Thermodynamic Studies on the Solvent Effects In Chromatography

Molecularly imprinted polymers (MIPs) are used as highly enantioselective .... Theory ...... Allender, C. J.; Heard, C. M.; Brain, K. R. Chirality 199...
0 downloads 0 Views 171KB Size
Anal. Chem. 2005, 77, 1708-1717

Thermodynamic Studies on the Solvent Effects In Chromatography on Molecularly Imprinted Polymers. 1. Nature of the Organic Modifier Hyunjung Kim and Georges Guiochon*

Department of Chemistry, University of Tennessee, Knoxville, Tennessee 37996-1600, and Division of Chemical Sciences, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831-6120

Molecularly imprinted polymers (MIPs) are used as highly enantioselective stationary phases in liquid chromatography. To optimize the binding performance of MIPs, different types of polar modifiers are frequently used. Previous studies have shown that the hydrogen-bonding donor parameter (HBD) of the modifier has a large influence on the binding performance of MIPs in chiral separations. This possibility is addressed in a detailed thermodynamic study of a Fmoc-L-tryptophan (Fmoc-LTrp) imprinted polymer, eluted with four different polar modifiers, i.e., THF, propan-2-ol, methanol, and acetic acid, which have different HBDs (0.00, 0.33, 0.43, and 0.61, respectively). Adsorption isotherm data for each enantiomer in each of these organic modifiers were acquired by frontal analysis over a 20 000 dynamic concentration range. Nonlinear regression of the isotherm data, along with independent calculation of the affinity energy distributions, identified four different types of binding sites coexisting for the enantiomers on the MIP. The exception was acetic acid, which has the highest HBD. In this case, three types of binding sites only coexist on the MIP. The isotherm parameters obtained from these data show the following: (1) The association energies of the two enantiomers with a given type of sites have a similar magnitude; however, the density of the sites is higher for the template than for its antipode. (2) The nature of the organic modifier has a larger influence on the density of high-energy sites than on the association constant of these sites. (3) The molecular size of the organic modifier has a larger influence on the site density (especially for Fmoc-D-Trp) than does HBD. (4) Using an organic modifier with a higher HBD reduces the enantioselectivity on each site. (5) High-energy sites are more enantioselective than low-energy ones. (6) Using an organic modifier with a high HBD causes a larger reduction in the density of high-energy sites approached by the template molecules. The development of chiral stationary phases (CSPs) has become an important area of research in pharmaceutical chemistry. The availability of highly selective CSPs having also a high * To whom correspondence should be addressed. Fax: 865-974-2667. E-mail: [email protected].

1708 Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

saturation capacity is required to support the development, the production, and the quality assurance of the compounds that are needed in enantiomerically pure form to produce pharmaceuticals that exhibit high target specificity, few side effects, and high safety. In many cases, a CSP exhibiting a high selectivity for a certain enantiomeric pair of interest is needed. Molecularly imprinted polymers (MIPs) constitute a new generation of tailor-made CSPs that provide some unprecedented advantages in terms of their chemical and physical stability, their high specificity toward the template present during their polymerization, and their relative ease of preparation. The most widely applied method of preparation of MIPs is based on the formation of noncovalent interactions between a target molecule (the template) and some suitable functional monomers. These prepolymer complexes are then immobilized into a polymer matrix by copolymerization with a high concentration of cross-linking monomers in the presence of an initiator. After polymerization, the template is extracted from the polymer matrix, leaving behind cavities that have complementary size, shape, and functionalities toward the template and can readily adsorb it. The major problem encountered in the synthesis of MIPs by this method is that the binding sites on the MIPs are heterogeneous. In other words, several different types of binding sites, having significantly or very different association constants and different densities, coexist on the MIP surface. Thus, there is a strong need for the optimization of the synthesis method, to improve the binding performance of the MIPs and their surface homogeneity and to shift toward higher values the energy of the different types of sites. Some of these optimization attempts include an increase of the concentration of template-functional complexes in prepolymerization mixtures, to increase the density of high-affinity sites,1 various posttreatments on the MIPs, to obtain more homogeneous sites2-4 or to shift the site distribution toward higher affinity sites,5 and the optimization of the experimental (1) Molecular imprinted polymers. Man made mimics of antibodies and their application in analytical chemistry; Sellergren, B., Ed.; Elsevier: Amsterdam, 2001. (2) Szabelski, P.; Kaczmarski, K.; Cavazzini, A.; Chen, Y.-B.; Sellergren, B.; Guiochon, G. J. Chromatogr., A 2002, 964, 99-111. (3) Stanley, B. J.; Szabelski, P.; Chen, Y.-B.; Sellergren, B.; Guiochon, G. Langmuir 2003, 19, 772-778. (4) Chen, Y.; Kele, M.; Sajonz, P.; Sellergren, B.; Guiochon, G. Anal. Chem. 1999, 71, 928-939. (5) Umpleby, R. J.; Rushton, G. T.; Shah, R. N.; Rampey, A. M.; Bradshaw, J. C.; Berch, J. K., Jr.; Shimizu, K. D. Macromolecules 2001, 34, 8446-8452. 10.1021/ac040155f CCC: $30.25

© 2005 American Chemical Society Published on Web 02/08/2005

conditions used in chromatography, such as adjusting the mobilephase composition to improve peak shape as well as selectivity on the MIPs.6-8 Studies made with mobile phases of different compositions established that the optimized binding performance of the MIPs is obtained when the mobile phase is based on the same solvent as the one used for polymerization, with some amount of an organic modifier. The role of the organic modifier in the binding performance of a Boc-L-phenylalanine imprinted polymer was systematically studied by Allender et al.7 in order to gain better understanding of the retention mechanism. These authors showed that the organic modifier modulates the retention factor of the template by competitively interacting with it for access to the imprinted binding sites on the MIPs. This conclusion was reached by showing a linear correlation between the capacity factor of the template and the hydrogen-bonding donor parameter (HBD) of different organic modifiers, with the concentration of these organic modifiers ranging between 0.3 and 0.6 M. However, the capacity factors, which are measured by analytical, i.e., linear chromatography, combine the contributions of all the sites coexisting on the MIPs. Thus, a more detailed understanding of the influence of the organic modifier on the MIPs could be obtained by separating these different contributions. For this purpose, it is necessary to measure isotherm data for the enantiomers in each organic modifier studied, over a sufficiently wide concentration range. These data allow the derivation of quantitative values characterizing the separate influence of the organic modifier on the saturation capacities and the adsorption constants of each type of sites present on the MIPs. The isotherm data for different templates on the corresponding MIPs were commonly determined by a conventional static method (“batch rebinding”)1. However, this method makes it very slow to acquire the isotherm data over a wide range of concentrations, and it is less accurate than chromatographic methods.9 Among chromatographic methods, frontal analysis (FA) is the most accurate method to obtain isotherm data for stationary phases that exhibit slow mass-transfer kinetics, which is often the case of MIPs.9 The FA method was successfully employed for the characterization of the binding sites of different MIPs.2-4 The isotherm parameters are obtained by the nonlinear fitting of the isotherm data to an isotherm model. Another independent approach to determine the best values of the isotherm parameters from the adsorption data is the calculation of the affinity energy distribution by the expectation-maximization method.10-12 In this method, the raw experimental isotherm data are directly inverted into a distribution of the adsorption constants, without introducing any arbitrary information except for the classical assumption that the local adsorption isotherm follows Langmuir behavior. The usefulness of this method to decipher the adsorption mechanism and to validate an isotherm model, selected based on the (6) Sellergren, B.; Shea, K. J. J. Chromatogr. 1993, 635, 31-49. Sellergren, B.; Shea, K. J. J. Chromatogr. 1993, 635 (1), 31-49. (7) Allender, C. J.; Heard, C. M.; Brain, K. R. Chirality 1997, 9, 238-242. (8) Yu, C.; Mosbach, K. J. Chromatogr., A 2000, 888, 63-72. (9) Guiochon, G.; Golshan-Shirazi, S.; Katti, A. Fundamentals of preparative and nonlinear chromatography; Academic Press: London, 1994. (10) Stanley, B. J.; Guiochon, G. J. Phys. Chem. 1993, 97, 8098-8104. (11) Stanley, B. J.; Guiochon, G. Langmuir 1994, 10, 4278-4285. (12) Stanley, B. J.; Bialkowski, S. E.; Marshall, D. B. Anal. Chem. 1993, 65, 259-267.

regression of the isotherm data, was demonstrated for MIPs3,13 as well as for other stationary phases.14,15 For this study, we prepared a Fmoc-L-tryptophan (Fmoc-L-Trp) imprinted polymer. Then, we investigated the influence of four different organic modifiers on the isotherm parameters for the enantiomers. The isotherm parameters were obtained from the isotherm data acquired by frontal analysis over a 20 000 dynamic concentration range, with subsequent isotherm regression and calculation of the affinity energy distribution. Acquiring the isotherm parameters provided the possibility to separate and quantify the adsorption contributions of the different organic modifiers on each of the sites present on the MIP. THEORY Isotherm Models. After the linear isotherm model, the second simplest isotherm model used in adsorption studies is the Langmuir isotherm,

q ) qsbC/(1 + bC)

(1)

where q and C are the equilibrium concentrations in the stationary and the mobile phases, respectively, qs is the monolayer saturation capacity, and b is the adsorption constant. The Langmuir model assumes that the surface of the adsorbent is homogeneous, being covered with only one type of adsorption site, that these sites are localized, and that there are no adsorbate-adsorbate interactions. The associated adsorption energy distribution is unimodal with an infinitely narrow mode. Simple extensions of the Langmuir isotherm model account for the adsorption behavior of many heterogeneous surfaces used in HPLC. These models assume that the surface is covered with patches of two or a few different types of independent adsorption sites. The main models used in this work belong to this group and are the bi-Langmuir (eq 2), the tri-Langmuir (eq 3) and the tetra-Langmuir (eq 4) isotherm models:

qs1b1C qs2b2C + 1 + b1C 1 + b2C

(2)

qs1b1C qs2b2C qs3b3C + + 1 + b1C 1 + b2C 1 + b3C

(3)

qs2b2C qs3b3C qs4b4C qs1b1C + + + 1 + b1C 1 + b2C 1 + b3C 1 + b4C

(4)

q)

q)

q)

where qs1, qs2, qs3, and qs4 are the saturation capacities for the first, the second, the third, and the fourth types of sites, respectively, and b1, b2, b3, and b4 are the corresponding adsorption constants. The energy distributions of these models display two, three, or four isolated peaks which are infinitely narrow. In practice, these modes are merely narrow, well resolved, and their small width does not cause any significant deviation from the Langmuirian behavior. (13) Kim, H.; Guiochon, G. Submitted for publication. (14) Stanley, B. J.; Krance, J. J. Chromatogr., A 2003, 1011, 11-22. (15) Go ¨tmar, G.; Zhou, D.; Stanley, B. J.; Guiochon, G. Anal. Chem. 2004, 76, 197-202.

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

1709

Affinity Energy Distribution. The heterogeneous nature of the affinity or selective sites on MIPs mandates that detailed investigations on the binding sites of MIPs are only possible by obtaining the affinity energy distribution of the template and its enantiomer. The affinity energy distribution (AED) gives the distribution of the values of the association constants for each possible adsorption site on the studied polymer. Among the several possible methods used to derive the AED from only the set of experimental isotherm data (q(C)), the expectation maximization (EM) method seems to be the most successful for the investigation of the binding sites on MIPs.3,13 The EM method allows the calculation of the adsorption energy distribution from the raw experimental isotherm data, without making any prior assumption regarding the isotherm model or the shape of the AED.10-12 The apparent, global, or overall adsorption isotherm (q(C)), which is measured experimentally, is related to the local isotherm and to the distribution of adsorption constants by the following equation:

q(C) )





-∞

f (ln K)θ(K, C)d ln K

(5)

where K is the adsorption constant, f (ln K) is the adsorption constant distribution (or density of the sites with association constant K), and θ(K, C) is the local adsorption isotherm, which is assumed here to follow Langmuir isotherm behavior. In the EM method, eq 5 is discretized into M

q(C) )

∑f (ln K)θ(K , C )∆(ln K) j

i

(6)

j)1

where Ci is the experimental value of the mobile-phase concentration, Kj is the adsorption constant, spanned continuously across the discretized grid, ∆(ln K) is the constant spacing between ln K values in the discretized space, and M is the total number of grid points in the space (i.e., the number of measurements made). All values of ln K in the discretized grid are divided up using two input parameters, Kmin and Kmax, and the number of points to be calculated for the distribution. The values of Kmin and Kmax are determined experimentally as the reciprocal of the largest, Cmax, and the smallest, Cmin, values of the mobile-phase concentration used in the measurements, respectively:

Kmin ) 1/Cmax

and

Kmax ) 1/Cmin

The EM method consists of the numerical calculation of the distribution (f (ln Kj)) by successive approximations. An initial estimate of the distribution is required. To minimize the amount of information introduced in the calculation of the distribution, the following constant initial distribution was used.

f (ln Kj) )

qhigh - qlow ln Kmax - ln Kmin

(7)

where qhigh and qlow are the highest and the lowest experimental values of the adsorbed amounts measured, respectively. In 1710

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

principle, they correspond to Cmax and Cmin, respectively. In this study, (0, 0) was included as the initial data point (i.e., qlow ) 0). Equation 7 provides a constant distribution across the entire adsorption constant range as the initial estimate. The EM method compares the ratio of the experimental value of q (qexp) and of the calculated value (qcal) at each data point and provides a vector that is used iteratively to correct the estimate of the isotherm. This vector is then convoluted with the local isotherm model to obtain the following correction factor (β) for the distribution: N

βj )

qexp(Ci) θ(Kj, Ci)∆(ln K) cal(Ci)

∑q i)1

(8)

where N is the number of data points and i and j are indices for the concentration and the adsorption constant, respectively. The distribution function is then corrected by multiplying f (ln Kj) by the correction factor given by eq 8 at each ln Kj

f (ln Kj)k ) β(ln Kj)f (ln Kj)k-1

(9)

where k is the iteration rank and f (ln Kj)k-1 is the previous estimate of the distribution. The iteration continues by recalculating eq 6 with the new estimate (expectation) and correcting the distribution using eqs 8 and 9 (maximization) until the change between two successive estimates is below a set threshold. EXPERIMENTAL SECTION Chemicals. Fmoc-L-Trp and Fmoc-D-tryptophan (Fmoc-D-Trp) were purchased from Novabiochem (San Diego, CA). Ethylene glycol dimethacrylate (EGDMA), 4-vinylpyridine (4-VPY), and 2,2azo-bis(isobutyronitrile) (AIBN) were obtained from Aldrich (Milwaukee, WI). 4-VPY and EGDMA were distilled under vacuum (60 mmHg, 75 °C, and 60 mmHg, 120 °C, respectively). All other chemicals and solvents used were commercially available, of analytical or HPLC grade, and were used as is. Preparation of the MIP Stationary Phase and Packing of the Column. The components of the polymerization mixtures were as follows: 1.58 mmol of Fmoc-L-Trp, 4.74 mmol of 4-VPY, 18.96 mmol of EGDMA, 0.474 mmol of AIBN, and 5.4 mL of acetonitrile (MeCN). The amount of solvent was 4/3 of the total volume of the monomers and the cross-lnking monomers. The solution was purged with N2 for 5 min in a scintillation vial and polymerized at 45 °C for 12 h. After polymerization, the bulk polymers were crushed, ground, and sieved to obtain particles within the size range of 25-38 mm. These polymer particles were slurry packed into a stainless column (100 × 4.6 mm). To remove any residues from the polymerization mixture and from the template from the polymers, the column packed with the MIP was exhaustively washed with methanol/acetic acid (4:1 v/v). Apparatus. The isotherm data were obtained by the frontal analysis method, using a Hewlett-Packard (Palo Alto, CA) HP 1090 liquid chromatograph. This instrument is equipped with a multisolvent delivery system (tank volumes, 1 dm3 each), an autosampler with a 250-µL sample loop, a diode-array UV detector, a column thermostat, and a computer data acquisition station. The

microcomputer of this system was used to program the series of breakthrough curves needed. Experimental Measurement of Breakthrough Curves, the Holdup, and the Extra Column Volumes. The two pumps of the HP1090 solvent delivery system were used to generate breakthrough curves at different mobile-phase concentrations of the two solutes in frontal analysis. One of the pumps delivered the pure mobile phase; the other pump delivered a solution of the compound studied. The concentration of this compound in the solution upstream of the breakthrough front was determined by its concentration in the sample solution and by the flow rate fractions delivered by the two pumps, at constant flow rate. This ratio was changed at set time intervals. The concentration range of the compound studied was between 0.005 and 100 mM, providing isotherm data points in a 20 000 dynamic linear range. The FA experiments were carried out using successively four different organic modifiers (THF, propan-2-ol, methanol, acetic acid), with the same base solvent (acetonitrile). For each different mobile phase, 33 consecutive breakthrough curves were recorded, with a sufficiently long delay time (30-100 min) between each breakthrough curve to allow the reequilibrium of the column with the pure mobile phase. The injection time of the sample solution was between 10 and 40 min to ensure that the composition of the eluate at the end of the run was the same as that of the plateau injected at the column inlet. The signal was detected, depending on the concentration range, between 260 and 310 nm, to avoid recording any signals having an absorbance above 1500 mAU. The holdup time, t0, of the MIP column was measured by injecting a small amount of acetone into the column. The phase ratio (F ) (volume of the stationary phase)/(volume of the mobile phase) in the column) was calculated from t0, the flow rate, and the geometrical volume of the column, giving F ) 0.307. The correction for the extracolumn volume was measured by injecting a small amount of the acetone from the pump into a “zero” dead volume connector instead of the column. A value of tx ) 0.911 min was obtained. The experimental data were corrected by subtracting tx. Calculation of Isotherm Data and Isotherm Model Selection. Adsorption isotherm data represent the amount of the bound substrate (q, mmol/L) for each mobile-phase concentration of the substrate (C, mmol/L). To calculate the amount of bound substrate from the breakthrough curve, the following equation was used:

Vequ - V0 Va

q)C

(10)

where Vequ is the elution volume of the equivalent area of the solute, V0 is the holdup volume, and Va is the volume of the stationary phase in the column. The value of Vequ for each breakthrough curve was calculated from the half-height method. The value of Va was calculated by subtracting V0 from the geometrical volume of the column. Nonlinear regression of the experimental data to the adsorption isotherm models was performed using Origin 6.0 (Northampton, MA). The experimental data were fit to each isotherm model with weights (1/q2) designed to put an even emphasis on each data point during the regression process. The best isotherm parameters

were selected by minimizing the residual sum of squares (RSS) for each isotherm model. The different adsorption isotherm models were compared using the Fisher test (Fcal) and the RSS values.17 Calculation of Affinity Energy Distribution. The EM algorithm programmed in Fortran (Lahey/Fujitsu Fortran 95, Incline Village, NV) was used to calculate the AEDs. One hundred grid points in the K-space were used to logarithmically digitize the range between Kmin ) 0.0001 mM-1 and Kmax ) 10 000 mM-1. The algorithm performed 106 iterations. The calculated AED from this algorithm shows the differential saturation at each point (qs(Kj)) as a function of ln Kj, where qs(Kj) ) f (lnKj)∆ ln K). To estimate the isotherm parameters from the calculated AED, the values of qs comprising one peak were added to obtain the saturation capacity of the corresponding type of sites, and the value of ln Kj for this peak was used to calculate the adsorption constant of the corresponding site. RESULTS AND DISCUSSION Adsorption Isotherms. The FA adsorption isotherm data of Fmoc-L-Trp and Fmoc-D-Trp were determined separately, in each of four different organic modifiers (at a constant concentration of 3.7 M), in a 20 000 linear dynamic concentration range of these solutes. These data are reported in Figure 1a-d. which correspond to different concentration ranges. The data for Fmoc-L-Trp (filled symbols) and Fmoc-D-Trp (open symbols) show clear separation of the two compounds in all concentration ranges. It is clear also that the isotherms are not linear even in the lowest concentration range, with any of the organic modifiers used, except acetic acid between 0.005 and 0.1 mM (Figure 1a). For Fmoc-L-Trp, the amount adsorbed at any given concentration increases in the order of acetic acid < propan-2-ol < THF < methanol. For Fmoc-D-Trp, the amount adsorbed increases in a slightly different order, acetic acid < THF < propan-2-ol < methanol. In both cases, the order of increasing amount adsorbed does not exhibit a linear correlation or even a qualitative correlation with the HBD of the organic modifiers. The highest amounts of either enantiomer adsorbed are observed with methanol. The amounts of enantiomers adsorbed from methanol (HBD ) 0.43) solutions are larger than those from either THF (HBD ) 0.00) or propan-2-ol (HBD ) 0.33) solutions, and both are markedly larger than the amount adsorbed from the acetic acid (HBD ) 0.61) solution.18 It seems that the size of the organic modifier molecule also influences the amount of either enantiomer adsorbed at a given concentration. The size of these molecules increases in the order methanol (Mw ) 32) < propan-2-ol or acetic acid (Mw ) 60) < THF (Mw ) 72). The small-size organic-modifier molecules do not seem to be able to experience molecular interactions with the MIP that would interfere with the interactions of the template or even of its enantiomer, which have both much larger molecules. However, the HBD of the organic modifier seems to influence somewhat the differential amount of Fmoc-L-Trp and Fmoc-D-Trp adsorbed at equilibrium. The difference between these two amounts decreases in the order THF > propan-2-ol or methanol > acetic acid, following the reverse of the order of the HBD (0.00, (16) Jaroniec, M.; Madey, R. Physical adsorption on heterogeneous solids; Elsevier: New York, 1988. (17) Quin ˜ones, I.; Guiochon, G. J. Chromatogr., A 1998, 796, 15-40. (18) Abraham, M. H. Chem. Soc. Rev. 1993, 22, 73-83.

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

1711

Figure 1. Adsorption equilibrium isotherms of Fmoc-L-Trp (filled symbols) and Fmoc-D-Trp (open symbols) in THF (circle), propan-2-ol (square), methanol (triangle up), and acetic acid (triangle down). The lines represent the best calculated tetra-Langmuir isotherm for Fmoc-L-Trp (solid lines) and Fmoc-D-Trp (dotted lines) in THF, propan-2-ol, and methanol. In acetic acid, the lines represent the best calculated tri-Langmuir isotherm for the two enantiomers. Concentration range: (a) between 0.005 and 0.1, (b) 0.005 and 1, (c) between 0.005 and 10, and (d) 0.005 and 100 mM.

0.33, 0.43, and 0.61, respectively). For example, for a mobile-phase concentration of 0.10 mM, the differential amounts adsorbed were 7.8, 2.0, 2.3, and 0.1 mM in THF, propan-2-ol, methanol, and acetic acid, respectively. Similar observations can be made at all concentrations (Figure 1b-d). In Figure 1b (concentration range, 0.005-1 mM), the isotherm data for the two enantiomers in acetic acid exhibit a slight curvature. The relative difference between the amounts of the two enantiomers adsorbed at equilibrium decreases with increasing concentrations of the two enantiomers. Finally, in the highest concentration range (between 0.005 and 100 mM, Figure 1d) the isotherms of Fmoc-L-Trp and Fmoc-DTrp in each organic modifier become close to each other. In summary, comparisons of the isotherms of the two enantiomers in the whole concentration range show that the amount of each enantiomer adsorbed does not correlate with the HBD of the organic modifiers. The amount of enantiomers adsorbed seems rather to increase with increasing size of the organic modifier molecule, especially for Fmoc-D-Trp. In contrast, the difference between the amounts of the two enantiomers adsorbed seems to follow the order of the HBD of the organic modifiers. Increasing this HBD seems to decrease this difference. In relative 1712 Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

terms, this difference decreases with increasing solute concentration. The lines in Figure 1a-d represent the best binding isotherm model corresponding to the set of symbols for the four different modifiers. The parameters of these isotherms were derived on the basis of statistical tests from the nonlinear regression of the isotherm data to each different isotherm model (i.e., RSS and Fisher parameters (Fcal)) and the independent calculation of the affinity energy distribution. Table 1 compares the statistical results obtained with these different models. In most cases, the lowest RSS and the highest Fcal were obtained with the tetra-Langmuir isotherm model for the two enantiomers. The exceptions were when the organic modifier was acetic acid for both enantiomers and THF for Fmoc-D-Trp, in which cases it was more probably a tri-Langmuir isotherm model. In these cases, when the statistical tests could not identify one unique isotherm model, the results of the calculation of the affinity energy distribution can be used to narrow down this selection process. Panels a-d in Figure 2 show the calculated affinity energy distributions for Fmoc-L-Trp (solid lines) and Fmoc-D-Trp (dotted lines) in THF, propan-2-ol, methanol, and acetic acid, respectively. Except for acetic acid, four

Table 1. Residual Sum of Squares (RSS) and Fisher Parameters (Fcal) bi-Langmuir

tri-Langmuir

tetra-Langmuir

substrate

organic modifier

RSS

Fcal

RSS

Fcal

RSS

Fcal

Fmoc-L-Trp

THF propan-2-ol methanol acetic acid THF propan-2-ol methanol acetic acid

272.8 630.2 2429.4 96.44 3113.2 6364.7 1620.2 596.8

144.3 160.4 108.7 630.8 35.6 116.95 142.8 114.1

23.5 73.3 236.1 28.7 21.4 36.71 283.3 10.8

2179.8 1277.9 1055.2 2009.9 2527.99 1587.24 758.3 5937.5

14.3 43.6 3.72 20.4 20.4 7.82 7.42 17.6

2438.6 1977.6 62846.1 1773.0 2373.91 8056.96 26731.9 3431.4

Fmoc-D-Trp

Table 2. Isotherm Parameters from Nonlinear Fitting of the Isotherm Data to the Selected Isotherm Model and from the Calculation of Affinity Energy Distribution for Fmoc-L-Trp and Fmoc-D-Trp on the Fmoc-L-Trp Imprinted Polymers in 3.75 M Concentration of Different Organic Modifier in Acetonitrile-Based Mobile Phasea isotherm parameters organic modifier

q1 (mM)

b1 (mM-1)

THF propan-2-ol methanol acetic acid

232(249) 315.9(336.7) 480(480) 406(386)

0.0129(0.0137) 0.0106(0.009) 0.0082(0.0073) 0.0039(0.0039)

THF propan-2-ol methanol acetic acid

317(301.6) 378(387) 563(552) 684(nd)b

0.0111(0.0111) 0.009(0.009) 0.0033(0.0032) 0.0012(nd)

b3 (mM-1)

q4 (mM)

b4 (mM-1)

Substrate: Fmoc-L-Trp 36.7(30.2) 0.08459(0.09) 3.627(3.47) 22.7(21.7) 0.1212(0.1369) 2.456(2.46) 73.3(85.6) 0.079(0.073) 4.99(5.76) 7.68(6.38) 0.257(0.194) 0.064(0.047)

3.104(3.162) 2.947(3.162) 2.73(2.08) 29.51(38.98)

0.509(0.612) 0.578(0.489) 0.579(0.724)

120.5(90.1) 46.610(48.0638) 119.83(90.063)

Substrate: Fmoc-D-Trp 6.95(6.75) 0.409(0.3899) 0.489(0.468) 17.9(17.6) 0.1403(0.1369) 0.729(0.737) 172.4(169.3) 0.0385(0.0389) 3.51(3.68) 98.3(121.2) 0.017(0.017) 0.047(0.041)

4.902(4.806) 6.243(5.926) 2.251(2.081) 9.81(3.16)

0.083(0.086) 0.155(0.153) 0.308(0.375)

48.36(48.06) 166.9(136.9) 132.8(90.1)

q2 (mM)

b2 (mM-1)

q3 (mM)

a The values in parentheses correspond to the isotherm parameters obtained from the calculation of affinity energy distribution. b nd, isotherm parameters from affinity energy distribution cannot be determined due to divergence.

distinct peaks on the AEDs are clearly seen for the two enantiomers. However, in acetic acid, which has the highest HBD, the AEDs exhibit only three peaks. In summary, the statistical tests and the affinity energy distributions provide consistent evidence that, within the concentration ranges investigated, four types of binding sites coexist for the enantiomers on the MIP in THF, propan-2-ol, and methanol and that three types of binding sites coexist for the enantiomers on the MIP in acetic acid. The affinity energy distributions shown in Figure 2a-d illustrate qualitatively the influence of the different organic modifiers on the retention of the MIP. In THF, propan-2-ol, and methanol (Figure 2a-c), there are four different types of binding sites. Each one interacts with the two enantiomers and has a binding energy of similar magnitude. However, there is a larger number of high-energy sites that interact with imprinted FmocL-Trp than with its antipode, Fmoc-D-Trp. The sites of the first type (1) interact in the same way with both enantiomers, at K1 ≡ 0.01 mM-1, although their binding constant is significantly lower with acetic acid. The sites of the second type (2) have a binding constant similar for both enantiomers, at K2 ≡ 0.1 mM-1. The average binding energies of the sites of types 1 and 2 seem to decrease slightly for Fmoc-D-Trp, in the order THF < propan-2-ol < methanol < acetic acid. The sites of the third sites (3) also have nearly the same binding constant for both enantiomers, at K3 ≡ 1 mM-1, and the energy of these sites does not seem to change with the different organic modifiers. The highest energy sites have also nearly the same binding constant for both enantiomers, at K4 ) ≡100 mM-1, and an energy that does not seem to change with the organic modifier. However, the density

of sites of each type clearly increases, especially for Fmoc-D-trp, in the order THF > propan-2-ol > methanol, in agreement with the previous observations on the amount of enantiomer adsorbed at equilibrium that were made in the discussion of the isotherm data (Figure 1a-d) and the influence of the size of the molecule of organic modifier on the density and energy of the different binding sites for Fmoc-D-trp on the MIP. However, for the MIP imprinted with Fmoc-L-Trp, the density of each type of site does not show consistent changes, indicating a mixed influence of both the HBD and the size of the organic modifier molecules on the binding sites. In contrast with what observed with the first three organic modifiers, THF, propan-2-ol, and methanol, there are only three types of binding sites on the MIP that interact in acetic acid. The highest energy sites (with K3 ≡ 40 mM-1) have a very low density for the template Fmoc-L-Trp and a negligible one for its enantiomer. These high-energy sites have a lower association constant and a lower density than those observed with the other organic modifiers. Similarly, the other two types of sites identified for the enantiomers have a range of association constants rather different from those measured with the other organic modifiers. Influence of the Organic Modifiers on the Isotherm Parameters. The isotherm parameters obtained from the nonlinear regression of the data and from the affinity energy distributions are summarized in Table 2. First, note the good agreement between the isotherm parameters obtained from these two independent methods. In acetic acid and in the other three modifiers, different types of sites coexist on the MIP (i.e., interactions are best accounted for by a tri-Langmuir isotherm Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

1713

1714

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

Figure 2. Affinity energy distributions for Fmoc-L-Trp (solid line) and Fmoc-D-Trp (dotted line) on the Fmoc-L-Trp imprinted polymer in different mobile phases. Acetonitrile with 3.75 M of modifier. (a) THF; (b) propan-2-ol; (c) methanol; (d) acetic acid. (a′-d′) are expanded views of (a-d), respectively. They show the high-energy sites that have a low density. The insets in these four figures are expanded views of the highest energy mode. The indices 1, 2, 3, and 4 represent the energy modes identified based on the affinity energy distribution for Fmoc-L-Trp.

Figure 3. Plots versus the HBD of the modifiers of (a) the enantiomeric selectivity ((qi,Lbi,L)/(qi,Dbi,D)), (b) the ratio of the numbers of sites (qi,L/qi,D), and (c) the ratio of the association constants (bi,L/bi,D) for the two enantiomers on each type of site. The indices 1, 2, 3, and 4 shown in the figure correspond to those in Figure 1a-d and Figure 2a-d.

model in acetic acid and a tetra-Langmuir model in the other modifiers). There is no clear correlation between the high-energy sites observed in the two cases. For this reason, we discuss separately the influence of the organic modifiers and of acetic acid on the isotherm parameters obtained for each type of site. In general, however, it is clear that, in acetic acid (the modifier with the highest HBD), we measured the lowest density and the lowest association constant for the highest energy sites found on the MIP. Second, we note in Table 2 that the density of each type of site for the template, Fmoc-L-Trp, increases in the order propan2-ol < THF < methanol. For Fmoc-D-Trp, the density of each type of site increases in the order THF < propan-2-ol < methanol. Both orders do not correlate with that of HBD for the modifiers. The concentration of organic modifiers used in this study is higher (3.7 M) than that (0.3-1.5 M) used in the similar previous study.7 This difference may explain the difference between the results obtained. It was necessary to use such a high concentration of organic modifier, especially for methanol, because when a lower concentration of 1 M of the organic modifiers was used, the retention of the Fmoc-Trp enantiomers was too high to allow accurate measurements by FA of the data points, especially in the lower concentration range. At the highest concentration of organic modifiers (1.5 M) used in the previous study, however, its results were in agreement with our own results. This previous

study provides data showing that the capacity factors for imprinted Boc-L-phenylalanine increases in the order propan-2-ol < THF < methanol at the highest concentration used (1.5 M). In contrast, at the lowest concentration of organic modifiers used (0.3 M), the capacity factors of Boc-L-pheylalanine increase with decreasing HBD, i.e., in the order methanol < propan-2-ol < THF. These results indicate and support our reasoning that the size of the molecules of the organic modifiers also influences the interactions between the substrates and the MIP and that this influence becomes more significant at high concentrations of the organic modifiers. In contrast to the trend observed regarding the density of each type of site in each organic modifier, no consistent trend was observed for the association constant on each type of site in the different organic modifiers, except for the lowenergy types of sites (types 1 and 2 in Figure 2a-d). The association constants of the enantiomers on the lowest energy sites decrease with increasing HBD of the organic modifiers. On the higher energy types of sites, no significant and consistent trend can be observed for the association constants of the enantiomers with the different organic modifiers. Thus, these results indicate that, for the higher energy sites, the nature of the organic modifier has a larger influence on the density of the high-energy sites than on the association constant on these sites. Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

1715

Figure 4. Correlation between the HBD of the organic modifiers and the fractional saturation capacities for (a) type of sites 3 and (b) type of sites 4, as identified from the nonlinear regressions (Figure 1a-d) and the affinity energy distributions (Figure 2a-d). Table 3. Fractional Differential Saturation Capacities (qx/qt) for Each Enantiomer on Each Site in the Different Organic Modifiersa THF (0.00)

propan-2-ol (0.33)

methanol (0.43)

site (Kmin - Kmax (mM-1))

Fmoc-L-Trp

Fmoc-D-Trp

Fmoc-L-Trp

Fmoc-D-Trp

Fmoc-L-Trp

Fmoc-D-Trp

1 (0.0015-0.015) 2 (0.015-0.15) 3 (1.5-5) 4 (40-200)

0.88 0.107 0.0112 0.0022

0.98 0.022 0.0015 0.00028

0.93 0.063 0.0088 0.0014

0.95 0.043 0.0018 0.00038

0.84 0.149 0.0036 0.001

0.76 0.233 0.0051 0.00052

a The values in parentheses next to the name of the modifier show HBD values. q represents the differential saturation capacity on each x corresponding site, and qt represents the sum of the differential saturation capacities on all of the identified sites.

Since we reported earlier that the trends regarding the amounts of each enantiomer adsorbed at equilibrium on each type of site, hence the difference between these two amounts are different for each one of these types of sites, we determined the selectivity of each type for Fmoc-L-Trp relative to Fmoc-D-Trp using the following equation:

Ri ) qi,Lbi,L/qi,Dbi,D

(11)

where i is the index of the type of site, as identified in Figure 2a-d, qi,L and qi,D are the corresponding saturation capacities, and bi,L and bi,D are the corresponding association constants. We attempted to correlate the enantiomeric selectivity with other solvent parameters such as the dielectric constant, the hydrogen bond donor parameters, the dipole moment, and the electrostatic factor.7 However, no correlation was observed between the enantiomeric selectivity and any of these solvent parameters, except for a correlation between the enantiomeric selectivities of the four types of sites and the HBD values of the organic modifiers, as shown in Figure 3a. Increasing the HBD of the organic modifier decreases the enantiomeric selectivity, especially those of the high-energy sites of types 3 and 4. Figure 3a shows also that the enantiomeric selectivity increases with increasing energy of the sites. Panels b and c in Figure 3 illustrate the correlation between the HBD of the modifiers and the ratio of the numbers of sites for the two enantiomers (qi,L/qi,D) and of their association constants (bi,L/bi,D), on each type of site. Comparing these figures clearly shows that the enantiomeric selectivity 1716

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

depends on the HBD of the modifier essentially through the influence of HBD on the relative numbers of sites for the two enantiomers while this influence on the relative association constant is negligible. The decrease in the relative number of sites available for the two enantiomers seems to be due to the decrease in the density of the high-energy sites with increasing HBD of the organic modifier. Figure 4 shows the correlation between the fractional saturation capacities (q′i ) qs,i/qT) of the high-energy sites (Figure 3) and the HBD of the modifiers. The fractional saturation capacity was calculated as the ratio of the saturation capacity for each type of site to the sum of the total saturation capacities of all the identified types of sites (Figure 2a-d), as qT ) qs1 + qs2 + qs3 + qs4. Table 3 summarizes these results and shows that for the types of sites 3 and 4, which exhibit significant enantiomeric selectivities (see Figure 3), q′ for the imprinted Fmoc-L-Trp decreases with increasing HBD of the organic modifier while it increases for its antipode, Fmoc-D-Trp, in the same time. These trends are illustrated in Figure 4. In conclusion, using a modifier of higher HBD decreases the enantiomeric selectivity of the higher energy types of sites because using an organic modifier having a higher HBD reduces the density of higher energy sites on which the three-dimensional structure of the template is memorized. The increasing fractional number of higher energy sites for Fmoc-DTrp might be due to the fact that the shape of these sites is not the three-dimensional complement of the Fmoc-D-Trp molecule and, thus, that these sites interact with Fmoc-D-Trp only by noncovalent interactions. Then, the small-size organic modifiers

cannot effectively interfere with the noncovalent interactions between the binding pockets on the MIP and the substrate while larger size organic modifiers can interfere. CONCLUSION The results reported in this paper provide thermodynamic explanations for the influence of the organic modifiers on the binding sites found on the MIP. A nonlinar regression of the isotherm data and the independent calculation of the affinity energy distribution allowed the identification of four types of binding sites on the MIP studied for the template and its enantiomer, in three different organic modifiers. However, in acetic acid, which has the highest HBD, the MIP exhibits only three types of binding sites, indicating that the distribution of the binding sites on the MIP can be controlled using different types of mobile phases. Each of the binding sites identified for the enantiomer on the MIP has association constants of similar magnitude, but the density of these sites is larger for the imprinted Fmoc-L-Trp than for its antipode Fmoc-D-Trp. Isotherm parameters obtained using the two independent methods show that the different organic modifiers have a larger influence on the density of each type of site than on the association constant for each type of site. The density of each type of site on the MIP decreases with increasing size of the organic modifier molecules. The effectiveness of the organic modifiers at interfering with the interactions between the enantiomers and the MIP seems to increase with their increasing molecular size. However, an almost linear correlation between the HBD of the organic modifiers and the enantiomeric selectivity on each site was observed. Using an organic modifier with a higher HBD decreases the enantiomeric selectivity on each type of site and a larger enantiomeric selectivity

is observed on the higher energy type of sites. This decreased enantiomer selectivity is explained by the reduced fractional number of the higher energy sites available for the template, Fmoc-L-Trp, when organic modifiers having a higher HBD are used. The results of this study raised important questions that should be answered to further understand the recognition process used with molecular imprinted polymers. First, the similar magnitude of the adsorption energy observed for both enantiomers on each type of site identified indicates that the binding sites made on the MIP are not independent for the two enantiomers. Acquiring and analyzing competitive binding isotherms from mixtures of the two enantiomers might help in clarifying this issue. Second, we observed that different types of binding sites exist on a MIP, depending on the nature of the organic modifier used (i.e., acetic acid versus propan-2-ol). Studying the dependence of the affinity distribution on the concentration of different organic modifiers and investigating the origin and importance of experimental errors made on the affinity distribution can clarify this issue. These observations will orient our forthcoming work. ACKNOWLEDGMENT This work was supported in part by Grant CHE-02-44693 of the National Science Foundation, by Grant DE-FG05-88-ER-13869 of the U.S. Department of Energy, and by the cooperative agreement between the University of Tennessee and the Oak Ridge National Laboratory. Received for review August 26, 2004. Accepted December 12, 2004. AC040155F

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

1717