This article presents the multicolumn countercurrent solvent gradient purification (MCSGP) process, which uses three chromatographic columns, and incorporates the principle of countercurrent operation and the possibility of using solvent gradients. A MCSGP prototype has been built using commercial chromatographic equipment. The application of this prototype for purifying a MAb from a clarified cell culture supernatant using only a commercial, preparative cation exchange resin shows that the MCSGP process can result in purities and yields comparable to those of purification using Protein A.
This article presents the multicolumn countercurrent solvent gradient purification (MCSGP) process, which uses three chromatographic columns, and incorporates the principle of countercurrent operation and the possibility of using solvent gradients. A MCSGP prototype has been built using commercial chromatographic equipment. The application of this prototype for purifying a MAb from a clarified cell culture supernatant using only a commercial, preparative cation exchange resin shows that the MCSGP process can result in purities and yields comparable to those of purification using Protein A. The second application example for the MCSGP prototype is the separation of three MAb variants using a preparative weak cation-exchange resin. Although the intermediately eluting MAb variant can only be obtained with 80% purity at recoveries close to zero in a batch chromatographic process, the MCSGP process can provide 90% purity at 93% yield. A numerical comparison of the MCSGP process with the batch chromatographic process, and a batch chromatographic process including ideal recycling, has been performed using an industrial polypeptide purification as the model system. It shows that the MCSGP process can increase the productivity by a factor of 10 and reduce the solvent requirement by 90%.
The increasing production volumes of biomolecules, especially therapeutic proteins, and a rising cost pressure from the market, has engendered a strong interest in the chromatographic purification step in the downstream processing of biomolecules.1–2 This is because the chromatography step is often irreplaceable and also a major cost driver in downstream downstream.
The generic purification problem to be solved, particularly in the area of chromatographic bioseparations, is a separation of the feedstock into three fractions: the early eluting, weakly adsorbing impurities; the desired product(s), and the late eluting, strongly adsorbing impurities. A literature overview of chromatographic three-fraction separation processes can be found elsewhere.3 Conventionally, the generic three-fraction purification problem is solved using batch column chromatography, often incorporating a linear variation of the eluting composition with respect to the elution time (also called solvent gradient chromatography). The column effluent is collected in several consecutive, small portions and some of these portions ideally contain the desired product with an average purity being higher than the required one, thereby making up the product fraction. To increase the yield of solvent gradient chromatography, product-rich portions collected from the column effluent, which do not fulfill the purity specifications, can be recycled to the feed point. The advantages of such a recycle have been discussed elsewhere in detail.7
Besides batch solvent gradient chromatography and its derivative including a recycle, the multicolumn countercurrent solvent gradient purification (MCSGP) process is a third suitable process developed recently to perform three-fraction separations of biomolecules.3–5 This process involves three columns and continuously purifies a complex feed mixture into three fractions, thereby incorporating solvent gradients.
Conventional multicolumn processes such as simulated-moving-bed (SMB) chromatography cannot be used for the purification of biomolecules, because an SMB can only perform a binary separation and solvent gradients can not be implemented properly. SMB chromatography is therefore more suitable for a binary separation of small molecules.
This work explains the principle of the MCSGP process and shows experimental results from the MCSGP prototype for purifying a MAb from clarified cell culture supernatant (cCCS) without using protein A resin, and a separation of three MAb variants.
A numerical comparison between the conventional batch chromatography process and the MCSGP process with respect to yield and process productivity is also performed for a polypeptide purification. The Pareto curves obtained from the optimization procedure are compared and the performance superiority of the MCSGP process is explained.
The generic problem in the chromatographic purification of biomolecules can be simplified to the chromatogram as shown in Figure 1.
Figure 1. Simplified chromatogram of generic chromatographic purification problem (time axis from right to left); S: strongly adsorbing, late eluting impurities; P: product; W: weakly adsorbing, early eluting impurities.
To facilitate the following explanation, the time axis of the chromatogram (Figure 1) has been reversed (i.e., it goes from right to left). The generic chromatogram can be cut into five fractions as indicated by the numbers on the time axis:
1 = strongly adsorbing impurities
2 = product contaminated by strongly adsorbing impurities
3 = product
4 = product contaminated with weakly adsorbing impurities
5 = weakly adsorbing impurities.
The aim of an ideal purification process would be to collect fraction 3, to drain fractions 1 and 5, and to recycle fractions 2 and 4. Transferring these tasks to a multicolumn chromatographic process results in a 6-column flowsheet (Figure 2), which shows the basic principle of the MCSGP process.
Figure 2 shows six columns. Each column is connected to a gradient pump. The outlet of columns 2 and 4 is recycled into columns 4 and 6, respectively. Fractions 1 to 5 as shown in the chromatogram in Figure 1 are eluted from a column with the same number (Figure 2). For example, the strongly adsorbing impurities (i.e., fraction 1 from Figure 1), are drained from column 1. In column 2, the product fraction contaminated with strongly adsorbing impurities, P+S (i.e., fraction 2) is eluted, mixed with the outlet of the gradient pump, and recycled to column 4. In column 3, the product is collected. The outlet of column 4, P+W, is recycled to column 6. In column 5, the raw feed mixture is fed to the column inlet and the weakly adsorbing impurities are eluted. In order to run this process continuously, after a certain time all columns are switched one position to the left (e.g., column 4 becomes column 3), and column 1, being void of any solutes at the end of the switch time, can be switched to the position of column 6.3–4 This column switching results in a countercurrent movement of liquid and solid flow. This countercurrent operation improves the efficiency of the process.
Figure 2. Basic principle of the MCSGP process
To reduce equipment costs, the MCSGP process used with just three columns. This reduction from six to three columns can be achieved by performing the draining or collecting of S, P, and W and the recycling of P+S and P+W, not in parallel (Figure 2), but in two sequential steps. This results in two separate flowsheets, which are applied to the three columns in an alternating fashion (Figure 3).
Figure 3. Flowsheet of the MCSGP process
After the columns have been switched one position to the left (where column 1 is switched to the position of column 3), the columns are operated in the countercurrent mode (i.e., recycling of P+S and P+W into the subsequent column takes place). After a certain time, tCC, the columns are connected in a different manner and then operated for a time, tB, in the batch mode (i.e., while the draining of S and W and the collecting of P take place). Then the columns are switched again and operate in the countercurrent mode. The column switching plus countercurrent and batch operation mode makes one cycle of time (tCC + tB) and this cycle is repeated.
Two prototypes of the MCSGP process have been built using commercially available chromatographic equipment from GE Healthcare. In order to realize the process (Figure 3), the following parts from the "Aekta Basic" series have been used for each prototype: three double pumps, one UV detector, one conductivity and pH sensor, and six multiposition valves.4 The system is controlled by Unicorn software. The prototype also includes the possibility of performing cleaning-in-place.
Purification of a MAb from cCCS using cation-exchange resins
The use of MAbs as therapeutic proteins is one of the strongest growing industry segments. Currently, only a few MAbs are on the market, but hundreds are in the pipelines. Because of the increasing production volumes, chromatographic downstream processing using key purification step Protein A affinity resins is becoming a bottleneck from an economic and engineering point of view.
Purifying MAbs from cCCS without using Protein A resin is difficult if only single-column chromatographic processes are used. Because the MCSGP process provides a very high separation efficiency, resins which are less specific than Protein A can be used.
Figure 4. SEC chromatogram (280 nm) of cCCS (black curve) and pure MAb standard (red line)
The aim of this work was to purify a MAb from a cCCS using the MCSGP process with three columns filled with a commercial, preparative cation-exchange resin (Merck FractoGel SO3-HiCap, M). The supernatant had a low titer of about 0.01 g MAb per liter cCCS and a feed purity of <1% at 280 nm. Figure 4 shows the SEC chromatogram of the cCCS (black line).
The impurities by far exceed the MAb, as revealed by the comparison with the SEC chromatogram of the pure MAb (red line in Figure 4). Using the MCSGP process (Figure 3), the MAb, being an intermediately eluting product on a cation-exchange resin, can be obtained with high purity from column 2 in the batch mode. The MAb collected from the MCSGP process has been analyzed on a SDS-PAGE and shows comparable purity levels to those of the MAb purified from the same cCCS with Protein A (Figure 5). The recovery of the MAb in the MCSGP process for the shown SDS-PAGE is 95%.
Figure 5. SDS-PAGE of MAb from cCCS purified by the MCSGP process and by Protein A
The results of comparable purities of the MCSGP-purified and the Protein A–purified MAb have been double checked with an SEC analysis (Figure 6).
Figure 6. SEC Chromatogram of Protein Aâpurified MAb (black line) and MCSGP-purified MAb (red line)
From the results above, we can conclude that the highly specific, but expensive Protein A purification step in the downstream processing of MAbs (together with the following batch column cation exchange step), could be replaced by the three-column MCSGP process using comparably cheap cation-exchange resins.
Separation of three monoclonal antibody variants
To verify the high separation efficiency of the MCSGP process, the separation of three variants of a MAb has been used as model system. The three MAb variants contain none, one, and two additional lysine groups at the C-terminal and cannot be separated with sufficient yield on a preparative resin in a single batch column (Figure 7).
Figure 7. Gradient elution of variants with a shallow gradient on a single column (Merck Fractogel EMD COO, 30m). Dashed line: summed concentration curve; markers: experimental data from offline analysis of fractions; solid lines: simulation; blue squares: MAb variant F1; red triangles: MAb variant F2; green circles: MAb variant F3.
Figure 7 shows that the maximum purity of the intermediately eluting MAb variant F2 at 0% yield is approximately 80%. If instead the MCSGP process is used with the same resin (Fractogel EMD COO, 30m), the yield can be increased to 93% at an even higher purity (of about 90%) of the MAb variant F2 . Figure 8 compares the analytical chromatograms of the MAb variant F2 purified with the MCSGP process and the feed mixture.
Figure 8. Analytical chromatograms of MAb variants. Blue: feed mixture; red: MAb variant F2 purified with MCSGP process
A highly pure fraction of the intermediately eluting MAb variant F2 can be obtained using the MCSGP process. If instead only a single-column process is used, it is impossible to obtain a yield of 93% at a purity of 90% (Figure 7).
To quantify the performance gain of the MCSGP process, the process has been compared numerically with the following two conventional chromatographic processes for the purification of biomolecules:
Figure 9. Multilinear solvent gradient
The purification of a polypeptide using reversed-phase chromatography has been used as a model system. For this purification problem, a detailed study has been performed to develop a suitable isotherm and to determine the necessary model parameters.4
Although conventionally linear solvent gradients are applied, recent publications have shown the advantage of nonlinear6 or multilinear gradients.7 In this work, a four-step multilinear solvent gradient (Figure 9) is used for the batch and the ideal recycle process.
To improve the process performance of solvent gradient batch chromatography, product-rich fractions can be recycled to the feed point. Figure 10 shows the flowsheet for ideal recycle process.
Figure 10. Batch chromatography with ideal recycling
The details of this process, including the fractionation methods and the product fraction, have been explained in detail elsewhere.7 This process is ideal because it is assumed that the perfect cut of the chromatogram can always be made. Therefore, all implemented recycling strategies will lead to a less than ideal performance.
The performance parameters investigated in this work have been defined as follows:
For all three processes analyzed, it has been assumed that the feed purity is 60% and the required product purity is 90%. The columns have a comparably low separation efficiency because in industrial applications preparative resins with larger particle diameters are used. Therefore, the modeling of the columns has been performed with only 5 theoretical plates per cm of column length. For the batch column processes, one column with the dimension 30 x 0.46 cm has been assumed for the MCSGP process instead of 3 columns, each 10 x 0.46 cm, so that the resin volume in all processes is the same. Additionally, a pressure drop constraint of the chromatographic columns has been taken into account for all processes.
Each process has been optimized separately, where the operation parameters, for example, column load and gradient shape, have been adjusted, so that yield and productivity are optimized for the given product purity of 90%. This optimization problem results in the Pareto curves (Figure 11).
Figure 11. Comparison of the MCSGP process with conventional single-column technologies with respect to yield and productivity for a product purity >90%
The Pareto curve of the batch process shows that the maximum yield that can be obtained is about 97%. The curve clearly shows that an increase in the productivity has to be paid for by a lower yield (e.g., reducing the yield from 80% to 60% approximately doubles the productivity from 0.0012 to 0.0027 g/min/L). Figure 11 shows that by using the ideal recycling process, yield can be pushed to 100%. This matches the purpose of recycling (i.e., increasing the yield by re-using product-rich fractions). The ideal recycling process also can improve productivity compared to the batch process (e.g., at 85% yield by a factor of 5, from 0.00082 to 0.0042 g/min/L). Productivity increases strongly for the MCSGP process with respect to the ideal recycling process (e.g., at 95% yield by a factor of 13, from 0.0025 to 0.034 g/min/L). The recycling is assumed to be ideal, hence practical recycling strategies will result in the Pareto curves falling below the one of ideal recycling.
An increase in the productivity by a factor of 13 would result in a resin volume of the MCSGP process being 13 times smaller than for the ideal recycling process if the same amount of product is produced per unit time. This also means that 13 times more product can be purified on the same resin volume per unit time or that the same amount on the same resin volume can be purified 13 times faster.
The results of Figure 11 also can be plotted in terms of the solvent requirement (Figure 12).
Figure 12. Comparison of the MCSGP-process with conventional single column technologies with respect to yield and solvent requirement for a product purity > 90%
At 95% yield, the solvent requirement per unit product purified for the batch process is about 3,000 L solvent per purified product. Using the ideal recycle process, the solvent requirement can be reduced by a factor of 37 to 80 L/g. Using the MCSGP process, even more solvent can be saved, and compared to the ideal recycling process, the solvent requirement can be reduced by a factor of 10 to 8 L/g.
The clear superiority of the MCSGP process with respect to yield, productivity, and solvent requirement compared to the conventional purification processes result from the principle of countercurrent operation, which strongly improves the process efficiency, because only a partial separation of the fraction is sufficient to get high yields and purities.
The performance increase as described above by using the MCSGP process has been encountered previously for binary separations of small molecules when the batch process has been compared to the countercurrent SMB process.8
The multicolumn countercurrent solvent gradient purification process for the purification of therapeutic proteins has been presented and the basic principles have been explained. The practical realization in a prototype using only three chromatographic columns has been done using commercially available chromatographic equipment. The successful application to two industrial purification problems is discussed. The first is the purification of a MAb from a clarified cell culture supernatant without using Protein A. Although the feed purity of the supernatant is very low, the experimental results show that the MCSGP process operated with a preparative cation-exchange resin can obtain purities with 95% recovery comparable to a Protein A purification. These results underline the potential of the MCSGP process to replace the Protein A purification step in the downstream processing of MAbs. Second, the separation of three MAb variants on a commercial, preparative weak cation-exchange resin was investigated. The use of batch chromatography for the purification of the intermediately eluting MAb variant yields only 80% purity at zero recovery, while the MCSGP process gives 90% purity and 93% yield. This application example clearly shows that the use of continuous, countercurrent processes such as the MCSGP process can strongly increase the separation efficiency. To evaluate the performance gain of the MCSGP process quantitatively with respect to the conventional processes (i.e., batch chromatography and batch chromatography with recycling), a numerical comparison using an industrial polypeptide purification problem has been performed. This analysis shows that the MCSGP process can increase the productivity by a factor of 10 and reduce the solvent requirement by 90%.
GUIDO STRÖHLEIN, LARS AUMANN, PhD, THOMAS MÜLLER-SPÄTH, ABHIJIT TARAFDER, PhD, and PROF. MASSIMO MORBIDELLI work at the Institute for Chemical and Bioengineering at the Swiss Federal Institute of Technology (ETH) Zurich, Switzerland, +41 44 633 45 26, guido.stroehlein@chem.ethz.ch
1. Nineteenth International symposium on preparative/process chromatography, ion exchange, adsorption/desorption processes and related separation techniques, PREP. Baltimore, USA; 2006, May 14–17.
2. 25th International symposium and exhibit on the separation of proteins, peptides and polynucleotides, ISPPP. Florida, USA; 2005, Nov 6–9.
3. Ströhlein G, Aumann L, Mazzotti M, Morbidelli M. A continuous, countercurrent multicolumn chromatographic process incorporating modifier gradients for ternary separations. J Chrom A, 2006; 1126(1–2):338–346.
4. Aumann L, Morbidelli M. A continuous multicolumn countercurrent solvent gradient purification process. Submitted to Biotechnology and Bioengineering.
5. Aumann L, Morbidelli M. European Patent, EP 05405327.7, EP 05405421.8; 2005.
6. Shan Y, Seidel-Morgenstern A. Optimization of gradient elution conditions in multicomponent preparative liquid chromatography, J Chrom A, 2005; 1093(1–2):47–58.
7. Tarafder A, Ströhlein G, Aumann L, Morbidelli M. Role of recycling in improving the performance of chromatographic solvent gradient bioseparations. International symposium on preparative and industrial chromatography and allied techniques, SPICA. Innsbruck, Austria; 2006, Oct 15–18.
8. Miller L, Grill C, Yan T, Dapremont O, Huthmann E, Juza M. Batch and simulated moving bed chromatographic resolution of a pharmaceutical racemate. J Chrom A, 2003; 1006(1–2):267–280.