Anal. Chem. 2004, 76, 5816-5822
Effect of Spacer Arm Length on Protein Retention on a Strong Cation Exchange Adsorbent Peter DePhillips
Merck Research Laboratories, Sumneytown Road, West Point, Pennsylvania 19486, and Department of Chemistry and Biochemistry, University of Delaware, Newark, Delaware 19716 Inger Lagerlund and Johan Fa 1 renmark
Amersham Biosciences AB, GE Healthcare, Bjo¨rkgatan 30, SE-751 84 Uppsala, Sweden Abraham M. Lenhoff*
Department of Chemical Engineering, University of Delaware, Newark, Delaware 19716
The retention of five proteins was compared on a set of three strong cation exchange adsorbents that differed in spacer arm chemical structure and length. The adsorbents included a commercial product, Amersham Biosciences SP Sepharose Fast Flow, containing a six-carbon spacer between the agarose matrix and the anionic ligand, and two custom-prepared materials. One of the custom adsorbents contained a spacer of about half the length of the SP Sepharose Fast Flow, and the other contained no spacer arm. The adsorbent with no spacer arm was found to be significantly more retentive for all of the test proteins examined, in both isocratic and gradient elution tests. Reducing the spacer arm length by half resulted in increased retention for four of the five proteins, but this increase was less than what was observed when the spacer arm was eliminated. Retention increases were obtained without increasing the density of the anionic charge groups and appear to result from an enhancement of electrostatic or secondary nonelectrostatic interactions, or both. The results indicate that spacer arm length may be a useful variable in manipulating stationary-phase retention properties. A wide variety of adsorbents, differing in manufacturer, base matrix, physical structure, coupling chemistry, and other structural parameters, are available for chromatographic separations. This remains true even when the discussion is focused on specific classes of separations, such as preparative cation exchange chromatography of proteins, which is what we deal with here. For this application, there are numerous media that are all intended for use in the same kinds of applications, but they display substantial ranges of performance in terms of such measures as retention, selectivity, capacity, and pressure drop. Determining the optimal stationary phase for a given separation thus requires, in principle, experimentally evaluating each of the materials for * To whom correspondence should be addressed. Fax +1-302-831-4466. E-mail
[email protected].
5816 Analytical Chemistry, Vol. 76, No. 19, October 1, 2004
the separation of interest. A preferred alternative is to develop a better understanding of the relationship between stationary-phase physical and chemical structural properties and the various performance measures of interest, to make performance more predictable once structural characterization has been undertaken. This approach would also facilitate stationary-phase design efforts, perhaps to the point of custom-synthesizing media for individual applications. We began such an effort in previous studies1,2 in which the retention of three test proteins was examined over a chemically diverse set of strong and weak cation exchangers, and key physical characteristics of this adsorbent set, based on the pore size distribution (PSD), were characterized by inverse size exclusion chromatography. This provided a data set from which the experimentally determined relative protein retention could be correlated to the chemical and physical properties of the adsorbents examined. Key determinants of protein retention for this adsorbent and protein set were found to be the adsorbent PSD and the anion type. Adsorbents with significant pore volume contained in pores of a diameter approximately the same as the solute size showed stronger protein retention, as did strong cation exchange SCX adsorbents (sulfate or sulfonate anion) relative to the weak (WCX) versions (carboxylate anion). An additional structural parameter that appeared to correlate with retention became apparent for the carbohydrate-based subset of strong cation exchangers examined, protein retention being greatest for those that lacked a significant spacer arm between the adsorbent and the anionic group. This observation is examined further here, in a more systematic fashion. The effect of spacer arm chemical structure on retention and binding affinity has previously been studied in detail for bioaffinity chromatography systems, where hydrocarbon spacer arms between the support and bioligand are a manifestation both of using a linker moiety activated on both ends and of a desire to improve the steric (1) DePhillips, P.; Lenhoff, A. M. J. Chromatogr., A 2000, 883, 39-54. (2) DePhillips, P.; Lenhoff, A. M. J. Chromatogr., A 2001, 93, 57-72. 10.1021/ac049462b CCC: $27.50
© 2004 American Chemical Society Published on Web 09/02/2004
Figure 1. Structures of the anionic charge groups and spacer arms derivatized to Sepharose and used in this study.
accessibility of the immobilized ligand. The spacer arm length has been found to have a direct effect on the binding and recovery of a variety of enzymes and carbohydrates3-9 in affinity chromatography separations by providing a likely site for hydrophobic and nonspecific adsorption. The addition of this hydrophobic component to adsorption has generally been regarded as unwanted or deleterious in bioaffinity systems, as the specificity of these mixed mode or “complex” affinity systems is reduced3,8,9. Attenuation of these unwanted hydrophobic interactions is typically achieved by modification of the spacer arm length, hydrophobicity, or both, with shorter or more hydrophilic spacer arms decreasing or eliminating these nonspecific hydrophobic interactions.3,5,7-9 The desire to obtain high selectivity in bioaffinity chromatography systems has made consideration of the chemical structure and the length of spacer arms an important aspect of the overall development of these separations. The present work examines the effect of spacer arm length and chemical structure on protein retention in cation exchange chromatography. For a set of custom-made agarose-based cation exchange adsorbents that differed in the length of the hydrocarbon spacer arm, measurements were made of the relative isocratic retention of a set of three model proteins. For the same adsorbents, the chromatographic behavior of five standard proteins was also studied in gradient elution using standard functional test methods for cation exchangers. EXPERIMENTAL SECTION Chromatographic Stationary Phases. Three strong cation exchange (SCX) adsorbents were used for this study. One of these, SP Sepharose Fast Flow, is a standard commercial product (from Amersham Biosciences AB, Uppsala, Sweden), having a chemical structure with a six-carbon spacer between the ligand (3) O′Carra, P.; Barry, S.; Griffin, T. FEBS Lett. 1974, 43, 169-175. (4) O′Flaherty, M.; McMahon, M.; Mulcahy, P. Protein Expression Purif. 1999, 15, 127-145. (5) Massoulie´, J.; Bon, S. Eur. J. Biochem. 1976, 68, 531-539. (6) Hirota, K.; Shimamura, M. J. Chromatogr. 1985, 319, 173-185. (7) McCourt, P. A. G.; Gustafson, S. Int. J. Biochem. Cell Biol. 1997, 29, 11791189. (8) Lowe, C. R. Eur. J. Biochem. 1977, 73, 265-274. (9) Barry, S.; O'Carra, P. Biochem. J. 1973, 135, 595-607.
and the matrix (Figure 1). The other two were custom-made for this study. One of them had a spacer length of about half the length of the standard SP Sepharose Fast Flow (Figure 1), through derivatization with allyl bromide and subsequent addition of bisulfite to the allylic group. The third derivative was prepared by reacting the Sepharose Fast Flow matrix with chlorosulfonic acid, giving an adsorbent having no spacer at all and with a sulfate ligand instead of a sulfonate (Figure 1). The physical properties of these adsorbents are shown in Table 1. The following abbreviations are used throughout to designate these adsorbents: S6C for the standard SP Sepharose Fast Flow, S3C for the half-length spacer arm version, and S0C for the no spacer arm version. Protein Samples and Procedures. The proteins used for the isocratic experiments, lysozyme (chicken egg white), R-chymotrypsinogen A (bovine pancreas) (aCT), and cytochrome c (bovine heart), were purchased from Sigma Chemical Co. (St. Louis, MO), and were used as received. The protein size and charge properties are summarized elsewhere.2 Protein solutions were prepared by dissolving 10 mg of protein/mL of 10 mM sodium phosphate at a pH of 7, and the resultant mixture was filtered through Millipore Millex-GV 0.22-µm filters (Bedford, MA). Of the proteins for the gradient elution chromatographic functional tests, β-lactoglobulin, ribonuclease A, apotransferrin, and cytochrome c were obtained from Sigma-Aldrich Sweden (Stockholm, Sweden), and wheat germ lectin was from Amersham Biosciences AB (Uppsala, Sweden). Determination of Ionic Capacity of the Cationic Adsorbents. Each gel was packed in a HR 10/10 column for 5 min at 8 mL/min and the bed height adjusted to ∼6 cm. A filter and an adaptor were attached to the column, and the packing was continued for another 5 min. The exact bed height was measured, and the packing procedure was finalized with 5 min of flow at 4 mL/min. To determine the ionic capacity, 50 mL of 1.0 M hydrochloric acid was run through the column at 4 mL/min, after which the column was washed with 50 mL of Milli-Q water. The protons were desorbed into a titration vessel using 48 mL of 1.0 M potassium nitrate and the eluate was titrated with 0.1 M sodium hydroxide. The ionic capacity was calculated from the column volume and the volume of sodium hydroxide at the equivalence point, with the values determined for the adsorbents used given in Table 1. Chromatographic Methods. Both isocratic and gradient retention measurements were performed on different sets of proteins. For the isocratic measurements, retention times were obtained for each protein and adsorbent pair at varying NaCl concentrations in 10 mM sodium phosphate buffer at pH 7. The column void time was determined by using the retention time of the protein at high NaCl concentrations (1 or 2 M). Other aspects of the adsorbent preparation, column packing, and isocratic chromatography procedures were identical to those described in greater detail previously.2 Two sets of gradient chromatographic functional tests were performed. For the first, at pH 4.0, the three adsorbents were packed in HR 16/10 columns at a flow rate of 8.5 mL/min for ∼5 min. Each column was attached to an FPLC system (Amersham Biosciences AB) with UV and conductivity monitors connected Analytical Chemistry, Vol. 76, No. 19, October 1, 2004
5817
Table 1. Physical Properties of the Stationary Phases
stationary phase SP Sepharose FF (6C) half-length spacer (3C) no spacer arm (0C)
spacer arm composition
base matrix
dp (µm)
ion exchange capacity (µmol/mL)
3-(2-hydroxypropoxy)1-propanesulfonic acid propanesulfonic acid sulfate
agarose
45-165
180
agarose agarose
45-165 45-165
180 77
at the outlet. The column was equilibrated using 0.02 M formic acid, pH 4.0 (buffer A), at a flow rate of 5 mL/min. A 0.5-mL sample of a solution of wheat germ lectin (10 mg/mL) and β-lactoglobulin (10 mg/mL) in buffer A was applied, and desorption was performed using a 200-mL linear salt gradient to 100% buffer B (buffer A + 0.75 M lithium chloride) followed by 50 mL of buffer B. The elution volumes for the three peaks of wheat germ lectin and one for β-lactoglobulin were recorded. For the second set of gradient functional tests, at pH 6.0, the adsorbents S6C and S0C were packed into small HiTrap columns (1 mL of gel), which were attached to an A¨ KTA Explorer PS1 system (Amersham Biosciences AB) equipped with UV and conductivity monitors. The columns were equilibrated using buffer A, comprising 0.03 M sodium hydrogen phosphate, 0.03 M sodium formate, and 0.03 M sodium acetate, pH 6.0, at a flow rate of 2 mL/min. The sample was applied as two separate sample solutions, sample 1 containing 10 mg/mL each of ribonuclease A and cytochrome c in buffer A, and sample 2 containing 10 mg/mL each of apotransferrin and cytochrome c in buffer A. The sample volume applied in each case was 100 µL. Sample application in each case was followed by washing with 10 mL of buffer A. Desorption of the samples was performed using a linear salt gradient 0-0.5 M NaCl in the same buffer followed by a step increase in the salt concentration to 1.0 M NaCl. RESULTS Relative Protein Retention in Isocratic Elution. The isocratic k′ values obtained for the three test proteins on the SP Sepharose Fast Flow and the S3C and S0C spacer arm variants were compared as a function of NaCl concentration, using loglog plots of k′ against the NaCl concentration. The data are shown for lysozyme in Figure 2, for aCT in Figure 3, and for cytochrome c in Figure 4. The customarily observed decrease in isocratic k′ values as the salt concentration increases is apparent in these graphs, resulting in a series of lines with negative slopes. As seen with previous protein retention experiments,2 these lines are largely parallel, with some flattening or curvature exhibited at high NaCl concentrations. This curvature has been noted with ion exchange adsorbents when operated with high-salt mobile phases; under these conditions, Coulombic interactions are largely screened, and retention is thought to occur through secondary interactions, such as van der Waals or hydrophobic interactions.10 For the comparison of retention among the adsorbents examined here, this high-salt, low-k′ environment is less relevant; therefore, retention comparisons are made in the linear region of the curves, typically encompassing k′ values of 1 and greater. (10) Melander, W. R.; El Rassi, Z.; Horva´th, Cs. J. Chromatogr. 1989, 469, 3-27.
5818 Analytical Chemistry, Vol. 76, No. 19, October 1, 2004
Figure 2. Log k′ vs log [NaCl] plots for lysozyme on the S6C (b), S3C (O), and S0C (9) strong cation exchangers.
For lysozyme, the S0C version showed much greater retention, with k′ increasing ∼15-fold compared to the full-length spacer version. The adsorbent retention order was S0C > S3C > S6C, with the k′ values of the S3C showing a measurable but small (∼20%) increase over the full-length spacer arm. For aCT and cytochrome c, elimination of the spacer arm significantly increased retention (∼2-3-fold) over the least retentive version, but the retention order of S0C > S3C > S6C was observed only for the cytochrome c, while the S6C was more retentive than the S3C version for aCT. Gradient Retention in Chromatographic Functional Tests. The elution volumes for the different protein peaks on the three adsorbents in gradient elution at pH 4.0 are given in Table 2. The elution volumes for all peaks increase in the order S6C < S3C < S0C, with the largest difference being between the S3C and the S0C. For the S0C derivative, the β-lactoglobulin was not eluted within the gradient but could be eluted when the concentration of LiCl in the elution buffer was increased from 0.75 to 1.5 M. The chromatograms from the different runs on S6C and S0C at pH 6.0 are presented in Figure 5. The ribonuclease peak appears at the same elution volume for both resins, right at the beginning of the gradient. Cytochrome c is eluted at approximately the same elution volume for both ion exchangers. Apotransferrin appears as a single peak early in the chromatogram (overlapping the ribonuclease peak) on S6C, but on S0C its behavior is different,
Table 2. Gradient Elution Retention Volumes for Test Proteins elution vol (mL) wgla stationary phase
peak 1
wgla peak 2
wgla peak 3
β-lactoglobulin
SP Sepharose Fast Flow (6C) half-length spacer (3C) no spacer arm (0C)
77 83 104
108 115 193
125 134 222
165 181 NEb
a wgl, wheat germ lectin. b NE, not elutable using 0.75 M LiCl but could be eluted with 1.5 M LiCl.
Figure 3. Log k′ vs log [NaCl] plots for aCT on the S6C (b), S3C (O), and S0C (9) strong cation exchangers.
Figure 4. Log k′ vs log [NaCl] plots for cytochrome c on the S6C (b), S3C (O), and S0C (9) strong cation exchangers.
with two peaks present, the first having the same position as for S6C and the second appearing much later in the gradient. Increasing the ion capacity of the S0C amplifies some of these effects (data not shown). The retention of cytochrome c is increased further, while for apotransferrin the first peak is weaker and the second grows and elutes even later in the gradient. DISCUSSION The removal of the hydrocarbon spacer arm from the SP Sepharose Fast Flow significantly increased the retention of all the proteins studied, in both the isocratic and the gradient experiments. In several cases, the increase in retention was substantial, especially with the complete removal of the spacer arm in S0C; for lysozyme, for instance, the isocratic k′ increased
by more than 1 order of magnitude. The common base matrix of the adsorbent set eliminates variation in the adsorbent morphology and physical properties such as PSD and phase ratios, which would otherwise have direct effects on chromatographic retention.2 Thus, the differences in retention observed in our experiments lie in the nature of the derivatization, i.e., in the ligand density and structure, which we consider separately. The adsorbents differ in ionic capacity, with the S0C adsorbent having a lower capacity than the S6C or S3C versions (Table 1). Colloidal retention models for ion exchange often predict retention differences as a result of such differences in adsorbent charge density,10-18 although this pattern is not followed under all conditions.12 However, no correlation of this kind can be seen in the data here. On the contrary, the complete removal of the spacer arm, as seen in the S6C versus S0C comparison, clearly increased chromatographic retention, despite the pronounced decrease in ionic capacity. In contrast to the picture presented by colloidal models, the generally accepted view of the role of ligand density19 is that increasing ligand density affects retention only up to the point where the ligand spacing is comparable to the dimensions of the adsorbate protein molecules. Based on the measured phase ratio of the base matrix used here,1 all the adsorbents, including S0C, have ligand densities high enough to satisfy this criterion. As noted earlier, additional S0C batches with somewhat higher ionic capacities were used in some of the gradient chromatographic functional tests at pH 6 and displayed even stronger retention. While this observation suggests that the ligand density criterion is not always strictly accurate, the results as a whole point to the role of ligand and spacer arm structure rather than ligand density as the principal determinant of relative retention in the adsorbents. The observed retention increase in the absence of a spacer arm runs counter to the accumulated experience in affinity chromatography systems, where such hydrocarbon spacer arms add a hydrophobic component that is considered to drive an observed increase in overall retention.3,5 It also differs from protein retention in hydrophobic interaction chromatography (HIC), (11) Kopaciewicz, W.; Rounds, M. A.; Fausnaugh, J.; Regnier, F. E. J. Chromatogr. 1983, 266, 3-21. (12) Ståhlberg, J.; Jo¨nsson, B.; Horva´th, Cs. Anal. Chem. 1991, 63, 1867-1874. (13) Ståhlberg, J.; Jo¨nsson, B.; Horva´th, Cs. Anal. Chem. 1992, 64, 3118-3124. (14) Ståhlberg, J.; Jo ¨nsson, B. Anal. Chem. 1996, 68, 1536-1544. (15) Haggerty, L.; Lenhoff, A. M. J. Phys. Chem. 1991, 95, 1472-1477. (16) Roth, C. M.; Lenhoff, A. M. Langmuir 1993, 9, 962-972. (17) Roth, C. M.; Unger, K. K.; Lenhoff, A. M. J. Chromatogr., A 1996, 726, 45-56. (18) Asthagiri, D.; Lenhoff, A. M. Langmuir 1997, 13, 6761-6768. (19) Wu, D.; Walters, R. R. J. Chromatogr. 1992, 598, 7-13.
Analytical Chemistry, Vol. 76, No. 19, October 1, 2004
5819
Figure 5. Chromatograms of the mixtures of apotransferrin and cytochrome c (upper and lower left panels) and ribonuclease A and cytochrome c (upper and lower right panels) on the S6C and S0C strong cation exchangers. The chromatographic gradient (shown on the chromatograms) consisted of a 10-bed volume wash after injection with a buffer of 0.03 M sodium phosphate, 0.03 M sodium formate, and 0.03 M sodium acetate at a pH of 6.0, followed by a linear gradient of 0-0.5 M NaCl and then a step increase to 1.0 M NaCl, both in the same buffer.
which is increased with increasing length of the alkyl chain ligand on the adsorbent.20,21 The most obvious explanation for the observed trends lies in the lengths of the respective spacer arms. An additional factor is the sulfonate ligand used in S0C as compared to the sulfate used in S6C and S3C, but the parts of these ligands presented to adsorbing molecules are too similar in structure and electronic properties for this to be a viable explanation for the very large increase in retention for S0C.22 Thus, the roles of the spacer arms themselves deserve closer investigation. However, obtaining a detailed physical understanding of the mechanisms determining the role of the spacer arms is greatly complicated by the potential for a number of entropic and enthalpic effects that contribute to overall retention and the often cooperative effects of these interactions.11,23 One such mechanism that may be promoted by the reduction in spacer arm length is an affinity-like interaction due to enhanced binding permitted by closer interaction between the proteins and the adsorbent. For the proteins used here, this seems possible only for lysozyme. Structural similarity between the agarose polymer (D-galactose β(1,4)-3,6-anhydro-L-galactose R(1,3)) and the cell wall polysaccharide substrate of lysozyme (N-acetylglucosamine β(1,4)-N-acetylmuramic acid) could permit some degree of affinity interaction, particularly as both polymers contain β (1,4) glycosidic linkages, a requirement for lytic activity in lysozyme.24 Of relevance here is whether such bioaffinity interactions with the uncharged agarose matrix could produce the retention increases observed, especially on the S0C adsorbent, where the (20) Hofstee, B. H. J. Biochem. Biophys. Res. Commun. 1974, 53, 1137-1144. (21) Shaltiel, S.; Er-El, Z. Proc. Nat. Acad. Sci. U.S.A. 1973, 70, 778-781. (22) Asthagiri, D.; Schure, M. R.; Lenhoff, A. M. J. Phys. Chem. B 2000, 104, 8753-8761. (23) Velayudhan, A.; Horva´th, Cs. J. Chromatogr. 1988, 443, 13-29. (24) Phillips, D. C. Sci. Am. 1966, 215, 78.
5820
Analytical Chemistry, Vol. 76, No. 19, October 1, 2004
elimination of the spacer arm would presumably make the agarose backbone more sterically accessible. Examination of previous studies suggests that uncharged agarose does not appear to have sufficient geometric complementarity with the lysozyme substrate to alter chromatographic retention to the degree observed here. Comparisons of the binding affinity of hen egg white lysozyme with a variety of glycosaminoglycans (GAGs), both sulfated (chondroitin sulfate, heparin) and nonsulfated (hyaluronic acid), have shown that sulfate groups are not the dominant factor in GAG-lysozyme interactions.25 In addition, although these GAGs display tight binding to lysozyme (Kd
