Anal. Ckem.1994,66, 2719-2725
Chiral Separations in Capillary Electrophoresis Using Human Serum Albumin as a Buffer Additive Ju Yang and Davld S. Hage’ Department of Chemistry, University of Nebraska, Lincoln. Nebraska 68588-0304 This study examined the theory and mechanisms of chiral separations in capillary electrophoresis based on the use of proteins as buffer additives. Human serum albumin (HSA) was used as the model ligand; D,L-tryptophan and (R,S)warfarin were used as the test analytes to be separated by this protein. Items examined in this work included the amount of HSA adsorbed to the capillary wall and the stability of this adsorbed protein layer. These were investigated by performing frontal analysis on the capillary with HSA and by injecting neutral markers through the capillary at different applied voltages before and after HSA treatment. The role of adsorbed HSA vs HSA in the buffer in determining the stereoselectivity of the CE system was also examined. Adsorbed HSA was the predominant agent involved in the separationof (R$)-warfarin, while HSA in the buffer had the most significant effect in the resolution of D,L-tryptophan. Two distinct separation mechanisms were proposed to explain these differences. Good agreement was obtained between the results predicted by these mechanisms and the experimental data. Under optimized conditions, both pairs of enantiomers were separated with baseline resolution in less than 12 min.
Table 1. Blndlng of Warfarln and Tryptophan E n a n t l o ” to HSA at pH 7.4 and 25 O C compound no. of sites site location binding const” (Mal)
(S)-warfarin (R)-warfarin L-tryptophan D-tryptophan
1 1 1 1
warfarin site warfarin site indole site not knownb
3.4 x 105 2.6 x 105 2.7 X 104 0.4 X 104
a Data taken from refs 17 and 18. Although the exact location of this site has not yet been identified, it is known that D-tryptophan docs not interact at the indole site but does have indirect interactions with the warfarin site.”
(1) Karger, B. L.; Cohen, A. S.;Guttman, A. J. Chromafogr. 1989, 492, 585. (2) McLaughlin, G.M.; Nolan, J. A.; Lindahl, J. L.; Palmieri, R. H.; Anderson, K. W.; Morris, S.C.; Morrison, J. A.; Bronzert, T. J. J. Liq. Chromufogr.
the protein can be either immobilized within the CE capillarygJOJ1or added to the running b~ffer.~JOA number of previous studies have examined the effects of different operating conditions on the resolution of protein-based CE separations,&’ but no in-depth examination of the mechanisms taking place in these separations has yet been reported. Such information would be valuable in providing guidelines for the development and optimization of new protein-based CE methods. This work will examine the theory of chiral CE separations based on protein buffer additives. Human serum albumin (HSA) is the model ligand used in these studies. The main advantage of using HSA as a model is that data are available for its interactions with a wide range of organic and inorganic compounds.I2 These interactions are believed to occur at five to six regions on HSA, the two most important of which are the warfarin- and indole-binding sites.13-15 Together, these two sites account for the majority of compound interactions with HSA.I3 It has been shown that binding at both of these sites can be stereoselective in nature,12J6 making them potentially useful in the separation of chiral molecules. Two sets of test analytes will be used in this study: D,Ltryptophan and (R,S)-warfarin. All of these compounds have well-characterized binding with HSA, with association equilibrium constants between lo3 and lo6 M-l at 25 OC (see Table 1). These values are representative of the range of association constants seen in the binding of HSA with many other c o m p o ~ n d s . l ~ One - ~ ~ attractive feature of these test
1992, 15, 961. (3) Gassmann, E.; Kuo, J. E.; Zarc, R. N. Science 1985, 230, 813. (4) Gozcl, P.; Gassmann, E.; Michelsen, H.; Zare, R. N. A n d . Chem. 1987,59, 44. (S)Guttman, A.; Paulus, A.; Cohen, A. S.;Grinberg, N.; Karger, B. L. J. Chromatogr. 1988, 448, 41. (6) Snopek, J.; Sioni, H.; Novotny, M.; Smolkova-Keulemansova, E.; Jelinek, I. J . Chromutogr. 1991, 559,215. (7) Altria, K. D.; Goodall, D. M.; Rogan, M. M. Chromofogruphia1992.34, 19. (8) Birnbaum, S.; Nilsson, S.Anal. Chem. 1992, 64,2872. (9) Barker, G.E.; Russo, P.; Hartwick, R. A. Anal. Chem. 1992.64, 3024. (IO) Lloyd, D. K. Presented at the Sixteenth International Symposium on Column Liquid Chromatography, 1992; Abstract 262.
(11) Li, S.;Lloyd, D. K. Anal. Chem. 1993, 65, 3684. (12) MOller, W. E.; Wollert, U. Phurmacology 1979. 19, 59. (1 3) Reidenberg, M. M., Erill, S.,Eds. Drug-Profein Binding, Pracgcr Publishers: New York, 1986. (14) Tillement, J.-P.;Houin,G.;Zini,R.;Uricn,S.;Alkngrcs,E.; Barre,J.;Lecomte, M.; D’Athis, P.; Sebille, B. Ado. Drug Res. 1984, 13, 59. (1 5) MBller, W. E.; Wollert, U. Nuuyn-Schmiedeberg’s Arch. Pharmacol. 1975, 288, 17. (16) Lagercrantz, C.; Larsson, T.; Karlsson, H. Anal. Biochem. 1979, 99, 352. (17) Yang, J.; Hage, D. S . J. Chromufogr. 1993, 645, 241. (18) Loun, B.; Hage, D. S.,manuscript in preparation. (19) Montero, M. T.; Estelrich, J.; Valls, 0. Inf. J. Pharm. 1990, 62, 21.
The use of capillary electrophoresis (CE) in the analysis of biological and pharmaceutical samples is an area of increasing One application of CE that has drawn particular attention is its use in the separation of chiral drugs and biological molecules. The first use of CE in chiral separations was demonstrated by Zare and co-workers, who resolved racemic mixtures of dansylated amino acids by adding Cu2+ and L-histidine or aspartame to CE running buffem3q4 More recently, chiral separations have been achieved through the use of cyclodextrins added to CE buffers or immobilized within CE ~apillaries.~-~ Another approach for obtaining chiral separations in CE is to use proteins as stereoselective binding agents. Examples include CE methods based on bovine serum a l b ~ m i n ,human ~,~ serum albumin,IO and al-acid glycoprotein.’ I In this method, ________
~~
0003-2700/94/036&2719$04.50/0 0 1994 Amerlcan Chemical Society
AnalyticalChemistry, Vol. 66, No. 17, September 1, 1994 2719
compounds is that they are all believed to have single-site binding with HSA. In addition, both of the major HSA binding regions, the indole and warfarin sites, are represented by the interactions of these analytes. Items that will be studied in the use of HSA as a buffer additive will include the amount of HSA deposited on the capillary wall and the stability of this adsorbed protein in the presence of an electric field. The role of the adsorbed HSA vs that in the running buffer in determining the stereoselectivity of the CE system will then be examined. The effects of HSA buffer concentration and applied voltage on the resolution of the warfarin and tryptophan enantiomers will also be studied. From this work, general guidelines and models will be developed that can be used to optimize chiral separations in CE based on either HSA or other protein additives. THEORY The following reaction can be used to describe the binding of an analyte (A) to a single site on a ligand (L) immobilized or adsorbed within a CE capillary: KI
A+L+A-L (1) In this equation, K1 is the association equilibrium constant for A with the immobilized ligand and A-L is the resulting complex. If there is no additional ligand present in solution, this capillary acts essentially as a chromatographic support in which compounds are moved through the column by their net electrophoretic mobilities. The capacity factor (k’) of A in this situation is given by the expressions22 k’= tr/tA - 1
(2)
= K P L / Vm (3) where tr is the migration time of A in the presence of L, t~ is the migration time of A in the absence of L, Vm is the capillary void volume, and mL is the total moles of active ligand present in the capillary. In this situation, the amount of active ligand can also be described in terms of its apparent concentration in the capillary (mL/ Vm). In eq 3 it is assumed that the loss of adsorbed ligand is negligible over the course of the analysis and that the amount of injected analyte is small vs the total number of available ligand-binding sites (i.e., linear elution conditions are present). If all or most of the ligand is present in the running buffer of the CE system, then the binding and migration of A is described by eqs 4-6. These expressions are similar to those K2
A+LeA-L
(4)
= K,[LI/(l + K,[LI)
(6)
used to describe the elution of compounds when they complex with cyclodextrins or micelles present in CE buffer^.^^-^^ In (20) Iwatsuru. M.; Nishigori, H.; Maruyama, K. Chem. Phorm. Bull. 1982, 30, 4489. (21) Loun, B.; Hage, D. S.J. Chromotogr. 1992, 579, 225. (22) Chaiken, I. M., Ed. Anofyticol Affinity Chromatography; CRC Press: Boca Raton, FL, 1987.
2720
Ana&t/ca/Chemistry, Vol. 66, No. 17, September 1, 1994
these equations, [L] is the buffer concentration of the ligand, K2 is the association equilibrium constant for the binding of A with L in solution, and k” is an index used to describe the relative migration rate of A through the capillary. The terms p~ and PAL represent the net electrophoretic mobilities of A in the absence of L and of A when it is completely complexed with ligand (Le., in the presence of high [L]). The term pr is the net electrophoretic mobility observed for A at intermediate levels of ligand. In eq 5, a k” value of 0 represents a migration velocity and elution time identical to that obtained by A in the absence of L. A k”va1ue of 1 represents a migration velocity and elution time that is the same as that seen with only the bound analyteligand complex. Values of k” between 0 and 1 can also occur, indicating that A is only partially bound to the ligand under the conditions used in the study. As in eqs 2 and 3, the relationship between eqs 5 and 6 assumes that the amount of injected analyte is small vs the total number of ligand-binding sites present in the CE system. Another assumption made in relating eqs 5 and 6 is that the values of p~ and PAL are constant with respect to ligand concentration. However, this assumption may not be valid if the addition of ligand to the running buffer causes a significant change in viscosity, thereby causing a shift in the mobilities observed for the analyte and analyte-ligand complex. Equations 5 and 6 can still be used in this situation by dividing the measured mobility of each peak by the mobility observed for a neutral compound (i.e., a marker for electroosmotic flow) injected under the same set of experimental conditions. Since the mobilities due to electroosmotic flow (pWm)and the electrophoretic migration of a solute are both inversely related to solvent v i s ~ o s i t ythe , ~ ~result is a series of normalizedvalues that are now independent of this parameter.26 The use of these normalized mobilities in eq 5 is convenient since it is equivalent to dividing both the numerator and denominator of this expression by pWm. The result is the same net relationship as shown previously. Equations 5 and 6 assume that the degree of solute binding with any adsorbed ligand present in the system is small compared with the binding of solute with ligand in the running buffer. However, even in this case it should be kept in mind that the presence of adsorbed ligand can still affect the net migration of solutes by the changes that it produces in the electroosmotic flow. Fortunately, this phenomenon does not have any significant effect on eq 5 since all of the net electrophoreticmobilities of the system (PLY, PA, PAL)are shifted by the same absolute amount. Since both the numerator and denominator of eq 5 represent a difference between two of these mobility terms, any shifts affecting only electroosmosis tend to cancel out and, thus, will not affect the final value of k” obtained with this expression. EXPER I MENTAL SECTION Reagents. The purified D-tryptophan, L-tryptophan, (R,S)warfarin racemic mixture, and human serum albumin (HSA; (23) Gareil, P.; Pernin, D.; Gramond,J.-P.;Guyon,F. J . High Resofur.Chromotogr. 1993, 16, 195. (24) Penn, S. G.; Goodall, D. M.; Loran, J. S. J. Chromatogr. 1993, 636, 149. (25) Camilleri, P., Ed. Copiflory Electrophoresis; CRC Press: Boca Raton, FL, 1993. (26) Yang, J.; Chaudhuri, S. R.; Hage, D. S., manuscript in preparation.
Catalog No. A-1653 or A-951 1) were purchased from Sigma (St. Louis, MO). Other chemicals used were from Fisher Scientific (Fair Lawn, NJ). All buffers and solutions were prepared using deionized water from a NANOpure water system (Barnstead, Dubuque, IA). Apparatus. All experiments were performed using an ISCO Model 3850 capillary electrophoresis system (Lincoln, NE) equipped with a circulating fan for temperature control. The CE capillaries were 50 pm i.d. X 70 cm untreated fused-silica columns (45-cm effective length) obtained from Polymicro Technologies, Inc. (Phoenix, AZ). Data were collected using a Chromlink interface and LCAdvantage software (LDC/ Milton Roy, Riviera Beach, FL). Methods. Samples were applied to the CE system using the vacuum injection mode supplied with the Model 3850 CE system. The applied voltage used in these studies varied from 10 to 30 kV (Le., from 0.14 to 0.43 kV/cm). The temperature monitored within the CE system during this work was 25 f 0.3 OC. Elution of D,L-tryptophan,acetone, and HSA was detected on-line at 280 nm. (R,S)-Warfarin was detected a t 307 nm. Since HSA absorbs at each of these wavelengths, there was some background signal due to the protein additive in both the tryptophan and warfarin studies. This was not a significant problem under the range of HSA concentrations used in this work (Le., 0-50 pM HSA). However, a decrease in the sensitivity and precision of the absorbance measurements was noted at higher HSA levels. All capillary columns were cleaned before protein application by applying 0.025 M sodium borate buffer (pH 10) for 1-2 h. HSA solutions were prepared in degassed 0.025 M phosphate buffer (pH 7.4) and sonicated for 5 min prior to application onto the C E system. As is true for samples containing any human-derived product, appropriate biohazard precautions should be taken in the handling and disposal of these types of solutions. The D-tryptophan, L-tryptophan, and (R,S)-warfarin solutions were prepared fresh daily, using 1 X lo4 M D-tryptophan, L-tryptophan, or racemic (R,S)warfarin dissolved in 0.025 M phosphate buffer (pH 7.4). A small amount of acetone was also added to each sample for monitoring electroosmotic flow. Injections of acetone onto an immobilized HSA HPLC column indicated that no detectable interactions were present between this compound and HSA. Samples were applied to the CE capillary using an injection time of 15 s. Migration times were calculated using the first statistical moment of each peak.27 Frontal analysis was performed by continuously applying 20-50 pM HSA in 0.025 M phosphate buffer (pH 7.4) to the capillary in the vacuum injection mode. Injection times of 1500 s or greater were used. After formation of the resulting breakthrough curve, 0.025 M phosphate buffer (pH 7.4) with no HSA additive was applied to wash away any excess or nonretained HSA in thecapillary. This process was continued for several additional cycles until no change in the response was observed. The amount of HSA required to saturate the CE capillary was determined by integrating each of the breakthrough curves2* and subtracting the results between (27) Grushka, E.; Myers, M.N.; Schettler, P. D.; Giddings, J. C. Anal. Chem. 1969, 41, 889. (28) Lund, U. J. Lip. Chromarogr. 1981, 4, 1933.
m
sn d
0
5
10
15
20
25
Time (min)
Figure 1. Frontal analysis using HSA applied to new and previously coated CE caplliaries. These results were obtained using 50 pM HSA.
the first and last runs. The flow rate of HSA solution during the vacuum injection was determined by similar application of a dilute solution of acetone to the capillary as a neutral, nonretained marker.
RESULTS AND DISCUSSION Determination of Adsorbed HSA. Previous studies have shown that protein adsorption in CE capillaries can be a significant problem at neutral or slightly basic pH.29+30If proteins are to be used as chiral buffer additives, then this adsorption should either be minimized or the effect of the adsorbed protein on the stereoselectivity of the system must be determined. To study the role of adsorbed HSA in such separations, it was first necessary to estimate how much HSA could absorb to a CE capillary. This was examined through frontal analysis by applying a known concentration of HSA to capillaries at a constant flow rate. Typical breakthrough curves for untreated and HSA-treated capillaries are shown in Figure 1. These curves were obtained with no electric field present and by use of vacuum injection for the application of HSA. Each time frontal analysis was performed on an untreated capillary, a relatively long breakthrough curve was initially obtained. But when the frontal analysis was repeated, all later runs showed similar curves occurring at somewhat shorter times. For example, duplicate runs on freshly cleaned capillaries gave initial breakthrough times of 18.62 f 0.01 ( f l SD) min, while the second and third runs on these capillaries gave an average breakthrough time of 16.8 f 0.3 min. One explanation for this is that HSA was irreversibly adsorbed to the capillary in its first application. This resulted in few or no adsorption sites being left for HSA in later runs, causing all subsequent breakthrough curves to appear at or near the void volume of the system. This explanation was supported by the fact that the similar application of a nonretained probe (acetone) on an uncoated capillary gave a breakthrough time of 16.5 min, a value statistically equivalent to that obtained in the second and third applications of HSA. From the frontal analysis results, the average amount of HSA adsorbed to two separate CE capillaries was determined (29) Towns, J. K.; Regnier, F. E. A n d . Chem. 1991, 63, 1126. (30) Green, J. S.;Jorgcnson, J. W.J. Chromarogr. 1989, 478, 63.
A~a!~ticaIChemistry,Voi. 66, No. 17, September 1, 1994
2721
I
x
II
I 0
1% x)
100
1%
Time (min)
Flgure 2. Changes In the mobility due to electroosmotic flow (p-) wlth tlme at dlfferent applled voltages for HSA-coated capillaries.
to be 2.9 (f0.5)X 10-8 mol/m2. For a totally smooth capillary, this represents a coverage of about 0.72 monolayer. This corresponded to an apparent concentration of adsorbed HSA in the capillary ( t n ~ / V , )of 2.7 p M . The conditions used in Figure 1 (Le., a neutral pH and low ionic strength buffer) represent a worst-case scenario with respect to protein adsorption. For example, it is known from previous work that such adsorption can be minimized by using highly acidic or basic solutionsor by adding high concentrations of salts to the running b ~ f f e r . 2 ~However, ~~0 these approaches will also significantly affect the binding and stereoselectivity of HSA. The buffer used in this study approximated physicological pH and ionic strength conditions,which produce optimum or near optimum binding for both of the sets of test solutes used in this work.17J8 This made it important to consider the protein adsorption that occurred when this particular buffer was used. Although the breakthrough studies indicated that HSA was binding irreversibly to the capillary in the absence of an electric field, this did not mean that irreversible binding was alsooccurring whenan electric field was present. Thestability of the adsorbed HSA in the presence of various applied voltages was examined by periodically injecting a neutral marker (acetone) to monitor the resulting shifts in electroosmotic flow. The data obtained are given in Figure 2. Figure 2 shows that all field strengths gave an approximately exponential loss of adsorbed HSA with time. The actual rate of loss varied with the electric field, with higher appliedvoltages producing faster losses of the protein coating. It was found that the loss of protein coating and the resulting changes in electroosmotic flow could be prevented by adding a small amount of HSA to the running buffer to replenish desorbed sites. For example, a newly coated HSA capillary in the presence of a 0.05 pM HSA solution showed less than 1% variation in the electroosmotic flow at 20 kV over a run time of 4 h. Under the same conditions but with no HSA in the buffer, a 20% change in the electroosmotic flow was observed in only 1 h. Chiral Separation of D,L-Tryptophan. D- and L-tryptophan were the first pair of model analytes studied in exploring the use of HSA in chiral CE separations. These analytes were of interest since the rapid separation of D- and L-tryptophan 2722
AnaiytIcalChemistry, Vol. 66, No. 17, September 1, 1994
~
0
5
10
15
2Q
HSA Buffer Conc. (vM)
Figure3. Changes Inthe net electrophoretic moblbs(circa) of acetone, @tryptophan, and L-tryptophan on an HSA-coated capUlary as HSA buffer concentratlon Is varied. Each result shown Is the average of three runs obtained at potentials ranglng from 15 to 25 kV. The error bars represent i l standard deviation of the mean.
has remained a difficult challenge for many years. Typical HPLC separations for these compounds have taken 15or more minutes to perform, with shorter analysis times being reported only in recent ~ o r k . ~ JFrom ’ earlier studies it is known that D- and L-tryptophan bind at two different sites on HSA and that the binding strengths of these sites differ by about 1 order of magnitude at room temperature (See Table 1.). The separation of D- and L-tryptophan was first examined by injecting each enantiomer separately onto the CE system in the presence of different field strengths and HSA buffer concentrations. Injections of acetone were also made under each set of conditions to monitor the electroosmotic flow. A 0.025 M phosphate buffer (pH 7.4) was used in this and all later studies since this has been shown to give the best separation of D- and L-tryptophan on immobilized HSA c01umns.l~ The net electrophoretic mobilities ( p ~ measured ~ ~ ) for acetone, D-tryptophan, and L-tryptophan at different HSA buffer concentrations are shown in Figure 3. The value of PNet for HSA under these same conditions was determined to be 1.01 X 10-6 m2/V min. A difference in the net migration rate of D- and L-tryptophan was observed at all HSA concentrations greater than 0 pM. Since the two enantiomers have the same inherent mobility when no HSA is present, the observed variation was attributed to differencesin their binding with HSA in the buffer. The minimum concentration of HSA needed to obtain a stable difference in the mobilities of D- and L-tryptophan was found to be 10 pM. This buffer concentration of HSA, or a higher amount, was used in all later work with D- and L-tryptophan. The effect of field strength on the separation of D- and L-tryptophan is shown in Figure 4, using an HSA buffer concentration of 10 pM. The degree of separation obtained for D- and L-tryptophan under these conditions varied considerably with the electric field used. At high applied voltages (25-30 kV), no resolution of the enantiomers was observed. As the voltage was decreased to 23 kV, D- and L-tryptophan began to separate, and at voltages below 23 kV, complete resolution was obtained. When 20 kV was used, Dand L-tryptophan were separated with baseline resolution (R,
R+S
HSA coeted Column + 20 I.IM HSA Buffer
HSA coeted Column No HSA In Buffer
1
0
I
I
0
I
I
I
4
2
I
I
4
I
I
E
6
I
.-
i I
0
5
10
I
I
4
I
I
6
I
I
I
E
I
10
I
I
I
12
,
,
14
I
10
Time (min) Figure 4. Elution of Dtryptophan (D), L-tryptophan (L), and acetone (A) in the presence of HSA in the running buffer at different applied voltages. The experlmentai conditions are given in the text. 30
2
15
HSA Buffer Conc. (pM) Flgure 5. Maximum applied voltages giving baseline resolution (Rs= 1.5) for D and L-tryptophan at various HSA buffer concentrations.
1 1.5) in less than 10 min. The much broader peaks seen for L-tryptophan than for D-tryptophan agree with behavior seen in earlier chromatographic studies.” Similar electric field studies were performed at other HSA buffer concentrations. Whenever 20 pM HSA or more was used, D- and L-tryptophan were separated with baseline resolution over the entire range of applied voltages examined (15-30 kV). However, at lower HSA concentrations the separation of D-and L-tryptophan could be obtained only when performed at or below a certain critical field strength. This is the same type of behavior as seen in Figure 4. Figure 5 shows the maximum applied voltages that gave baseline resolution for D- and L-tryptophan at the various HSA buffer concentrations studied. The importance in these separations of HSA in the buffer vs HSA adsorbed onto the capillary wall was studied by injecting mixtures of D- and L-tryptophan onto a capillary pretreated with HSA but using only 0.025 M phosphate buffer (pH 7.4) as the running buffer. Under these conditions, no separation was observed for all applied voltages tested (10-30 kV). This indicated that HSA in the buffer, and not adsorbed HSA, was responsible for separating D- and L-tryptophan.
Time (min) Figure 6. Injections of (&warfarin (R), (*warfarin (S), and acetone (A) onto a freshly coated HSA capillary with and without HSA present in the running buffer. The applied voltage was 23 kV.
Chiral Separation of (R,S)-Warfarin. The second model analyte system used in this study was (R,S)-warfarin. This system differs from D,L-tryptophan in several ways. As shown in Table 1, (R)and (S)-warfarin have much stronger binding to HSA than D,L-tryptophan and have a smaller difference in their relative binding strengths. Also, the warfarin enantiomers have the same binding site, as opposed to D- and L-tryptophan, which bind to two distinct sites. Initial studies examining theseparation of (R)and @)-warfarin using HSA in the CE running buffer were performed in the same manner as described for D,L-tryptophan. The concentrations of HSA used in these studies ranged from 5 to 50 pM and the applied potential ranged from 10 to 30 kV. A running buffer containing 0.025 M phosphate (pH 7.4) was again chosen in order to allow a direct comparison of the results with the data in Table 1. No separation of (R)-and (S)-warfarin was obtained at any of the HSA buffer concentrations or electric field strengths tested. The top portion of Figure 6 gives an example of one such study using the same conditions shown earlier to give baseline resolution for D- and L-tryptophan. Since (R)and (,!?)-warfarin could not be separated when moderate amounts of HSA were added to the running buffer, the use of HSA coated on the capillary wall was next considered. These results are shown in the lower portion of Figure 6 . With only pH 7.4 phosphate buffer in the system, it was found that baseline resolution of (R)and (S)-warfarin could easily be obtained. At an applied voltage of 23 kV, the enantiomers were separated in approximately 12 min on a freshly coated HSA capillary. The slightly wider peak seen for (S)-warfarin vs (R)-warfarin is similar to that seen when these compounds are injected onto immobilized HSA chromatographic supports.’* Recall that one problem with the use of adsorbed HSA is that there is a gradual loss of protein from the capillary in the presence of an electric field. One way of minimizing this loss is to avoid high applied voltages. For example, at least five consecutive baseline separations of (R,S)-warfarin could be performed at 23 kV, but the resolution rapidly decreased when the potential was increased to 30 kV. The rate of change in the resolution paralleled the rate of protein desorption seen Ana!vticalChemistry, Vol. 66, No. 17, September 1, 1994
2723
2.0
I
I
1.0
,
1
1,5
/
L-Tryptophan
I
D-Tryptophan 0.0
0
2
4
Apparent HSA Conc. (pM)
7 0
I
I
10
20
30
40
50
HSA Buffer Conc. (pM)
Flguro 7. Predicted capacity factors for (R,S)-warfarln and D,Ltryptophan in the presence of various amounts of Immobilized HSA. These plots were generated using eq 3 and the data in Table 1.
Flguro 8. Predicted relative migration indexes for (R,S).warfarin and o,L-tryptophan in the presence of various amounts of HSA In solution. These plots were generated using eq 8 and the data in Table 1.
earlier in Figure 2. At any of the potentials studied, the protein coating could be regenerated by periodically applying more HSA to the system or by adding a trace amount of HSA to the running buffer. However, care must be taken in using the later approach since the presence of too much HSA in the buffer can result in a loss of resolution. It was found that a running buffer containing 0.05 pM HSA was sufficient to maintain a stable separation of ( R ) -and (S)-warfarin (R, = 1.O) over the course of several hours. At slightly higher HSA buffer levels (0.4 pM),no observable separation was obtained. Comparison of Separation Mechanisms. The work with warfarin and tryptophan indicated that at least two different mechanisms were involved in the resolution of chiral compounds by HSA. For instance, the separation of ( R ) - and (S)-warfarin depended on HSA adsorbed to the capillary wall, but HSA in the buffer was required to separate D- and L-tryptophan. These two types of behavior were examined in more detail using eqs 3 and 6 and the data given in Table 1. Based on this information, it was possible to predict what types of retention or migration would have been expected for the warfarin and tryptophan enantiomers under the experimental conditions used in this study. Figure 7 shows the results predicted for when only adsorbed HSA was present in the CE capillary. The values predicted for the capacity factors of ( R ) - and (S)-warfarin with an apparent concentration of adsorbed HSA equal to 2.7 pM were within 20-30’31 of the initial capacity factors observed experimentally. The separation factor measured for (R)-and (S)-warfarin (a,where a = k’s/k’R) was within 10% of the value predicted from the data in Table 1 (1.4 f 0.1 vs 1.3 predicted). This indicated that there was good agreement between the experimental results and those predicted by a model in which ( R ) - and (S)-warfarin were interacting with the adsorbed HSA. Similar chromatographic interactions between protein solutes and modified CE capillaries have been reported in studies by Maa et aL31 In this work, the expected retention and values for k’ were greatest for those compounds with the highest affinities for HSA (Le., ( R ) -and (S)-warfarin). Although the association
constants for ( R ) -and (S)-warfarin are much closer together than those for D- and L-tryptophan, this model predicts that the stronger binding of (R)-and (S)-warfarin to HSA would produce a larger difference in their k’ values. These results agree with the observation made experimentally that adsorbed HSA could be used to separate (R)-and @)-warfarin but not D- and L-tryptophan. The system in Figure 7 could potentially be adapted for the separation of lower affinity compounds by increasing the apparent concentration of adsorbed HSA. In this study, the separation of (R)-and (S)-warfarin was achieved with capacity factors at or near 1.0. According to eq 3, a 10-100-fold increase in the apparent HSA concentration would be needed to obtain similar k’ values for D- and L-tryptophan. A 2-5fold increase in mL/V , could be obtained by using a smaller diameter capillary (Le., one that has a larger surface area/ volume ratio). An even larger increase (50-100-fold) could be reached by using capillaries packed with immobilized HSA supports. Figure 8 shows the results predicted for the separation of warfarin and tryptophan enantiomers, assuming that HSA was present only in solution. In the earlier experimental studies, some additional adsorbed HSA was also present, but its apparent concentration (2.7 pM) was much lower than the concentration of HSA in the buffer (5-50 pM). Assuming that the effect of this adsorbed HSA was small, Figure 8 indicates that high-affinity compounds, like ( R ) - and (S)warfarin, would be tightly bound to HSA over most of the concentration range shown (Le., have k“ values approaching 1). The result is that these compounds would tend to travel through the capillary with HSA, making them difficult to resolve. For lower affinity compounds, like D- and Ltryptophan, Figure 8 predicts a much different type of behavior. The smaller binding constants of these analytes would produce either intermediate rates of migration (L-tryptophan) or a net electrophoretic mobility only slightly different from that of the free analyte (D-tryptophan). This behavior would make it relatively easy to separate such compounds. These results agree with the experimental observation that D- and Ltryptophan could be separated at a variety of HSA buffer concentrations but that no separationof (R)-and @+warfarin
(31) Maa, Y.-F.; Hyver, 1991, 14, 65.
2724
K. J.; Swcdberg, S. A. J. High Resolut. Chromatogr.
AnatyticaIChemlstry, Vol. 66, No. 17, September 1, 1994
could be obtained when even a small amount of HSA (0.4 pM)was present in the running buffer.
CONCLUSION This work examined the mechanisms involved in the CE separation of D,L-tryptophan and (R,S)-warfarin using HSA as a buffer additive. Two possible separation mechanisms were considered: the first based on HSA adsorbed to the capillary wall and the second based on HSA in the running buffer. Equations were derived to describe the relative retention or migration of the analytes in both cases. Adsorbed HSA was found to be the predominant agent involved in the separation of (R,S)-warfarin, while HSA in the buffer had the most significant effect in the separation of D,L-tryptophan. These results showed good agreement with the predicted response and could be directly related to the different binding strengths of these compounds with HSA. Both the amount of protein and the applied voltage were considered in obtaining separations of (R,S)-warfarin and D,L-tryptophan. For separations based on adsorbed HSA, the amount of protein that could be placed on the capillary wall was estimated to be about 0.7 monolayer. This protein slowly desorbed in the presence of an electric field, with the rate of desorption increasing with field strength. However, the adsorbed protein could be stabilized by the addition of a trace amount of HSA to the running buffer. Separations based on HSA in the buffer were found to have a complex
dependence between the applied voltage and protein concentration. In this case, all voltages below a certain critical level gave baseline resolution, with the critical voltage increasing with protein concentration. Under the final optimum conditions used, both pairs of enantiomers were separated with baseline resolution in less than 12 min. The results of this study should be useful in the development and optimization of protein-based separations in CE, since no previous work has provided an in-depth examination of the mechanisms taking place in these types of methods. The data presented in this report indicate several general guidelines that can be used in the design of such separations. Although this work concentrated on the separation of warfarin and tryptophan enantiomers, the same techniques could be applied to the resolution of other chiral compounds that bind to HSA or alternative proteins.
ACKNOWLEDGMENT This work was supported by the National Institutes of Health under Grant GM4493 1. The Model 3850 CE system used in this study was the generous gift of ISCO, Inc. (Lincoln, NE). Received for review October 12, 1993. Accepted M a y 28, 1994.' Abstract published in Advance ACS Absrrclcfs, July 15, 1994.
Analytical Chemism, Vol. 66, No. 17, September 1, 1994
2725
