chromatography Simulated moving bed Histidine-tagged recombinant protein purification Bacterial expression Adsorption equilibrium constants

a b s t r a c t This work describes the adsorption–desorption behavior of a histidine-tagged single-chain Fragment variable antibody (D1.3 scFv) on a commercial immobilized metal ion affinity chromatography (IMAC) column. A clarified cell culture supernatant originating from Bacillus megaterium was characterized using single column experiments in a pH-gradient elution mode. The cell culture supernatant containing the antibody fragment D1.3 scFv could be treated in the chromatographic separation process as a pseudobinary mixture. Adsorption equilibrium constants of the antibody fragment fraction (ABF) and the nonspecifically retained protein impurity fraction (IMP) were determined experimentally at constant pH by reinjecting pulses of pooled fractions collected in preliminary batch gradient elution runs. Based on the estimated adsorption equilibrium constants a possible multicolumn open-loop three-zone two-step pHgradient simulated moving bed (SMB) process is suggested and designed, which possesses the potential to isolate continuously the antibody fragment fraction (ABF) containing the single-chain antibody fragment D1.3 scFv. © 2012 Elsevier B.V. All rights reserved.

1. Introduction In recent years, several therapeutic antibody fragments have been approved by the US Food and Drug Administration (FDA) and are already in the market (Scott and Clarke, 2009). Among the novel therapeutic recombinant antibody fragments under development, small-sized and engineered fragments such as the single-chain Fragment variable formats (scFv) offer new alternatives to the full-length monoclonal antibodies (Brereton et al., 2007). Singlechain antibody fragments are becoming essential tools for research, diagnostic and therapy due to their improved tissue perfusion or increased specificity for a defined antigen (Hust and Dübel, 2004; Weisser and Hall, 2009). Novel high-yield and low cost bacterial expression systems are being developed for the production of recombinant proteins like antibody fragments (David et al., 2010). In addition to that, gram

∗ Corresponding author at: Max Planck Institute for Dynamics of Complex Technical Systems, Sandtorstr. 1, 39106 Magdeburg, Germany. E-mail address: [email protected] (A. Seidel-Morgenstern). 0168-1656/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jbiotec.2012.08.022

positive bacteria like Bacillus megaterium are able to secret proteins in high amount into the culture supernatant making the cell disruption step unnecessary and increasing the target protein yield (Jordan et al., 2007; David et al., 2011). Expression technologies allow the fusion of polyhistidine tags to recombinant proteins, without affecting the structure or function of the tagged protein but facilitating its purification (Hochuli et al., 1988; Clemmitt and Chase, 2000). The optimization and success of the purification of therapeutic recombinant proteins depend strongly on the development and implementation of downstream processes consisting of a minimum number of steps. In particular, efficient continuous operation modes are attractive (Gottschalk, 2009; Carta and Jungbauer, 2010). In immobilized metal ion affinity chromatography (IMAC) the selected ligands attached to the solid stationary phase are divalent transition metal ions, usually copper, nickel, zinc or cobalt (Knecht et al., 2009). The corresponding technique was developed by Porath in the mid-1970s and exploits the ability of some exposed amino acid residues on the protein surface, such as the imidazole group of histidine, to form chelates with the immobilized metal ions (Porath et al., 1975). The elution of the adsorbed proteins is performed under gentle, non-denaturating conditions and by two different

234

C.A. Martínez Cristancho et al. / Journal of Biotechnology 163 (2013) 233–242

mechanisms: by lowering the pH (protonation of the imidazole group of histidine) or by the introduction of a competing agent (pure imidazole) (Porath, 1992). IMAC columns have high capacity, stability, and reproducibility at similar costs as ion exchange columns (Chaga, 2001). A possible drawback of IMAC columns is the leakage of metal ions during elution. For the purification of histidine-tagged single-chain antibody fragments using IMAC currently discontinuous single column processes are dominant (Skerra et al., 1991; Essen and Skerra, 1993). However, there is an increasing interest in using more productive multicolumn configurations, in which biological multi-component mixtures are fed uninterruptedly and the target biomolecules are withdrawn continuously at two or more characteristic outlet ports. Various alternative configurations have been suggested and were successfully applied for protein purification, such as the multicolumn countercurrent solvent gradient purification (MCSGP) process (Müller-Späth et al., 2008), size exclusion simulated moving bed (SE-SMB) chromatography (Ottens et al., 2006; Buhlert et al., 2009), and the open-loop simulated moving bed (SMB) chromatography (Kessler et al., 2007; Gueorguieva et al., 2011) as an adaptation of the classical SMB principle (Broughton and Gerhold, 1961). In contrast to SMB, the MCSGP process offers the possibility to isolate more than two fractions. It is an attractive concept to realize purifications requiring solvent gradients exploiting an elegant but complex combination between continuous steps connections of all columns and batch-wise elution steps (Krättli et al., 2011). The focus of this work was the purification of histidine-tagged single-chain antibody fragments using immobilized metal ion affinity chromatography (IMAC). For this reason, gradient elution chromatography was studied using stepwise pH-gradients. The antibody fragment fraction (ABF) was the most retained component in the cell culture supernatant due to the presence of a histidine affinity tail. The clarified cell culture supernatant can be considered as a pseudo-binary mixture by lumping of all undesirable proteins eluting before the target fraction as impurities. It is rather clear that the mobile phase pH will dominate the elution of the antibody fragment fraction (ABF) and the closest eluting protein impurity fraction (IMP). The influence of the mobile phase pH on the adsorption equilibrium constants for both components (ABF and IMP) was determined in this work carrying out single batch pulse experiments under isocratic pH elution conditions. Based on the dependence of the adsorption equilibrium constants on the mobile phase pH, an open-loop three-zone two-step pH-gradient simulated moving bed (SMB) process will be suggested for the continuous chromatographic purification of the antibody fragment fraction (ABF). The process was investigated theoretically through the simulation of an equivalent true moving bed (TMB) process using an equilibrium stage model. The theoretical study of the continuous multicolumn countercurrent process delivered a region of applicable operating parameters. Additionally, concentration profiles were simulated within a cascade of three columns for selected operating points. Finally, the productivity, the specific buffer consumption and final product concentration of collecting the antibody fragment fraction (ABF) continuously at the extract port of the theoretical SMB unit is evaluated and compared with the corresponding performance of a standard experimental batch process.

The B. megaterium strain YYBm1 carrying the plasmid pEJBMD1.3scFv was used for the bacterial expression (Yang et al., 2007). The preculture were carried out with fructose concentration of 5 g/L in shaking flasks. Lab scale fermentations were carried out in a 5-L Bioreactor using an adapted oscillating fed-batch strategy. It is worth noting that B. megaterium allows the production of functional single-chain antibody fragments by direct secretion of proteins into the culture medium. More details regarding the production of the single-chain antibody fragment D1.3 scFv are given in the literature (David et al., 2011). The single-chain antibody fragment D1.3 scFv, a 27 kDa recombinant protein, has a concentration around 0.025 mg/ml in the supernatant. 2.2. Preparation of samples prior to chromatography The non-clarified cell culture supernatant coming from the bioreactor were pretreated and clarified prior to chromatography. First, the supernatant was centrifuged for 30 min at 4 ◦ C and 4000 rpm, in order to remove the biomass present in the harvested cell culture fluid (Thermo Scientific Heraeus® Biofuge Stratos Centrifuge, Heraeus Holding GmbH, Hanau, Germany). Next the supernatant was separated carefully from the precipitate. The particulate matter still present was removed by filtration (Pall® Acrodisc® 32 mm syringe filter with 0.2 ␮m Supor® Membrane, Pall Life Sciences, Ann Arbor, MI, USA). The clarified cell culture supernatant was aliquoted, stored frozen at –30 ◦ C, in order to preserve completely the protein structure and activity. Aliquoted samples were thawed before the chromatographic runs. 2.3. Selected stationary phase Prepacked and ready-to-use chromatographic commercial columns were used (HisTrapTM FF crude, GE Healthcare BioSciences AB, Uppsala, Sweden). Column volumes were 1 ml (2.5 cm × 0.7 cm i.d.). The columns were precharged with Ni SepharoseTM 6 Fast Flow (highly cross-linked agarose beads with nickel as metal ion). Average particle size was 90 ␮m. After use the columns were always washed, first with water and then with a solution of 20% ethanol. The columns were stored at 4 ◦ C. 2.4. Loading and elution buffers In order to generate a descending stepwise pH-gradient in the IMAC columns, loading and elution buffers are needed. The loading and elution buffers were prepared by dissolving sodium dihydrogen phosphate monohydrate (NaH2 PO4 ·H2 O, Merck KGaA, Darmstadt, Germany), disodium monohydrogen phosphate anhydrous (Na2 HPO4 , Merck KGaA, Darmstadt, Germany) and sodium chloride (NaCl, Sigma–Aldrich Chemie GmbH, Steinheim, Germany) in water. The pH values were adjusted either to 6.8 (loading buffer) or 4.2 (elution buffer) with either concentrated solutions of sodium hydroxide or hydrochloric acid. Buffers were filtered prior to the chromatographic runs (cellulose nitrate filters 0.20 mm, Sartorius Stedim Biotech, Göttingen, Germany). The water used in all experiments was purified in a Milli-Q® ultrapure water purification system (Milli-Q® gradient, Millipore SAS, Molsheim, France).

2. Materials and methods 2.5. Instruments 2.1. Bacterial expression of an antibody fragment The lysozyme specific histidine-tagged single-chain Fragment variable antibody format (D1.3 scFv) was expressed in B. megaterium (David et al., 2011).

Batch chromatographic experiments were carried out in an ÄKTA Purifier controlled by Unicorn software (GE Healthcare BioSciences AB, Uppsala, Sweden). The pH-gradient was monitored online at the column outlet using an 88 ␮l flow cell equipped with

C.A. Martínez Cristancho et al. / Journal of Biotechnology 163 (2013) 233–242

235

a flow-through pH electrode (GE Healthcare Bio-Sciences AB, Uppsala, Sweden). Offline pH measurements of buffers were done using a Mettler Toledo pH-meter (SevenEasy pH, Mettler Toledo GmbH, Schwerzenbach, Switzerland). 2.6. Analysis by gel electrophoresis Prior to electrophoresis the samples were first ultrafiltrated (Amicon® Ultra-4 Centrifugal filters with Ultracel® -10, Millipore Ireland Ltd, Cork, Ireland) for 30 min at 4 ◦ C and 4000 rpm (Thermo Scientific Heraeus® Biofuge Stratos Centrifuge, Heraeus Holding GmbH, Hanau, Germany). Following this, the samples were mixed with the loading buffer, heated up to 99 ◦ C and shaked at 750 rpm for 10 min (Thermomixer comfort, Eppendorf AG, Hamburg, Germany). The sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS–PAGE) was done using a 12% polyacrylamide gel according to the method of Laemmli (Laemmli, 1970). Samples were compared to the standard protein marker (PageRulerTM Unstained Protein Ladder, Fermentas GmbH, St. Leon-Rot, Germany). Clarified cell culture supernatant and purified samples collected were run in the SDS-PAGE gels at 50 mA for 1 h (Biometra® Power Pack P25 T, Biomedizinische Analytik GmbH, Göttingen, Germany) and stained with Coomassie blue (Coomassie® Brilliant Blue R-250, AppliChem GmbH, Darmstadt, Germany).

Fig. 1. Chromatographic elution profile of diluted clarified cell culture supernatant containing the single-chain antibody fragment using a stepwise pH-gradient. The recorded pH at the column outlet is depicted as a dashed line. Fractions collected are indicated between dotted lines (IMP) and continuous lines (ABF).

collected fractions of IMP and ABF were reinjected afterwards into the column at constant pH conditions in order to evaluate possible isotherm non-linearities.

2.7. Column characterization

3. Results

The total column porosity was estimated experimentally using small pulses (50 ␮l) of unretained protein (collected in preliminary runs) under isocratic conditions at pH 4.2. Both hold-up times of the column and the apparatus were determined in this manner.

3.1. Chromatographic elution behavior

2.8. Single-column batch chromatography using IMAC The purification of the antibody fragment fraction by immobilized metal ion affinity chromatography was carried out using stepwise downward pH-gradients using a non-retained buffer system. An IMAC column was equilibrated using 10 column volumes of a high pH mobile phase (pH = 6.8) which promotes the binding of the histidine-tagged single-chain antibody. After the equilibration of the column with the loading buffer, 10 ml of diluted clarified cell culture supernatant were injected using a 50 ml SuperloopTM (GE Healthcare Bio-Sciences AB, Uppsala, Sweden). For single-column batch experiments, the feed mixture containing the single-chain antibody fragment was diluted to 1:1 in the loading buffer. Therefore, the concentrations of the single-chain antibody fragment for the batch experiments were 0.0125 mg/ml, just half of the one given in section 2.1 for the clarified cell culture supernatant. Following this, 25 column volumes were used to wash away the non-binding proteins. The stepwise pH-gradient was realized by a sudden change from 0% to 100% of the elution buffer (pH = 4.2) maintaining 30 column volumes to elute the target protein. The flow rate in all batch runs was kept constant at 1.60 ml/min. UV-signals for the chromatograms were measured at 214 nm. Numerous fractions were collected during sample load, wash out and elution over the course of the chromatographic runs for subsequent SDS-PAGE analysis. 2.9. Determination of the adsorption equilibrium constants as a function of the pH value The experimental strategy to estimate adsorption equilibrium constants for the fractions of interest was based on the evaluation of retention times of pulse injections of certain fractions collected in preliminary batch runs (see Section 3.1). Different volumes of

A characteristic single-column batch chromatogram is shown in Fig. 1. The proteins in the clarified cell culture supernatant migrate differently as they pass through the IMAC-column. Basically, there are three characteristic fractions displaying different degrees of affinity to the stationary phase. The first peak eluting during the sample loading and washout periods represents the large amount of unretained proteins (UNR) passing rapidly through the column due to the lack of exposed histidine clusters or residues. As the pH decreases in the beginning of the stepwise gradient period, some of the retained proteins eluted faster. These non-specifically bound proteins can be lumped together forming a closest eluting protein impurity fraction (IMP). Finally, as the pH value is further decreased the antibody fragment fraction (ABF) elutes as the most retained protein fraction. In Fig. 1, the fraction between the two dotted lines corresponds to the protein impurity (IMP) and the fraction between the two continuous lines corresponds to the target antibody fragment fraction (ABF). The protein impurity fraction (IMP) eluted approximately at pH 5.90 and the antibody fragment fraction (ABF) approximately at pH 4.60. The purification of the antibody fragment fraction (ABF) can be considered as a pseudo-binary separation problem. The pseudobinary feed mixture consists of the unretained proteins (UNR) and the protein impurity (IMP) lumped together forming the faster eluting fraction and the antibody fragment fraction (ABF) as the slowest eluting fraction. It should be pointed out that there are two different views characterizing the feed mixture. At first the mixture can be considered as pseudo-ternary due to the number of main components present in the feed mixture. The second view regards the mixture as pseudo-binary related to the number of fractions achievable in the continuous SMB separation process (i.e. raffinate and extract). It can be seen in Fig. 1 that theoretically ideal planned squaredshaped pH step gradients are in reality distorted and delayed due to dispersion and dead volumes between the injection valve and the pH-meter located at the column outlet.

236

C.A. Martínez Cristancho et al. / Journal of Biotechnology 163 (2013) 233–242 Table 1 Estimated adsorption equilibrium constants (Eq. (2)) for ABF, IMP and CRI. pH

CHPO

5.2 5.8 6.0

1.0 3.0 5.0

4

2−

[mM]

HIMP

HABF

HCRI

7.65 9.58 13.28

20.80 36.61 52.24

11.24 22.21 30.18

stationary phase, qi . For diluted protein solutions, as studied here, linear isotherms characterized by single adsorption equilibrium constants are typically applicable. This approach was supported by the observation that in the concentration range covered retention times and peak shapes were not affected significantly by the injection volumes. The initial slopes of the isotherms, i.e. the adsorption equilibrium constants Hi , were calculated based on the measured retention times (tR,i ) according to the following expression: 1−ε · ε

Hi = Fig. 2. SDS-PAGE analysis of the discontinuous elution profile corresponding to Fig. 1 and showing the fractions of the proteins of interest as indicated: lane (IMP) protein impurity fraction; lane (M) protein marker with molecular masses given in kilodaltons (kDa); lane (ABF) antibody fragment fraction.

The results of SDS-PAGE analysis given in Fig. 2 reveal that the IMP fraction contains different proteins with larger and lower molecular weights compared to the target product D1.3 scFv. In contrast, the ABF fraction is characterized essentially by the 27 kDa band corresponding to the target single-chain antibody fragment and two bands around 15 kDa. These bands correspond to the variable domains of the antibody fragment, the so-called VH and VL domains. The presence of the VH and VL domains in the ABF fraction could be explained due to the disruption of the peptide linker that join the variable domains of the antibody fragment during the purification or even before during the bacterial fermentation. These VH and VL bands can be found both in the ABF and IMP fractions. The variable domains eluted during the whole elution step. It should be emphasized that, despite the presence of the variable domains, the ABF fraction accessible through this IMAC-technique is a very valuable product. Thus, it is an attractive target fraction for either discontinuous or continuous chromatographic separation. 3.2. Column porosity As a result of preliminary quantitative evaluation, the holdup time of the apparatus (t0,a ) was found to be equal to 0.19 min using a volumetric flow rate of 1 ml/min. The total hold-up time of the column and the apparatus (t0 ) was 0.54 min. The difference between both values describes the hold-up time of just the column (t0,c = t0 − t0,a ). The following equation was used to estimate the total porosity of the column or the total void volume of the column (ε) using the volumetric flow rate (V˙ ) and the column volume (Vc ): ε=

V˙ · to,c Vc

(1)

The corresponding total porosity of the column is ε = 0.35. 3.3. Effect of the pH on the adsorption equilibrium constants Knowledge regarding the adsorption equilibrium is most instructive when designing chromatographic processes (Guiochon et al., 2006). An adsorption isotherm relates, at a certain constant temperature, the concentration of a component (i) in solution, Ci , to the concentration of the same component adsorbed on the

t

R,i

to,c



−1

i = IMP, ABF

(2)

The retention times were estimated using the first central moments of measured peaks (1,i ) evaluating the discrete UVsignals recorded during isocratic elution (Guiochon et al., 2006). The specific pH values are related to the concentration of the monohydrogen phosphate ion CHPO 2− and its conjugated acid, the dihydrogen phosphate ion CH

4

2 PO4

1−

(Deutscher, 1990). The

adsorption equilibrium constants depend strongly upon the monohydrogen phosphate ion concentrations, i.e. the pH. qi = Hi (pH(CHPO

4

2−

)) · Ci

i = IMP, ABF

(3)

The pH range for the isocratic experiments was chosen between the loading (pH = 6.8, CHPO 2− = 7.0 mM) and elution (pH = 4.2, CHPO

4

4

2−

= 0.1 mM) buffer conditions. Selected elution profiles are

depicted at three distinct pH values in Fig. 3. At low monohydrogen phosphate ion concentrations or non-adsorbing conditions (pH = 5.2), both fractions ABF and IMP elute relatively fast and are less separated from each other (Fig. 3(a)). As the monohydrogen phosphate ion concentration increases (pH = 5.8 and 6.0), the binding of the proteins becomes stronger resulting in larger retention times and broader but more separated peaks (Fig. 3(b) and (c)). The peaks corresponding to the protein impurity fraction (IMP) are sharper and elute faster than those of the antibody fragment fraction (ABF). Thus, the monohydrogen phosphate ion concentration (i.e. the pH) has a strong influence on the retention times and peak shapes of ABF and IMP fractions. Two characteristic adsorption equilibrium constants HIMP and HABF were estimated using Eq. (2). They provide a good initial basis to design a continuous binary separation process bringing the corresponding fractions to the raffinate port (UNR and IMP) and the extract port (ABF) of a potential SMB process. Nevertheless, it is advisable to design a more conservative SMB process considering a closer hypothetical border in the isocratic elution profiles instead of HIMP . A suitable, arbitrary selected critical retention time (tR,CRI ) indicated by the arrows in Fig. 3 marks the position of a critical pseudo-impurity (CRI) allowing to estimate a corresponding hypothetical adsorption equilibrium constant HCRI . This adsorption equilibrium constant marks a border allowing to design in a more reliable way the modified pseudo-binary SMB separation of faster eluting fractions (UNR up to CRI) and the antibody fragment fraction (ABF). A design for a continuous separation based on HCRI (instead of HIMP ) should guarantee the capture of all the protein impurities which possess a retention time before the imaginary border (tR,CRI ) in the raffinate port in the SMB unit. All three sets of estimated adsorption equilibrium constants are summarized in Table 1.

C.A. Martínez Cristancho et al. / Journal of Biotechnology 163 (2013) 233–242

HCRI = 7.19 + 4.74 · CHPO

4

2−

237

(6)

Fig. 4 illustrates these linear dependences. Comparing the slopes it can be seen that the effect of the monohydrogen phosphate ion concentration on the adsorption equilibrium constant of the antibody fragment fraction (ABF) is larger than on the other two adsorption equilibrium constants (Fig. 4). Thus, the antibody fragment fraction (ABF) is more sensitive to changes in the monohydrogen phosphate ion concentration (pH). The linear relationships given in Eqs. (4)–(6) are the key thermodynamic information regarding the adsorption–desorption behavior of the fractions of interest. This information will be used below to design an efficient multicolumn countercurrent continuous purification process. 3.4. Performance parameters: batch experiments To compare below experimentally determined batch elution results with theoretically predicted continuous multicolumn chromatographic SMB separation several performance parameters like productivity, specific buffer consumption and final concentration of the target product were evaluated. The reference productivity of the batch process was estimated using the following equation: Pr Batch =

mABF CABF · VFraction = tcycle · Vc · (1 − ε) tcycle · Vc · (1 − ε)

(7)

For the concentration of ABF, CABF = 0.0075 mg/ml in the fractionated volume VFraction = 4.5 ml, the total cycle time for a batch experiment is tcycle = 50 min, the column volume, Vc = 1 ml and the porosity given above a productivity of PrBatch = 1.49 mgABF /day/mlstat phase is obtained. A specific buffer consumption of 2.37 l/mgABF was determined as total buffer consumption (loading buffer plus elution buffer) during the different steps of the batchwise gradient elution mode (sample injection, wash-out, gradient and regeneration). For the batch process, a product concentration of ABF, CABF = 0.0075 mg/ml, was estimated using the BCA assay. 4. Continuous multicolumn chromatographic separation process

Fig. 3. Comparison of isocratic elution profiles for pooled fractions of IMP and ABF at different pH values. Characteristic retention times are indicated with arrows (IMP and ABF). A retention time for a critical pseudo-impurity (CRI) was estimated in order to assure that the ABF fraction is not contaminated by the earlier eluting impurities. (a) pH = 5.2, CHPO 2− = 1 mM. (b) pH = 5.8, CHPO 2− = 3 mM. (c) pH = 6.0, CHPO 2− = 4 4 4 5 mM.

It was found that these adsorption equilibrium constants can be described by the following linear functions of the monohydrogen phosphate ion concentration CHPO 2− in the pH range investigated: 4

HABF = 13.16 + 7.86 · CHPO

4

HIMP = 5.49 + 1.56 · CHPO

4

2−

2−

(4) (5)

Continuous multicolumn configurations are attractive in order to achieve a more productive and efficient configuration than the discontinuous gradient elution process. The most exploited continuous chromatographic separation mode is the classical isocratic simulated moving bed (SMB) process. The standard SMB configuration exploits several chromatographic columns in four zones arranged in a closed mobile phase loop. There are two inlet (feed and desorbent) and two outlet streams (extract and raffinate). This four-zone arrangement gives each zone a very specific task. The separation of the pseudo-binary mixture into two fractions takes place in zones II and III. Regeneration of the stationary and mobile phases is carried out in zones I and IV, respectively (Broughton and Gerhold, 1961). The use of continuous multicolumn countercurrent gradient chromatography has offered attractive opportunities in the field of protein separations due to the possibility of tuning the solvent strength within the simulated moving bed process. A rather simple SMB configuration could be made consisting of just three zones eliminating zone IV. This open-loop three-zone process provides an attractive option when cheap aqueous buffers are used as mobile phase and solvent recycling is not mandatory (Palani et al., 2011). In this work, an open-loop three-zone two-step pH-gradient SMB unit is suggested for the separation problem described and theoretically designed, based on the concrete thermodynamic data.

238

C.A. Martínez Cristancho et al. / Journal of Biotechnology 163 (2013) 233–242

Fig. 4. Linear dependence of the characteristics adsorption equilibrium constants upon the monohydrogen phosphate ion concentration (CHPO

4

The configuration exploits two different fluid phases with two distinct monohydrogen phosphate ion concentrations (or pH values) at the two inlet ports. This establishes two different pH conditions in zones I and II, and in zone III as depicted in Fig. 5(a) and (b). The purification of the ABF fraction from fractions UNR and IMP is possible by tuning the adsorption–desorption behavior within the SMB unit imposing non-adsorbing conditions in zones I and II, and adsorbing conditions in zone III. The unretained proteins fraction (UNR) is the fastest eluting component and will travel together with the protein impurity fraction (IMP) toward the raffinate port within the SMB unit (Figs. 5(a) and (b)). The goals of the calculations carried out are twofold. At first, a simple and effective strategy is described to design the proposed SMB process, in particular to estimate possible ranges for the operating parameters. Another goal is to evaluate theoretically reachable SMB process productivities and to compare them with the performance of batch chromatography. The basis for the SMB design applied is the specification of dimensionless flow-rate ratios for each zone. These ratios are defined by the varying volumetric fluid phase flow rates V˙ K and the volumetric solid phase flow rate V˙ s for each zone k (Storti et al., 1993): mk =

V˙ K V˙ s

k = I, II, III

(8)

Exploiting the analogy to a true moving bed process for the open-loop three-zone two-step pH-gradient configuration considered in Fig. 5(a), the following conditions on the flow-rate ratios are required to achieve complete separation of the two components present in a pseudo-binary mixture under linear conditions (Mazzotti et al., 1997; Abel et al., 2002). Hereby, for the open-loop three-zone two-step pH-gradient process considered, the lower and upper boundaries of the dimensionless flow rate ratios mk depend upon monohydrogen phosphate ion concentration of the mobile phase CHPO 2− k in the following manner: 4

HIMP/CRI III (CHPO

4

2−

HIMP/CRI I–II (CHPO

4

III

2−

) < mIII < HABF III (CHPO

4

I–II

2−

III

)

(9)

HABF I–II (CHPO

4

4

2−

I–II

)

(10)

I–II

) < mI

(11)

As depicted in Fig. 5(b), it holds CHPO CHPO

4

2−

Raffinate

= CHPO

4

2−

III

4

< CHPO

4

2−

2−

Feed

Desorbent

= CHPO

4

2−

I–II

<

.

A simple equilibrium stage model capable to describe the hypothetical true moving bed (TMB) concept (Storti et al., 1993) was used in this paper to theoretically investigate the open-loop threezone two-step pH-gradient configuration. The identification of regions of operating parameters allowing complete separation of the pseudo-binary mixture in its components (Fig. 5(a)) was based on the equilibrium stage model described by Beltscheva et al. (2003). The internal concentration profiles of the antibody fragment fraction (ABF), the hypothetical critical impurity (CRI) and the protein impurity fraction (IMP) were predicted using this model. The model took ideal pH profiles into account. Deviations from the ideal stepwise gradient shown in Fig. 5(b) were neglected. Finetuning will probably be needed in experimental validation of the discussed continuous multicolumn configuration. For the sake of simplicity, the steady state of the TMB was calculated for identical number of equilibrium stages (N) in each zone. As an estimate N was set to be 20. This number of stages was chosen particularly low in order to provide a conservative SMB design. To fully specify operating points of the TMB unit, the flow rate ratio mI was calculated introducing a safety factor of 1.05 and the feed flow rate was set in agreement with possibilities of an available lab-scale unit to be 0.64 ml/min. The extract and raffinate purity requirements were defined in the calculations to be both above 99.9%. The simulations were carried out using the same buffer compositions for the inlet ports (feed and desorbent port) as in the batch gradient elution experiments. That means, the buffer composition of the feed inlet stream was set equal to the loading buffer and the buffer composition of the desorbent inlet stream was set equal to the elution buffer. Complete separation regions were determined by screening all values possible for the monohydrogen phosphate ion concentrations in zone III (CHPO 2− III ), i.e. between CHPO 2− Feed and CHPO

4

) < mII < HABF I–II (CHPO

2−

2− ).

2−

Desorbent

4

. Each CHPO

4

2−

III

4

value generated corresponding

adsorption equilibrium constants for the components in zone III. In addition, every single scanned (mII ,mIII )-pair must fulfill the

C.A. Martínez Cristancho et al. / Journal of Biotechnology 163 (2013) 233–242

239

Fig. 5. Open-loop three-zone two-step pH-gradient SMB process proposed for the purification of recombinant antibody fragments using immobilized metal ion affinity chromatography. (a) Schema of the open-loop three-zone two-step pH-gradient SMB process. (b) pH- and monohydrogen phosphate ion concentration profiles within the SMB.

constraints in Eqs. (9) and (10). The (mII ,mIII )-pairs fulfilling the boundary constraints specify a system of linear of equations representing mass balances of each component. The concentrations of each component at the beginning and end of each zone of the TMB unit can be found solving the linear system. Only the (mII ,mIII )-pairs fulfilling the constraints were used to depict the complete separation region by simulation. Details are given in Beltscheva et al. (2003). The shape of the complete separation region in the (mII ,mIII )plane depends on the external salt concentrations CHPO 2− Desorbent and CHPO

4

2−

Feed

4

and on the retention behavior of the components

described by the adsorption equilibrium constants. Systematic simulations were done using two sets of these constants: (A) the adsorption equilibrium constants of fractions ABF and IMP and (B) the adsorption equilibrium constants of fraction ABF and the hypothetical critical component CRI. When comparing the complete separation regions in Fig. 6(a), it can be observed that the size region belonging to set A is larger than the one for set B and comprises the later one. This is due to

the smaller difference between the values of the adsorption equilibrium constants of ABF and CRI, in comparison to the difference between the values of the adsorption equilibrium constants of ABF and IMP. The smaller region describes a more conservative design space for the continuous process, justifying the assumption of a hypothetical critical impurity. After identifying the complete separation regions, two specific operating points indicated as I and II in Fig. 6(a) were selected for closer inspection. Hereby point II is located within both regions, whereas point I is in region A but not in B. For the sake of illustration, the concentrations of ABF, IMP, and CRI in the feed mixture for the simulation studies were assumed to be all equal to 0.0125 mg/ml (CFeed,ABF = CFeed,IMP = CFeed,CRI = 0.0125 mg/ml). This probably overestimates the competition between CRI and ABF but provides, thus, a valuable additional safety margin and a lower bound for the process productivities. The internal concentration profiles predicted with the equilibrium stage model corresponding to the two different operating points I and II are summarized in Table 2 and shown in Figs. 6(b) and (c). The feed concentrations of ABF, IMP and CRI are depicted as

240

C.A. Martínez Cristancho et al. / Journal of Biotechnology 163 (2013) 233–242

Table 2 Results of the simulations for two operating points obtained with the equilibrium stage model of an open-loop three-zone two-step pH-gradient SMB process as illustrated in Fig. 6. Parameters used for the simulations: V˙ Feed = 0.64 ml/min, CFeed = 0.0125 mg/ml, CHPO 2− Feed = 7.0 mM, CHPO 2− Desorbent = 0.1 mM and mI = 14.6. 4

I II

4

mII

mIII

V˙ S [ml/min]

V˙ Desorbent [ml/min]

V˙ Raffinate [ml/min]

V˙ Extract [ml/min]

˙ ABF m [mg/min]

CABF Extract [mg/ml]

PuABF(IMP) Extract [%]

PuABF(CRI) Extract [%]

9.0 12.5

35.0 30.0

0.025 0.037

0.360 0.535

0.862 1.097

0.139 0.078

0.007 0.005

0.05 0.07

99.8 99.9

84.0 99.2

dashed horizontal lines. The position of the inlet and outlet streams with the enriched components are indicated and marked on the xaxis. The internal concentration profiles of ABF, CRI, and IMP given in Figs. 6(b) and (c) differ strongly comparing operating points I and II. It is worth realizing, that the antibody fragment fraction (ABF) traveling toward the extract port in Fig. 6(b) is contaminated for operating point I by the critical impurity (CRI). This is in agreement with the theory because this operating point is outside the complete separation region for ABF and CRI. In contrast, the separation between ABF and IMP is predicted to be successful for both operating points I and II. This is the case also for the separation between ABF and CRI for the operating point II. There are also differences between the two operating points regarding the enrichment of the antibody fragment fraction, which can be defined as the ABF concentration achieved at the extract port over the ABF concentration in the feed. The mass flow rate of the antibody fragment fraction (ABF) leaving the unit corresponding to the operating point I is larger than the one for operating point II. But less pure antibody fragment fraction is obtained with operating point I, because this point is outside the complete separation region of ABF-CRI as discussed above. The essential parameters and results of the SMB simulations carried out are summarized in Table 2. The productivity of the continuous process was estimated using the following equation: PrSMB =

˙ ABF m CABF Extract · V˙ Extract = Nc · Vc · (1 − ε) Nc · Vc · (1 − ε)

(12)

where CABF Extract is the final concentration of the ABF fraction at the extract port. V˙ Extract is the volumetric flow rate at the extract port. Nc is the number of columns used in the SMB open-loop configuration. Again Vc and ε are the column volume and total porosity of the column, respectively. The predicted productivities of operating points I and II were PrSMB = 5.17 and 3.69 mgABF /day/mlstat phase , respectively, thus exceeding significantly the value of the discontinuous chromatographic process (PrBatch = 1.49 mgABF /day/mlstat phase ). Predicted specific buffer consumptions of operating points I and II were 0.10 and 0.17 l/mgABF , respectively. The specific buffer consumptions of the continuous process are significantly smaller than for the batch process. The predicted final concentration of the ABF fraction in the extract of the continuous process is several times higher compared to the batch process and exceeds the feed concentration. A comparison between the determined performance parameters of the theoretically studied SMB-process and the experimentally investigated discontinuous process is summarized in Table 3. Table 3 Performance parameters comparison between the experimentally investigated batch IMAC-process and the theoretically studied continuous SMB IMAC-process (CFeed = 0.0125 mg/ml). Fig. 6. Simulation results for the open-loop three-zone two-step pH-gradient SMB process illustrated in Fig. 5. CFeed,i = 0.0125 mg/ml; i = ABF, IMP and CRI. (a) Two different separation regions in the mII ,mIII -plane according to the conditions in Fig. 5 for the separation of ABF-IMP (larger region, delimited by empty circles) and of ABF-CRI (smaller region contained within the larger region, delimited by full circles) and two operating points I and II. (b) Predicted internal concentration profile for ABF, IMP and CRI according to operating point I. (c) Predicted internal concentration profile for ABF, IMP and CRI according to operating point II.

Batch, experiment SMB, theoretical Operating point I SMB, theoretical Operating point II

Productivity [mgABF /day/mlstat phase ]

Specific buffer consumption [l/mgABF ]

Product Concentration ABF [mg/ml]

1.49 5.17

2.37 0.10

0.0075 0.050

3.69

0.17

0.070

C.A. Martínez Cristancho et al. / Journal of Biotechnology 163 (2013) 233–242

5. Conclusions In this work the use of immobilized metal ion affinity (IMAC) in discontinuous and continuous gradient elution chromatography was studied experimentally and theoretically as an option to isolate high-value recombinant therapeutic proteins like antibody fragments from complex biological mixtures. The continuous separation approach described can be extended to purify other tagged or even not tagged recombinant proteins being either the slowest or fastest in the elution train. The feed mixture was considered as pseudo-binary with the target being in the slowest eluting fraction. This allows applying a continuous purification process using the proposed open-loop three-zone two-step pH-gradient SMB, which produces only two outlet streams. At the raffinate port, UNR and IMP will be recovered together as shown in Fig. 5. The valuable purified target product, the ABF fraction, can be collected at the extract port. Thermodynamic adsorption data of a specific antibody fragment fraction (ABF) and a protein impurity fraction (IMP) were determined from measured retention times at various isocratic pH elution conditions. Adsorption equilibrium constants were fitted by linear relationships to the monohydrogen phosphate ion concentration in the buffer used. In batch operation the antibody fragment could be successfully isolated under pH-gradient conditions. The experimental results of the discontinuous single-step purification by immobilized metal ion affinity chromatography were used as a basis for the design of a more productive continuous multicolumn countercurrent chromatographic purification process based on the simulated moving bed (SMB) principle. An open-loop three-zone two-step pH-gradient configuration was suggested and investigated theoretically using an equilibrium stage true moving bed model. A systematic simulation study was carried out to select possible operating conditions exploiting the influence of the monohydrogen phosphate ion concentration (or mobile phase pH) on the adsorption equilibrium constants. The model predicted for ideal pH-gradients regions of operating parameters (flow rates) providing complete separation and generated the corresponding internal concentration profiles of the antibody fragment fraction (ABF), the protein impurity fraction (IMP) and a critical pseudo-impurity (CRI) along the unit. The selection of a characteristic retention time of this critical pseudo-impurity was found to be very helpful regarding the choice of suitable operating conditions. An enrichment of the target protein fraction (ABF) was found to be theoretically possible using the multicolumn continuous gradient process. Productivities for isolating the antibody fragment fraction around 5 mg/day/mlstat phase can be expected from the simulated continuous multicolumn chromatographic process, which is currently under experimental validation. When planning concrete SMB validation experiments the real shape of the pH profiles should be considered instead of the ideal stepwise form assumed for the estimation of applicable operating conditions. An option is to refine the model by fitting the pH profiles measured and to take the real fluctuations into account. The considered open-loop three-zone two-step pH-gradient SMB process can be widely applied whenever a gradient elution chromatographic mode is needed, as in ion exchange, hydrophobic interaction, mixed mode or affinity chromatography. A quantitative comparison with other possible alternatives, as with the MCSGP process still needs to be performed.

Acknowledgement The authors gratefully acknowledge the financial support granted by the Deutsche Forschungsgemeinschaft (DFG) in the

241

frame of the project “From Gene to Product” (Sonderforschungsbereich 578, Project C2).

References Abel, S., Mazzotti, M., Morbidelli, M., 2002. Solvent gradient operation of simulated moving beds. I. Adsorption equilibrium constants. Journal of Chromatography A 944, 23–39. Beltscheva, D., Hugo, P., Seidel-Morgenstern, A., 2003. Linear two-step gradient counter-current chromatography analysis based on a recursive solution of an equilibrium stage model. Journal of Chromatography A 989, 31–45. Brereton, H.M., Chamberlain, D., Yang, R., Tea, M., McNeil, S., Coster, D., Williams, K., 2007. Single chain antibody fragments for ocular use produced at high levels in a commercial wheat variety. Journal of Biotechnology 129, 539–546. Broughton, D.B., Gerhold, C.G., 1961. Continuous sorption process employing fixed beds of sorbent and moving inlets and outlets. U.S. Patent No. 2,985,589. Buhlert, K., Lehr, M., Jungbauer, A., 2009. Construction and development of a new single-column simulated moving bed system on the laboratory scale. Journal of Chromatography A 1216, 8778–8786. Carta, G., Jungbauer, A., 2010. Protein Chromatography: Process Development and Scale Up. John Wiley and Sons. Chaga, G., 2001. Twenty-five years of immobilized metal ion affinity chromatography: past, present and future. Journal of Biochemical and Biophysical Methods 49, 313–334. Clemmitt, R.H., Chase, H.A., 2000. Facilitated downstream processing of a histidinetagged protein from unclarified E. coli homogenates using immobilized metal affinity expanded-bed adsorption. Biotechnology and Bioengineering 67, 206–216. David, F., Steinwand, M., Hust, M., Bohle, K., Ross, A., Dübel, S., Franco-Lara, E., 2011. Antibody production in Bacillus megaterium: strategies and physiological implications of scaling from microtiter plates to industrial bioreactors. Biotechnology Journal 6, 1516–1531. David, F., Westphal, R., Bunk, B., Jahn, D., Franco-Lara, E., 2010. Optimization of antibody fragment production in Bacillus megaterium: the role of metal ions on protein secretion. Journal of Biotechnology 150, 115–124. Deutscher, M.P., 1990. Guide to Protein Purification: Methods in Enzymology, vol. 182. Academic Press. Essen, L., Skerra, A., 1993. Single-step purification of a bacterially expressed antibody Fv fragment by immobilized metal affinity chromatography in the presence of betaine. Journal of Chromatography A 657, 55–61. Gottschalk, U., 2009. Process Scale Purification of Antibodies. John Wiley and Sons. Gueorguieva, L., Palani, S., Rinas, U., Jayaraman, G., Seidel-Morgenstern, A., 2011. Recombinant protein purification using gradient-assisted simulated moving bed hydrophobic interaction chromatography. Part II: Process design and experimental validation. Journal of Chromatography A 1218, 6402–6411. Guiochon, G., Felinger, A., Shirazi, D.G., Katti, A.M., 2006. Fundamentals of Preparative and Nonlinear Chromatography. Elsevier. Hochuli, E., Bannwarth, W., Döbeli, H., Gentz, R., Stüber, D., 1988. Genetic approach to facilitate purification of recombinant proteins with a novel metal chelate adsorbent. Nature Biotechnology 6, 1321–1325. Hust, M., Dübel, S., 2004. Mating antibody phage display with proteomics. Trends in Biotechnology 22, 8–14. Jordan, E., Hust, M., Roth, A., Biedendieck, R., Schirrmann, T., Jahn, D., Dübel, S., 2007. Production of recombinant antibody fragments in Bacillus Megaterium. Microbial Cell Factories 6, 2. Kessler, L., Gueorguieva, L., Rinas, U., Seidel-Morgenstern, A., 2007. Step gradients in 3-zone simulated moving bed chromatography: application to the purification of antibodies and bone morphogenetic protein-2. Journal of Chromatography A 1176, 69–78. Knecht, S., Ricklin, D., Eberle, A.N., Ernst, B., 2009. Oligohis-tags: mechanisms of binding to Ni2+ -NTA surfaces. Journal of Molecular Recognition 22, 270–279. Krättli, M., Ströhlein, G., Aumann, L., Müller-Späth, T., Morbidelli, M., 2011. Closed loop control of the multi-column solvent gradient purification process. Journal of Chromatography A 1218, 9028–9036. Laemmli, U., 1970. Cleavage of structural proteins during the assembly of the head of bacteriophage t4. Nature 227, 680–685. Mazzotti, M., Storti, G., Morbidelli, M., 1997. Optimal operation of simulated moving bed units for nonlinear chromatographic separations. Journal of Chromatography A 769, 3–24. Müller-Späth, T., Aumann, L., Melter, L., Ströhlein, G., Morbidelli, M., 2008. Chromatographic separation of three monoclonal antibody variants using multicolumn countercurrent solvent gradient purification (MCSGP). Biotechnology and Bioengineering 100, 1166–1177. Ottens, M., Houwing, J., Van Hateren, S.H., Van Baalen, T., Van der Wielen, L.A.M., 2006. Multi-component fractionation in SMB chromatography for the purification of active fractions from protein hydrolysates. Food and Bioproducts Processing 84, 59–71. Palani, S., Gueorguieva, L., Rinas, U., Seidel-Morgenstern, A., Jayaraman, G., 2011. Recombinant protein purification using gradient-assisted simulated moving bed hydrophobic interaction chromatography. Part I: Selection of chromatographic system and estimation of adsorption isotherms. Journal of Chromatography A 1218, 6396–6401. Porath, J., 1992. Immobilized metal ion affinity chromatography. Protein Expression and Purification 3, 263–281.

242

C.A. Martínez Cristancho et al. / Journal of Biotechnology 163 (2013) 233–242

Porath, J., Carlsson, J., Olsson, I., Belfrage, G., 1975. Metal chelate affinity chromatography, a new approach to protein fractionation. Nature 258, 598–599. Scott, M., Clarke, N., 2009. Next generation antibody therapeutics: antibody fragments, dual-targeting strategies, and beyond. European Pharmaceutical Review 5, 22–29. Skerra, A., Pfitzinger, I., Plückthun, A., 1991. The functional expression of antibody Fv fragments in E. coli: improved vectors and a generally applicable purification technique. Nature Biotechnology 9, 273–278.

Storti, G., Mazzotti, M., Morbidelli, M., Carrà, S., 1993. Robust design of binary countercurrent adsorption separation processes. AIChE Journal 39, 471–492. Weisser, N.E., Hall, J.C., 2009. Applications of single-chain variable fragment antibodies in therapeutics and diagnostics. Biotechnology Advances 27, 502– 520. Yang, Y., Malten, M., Grote, A., Jahn, D., Deckwer, W.D., 2007. Codon optimized Thermobifida fusca hydrolase secreted by Bacillus megaterium. Biotechnology and Bioengineering 96, 780–794.