Purification of Equine Chorionic Gonadotropin (eCG) using Magnetic Ion Exchange Adsorbents in Combination with High-Gradient Magnetic Separation Christine M€ uller Dept. of Chemical Engineering, Loughborough University, Loughborough, United Kingdom LE11 3TU

Elena Heidenreich and Matthias Franzreb Karlruhe Inst. of Technologie, Inst. of Functional Interfaces, Karlsruhe, Germany 76344

Katrin Frankenfeld fzmb GmbH, Bad Langensalza, Germany 99947 DOI 10.1002/btpr.2007 Published online November 13, 2014 in Wiley Online Library (wileyonlinelibrary.com)

Current purification of the glycoprotein equine chorionic gonadotropin (eCG) from horse serum includes consecutive precipitation steps beginning with metaphosphoric acid pH fractionation, two ethanol precipitation steps, and dialysis followed by a numerous of fixed-bed chromatography steps up to the specific activity required. A promising procedure for a more economic purification procedure represents a simplified precipitation process requiring only onethird of the solvent, followed by the usage of magnetic ion exchange adsorbents employed together with a newly designed ‘rotor-stator’ type High Gradient Magnetic Fishing (HGMF) system for large-scale application, currently up to 100 g of magnetic adsorbents. Initially, the separation process design was optimized for binding and elution conditions for the target protein in mL scale. Subsequently, the magnetic filter for particle separation was characterized. Based on these results, a purification process for eCG was designed consisting of (i) pretreatment of the horse serum; (ii) binding of the target protein to magnetic ion exchange adsorbents in a batch reactor; (iii) recovery of loaded functionalized adsorbents from the pretreated solution using HGMF; (iv) washing of loaded adsorbents to remove unbound proteins; (v) elution of the target protein. Finally, the complete HGMF process was automated and conducted with either multiple single-cycles or multicycle operation of four sequential cycles, using batches of pretreated serum of up to 20 L. eCG purification with yields of approximately 53% from single HGMF cycles and up to 80% from multicycle experiments were reached, with C 2014 American purification and concentration factors of around 2,500 and 6.7, respectively. V Institute of Chemical Engineers Biotechnol. Prog., 31:78–89, 2015 Keywords: equine chorionic gonadotropin, high gradient magnetic fishing, serum, purification

Introduction Equine chorionic gonadotropin (eCG) is a 60 kDa heterodimeric glycoproteine hormone composed of two noncovalently bound subunits. The a subunit of eCG shows 68 to 79% homology between different mammalian species whilst the b subunit acts as the “hormone specific” subunit.1 The average molecular weights of the a and b subunits are 16.96 and 43.72 kDa, respectively. eCG is the most heavily glycosylated glycoprotein hormone with 45% carbohydrate, of which 10% corresponding to sialic acid. This contributes to the very low isoelectric point (pI) of 1.8 of the hormone, a unique property which can be used for the purification.2 The a subunit bears two complex-type N-linked oligosaccharide chains located at asparagines, whereas the b subunit possess a carboxy-terminal peptide (CTP) which is O-glycosylated in addition to the N-glycans.3 Correspondence concerning this article should be addressed to M. Franzreb at [email protected]. 78

eCG can be detected in mare sera between day 40 and 130 of pregnancy. Although, during pregnancy there are detectable variations in the concentration and biological activity of eCG, the concentration of eCG reaches the highest level between day 60 to 90.4 In this period, levels of about 60 to 80 IUOHR mL21 can be attained in bioassays based on rats (Ovarian hyperemia reaction).5 Commercial eCG compounds are widely used to control the reproductive cycle in animal breeding as well as in veterinary medicine, as a means of inducing ovulation and superovulation. Depending on the species, dosages of 500 to 3,000 IUOHR are required, with a corresponding specific activity of >1,000 IUOHR mg21. Due to the long elimination half-life of the hormone eCG only one dosage is required.6 The plasma elimination half-life of eCG was determined to be about 6 days in horses7 and 5 to 15 days in cattle.8 The current state of the art for the purification of eCG from serum was developed in 1967 by Meyer et al.9 and can be divided into two steps. Firstly, an intermediate product is generated by precipitation and filtration to remove major C 2014 American Institute of Chemical Engineers V

Biotechnol. Prog., 2015, Vol. 31, No. 1

contaminants, such as albumin. Secondly, the final extract is generated using packed bed chromatography. To obtain the intermediate, fractionation with metaphosphoric acid and a two-step ethanol precipitation with 50 and 75% (v/v) at 4 C are employed. Subsequently, the pellet is dissolved and desalted by gelfiltration, to obtain a suitable solution for three ion exchange chromatography steps immediately succeeding one another. In contrast the procedure using magnetic adsorbents does not require the second ethanol precipitation, secondly only one further purification step using magnetic ion exchange adsorbents is required to obtain a product with a comparable purity. As can be seen from this example, bio-separation technology used in industry today is often based on principles first discovered over 70 years ago, indicating that improvements are needed at all stages of processing, i.e., from pretreatment of raw materials and during subsequent purification and modification to yield the final product. One possibility for such an improvement is the usage of magnetic adsorbents as a new option in protein purification in particular from a crude solution. Functional magnetic particles have the potential to enhance the physical and chemical properties of bio-separation processes, i.e., (micro-) particles have high specific surface areas, rapid binding kinetics, and unique physical and chemical properties. The possibility of using magnetic microadsorbents to isolate biomolecules directly from crude suspension has been known for some time in different areas of biotechnology. In a typical HGMF-process (High Gradient Magnetic Fishing), the target molecule is bound to the functionalized magnetic adsorbents in a batch reactor. Afterwards, the product loaded adsorbents are captured within a magnetic filter and separated from the nonmagnetic components which will pass through the filter unhindered. Finally, the product is further cleaned and recovered by washing and elution steps taking place within the magnetic filter. Each washing or elution step consists of four operations, namely: (i) filling the magnetic filter with new buffer while keeping the adsorbents separated, (ii) resuspension of the adsorbents in the buffer with the magnet turned off, (iii) recapturing of the adsorbents by switching the magnet on again, (iv) discharge of the spent washing buffer or the elution buffer containing the product. Several small-scale bioseparation experiments applying magnetic microadsorbents have been reported e.g. whey proteins,10,11 cell homogenates,12 plant extracts,13 enrichment of glycoproteins from serum for bioanalytics.14 Recently our group also published two HGMF examples using a small prototype of a “rotor stator” HGMF for lactoferrin purification from crude whey15 and eCG purification directly from serum using magnetic affinity adsorbents.16 Purification of eCG by magnetic affinity adsorbents was shown in 40 ml scale over 60 cycles. Over the first 12 cycles a loss of binding capacity of less than 10% could be achieved, resulting in a product yield of the process of around 50%. However, in the following cycles the yield decreased drastically and reached less than 30% of the original yield after 30 cycles.16 Therefore, the use of magnetic affinity adsorbents wasn’t considered for the larger scale required for the operation of the magnetic separator. With respect to eCG purification, we also reported the optimisation of the prepurification of serum required and the magnetic process in mL scale.17 In this study, for the first time we describe the integration of a magnetic separation step at pilot scale into a fully scal-

79

able downstream processing (DSP) scheme for the purification of the commercially relevant protein eCG. The reported purification procedure of eCG differs in both prepurification and final purification to the previously described conventional purification by Meyer et al.9 First, the initial generation of the intermediate is reduced by two steps decreasing the amount of ethanol necessary by 66% and omitting the corresponding solid-liquid separation. Second, the final purification of the resulting intermediate is accomplished using magnetic adsorbents in a pilot scale HGMF device instead of packed bed chromatography. Apart from the experimental results obtained for the recovery of eCG from horse sera using magnetic adsorbents, we also present data for the efficiency of the process by direct comparison to conventional chromatographic purification of the biomolecule.

Materials and Methods Extraction and purification of eCG Blood was collected from Haflinger mares between 50 and 100 days of gestation. After separation of erythrocytes and fibrin from the whole blood, the serum was stored at 220 C without additional preservatives. The serum collected showed activities of 100 to 400 MU mL21 (mouse units) determined by using immature female mice. From these batches, only serum containing 200 MU mL21 was used. In order to reduce the amount of bioassays necessary, for routine measurements they were replaced by a commercial immunoassay (ELISA) from Senova, Jena, Germany. The 200 MU mL21 sera resulted in activities in the immunoassay of 30 to 43 IUELISA mL21. The reason for the variation can be twofold: first, the activity of the serum donated by a pregnant horse is monitored using bioassays on a weekly basis only. However, depending on the state of pregnancy and other factors, day to day variation exists. Second, bio- and immunoassays use different epitopes of eCG. Therefore, an alteration of the molecule structure during treatment may influence one type of assay while the other shows no deviation. Nevertheless, in an extensive study Lecompte et al.18 showed that both types of assays are well correlated as long as the sialic acid content of the eCG is not changed and a suitable standard is used for the calibration.18 A detailed explanation of the different assay types is given in the following section. For prepurification the serum was treated in several batches of 2 L volume as follows. In the first step the pH of the serum was adjusted to 3.0 by addition of fresh 0.5 M meta-phosphoric acid, followed by a centrifugation at 9,000g and discarding the resulting precipitate. The pH of the supernatant was then adjusted to 4.5 with 1 M NaOH before adding an equal volume of prechilled (220 C) ethanol. After 2 h at 4 C, the precipitate was again removed by centrifugation, the supernatant was concentrated fourfold by ultrafiltration and followed by a discontinuous diafiltration against H2OMILLIQ using a cellulose membrane (10000 MWCO Hydrosart, Sartorius, G€ottingen, Germany) until a conductivity of 0.7 to 1 mS cm21 was reached. Starting with the intermediate resulting from the described prepurification, a final purification was performed using magnetic anion exchanger particles with N,N-diethyl-ammonium group as functional group (DEAP) within an HGMF process as described above. For the production of the magnetic anion

80

Biotechnol. Prog., 2015, Vol. 31, No. 1

Figure 1. Reaction scheme for the functionalization of carboxylated magnetic particles with DEAP.

exchange adsorbents, magnetic base particles with a standard bead size of 1 to 3 mm (carboxylated M-PVA, PerkinElmer chemagen Technologie GmbH, Baesweiler, Germany) were used. The particles were dissolved in H2Odd and 1-ethyl-3(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC) was added in a molar ratio of 1:2. For the reaction the particles were stirred for 30 min at room temperature. Afterwards, functionalization was completed by adding N,Ndiethyl-1,3-propanediamine (DEAP) in a 1:2 molar ratio whilst stirring for 4 to 5 h at room temperature. Finally the magnetic adsorbents were washed three times with H2Odd and stored at 4 C (Innovent Technologieentwicklung, Jena, Germany) (Figure 1). The degree of activation with DEAP was measured by titration using Tashiro’s indicator solution. The production of the indicator solution was done by dissolving 0.2 g of methyl red and 0.1 g of methylene blue in 100 mL of ethanol. The magnetic adsorbents used in this work had an average degree of activation with DEAP of 1,300 mmol g21. Characterization of eCG Bioassay for eCG Concentration. Ovarian hyperemia reaction (OHR) in immature rats (IDT Biologika GmbH, Dessau-Roßlau, Germany) was used for estimation of biological activity of eCG present in selected fractions after purification of eCG using DEAP adsorbents in a HGMF-process. The biological activity of the sera used was determined using immature female mice (IDT Biologika GmbH, DessauRoßlau, Germany). A fixed correlation exists between the biological activities determined in mice and rats: 3.8 MU mL21 5 1 IUOHR mL21. Therefore the used serum of 200 MU mL21 corresponded to an average of 53 IUOHR mL21. Immunoassay for eCG Concentration. In order to reduce the amount of bioassays necessary for the routine measurements, they were replaced by a commercial immunoassay (ELISA) for eCG (Senova, Jena, Germany). Procedures were performed according to the manufacturer’s protocol. Total Protein. The protein content was determined by the method of Bradford using bovine serum albumin as the standard. Chromatography Analogous to the packed bed chromatography procedure for eCG purification from horse serum described in literature,2,9,19 selected anion exchange (AEX) materials in 1 mL € packed bed plastic columns connected to an AKTApurifier

100 workstation (GE Healthcare, Uppsala, Sweden), were tested. For the purification, weak (DEAE Sepharose Fast Flow) and strong (Q Sepharose Fast Flow (Q FF), and Q Sepharose XL (Q XL)) AEX materials were chosen. For analysis a lyophilisate was obtained following the method of Meyer et al.9 The lyophilisate was resolved in equilibration buffer (20 mM sodium acetate, pH 4.5) to reach an eCG solution with approximately 200 IUELISA mL21. Afterwards the extract was filtered through a 0.22 mm syringe filter (Millipore, Darmstadt, Germany) and loaded onto a packed bed, which was previously equilibrated with 20 mM sodium acetate (pH 4.5). After washing with 5 CV of the same buffer, the column was developed with a linear gradient of increased salt (0–1 M NaCl) over 20 CVs. 1 mL fractions collected during chromatography (Fraction Collector Frac950, GE Healthcare, Uppsala, Sweden) were retained for analysis of protein composition, eCG and protein contents (data not shown). The strong AEX material, Q FF, showed the most promising results (data not shown) and was subsequently used for purification of the eCG solution achieved after ultra/diafiltration following the modified protocol of Meyer et al.,9 to create identical starting conditions for packed bed chromatography and HGMF operation. Chromatography was performed using Q FF in a 20 mL packed bed plastic column (GE Healthcare, Uppsala, Sweden), using the above described buffer system. After equilibration of the column, an eCG extract with a total of approximately 8,800 IUELISA was loaded on the packed bed. After washing with 5 CV, the column was developed with an increasing linear salt gradient (0–1 M NaCl) over 20 CVs; 5 mL fractions collected during chromatography were retained for analysis of protein composition, eCG, and protein contents. Binding isotherm of DEAP adsorbents Protein binding capacity of the DEAP adsorbents for eCG was analyzed in small scale batch binding tests using the intermediate product after the pretreatment ultradiafiltration step. The protein concentrate was diluted in 10 mM sodium acetate buffer at pH 4.5 and a concentration of 0.018 mg mL21 (180 IUELISA mL21) was used. DEAP particles were equilibrated in the same buffer with a particle concentration of 0.4 to 20 g L21 in 1 mL. Particles were washed three times with 0.5 mL equilibration buffer, then replaced with 1 mL protein solution. The particles were incubated at room temperature for 20 min with shaking (1,000 rpm, Thermomixer Comfort shaker, Eppendorf, Hamburg, Germany), and the eCG content was measured from the supernatant using

Biotechnol. Prog., 2015, Vol. 31, No. 1

the described immunoassay. The protein bound to the support was calculated from the difference between the protein concentration in the initial solution and in the supernatant after binding. The resulting adsorption data for these experiments were then fitted to the Langmuir equation: Q5

Qmax
C Kd 1C

where Q* and C* represent the equilibrium values of adsorbed and liquid-phase protein, respectively; Qmax is the maximum protein binding capacity of the support; and Kd is the dissociation constant. Data was fitted to the model using SigmaPlot software version 9.0 (Systat Software Inc., San Jose, CA). Sorption kinetics of DEAP adsorbents Small-scale kinetic studies were carried out in 1.5 mL reaction tubes using the intermediate product after the ultradiafiltration step of the pretreated serum with defined amounts (5 g L21) of DEAP supports. The adsorption kinetics was determined in 10 mM sodium acetate buffer at pH 4.5 as equilibration and binding buffer. The particles were incubated with protein solution for between 30 s to 20 min. The eCG content in the supernatant was measured by immunoassay. For desorption studies, the particles were incubated with protein solution for 20 min. Bound proteins were desorbed afterwards with 10 mM sodium phosphate buffer containing 0.3 M NaCl at pH 8 for 60 s up to 20 min. All the above operations were carried out at room temperature with shaking at 1,000 rpm. Reusability of DEAP particles Long-term stability of the supports was investigated over a period of 70 cycles in 1.5 mL reaction tubes using the intermediate product after ultra-diafiltration. Initially the supports were washed three times with equilibration buffer (10 mM sodium acetate buffer at pH 4.5) and then incubated for 20 min with protein solution. Subsequently, the supernatant was discarded and the particles were washed again three times for 5 min with washing buffer (10 mM sodium acetate buffer at pH 4.5). For elution, the particles were incubated twice in a 10 mM sodium phosphate buffer containing 0.3 M NaCl at pH 8 for 20 min. Before adding another eCG solution, the particles were washed again three times with washing buffer. All of the described operations were carried out at room temperature with shaking at 1,000 rpm. Magnetic particle concentration Magnetic particle concentrations were determined by dry weight measurements. Samples containing magnetic particles were washed with deionized water to release entrained salts or solids and then applied to 1.5 mL weighed glass snap ring vial (VWR International GmbH, Darmstadt, Germany). Afterwards particles were dried over night at 60 C (VE400 W€armeschrank, Memmert, Schwabach, Germany). After cooling to room temperature in a desiccator jar, the glass vessels were re-weighed and the amount of particles in each sample was calculated. Automated HGMF separator The magnetic separation set-up employed in this study (Figure 2a) is a cooperative development of the Karlsruhe

81

Institute of Technology (KIT; Eggenstein-Leopoldshafen, Germany), ABBIS-bio process automation (Wiesbaum, Germany) and PerkinElmer chemagen Technologie GmbH (Baesweiler, Germany). The apparatus (Figure 2c) consisted of a (i) stirred batch adsorption reactor (B1); (ii) five additional tanks; (iii) a single two directional variable speed peristaltic pump; (iv) two computer controlled six-way valves; (v) a bubble detection (BD); (vi) a water-cooled solenoid which can be turned “on” and “off” with a field strength up to 0.25 T; and (vii) a specially designed “rotor stator” magnetic filter (internal diameter 5 86 mm; working volume 5 980 mL) with associated stirrer (MF). Figure 2b shows a three-dimensional sectional drawing of the “rotor stator” magnetic filter surrounded by the water-cooled solenoid. During loading of the magnetic filter with the field “on,” the particle/feedstock suspension is pumped through V1.1. The loaded magnetic adsorbents are collected in the magnetic filter and the supernatant is pumped out via V2.5. The following purification procedure of the target protein runs automatically by a computer controlling the operation of the valves, stirrer, solenoid and pump. Following the separation of the loaded magnetic adsorbents, the remaining unbound serum proteins in the filter chamber are displaced by washing buffer (V1.2) in three sequential washing steps. During this washing procedure the magnetic field is switched ‘off’ and the rotor-stator filter matrix is operated at 850 rpm (see Ref. [15 for a more detailed description of the rotor-stator system). In this way, strong shear forces are generated within the filter matrix, resulting in efficient detachment, deagglomeration and washing of the adsorbents. After each washing step, the magnetic field is switched “on” again, to recapture the magnetic particles. Recapturing of the particles is completed by pumping the liquid in the magnetic filter chamber and the in- and outlet tubing for 120 s in the opposite direction then during regular operation using a closed recycle loop (V1.6-V2.6). Each wash fraction is discharged through V2.1. The elution of the bound product occurs in up to three sequential steps drawing elution buffer via V1.3 and collecting the elution fractions containing the target protein in tank B6 (V2.4). Finally, the elution buffer remaining in the magnetic filter chamber is displaced by two further washing steps. All washing and elution steps follow the same sequence for buffer displacement, resuspension, and recapturing of the magnetic adsorbents as described in detail for the first washing step, except for some variation in the buffer volumes. Under the conditions employed, the magnetic field gradient generated within the filter chamber resulted in an efficiency of particle retention very close to 100%, meaning no magnetic particle breakthrough was observed during any of the process steps described. According to the manufacturer, the capacity of the magnetic filter amounts to 100 g of MPVA supports. In the case of the eCG purification HGMF process, the maximum working capacity used was 60 g of DEAP-functionalized magnetic M-PVA adsorbents.

Recovery of eCG in a single-cycle HGMF operation Equilibrated DEAP adsorbents (10 mM sodium acetate buffer at pH 4.5) were suspended in the pretreated intermediate obtained after the ultra/diafiltration step in B1 and mixed at room temperature for 20 min using an overhead stirrer. Subsequently, the suspension was pumped through the magnetic filter at a linear flow rate of 1,020 mL min21, while

82

Biotechnol. Prog., 2015, Vol. 31, No. 1

C ABBIS bio proFigure 2. a: Photograph of the “rotor-stator” magnetic separator. b: Filtration chamber with “rotor-stator” discs (V cess automation, Wiesbaum, Germany). c: Schematic illustration of the automated high gradient magnetic separation (HGMF) system. B1-6: tanks, V1.1-V2.6: valves, P: pump, BD: bubble detection, MF: magnetic filter.

the magnetic field was switched “on.” The loaded adsorbents were collected in the magnetic filter, whereas the supernatant containing unbound biomolecules passed through the filter and was collected as the “flow through” fraction. With the magnetic field on, a first washing step was conducted using 2,700 to 3,200 mL wash buffer (10 mM sodium acetate buffer at pH 4.5) in a liquid displacement, resuspension, recapturing sequence described above. The wash procedure was repeated twice and fractions “Wash 2” and “Wash 3” were collected for further analysis. The filter was then filled with 1,800 mL elution buffer (10 mM sodium phosphate buffer containing 0.3 M NaCl at pH 8) over 180 s and simultaneously rinsing out approximately 1 L of wash buffer, which remained in the filter after the washing steps, and 800 mL of elution buffer, which passed the filter. Despite this, the fraction collected was named “Elution 1.” Three elution

steps were performed in total and the supports were resuspended and collected as described above. The elution fractions were collected individually in tank 6. Once elution was completed, the particles were washed twice again to prepare them for the next sorption step by pumping 2,700 to 3,200 mL wash buffer through the filter, and the fractions “Wash 4” and “Wash 5” were collected. The washing procedure was performed following the same procedure carried out for wash steps 1 to 3. Once the magnetic adsorbents were resuspended after the fifth washing step, the magnetic field was kept ‘off’ and the particles were rapidly flushed out (1,100 mL min21, pump direction: backwards) while stirring the system. After emptying the filter it was filled again with wash buffer and rinsed for a second time. This procedure was repeated four times until no further magnetic particles were visible in the effluent while emptying the filter.

Biotechnol. Prog., 2015, Vol. 31, No. 1

83

Afterwards, the collected supports were concentrated over night by magnetic settling on a 0.5 T permanent magnet block, the supernatant was removed and the supports were stored in 20% EtOH at 4 C. One complete automated cycle, including the 20 min incubation time in the stirred batch reactor, took approximately 110 min, when 20 L of feedstock was used. Recovery of eCG in multicycle HGMF operation Operation in multicycle mode was executed similar to the above described single-cycle mode. However, following a complete single-cycle, the magnetic adsorbents were flushed out of the system using another eCG feedstock instead of washing buffer and were directly collected into the stirred batch tank (B1) for 20 min incubation. The multicycle experiment was carried out over four cycles, using 4,589 mL eCG solution and a particle concentration of 4.5 g L21 in each cycle. At the end of the final cycle the magnetic adsorbents were flushed out of the filter and treated in the same way as described above.

Results and Discussion DEAP particle characterization M-PVA particles functionalized with N,N-diethyl-ammonium (DEAP) groups were characterized by examining their ability to bind eCG from the pretreated intermediate after the ultra/diafiltration step. The adsorption isotherm exhibited Langmuir adsorption behavior. The adsorption capacity Qmax and dissociation constant Kd deduced from the Langmuir isotherm (Eq. 1) is shown in Table 1. Due to the high ionic strength (10–13 mS cm21) direct capture of eCG from untreated serum was not possible. Even after Table 1. Adsorption Capacity Qmax and Dissociation Constant Kd of eCG on DEAP-Functionalized Adsorbents at Room Temperature Qmax (mg g21) eCG in ultra/diafiltrate

3.2

Kd (g L21) 23

1.2 3 10

Qmax/Kd (L g21) 0.27

removal of the main contaminant albumin and treatment by an ultra/diafiltration step, the maximum binding capacity for eCG from pretreated intermediate only reached 3.2 mg g21. In comparison, previous experiments using an eCG extract of higher purity, e.g. re-solved lyophilisate after extraction of eCG from serum following the protocol of Meyer et al.,9 showed binding capacities up to 38 mg21. The comparatively low binding capacity for eCG in case of the intermediate indicates competing adsorption of other proteins which are still present in much higher concentrations than eCG. However, with a concentration of 0.44 to 0.58 mg mL21 the total protein concentration in the supernatant after adsorption showed no change in the context of measurement accuracy. An important aspect in the characterization of an ion exchanger is its total ionic capacity. The total ionic capacities of the commercial weak AEX support, DEAE (GE Healthcare, Uppsala, Sweden) and the magnetic DEAP adsorbents were measured as 3 to 4 mmol g21 and 1.3 mmol g21, respectively. Optimisation of the DEAP particle concentration as well as the optimal adsorption and desorption conditions are crucial parameters for the HGMF process. The influence of DEAP particle concentration between 0.2 and 6 g L21 on the adsorption of eCG from pretreated intermediate (Figure 3a) as well as the adsorption and desorption kinetics (Figure 3b) were tested. As can be seen in Figure 3a, adsorption of eCG increases almost linearly up to a particle concentration of 2.5 g L21. At higher particle concentrations, the eCG in the intermediate is almost completely adsorbed by the DEAP supports. Due to the limited mass of particles that can be retained in the magnetic filter, the particle concentration in the batch reactor directly influences the volume of pretreated intermediate that can be processed in one cycle. At a particle concentration of 3 g L21, the maximum working capacity of the filter would be reached after filtering a volume approximately 67% larger than if a particle concentration of 5 g L21 was used. Nevertheless, in the pilot plant HGMF experiments, particle concentrations of 3 to 4.5 g L21 were employed. Depending on the stadium of pregnancy of the mares, high variations of the eCG contents in the sera are detectable. Usage of higher particle concentrations ensures near complete adsorption of eCG, including sera with maximum levels of eCG.

Figure 3. a: Effect of DEAP particle concentration on the adsorption of eCG from pretreated intermediate, binding buffer: 10 mM sodium acetate buffer at pH 4.5. b: Binding and elution kinetics on DEAP supports, binding buffer: 10 mM sodium acetate buffer at pH 4.5, elution buffer: 10 mM sodium phosphate buffer containing 0.3 M NaCl at pH 8.

84

Biotechnol. Prog., 2015, Vol. 31, No. 1

defrosted eCG solutions than eCG solution at room temperature, likely to be observed after prolong use. HGMF-processes

Figure 4.

Purification of eCG from pretreated intermediate reusing magnetic DEAP adsorbents (70 cycles); initial eCG concentration: 187 IUELISA; initial particle concentration: 5 g L21.

Figure 3b shows the adsorption and elution kinetics for eCG on magnetic DEAP adsorbents over time. A near complete adsorption of eCG occurs by 60 s. In contrast, the results of the desorption studies show an increase of the elution efficiency depending on the incubation time, even after 200 s. Approximately 80% elution of the bound eCG could be reached after 4 min, a control experiment after 10 min and 20 min showed an elution efficiency of around 83% and 89%, respectively. The experiments indicate, the repeated elution steps are one of the most time-consuming periods in the HGMF process, due to the slow desorption kinetics. Although only a slight increase of eCG in the elution fraction could be observed after 4 min, we chosen a timeframe of 20 min in the following HGMF-processes, to obtain the maximum yield in the process.

Preparation of Pretreated Intermediate. Two-liter batches of serum (200 MU) were employed for each pretreatment step. To produce a suitable feedstock for the DEAP particles in the HGMF process, pretreatment was necessary to reduce the quantity of competing proteins and the ionic strength of the serum. Prepurification of eCG from serum was performed by a simplified process scheme, following the method of Meyer et al.9 in the beginning; however, avoiding the second ethanol precipitation and therefore using 66% less organic solvent. Table 2 shows the results of the simplified purification process for five independent batches. The first precipitation step with meta-phosphoric acid resulted in an eCG activity loss of approximately 11.2%. The following precipitation with 50% (v/v) ethanol and the ultra/diafiltration step resulted in a total loss of approximately 21.7% eCG activity in the intermediate. The loss caused by the EtOH precipitation step alone could not be determined, due to the strong influence of EtOH on the ELISA. An average purification factor of 406 with a concentration factor of 3.4 could be achieved. The product yield was approximately 78.3% of the eCG initially present in serum. eCG shows the unique property of having a very low isoelectric point of 1.8. In the first precipitation step using metaphosphoric acid (pH 3) this fact is used. In the short time of 5 min chosen for the precipitation, the loss of eCG activity was very low in each experiment, corresponding with a low deviation of the step yield. Higher losses in activity could be observed in the following precipitation step using 50% ethanol (v/v). The duration of this step is 2 h, leaving time for a substantial degradation and an increased fluctuation of the resulting yield. In addition it showed, that the varying salt content of the serum (conductivities swayed between 10 and 13 mS cm21) has a stronger influence on the ethanol-precipitation step than on the precipitation using meta-phosphoric acid.

Reuse of DEAP particles The potential of industrial use of the HGMF process strongly depends on the possibility of recycling the DEAP particles for as many cycles as possible. Therefore, the reuse of the supports was investigated for 70 cycles of eCG isolation from pretreated intermediate. Figure 4 shows the yield of eCG in the elution fraction and the DEAP particle concentration per cycle. Particle loss due to recycling occurred because of an incomplete particle recovery after each cycle. To minimize this particle loss, the pipette tips were not discharged after each step, by using a different pipette with a fixed tip for each step. This meant that, the loss of particles attached to the walls of discharged tips could be avoided, resulting in a loss of only 6.4% over 70 cycles. During the first 54 cycles, 61.2% of the eCG initially available in the feedstock could be recovered in the elution fraction, followed by a decrease of the average recovery in the cycles 55 to 65 and an increase in return in cycle 68 and 70. However, the results clearly show that a stable binding capacity of the DEAP particles is maintained through 54 cycles and that the particles are suitable for frequent recycling for a minimum of 70 cycles. The undulating pattern of eCG yield shown in Figure 4 is likely to be due variable change in the temperature of the feedstock. Better adsorption was observed in freshly

Recovery of eCG in a single-cycle HGMF process The results for eCG recovery from the intermediate in two different single cycles following the HGMF procedure described in the methods section are summarized in Tables 3 and 4. A particle concentration of 4.5 and 3 g L21 was selected, processing 2,285 mL and 19,840 mL feedstock respectively. The, in comparison, low particle concentration of 3 g L21 is due to the processed volume of nearly 20 L feedstock and the limited amount of DEAP supports available of 60 g. In previous work, we already had demonstrated a possibility of small scale benchmarking experiments for the purification of eCG from feedstocks up to 100 mL using 5 g L21 DEAP supports and this study simplified prepurification of the serum.17 Using the HGMF system, we processed 2,285 mL and 19,840 mL of feedstock, respectively, which represents approximately 23-fold and 200-fold volume increases of the laboratory scale experiments, respectively. Table 3 shows the results of the first single cycle HGMF experiment using 4.5 g L21 DEAP supports and 2,285 mL prepurified intermediate as feedstock, originating from 10 L serum after the ultra/diafiltration step. In the adsorption step of the HGMF process, a binding capacity of 1.5 mg eCG per g DEAP supports was reached,

Biotechnol. Prog., 2015, Vol. 31, No. 1

85

Table 2. Preparation of an eCG Intermediate for Further HGMF Purification Step

Protein (mg mL21)

eCG (IUELISA mL21)

Volume (mL)

Yield (%)

Purification Factor

Concentration Factor

57.5 6 2.93 6.18 6 0.76 0.5 6 0.09

33.6 6 4.2 26.1 6 3.4 112 6 8.9

2,000 2,290 6 65 470 6 31

100 88.8 6 5.27 78.3 6 17.3

1 7.31 6 1 406 6 145

1 0.78 6 0.04 3.4 6 0.64

Serum HPO3 ppt UF/DF

Table 3. Purification Table for the Isolation of eCG From Pretreated Intermediate Using 4.5 g L21 DEAP Supports in a Single Cycle HGMF Operation Step UF/DF Flow-through Wash I Wash II,III Elution I Elution II Elution III Mass balance (%)

Protein (mg mL21)

eCG (IUELISA mL21)

Volume (mL)

HGMF-Yield (%)

HGMF-Purification Factor

HGMF-Concentration Factor

0.48 0.26 0.01 0 0.01 0.05 0.01 77.4

93.3 23 0.58 0 2.85 77.7 0 97

2,285 2,700 2,700 2,700 1,800 1,800 1,800

100

1

1

2.41 65.6 0

1.47 8.0 0

0.03 0.83 0

Table 4. Purification Table for the Isolation of eCG from Pretreated Intermediate Using 3 g L21 DEAP Supports in a Single Cycle HGMF Operation Step UF/DF Flow-through Wash I Wash II,III Elution I Elution II Elution III Elution IV Mass balance (%)

Protein (mg mL21)

eCG (IUELISA mL21)

Volume (mL)

HGMF-Yield (%)

HGMF-Purification Factor

HGMF-Concentration Factor

0.26 0.23 0.02 0.02 0.01 0.04 0.01 0.03 101.7

53.5 14 0.45 0 13.4 302 66.5 11.9 95.1

19,840 21,400 2,700 2,700 1,800 1,800 1,800 1,800

100

1

1

2.3 51.3 11.3 2

5.9 37.8 29.1 2.09

0.25 5.65 1.24 0.22

which corresponds well with the theoretical value calculated by the Langmuir parameters given in Table 1. However, a relatively high fraction of eCG remained unbound, corresponding to a loss within the flow-through of 29%. Therefore, despite a good elution yield of 95.8% of bound eCG, the yield of the HGMF process amounts to 68%. Due to backmixing of the third washing buffer and the first elution buffer in the separator, the sodium chloride concentration in the system does not change sharply between the washing and the elution step. As a consequence, approx. the first half of the draining elution volume contains washing buffer and, hence, decreasing the eCG concentration in the first elution fraction. In the following second elution step, a higher eCG concentration of 77.7 IUELISA mL21 was achieved, due to a higher eluting power of the buffer. The purification factor and the concentration factor for eCG in the second elution fraction were 8.0 and 0.83, respectively. The mass balance for total protein and eCG closed to 77.4% and 97%, respectively. In combination with the yield of 78% (see Table 2) in the ultra/diafiltrate after prepurification of eCG from serum, the theoretical yield of the entire process is 53%. To determine the specific activity, the second elution fraction was dialyzed to a conductivity of 1 mS cm21 and lyophilized. On the basis of the dry weight, a specific activity of 136 IUELISA mg21 was determined. While these results show the general operability of the developed purification scheme, the achieved activities in the eluate and corresponding lyophilisate are not satisfactory. This is true as the laboratory scale experiments mentioned previousely17 led to higher specific activities in the final product.

The main difference between the labscale experiments and the first HGMF run is the low particle concentration during elution. Therefore, a second HGMF run was conducted using larger amounts of supports and pretreated intermediate. Table 4 shows the results, using 3 g L21 DEAP supports and 19,840 mL of intermediate achieved by pretreatment of approximately 80 L of serum. Again, incomplete binding resulted within an eCG loss in flow-through of 28%, limiting the yield of the HGMF process to 66.9%. However, the purification and concentration factor for eCG in the second elution fraction were much higher than in the first run, reaching 37.8 and 5.65, respectively. The resulting binding capacity of the DEAP supports was 1.3 mg eCG per g, and in total 93.1% of the bound eCG was eluted in three elution steps. The mass balance for total protein and eCG was 101.7% and 95.1%, respectively. To identify the specific activity, the second elution fraction was again dialyzed to a conductivity of 1 mS cm21 and lyophilized. On the basis of dry weight, a specific activity of 598 IUELISA mg21 was determined, which is more than four times higher than from the first run. Assuming an yield of for example, 78% (see Table 2) in the ultra/diafiltrate, as performed in the HGMF operation before, a theoretical yield of the entire process of 52% is resulting in this case. Considering the high amount of 60 g magnetic particles and the resulting high particle concentration in the elution step, the concentration of eCG reached in the second elution fraction as well as the specific activity of the lyophilisate were below the expectation of 800 IUELISA mg21. The main reason can be found in the fact that in this second experiment the eluted eCG was not concentrated almost completely within elution volume II but within elution

86

Biotechnol. Prog., 2015, Vol. 31, No. 1

Table 5. Purification Table for the Isolation of eCG from Pretreated Intermediate in a Multicycle HGMF Operation Using 4.5 g L21 DEAP Supports Step UF/DF Flow-trough (1) Wash I (1) Wash II,III (1) Elution I (1) Elution II (1) Elution III (1) Mass balance (%) Flow-trough (2) Wash I (2) Wash II,III (2) Elution I (2) Elution II (2) Elution III (2) Mass balance (%) Flow-trough (3) Wash I (3) Wash II;III (3) Elution I (3) Elution II (3) Elution III (3) Mass balance (%) Flow-trough (4) Wash I (4) Wash II,III (4) Elution I (4) Elution II (4) Elution III (4) Mass balance (%)

Protein (mg mL21)

eCG (IUELISA mL21)

Volume (mL)

HGMF-Yield (%)

HGMF-Purification Factor

HGMF-Concentration Factor

0.45 0.22 0.04 0.02 0.01 0.08 0.03 103 0.19 0.01 0.02 0.06 0.1 0.03 93.5 0.17 0.01 0 0.02 0.08 0.03 60.8 0.18 0.04 0.02 0.02 0.06 0.04 76.6

81 1.6 0.62 0 1.19 148 29.2 90.7 1.6 0.3 0 6.11 150 21.9 89.9 1.8 0.3 0 3.38 185 31.9 107 2.1 0.45 0 7.82 179 29.5 109

4,589 6,200 3,200 3,200 1,800 1,800 1,800

100 2.59 0.53 – 0.6 71.4 15.3

1

1

1.19 10.3 5.85

0.02 1.82 0.36

7,500 3,200 3,200 1,800 1,800 1,800

3.23 0.01 – 3.0 73.1 10.6

0.62 8.07 3.64

0.08 1.86 0.27

5,750 3,200 3,200 1,800 1,800 1,800

2.79 0.01 – 1.6 89.7 15.4

0.85 13.2 5.31

0.04 2.29 0.39

6,750 3,200 3,200 1,800 1,800 1,800

3.89 0.39 – 3.8 86.7 14.3

2.61 17.9 3.84

0.1 2.21 0.36

volumes II and III. Due to less complete elution at higher product concentrations in solution and due to backmixing effects of approximately 20%, the expected eCG amount in the elute shifts to the next volume. A partial solution could be an increase of the concentration of the elute buffer, in order to achieve a more complete elution within one step. The maximum capacity of the filter used in this work amounts to at least 100 g M-PVA particles. It must be assumed that in case of an exploitation of the system under optimum conditions, a further improvement of the results is possible. Based on the use of 100 g M-PVA-DEAP supports, it can be estimated that the requested specific activity of >1,000 IUOHR mg21 should be possible. Recovery of eCG in a multicycle HGMF process In an industrial scenario, many cycles of the HGMF process would be conducted by recycling DEAP supports. In order to investigate the performance of such a scenario, four sequential purification cycles were performed at a particle concentration of 4.5 g L21 using 4,589 mL of intermediate each time. Between each cycle, the magnetic supports were flushed out of the magnetic filter with 4,589 mL of new feedstock with binding buffer to transfer all of the new feedstock into the stirred batch reactor. In this batch reactor the particles were incubated for 20 min with the feedstock before the suspension was pumped through the magnetic filter. Some variations in the volume of flow-through resulted from different volumes of binding buffer being used for rinsing the particle/ feedstock suspension into the batch reactor. In the multicycle HGMF process, a binding capacity of 1.8 mg eCG per g DEAP supports was reached, at an eCG concentration of only 1.6 to 1.8 IUELISA mL21 in the flow-through, proving a high affinity of the DEAP adsorbents. A possible reason for the

higher binding capacity in the multicycle HGMF process might be the use of a high quality particle batch. The yields, concentration and purification factors observed for eCG gently increased with the number cycles. The first two cycles showed an accumulated yield of the elutes of 72% and 88% (Table 5), respectively. In the following two cycles, the accumulated yield of the elutes of the HGMF process reached numbers of 95% and almost 100%, respectively. In the first two cycles, no complete elution of the bound eCG was reached (% elution 88.5% (1); 89.5% (2)), however, due to recycling of the DEAP supports, the eCG remaining on the particles seemed to be eluted in the preceding cycles 3 and 4 (% elution 109% (3); 104% (4)), (Table 5). In the second elution fraction, a purification factor of 8 to 18 and a concentration factor of 1.8 to 2.3 were reached. The specific activity of the second elution fraction of cycle 1 and 2 was determined gravimetrically after dialyzation and lyophilization. However, in this case dialysis was operated until the samples showed conductivities in the range of 0.5 to 0.7 mS cm21. Additionally dialyzed samples of the multicycle HGMF operation were tested for their biological activity in a bioassay based on rats (OHR). Based on dry weight, a specific activity of 494 IUELISA mg21 (cycle 1) and 928 IUELISA mg21 (cycle 2) was observed. In vivo, a corresponding specific activity of 600 IUOHR mg21 (cycle 1) and 1273 IUOHR mg21 (cycle 2) was observed. In comparison the in vitro and in vivo activity of the sample of cycle 2 is almost twice the one of the sample of cycle 1, which is due to the stronger desalination in the case of the cycle 2 sample. However, care has to be taken, because very low salt concentrations result in deactivation of eCG (results not shown). In order to better understand the consequence of eCG biological activity during the entire purification process, different samples of the procedure were additionally investigated by

Biotechnol. Prog., 2015, Vol. 31, No. 1

87

Table 6. Purification Table for the Isolation of eCG From Serum by Simplified Pretreatment and HGMF Based on Data From Bioassays for eCG and BCA Measurements of Total Protein Step Serum UF/DF Elution II (1) Elution II (2)

Protein (mg mL21)

eCG (IUOHR mL21)

Volume (mL)

Yield (%)

Purification Factor

Concentration Factor

52 0.45 0.13 0.17

51.2 105 340 343

21,450 4,589 1,800 1,800

100 44 55.8 56.3

1 237 2,656 2,049

1 2.05 6.6 6.7

bioassays based on rats. Table 6 summarizes the main results exemplary for cycles (1) and (2) of the multicycle HGMF test. From the data it can be seen, that the precipitation with meta-phosphoric acid and EtOH resulted in a partial loss of the available activity, which cannot be explained by the physical loss of the target alone. eCG is stable in a pH range of 4 to 8, but a pH of less than 3, which occurs during the precipitation step using meta-phosphoric acid, the noncovalent linkage between the subunits is gradually decreased and a loss of sialic acid residues can occur.19 In case of sialic proteins, the sialic acids in the molecule maintain the hormone in circulation. A removal of terminal sialic acid leads to simplified access of the enzyme neuramidase destroying biological activity in vivo.20 The following precipitation using 50% EtOH (v/v) effects the intramolecular hydrophobic interactions which may result in a partial degradation of the protein.21 Therefore, literature indicates that both precipitation steps may affect the biological activity of the hormone. However, the low biological activity of 105 IUOHR mL21 after EtOH precipitation and UF/DF and the resulting calculated yield at this point of only 44% in comparison to the final yield after the magnetic sorbent purification step of approximately 56%, suggests that the bio assay of the intermediate still containing some EtOH after the UF/DF step is prone to some systematic error. Nevertheless, the values of the bioassays of samples before and after the HGMF process show that the use of magnetic sorbents does not interfere with the biological activity of the target (Table 6). In the yield calculated from the bioassay, only the second elution fraction containing the highest amount of eCG is taken into account, while the first and third elution fractions are ignored despite containing eCG. If eCG from these fractions is considered, the total yield of each cycle would amount to approximately 60 to 70%. This result is similar to the total yield of eCG measured by immunoassay of approximately 67% (cycle 1) and 64% (cycle 2). In animal breeding as well as in veterinary medicine, commercial eCG compounds are used to control the reproductive cycle. Depending on the animal, dosages of 500 to 3,000 IUOHR are required, with a required specific activity of the eCG product of >1,000 IUOHR mg21 in vivo.6 The second cycle of the multicycle HGMF operation with 1273 IUOHR mg21 already conforms to the requirement. An assembly of all three elution fractions from one cycle would result in a higher total eCG yield, but would also result in a decrease of the specific activity of the final eCG product. In addition, an increase of the recovery and purification factors was detectable from cycle to cycle by recycling the DEAP supports in the HGMF operation (Table 5), correlating with an increase of the biological activity.

Benchmarking This study has confirmed that the HGMF process produces eCG with the required specific activity. An important aspect

to be addressed is how the HGMF procedure compares with conventional chromatographic procedures. Therefore, different conventional resins of the company GE Healthcare have been screened for their suitability to purify eCG from pretreated intermediate. In the screening, Q XL, DEAE, and Q FF 1 mL HiTrap columns have been tested. While from these the weak basic ligand of the DEAE resins comes closest to the DEAP ligands used for the magnetic adsorbents, the strong basic resins Q FF and Mono Q showed superior performance. The reason for this difference in ligand preference between sepharose based beads and the magnetic beads having a PVA matrix is not clear, and out of the scope of this work. On the basis of this screening, in the benchmark the optimum ligand of the respective processes was used, saying DEAP in case of the magnetic adsorbents and QFF in case of fixed bed chromatography. In the benchmarking experiment conducted, running HGMF and anion exchange chromatography, the same feedstock was applied. The feedstock (intermediate after UF/DF) received the same pretreatment, expect of a subsequent 0.22 mm filtration step of the feedstock before Q FF chromatography to ensure the feedstock is free of fine suspended solids. The 0.22 mm filtration step resulted in unavoidable losses of intermediate volume, corresponding to a reduction of the eCG amount available of 12%. In both cases, serums with 200 MU were used for intermediate preparation. The data of the comparison shows that the usage of magnetic DEAP adsorbents resulted in a 22% higher yield (total of elution I and II) in comparison to the chromatography with commercial Q FF material. For the first elution step in the magnetic process, it was possible to increase the purification factor by a factor 5.5 and the concentration factor by a factor 4.3. For Q FF matrices a specific activity of eCG of 581 IUELISA mg21 (693.5) could be reached. Purification of eCG using magnetic DEAP adsorbents resulted in a product with a specific activity of 1951 IUELISA mg21 (6305). The standard deviation was calculated with triplicate and duplicate results, respectively. The, in comparison, low specific activity of eCG with Q FF material is astonishing taking into account the fact that chromatography realizes multiple separation stages, while HGMF uses a batch adsorption step only. A possible explanation could be that the available chromatography column has been roughly oversized by a factor of 2, and therefore binding to many competing proteins which were later eluted together with the target protein, eCG. A further advantage of the magnetic process is the shortening of the duration of the purification process. The duration of the HGMF purification process described in this study is only 70 min, excluding the loading of the magnetic filter with the magnetic particle/protein suspension. For a particle concentration of 5 g L21 in the feedstock correlating to a total amount of 100 g of magnetic particles, the loading step itself will take 20 min, if for example 20 L of the suspension would be fed at a flow rate of 1 L min21. Therefore, the total duration of the purification process is 90 min. In comparison, the operation time for an

88

Biotechnol. Prog., 2015, Vol. 31, No. 1

Table 7. Comparison of HGMF Using Magnetic Adsorbents (DEAP, 5 g L21) With Conventional Adsorption Chromatography (QFF, 20 mL Packed Bed) for the Recovery of eCG From Serum Step Serum HPO3 ppt UF/DF Filtrate Elution (Q FF) Serum HPO3 ppt UF/DF Supernatant Elution I (DEAP) Elution II (DEAP)

Protein (mg mL1)

eCG (IUELISA mL21)

Volume (mL)

Yield (%)

Purification Factor

Concentration Factor

64 7.24 0.49 0.46 0.29 6 0.09 66 6 6.1 8.5 6 2.1 0.8 6 0.5 0.33 6 0.02 0.6 6 0.08 0.17 6 0.007

43.8 31.2 186 186 247 6 56.6 49 6 8.6 41 6 14 184 6 14 8 6 5.4 774 6 46 120 6 11

303 363 53 46.6 25 6 5 378 441 77 77 15 15

100 85.4 74.5 65.5 45.2 6 1.1 100 90 6 9.7 71 6 9.6

1 6.29 556 592 1,286 6 198 1 6.4 310 6 200

1 0.71 4.25 4.25 5.64 6 1.3 1 0.8 6 0.2 3.7 6 0.8

58 6 9.2 9 6 0.84

1,738 6 138 924 6 49

16 6 0.9 2.5 6 0.23

Figure 5. a: Correlation between particle concentration prevailing during elution in the magnetic filter and the resulting eCG concentration (IUELISA mL21) in the elution fraction; the eCG concentrations in the feedstocks applied are noted above the respective bars. b: Correlation between the product of the particle concentration prevailing during elution in the magnetic filter times the eCG concentration in the feedstocks applied (IUELISA mL21) versus the resulting eCG concentration (IUELISA mL21) in the elution fraction.

equal volume of feedstock purified by conventional chromatography with a column of 100 mL (with a comparable capacity for eCG binding), would be many times higher. Based on a flow rate of 10 CV h21, equals to 1.0 L h21, the duration for the purification would result in an operation time of 20 h for the loading step alone. It has to be stated that the purpose of this benchmarking has been to show that the use of magnetic adsorbents is able to deliver a product stream with a purity and concentration equal or even better to the one of conventional chromatography, while showing clearly superior kinetics. As stated above, we expect that the purification and concentration factor of the product resulting from conventional chromatography could have been improved by using a smaller column or an increased amount of feedstock. However, for the benchmark a limit of the possible duration of the chromatographic run of 24 h has been set, in order to keep it realistic. Parallel to our studies we implemented both process variants in a simplified form in the simulation software SuperProDesigner. Assuming the same yield for both processes, the process using QFF had a payback time of 3 years, while in case of the magnetic bead process 4 years are needed. However, due to uncertainties regarding the particle performance and the maximum number of cycles they can be reused, these absolute numbers only show that the magnetic bead process can be competitive. The simulations showed that the price of the

feedstock (horse serum) is dominating the operating costs. Therefore the achieved overall yield is the most important parameter with respect to the economics. For the magnetic beads a price of 10 e/g was assumed, however, also a price of 20 e/g would not change the results substantially (Table 7). Although the high specific eCG activities of 1951 IUELISA mg21 of the HGMF process in the small scale benchmark may seem to be in contrast to the results obtained at larger scales, it can be shown that they completely fall into our expectations. The reasons for the differences in specific activity of the final eCG product are the varying concentrations of eCG in the used intermediates and the particle concentration prevailing during the elution step. Plotting the resulting specific activity in the eluate versus the prevailing particle concentration during this step shows a clear trend (Figure 5a). However, the small scale HGMF process of the benchmark experiment seems to deviate, by showing the highest specific activity while not having the highest prevailing particle concentration. For explanation the second mentioned criterion, the specific activity of the intermediate used, must be taken into account. A new criterion can be calculated by a simple linear combination of specific activity in the intermediate and particle concentration prevailing during elution. Plotting specific activity in the eluates versus this criterion delivers a surprisingly linear relationship

Biotechnol. Prog., 2015, Vol. 31, No. 1

(Figure 5b), showing that the performance of the HGMF process follows the same rules at largely different scales. It must be stated that the linear correlation only holds true when the particle concentration in the binding step is constant and large enough to almost completely capture the eCG available in the feedstock. However, as long as this is guaranteed, the HGMF process is robust and easily predictable.

Conclusions It could be shown that the developed magnet technologybased separation process is qualified for the purification of the glycoprotein hormone equine chorionic gonadotropin (eCG) from pregnant mare sera in large scale. The process consists of a simplified pretreatment including two precipitation steps removing the majority of unwanted proteins, followed by an ultra/diafiltration step in order to produce an intermediate suitable for ion exchange sorbents. The subsequent HGMF process was conducted in a rotor-stator type magnetic filter, specially designed for biotechnological applications. The magnetic filter was operated at feed rates of 1 L min21 up to particle loadings of 60 g of magnetic microparticles per liter of filter matrix. The system showed only negligible particle losses (1,000 IUOHR mg21. While this shows that regulative thresholds can be met, the economics of the process mainly depend on the price and quality of the magnetic microadsorbents as well as the number of cycles they can be used. An initial estimation is that process yields of 70 to 80% can be achieved in multicycle operation, and particle usage of around 100 cycles. The HGMF process described in this study could be a better alternative for eCG purification than the classical purification procedure of Gospodarowicz and Papkoff used nowadays.10

Acknowledgment The authors thank their industrial partners (ABBIS, INLAB GmbH) of the joint project for the fruitful discussion. The authors also thank PerkinElmer chemagen Technologie GmbH AG in Baesweiler for the preparation of the carboxylated magnetic polyvinyl alcohol beads and Innovent Technologieentwicklung in Jena for the functionalization of the magnetic adsorbents. Funding by the Federal Ministry of Economics and Technology (BMWi) is acknowledged.

89

Literature Cited 1. Allen WR, Stewart F. In: McKinnon AO, Voss JL, editors. Equine Reproduction. Vol. 9. Chichester: Blackwell Publishing Asia; 1993;81-84. 2. Christakos S, Bahl OP. Pregnant mare serum gonadotropin. Purification and physicochemical, biological, and immunological characterization. J Biol Chem. 1979;254:4253-4261. 3. Legardinier S, Cahoreau C, Klett D, Combarnous Y. Involvement of equine chorionic gonadotropin (eCG) carbohydrate side chains in its bioactivity; lessons from recombinant hormone expressed in insect cells. Reprod Nutr Dev. 2005;45:255-259. 4. Virmani M, Gupta AK, Garg SK. Extraction, purification and characterization of equine chorionic gonadotropin (eCG) from pregnant mare’s serum. Indian J Anim Sci. 2003;73:1224-1228. 5. Allen WR. The immunological measurement of pregnant mare serum gonadotropin. J Endocrinol. 1969;43:593-598. 6. Frey HH, L€oscher W. Lehrbuch der Pharmakologie und Toxikologie f€ ur die Veterin€ armedizin. 3rd ed. Stuttgart: Enke Verlag; 2009. 7. Catchpole HR, Cole HH, Pearson PG. Studies of the rate of disappearance and fate of mare gonadotropic hormone following intravenous injection. Am J Physiol. 1935;112:21-22. 8. Menzer C, Schams D. Radioimmunoassay for PMSG and its application to in-vivo studies. J Reprod Fertil. 1979;55:339-343. 9. Meyer A, Berensmeier S, Franzreb F. Direct capture of lactoferrin from whey using magnetic micro-ion exchangers in combination with high-gradient magnetic separation. React Funct Polym. 2007;67:1577-1588. 10. Gospodarowicz D, Papkoff H. A simple method for the isolation of pregnant mare serum gonadotropin. Endocrinology. 1967;80: 699-702. 11. Heeboll-Nielsen A, Dalkiaer M, Hubbuch JJ, Thomas ORT. Superparamagnetic adsorbents for high-gradient magnetic fishing of lectins out of legume extracts. Biotechnol Bioeng. 2004;87: 311-323. 12. Hubbuch JJ, Matthiesen DB, Hobley TJ, Thomas ORT. High gradient magnetic separation versus expanded bed adsorption: a first principle comparison. Bioseparation. 2001;10:99-112. 13. Heeboll-Nielsen A, Justesen SFL, Thomas ORT. Fractionation of whey proteins with high-capacity superparamagnetic ionexchangers. J Biotechnol. 2004;113:247-262. 14. Loo D, Jones A, Hill MM. Lectine magnetic bead array for biomarker discovery. J Proteome Res. 2010;9:5496-5500. 15. Brown GN, M€uller C, Theodosiou E, Franzreb M, Thomas ORT. Multicycle recovery of lactoferrin and lactoperoxidase from crude whey using fimbriated high-capacity magnetic cation exchangers and a novel ‘rotor–stator’ high-gradient magnetic separator. Biotechnol Bioeng. 2013;110:1714-1725. 16. M€uller C, Preußer-Kunze A, Wagner K, Franzreb M. Gonadotropin purification from horse serum applying magntic beads. Biotechnol J. 2011;6:392-395. 17. M€uller C, Wagner K, Frankenfeld K, Franzreb M. Simplified purification of equine chorionic gonadotropin (eCG)-an example of the use of magnetic microadsorbents for the isolation of glycoproteins from serum. Biotechnol Lett. 2011;33:929-936. 18. Lecompte F, Roy F, Combarnous Y. International collaborative calibration of a preparation of equine chorionic gonadotropin (eCG NZY-01) proposed as a new standard. J Reprod Fertil. 1998;113:145-150. 19. Gonzalez G, Castro B, Massaldi H. Extraction of equine chorionic gonadotropin from pregnant mare plasma by direct adsorption on chromatographic media. Biotechnol Bioeng. 1998;57:22-25. 20. Sairam MR. Role of carbohydrates in glycoprotein hormone signal transduction. FASEB J. 1989;3:1915-1926. 21. Pingoud A, Urbanke C. Arbeitsmethoden der Biochemie. Berlin, NY: Walter de Gruyter; 1997. 22. Gomes CSG. Advances in High-Gradient Magnetic Fishing for Bioprocessing. PhD Thesis. Denmark: Technical University of Denmark; 2006. Manuscript received Jun. 11, 2014, and revision received Aug. 9, 2014.