- Email: [email protected]
Influence of mixed electrolytes and pH on adsorption of bovine serum albumin in hydrophobic interaction chromatography
Journal of Chromatography A, 1521 (2017) 73–79
Contents lists available at ScienceDirect
Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma
Influence of mixed electrolytes and pH on adsorption of bovine serum albumin in hydrophobic interaction chromatography Eva Hackemann, Hans Hasse ∗ Laboratory of Engineering Thermodynamics (LTD), University of Kaiserslautern, Kaiserslautern, Germany
a r t i c l e
i n f o
Article history: Received 18 April 2017 Received in revised form 15 August 2017 Accepted 2 September 2017 Available online 14 September 2017 Keywords: BSA Hydrophobic interaction chromatography (HIC) Mixed electrolytes Modeling pH
a b s t r a c t Using salt mixtures instead of single salts can be beneficial for hydrophobic interaction chromatography (HIC). The effect of electrolytes on the adsorption of proteins, however, depends on the pH. Little is known on that dependence for mixed electrolytes. Therefore, the effect of the pH on protein adsorption from aqueous solutions containing mixed salts is systematically studied in the present work for a model system: the adsorption of bovine serum albumin (BSA) on the mildly hydrophobic resin Toyopearl PPG600M. The pH is adjusted to 4.0, 4.7 or 7.0 using 25 mM sodium phosphate or sodium citrate buffer. Binary and ternary salt mixtures of sodium chloride, ammonium chloride, sodium sulfate and ammonium sulfate as well as the pure salts are used at overall ionic strengths between 1500 and 4200 mM. The temperature is always 25 ◦ C. The influence of the mixed electrolytes on the adsorption behavior of BSA changes completely with varying pH. Positive as well as negative cooperative effects of the mixed electrolytes are observed. The results are analyzed using a mathematical model which was recently introduced by our group. In that model the influence of the electrolytes is described by a Taylor series expansion in the individual ion molarities. After suitable parametrization using a subset of the data determined in the present work, the model successfully predicts the influence of mixed electrolytes on the protein adsorption. Furthermore, results for BSA from the present study are compared to literature data for lysozyme, which are available for the same adsorbent, temperature and salts. By calculating the ratio of the loading of the adsorbent for both proteins particularly favorable separation conditions can be selected. Hence, a model-based optimization of solvents for protein separation is possible. © 2017 Elsevier B.V. All rights reserved.
1. Introduction Hydrophobic interaction chromatography (HIC) is an attractive method for purifying proteins. The hydrophobic interaction of an adsorbent and a protein is favored by high salt concentrations [1,2]. In HIC typically a single salt is used for enhancing the adsorption. The choice of the salt type and concentration depends on the protein-adsorbent system. There is much literature on the influence of single salts on HIC, for a review see e.g. Lienqueo et al. [3]. It has been recently shown that using mixed electrolytes has a great potential in HIC [4–9]. Higher equilibrium loadings [4] and dynamic binding capacities [5] can be achieved. Also the retention and selectivity can be improved [6]. But the mechanisms determining the influence of mixed electrolytes on HIC are still only poorly understood. HIC is strongly influenced by the pH of the solu-
∗ Corresponding author. E-mail address: [email protected] (H. Hasse). http://dx.doi.org/10.1016/j.chroma.2017.09.024 0021-9673/© 2017 Elsevier B.V. All rights reserved.
tion [3,10–14]. Depending on the pH, the protein’s net charge and also its conformation can change significantly. Bovine serum albumin (BSA) is known to change its conformation drastically, but in a reversible manner, going from pH 2.7 to pH 10.0 [15,16]. Furthermore, it is known that hydrophobic interactions are stronger when solution pH is close to the isoelectric point of the protein [3]. Due to the vanishing net charge of the protein near its isoelectric point, the electrostatic repulsion between the protein molecules becomes small, favoring a closer packing on the adsorbent surface [3]. Xia et al. [13] found that when the buffer pH is close to the isoelectric point of the protein, more water is released during adsorption. They also report that the number of water molecules released upon a pH change increases with increasing salt concentration and that kosmotropic salts lead to a more important release of water than chaotropic salts. Baumann et al. [14] conclude that pH-induced reversible structural changes and protein reorientation upon binding can increase the dynamic binding capacity. There are also reports on calorimetric studies of the effect of pH on protein adsorption on HIC materials [12,13]. To the best of our
74
E. Hackemann, H. Hasse / J. Chromatogr. A 1521 (2017) 73–79
knowledge, the influence of pH on protein adsorption from mixed electrolyte solutions has not yet been studied for HIC materials. We have recently presented studies on the influence of mixed electrolytes on the adsorption of lysozyme, polyethylene glycol (PEG) and di-PEGylated lysozyme at different ionic strengths on the mildly hydrophobic adsorbent Toyopearl PPG-600M [4,9]. However, in those studies the pH was 7.0 throughout. Those studies are extended in the present work to BSA. For BSA, the pH can be varied so that the protein’s net charge goes from positive over neutral to negative, so that BSA lends itself to study the influence of pH on protein adsorption. The isoelectric point of BSA is about 4.7 in water for low salt concentrations. However, it may vary by several digit points with increasing salt concentration [17,18]. In the present study the pH was set to 4.0, 4.7 and 7.0, corresponding to a positive net charge at pH 4.0, and a negative net charge at pH 7.0, while the net charge at pH 4.7 is near zero. Furthermore, BSA is also known to change its conformation upon adsorption [19–21]. The adsorbent is, as in our previous studies [4,9], the mildly hydrophobic resin Toyopearl PPG-600M. The studied salts are sodium chloride, ammonium chloride, sodium sulfate and ammonium sulfate. All binary mixtures and some ternary mixtures of these salts are used, as well as the pure salts. The temperature is always 25 ◦ C. In our recent studies [4,9], we have introduced and successfully applied a mathematical model for the description of the influence of electrolytes on protein adsorption. It is based on a Taylor series expansion in the molarities of the individual ions. That model is also applied here to correlate and interpret the new experimental data. 2. Materials and methods 2.1. Materials BSA was obtained from Sigma–Aldrich (St. Louis, USA) with a purity of 98.5%. The hydrophobic resin Toyopearl PPG-600M was obtained from Tosoh Bioscience (Stuttgart, Germany). The components used for the buffer preparation, namely citric acid (C6 H8 O7 ), trisodium citrate dihydrate (Na3 C6 H5 O7 ·2H2 O), sodiumdihydrogen phosphate dihydrate (NaH2 PO4 ·2H2 O), disodiumhydrogen phosphate dihydrate (Na2 HPO4 ·2H2 O), sodium chloride (NaCl), ammonium chloride (NH4 Cl), sodium sulfate (Na2 SO4 ) and ammonium sulfate ((NH4 )2 SO4 ) were of analytical grade and were obtained from Carl Roth (Karlsruhe, Germany). High purity water was produced in a Milli-Q system from Millipore (Merck, Schwalbach, Germany). 2.2. Batch adsorption isotherms Batch adsorption measurements were carried out on a fully automated liquid handling station from Tecan (Maennedorf, Switzerland). The experiments and the analytics were performed as described previously in detail by Werner and Hasse [4]. The batch experiments were carried out with 50.9 l adsorbent (corresponding to about 32 mg [9]) and 500 l BSA solution. BSA was dissolved in 25 mM sodium phosphate buffer (pH 7.0) or 25 mM sodium citrate buffer (pH 4.0 and pH 4.7), to which salts were added gravimetrically with an AG 204 scale from Mettler Toledo (Columbus, USA). The pH was adjusted using sodium hydroxide, phosphoric acid or citric acid after the salts were added. The pH was measured with a Metrohm (Limburg, Germany) 780 pH meter. The salts were: sodium chloride, ammonium chloride, sodium sulfate and ammonium sulfate. Binary and ternary salt mixtures as well as pure salts were used. The overall ionic strength was varied between 1500 mM and 4200 mM. The temperature during the adsorption experiments was 25 ◦ C. For tempering a shaker from Inheco (Mar-
tiensried, Germany) with a special plate adapter was used. The equilibration time was 6 h. Preliminary experiments showed that equilibrium is always reached within that time. The BSA concentration in the equilibrated solution was analyzed by UV absorption at 280 nm. The relative uncertainty of the equilibrium loading determined in these experiments is about 10%, except for very small loadings [22]. 2.3. Experimental program in reduced coordinates In a previous work of our group [4], a model was developed to describe the loading of HIC resins with protein as a function of individual ion concentrations. The relation is established by a multi-dimensional Taylor series expansion. In the present work the investigated salts consist of the cations Na+ and NH 4+ and the anions Cl− and SO 42−. The concentrations are given in molarities ci . The degree of freedom can be reduced by introducing side conditions. One is given naturally by the electroneutrality condition:
ci zi = 0
(1)
i
where zi is the charge number of the ion i. Furthermore, for one series of experiments at constant ionic strength I, there is a second side condition: I=
1 2 ci zi = const. 2
(2)
i
In Eqs. (1) and (2), the summation is over all ions, i = Na+ , NH+ , Cl− 4
and SO2− 4 . According to this, all salt mixtures in the present study are fully specified by one anion and one cation concentration and the constant ionic strength. Na+ and Cl− concentrations are chosen for that purpose. Fig. 1 gives an overview of the studied salt mixtures in these coordinates. Depending on the pH of the solution different overall ionic strengths were chosen. In the diagrams in Fig. 1, the pure salts are shown in the four corners. The six line segments between the four corners correspond to binary salt mixtures. The points which are not on lines represent experiments with ternary salt mixtures, which were only studied for pH 4.7 and I = 2700 mM. The gray areas do not belong to the parameter space as the overall ionic strength would be above 1500 mM (A), 2700 mM (B) or 4200 mM (C) respectively. 2.4. Correlation of adsorption isotherms For modeling equilibrium adsorption the model does not have to be physically justified and any suitable equation can be used for the correlation. We have chosen the semi-empirical colloidal model of Oberholzer and Lenhoff [23] for that purpose, which correlates experimental isotherms well. It can be written in the following form: cp =
qp Kpads
· exp ˇ
qp /q0 · exp
−
qp /q0
(3)
where cp is the protein concentration in the liquid phase, qp is the protein loading of the adsorbent and Kpads is the adsorption equilibrium constant. ˇ and are fitting parameters which are made dimensionless by introducing q0 = 1 mg ml−1 . 2.5. Mathematical model Werner and Hasse [4] introduced a mathematical model for describing the dependence of properties zp related to the loading of
E. Hackemann, H. Hasse / J. Chromatogr. A 1521 (2017) 73–79
75
Fig. 1. Overview of the compositions of the salt mixtures which are used in the present work. The measured compositions are indicated by the dots. In the reduced parameter space, the salt mixture is fully specified by one cation and one anion concentration. Na+ and Cl− are chosen for that purpose. The pure salts are represented by the corners. The line segments between the corners represent the binary salt mixtures. The pH and overall ionic strength are (A) pH 4.0, I = 1500 mM, (B) pH 4.7, I = 2700 mM, (C) pH 7.0, I = 4200 mM. The gray areas do not belong to the parameter space.
HIC adsorbents with the protein p on the molarities of single ions ci present in the protein solutions: ln
c zp 1 ci cj ki i + kij = k0 + c0 2 c0 c0 unit of z i
i
(4)
j
k0 , ki and kij are the model parameters. To yield model parameters which are dimensionless c0 = 1000 mM is introduced. The sum over i and j covers all ions which are present in the solution. If no salt is present in the buffer, the number of zp is given by the parameter k0 . The influence of the ions is described by a multidimensional Taylor series expansion around that state. Numbers for the parameters ki , kij > 0 correspond to an increase of zp with increasing ion concentrations, whereas ki , kij <0 correspond to a decrease. The linear term describes the influence of individual ion concentrations on the adsorption. The quadratic term has been found sufficient to describe ion cross-interactions, if present [4,9]. Werner and Hasse [4] correlated experimental data on the loading of the adsorbent at a liquid phase concentration of the protein of 0.01 mM with that model, i.e. zp = qp (cp = 0.01 mM). In Hackemann et al. [9] it was shown that it is also possible to use the correlation proposed by Werner and Hasse [4] for describing the adsorption equilibrium constant, i.e. zp = Kpads . As discussed in [4], the number of parameters used in Eq. (4) can be reduced. The cross-interaction parameters are assumed to be symmetric: kij = kji
(5)
Interactions of ions with themselves are not considered. Therefore the kii and kjj are set to zero as they describe the same effect as ki : kii = kjj = 0
(6)
Furthermore, cross-interactions of anions and cations are set to zero, because it is assumed that for describing the properties of the solution containing a pure salt the linear part of Eq. (4) is sufficient. Further information and the procedure for the determination of model parameters can be found in Werner and Hasse [4]. 3. Results and discussion 3.1. Equilibrium adsorption isotherms In the present work, equilibrium adsorption isotherms were measured for different combinations of pH and ionic strength.
The results are documented in the Appendix, were the correlation parameters Kpads , ˇ and for all isotherms are reported. In Fig. 2 some examples of results from the present study are shown. It depicts selected equilibrium adsorption isotherms and corresponding model fits based on Eq. (3) for BSA for experiments at different conditions. In Fig. 2A, results at pH 4.7 and 7.0 for sodium chloride (4200 mM) are shown. At pH 4.7 the adsorption is much stronger than at pH 7.0. This can be explained by electrostatic effects. The isoelectric point of BSA is about 4.7. Hence, at pH 4.7 BSA carries almost no net charge, whereas at pH 7.0 it is negatively charged. The negative net charge is unfavorable for adsorption due to the repulsion of the adsorbed proteins [3,13]. Fig. 2B shows corresponding results at pH 4.0 and pH 4.7 for sodium chloride (2400 mM), respectively. The loading at pH 4.0 is much higher than that at pH 4.7. This is a consequence of the well-known influence of the pH on the structure of BSA [24,25]. At pH 7.0 BSA has a heard-shaped conformation, called N-Albumin. At pH 4.0 the protein can be partially expanded in a cigar-like shape, called F-Albumin. The transition to F-Albumin results in a higher accessible surface and in an exposure of core hydrophobic regions. As a result, the hydrophobic interaction at pH 4.0 is much higher than at pH 4.7, despite the adverse charge effect (BSA carries a positive net charge at pH 4.0). In conclusion, there is a strong influence of the buffer pH on the adsorption strength resulting both from configurational changes and electrostatic effects. In the following, the study is extended to mixed electrolyte systems. In Fig. 2C the equilibrium adsorption isotherms and corresponding model fits are shown for experiments at pH 4.0 with different mixtures of ammonium chloride and ammonium sulfate at I = 1500 mM. The lowest loadings are observed for pure ammonium sulfate. With pure ammonium chloride, much higher loadings are found. However, even higher loadings are observed for two salt mixtures. This shows positive cooperative effects of mixed electrolytes on the adsorption of BSA. Fig. 2D shows the results for the corresponding experiments at pH 7.0 with mixtures of sodium chloride and sodium sulfate at I = 4200 mM. The highest loading is achieved for pure sodium sulfate. Using pure sodium chloride leads to much lower loadings. But there are some mixtures of these two salts, which yield even lower loadings. Hence, also negative cooperative effects of using mixed electrolytes may occur. Based on the entire set of experimental data taken in the present work, it can be concluded that, depending on the pH, the type of salts, and the overall ionic strength, cooperative effects of using mixed electrolytes may occur, and these cooperative effects can be either
76
E. Hackemann, H. Hasse / J. Chromatogr. A 1521 (2017) 73–79
Fig. 2. Equilibrium adsorption isotherms and corresponding model fits. (A) pH = 4.7 and 7.0, pure sodium chloride, I = 4200 mM, (B) pH = 4.0 and 4.7, pure sodium chloride, I = 2400 mM, (C) pH 4.0, mixtures of ammonium chloride and ammonium sulfate, overall ionic strength I = 1500 mM, (D) pH 7.0, mixtures of sodium chloride and sodium sulfate, overall ionic strength I = 4200 mM. The dotted line indicates cp = 0.01 mM which was chosen for further comparison.
positive or negative. In the following, an overview of these effects in the studied systems is presented. To keep it simple, the discussion is not lead for the entire isotherms but only for one point on the isotherms, namely that for cp = 0.01 mM, cf. dotted line in Fig. 2A–D.
3.2. Parametrization of the model The model presented in Section 2.5 (Eq. (4)) was used for correlating the data for the BSA loading of the adsorbent for a BSA molarity in the liquid phase of cp = 0.01 mM, i.e. zp = qp (cp = 0.01 mM). The data obtained from the correlation of the individual isotherms (Eq. (3)) was used for calculating those numbers, as direct experimental data for exactly 0.01 mM are generally not available. For the adjustment of the model parameters, only data for the four binary systems, which border the reduced parameter space, namely sodium chloride–ammonium chloride, sodium chloride–sodium sulfate, sodium sulfate–ammonium sulfate, and ammonium chloride–ammonium sulfate (cf. Fig. 1) were used. The parameter k0 describes lnqp (cp = 0.01 mM) in the absence of salt in the buffer solution. Under these conditions, the loading qp (cp = 0.01 mM) is negligible and cannot be determined in batch
experiments. Therefore, k0 is arbitrarily set to −6.91, corresponding to qp (cp = 0.01 mM) = 0.001 mM in the absence of salts. The choice of the number for k0 has no significant effect on the other parameters or the results, as long as the resulting qp (cp = 0.01 mM) is small enough. If experimental data for qp in the absence of salts is available, it should be used. There are six cross-interactions between ions. The crossinteractions between ions with different signs of the charge are set to zero for the reasons discussed above. Only two parameters remain for describing cooperative effects, namely kNa+ ,NH+ and 4
kCl− ,SO2− respectively. Preliminary evaluation of the present data 4
has shown that the parameter kNa+ ,NH+ can be set to zero with4
out significant loss in the accuracy of the correlation. Hence only one cross-interaction parameter remains per system for describing cooperative effects, namely kCl− ,SO2− . The model parameters 4
are presented in Table 1. As explained in [4], only three of the four parameters of the linear term are linearly independent. As recommended in [4], two parameters, one for an cation and one for a anion, are constrained to be equal. NH 4+ and Cl− are selected for that purpose.
E. Hackemann, H. Hasse / J. Chromatogr. A 1521 (2017) 73–79 Table 1 Parameters of the mathematical model (Eq. (4)) for the adsorption of BSA on Toyopearl PPG-600M at 25 ◦ C. k0 is set to −6.91 throughout. All parameters not listed below are set to zero.
I / mM pH
A 1500 4.0
B 2700 4.7
C 4200 7.0
kNa+ kNH+
1.97 1.66
0.95 0.76
0.57 0.26
kCl− kSO2−
1.66 3.50
0.76 3.39
0.26 2.71
4.59
0.00
−0.61
4
4
kCl− ,SO2− 4
3.3. Comparison of experimental and model results In the following, the experimental results for lnqp (cp = 0.01 mM) are compared to the model results obtained using the parametrization described in Section 3.2. In the remainder of this paragraph qp (cp = 0.01 mM) will simply be referred to as ‘loading’. In Fig. 3 results for the loading are shown as a function of the salt composition, which is represented here in the reduced parameter space introduced in Section 2.3. Three cases are shown: (A) pH 4.0 and I = 1500 mM, (B) pH 4.7 and I = 2700 mM, (C) pH 7.0 and I = 4200 mM. The red dots are the experimental values. The small gray dots represent the results from the mathematical model. They are points on a surface which will be called ‘model surface’ here. The blue crosses are vertical projections of the experimental data on that model surface. A good fit of the experimental results is obtained in all three systems A–C. The deviations between the experimental data and the model results are mostly within the experimental uncertainty. The discrepancies for the mixed salts are not above these for the pure salts. Not only the four binary systems which were used for adjusting the model parameters are well described but also the two other binary systems. The same holds for data for ternary salt mixtures for which no adjustment of model parameters was made. In Fig. 3A the results for pH 4.0 and I = 1500 mM are shown. In some of the binary salt systems linear behavior is observed whereas in others, positive cooperative effects occur. They are due to the cross-interactions between chloride and sulfate ions. The highest loadings are observed in binary mixtures containing high amounts of sodium chloride. The lowest loadings can be found in pure ammonium sulfate. Fig. 3B shows the results for pH 4.7, which is near the isoelectric point of BSA and I = 2700 mM. At those conditions, no important cooperative effects are observed, the linear terms in Eq. (4) are sufficient to describe the experimental results well. Independent of which salt or salt mixture is used the loading is of the order of 0.1 mM. Fig. 3C shows the results for pH 7.0 and I = 4200 mM. For those conditions, only negative cooperative effects are observed. There is a negative influence of using Cl− and SO 42− in combination. The highest loading is observed in pure sodium sulfate, the lowest loading in pure ammonium chloride. The results of the fit are discussed in the following, keeping in mind that the magnitude of the numbers for the ki parameters gives information on the influence of an individual ion on the loading whereas the magnitude of the numbers for kij give information on the influence of the cooperative ion effects. Positive numbers indicate a salting-out effect, negative numbers indicate a salting-in effect. The absolute numbers should not be overinterpreted as they depend on the way the normalization is carried out (here: kNH+ = kCl− throughout). However, it is 4
still possible to compare ratios between numbers for cations and anions. From this it is clear that for all three systems A, B, C, the salting-out effect of SO 42− is much stronger than that of Cl− , in perfect agreement with the well-known Hofmeister series [26]. The results also indicate a slightly stronger salting-out effect of Na+
77
compared to NH 4+. This is not in agreement with the Hofmeister series [26]. There is a strong positive cooperative effect of the combination of Cl− and SO 42− for pH 4.0 for which BSA has a positive net charge. This is interpreted here as a consequence of different types of positive sites on BSA, where certain types are more favorable for Cl− while other types are more favorable for SO 42−. The small negative cooperative effects which are observed for pH 7.0, for which BSA has a negative net charge, are interpreted here as a consequence of the influence of the solvent, i.e. we speculate that having both Cl− and SO 42− in the solution together with the corresponding cations and not only one type of anions, yields a more favorable solubility for BSA. At the isoelectric point of the protein at pH 4.7 nearly no cooperative effects occur as the protein has almost no net charge and therefore the interactions of the charged sites of BSA and the ions in the solution are less important than in the cases when BSA carries a net charge. Overall, summing up all parameters for the three cases shows that the influence of electrolytes on the BSA adsorption is strongest for pH 4.0. 3.4. Comparison of BSA and lysozyme It is interesting to compare the results from the present study on BSA with those for lysozyme from [4] as in both cases the same adsorbent, temperature and salts were used. The comparison is carried out for pH 7.0 as only for that pH data from both studies is available. Lysozyme carries a positive net charge at that pH whereas BSA carries a negative net charge. Furthermore an ionic strength must be selected for the comparison, which is carried out by comparing model results for both systems. For convenience, I = 3000 mM is chosen, which is in the range where experimental data is available for both systems. The results of the comparison are shown in Fig. 4 in the format which was already used above. In Panel A the results for BSA (blue) and lysozyme (red) are compared. The adsorption of lysozyme is stronger than that of BSA for all salt mixtures except those containing predominantly ammonium sulfate. Positive cooperative effects are observed for lysozyme, as they were observed in the present study for BSA at conditions where BSA is positively charged. However, for pH 7.0 for BSA negative cooperative effects occur, as discussed above. For process design the separation of proteins is of predominant importance. The information presented in Fig. 4A can be used for selecting conditions for which the separation of BSA and lysozyme is expected to be particularly favorable. For this, the ratio of the loading of the adsorbent for both proteins is calculated. The results are presented in Fig. 4B. They indicate that mixtures of sodium chloride and sodium sulfate with a strong excess of sodium chloride are particularly favorable. Of course, this optimization covers only one aspect of the separation process. It is mainly presented here to demonstrate the possibilities a model-based optimization offers in solvent design for protein separation in HIC. 4. Conclusion In the present work the combined influence of mixed electrolytes together with a buffer pH variation on hydrophobic interaction chromatography (HIC) is explored using bovine serum albumin BSA as a model protein. The pH is adjusted so that BSA carries either positive, negative or no net charge. Furthermore, BSA undergoes a structural change as result of the pH variation. Four salts, sodium chloride, ammonium chloride, sodium sulfate and ammonium sulfate, containing four different ions, are used. In the study, binary and ternary salt mixtures are used as well as the pure salts. For each studied pH, the salt composition is varied keeping the overall ionic strength constant at a suitably chosen value.
78
E. Hackemann, H. Hasse / J. Chromatogr. A 1521 (2017) 73–79
Fig. 3. BSA loading of the adsorbent at a BSA concentration in the solvent of 0.01 mM over the reduced parameter space describing the composition of the salt mixture. (A) pH 4.0 and I = 1500 mM, (B) pH 4.7 and I = 2700 mM, (C) pH 7.0 and I = 4200 mM. The big red dots are the experimental values. The small gray dots represent the surface resulting from the mathematical model adjusted to experimental data. The blue crosses are vertical projections of the experimental data on the model surface.
Fig. 4. BSA and lysozyme loading of the adsorbent at a protein concentration in the solvent of 0.01 mM over the reduced parameter space at pH 7.0 and I = 3000 mM. (A) BSA (red) and lysozyme (blue) show opposed behavior. (B) Difference lnqlysozyme − lnqBSA . The best separation is found for a mixed salt system of 2250 mM sodium chloride and 250 mM sodium sulfate.
Mixed salts can have positive, negative, or no cooperative effects. The magnitude and sign of the effect depends on the net charge of the protein. A mathematical model is used to correlate the data and interpret the observations. Despite the simplicity of that model, which is basically a multidimensional Taylor series expansion in the individual ion molarities, it correlates the comprehensive data taken in the present study well, with comparatively few parameters. It also gives good predictions for salt mixtures which were not included in the parameter fit. Particularly strong positive cooperative effects are observed when both Cl− and SO 42− are present in the solution and BSA is positively charged. This is attributed to the fact that different types of positive sites on BSA exist and some may be more favorable for Cl− while others may be more favorable for SO 42−. Whether this is a consequence of the different charge number of the chloride and the sulfate ions needs to be studied in future work. The present study also supports the approach of interpreting the influence of salts on HIC as resulting from the influence of sin-
gle ions, which opens more room for predictions as compared to interpretations relying on a salt, i.e. a certain combination of cation and anion. Finally, by comparing results for BSA from the present study and results for lysozyme from a previous study, it is demonstrated that the model can be used for selecting suitable conditions for protein separations. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.chroma.2017.09. 024. References [1] W. Melander, C. Horváth, Salt effects on hydrophobic interactions in precipitation and chromatography of proteins: an interpretations of the lyotropic series, Arch. Biochem. Biophys. 183 (1977) 200–215.
E. Hackemann, H. Hasse / J. Chromatogr. A 1521 (2017) 73–79 [2] T.W. Perkins, D.S. Mak, T.W. Root, E.N. Lightfoot, Protein retention in hydrophobic interaction chromatography: modeling variation with buffer ionic strength and column hydrophobicity, J. Chromatogr. A 766 (1997) 1–14. [3] M.E. Lienqueo, A. Mahn, J.C. Salgado, J.A. Asenjo, Current insights on protein behaviour in hydrophobic interaction chromatography, J. Chromatogr. B 849 (2007) 53–68, http://dx.doi.org/10.1016/j.jchromb.2006.11.019. [4] A. Werner, H. Hasse, Experimental study and modeling of the influence of mixed electrolytes on adsorption of macromolecules on a hydrophobic resin, J. Chromatogr. A 1315 (2013) 135–144, http://dx.doi.org/10.1016/j.chroma. 2013.09.071. [5] E. Mueller, J. Vajda, D. Josic, T. Schröder, R. Dabre, T. Frey, Mixed electrolytes in hydrophobic interaction chromatography, J. Sep. Sci. (2013) 1327–1334, http://dx.doi.org/10.1002/jssc.201200704. [6] Z.E. Rassi, L.F. de Ocampo, M.D. Bacolod, Binary and ternary salt gradients in hydrophobic-interaction chromatography of proteins, J. Chromatogr. A 499 (1990) 141–152, http://dx.doi.org/10.1016/S0021-9673(00)96969-3. [7] A.M. Senczuk, R. Klinke, T. Arakawa, G. Vedantham, Y. Yigzaw, Hydrophobic interaction chromatography in dual salt system increases protein binding capacity, Biotechnol. Bioeng. 103 (5) (2009) 930–935, http://dx.doi.org/10. 1002/bit.22313. [8] K. Baumgartner, S.A. Oelmeier, J. Hubbuch, The influence of mixed salts on the capacity of hic adsorbers: a predictive correlation to the surface tension and the aggregation temperature, Biotechnol. Prog. (2016), http://dx.doi.org/10. 1002/btpr.2166. [9] E. Hackemann, A. Werner, H. Hasse, Influence of mixed electrolytes on the adsorption of lysozyme, PEG, and PEGylated lysozyme on a hydrophobic interaction chromatography resin, Biotechnol. Prog. 33 (4) (2017) 1104–1115 https://doi.org/10.1002/btpr.2474. [10] S. Hjertén, K. Yao, K.-O. Eriksson, B. Johansson, Gradient and isocratic high-performance hydrophobic interaction chromatography of proteins on agarose columns, J. Chromatogr. 359 (1986) 99–109. [11] J. Fausnaugh, F. Regnier, Solute and mobile phase contributions to retention in hydrophobic interactions chromatography of proteins, J. Chromatogr. 359 (1986) 131–146. [12] R.A. Alberty, Maxwell relations for thermodynamic quantities of biochemical reactions, J. Am. Chem. Soc. 91 (14) (1969) 3899–3903, http://dx.doi.org/10. 1021/ja01042a037. [13] F. Xia, D. Negrath, S.M. Cramer, Effect of pH changes on water release values in hydrophobic interaction chromatographic systems, J. Chromatogr. A 1079 (2005) 229–235.
79
[14] P. Baumann, K. Baumgartner, J. Hubbuch, Influence of binding pH and protein solubility on the dynamic binding capacity in hydrophobic interaction chromatography, J. Chromatogr. A 1396 (2015) 77–85, http://dx.doi.org/10. 1016/j.chroma.2015.04.001. [15] D.C. Carter, J.X. Ho, Structure of serum albumin, in: F.M. Richards, C.B. Anfinsen, J.T. Edsall, D.S. Eisenberg (Eds.), Lipoproteins, Apolipoproteins, and Lipases Advances in Protein Chemistry, vol. 45, Academic Press, 1994, pp. 153–203, http://dx.doi.org/10.1016/S0065-3233(08)60640-3. [16] T.J. Peters, 2 – The albumin molecule: its structure and chemical properties, in: T. Peters (Ed.), All About Albumin, Academic Press, San Diego, 1995, pp. 9–11, http://dx.doi.org/10.1016/B978-012552110-9/50004-0. [17] A. Salis, M. Bostroem, L. Medda, F. Cugia, B. Barse, D.F. Parsons, B.W. Ninham, M. Monduzzi, Measurements and theoretical interpretation of points of zero charge/potential of BSA protein, Langmuir 27 (18) (2011) 11597–11604, http://dx.doi.org/10.1021/la2024605, pMID: 21834579. ´ U. Salgyn, ´ S. Bahadir, Zeta potentials and isoelectric points of [18] S. Salgyn, biomolecules: the effects of ion types and ionic strengths, Int. J. Electrochem. Sci. 7 (2012) 12404–12414. [19] R. Ueberbacher, E. Haimer, R. Hahn, A. Jungbauer, Hydrophobic interaction chromatography of proteins V. Quantitative assessment of conformational changes, J. Chromatogr. A 1198–1199 (2008) 154–163. [20] R. Ueberbacher, A. Rodler, R. Hahn, A. Jungbauer, Hydrophobic interaction chromatography of proteins: thermodynamic analysis of conformational changes, J. Chromatogr. A 1217 (2010) 184–190. [21] A. Jungbauer, C. Machold, R. Hahn, Hydrophobic interaction chromatography of proteins: III. Unfolding of proteins upon adsorption, J. Chromatogr. A 1079 (2005) 221–228, http://dx.doi.org/10.1016/j.chroma.2005.04.002. [22] A. Werner, T. Blaschke, H. Hasse, Microcaloric study of the adsorption of PEGylated lysozyme and peg on a mildly hydrophobic resin: influence of ammonium-sulfate, Langmuir 28 (2012) 11376–11383. [23] M.R. Oberholzer, A.M. Lenhoff, Protein adsorption isotherms through colloidal energetics, Langmuir 15 (1999) 3905–3914. [24] C. Leggio, L. Galantini, N.V. Pavel, About the albumin structure in solution: cigar expanded form versus heart normal shape, Phys. Chem. Chem. Phys. 10 (2008) 6741–6750, http://dx.doi.org/10.1039/B808938H. [25] K. Baler, O.A. Martin, M.A. Carignano, G.A. Ameer, J.A. Vila, I. Szleifer, Electrostatic unfolding and interactions of albumin driven by pH changes: a molecular dynamics study, J. Phys. Chem. B 118 (4) (2014) 921–930, http://dx. doi.org/10.1021/jp409936v, pMID: 24393011. [26] F. Hofmeister, Zur lehre von der wirkung der salze, Arch. Exp. Pathol. Pharm. 24 (4) (1888) 247–260, http://dx.doi.org/10.1007/BF01918191.
Contents lists available at ScienceDirect
Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma
Influence of mixed electrolytes and pH on adsorption of bovine serum albumin in hydrophobic interaction chromatography Eva Hackemann, Hans Hasse ∗ Laboratory of Engineering Thermodynamics (LTD), University of Kaiserslautern, Kaiserslautern, Germany
a r t i c l e
i n f o
Article history: Received 18 April 2017 Received in revised form 15 August 2017 Accepted 2 September 2017 Available online 14 September 2017 Keywords: BSA Hydrophobic interaction chromatography (HIC) Mixed electrolytes Modeling pH
a b s t r a c t Using salt mixtures instead of single salts can be beneficial for hydrophobic interaction chromatography (HIC). The effect of electrolytes on the adsorption of proteins, however, depends on the pH. Little is known on that dependence for mixed electrolytes. Therefore, the effect of the pH on protein adsorption from aqueous solutions containing mixed salts is systematically studied in the present work for a model system: the adsorption of bovine serum albumin (BSA) on the mildly hydrophobic resin Toyopearl PPG600M. The pH is adjusted to 4.0, 4.7 or 7.0 using 25 mM sodium phosphate or sodium citrate buffer. Binary and ternary salt mixtures of sodium chloride, ammonium chloride, sodium sulfate and ammonium sulfate as well as the pure salts are used at overall ionic strengths between 1500 and 4200 mM. The temperature is always 25 ◦ C. The influence of the mixed electrolytes on the adsorption behavior of BSA changes completely with varying pH. Positive as well as negative cooperative effects of the mixed electrolytes are observed. The results are analyzed using a mathematical model which was recently introduced by our group. In that model the influence of the electrolytes is described by a Taylor series expansion in the individual ion molarities. After suitable parametrization using a subset of the data determined in the present work, the model successfully predicts the influence of mixed electrolytes on the protein adsorption. Furthermore, results for BSA from the present study are compared to literature data for lysozyme, which are available for the same adsorbent, temperature and salts. By calculating the ratio of the loading of the adsorbent for both proteins particularly favorable separation conditions can be selected. Hence, a model-based optimization of solvents for protein separation is possible. © 2017 Elsevier B.V. All rights reserved.
1. Introduction Hydrophobic interaction chromatography (HIC) is an attractive method for purifying proteins. The hydrophobic interaction of an adsorbent and a protein is favored by high salt concentrations [1,2]. In HIC typically a single salt is used for enhancing the adsorption. The choice of the salt type and concentration depends on the protein-adsorbent system. There is much literature on the influence of single salts on HIC, for a review see e.g. Lienqueo et al. [3]. It has been recently shown that using mixed electrolytes has a great potential in HIC [4–9]. Higher equilibrium loadings [4] and dynamic binding capacities [5] can be achieved. Also the retention and selectivity can be improved [6]. But the mechanisms determining the influence of mixed electrolytes on HIC are still only poorly understood. HIC is strongly influenced by the pH of the solu-
∗ Corresponding author. E-mail address: [email protected] (H. Hasse). http://dx.doi.org/10.1016/j.chroma.2017.09.024 0021-9673/© 2017 Elsevier B.V. All rights reserved.
tion [3,10–14]. Depending on the pH, the protein’s net charge and also its conformation can change significantly. Bovine serum albumin (BSA) is known to change its conformation drastically, but in a reversible manner, going from pH 2.7 to pH 10.0 [15,16]. Furthermore, it is known that hydrophobic interactions are stronger when solution pH is close to the isoelectric point of the protein [3]. Due to the vanishing net charge of the protein near its isoelectric point, the electrostatic repulsion between the protein molecules becomes small, favoring a closer packing on the adsorbent surface [3]. Xia et al. [13] found that when the buffer pH is close to the isoelectric point of the protein, more water is released during adsorption. They also report that the number of water molecules released upon a pH change increases with increasing salt concentration and that kosmotropic salts lead to a more important release of water than chaotropic salts. Baumann et al. [14] conclude that pH-induced reversible structural changes and protein reorientation upon binding can increase the dynamic binding capacity. There are also reports on calorimetric studies of the effect of pH on protein adsorption on HIC materials [12,13]. To the best of our
74
E. Hackemann, H. Hasse / J. Chromatogr. A 1521 (2017) 73–79
knowledge, the influence of pH on protein adsorption from mixed electrolyte solutions has not yet been studied for HIC materials. We have recently presented studies on the influence of mixed electrolytes on the adsorption of lysozyme, polyethylene glycol (PEG) and di-PEGylated lysozyme at different ionic strengths on the mildly hydrophobic adsorbent Toyopearl PPG-600M [4,9]. However, in those studies the pH was 7.0 throughout. Those studies are extended in the present work to BSA. For BSA, the pH can be varied so that the protein’s net charge goes from positive over neutral to negative, so that BSA lends itself to study the influence of pH on protein adsorption. The isoelectric point of BSA is about 4.7 in water for low salt concentrations. However, it may vary by several digit points with increasing salt concentration [17,18]. In the present study the pH was set to 4.0, 4.7 and 7.0, corresponding to a positive net charge at pH 4.0, and a negative net charge at pH 7.0, while the net charge at pH 4.7 is near zero. Furthermore, BSA is also known to change its conformation upon adsorption [19–21]. The adsorbent is, as in our previous studies [4,9], the mildly hydrophobic resin Toyopearl PPG-600M. The studied salts are sodium chloride, ammonium chloride, sodium sulfate and ammonium sulfate. All binary mixtures and some ternary mixtures of these salts are used, as well as the pure salts. The temperature is always 25 ◦ C. In our recent studies [4,9], we have introduced and successfully applied a mathematical model for the description of the influence of electrolytes on protein adsorption. It is based on a Taylor series expansion in the molarities of the individual ions. That model is also applied here to correlate and interpret the new experimental data. 2. Materials and methods 2.1. Materials BSA was obtained from Sigma–Aldrich (St. Louis, USA) with a purity of 98.5%. The hydrophobic resin Toyopearl PPG-600M was obtained from Tosoh Bioscience (Stuttgart, Germany). The components used for the buffer preparation, namely citric acid (C6 H8 O7 ), trisodium citrate dihydrate (Na3 C6 H5 O7 ·2H2 O), sodiumdihydrogen phosphate dihydrate (NaH2 PO4 ·2H2 O), disodiumhydrogen phosphate dihydrate (Na2 HPO4 ·2H2 O), sodium chloride (NaCl), ammonium chloride (NH4 Cl), sodium sulfate (Na2 SO4 ) and ammonium sulfate ((NH4 )2 SO4 ) were of analytical grade and were obtained from Carl Roth (Karlsruhe, Germany). High purity water was produced in a Milli-Q system from Millipore (Merck, Schwalbach, Germany). 2.2. Batch adsorption isotherms Batch adsorption measurements were carried out on a fully automated liquid handling station from Tecan (Maennedorf, Switzerland). The experiments and the analytics were performed as described previously in detail by Werner and Hasse [4]. The batch experiments were carried out with 50.9 l adsorbent (corresponding to about 32 mg [9]) and 500 l BSA solution. BSA was dissolved in 25 mM sodium phosphate buffer (pH 7.0) or 25 mM sodium citrate buffer (pH 4.0 and pH 4.7), to which salts were added gravimetrically with an AG 204 scale from Mettler Toledo (Columbus, USA). The pH was adjusted using sodium hydroxide, phosphoric acid or citric acid after the salts were added. The pH was measured with a Metrohm (Limburg, Germany) 780 pH meter. The salts were: sodium chloride, ammonium chloride, sodium sulfate and ammonium sulfate. Binary and ternary salt mixtures as well as pure salts were used. The overall ionic strength was varied between 1500 mM and 4200 mM. The temperature during the adsorption experiments was 25 ◦ C. For tempering a shaker from Inheco (Mar-
tiensried, Germany) with a special plate adapter was used. The equilibration time was 6 h. Preliminary experiments showed that equilibrium is always reached within that time. The BSA concentration in the equilibrated solution was analyzed by UV absorption at 280 nm. The relative uncertainty of the equilibrium loading determined in these experiments is about 10%, except for very small loadings [22]. 2.3. Experimental program in reduced coordinates In a previous work of our group [4], a model was developed to describe the loading of HIC resins with protein as a function of individual ion concentrations. The relation is established by a multi-dimensional Taylor series expansion. In the present work the investigated salts consist of the cations Na+ and NH 4+ and the anions Cl− and SO 42−. The concentrations are given in molarities ci . The degree of freedom can be reduced by introducing side conditions. One is given naturally by the electroneutrality condition:
ci zi = 0
(1)
i
where zi is the charge number of the ion i. Furthermore, for one series of experiments at constant ionic strength I, there is a second side condition: I=
1 2 ci zi = const. 2
(2)
i
In Eqs. (1) and (2), the summation is over all ions, i = Na+ , NH+ , Cl− 4
and SO2− 4 . According to this, all salt mixtures in the present study are fully specified by one anion and one cation concentration and the constant ionic strength. Na+ and Cl− concentrations are chosen for that purpose. Fig. 1 gives an overview of the studied salt mixtures in these coordinates. Depending on the pH of the solution different overall ionic strengths were chosen. In the diagrams in Fig. 1, the pure salts are shown in the four corners. The six line segments between the four corners correspond to binary salt mixtures. The points which are not on lines represent experiments with ternary salt mixtures, which were only studied for pH 4.7 and I = 2700 mM. The gray areas do not belong to the parameter space as the overall ionic strength would be above 1500 mM (A), 2700 mM (B) or 4200 mM (C) respectively. 2.4. Correlation of adsorption isotherms For modeling equilibrium adsorption the model does not have to be physically justified and any suitable equation can be used for the correlation. We have chosen the semi-empirical colloidal model of Oberholzer and Lenhoff [23] for that purpose, which correlates experimental isotherms well. It can be written in the following form: cp =
qp Kpads
· exp ˇ
qp /q0 · exp
−
qp /q0
(3)
where cp is the protein concentration in the liquid phase, qp is the protein loading of the adsorbent and Kpads is the adsorption equilibrium constant. ˇ and are fitting parameters which are made dimensionless by introducing q0 = 1 mg ml−1 . 2.5. Mathematical model Werner and Hasse [4] introduced a mathematical model for describing the dependence of properties zp related to the loading of
E. Hackemann, H. Hasse / J. Chromatogr. A 1521 (2017) 73–79
75
Fig. 1. Overview of the compositions of the salt mixtures which are used in the present work. The measured compositions are indicated by the dots. In the reduced parameter space, the salt mixture is fully specified by one cation and one anion concentration. Na+ and Cl− are chosen for that purpose. The pure salts are represented by the corners. The line segments between the corners represent the binary salt mixtures. The pH and overall ionic strength are (A) pH 4.0, I = 1500 mM, (B) pH 4.7, I = 2700 mM, (C) pH 7.0, I = 4200 mM. The gray areas do not belong to the parameter space.
HIC adsorbents with the protein p on the molarities of single ions ci present in the protein solutions: ln
c zp 1 ci cj ki i + kij = k0 + c0 2 c0 c0 unit of z i
i
(4)
j
k0 , ki and kij are the model parameters. To yield model parameters which are dimensionless c0 = 1000 mM is introduced. The sum over i and j covers all ions which are present in the solution. If no salt is present in the buffer, the number of zp is given by the parameter k0 . The influence of the ions is described by a multidimensional Taylor series expansion around that state. Numbers for the parameters ki , kij > 0 correspond to an increase of zp with increasing ion concentrations, whereas ki , kij <0 correspond to a decrease. The linear term describes the influence of individual ion concentrations on the adsorption. The quadratic term has been found sufficient to describe ion cross-interactions, if present [4,9]. Werner and Hasse [4] correlated experimental data on the loading of the adsorbent at a liquid phase concentration of the protein of 0.01 mM with that model, i.e. zp = qp (cp = 0.01 mM). In Hackemann et al. [9] it was shown that it is also possible to use the correlation proposed by Werner and Hasse [4] for describing the adsorption equilibrium constant, i.e. zp = Kpads . As discussed in [4], the number of parameters used in Eq. (4) can be reduced. The cross-interaction parameters are assumed to be symmetric: kij = kji
(5)
Interactions of ions with themselves are not considered. Therefore the kii and kjj are set to zero as they describe the same effect as ki : kii = kjj = 0
(6)
Furthermore, cross-interactions of anions and cations are set to zero, because it is assumed that for describing the properties of the solution containing a pure salt the linear part of Eq. (4) is sufficient. Further information and the procedure for the determination of model parameters can be found in Werner and Hasse [4]. 3. Results and discussion 3.1. Equilibrium adsorption isotherms In the present work, equilibrium adsorption isotherms were measured for different combinations of pH and ionic strength.
The results are documented in the Appendix, were the correlation parameters Kpads , ˇ and for all isotherms are reported. In Fig. 2 some examples of results from the present study are shown. It depicts selected equilibrium adsorption isotherms and corresponding model fits based on Eq. (3) for BSA for experiments at different conditions. In Fig. 2A, results at pH 4.7 and 7.0 for sodium chloride (4200 mM) are shown. At pH 4.7 the adsorption is much stronger than at pH 7.0. This can be explained by electrostatic effects. The isoelectric point of BSA is about 4.7. Hence, at pH 4.7 BSA carries almost no net charge, whereas at pH 7.0 it is negatively charged. The negative net charge is unfavorable for adsorption due to the repulsion of the adsorbed proteins [3,13]. Fig. 2B shows corresponding results at pH 4.0 and pH 4.7 for sodium chloride (2400 mM), respectively. The loading at pH 4.0 is much higher than that at pH 4.7. This is a consequence of the well-known influence of the pH on the structure of BSA [24,25]. At pH 7.0 BSA has a heard-shaped conformation, called N-Albumin. At pH 4.0 the protein can be partially expanded in a cigar-like shape, called F-Albumin. The transition to F-Albumin results in a higher accessible surface and in an exposure of core hydrophobic regions. As a result, the hydrophobic interaction at pH 4.0 is much higher than at pH 4.7, despite the adverse charge effect (BSA carries a positive net charge at pH 4.0). In conclusion, there is a strong influence of the buffer pH on the adsorption strength resulting both from configurational changes and electrostatic effects. In the following, the study is extended to mixed electrolyte systems. In Fig. 2C the equilibrium adsorption isotherms and corresponding model fits are shown for experiments at pH 4.0 with different mixtures of ammonium chloride and ammonium sulfate at I = 1500 mM. The lowest loadings are observed for pure ammonium sulfate. With pure ammonium chloride, much higher loadings are found. However, even higher loadings are observed for two salt mixtures. This shows positive cooperative effects of mixed electrolytes on the adsorption of BSA. Fig. 2D shows the results for the corresponding experiments at pH 7.0 with mixtures of sodium chloride and sodium sulfate at I = 4200 mM. The highest loading is achieved for pure sodium sulfate. Using pure sodium chloride leads to much lower loadings. But there are some mixtures of these two salts, which yield even lower loadings. Hence, also negative cooperative effects of using mixed electrolytes may occur. Based on the entire set of experimental data taken in the present work, it can be concluded that, depending on the pH, the type of salts, and the overall ionic strength, cooperative effects of using mixed electrolytes may occur, and these cooperative effects can be either
76
E. Hackemann, H. Hasse / J. Chromatogr. A 1521 (2017) 73–79
Fig. 2. Equilibrium adsorption isotherms and corresponding model fits. (A) pH = 4.7 and 7.0, pure sodium chloride, I = 4200 mM, (B) pH = 4.0 and 4.7, pure sodium chloride, I = 2400 mM, (C) pH 4.0, mixtures of ammonium chloride and ammonium sulfate, overall ionic strength I = 1500 mM, (D) pH 7.0, mixtures of sodium chloride and sodium sulfate, overall ionic strength I = 4200 mM. The dotted line indicates cp = 0.01 mM which was chosen for further comparison.
positive or negative. In the following, an overview of these effects in the studied systems is presented. To keep it simple, the discussion is not lead for the entire isotherms but only for one point on the isotherms, namely that for cp = 0.01 mM, cf. dotted line in Fig. 2A–D.
3.2. Parametrization of the model The model presented in Section 2.5 (Eq. (4)) was used for correlating the data for the BSA loading of the adsorbent for a BSA molarity in the liquid phase of cp = 0.01 mM, i.e. zp = qp (cp = 0.01 mM). The data obtained from the correlation of the individual isotherms (Eq. (3)) was used for calculating those numbers, as direct experimental data for exactly 0.01 mM are generally not available. For the adjustment of the model parameters, only data for the four binary systems, which border the reduced parameter space, namely sodium chloride–ammonium chloride, sodium chloride–sodium sulfate, sodium sulfate–ammonium sulfate, and ammonium chloride–ammonium sulfate (cf. Fig. 1) were used. The parameter k0 describes lnqp (cp = 0.01 mM) in the absence of salt in the buffer solution. Under these conditions, the loading qp (cp = 0.01 mM) is negligible and cannot be determined in batch
experiments. Therefore, k0 is arbitrarily set to −6.91, corresponding to qp (cp = 0.01 mM) = 0.001 mM in the absence of salts. The choice of the number for k0 has no significant effect on the other parameters or the results, as long as the resulting qp (cp = 0.01 mM) is small enough. If experimental data for qp in the absence of salts is available, it should be used. There are six cross-interactions between ions. The crossinteractions between ions with different signs of the charge are set to zero for the reasons discussed above. Only two parameters remain for describing cooperative effects, namely kNa+ ,NH+ and 4
kCl− ,SO2− respectively. Preliminary evaluation of the present data 4
has shown that the parameter kNa+ ,NH+ can be set to zero with4
out significant loss in the accuracy of the correlation. Hence only one cross-interaction parameter remains per system for describing cooperative effects, namely kCl− ,SO2− . The model parameters 4
are presented in Table 1. As explained in [4], only three of the four parameters of the linear term are linearly independent. As recommended in [4], two parameters, one for an cation and one for a anion, are constrained to be equal. NH 4+ and Cl− are selected for that purpose.
E. Hackemann, H. Hasse / J. Chromatogr. A 1521 (2017) 73–79 Table 1 Parameters of the mathematical model (Eq. (4)) for the adsorption of BSA on Toyopearl PPG-600M at 25 ◦ C. k0 is set to −6.91 throughout. All parameters not listed below are set to zero.
I / mM pH
A 1500 4.0
B 2700 4.7
C 4200 7.0
kNa+ kNH+
1.97 1.66
0.95 0.76
0.57 0.26
kCl− kSO2−
1.66 3.50
0.76 3.39
0.26 2.71
4.59
0.00
−0.61
4
4
kCl− ,SO2− 4
3.3. Comparison of experimental and model results In the following, the experimental results for lnqp (cp = 0.01 mM) are compared to the model results obtained using the parametrization described in Section 3.2. In the remainder of this paragraph qp (cp = 0.01 mM) will simply be referred to as ‘loading’. In Fig. 3 results for the loading are shown as a function of the salt composition, which is represented here in the reduced parameter space introduced in Section 2.3. Three cases are shown: (A) pH 4.0 and I = 1500 mM, (B) pH 4.7 and I = 2700 mM, (C) pH 7.0 and I = 4200 mM. The red dots are the experimental values. The small gray dots represent the results from the mathematical model. They are points on a surface which will be called ‘model surface’ here. The blue crosses are vertical projections of the experimental data on that model surface. A good fit of the experimental results is obtained in all three systems A–C. The deviations between the experimental data and the model results are mostly within the experimental uncertainty. The discrepancies for the mixed salts are not above these for the pure salts. Not only the four binary systems which were used for adjusting the model parameters are well described but also the two other binary systems. The same holds for data for ternary salt mixtures for which no adjustment of model parameters was made. In Fig. 3A the results for pH 4.0 and I = 1500 mM are shown. In some of the binary salt systems linear behavior is observed whereas in others, positive cooperative effects occur. They are due to the cross-interactions between chloride and sulfate ions. The highest loadings are observed in binary mixtures containing high amounts of sodium chloride. The lowest loadings can be found in pure ammonium sulfate. Fig. 3B shows the results for pH 4.7, which is near the isoelectric point of BSA and I = 2700 mM. At those conditions, no important cooperative effects are observed, the linear terms in Eq. (4) are sufficient to describe the experimental results well. Independent of which salt or salt mixture is used the loading is of the order of 0.1 mM. Fig. 3C shows the results for pH 7.0 and I = 4200 mM. For those conditions, only negative cooperative effects are observed. There is a negative influence of using Cl− and SO 42− in combination. The highest loading is observed in pure sodium sulfate, the lowest loading in pure ammonium chloride. The results of the fit are discussed in the following, keeping in mind that the magnitude of the numbers for the ki parameters gives information on the influence of an individual ion on the loading whereas the magnitude of the numbers for kij give information on the influence of the cooperative ion effects. Positive numbers indicate a salting-out effect, negative numbers indicate a salting-in effect. The absolute numbers should not be overinterpreted as they depend on the way the normalization is carried out (here: kNH+ = kCl− throughout). However, it is 4
still possible to compare ratios between numbers for cations and anions. From this it is clear that for all three systems A, B, C, the salting-out effect of SO 42− is much stronger than that of Cl− , in perfect agreement with the well-known Hofmeister series [26]. The results also indicate a slightly stronger salting-out effect of Na+
77
compared to NH 4+. This is not in agreement with the Hofmeister series [26]. There is a strong positive cooperative effect of the combination of Cl− and SO 42− for pH 4.0 for which BSA has a positive net charge. This is interpreted here as a consequence of different types of positive sites on BSA, where certain types are more favorable for Cl− while other types are more favorable for SO 42−. The small negative cooperative effects which are observed for pH 7.0, for which BSA has a negative net charge, are interpreted here as a consequence of the influence of the solvent, i.e. we speculate that having both Cl− and SO 42− in the solution together with the corresponding cations and not only one type of anions, yields a more favorable solubility for BSA. At the isoelectric point of the protein at pH 4.7 nearly no cooperative effects occur as the protein has almost no net charge and therefore the interactions of the charged sites of BSA and the ions in the solution are less important than in the cases when BSA carries a net charge. Overall, summing up all parameters for the three cases shows that the influence of electrolytes on the BSA adsorption is strongest for pH 4.0. 3.4. Comparison of BSA and lysozyme It is interesting to compare the results from the present study on BSA with those for lysozyme from [4] as in both cases the same adsorbent, temperature and salts were used. The comparison is carried out for pH 7.0 as only for that pH data from both studies is available. Lysozyme carries a positive net charge at that pH whereas BSA carries a negative net charge. Furthermore an ionic strength must be selected for the comparison, which is carried out by comparing model results for both systems. For convenience, I = 3000 mM is chosen, which is in the range where experimental data is available for both systems. The results of the comparison are shown in Fig. 4 in the format which was already used above. In Panel A the results for BSA (blue) and lysozyme (red) are compared. The adsorption of lysozyme is stronger than that of BSA for all salt mixtures except those containing predominantly ammonium sulfate. Positive cooperative effects are observed for lysozyme, as they were observed in the present study for BSA at conditions where BSA is positively charged. However, for pH 7.0 for BSA negative cooperative effects occur, as discussed above. For process design the separation of proteins is of predominant importance. The information presented in Fig. 4A can be used for selecting conditions for which the separation of BSA and lysozyme is expected to be particularly favorable. For this, the ratio of the loading of the adsorbent for both proteins is calculated. The results are presented in Fig. 4B. They indicate that mixtures of sodium chloride and sodium sulfate with a strong excess of sodium chloride are particularly favorable. Of course, this optimization covers only one aspect of the separation process. It is mainly presented here to demonstrate the possibilities a model-based optimization offers in solvent design for protein separation in HIC. 4. Conclusion In the present work the combined influence of mixed electrolytes together with a buffer pH variation on hydrophobic interaction chromatography (HIC) is explored using bovine serum albumin BSA as a model protein. The pH is adjusted so that BSA carries either positive, negative or no net charge. Furthermore, BSA undergoes a structural change as result of the pH variation. Four salts, sodium chloride, ammonium chloride, sodium sulfate and ammonium sulfate, containing four different ions, are used. In the study, binary and ternary salt mixtures are used as well as the pure salts. For each studied pH, the salt composition is varied keeping the overall ionic strength constant at a suitably chosen value.
78
E. Hackemann, H. Hasse / J. Chromatogr. A 1521 (2017) 73–79
Fig. 3. BSA loading of the adsorbent at a BSA concentration in the solvent of 0.01 mM over the reduced parameter space describing the composition of the salt mixture. (A) pH 4.0 and I = 1500 mM, (B) pH 4.7 and I = 2700 mM, (C) pH 7.0 and I = 4200 mM. The big red dots are the experimental values. The small gray dots represent the surface resulting from the mathematical model adjusted to experimental data. The blue crosses are vertical projections of the experimental data on the model surface.
Fig. 4. BSA and lysozyme loading of the adsorbent at a protein concentration in the solvent of 0.01 mM over the reduced parameter space at pH 7.0 and I = 3000 mM. (A) BSA (red) and lysozyme (blue) show opposed behavior. (B) Difference lnqlysozyme − lnqBSA . The best separation is found for a mixed salt system of 2250 mM sodium chloride and 250 mM sodium sulfate.
Mixed salts can have positive, negative, or no cooperative effects. The magnitude and sign of the effect depends on the net charge of the protein. A mathematical model is used to correlate the data and interpret the observations. Despite the simplicity of that model, which is basically a multidimensional Taylor series expansion in the individual ion molarities, it correlates the comprehensive data taken in the present study well, with comparatively few parameters. It also gives good predictions for salt mixtures which were not included in the parameter fit. Particularly strong positive cooperative effects are observed when both Cl− and SO 42− are present in the solution and BSA is positively charged. This is attributed to the fact that different types of positive sites on BSA exist and some may be more favorable for Cl− while others may be more favorable for SO 42−. Whether this is a consequence of the different charge number of the chloride and the sulfate ions needs to be studied in future work. The present study also supports the approach of interpreting the influence of salts on HIC as resulting from the influence of sin-
gle ions, which opens more room for predictions as compared to interpretations relying on a salt, i.e. a certain combination of cation and anion. Finally, by comparing results for BSA from the present study and results for lysozyme from a previous study, it is demonstrated that the model can be used for selecting suitable conditions for protein separations. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.chroma.2017.09. 024. References [1] W. Melander, C. Horváth, Salt effects on hydrophobic interactions in precipitation and chromatography of proteins: an interpretations of the lyotropic series, Arch. Biochem. Biophys. 183 (1977) 200–215.
E. Hackemann, H. Hasse / J. Chromatogr. A 1521 (2017) 73–79 [2] T.W. Perkins, D.S. Mak, T.W. Root, E.N. Lightfoot, Protein retention in hydrophobic interaction chromatography: modeling variation with buffer ionic strength and column hydrophobicity, J. Chromatogr. A 766 (1997) 1–14. [3] M.E. Lienqueo, A. Mahn, J.C. Salgado, J.A. Asenjo, Current insights on protein behaviour in hydrophobic interaction chromatography, J. Chromatogr. B 849 (2007) 53–68, http://dx.doi.org/10.1016/j.jchromb.2006.11.019. [4] A. Werner, H. Hasse, Experimental study and modeling of the influence of mixed electrolytes on adsorption of macromolecules on a hydrophobic resin, J. Chromatogr. A 1315 (2013) 135–144, http://dx.doi.org/10.1016/j.chroma. 2013.09.071. [5] E. Mueller, J. Vajda, D. Josic, T. Schröder, R. Dabre, T. Frey, Mixed electrolytes in hydrophobic interaction chromatography, J. Sep. Sci. (2013) 1327–1334, http://dx.doi.org/10.1002/jssc.201200704. [6] Z.E. Rassi, L.F. de Ocampo, M.D. Bacolod, Binary and ternary salt gradients in hydrophobic-interaction chromatography of proteins, J. Chromatogr. A 499 (1990) 141–152, http://dx.doi.org/10.1016/S0021-9673(00)96969-3. [7] A.M. Senczuk, R. Klinke, T. Arakawa, G. Vedantham, Y. Yigzaw, Hydrophobic interaction chromatography in dual salt system increases protein binding capacity, Biotechnol. Bioeng. 103 (5) (2009) 930–935, http://dx.doi.org/10. 1002/bit.22313. [8] K. Baumgartner, S.A. Oelmeier, J. Hubbuch, The influence of mixed salts on the capacity of hic adsorbers: a predictive correlation to the surface tension and the aggregation temperature, Biotechnol. Prog. (2016), http://dx.doi.org/10. 1002/btpr.2166. [9] E. Hackemann, A. Werner, H. Hasse, Influence of mixed electrolytes on the adsorption of lysozyme, PEG, and PEGylated lysozyme on a hydrophobic interaction chromatography resin, Biotechnol. Prog. 33 (4) (2017) 1104–1115 https://doi.org/10.1002/btpr.2474. [10] S. Hjertén, K. Yao, K.-O. Eriksson, B. Johansson, Gradient and isocratic high-performance hydrophobic interaction chromatography of proteins on agarose columns, J. Chromatogr. 359 (1986) 99–109. [11] J. Fausnaugh, F. Regnier, Solute and mobile phase contributions to retention in hydrophobic interactions chromatography of proteins, J. Chromatogr. 359 (1986) 131–146. [12] R.A. Alberty, Maxwell relations for thermodynamic quantities of biochemical reactions, J. Am. Chem. Soc. 91 (14) (1969) 3899–3903, http://dx.doi.org/10. 1021/ja01042a037. [13] F. Xia, D. Negrath, S.M. Cramer, Effect of pH changes on water release values in hydrophobic interaction chromatographic systems, J. Chromatogr. A 1079 (2005) 229–235.
79
[14] P. Baumann, K. Baumgartner, J. Hubbuch, Influence of binding pH and protein solubility on the dynamic binding capacity in hydrophobic interaction chromatography, J. Chromatogr. A 1396 (2015) 77–85, http://dx.doi.org/10. 1016/j.chroma.2015.04.001. [15] D.C. Carter, J.X. Ho, Structure of serum albumin, in: F.M. Richards, C.B. Anfinsen, J.T. Edsall, D.S. Eisenberg (Eds.), Lipoproteins, Apolipoproteins, and Lipases Advances in Protein Chemistry, vol. 45, Academic Press, 1994, pp. 153–203, http://dx.doi.org/10.1016/S0065-3233(08)60640-3. [16] T.J. Peters, 2 – The albumin molecule: its structure and chemical properties, in: T. Peters (Ed.), All About Albumin, Academic Press, San Diego, 1995, pp. 9–11, http://dx.doi.org/10.1016/B978-012552110-9/50004-0. [17] A. Salis, M. Bostroem, L. Medda, F. Cugia, B. Barse, D.F. Parsons, B.W. Ninham, M. Monduzzi, Measurements and theoretical interpretation of points of zero charge/potential of BSA protein, Langmuir 27 (18) (2011) 11597–11604, http://dx.doi.org/10.1021/la2024605, pMID: 21834579. ´ U. Salgyn, ´ S. Bahadir, Zeta potentials and isoelectric points of [18] S. Salgyn, biomolecules: the effects of ion types and ionic strengths, Int. J. Electrochem. Sci. 7 (2012) 12404–12414. [19] R. Ueberbacher, E. Haimer, R. Hahn, A. Jungbauer, Hydrophobic interaction chromatography of proteins V. Quantitative assessment of conformational changes, J. Chromatogr. A 1198–1199 (2008) 154–163. [20] R. Ueberbacher, A. Rodler, R. Hahn, A. Jungbauer, Hydrophobic interaction chromatography of proteins: thermodynamic analysis of conformational changes, J. Chromatogr. A 1217 (2010) 184–190. [21] A. Jungbauer, C. Machold, R. Hahn, Hydrophobic interaction chromatography of proteins: III. Unfolding of proteins upon adsorption, J. Chromatogr. A 1079 (2005) 221–228, http://dx.doi.org/10.1016/j.chroma.2005.04.002. [22] A. Werner, T. Blaschke, H. Hasse, Microcaloric study of the adsorption of PEGylated lysozyme and peg on a mildly hydrophobic resin: influence of ammonium-sulfate, Langmuir 28 (2012) 11376–11383. [23] M.R. Oberholzer, A.M. Lenhoff, Protein adsorption isotherms through colloidal energetics, Langmuir 15 (1999) 3905–3914. [24] C. Leggio, L. Galantini, N.V. Pavel, About the albumin structure in solution: cigar expanded form versus heart normal shape, Phys. Chem. Chem. Phys. 10 (2008) 6741–6750, http://dx.doi.org/10.1039/B808938H. [25] K. Baler, O.A. Martin, M.A. Carignano, G.A. Ameer, J.A. Vila, I. Szleifer, Electrostatic unfolding and interactions of albumin driven by pH changes: a molecular dynamics study, J. Phys. Chem. B 118 (4) (2014) 921–930, http://dx. doi.org/10.1021/jp409936v, pMID: 24393011. [26] F. Hofmeister, Zur lehre von der wirkung der salze, Arch. Exp. Pathol. Pharm. 24 (4) (1888) 247–260, http://dx.doi.org/10.1007/BF01918191.