Adsorption on Molecularly Imprinted Polymers of Structural Analogues

These two factors control also the MIP affinity toward the enantiomers of the structural analogues that have a stereochemistry different from that of ...
0 downloads 4 Views 189KB Size
Download PDF
Anal. Chem. 2005, 77, 6415-6425

Adsorption on Molecularly Imprinted Polymers of Structural Analogues of a Template. Single-Component Adsorption Isotherm Data 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

The equilibrium adsorption isotherms on two otherwise identical polymers, one imprinted with Fmoc-L-tryptophan (Fmoc-L-Trp) (MIP), the other nonimprinted (NIP), of compounds that are structural analogues of the template were acquired by frontal analysis (FA) in an acetonitrile/ acetic acid (99/1 v/v) mobile phase, over a wide concentration range (from 0.005 to 50 mM). These analogues were Fmoc-L-tyrosine, Fmoc-L-serine, Fmoc-L-phenyalanine, Fmoc-glycine (Fmoc-Gly), Fmoc-L-tryptophan pentafluorophenyl ester (Fmoc-L-Trp(OPfp)), and their antipodes. These substrates have different numbers of functional groups able to interact with the 4-vinylpyridine groups of the polymer. For a given number of the functional groups, these substrates have different hydrophobicities of their side groups (as indicated by their partition coefficients (log Pow) in the octanol-water system (e.g., from 4.74 for Fmoc-Trp to 2.53 for Fmoc-Gly)). Statistical results from the fitting of the FA data to Langmuirian isotherm models, the calculation of the affinity energy distribution, and the comparison of calculated and experimental band profiles show that all these sets of FA data are best accounted for by a tri-Langmuir isotherm model, except for the data of Fmoc-L-Trp(OPfp) that are best modeled by a simple Langmuir isotherm. So, all compounds but Fmoc-L-Trp(OPfp) find three different types of adsorption sites on both the MIP and the NIP. The properties of these different types of sites were studied systematically. The results show that the affinity of the structural analogues for the NIP is controlled mostly by the number of the functional groups on the substrates and somewhat by the hydrophobicity of their side groups. These two factors control also the MIP affinity toward the enantiomers of the structural analogues that have a stereochemistry different from that of the template. In contrast, the affinity of the highest affinity sites of the MIP toward the enantiomers of these structural analogues that have the same stereochemistry as the template is highest for the imprinted molecule (Fmoc-L-Trp). The separation of the template from the substrates with the same stereochemistry is influenced by the number of the functional groups on the substrates that can interact with * To whom correspondence should be addressed. Fax: 865-974-2667. Email: guiochonutk.edu. 10.1021/ac050914+ CCC: $30.25 Published on Web 09/08/2005

© 2005 American Chemical Society

the highest affinity sites on the MIP. The separation of the enantiomers of the analogues of the substrates was also achieved on the MIP, and these enantiomeric separations are influenced by the hydrophobicity of the substrates. Molecular imprinting is used to prepare adsorbents that are highly selective for the template used. These adsorbents are mostly employed as packing materials for high-performance liquid chromatography (HPLC), for solid-phase extraction, or for the manufacturing of sensors.1-3 The most common method of synthesis of these materials uses noncovalent interactions between a target molecule (the template) and some suitable functional groups. These interactions allow the formation of templatefunctional monomer complexes in solution. These complexes are then immobilized into a polymer matrix by copolymerization with a high concentration of cross-linking monomers, resulting in molecular imprinted polymers (MIPs). Sites having size, shape, and functionalities that are complementary toward those of the template are formed on the MIP surface by extracting the template from the MIP matrix after the polymerization process. In contrast with the expected high selectivity for the template that is suggested by the MIP concept, the literature reports that MIPs bind to compounds where the molecules of which have a structure closely related to that of the template and do so with an affinity that is close to their affinity for the template.4-11 It was also reported that MIPs can separate the enantiomers of compounds that are structurally related to a template12-14 (a phenom(1) Turiel, E.; Martin-Esteban, A. Anal. Bioanal. Chem. 2004, 378, 1876-1886. (2) Kandimalla, V. B.; Ju, H. Anal. Bioanal. Chem. 2004, 380, 587-605. (3) Wulff, G. Chem. Rev. 2002, 102, 1-27. (4) Sun, R.; Yu, H.; Luo, H.; Shen, Z. J. Chromatogr., A 2004, 1055, 1-9. (5) Turiel, E.; Martin-Esteban, A.; Fernandez, P.; Perez-Conde, C.; Camara, C. Anal. Chem. 2001, 73, 5133-5141. (6) Turiel, E.; Perez-Conde, C.; Martin-Esteban, A. Analyst 2003, 128, 137141. (7) Cacho, C.; Martin-Esteban, A.; Perez-Conde, C.; Camara, C. J. Chromatogr., B 2004, 802, 347-353. (8) Pap, T.; Horvai, G. J. Chromatogr., A 2004, 1034, 99-107. (9) Wu, L.; Li, Y. J. Mol. Recognit. 2004, 17, 567-574. (10) O’Mahony, J.; Molinelli, A.; Nolan, K.; Smyth, M. R.; Mizaikoff, B. Biosens. Bioelectron. 2005, 20, 1884-1893. (11) Martin, P. D.; Wilson, T. D.; Wilson, I. D.; Jones, G. R. Analyst 2001, 126, 757-759. (12) Sabourin, L.; Ansell, R. J.; Mosbach, K.; Nicholls, I. A. Anal. Commun. 1998, 35, 285-287. (13) Allender, C. J.; Brain, K. R.; Heard, C. M. Chirality 1997, 9, 233-237.

Analytical Chemistry, Vol. 77, No. 19, October 1, 2005 6415

enon that is referred to as cross-reactivity). However, a clear understanding of the origin of cross-reactivity is still lacking, and the literature also contains contradictory results. For example, Sun et al.4 have investigated the cross-reactivity on a Boc-L-tryptophan MIP of substrates such as Cbz-L-Trp and Ace-L-Trp, which are structurally similar to the template, by measuring the retention factors at infinite dilution of these substrates on the MIP. In their study, they found that the largest capacity factor was obtained for the template but that higher capacity factors were observed for bulkier substrates (i.e., Cbz-L-Trp) than for less bulky ones (i.e., Ace-L-Trp). They attributed this result to an increased hydrophobic interaction of the bulkier substrate with the cavities of the MIP. In contrast to the results of this study, rather different observations were reported in studies of the cross-reactivity on a propazine MIP.5,6 In this case, the isotherm data of the substrates were measured by batch rebinding and the best isotherm parameters obtained by fitting the isotherm data to the LangmuirFreundlich (LF) isotherm model. The authors concluded that the number of adsorption sites on the MIP estimated from the LF isotherm model increases with decreasing molecular size of the substrate. They explained their results by the easier accessibility of the molecules of smaller substrates to the binding sites of the MIP. These contradictory results might originate from the use, to characterize the heterogeneous binding sites of the MIP, of a method that gives only average characteristics of the heterogeneous sites of the MIP (i.e., the retention factors or initial slope of the LF isotherm model). To better understand the crossreactivity observed on MIPs, we need a method that permits the separate characterization of each type of binding site that can be identified on the heterogeneous surface of a MIP. In previous studies,15-18 we showed that the combination of the nonlinear regression of isotherm data to an isotherm model, of the independent calculation of the affinity energy distribution, and of the comparison of the calculated and experimental band profiles in overloaded elution provides means to select the most appropriate isotherm model and to characterize each of the identified types of binding sites on the heterogeneous surface of a MIP. The purpose of this paper is to apply these methods to characterize heterogeneous binding sites on a MIP, in order to better understand the cross-reactivity that is frequently observed on MIPs. For this purpose, we used a Fmoc-L-tryptophan (FmocL-Trp) imprinted polymer and its reference polymer (i.e., the nonimprinted polymer or NIP) and acquired on these polymers the adsorption isotherm data for a series of structural analogues of the template and their enantiomers. We used HPLC frontal analysis (FA) for this data acquisition. The isotherm model that best accounts for each set of isotherm data was selected using (1) the classical statistical tests on the results of the nonlinear regression of the isotherm data to several possible isotherm models, (2) the independent calculation of the affinity energy distribution, and (3) the comparison of calculated and experimental band profiles of overloaded elution bands. The numerical values of the isotherm parameters of each substrate provided by these methods were then used to analyze and compare the adsorption

behavior of each substrate on the MIP and on the corresponding NIP. The results of this study illustrate the influence of several factors on the cross-reactivity of the structural analogues of the template and of their enantiomers.

(14) Spivak, D. A.; Simon, R.; Campbell, J. Anal. Chim. Acta 2004, 504, 23-30. (15) Kim, H.; Guiochon, G. Anal. Chem. 2005, 77, 1708-1717. (16) Kim, H.; Guiochon, G. Anal. Chem. 2005, 77, 1718-1726. (17) Kim, H.; Guiochon, G. Anal. Chem. 2005, 77, 2494-2504. (18) Kim, H.; Kaczmarski, K.; Guiochon, G. Submitted.

(19) Guiochon, G.; Shirazi, S. G.; Katti, A. M. Fundamentals of Preparative and Nonlinear Chromatography; Academic Press: Boston, MA, 1994. (20) Stanley, B. J.; Guiochon, G. Langmuir 1995, 11, 17035-1743. (21) Stanley, B. J.; Guiochon, G. Langmuir 1994, 10, 4278-4285. (22) Stanley, B. J.; Guiochon, G. J. Phys. Chem. 1993, 97, 8098-8104.

6416

Analytical Chemistry, Vol. 77, No. 19, October 1, 2005

THEORETICAL BACKGROUND To study the interactions between a MIP, the template, and its enantiomer, we use a systematic approach developed for the study of heterogeneous surfaces.15-18 This approach involves the determination of the adsorption isotherms of the two enantiomers on the MIP and the corresponding NIP in a wide concentration range, the modeling of these isotherm data, and the calculation of the adsorption affinity distribution (AED) on the two polymer surfaces. The isotherm data are acquired by FA, a chromatographic technique that is accurate and fast.19 The isotherm modeling is done by fitting the data to a variety of isotherm models19 and using a series of conventional statistic tests to compare the quality of the fits. The AED gives the density of adsorption sites as a function of the value of their equilibrium constant.20-22 It characterizes the degree of heterogeneity of a surface. We briefly report in this section the fundamental basis of these basic tools and refer the readers to more elaborate treatments.15 - 22 Isotherm Models. The models that best fit our data are derived from the Langmuir isotherm model. Three models fit the different sets of adsorption data acquired by FA, the bi-Langmuir (eq 1), the tri-Langmuir (eq 2), and the tetra-Langmuir (eq 3) isotherms.

qs2b2C qs1b1C + 1 + b1C 1 + b2C

(1)

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

(2)

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

(3)

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 adsorption sites, respectively; and b1, b2, b3, and b4 are the corresponding adsorption constants. The basic rational behind these models is that, although the adsorbent surface is heterogeneous, it is actually paved with series of domains (two to four) that are nearly homogeneous. Affinity Energy Distribution. Because the surface of a MIP is heterogeneous, a detailed investigation of the properties of the different binding sites benefits from the knowledge of the AED of the template and of its enantiomer and both the MIP and the NIP. There are several methods to derive the AED from a set of experimental isotherm data (q(C)). The expectation maximization method (EM)20-22 allows this calculation directly, from the raw experimental isotherm data, without making any prior assumption regarding either the isotherm model or the shape of the AED. We used it.

Modeling of Overloaded Band Profiles. An empirical but important and sensitive test of the validity of an isotherm model for a compound in a phase system consists of comparing the band profiles recorded for samples of widely different sizes with those calculated using a proper chromatographic model and the selected isotherm. The results of the calculation are quite sensitive to the isotherm model, and discrepancies with experimental data are informative.19 The profiles of the overloaded elution bands of the two enantiomers studied were calculated using the lumped pore diffusion model (POR), selected for reasons explained elsewhere.23 The POR model takes into account the finite rate of the masstransfer kinetics involved in the chromatographic process. Detailed theoretical information on this model is available elsewhere.23-26 EXPERIMENTAL SECTION The equipment, the methology, and the procedure used in this work were previously described.15-18 Only the particular features of the experimental conditions that are critical to understand our present study are given below. Experimental Procedure for Frontal Analysis: Acquisition of the Breakthrough Curves. The isotherm data of the different substrates on the two polymer columns were acquired by FA at 25 ( 2 °C, using a mobile phase made of acetonitrile with 1% acetic acid. The conventional wide rectangular injection procedure of FA was followed.19 A solution of the studied compound at a known concentration in the mobile phase is pumped through the column until the plateau of the breakthrough curve begins to elute (1040 min in our experiments). The composition of this stream is adjusted through the gradient function of the chromatograph by setting the ratio of pure mobile phase and a mother solution mixed and pumped to the column. After at least 5 min, the adsorbed substrate is washed off the column with a stream of the pure mobile phase until the absorbance of the eluent has decreased to the same baseline as before the beginning of the first measurement (60-120 min). The feed composition is then increased, and a new breakthrough curve is acquired. Depending on the compound used and its concentration range, the detector signal is recorded at between 260 and 310 nm, to avoid signals corresponding to more than 1500 mAU. Finally, before changing the solute and acquiring a new series of breakthrough curves, the column is washed with a 4/1 methanol/acetic acid solution for at least 2 h at a flow rate of 1.0 mL/min. Column Characteristics. A polymer imprinted with Fmoc-Ltryptophan (dp, 25-38 µm) was prepared by thermal polymerization of 4-vinylpyridine (as a functional monomer) and ethylene glycol dimethacrylate (as a cross-linking monomer) in a solution of the template in acetonitrile (porogen).15 The NIP was prepared by the same procedure, omitting the template. Stainless steel columns (10 cm × 0.46 cm) were packed with each polymer. The holdup time for the columns (t0), at a flow rate of 1.0 mL/ min, was measured by injecting a small amount of acetone into the column with the autosampler. After substrating the contribution of the extracolumn volume from the autosampler to the (23) Kim, H.; Kaczmarski, K.; Guiochon, G. Chem. Eng. Sci. 2005, 60, 54255444. (24) Kaczmarski, K.; Antos, D.; Sajonz, H.; Sajonz, P.; Guiochon, G. J. Chromatogr., A 2001, 925, 1-17. (25) Antos, D.; Kaczmarski, K.; Piatkowski, W.; Seidel-Morgenstern, A. J. Chromatogr., A 2003, 1006, 61-76. (26) Kim, H.; Guiochon, G. J. Chromatogr., A, in press.

column (tx ) 0.007 min), the total porosity of the column (volume of mobile phase divided by geometrical volume of the column) was calculated from t0. The values obtained were 0.75 for the MIP column and 0.78 for the NIP column. Calculation of the Isotherm Data and Isotherm Model Selection. Adsorption isotherm data represent the amount of substrate (q, mmol/L) bound to the column, in equilibrium with a known mobile-phase concentration of the substrate (C, mmol/ L). This amount is derived from the retention time of the breakthrough curve using the equal area method.15-18 The isotherm data were fitted to a series of isotherm models.15-18 Depending on the data set, the best isotherm was one of those in eqs 1-3. The qualities of the fits to each isotherm model were compared by calculating the standard deviation of each isotherm parameter, the Fisher parameter, and the residual sum of squares. Calculation of Affinity Energy Distribution. The EM algorithm was used to calculate the AEDs of each substrate studied. The AEDs are reported later as the number of sites, qs(Kj, versus the corresponding association constant, ln Kj. The calculation give Kj in the range between Kmin ) 0.000 01 mM-1 and Kmax ) 200 mM-1 corresponding to the range of concentrations within which the adsorption data were acquired. The isotherm parameters were calculated from the AED. The saturation capacity on each identified type of site, corresponding to a mode of the AED, was calculated by adding the values of qs comprising the corresponding peak. The adsorption constant of the corresponding site was calculated from the value ln Kj for the peak maximum. Experimental and Calculated Peak Profiles. The experimental elution band profiles of large samples of each substrate on the MIP and on the NIP were recorded at the end of the acquisition of each set of FA data. Solutions of the substrate were injected from the pump, at six different concentrations ranging approximately between 0.1 and 50 mM. The band profiles were recorded between 260 and 310 nm and the absorbance data (mAU) converted into concentrations (mM), using a calibration curve derived from the plateaus of the breakthrough curves. The corresponding profiles were calculated using the POR model and the isotherm parameters.26 RESULTS AND DISCUSSION The structural analogues of the template have a similar structural outlook but different numbers of functional groups able to interact with the functional 4-vinylpyridine groups of the polymers: There are two such functional groups in Fmoc-Ltyrosine (Fmoc-L-Tyr) (carboxylic acid and phenol groups) and Fmoc-L-serine (Fmoc-L-Ser) (carboxylic acid and aliphatic hydroxy groups); one carboxylic acid group in Fmoc-L-Trp, Fmoc-Lphenyalanine (Fmoc-L-Phe), and Fmoc-glycine (Fmoc-Gly); and none in Fmoc-L-tryptophan pentafluorophenyl ester (Fmoc-L-Trp(OPfp). For the substrates that have one functional group, the octanol-water partition coefficient (log Pow) decreases in the following order: Fmoc-L-Trp > Fmoc-L-Phe > Fmoc-Gly. The symbols in Figure 1a-c show the FA adsorption isotherm data on the NIP for Fmoc-L-Trp (b), Fmoc-L-Tyr (9), Fmoc-L-Ser (2), Fmoc-L-Phe (O), Fmoc-Gly (0), and Fmoc-L-Trp(OPfp) (4). The three figures cover the whole concentration range, Figure 1a between 0.005 and 0.12 mM, Figure 1b between 0.005 and 1.4 mM, and Figure 1c between 0.005 and 70 mM. The overall affinity (i.e., the amount of adsorbed substrate) on the NIP decreases in Analytical Chemistry, Vol. 77, No. 19, October 1, 2005

6417

Figure 1. Adsorption equilibrium isotherms of structural analogues of Fmoc-L-Trp template on the NIP and the MIP: Fmoc-L-Trp (b); FmocL-Tyr (9); Fmoc-L-Ser (2); Fmoc-L-Phe (O); Fmoc-Gly (0); Fmoc-L-Trp(OPfp) (4). The lines represent the best calculated Tri-Langmuir isotherm for each substrate. (a) Isotherms on the NIP. Concentrations between 0.005 and 0.12 mM. (b) Isotherms on the NIP. Concentrations between 0.005 and 1.4 mM. (c) Isotherms on the NIP. Concentrations between 0.005 and 70 mM. (d) Isotherms on the MIP. Concentrations between 0.005 and 0.12 mM. (e) Isotherms on the MIP. Concentrations between 0.005 and 1.2 mM. (f) Isotherms on the MIP. Concentrations between 0.005 and 70 mM.

the following order: Fmoc-L-Tyr > Fmoc-L-Ser > Fmoc-L-Trp > Fmoc-L-Phe > Fmoc-Gly > Fmoc-L-Trp(OPfp). The differences in overall affinity for these substrates are larger at low concentrations (Figure 1a) and decrease with increasing concentration (Figure 1c). The symbols in Figure 1d-f show the FA adsorption isotherm data of the L-enantiomers of the different substrates on the MIP, in the same format as in Figure 1a-c for the FA data on the NIP. First, we note that, at low concentrations, below 0.02 mM, the affinities of Fmoc-L-Tyr and Fmoc-L-Trp are similar (Figure 1d) but that the amount of Fmoc-L-Tyr adsorbed increases faster with increasing concentration than that of Fmoc-L-Trp (Figure 1e). At concentrations above 0.02 mM, the overall affinity of the substrates decreases in the following order: Fmoc-L-Tyr (9) > Fmoc-L-Trp (b) > Fmoc-L-Ser (2) > Fmoc-L-Phe (O) > Fmoc-Gly (0) > FmocL-Trp(OPfp) (4). Most importantly, the separation of the enantiomers of the structural analogues of the template was obtained on the FmocL-Trp MIP. This is illustrated in Figure 2a-c where the data for the L-enantiomers in Figure 1d-f are replotted together with the data for the corresponding D-enantiomers. The FA data for the L-enantiomers are shown as solid symbols and those for the D-enantiomers as open symbols. The overall affinities of the substrates that have the same stereochemistry as the template (i.e., the L-enantiomers) decrease in the order Fmoc-L-Tyr > FmocL-Trp > Fmoc-L-Ser > Fmoc-L-Phe. On the other hand, those of the substrates that have the opposite stereochemistry decrease 6418 Analytical Chemistry, Vol. 77, No. 19, October 1, 2005

in the same order to their affinities on the NIP: Fmoc-D-Tyr > Fmoc-D-Ser > Fmoc-D-Trp > Fmoc-D-Phe. The highest enantiomeric separation of the substrates was obtained in the low concentration range (see Figure 2a). Selection of Best Isotherm Model. The solid lines in Figures 1 and 2 show the best isotherm model accounting for the FA data. They correspond to tri-Langmuir isotherm models for all the compounds studied, except for Fmoc-L-Trp(OPfp) on the NIP and the MIP, in which cases the Langmuir isotherm model best account for the adsorption data. These best isotherm models were selected based on the following combination of reasons: (1) the classical statistical tests obtained for regression data; (2) the independent calculation of the AED from the FA data; (3) the comparison of the calculated and experimental band profiles. (1) Statistical Tests from Regression Data. Figures 1c, 1f, and 2c clearly show that the isotherm data of all the substrates on the two stationary phases are convex upward, hence Langmuirian. The only question was whether a multi-Langmuir or a more sophisticated isotherm model would fit them best. The goodness of the fit to each isotherm model in eqs 1-3 was compared using the results from the statistical tests made on the regression data. The Fisher parameter and the best estimates of the parameters of the tri-Langmuir isotherm model for each substrate on the NIP are summarized in Table 1. The best estimates of the parameters of the tri- and the tetra-Langmuir isotherm models for each substrate on the MIP are summarized in Tables 2 and 3, respectively. To compare the goodness of the

Figure 2. Adsorption equilibrium isotherms of enantiomers of structural analogues of Fmoc-L-Trp template on the MIP. The closed symbols and the open symbols represent the isotherm data for L-enantiomer and D-enantiomer, respectively: Fmoc-Trp (b, O); Fmoc-Tyr (9, 0); FmocSer (2, 4); Fmoc-Phe (1, 3). The solid and dotted lines represent the best calculated Tri-Langmuir isotherm for L-enantiomer and D-enantiomer on the MIP, respectively. Concentrations between (a) 0.005 and 0.12, (b) 0.005 and 1.6, and (c) 0.005 and 60 mM. Table 1. Isotherm Parameters and Fisher Parameters (Fcal) for Each Substrate on the NIP, Estimated by Fitting Data to the Tri-Langmuir Isotherm Model and from the Calculation of Affinity Energy Distributiona isotherm parameters b (mM-1)

q (mM) substrate

site 1

site 2

site 3

site 1

site 2

site 3

524 (533) ( 13 456.5 (406, 212) ( 9.1 598 (nd, 278) ( 18 439 (412) ( 87 745(1270, 204.1) ( 43 62.0 (nd,b 16.1) ( 14

69.8 (71.9) ( 30 46.9(28.3) ( 6.3 20.91 (10.49) ( 2.3 31.8(25.8) ( 11 38.9(10.8) ( 5.6 ndc (0.0232)

0.114(0.111) ( 0.059 1.408(1.37) ( 0.25 0.216(0.029) ( 0.086 0.0757(0.0455) ( 0.026 0.04088(0.0129) ( 0.011

0.0118(0.01205) ( 0.0066 0.0222(0.00757, 0.0486) ( 0.0017 0.0168 (nd, 0.0385) ( 9.9 × 10-4 0.0183 (0.0192) ( 0.0065 0.00887(0.00236, 0.0305) ( 0.00094 0.0194 (nd,b 0.0385) ( 0.0046

0.157 (0.156) ( 0.045 0.297 (0.394) ( 0.038 0.53 (1.00) ( 0.061 0.278 (0.313) ( 0.062 0.164(0.394) ( 0.017 ndb (0.628)

33.9(32.7) ( 20 7.27(6.43) ( 0.95 19.8(210.5) ( 7.1 47.0(104.8) ( 18 53.8(266) ( 16

Fcal 104

Fmoc-L-Trp

1.031×

Fmoc-L-Tyr

1.71× 104

Fmoc-L-Ser

4170

Fmoc-L-Phe

7420

Fmoc-Gly

3.107× 104

Fmoc-L-Trp(Opfp)

1180

a The values from the affinity energy distribution are shown in parentheses and the standard deviation of each parameter was obtained from the nonlinear regression. b nd, not determined due to the divergence at the lowest energy site on the AED. c nd, not determined due to unrealistically large standard deviations associated with the corresponding isotherm parameters from regression fitting.

fits of the data to each isotherm model (which has different numbers of parameters), we used the Fisher parameter (Fcal) defined as

Fcal )

∑(q (n - 1)∑(q (n - l)

ex,i

- qjex)2

ex,i

- qt,i)2

(4)

where qjex is the mean value of the set of experimental data, qex,i, qt,i is the value of q derived from the best isotherm model, l is the number of adjusted parameters of the model, and n is the number of data points acquired. Using the F-test to compare, for each substrate, the Fcal value obtained for each isotherm model, we decide which isotherm model provides the best fit, independently of the number of parameters. For example, the critical value to decide whether the tri-Langmuir isotherm model (six parameters) gives a better fit than the bi-Langmuir isotherm model (four parameters) at the 95% confidence level is 6.61. Thus, if the ratio of the Fcal values for the tri- and the bi-Langmuir isotherm models is larger than the critical value of 6.61, the tri-Langmuir model gives a significantly better fit of the isotherm data than the biLangmuir model. These calculations for the data acquired on the

NIP and using the Fcal values show that the tri-Langmuir isotherm model accounts best for the adsorption data of Fmoc-L-Trp (12.07 > 6.61), Fmoc-L-Tyr (24.02 > 6.61), Fmoc-L-Phe (10.51 > 6.61), and Fmoc-Gly (9.07 > 6.61). Furthermore, the standard deviations of the different isotherm parameters that were calculated from the fitting of the isotherm data to the tri-Langmuir isotherm model (Table 1) are reasonably lower than the isotherm parameters themselves. Although for Fmoc-L-Ser, higher Fisher parameters were obtained for the tri-Langmuir than for the bi-Langmuir isotherm model, the ratio of the Fisher parameters (1.21) for these two models is not larger than the critical value required to confirm a significant improvement of the fitting to the tri- than to the biLangmuir model. For Fmoc-L-Trp(OPfp), the fitting of the isotherm data to the bi-Langmuir isotherm model gives numerical values of the parameters that are too uncertain, the standard deviations of the parameters being 100 times larger than the parameters. For all the substrates studied on the MIP, the statistical tests show that the fit of the data is significantly better to the tri-Langmuir isotherm model than to the bi-Langmuir isotherm model. The critical ratio of the F values for the tetra- (eight parameters) and the tri-Langmuir (six parameters) isotherm models is 4.15 at the 95% confidence level. For most of the Analytical Chemistry, Vol. 77, No. 19, October 1, 2005

6419

Table 2. Isotherm Parameters and Fisher Parameters (Fcal) for Each Substrate on the Fmoc-L-Trp MIP, Estimated by Fitting Data to theTetra-Langmuir Isotherm Model and from the Calculation of Affinity Energy Distributiona isotherm parameters b (mM-1)

q (mM) substrate

Fcal

site 1

site 2

site 3

site 4

site 1

site 2

site 3

site 4

Fmoc-L-Trp

1.601×104

Fmoc-L-Tyr

9480

Fmoc-L-Ser

2.66×105

Fmoc-L-Phe

3.53×104

Fmoc- Gly

5.70×105

0.546(0.590) ( 0.067 0.114(0.05) ( 0.026 0.121(0.1204) ( 0.0099 0.0511(0.025) ( 0.083 0.0877(0.108) ( 0.032

0.182(0.156) ( 0.096 0.538(0.629) ( 0.071 0.0688(0.0486) ( 0.044 0.516(0.39) ( 0.083 0.0466(0.0385) ( 0.11 ndb (0.0739)

119(87) ( 18 450(1350) ( 110 196(167) ( 25 307(900) ( 130 69.3(41.3) ( 23

Fmoc-D-Tyr

2.95×104

Fmoc-D-Ser

2.80×104

Fmoc-D-Phe

884

127(111) ( 65 161(164) ( 65 37.3(393) ( 3.9 121(235) ( 85

3.35(3.63) ( 0.87 9.13(9.51) ( 3.0 0.783 (22.6) ( 0.43 7.15(5.22) ( 4.18

0.0463(0.0585) ( 0.015 0.144(0.161) ( 0.035 0.0417(0.218) ( 0.027 0.0354(0.0107) ( 0.025

0.0177(0.0167) ( 0.0062 0.0285(0.0305) ( 0.0020 0.01071(0.00955) ( 0.007 0.0245(0.0257) ( 0.0027 0.0078(0.0060) ( 0.017 0.0177(0.0026) ( 0.0045 0.0152(0.0138) ( 0.0064 0.0118(0.01205) ( 0.0058 0.01806(0.00298) ( 0.00086 0.0127 (nd) ( 0.0693

2.84(2.54) ( 1.3 15.0(30) ( 4.6 1.053(1.0) ( 0.15 21.3(48.1) ( 11 0.920(0.792) ( 0.65

1059

82. (90.6) ( 46 52.2(46.9) ( 5.9 155(206.1) ( 150 26.7(24.5) ( 4.5 139(181) ( 577 ndb(7.53)

5.42(6.36) ( 2.9 1.44(0.93) ( 0.38 15.2(15.6) ( 3.2 0.376(0.249) ( 0.13 6.25(7.03) ( 6.3

Fmoc-L-Trp(Opfp) Fmoc-D-Trp

614(617) ( 47 585(584) ( 12 743(753) ( 50 535(544) ( 31 821(830) ( 190 102.4 (529) ( 25 596(633) ( 45 715(719) ( 97 784(1240) ( 14 568 (ndc) ( 550

0.119(0.107) ( 0.048 0.125(0.123) ( 0.0433 0.41(0.0385) ( 0.056 0.0676(0.0481) ( 0.22

3.48(3.054) ( 0.87 1.95(2.0092) ( 0.48 7.73(0.628) ( 4.5 1.18(1.69) ( 0.42

174(126) ( 53 83.5(65.8) ( 17 220(32.7) ( 160 104.8(169) ( 280

3.29×105

a The values from the affinity energy distribution are shown in parentheses and the standard deviation of each parameter was obtained from the nonlinear regression. b nd, not determined due to unrealistically large standard deviations associated with the corresponding isotherm parameters from regression fitting. c nd, not determined due to the divergence at the lowest energy site on the AED.

Table 3. Isotherm Parameters for Each Substrate on the Fmoc-L-Trp MIP, Estimated by Fitting Data to the Tri-Langmuir Isotherm Model isotherm parameters b (mM-1)

q (mM) substrate

Fcal

site 1

site 2

site 3

site 1

site 2

site 3

Fmoc-L-Trp Fmoc-L-Tyr Fmoc-L-Ser Fmoc-L-Phe Fmoc-Gly Fmoc-D-Trp Fmoc-D-Tyr Fmoc-D-Ser Fmoc-D-Phe

3390 5730 6440 2.71×104 9870 2230 3720 1.38×104 5.93×104

550.2 ( 17 562 ( 17 696 ( 13 496 ( 25 754 ( 34 526 ( 13 623 ( 17 764 ( 14 496 ( 17

23.1 ( 3.6 32.7 ( 3.8 25.2 ( 1.7 19.02 ( 2.6 12.7 ( 3.0 10.4 ( 1.9 38.7 ( 4.8 29.5 ( 2.2 9.81 ( 1.07

0.7054 ( 0.057 0.351 ( 0.037 0.143 ( 0.012 0.165 ( 0.015 0.106 ( 0.027 0.128 ( 0.036 0.3607 ( 0.0704 0.178 ( 0.026 0.04106 ( 0.0093

0.0313 ( 0.0023 0.0353 ( 0.0025 0.0223 ( 0.00083 0.02901 ( 0.0026 0.015 ( 0.0012 0.0344 ( 0.0025 0.0289 ( 0.0021 0.0197 ( 0.00078 0.0287 ( 0.0015

0.884 ( 0.13 1.004 ( 0.11 0.731 ( 0.044 0.752 ( 0.084 0.553 ( 0.12 1.270 ( 0.24 0.638 ( 0.072 0.567 ( 0.038 0.9607 ( 0.092

89.8 ( 10 148 ( 18 158 ( 21 116 ( 12 59.5 ( 15 75.8 ( 19 41.2 ( 7.2 57.1 ( 9.2 93.4 ( 20

substrates studied on the MIP, the ratio of these parameters is higher than the critical value, suggesting that the fitting of the data to the tetra-Langmuir isotherm model gives results significantly better than that to the tri-Langmuir isotherm model. However, for Fmoc-L-Tyr (1.65 < 4.15), Fmoc-L-Phe (1.30 < 4.15), and Fmoc-D-Ser (2.03 < 4.15), the improvement of the fitting is not significant. For Fmoc-D-Phe, the statistical test shows that the tri-Langmuir isotherm model accounts better for the isotherm data than the tetra-Langmuir isotherm model. Furthermore, unrealistically high values were obtained for the standard deviations of each of the tetra-Langmuir isotherm parameters (Table 2) of Fmoc-DPhe on the MIP. This result confirms that this model is not the best one for this compound. (2) Repeatability of Fmoc-L-Trp Data on the MIP. The repeatability and reproducibility of the isotherm parameters obtained in this study were tested using the adsorption isotherm data of the template, Fmoc-L-Trp on the MIP (Figure 3). Three sets of isotherm data were obtained for the template at the beginning, in the middle, and at the end of the series of experiments done to acquire the FA adsorption isotherm data for 6420 Analytical Chemistry, Vol. 77, No. 19, October 1, 2005

the different substrates studied here. We observed a good agreement between these three series of isotherm data for FmocL-Trp on the MIP that were obtained over a three-month period, in the low concentration range (0.005-0.14 mM; Figure 3a), the intermediate concentration range (0.005-1.5 mM; Figure 3b), and the high concentration range (0.005-50 mM; Figure 3c). For each parameter, the RSD is smaller than 15%. (3) Calculation of the Affinity Energy Distributions. In a previous section, we discussed the statistical results of the regression of the FA isotherm data to different Langmuirian isotherm models. These results demonstrate that the surfaces of the MIP and the NIP are not homogeneous. At least three different types of sites could be identified based on the modeling of the FA data and the statistical tests on the NIP in an acetonitrile/ acetic acid (99/1 v/v) mobile phase. A similarly complex surface, covered with at least three different types of sites, was shown to exist on the MIP. These different types of sites interact with all the substrates studied, except Fmoc-L-Trp(OPfp), for which a Langmuir isotherm model accounts best for the isotherm data.

Figure 3. Adsorption isotherm data of Fmoc-L-Trp acquired at the beginning (O), in the middle (0), and at the end (4) of this study, over a three-month period. Concentrations between (a) 0.005 and 0.12, (b) 0.005 and 1.2, and (c) 0.005 and 50 mM.

The results of the isotherm modeling are now compared to those of the calculation of the AED of these substrates from the same data. The AEDs were derived independently from the same experimental isotherm data. Using the EM method, we calculated the AEDs without any prior assumption regarding the types of surface heterogeneity. During the calculation of the AEDs, we increased the number of iterations from 1 × 106 to 1 × 108, to improve the resolution between the peaks on the AED. The resolution between the two low-energy types of sites increases with increasing number of iterations. At the same time, a very small additional peak appears in the AEDs of Fmoc-D-Trp and Fmoc-D-Phe on the MIP (data not shown). This very high adsorption-energy peak would correspond to an extremely small fraction of the polymer surface, and it was not possible to acquire experimental data validating its existence. In further discussions, we consider that this additional peak in the AEDs of Fmoc-D-Trp and Fmoc-D-Phe is artifactual and does not correspond to any actual characteristic of the surface of the MIP. Figure 4 shows the AEDs of Fmoc-Trp and Fmoc-Tyr on the NIP and on the MIP as representative examples. The saturation capacities and the association constants of each type of site identified were calculated as explained earlier. The values obtained for the NIP and the MIP are included in parentheses in Tables 1 and 2, respectively. The AEDs of the different substrates on the NIP (Figure 4a and b) exhibit three distinct peaks, a result in agreement with the selection of a tri-Langmuir isotherm model made independently, as explained earlier. The AEDs of the different substrates on the MIP (Figure 4c and d) show four different distinctive peaks, in agreement with the selection of tetraLangmuir isotherms, and suggesting that four different types of sites exist on the MIP for these substrates, within the investigated concentration range. However, the data in Table 2 show that the number of experimental data points acquired (at least 30 data points) is insufficient to permit an accurate calculation of the eight different isotherm parameters that are needed for a tetra-Langmuir isotherm model. The best tri-Langmuir isotherm parameters for each different substrate on the MIP are listed in Table 3. Comparing Tables 3 and 2 (the best tetra-Langmuir isotherm parameters), we can see that the parameters corresponding to the two low-energy sites (sites 1 and 2 shown in Figure 4c and d) in the tetra-Langmuir isotherm model are averaged as the

parameters for the lowest energy sites (sites 1 in Table 3) in the tri-Langmuir isotherm model. (4) Comparison of Calculated and Experimental Peak Profiles. A final check of the validity of the selection of the isotherm model of the different substrates on the NIP and the MIP is afforded by a comparison of the experimental overloaded band profiles with those calculated using an appropriate model of chromatography and these isotherms.18,19 In these earlier studies, it was shown that the POR model accounts very well for the mass-transfer kinetics of the substrates on the stationary phases used in this study. The details of the calculation involved with the use of the POR model and of the determination of the mass-transfer coefficients needed can be found elsewhere.19 In this paper, we use the mass-transfer coefficient derived previously and calculate the elution band profiles using each one of several different isotherm models. These calculated band profiles are then compared to experimental band profiles. As an example, we compare in Figure 5 the calculated and experimental band profiles of Fmoc-L-Tyr on the NIP (Figure 5a) and Fmoc-L-Tyr on the MIP (Figure 5b) at 25 ( 2 °C, for samples of the same volume and two widely different concentrations, 0.1 and 40 mM. For Fmoc-L-Tyr on the NIP (Figure 5a), the band profiles were calculated using two different isotherm models, the bi-Langmuir and the tri-Langmuir models. The profiles derived from the bi-Langmuir model (solid lines) deviate significantly from the experimental profiles for each substrate (symbols), especially at low concentrations. In contrast, there is a good agreement between the experimental band profiles and those calculated with the tri-Langmuir isotherm model (dotted lines). For Fmoc-L-Tyr on the MIP (Figure 5b), the band profiles calculated either with the tri-Langmuir (solid lines) or the tetra-Langmuir (dotted lines) show similarly good agreement with the experimental band profiles. In summary, the selection of the isotherm accounting best for the isotherm data was decided by the following three independent methods: (1) the statistical results from the regression of the FA isotherm data; (2) the affinity energy distribution derived from the same isotherm data; (3) the comparison of experimental and calculated band profiles at low and high concentrations. Considering the consistency of the results obtained from these three methods, within the concentration range investigated and given Analytical Chemistry, Vol. 77, No. 19, October 1, 2005

6421

Figure 4. Affinity energy distributions for selected substrates (a) Fmoc-L-Trp on the NIP, (b) Fmoc-L-Tyr on the NIP, (c) Fmoc-L-Trp (solid lines) and Fmoc-D-Trp (dotted lines) on the MIP, and (d) Fmoc-L-Tyr (solid lines) and Fmoc-D-Tyr (dotted lines)on the MIP. (a-d) are expanded views of the corresponding affinity energy distributions. They show the high-energy sites, which have a low density. The insets in (c) and (d) 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 L-enantiomers, and the indices 1′, 2′, 3′, and 4′ represent the energy modes for D-enantiomers.

Figure 5. Comparison between calculated and experimental peak profiles for the substrates (a) Fmoc-L-Tyr on the NIP and (b) Fmoc-L-Tyr on the MIP. The experimental peak profiles were represented by symbols. For Fmoc-L-Tyr on the NIP, the calculated peak profiles were calculated using the bi-Langmuir (solid lines) or tri-Langmuir (dotted lines) isotherm model. For Fmoc-L-Tyr on the MIP, the calculated peak profiles were calculated using the tri-Langmuir (solid lines) or tetra-Langmuir (dotted lines) isotherm model. The y-axis was normalized by the inlet concentration to show peak profiles obtained with two widely different concentrations of the substrate. The injection time was 1 min, and the following inlet concentrations were used from right to the left on the each figure: for Fmoc-L-Tyr on the NIP, 0.25 and 50.76 mM; for Fmoc-L-Tyr on the MIP, 0.125 and 50.32 mM.

the limited number of data points obtained, we can conclude that the tri-Langmuir isotherm model best accounts for the isotherm data for each different substrate on the NIP as well as on the MIP, except for the data of Fmoc-L-Trp(OPfp) for which the Langmuir isotherm accounts best. 6422 Analytical Chemistry, Vol. 77, No. 19, October 1, 2005

Isotherm Parameters for Structural Analogues on the MIP and on the NIP. The sets of isotherm parameters derived for each substrate from the regression of the adsorption data on the NIP and the MIP (Tables 1 and 3, respectively) were used to calculate the overall affinity (the sum of the products of the

saturation capacity (qs) and of the association constant (b) of each type of site) and the affinity for each type of site identified, for each substrate. The overall affinities ((qb)t) of the different substrates are higher on the MIP than on the NIP. For example, the overall affinities of Fmoc-L-Trp on the NIP and the MIP were 63.3 and 3.86, respectively. On both polymers, Fmoc-L-Trp(OPfp), which has no functional group that could interact strongly with the polymers, interacts only with the lowest energy sites of type 1. On the NIP, the substrates interact mostly with the sites of the two lower energy types (types 1 and 2), although adsorption on the highest energy type of sites contributes to between 15 and 30% of the overall affinity. The overall affinity on the NIP decreases in the following order:

Fmoc-L-Tyr ((qb)t ) 34.3) > Fmoc-L-Ser ((qb)t ) 25.4) > Fmoc-L-Trp ((qb)t ) 21.006) > Fmoc-L-Phe ((qb)t ) 20.43) > Fmoc-Gly ((qb)t ) 15.2) The same trend is observed on the MIP for the substrates that have a stereochemistry different from that of the template, FmocL-Trp. The two lower energy sites (sites 1 and 2) have similar contributions and dominate the overall affinity (although the third type of sites accounts again for between 15 and 25% of the overall affinity). The overall affinity decreases in the following order:

Fmoc-D-Tyr ((qb)t ) 57.6) > Fmoc-D-Ser ((qb)t ) 41.9) > Fmoc-D-Trp ((qb)t ) 41.005)> Fmoc-D-Phe ((qb)t ) 27.5) On the other hand, for the substrates that have the same stereochemistry as the template, the type of site with the highest energy (sites of type 3) dominates the overall affinity. The overall affinity decreases in the following order:

Fmoc-L-Tyr ((qb)t ) 104.6) > Fmoc-L-Trp ((qb)t ) 100.99) > Fmoc-L-Ser ((qb)t ) 56.5) > Fmoc-L-Phe ((qb)t ) 47.8) The differences in the overall affinities of Fmoc-L-Tyr and FmocL-Trp on the MIP are very small (3%). On the other hand, the affinity on the highest energy sites (site 3) on the MIP decreases in the following order:

Fmoc-L-Trp ((qb)3 ) 63.3) > Fmoc-L-Tyr ((qb)3 ) 51.9) > Fmoc-L-Ser ((qb)3 ) 22.6)> Fmoc-L-Phe ((qb)3 ) 19.1) The affinity of Fmoc-L-Trp is 19% higher than that of Fmoc-LTyr on the highest energy type of site on the MIP. These results show that the affinity of the different substrates for the NIP increases with increasing number of functional groups on the substrates and with increasing hydrophobicity of the substrates, the former factor being more important than the latter. Similarly, on the MIP, these two factors control the affinity of the substrates that have a stereochemistry different from that of the template. On the other hand, when the substrates have the same stereochemistry as the template, the imprinting process results

in a higher affinity for the template (Fmoc-L-Trp) than for the other substrates. This rule applies even to the substrates that have a larger number of functional groups that can interact with the pyridine functional groups on the highest energy type of binding site of the MIP. These changes in the affinity of the different substrates for the highest energy sites of the MIP originate mainly from the changes in the numbers of sites of this type and in the corresponding association constants. Figure 6 compares the relative changes in the affinity of the different substrates for the highest energy sites of the MIP, in the numbers of these sites. The x-axis is arbitrary, chosen to present the data in order of decreasing affinity for the type 3 sites. Figure 6 shows that, for both the L-enantiomers (Figure 6a and b) and the D-enantiomers (Figure 6c and d), the affinity for the highest energy sites is influenced mostly by the number of such sites rather than by the corresponding association constant. The cross-reactivities of the structural analogues of the template on the MIP were studied by calculating their overall selectivity and their selectivity on each type of site identified compared to the template (Fmoc-L-Trp). This last value was defined as the ratio of the affinity of a type of site for Fmoc-L-Trp to the corresponding affinity for the substrate. The results are summarized in Table 4. As in the case of the affinity, the selectivity on the highest energy sites (sites 3) dominates the overall selectivity. The data in Table 4 show that the selectivity toward Fmoc-L-Trp is higher for the enantiomers of the other substrates that have a different stereochemistry. For example, the selectivity toward Fmoc-L-Trp is higher for Fmoc-D-Tyr (4.25) than for FmocL-Tyr (1.22). For the D-enantiomers, the selectivity for the different substrates relative to the template on the highest energy sites decreases in the following order:

Fmoc-D-Phe > Fmoc-D-Trp = Fmoc-D-Ser > Fmoc-D-Tyr For the L-enantiomers, this selectivity on the highest energy sites decreases within the errors in the following order:

Fmoc-L-Phe = Fmoc-L-Ser > Fmoc-L-Tyr An “imprinting factor” was also calculated. This factor is defined as the ratio of the overall selectivities toward the template on the MIP and on the NIP. This ratio cancels out factors that are independent of the MIP, such as substrate-solvent interactions or substrate-substrate interactions. These values show the same trend. The imprinting effect on the MIP decreases in the following order:

Fmoc-Gly (2.45 ( 0.90) > Fmoc-L-Phe (1.60 ( 0.66) = Fmoc-L-Ser (1.60 ( 0.55) > Fmoc-L-Tyr (1.26 ( 0.45) These results show that the cross-reactivities on the MIP of the structural analogues increase with increasing number of functional groups in the molecule of the substrates that can interact with the vinlypyridine on the highest energy sites of the MIP. Table 5 summarizes the separation factors of the different enantiomeric pairs studied, template and structural analogues. The Analytical Chemistry, Vol. 77, No. 19, October 1, 2005

6423

Figure 6. Changes in (a) the affinity (q3b3) of different L-enantiomers (as indicated in the graph) on the highest energy site (site 3 in Table 3) of the MIP. (b) The number of the highest energy site (q3) and the corresponding association constant (b3) of the MIP for the different L-enantiomers. (c) The affinity of different D-enantiomers (as indicated in the graph) on the highest energy site (site 3 in Table 3) of the MIP. (d) The number of the highest energy site (q3) and the corresponding association constant (b3) of the MIP for the different D-enantiomers. Table 4. Selectivity toward the Template Fmoc-L-Trp Compared to the Other Substrates on the MIP MIP

substrates

no. of functional groups

log Pow

(qb)t,L-trp/ (qb)t,substrates

(qb)1,L-trp/ (qb)1,substrates

(qb)2,L-trp/ (qb)2,substrates

(qb)3,L-trp/ (qb)3,substrates

Fmoc-L-Tyr Fmoc-L-Ser Fmoc-L-Phe Fmoc-Gly Fmoc-L-Trp(Opfp) Fmoc-D-Trp Fmoc-D-Tyr Fmoc-D-Ser Fmoc-D-Phe

2 2 1 1 0 1 2 2 1

4.17 1.48 4.65 2.53 7.61 4.74 4.17 1.48 4.65

0.965 ( 0.069 1.786 ( 0.103 2.11 ( 0.16 4.098 ( 0.47 9.501 ( 3.4 2.46 ( 0.27 1.75 ( 0.16 2.408 ( 0.16 3.67 ( 0.28

0.868 ( 0.096 1.110 ( 0.099 1.20 ( 0.16 1.52 ( 0.19 9.501 ( 3.4 0.952 ( 0.11 0.956 ( 0.11 1.14 ( 0.104 1.21 ( 0.12

0.622 ( 0.17 1.11 ( 0.26 1.43 ( 0.40 2.91 ( 1.1

1.22 ( 0.039 2.804 ( 0.087 3.31 ( 0.089 10.044 ( 1.3

1.55 ( 0.52 0.827 ( 0.225 1.22 ( 0.29 2.17 ( 0.56

6.53 ( 0.93 4.25 ( 0.30 6.23 ( 0.32 16.5 ( 1.6

enantiomeric separation factor (R) is the ratio of the affinity for the L- and the D-enantiomers. The enantiomeric separations on the highest energy sites have a dominant contribution to the overall enantiomeric separation. The highest enantiomeric separation on this type of site was obtained for the template, Fmoc-LTrp. For the other substrates, the enantiomeric separation decreases in the following order:

Fmoc-Phe > Fmoc-Tyr > Fmoc-Ser These results show that the cross-reactivity of the enantiomers of the structural analogues increases with increasing hydrophobicity of the enantiomers. 6424

Analytical Chemistry, Vol. 77, No. 19, October 1, 2005

CONCLUSION The systematic acquisition of FA adsorption isotherm data of the template, its enantiomer, and a group of structural analogues on a MIP and the corresponding NIP permits the investigation of the cross-reactivities of these analogues. The nonlinear regression of the isotherm data, the independent calculation of the affinity energy distribution, and the comparison of calculated and experimental band profiles show that the tri-Langmuir isotherm model best accounts for all of the systems studied, within a wide concentration range, except for Fmoc-L-Trp(OPfp), which does not exhibit any significant affinity for the polymer used. In this case, the Langmuir isotherm model accounts best for the isotherm data. Accordingly, the surfaces of the two polymers exhibit three

Table 5. Enantiomer Separation (r) of Different Substrates on the MIP

substrates

no. of functional groups

log Pow

R

R1

R2

R3

Fmoc-Trp Fmoc-Tyr Fmoc-Ser Fmoc-Phe

1 2 2 1

4.74 4.17 1.48 4.65

2.46 ( 0.27 1.82 ( 0.17 1.35 ( 0.075 1.74 ( 0.15

0.952 ( 0.11 1.102 ( 0.12 1.031 ( 0.062 1.011 ( 0.12

1.55 ( 0.52 1.33 ( 0.31 1.101 ( 0.15 1.52 ( 0.35

6.53 ( 0.93 3.50 ( 0.26 2.22 ( 0.12 4.99 ( 0.49

MIP

different types of adsorption sites for the Fmoc derivatives of amino acids. Each type of site is nearly homogeneous. The detailed study of the best estimates of the isotherm parameters of the tri-Langmuir isotherms shows that the interactions between the nonimprinted polymer and the substrates increase with increasing number of functional groups and with increasing hydrophobicity of the substrates, the former being the more important. The polymer imprinted with Fmoc-L-Trp exhibits highest energy sites that have the largest affinity for Fmoc-L-Trp, due to the number of these sites being greater for Fmoc-L-Trp than for its structural analogues while the corresponding association constants for the substrates are similar to that for the template. A considerable extent of cross-reactivity of the structural analogues is observed, particularly for the structural analogues that have the same stereochemistry as the template. The crossreactivity of the structural analogues increases with increasing number of functional groups in the substrate molecules, groups

that can interact with the pyridine functional groups of the highest energy sites of the MIP. In contrast, the cross-reactivity of the enantiomers of these substrates increases with increasing hydrophobicity of the substrates. 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 May 25, 2005. Accepted August 6, 2005. AC050914+

Analytical Chemistry, Vol. 77, No. 19, October 1, 2005

6425