Displacement chromatography is rapidly emerging as a powerful preparative bioseparation technique due to the high throughput and product purity associated with the process (1-4). This technique offers several advantages in preparative chromatography as compared to the traditional elution mode. The displacement process takes advantage of the non-linearity of the isotherms such that a larger feed can be separated on a given column with the purified components recovered at significantly higher concentrations. Furthermore, the tailing observed in preparative elution is greatly reduced in displacement chromatography due to the self-sharpening boundaries developed during the process. This is illustrated in Figure 1, which exaggerates the difference between the elution profiles of the two techniques. In preparative elution the feed components are diluted during the separation, but the feed components can be significantly concentrated during the displacement procedure. These advantages combine to make displacement chromatography an extremely attractive preparative technique for the isolation of biomolecules from the dilute solutions often encountered in biotechnology processes.
In displacement chromatography, a front of displacer solution traveling behind the feed drives the separation of the feed components into adjacent pure zones, which move with the same velocity as the displacer front. The column is first equilibrated with a carrier solution in which the components to be separated have a relatively high affinity for the stationary phase (Figure 2). A feed mixture is then pumped into the column followed by a displacer solution. During the introduction of the feed, the components saturate the stationary phase at the top of the column and frontal chromatography occurs. The displacer is selected such that it has a higher affinity for the stationary phase than any of the feed components (4, 5). As the displacer front traverses the column the feed components are displaced and separated as they compete for the adsorption sites on the stationary phase. Under appropriate conditions, a displacement train is formed containing the feed components as adjacent bands, all moving with the velocity of the displacer front (ud), which can be determined directly from a material balance to be (4, 6):
ud = uo / [1 + F(qd/cd)] (1)where uo is the mobile phase velocity, F is the phase ratio (volume of resin/column void volume), qd and cd are the stationary phase and mobile phase concentration of the displacer, respectively. This isotachic condition is given by the expression:
q1/c1 = q2/c2 = ... = qd/cd (2)The order of the zones corresponds to the increasing affinity of the components for the stationary phase. The concentration of each component in the final displacement train is determined solely by its adsorption isotherm and the concentration and isotherm of the displacer (4, 7) as shown in Figure 3. Figure 3A shows typical isotherms of three feed components and a displacer along with the displacement operating line. The operating line is a chord drawn from the origin to the point on the displacer isotherm, which corresponds to the displacer mobile phase concentration (in this example 20 mg/ml). The operating line actually represents a line of constant chromatographic velocity. The intersection of the operating line with a feed component isotherm indicates (dictates) at what concentration that feed component will be displaced (this is determined through equations 1 and 2). Figure 3B is the square wave displacement train that would result from the combination of isotherms and operating line shown in Figure 3A. The dotted lines drawn from Figure 3A to 3B indicate how the lower figure is constructed. For example, the operating line intersects the isotherm of feed component 1 at approximately 8 mg/ml and thus component 1 would be displaced at a concentration of 8 mg/ml. The width of a displacement zone is determined solely through a mass balance. The higher the feed load of a component, the wider its displacement zone will become. However, regardless of the quantity of a component applied to the column, the concentration of its displacement zone will always obtain the value pre-determined by the operating line/isotherm intersection point (all other operating parameters constant). This figure serves only to demonstrate some of the theory behind displacement chromatography. In practice, it would be rare to measure the adsorption isotherms and use this construction to predict displacement separations. Because the concentration of each displacement zone is determined through these thermodynamic forces, displacement systems can often result in significant concentration of the feed components during the separation process. Upon the emergence of displacer, the column is regenerated by removal of the displacer with an appropriate solvent sequence followed by re-equilibration with the carrier.
Anion-exchange displacement chromatography of proteins has been studied by several investigators. Peterson and coworkers (8-13) have used carboxymethyldextrans as displacers for various protein mixtures. Chondroitin sulfate has been employed by Horvath and coworkers to displace b-galactosidase (14) and b-lactoglobulins (15, 16). Jen and Pinto have performed protein displacements using relatively low molecular weight dextran sulfate (17) and poly(vinylsulfonic acid) (18) as displacers. Ghose and Mattiasson (19) have examined the purification of lactate dehydrogenase using a carboxymethylstarch displacer. Cramer and co-workers (20-25) have examined a variety of cation- and anion-exchange displacement systems.
Displacement versus Traditional Modes
Displacement chromatography achieves separations through a fundamentally different mode as compared to desorption chromatography (e.g., step and linear gradient chromatography). In desorption, the binding affinity of an adsorbed component is dramatically reduced by a change in the equilibrium conditions of adsorption (e.g., changes in pH, salt, organic modifier). With displacement chromatography, the binding affinity of the solute component itself is not significantly altered due to introduction of the displacer. Instead, the displacer, having a higher affinity than any of the feed components, competes effectively for adsorption sites on the stationary phase. The displacement process involves mathematically non-linear competition of solutes for binding sites, while the desorption process involves a relative change in binding affinities of the solutes. The implications of this distinction are quite significant in that displacement chromatography can potentially purify components from mixtures having low separation factors. In the case of desorption chromatography, large separation factors are generally required to give satisfactory resolution.
Experimental
Materials
Displacement experiments were performed using POROS HQ/M (quaternary amine, 20 mm) columns (10.0 mm x 100 mm, 4.6 mm x 100 mm, PerSeptive Biosystems, Inc., Cambridge, MA). Sodium chloride, sodium hydroxide, dithiothreitol, and sodium phosphate were purchased from Sigma (St. Louis, MO). Methanol, ethanol, and acetonitrile were obtained from J.T. Baker (Danvers, MA). Dextran sodium sulfate (40 kDa) was purchased from ICN Biomedicals. Polydiallyl dimethyl ammonium chloride (polyDADMAC) and o-toluidine blue indicator were obtained from Nalco Chemical Company (Naperville, IL). The HPLC systems used in this study were a BioCAD 20 (20 ml/min maximum flowrate) and a BioCAD 60 (PerSeptive Biosystems, Cambridge, MA), which included an Advantec model SF-2120 fraction collector (Advantec Toyo Kaisha, Ltd, Japan). Analysis of displacement fractions was performed on a Beckman P/ACE 5510 capillary electrophoresis system using eCAP ssDNA 100 gel, 47 cm capillaries (Beckman Instruments, Inc., Fullerton, CA). The running buffer of Tris-Borate/7 M urea was also obtained from Beckman.
Methods
Displacement Chromatography. A POROS HQ/M column was sequentially perfused with carrier, feed, displacer, and regenerant solutions. The particular carrier composition, fraction size, flowrate, and feed load for each displacement is listed in the figure legends. The displacer was 7 mg/ml dextran sulfate, and the following volumes were employed: 4.6 mm x 100 mm column, 7 ml; 10 mm x 100 mm column, 35 ml. The regenerant was 3.0 M sodium chloride in 0.5 M sodium hydroxide, and 30 column volumes (CV) (unoptimized volume and composition) were used to assure column regeneration.
Capillary Electrophoresis Analysis.
A capillary cartridge was prepared using an eCAP ssDNA 100 gel-filled capillary, which was trimmed to 47 cm in length. The cartridge was loaded onto the Beckman P/ACE capillary electrophoresis unit, and the capillary ends were placed in Tris-Borate/7 M urea running buffer. Displacement fractions were desalted and diluted with de-ionized water to yield samples with about 5 O.D. units/ml. Voltage injection (7.5 kV) was used to load sample onto the capillary, and injection times were typically 1-3 seconds. The capillary was run at 37 C using a 14.1 kV voltage potential, which yielded an analysis time of 35 minutes. The separation was monitored at 254 nm.
Analysis for Dextran Sulfate. Dextran sulfate concentration was measured using a colloidal titration assay obtained from Nalco Chemical Company. For analysis of dextran sulfate, two drops of o-toluidine indicator were added to 100 ml of distilled water, and the subsequent addition of dextran sulfate produced a colorimetric change. Titrating against polyDADMAC produced another colorimetric change. A linear calibration plot, which was generated using known amounts of dextran sulfate, was used to quantify the unknown concentrations of dextran sulfate in displacement fractions.
Estimation of Oligonucleotide Concentration. Oligonuc-leotide concentrations were estimated from absorbance measurements at 260 nm using a conversion factor of 22 O.D. = 1 mg oligonucleotide. Feed solutions and displacement fractions were diluted in the range of 1:100 to 1:1000 to obtain absorbance readings of less than 1 AU.
Results and Discussion
Displacement Histograms
Displacement histograms are constructed from analytical data obtained for each displacement fraction. A displacement histogram is a discreet representation of a continuous separation process and at first exposure is difficult to read. Figure 4 serves to remove some of the confusion that surrounds displacement histograms. The UV-absorbance trace (grey) is overlaid with the fraction analysis (black). Here it can be seen how a histogram is an approximation of the actual separation.
Development of Oligonucleotide Displacement Separations Using two unrelated 20-base phosphorothioate oligonuc-leotides (20A and 20B) as model systems, approximately thirty displacements were performed to identify optimal separation conditions. These samples were provided to PerSeptive Biosystems as part of collaborative efforts and thus the sequences can not be disclosed. However, both samples are currently in clinical trials and represented real world separation problems. Prior to displacement purification, all the oligonucleotides employed in this report were subjected to reversed-phase (RP) purification and on-column detritylation with 2% TFA. The partially purified oligonucleotides were then lyophilized. The composition of the oligonucleotide following this RP step was typically 88-90% full-length, 8-10% N-1, and 2-4% N-2 and shorter.
The optimized displacement purification of oligonucleotide 20A is presented in Figure 5. In this semi-preparative displacement, 51 mg of crude 20A were purified on a 10 mm x 100 mm POROS HQ/M column. The purity as determined by capillary gel electrophoresis (CGE) analysis is listed above each fraction. This separation has a product yield of 60% at 96% purity, and the mass balance recovery was greater than 95%. Shorter failure sequences (N-2, N-3, N-x) are concentrated into the initial fractions and are effectively eliminated from the full-length product zone. Unfortunately, while N-1 is also significantly concentrated into the initial fractions, N-1 tails into the zone of full-length material at a relatively constant level of about 2 to 3%. Despite this tailing, displacement purification still realizes a high product yield and high product purity that is generally unobtainable using traditional linear gradient techniques.
The mobile phase employed for this displacement was 100 mM NaCl, 10 mM NaOH and 5% (v/v) methanol. The denaturing condition provided by pH 12 (from 10 mM NaOH) makes this mobile phase appropriate for both C- and G-rich sequences. As will be illustrated later, this mobile phase seems to offer nearly "universal" conditions for the displacement purification of phosphorothioate oligonucleotides. The displacement portion of this separation required only 9 minutes to purify 51 mg of oligonucleotide. Total cycle time for the separation, including regeneration (20 column volumes) and re-equilibration (10 column volumes), would be under 30 minutes.
The analytical-scale displacement of the oligonucleotide 20B is shown in Figure 6. Unfortunately, the supply of 20B was exhausted before the optimal conditions were obtained and thus a semi-preparative displacement was never performed. However, the performance of the separation is still remarkable. The product yield was 60% at 96% purity with a mass balance recovery of greater than 95%. Compared to 20A, this sample presented a unique separation problem. Not only were N-1 failure sequences difficult to remove from the product, but "N-2" failures also persisted in contaminating the product. Dithiothreitol (DTT) was utilized as a mobile phase modifier in an attempt to attenuate any sulfur-sulfur interactions that could be occurring between failures and product. The introduction of DTT into the mobile phase had a tremendous effect and resulted in near complete removal of these "N-2" species from the full-length product zone. Unfortunately, DTT did not improve the removal of N-1 failures in either this oligonucleotide or 20A.
At this stage of the optimization, a buffer (Na2HPO4) was employed in place of sodium hydroxide to obtain denaturing conditions (pH 12). When the sample 20A was chromatographed with phosphate buffer instead of sodium hydroxide, the separation performance was identical. Based on experience with 20A, it is doubtful that the higher salt concentration used for 20B (560 vs. 100 mM NaCl) improved the degree of separation. Moreover, the higher salt concentration decreased the effective column capacity for oligonucleotide, such that with 20B the feed load was only 5.5 mg/ml column volumes as compared to 6.5 mg/ml column volumes for 20A. Therefore, it is reasonable to assume that the mobile phase conditions for 20A would also be appropriate for sample 20B and would yield similar results.
Optimization of Displacement Conditions
During the optimization of the displacements of these 20-base oligonucleotides, N-1 persistently tailed into the zone of full-length product. Regardless of experimental conditions, the N-1 content could not be reduced to less than 2%. A multitude of mobile phase compositions, summarized in Table 1, were employed to alleviate this tailing with no success. High mobile phase salt concentrations were utilized to increase the rate of adsorption/desorption kinetics on the hypothesis that the N-1 tailing was due to slow kinetics. No improvements in purity were noted at elevated salt concentrations. A displacement was also attempted at elevated temperature (60 C) in order to enhance the rate of adsorption/desorption. Unfortunately, the N-1 tailing remained relatively constant even at 60 C indicating that the tailing was probably not due to slow adsorption/desorption kinetics. The remaining mobile phase conditions were employed in an attempt to minimize any secondary structure or aggregation effects that may occur in oligonucleotides. Other than the previously mentioned effect of DTT, these conditions had no effect on the separation performance. In fact, pure fractions (about 96% purity) from a displacement were pooled and re-displaced with only marginal improvements (about 0.5%) in product purity. Thus, the N-1 failures that contaminated the product zone were essentially chromatographically identical to the full length product.
Table 1: Mobile Phase Modifications Attempted During 20-mer Optimization Displacements
7 M urea organic modifier (0-20%) guanidine-HCl 5% acetonitrile modifier high salt (up to about 1M) 5% methanol modifier dithiothreitol 5% ethanol modifier 60 C
The only major improvement that occurred in the displacement protocol was the addition of organic modifier to the mobile phase. Displacements without organic modifier typically exhibited displacement mass balance recoveries of only 70-80%. The remaining 20-30% of oligonucleotide was recovered from the column during the regeneration cycle. With the addition of 5% (v/v) organic modifier to the mobile phase, displacement mass balance recoveries exceeded 95%. It is interesting to note that organic modifier had no effect on the purity of displaced oligonucleotide. With or without organic modifier, the oligonucleotide that was displaced generally obtained purity levels in the range of 96-98%. The mode of action of organic modifier seems to be the elimination of non-specific (i.e., hydrophobic) interactions between oligonucleotide and the adsorbent surface. With these non-specific interactions minimized, the oligonucleotide, binding primarily in an electrostatic mode, can be efficiently and completely displaced. Without organic modifier, non-specific interactions seem to interfere with the displacement process leading to less than complete displacement of oligonucleotide from the column. Increased organic modifier content (10 and 20%) and selection of organic (methanol, ethanol and acetonitrile) had no effect on separation performance.
Additional Oligonucleotide Displacements
In a second collaboration, we were asked to purify two phosphorothioate oligonucleotides (an 18-mer and a 25-mer) from 100 mmole scale syntheses. Because the sample size was limited, optimization of the displacement technique was not performed because we wanted to maximize product recovery. Optimization displacements on the analytical-scale actually yield little product because of the amounts needed for subsequent analytical techniques: for each 200 ml displacement fraction, 10 ml is used for concentration determination, 50-70 ml for displacer analysis, and 100 ml for CGE analysis. Thus, one analytical-scale displacement was performed for each sample using conditions optimized for the oligonucleotide 20A. Because these preliminary displacements were promising, the separations were scaled to semi-preparative displacements (about 50 mg on a 10 mm x 100 mm POROS HQ/M column). The semi-preparative displacements for the 25-mer and the 18-mer are shown in Figures 7 and 8, respectively. The 18-mer had a yield of 55% at 95% purity while the 25-mer had a slightly lower yield of 45% at 96% purity. The yield of the 25-mer is lower because the dextran sulfate displacer does not have sufficient affinity to efficiently displace a 25-mer. Methods for increasing displacer affinity are currently being investigated, to improve displacement performance for long oligonucleotides.
Final Purification
Following displacement chromatography, polishing of the oligonucleotide product is achieved through a desalting step. From CGE and displacer analysis, fractions meeting the purity constraints are pooled and desalted using an appropriate method. Fractions contaminated with displacer may be included with the product pool provided that the desalting step removes dextran sulfate. Preliminary experiments indicate that a reversed-phase (RP) desalting step may result in significant removal of displacer from the oligonucleotide. Unfortunately, the RP desalting step does not increase the purity of the oligonucleotide product.
Application of Displacement Chromatography to Phosphodiester Oligonucleotides
While we have not yet used displacement chromatography to purify phosphodiester oligonucleotides, the technique should be able to purify them efficiently without major modifications. Most likely, an adjustment to a pH more appropriate for phosphodiester oligonucleotides would be the only change needed. Displacer and mobile phase salt concentration would probably be very similar to those used for phosphorothioate oligonucleotides.
Displacer Selection and Contamination
Displacer selection has not yet been reduced to an exact science, but there are several outstanding candidate displacers in the literature from which to choose. Most of our experience has been with heparin and dextran sulfate in anion-exchange systems. We suggest these compounds as displacers for a wide range of biomolecules.
Displacer contamination of the product may occur, which necessitates that fractions be assayed for the displacer. The displacer may contaminate a portion (about 3-5%) of the product, which may be discarded to avoid complications with FDA approval of therapeutic products. Displacement purification significantly improves product yield such that, even with a 5% loss due to displacer contamination, the overall product recovery will probably exceed that obtained using traditional techniques. Moreover, separation processes following displacement purification could be designed to eliminate displacer from the product (e.g., by following an anion-exchange displacement with a cation-exchange polishing step). To aid in displacer selection and eliminate contamination as an issue, researchers are currently developing low molecular weight displacers, displacers with secondary affinity to facilitate removal, and non-toxic displacer compounds (26-28).
Conclusions Displacement chromatography appears to offer the potential of nearly universal conditions for the purification of phosphorothioate oligonucleotides. Mobile phase conditions of 100 mM NaCl, 10 mM NaOH, 5% methanol seem to be an appropriate starting point for the displacement purification of any phosphorothioate oligonucleotide. Dextran sulfate (40 kDa) has been shown to be a very efficient displacer for oligonucleotides up to 20 bases long, but displacement performance declines slightly when the method is used to purify longer oligonucleotides. However, the product yield and purity are still superior to those obtained using traditional modes of chromatography.
The primary advantage of displacement chromatography as a separation method is an increased yield of highly pure product. Displacement separations generate pure product (96-98%) at 50-60% yield whereas traditional techniques obtain similar purity but at significantly lower yields (30-40%).
The improved product yield and apparent "universality" of separation conditions make displacement chromatography attractive for semi-preparative through preparative-scale purification of phosphorothioate oligonucleotides. In a separate publication, the preparative displacement purification of gram quantities of a 24-mer phosphorothioate oligonucleotide will be discussed (29).
References
Feed (or Feed Component): A general term used to describe the compounds to be purified.
Displacer: A molecule that has a higher affinity for the adsorbent phase than any of the feed components. For ion-exchange displacement chromatography, displacers are typically polyelectrolytes (10-50 kDa).
Displacer Front: As a column is perfused with displacer, a boundary will develop within the column. On the downstream side of this boundary there will be only feed components. The upstream side of the boundary will contain only displacer. This boundary is deemed the displacer front.
Displacement Train: The series of square-wave zones of pure feed components that form adjacent to the displacer front.
Isotachic Condition: Each square-wave zone in the displacement train is moving at the same chromatographic velocity as the displacer front.
Isotherm: An isotherm describes the equilibrium relationship between the mobile phase concentration (c) of a component and its corresponding stationary phase concentration (q).
Frontal Chromatography: With high feed loads, the introduction of a multi-component feed mixture to the column will result in partial resolution of the components. The highest affinity feed component preferentially binds to the adsorbent phase pushing lower affinity species ahead in the column. The process continues with the next highest affinity component until a zone of the lowest affinity species forms at the very front of the feed slug. However, all the feed components exhibit a high degree of tailing, such that complete resolution of the feed components is not obtained. This phenomena is termed Frontal Chromatography and is also described as Sample Self Displacement. The introduction of a displacer eliminates this tailing resulting in full resolution of the feed components into adjacent zones of pure material.
Return to the The ABRF Home Page